Studies of the Off-Field Behaviours of Sportspeople: Are Sponsors’ Objectives at Risk?

Studies of the off-field behaviours of sportspeople: Are sponsors’ objectives at risk?

Ellen Bloxsome B. Arts (Sociology) hons. Murdoch University B. Comm. (Marketing) Murdoch University

This thesis is submitted for the degree of Doctor of Philosophy in the Faculty of Business at the Queensland University of Technology.

June 2012

Keywords: Sport Sponsorship, Objectives, Athlete Off-Field Behaviours, Consumer evaluations, Identification, Balance

Statement of Original Authorship

The work contained in this thesis has not been previously submitted to meet requirements for an award at this or any other higher education institution. To the best of my knowledge and belief, the thesis contains no material previously published or written by another person except where reference is made.

______

Ellen Bloxsome

Abstract

The research reported here addresses the problem of athlete off-field behaviours as they influence sports’ sponsors, particularly the achievement of sponsorship objectives. The question arises because of incidents of sponsorship contract cancellation following news-media reporting of athletes’ off-field behaviours. Two studies are used to investigate the research question; the first establishes the content of news-media reports, and the second tests the effects of news’ reports on athlete, team and sponsor evaluations using an experimental design.

Key assumptions of the research are that sponsorship objectives are principally consumer-based and mediated. Models of sponsorship argue that sponsors aim to reach and influence consumers through sponsees. Assuming this pathway exists is central to sponsorship activities. A corollary is that other mediators, in this case the news-media, may also communicate (uncontrollable) messages such that a consumer audience may be told of negative news that may then be associated with the sponsor. When sponsors cancel contracts it is assumed that their goal is to control the links between their brand and a negative referent.

Balance theory is used to discuss the potential effects of negative off-field behaviours of athletes on sponsor’s objectives. Heider’s balance theory (1958) explains that individuals prefer to evaluate linked individuals or entities consistently. In the sponsorship context this presents the possibility that a negative evaluation of the athlete’s behaviour will contribute to correspondingly negative evaluations of the athlete’s team and sponsors.

A content analysis (Study 1) was used to survey the types of athlete off-field behaviours commonly reported in a newspaper. In order to provide a local context for the research, articles from the Courier Mail were sampled and teams in the National Rugby League (NRL) competition were the focus of the research. The study identified nearly 2000 articles referring to the NRL competition; 258 of those refer to off-field incidents involving athletes. The various types of behaviours reported include assault, sexual assault allegations, driving under the influence of alcohol, illicit drug use, breaches of club rules, and positive off-field activities (i.e., charitable activities).

An experiment (Study 2) tested three news’ article stimuli developed from the behaviours identified in Study 1 in a between-subjects design. A measure of Identification with the Team was used as a covariate variable in the Multivariate Analysis of Covariance analysis. Social identity theory suggests that when an individual identifies with a group, their attitudes and behaviours towards both in- and out-group members are modified. Use of Identification with the Team as a covariate acknowledges that respondents will evaluate behaviours differently according to the attribution of those behaviours to an in- or out-group member.

Findings of the research suggest that the news’ article stimuli have significant, large effects on evaluations of athlete off-field behaviour and athlete Likability. Consistent with pretest results, charitable fundraising is regarded as extremely positive; the athlete, correspondingly, is likable. Assault is evaluated as extremely negative, and the athlete as unlikable. DUI scores reveal that the athlete’s behaviour is very

4 negative; however, the athlete’s likability was evaluated as neutral. Treatment group does not produce any significant effects on team or sponsor variables.

This research also finds that Identification with the Team has significant, large effects on team variables (Attitude toward the Brand and Corporate Image). Identification also has a significant large effect on athlete Likability, but not on Attitude toward the Act. Identification with the Team does not produce any significant effects on sponsor variables.

The results of this research suggest that sponsor’s consumer-based objectives are not threatened by newspaper reports linking athlete off-field behaviour with their brand. Evaluations of sponsor variables (Attitude toward the Sponsor’s Brand and Corporate Image) were consistently positive. Variance in that data, however, cannot be attributed to experimental stimuli or Identification with the Team. These results argue that respondents may regard sponsorships, in principle, as good.

Although it is good news for sponsors that negative evaluations of athletes will not produce correspondingly negative evaluations of consumer-based sponsorship objectives, the results indicate problems for sponsorship managers. The failure of Identification with the Team to explain sponsor variable variance indicates that the sponsor has not been evaluated as a linked entity in a relationship with the sporting team and athlete in this research. This result argues that the sponsee-mediated affective communication path that sponsors aim use to communicate with desirable publics is not necessarily a path available to them.

Acknowledgements

After having written several hundred pages of thesis and deleting more than I care to think about; having consistently (and hopefully, mildly) annoyed my supervisors with intransigence, relcalcitrance and all manner of other (hopefully, minor) offences, I’m a little surprised to find myself near the end of this process. As I’ve procrastinated and mused, and scribbled my way through this peculiar brand of purgatory, I’ve had plenty of help. And although I’m a little hysterical at this point, my appreciation of those who have helped along the way is sincere.

Thanks, firstly, must therefore go to my supervisory team: Larry Neale, Ian Lings and Nigel Pope. I promise to (try to) be less horrible in future.

My thanks also go to many people in academe who helped in a multitude of ways. Trina Robbie was among the first people I met at QUT and remains a fantastic first(ish) impression. Trina, and later, Carol and Dennis have been the keepers of all knowledge related to the forms, the dates, the milestones, and the people to whom one should speak. The consistently speedy and accurate access to that knowledge has been more than merely useful.

To the people who facilitated data collection by giving me access to their classes, handing out forms, or giving advice on how to proceed; you have my gratitude for your time and your effort: Dan Funk, Dominique Greer, Kim Johnston, Kerri Kuhn, Ian Lings, Anita Love, John McDonnell, Larry Neale, Nigel Pope, Janet Ransley, and Christine Voge.

I must acknowledge the assistance of the members of various panels over the course of my enrolment who tried to improve my research, or my mind. Thank you: Jennifer Bartlett, Stephen Cox, Paula McDonald, Alan McKee, Clinton Weeks and my supervisors. And, thanks to Kate, Tifani and other Z701 people who provided the chocolates and the moral support while panels deliberated.

For the wine, food, and discussions on how to best change the universe my thanks go to Anita, Jannie, Kerri, and Nigel.

Finally, I would like to express my gratitude to, and for, my parents and my brothers. I couldn’t begin to list the many and strange ways they have been inspiring, or the occasions and value of the support they have provided.

Table of Contents

Statement of Original Authorship ...... 3 Abstract ...... 4 Acknowledgements ...... 6 List of Figures ...... 10 List of Tables ...... 11

Chapter 1: Off-field behaviours and sport sponsorsorship ...... 13 A dearth of negatives ...... 14 Uncontrollable factors in sport sponsorship contracts...... 17 Defining the off-field context: scandalous morality ...... 26 Research Agenda and Research Questions ...... 32 Conclusions ...... 36

Chapter 2: Review of Literatures ...... 37 A Background on Sponsorship ...... 38 Sponsorship defined ...... 38 Size of the Sponsorship Market ...... 42 Superlative Sport ...... 44 Sponsorship Objectives ...... 57 Theories that explain consumer responses to sponsorship ...... 74 The Information Processing Model ...... 76 Impression formation processes ...... 82 Balance theory ...... 85 Identification ...... 92 Summary of Theory Literatures ...... 124

Chapter 3: Conceptual Framework ...... 129 Conclusions from the Introduction & Literature Review ...... 131 Research Considerations: Athlete off-field behaviours, respondents & sampling .. 133 Study 1: Content Analysis of Local News ...... 136 Pretest: Valence of Behaviours ...... 136 Pretest: Identification scale ...... 137 Study 2: Experimental test of news’ stimuli on consumer evaluations ...... 138 Variables used in Study 2 ...... 139 Hypotheses & Research Models ...... 143

Chapter 4: Study 1: Content analysis of local news ...... 148 Aims...... 148 Method ...... 150 Findings ...... 156 Discussion ...... 168

Chapter 5: Pretests of Variables ...... 176 Pretest 1: Valence of Behaviours ...... 177 Literature & Aims ...... 177 Method ...... 181 Findings ...... 183 Discussion ...... 192 Pretest: Team Identification ...... 197 Literature & Aims ...... 197 Method ...... 204 EFA Findings: State of Origin data ...... 210 Discussion: State of Origin data ...... 211 EFA Findings: Brisbane Broncos data ...... 213 Discussion: Model Specifications ...... 214 CFA Findings: Brisbane Broncos data ...... 215 Conclusions: Multidimensional & Unidimensional Identification ...... 225

Chapter 6: Study 2: Test of news’ article stimuli effects on attitudes ...... 226 Hypotheses ...... 227 Method ...... 230 Findings ...... 239 Confound Model Specification ...... 247 Final analysis: confounds removed ...... 252 Hypothesis 1: Effects of news’ on evaluations...... 257 Hypothesis 2: Effects of Identification ...... 260 Hypothesis 3: the balanced model ...... 262 Discussion of Results ...... 270

Chapter 7: Conclusions & Future Research ...... 272 Summary of Findings: Research Questions ...... 273 Limitations...... 277 Opportunities for Future Research ...... 279

Contributions of the Research ...... 284 Conclusions ...... 293

Appendices ...... 296 Chapter 1 ...... 297 Chapter 2 ...... 298 Chapter 5 ...... 301 Chapter 6 ...... 306

References ...... 316

List of Figures Figure 2.1: The Sponsorship Model ...... 39 Figure 2.2: Annual change in spending on marketing communications ...... 43 Figure 2.3: Top 20 rating television programs by genre: Australia 2001-09 ...... 48 Figure 2.4: Sports attendance by frequency of attendance, 2009-10 ...... 51 Figure 2.5: Performing arts attendance by frequency of attendance, 2005-06 ...... 52 Figure 2.6: Balanced and unbalanced triads ...... 87 Figure 3.7: The research process ...... 130 Figure 3.8: Model A: Hypotheses 1 & 2 ...... 145 Figure 3.9: Model B: Hypothesis 3 ...... 146 Figure 3.10: Model B (ii): Hypothesis 3 ...... 147 Figure 4.11: Article type: NRL in the Courier Mail January-June 2009 ...... 157 Figure 4.12: Competition: match play, injuries, on-field & off-field reports ...... 160 Figure 5.13: A-act relationships between scenarios ...... 196 Figure 5.14: Genealogy of Identification scales ...... 198 Figure 5.15: Standardised Factor Model: Model 1 ...... 221 Figure 6.16: Model A: Hypotheses 1 & 2 ...... 227 Figure 6.17: Model B: Hypothesis 3 ...... 229 Figure 6.18: Model B (ii): Hypothesis 3 ...... 229 Figure 6.19: Graph of variable means: Study 2 ...... 252 Figure 6.20: Balance: Hypothesised & Actual Effects of Group ...... 264 Figure 6.21: Balance: Effects of Group on Evaluations of the Athlete ...... 265 Figure 6.22: Balance: Hypothesised and Actual Effects of Identification ...... 267 Figure 6.23: Balance: Effects of Identification on Likability & Team variables..... 268

List of Tables

Table 1.1: Old vs Modern News-values ...... 22 Table 1.2: Research questions & goals ...... 34 Table 2.3: Top 5 Popular sports in Australia & Queensland, 2009-10 ...... 47 Table 2.4: Sports programs ranked by audience size (000’s), 2001-09 ...... 50 Table 2.5: The Hierarchy of Effects and Sponsorship Objectives ...... 61 Table 2.6: Consequences of Identification ...... 102 Table 2.7: Identification Scale Dimensions ...... 115 Table 3.8: Research Questions ...... 129 Table 4.9: Variable Descriptions & Coding ...... 152 Table 4.10: Frequencies: Article type by month ...... 156 Table 4.11: Multiple t-test comparison: Article type codes ...... 157 Table 4.12: Frequencies: Binary variables by Month ...... 158 Table 4.13: Cochran Q test of difference: Binary variables...... 159 Table 4.14: Cochran Q test of month-by-month differences: Binary variables ...... 159 Table 4.15: Frequencies: Number of Articles & Number of incidents ...... 161 Table 4.16: Test of Difference: Numbers of Articles & Incidents ...... 161 Table 4.17: NRL off-field incidents reported by the Courier Mail ...... 163 Table 4.18: Number of Incidents & Total Reports by Year ...... 164 Table 4.19: Categories of Behaviour ...... 165 Table 4.20: Categorisation consideations: ASOC and Morality ...... 166 Table 4.21: Categories, Incidents, & Incident reports ...... 167 Table 4.22: Representation: Court offences & NRL off-field incidents ...... 172 Table 5.23: Behaviour valence: Scenarios from Study 1 ...... 182 Table 5.24: Scenario: Moral Cue Diagnosticities and Scores ...... 184 Table 5.25: Repeated measures ANOVA: Differences in Diagnosticity ...... 184 Table 5.26: Pairwise comparisons of Moral Cue Diagnosticities ...... 185 Table 5.27: Behaviour EFA: Common Structue ...... 188 Table 5.28: Scenarios: A-act descriptive statistics & scores...... 189 Table 5.29: Repeated measures ANOVA: differences in A-act ...... 190 Table 5.30: Pairwise comparisons of A-act scores ...... 191 Table 5.31: Pearson correlation: Cue Diagnosticity & A-act ...... 193 Table 5.32: Identification: pretest items & origins ...... 205 Table 5.33: EFA: State of Origin solution ...... 211 Table 5.34: EFA: Brisbane Broncos solution ...... 214 Table 5.35: Model 1: Large residual covariances ...... 216 Table 5.36: Model 1: Error variable Modification Indices & Variances ...... 217 Table 5.37: Model 1, Fit 2: Regression estimates ...... 220 Table 5.38: Model 1, Fit2: Latent variable correlations...... 222 Table 5.39: Model 2, Fit 2: Standardised Model ...... 225 Table 6.40: Pretest A-act scores ...... 231 Table 6.41: Assumptions: Missing data by case ...... 241 Table 6.42: Assumptions: Normal distribution: Group A ...... 242 Table 6.43: Assumptions: Normal distribution: Group B ...... 243 Table 6.44: Assumptions: Normal distribution: Group C ...... 243 Table 6.45: Assumptions: Outliers by case, group and variable ...... 245 Table 6.46: Assumptions: Pearson correlations: Dependent variables ...... 245 Table 6.47: Assumptions: Scale reliabilities: Cronbach’s alpha ...... 246 Table 6.48: Confounds: Tests of Main Effects of Gender & Language ...... 250

Table 6.49: Confounds: Between-subjects Effects of Language*Group ...... 251 Table 6.50: Confounds: Comparison of Means: Athlete variables ...... 251 Table 6.51: Assumptions: Box’s M ...... 253 Table 6.52: Assumptions: Levene’s test of error variances ...... 254 Table 6.53: Assumptions: Homogeneity of Regression ...... 255 Table 6.54: Omnibus F for the Corrected Model ...... 255 Table 6.55: Multivariate Tests of Main Effects ...... 258 Table 6.56: Between-subjects effects of Group, Covariate & Interactions...... 259 Table 6.57: Pairwise comparison of means: Person variables ...... 263 Table 6.58: Paired t-tests: Person variables by Group ...... 266 Table 6.59: Test of difference: Team variables ...... 266 Table 6.60: Tests of difference: DUI means ...... 268 Table 6.61: Summary of Balance models ...... 269

Chapter 1: Off-field behaviours and sport sponsorsorship

In 2009, worldwide sponsorship spending was estimated to reach $44 billion (IEG, 2009a). By 2012, despite the global financial crises, IEG proposed that the global sponsorship industry was forecast to grow “…4.9 per cent to $51 billion” (IEG, 2012) during the course of one year. Historically, the majority of sponsorship investment is paid to sports properties; researchers report that sports receive anywhere between 60 and 85% of all sponsorship spending (see: ABS, 1999; Fenton, 2009; IEG, 2009a).

Sponsorship research has traditionally focused on means to improve practice; in the form of measurement of outcomes, positive image transfer and brand recognition (Chien, Cornwell & Stokes, 2005; Grohs, Wagner, & Vsetecka, 2004; Gwinner & Eaton, 1999). Risk assessment, or the potential negatives associated with consumer evaluations of sponsorship are significantly weaker areas of research, although emerging themes in research suggest that attention to potentially negative sponsorship issues is increasing (see: Parker, 2007; Pope, Voges & Brown, 2009; Wilson, Stavros, and Westberg, 2008).

This chapter provides a general introduction to the research project. It argues that negatives, or risks, associated with sport sponsorship deserve academic attention and research. Behaviours of sports consumers in response to team wins (basking in reflected glory, or, BIRGing: Cialdini, Borden, Thorne, Walker, Freeman & Sloan, 1976) and losses (cutting off reflected failure, or, CORFing: Snyder, Lassegard & Ford, 1986), the actions and ideologies of the news-media, and off-field behaviours of sportspeople are reviewed as uncontrollable factors in the sponsorship contract. BIRGing, CORFing and the news-media are presented as comprehensively researched and theorized subjects. The behaviours of sportspeople, and consumer responses to such within the sponsorship context, are less well understood and therefore present a gap in academic knowledge.

The means to approach off-field behaviours are reviewed in this chapter. Off-field behaviours have been characterised as scandals, and transgressions. They might also be regarded as moral hazards or moral risks in the sponsorship context. Despite the availability of terms to describe off-field behaviours as potential negatives in the sponsorship environment, this chapter concludes that less valenced terms will allow a broader scope for the research project; including the possibility of understanding the impact of positive off-field behaviours on consumer impressions of sponsors.

The chapter concludes by presenting the research questions for this thesis and an outline of the following chapter.

A dearth of negatives

Javalgi, Traylor, Gross and Lampman (1994) introduced the possibility that corporate sponsorship does not produce a ‘halo effect’, or positive evaluations of all elements of the company given consumer awareness of the sponsorship. Instead, they note that individual elements of corporate image may be differently affected by awareness, and, that “corporate sponsorships might exacerbate a negative image” (Javalgi et al, 1994, p57). Despite this early indication that corporate sponsorship might prove problematic for some organisations, little sponsorship research published between 1994 and 2007 specifically considers the possibility of negative sponsorship outcomes in terms of consumer evaluations.

Dean (2002) presents the possibility that consumers attribute both altruistic and ant- altruistic motives to corporate sponsorship and that those attributions contribute to evaluations of corporate community relations (CCR). Although the structural path from anti-altruism to CCR was non-significant, Dean’s (2002) conclusions support that consumers hold positive and negative attributions about sponsorship concurrently.

Other researchers also investigate consumer perceptions of corporate sponsorship as altruistic or commercial (see: Dardis, 2009; Rifon, Choi, Trimble and Li, 2004). Each of these research projects studies cause-related sponsee contexts (real or fictitious).

The use of measures of perceived altruism in these projects reflects the likely objectives of cause-related sponsorships as the development of community relations and perceived altruism.

The literature that has emerged more recently to question sponsorship risks does so with a keen sense of corporate sponsors’ desires to achieve measurable objectives. In the industry-focused literature Cameron (2009), Hutchinson and Bouchet (2010), O’Reilly and Madill (2009) and Pearsall (2010) argue the importance of metrics. They also indicate that the objectives of logo counts, media exposure and sponsor recall are becoming less popular (Cameron, 2009). The development in sponsorship metrics is towards objectives that are less tangible, such as sponsor-sponsee relationships, commitment and consumer brand experiences (Cameron, 2009; O’Reilly & Madill, 2009). However, although the appearance is a move towards qualitative objectives; they remain conceptually related to hierarchies of effects (Lavidge& Steiner, 1961; Palda, 1966) in developing positive evaluations (or attitudes) which are thought to contribute to consumer behaviour.

The importance of relationships in the industry literature is consistent with that emphasised in the academic literature. Value, according to Farrelly, Quester and Burton (2006) should be assessed according to the success of sponsor-sponsee relationships. Success, in that research, is defined as “…satisfaction in terms of whatever objectives had been set for the agreement” (Farrelly, Quester & Burton, 2006, p1019). For Wilson, Stavros and Westberg (2008) managing the sponsor- sponsee relationship involves open communication, sponsee policy development and the monitoring of expectations.

Within this literature, the focus is primarily on the sponsorship agreement and manager’s perceptions, with consumer research less prevalent. Parker (2007) and Pope, Voges and Brown (2009) provide some insight into consumer evaluations of sponsorship negatives. Parker (2007) investigates the direct effects of sponsor misbehavior on consumer evaluations in a university sports team context, and finds that consumers’ level of team identification is unaffected by sponsor misbehavior. Consumer attitudes toward the sponsor, however, are affected by sponsor misbehavior (Parker, 2007).

Pope, Voges and Brown (2009), studied consumer evaluations of sponsor’s brand quality and corporate image given varied frequency of sponsorship stimulus information (repeated, single-exposure, or control), team win/lose information, and brand use during the event information. They found that when sponsorship information is present (versus absent) and frequent (versus single-exposure), consumers provide more positive evaluations of sponsor’s brand quality and corporate image. Pope, Voges and Brown (2009) also found that evaluations of corporate image remain high regardless of win/lose information when sponsorship stimulus information is present (versus absent). Evaluations of sponsor’s brand quality, however, appear to vary with information about the sponsored team’s win/lose information in competition (Pope, Voges & Brown, 2009), such that winning appears to produce higher evaluations of sponsor’s brand quality evaluations whether the product is used in competition or not. Losing, therefore, presents a risk to sponsors that is uncontrollable; the authors suggest that, “…sponsorship should be primarily considered for successful teams only” (Pope, Voges & Brown, 2009, p16).

The studies of consumer evaluations of sponsors’ objectives in the presence of uncontrollable, or risk-laden, sponsorship information presents new research opportunities. Management perspectives on sponsorship, and sponsorship risks is, relatively, well-established (see: Farrelly, Quester & Burton, 2006; Hughes & Shank, 2005, 2008; Westberg, Stavros & Wilson, 2011; Wilson, Stavros & Westberg, 2009). As industry analysts argue that sponsor’s objectives are evolving towards measures of consumer’s sponsorship experiences and evaluations, a concomitant academic research agenda that studies consumer evaluations of sponsorship in the presence of uncontrollable (by sponsors) situations and risky sponsorship contexts is warranted.

The research project reported here investigates news-media reports of athlete's off- field behaviours and the impact of such reports on consumer’s impressions of sportspeople, their teams, and the team’s sponsor. In order to justify this agenda, the following section reviews three issues that present risks to sponsorship success: consumer behaviour in the form of basking in reflected glory (BIRGing) and cutting off reflected failure (CORFing), news-media reporting, and athlete off-field behaviours.

Uncontrollable factors in sport sponsorship contracts

The potential rewards of sports sponsorship are relatively clear. Audience reach, via attendance and television ratings metrics can be measured. Affect for, and commitment to, sports is relied upon in sport sponsorship. The risks in sports sponsorship are less easy to anticipate, harder to measure the impact of, and often far outside of the sponsors’ scope to control. This section identifies predictable risks in sports sponsorship, such as BIRGing and CORFing. It also identifies the less controllable risks in news-media reporting and the off-field behaviours of sportspeople.

BIRGing and CORFing

Basking in reflected glory (BIRGing) (Cialdini, Borden, Thorne, Walker, Freeman & Sloan, 1976; Cialdini & de Nicholas, 1989; Cialdini & Richardson, 1980) refers to the tendency of individuals to take pleasure in, and emphasise their association with the successes of others, even in situations where they cannot claim to have contributed to the success (Cialdini et al, 1976; Hirt, Zillman, Erickson & Kennedy, 1992). Cutting off reflected failure (CORFing) (Snyder, Lassegard & Ford, 1986) represents the alternative, that individuals deny association with an unsuccessful other, for fear that it will reflect badly on themselves, effectively, to threaten their self-concept (Sherman et. al., 2007).

BIRGing and CORFing are explained with reference to the social identity theory of identification and balance theory. Social identity theory reasons that identification involves self-categorisation (Tajfel, 1974). Consequences of identification include that an individual will seek to maintain positive self-esteem and that evaluations of in- and out-group members will be biased (Ashforth & Mael, 1989; Tajfel, 1974). The BIRG/CORF research finds that “higher fan identification resulted in increased tendencies to BIRG and decreased tendencies to CORF” (Wann & Branscombe, 1990, p103). This conclusion is supported by Spears, Doosje and Ellemers (1997) who report that highly identified individuals don’t CORF and that low identification

17 individuals do. As such, high identification encourages the individual to maintain their support of their team in situations of failure. Individuals with lower levels of identification, in contrast, are associated with higher levels of CORFing, or the psychological “distancing” (Sherman et al, 2007, p1109) from failure, and a lesser inclination to BIRG (Wann and Branscombe, 1990).

Balance theory explanations of BIRGing and CORFing have been used by numerous authors (see: Cialdini et al, 1976; Hirt, Zillman, Erickson & Kennedy, 1992; Madrigal, 1995; Miller, 2009). Balance theory suggests that cognitive consistency governs judgments of objects and the relationships between objects (Crandall, Silvia, N'Gbala,Tsang, & Dawson, 2007). In the context of a university sports team, Cialdini et al (1976) explain how evaluations of a student and their university sporting team might be balanced. A balanced, and BIRGing, view of the student and university team occurs: ...if observers perceive a positive unit relationship (e.g., university affiliation) between a student and a successful football team and if observers generally evaluate successful teams positively, then in order to keep their cognitive systems in balance, the observers would have to evaluate the student positively as well. Cialdini et al, 1976, p369 The same application is appropriate for perceptions of CORFing. If the observer regards the team as negative, and the student positive, the observer faces an imbalanced system, which contributes to motivation to distance the self from negative objects or entities.

BIRGing and CORFing research has most commonly been applied to sporting wins and losses (Bernache-Assolant, Lacassagne & Braddock, 2007; Boen, Vanbeselaere & Feys, 2002; Campbell, Aiken & Kent, 2004; Hirt, Zillman, Erickson & Kennedy, 1992), and the winning and losing of political elections (Miller, 2009; Wann, Hamlet, Wilson & Hodges, 1995). The focus on winning does not account for all possible applications of BIRGing. Cialdini et al (1976), for example, provide the following as an example of BIRGing:

At a women's movement forum attended by one of the authors, there was a round of feminine applause when it was announced that Madame Curie was a woman and Lee Harvey Oswald was not. Cialdini et al., 1976, p366 The above illustrates that BIRGing need not exclusively apply to winning. The potential associations that an individual may feel toward symbols or representatives of success; or even, the lack of association with a negative symbol (Oswald) may be the basis of BIRGing. Similarly, CORFing need not only apply to failure as loss.

BIRGing and CORFing have been substantively researched. Each concept represents an interesting observation. However, it is the theories underlying BIRGing and CORFing, identification and balance, rather than the concepts themselves which present the greatest opportunities for future research.

The News-media

This section reviews the role of the news-media as it may constitute a risk to sponsorship contracts. Some regard the making of the news as determined by market forces; failing and superseding business models and the goal of representing topics of consumer interest (Beale, 2006; Denham, 2004; Franklin, 2008). Others regard the making of news as the determining of public opinion, and the role of journalists as deeply ideological; driven by a public trust, belief in democracy and the rights of citizens to information (Dueze, 2005; Glasser, 2009). Included, are questions on the role of the journalist; to report, or to interpret (Baum & Potter, 2008; Chan & Suen, 2009; Ross, 2007), and that to have reported is to have interpreted (Fortunato, 2008; Kiousis & Wu, 2008; Wanta, Golan & Lee, 2004). These divergent perspectives suggest that consumers’ desires drive news-media production, but also that consumers’ are served by expert producers of news who know the public interest. To consider the impact of the news-media, agenda-setting theory and news-values are considered.

Agenda-Setting Theory

The basic premise of agenda-setting theory is that “the media can affect what issues the public think about and how they think about those issues” (Fortunato, 2008, p118). This claim originates in a much cited (see: Kiousis & Wu, 2008; Wanta, Golan & Lee, 2004; Weaver, 2007) statement from Cohen (1963) that suggests that the media tells news consumers which issues to think about, if not how we should be thinking about each of those issues. First-level agenda-setting, or classical agenda- setting states that media coverage of an issue suggests the salience, or importance of that issue to the public (Kiousis & Wu, 2008; McCombs & Shaw, 1972; Wanta, Golan & Lee, 2004). Effectively, the selection of an issue for production as a news story implied the importance of that issue, and varying levels of coverage suggested the relative importance of different stories (Wanta, Golan & Lee, 2004). Cohen's early argument has since been superseded by more complex analyses of media influence.

Second-level agenda-setting suggests it is insufficient to suppose that the reporting of the object of the story itself has an influence on public perceptions of salience. It is necessary, instead, to think also about the attributes of the object, which include both substantive and affective elements (Kiousis & Wu, 2008; McCombs, Llamas, Escobar-Lopez & Rey, 1997; Wanta, Golan & Lee, 2004). The “substantive elements refer to those characteristics of communication messages that help us to cognitively structure and discern among various topics” (Kiousis & Wu, 2008, p62). These substantive elements have sometimes been equated with frames, hence the inclusion of frame theory within agenda-setting theory (Fortunato, 2008; Kiousis & Wu, 2008; Weaver, 2007).

To frame, according to Entman, is to “select some aspects of a perceived reality and make them more salient … to promote a particular problem definition, causal interpretation, moral evaluation, and/or treatment recommendation” (1993, p52). Frames have alternatively been represented as “cognitive structures consisting of systems of classification and rules of interpretation” (Paterson, 2007, p1087-1088); or “the central organising idea for news content that supplies a context and suggests what the issue is through the use of selection, emphasis, exclusion, and elaboration”

(Tankard, Hendrickson, Silberman, Bliss & Ghanem,1991, p3). Examples of frames include the ‘Cold War frame’ (Weaver, 2007, p142); 'conflict' and 'US involvement' frames in foreign relations (Kiousis & Wu, 2008); or 'social policy' 'sport/celebrity stories', and others (Paterson, 2007).

The other element that contributes to second-level agenda-setting is the affective element which refers to “the valence dimension of attribute salience” (Kiousis & Wu, 2008, p62). Affective attributes are measured according to the positive, neutral or negative tone used in news stories (Kiousis & Wu, 2008; Wanta, Golan & Lee, 2004). The impact of affective valence is studied using content analyses (frequencies of value-laden terms), correlated with object-affect scores in survey research (see: Kiousis & Wu, 2008; Wanta, Golan & Lee, 2004). The results of these correlations indicate that the larger the quantity of positive information about an object news- consumers receive, the more likely those consumers will have a positive view of the news-object (Wanta, Golan & Lee, 2004).

Overall, second-level agenda-setting theory now significantly disagrees with Cohen (1963). According to Wanta, Golan and Lee, “first-level agenda setting suggests media coverage influences what we think about, second-level agenda setting suggests media coverage influences how we think” (2004, p367).

News Values

Media frames suggest how news stories might be categorized by subject, or how they might be understood with reference to context. News-values contend that substantive story elements are inherently more or less newsworthy.News-values answer the editorial questions of ‘which stories?’ and ‘why?’ The most frequently cited (see: Harcup & O’Neill, 2001; Lee & Choi, 2009; Niblock & Machin, 2007) work on this topic is that of Galtung and Ruge (1965) who explained why particular international events were reported in domestic newspapers. They identified twelve ‘news criteria’ that represent an underlying framework for assessing newsworthiness.

News criteria, for Galtung and Ruge (1965), are likened to the problem of selective attention in the communication process. They suggest that “since we cannot register

21 everything, we have to select, and the question is what will strike our attention” (Galtung & Ruge, 1965, p65). The criteria developed by Galtung & Ruge (1965) are provided in Table 4.12 alongside the modern news values developed by Harcup and O’Neill (2001).

Table 1.1: Old vs Modern News-values Galtung & Ruge (1965) Harcup & O’Neill (2001) Frequency The power elite Threshold Celebrity Unambiguity Entertainment Meaningfulness Surprise Consonance Bad news Unexpectedness Good news Continuity Magnitude Compositional balance Relevance Elite nations Follow-up Elite people Newspaper agenda Personification Negativity

Despite the similarities across the lists of news values (continuity/follow-up; magnitude/threshold; unexpectedness/surprise; elites, etc), the addition of both celebrity and entertainment in the modern list is informative. The additions suggest that a person is not interesting only in relation to their potential to influence government, business or international affairs. Other ‘modern’ lists of news values have been produced by Herbert (2000), and Hetherington (1985). Each of these lists contains topics like sex, surprise or novelty, crime or scandal, ‘human interest’, humour, or animals. Such news values suggest the importance of such debates as the ‘tabloidization’ of news-media reporting (Bek, 2004; Franklin, 2008; Johansson, 2008).

The application of news values to the question of newsworthy issues in sport sponsorship is to ask how and why athletes and their off-field behaviours might be reported. The result of such a question is to acknowledge that off-field incidents involving athletes appeal to numerous news values.

In the realm of modern news values, professional athletes are celebrities (see: Martin

& Bush, 2000; Crepeau, 1981; Jones & Schumann, 2000; Stone, Joseph & Jones, 2003). The consumption of celebrity news, however, is Janus-faced. It is seen as “entertaining”, as “fun”, and as performing a “cheering-up function” for news consumers (Johansson, 2008, p407), but it can also be seen as “…an attack on social privilege” (Johansson, 2008, p408). The attitudes of consumers reflect a delight in not just the sports results, but what Johansson calls ‘transfer gossip’, or the “features that would go beyond the sport itself” (Johansson, 2008, p405). Among Johansson’s many transcript excerpts is the quote from “Daniel, 35”: …No one really wants to know if … sacked their manager. But everyone wants to know what Ferguson’s done, or what Chelsea are doing. Or someone putting coke up their nose. These… That’s the sort of thing you wanna know what’s going on in football. Johansson, 2008, p405 Whether knowing someone is using cocaine is an example of entertainment or a delight in the fallibility of the otherwise privileged is not clear. A possibility is that in collecting such pieces of news, the news consumer is developing an associative schema of what it is to be a sportsperson, or celebrity. At present, there is no research that reports the implications of consumer desires for such information for use in judgments of teams and sponsors.

Off-field behaviours

Where sponsors are not capable of controlling BIRGing and CORFing, and often have little influence over news-media information dissemination; they do have the ability to exercise some control over off-field risks associated with their sponsees. It is the recognition of the risk, as well as the ability to exert some influence, that makes off-field risks an important subject for sponsorship research.

Sponsorship research has not delved into the off-field risks posed to sponsors in much depth. This situation is beginning to change. Interest in non-sport influences on the sponsor’s brand has been approached from a managerial perspective (Hughes & Shank, 2005, 2008; Kahuni, Rowley & Binsardi, 2009; Wilson, Stavros & Westberg, 2008), in relation to celebrity brand DNA (Johnson Morgan, Summers &

Sassenberg, 2008; Sassenberg & Johnson Morgan, 2010), and considering the influence of team identification on consumer perception (Fink, Parker, Brett & Higgins, 2009; Parker, 2007), or from a public relations’ perspective (Brazeal, 2008; Dimitrov, 2008; Fortunato, 2008). Many of the papers that discuss off-field behaviours are conceptual; those that report research are discussed below.

Hughes and Shank (2005, 2008) provide the earliest insight into sponsorship ‘scandals’; use of the term scandal will be discussed in the following section. These authors interviewed 10 sponsors and media representatives to develop four characteristics of scandals. Their respondents expressed concerns about the criminality of behaviours, sports integrity, impact on the game, and the level of sport at which the ‘scandal’ occurs (professional vs amateur) (Hughes & Shank, 2005). Overall, their findings indicate concern for public perception that is driven by audience awareness of the issues, and whether the business of sport and sponsorship will continue to operate without sanction (legal or otherwise).

Hughes and Shank (2008), in other research, found that scandals in college athletics had a variety of influences on charitable donations, and alumni gifts to sports programs. Data related to fifteen US universities showed that charitable support was often unaffected by scandals, particularly when those scandals related to staff, rather than student athletes. Similarly, gifts or bequests from alumni were quick to recover from scandals, or suffered no downturn (Hughes & Shank, 2008). Enrolments and applications to study at the universities also appeared robust (Hughes & Shank, 2008). These results suggest the importance of understanding the attributions that people make about the responsibility for various sports-related scandals.

Other research that focuses on managers’ perspectives, include Kahuni, Rowley and Binsardi (2009), and Wilson, Stavros and Westberg (2008). Kahuni, Rowley and Binsardi (2009) conducted a content analysis of news coverage relating to allegations of spying in Formula 1, by the Vodafone McLaren-Mercedes F1 team. They found that coverage typically referred to the spy allegations using the ‘Vodafone McLaren- Mercedes’ name (Kahuni, Rowley & Binsardi, 2009). They concluded that title sponsorship presents a credible risk to sponsor’s brands in cases of negative image spillover (Kahuni, Rowley & Binsardi, 2009). However, they also suggest that

Vodafone’s lack of public outrage could be interpreted as team support, and their strategy as ‘business as usual’ (Kahuni, Rowley & Binsardi, 2009, p60).

Wilson, Stavros and Westberg (2008) also present support for the ‘business as usual’ strategy. Following interviews with Australian sports marketing executives, they conclude that sponsors want teams to provide full and fast disclosure of on- and off- field transgressions (Wilson, Stavros & Westberg, 2008). Policies and programs to manage player behaviour and compliance are regarded as increasingly important to sponsors, and media coverage of transgressions is regarded as a powerful influence on public opinion (Wilson, Stavros & Westberg, 2008).

Sponsorship research on consumer perceptions of off-field behaviours of sportspeople is currently very limited. Parker (2007) provides the only insight into consumer responses to off-field behaviours thus far. Parker’s PhD thesis (2007) investigated the effect of sponsor misbehaviour on sport consumers’ team identification scores (categorised as high or low) and attitude toward the sponsor. As reported earlier, this research found that subjects’ with high (versus low) team- identification scores had more positive attitudes toward the sponsor (Parker, 2007). It also confirmed that negative sponsor information influenced attitudes toward the sponsor (Parker, 2007).

Research conducted by Fink et al (2009) studies the negative off-field behaviours of athletes as they influence identification with the athlete’s team. It does this using a sport management perspective, and does not mention sponsorship. The 2 (Identification: high/ low) x 2 (Leadership response: weak/ strong) design evaluated the effect of management response on team identification (Fink et al, 2009). Results of this research show that level of identification for pretest low identification subjects did not change significantly given a negative information stimulus; high identification subjects, however, produce mixed responses (Fink et al, 2009). These mixed results indicate that in the presence of a strong ‘leadership response’, (information that team managers act quickly, condemn poor behaviour and act openly), team-identification scores are stable (Fink et al, 2009). However, a weak ‘leadership response’, including slow response, denial of responsibility and opaque discipline policies, contribute to a significant fall in team-identification scores (Fink

25 et al, 2009). These results support the public relations’ approaches to crisis management that advocate speed, transparency and public apology (see: Brinson & Benoit, 1999; Fortunato, 2008).

In its entirety, the field of research on the off-field behaviours of sportspeople suggests the need for theoretical work in this area. Research that evaluates the opinions of sponsorship managers and media representatives suggests that sponsors are increasingly wary of sports sponsorship when teams lack appropriate policies to manage player behaviour (Hughes & Shank, 2005; Wilson, Stavros & Westberg, 2009). This remains an important area for research, especially given the small sample sizes used to date. Consumer research, however, has been an even smaller area for research, and this should change. Understanding of consumer responses to the off- field behaviours of sportspeople is currently limited to one example (see: Fink et al, 2009) which does not consider the implications for sponsors.

Defining the off-field context: scandalous morality

Having recognised the need for consumer research addressing the question of the effects of athlete off-field behaviours on sponsorship objectives, an important next step is to consider how the off-field context should be defined. Previous research has referred to the activities in a variety of ways. They have been called ‘scandals’ by some (Dann, 2007; Fortunato, 2008; Hughes & Shank, 2005, 2008), ‘transgressions’ by others (Sassenberg & Johnson Morgan, 2010; Wilson, Stavros & Westberg, 2008). They might otherwise be thought of as ‘moral hazards’ (see: Baker, 1996- 1997), or moral risks, as evidenced in morals clauses in sponsorship contracts (George, 2009; Kressler, 2005-06; Reed, Bhargava, Gordon & Kjaer, 2010). Each of these terms is considered briefly here.

Crompton suggests that “implicit in any arrangement designed to be mutually beneficial is the risk that anticipated benefits may not be realised or, worse, that the relationship may lead to negative outcomes” (Crompton, 1994, p65). Sponsorship, in this sense, is not different from any other business agreement. Sport sponsorship, however, might present a more volatile example than other sponsorship contexts.

Johnson Morgan, Summers and Sassenberg argue that the “things that make a sport celebrity attractive as a marketing investment” (2008, p2) are the very elements that make sportspeople and sporting clubs risky propositions for sponsors. Media attention paid to sportspeople, money, lifestyle factors and the attention of fans (Dann, 2007; Hughes & Shank, 2008; Johnson Morgan, Summers & Sassenberg, 2008) all contribute to the potential for ‘scandals’.

Scandals and Transgressions

This focus on ‘scandal’ as a description of events occurring on-field, off-field and in event management that create risks to sponsors objectives is a curious one. The term, ‘scandal’, is used by a number of authors from both marketing and public relations disciplines (see: Dann, 2007; Hughes & Shank, 2005, 2008; Fortunato, 2008; Kahuni, Rowley & Binsardi, 2009; Reisinger, Grohs & Eder, 2006; Sassenberg & Johnson Morgan, 2010). Despite its use, ‘scandal’, is a term that is not often explained. Hughes and Shank (2005) developed criteria to define a scandal based on four characteristics that were consistently referred to in in-depth interviews. They found that scandals were associated with: …an action that was either illegal or unethical, involves multiple parties over a sustained period of time, and whose impact affected the integrity of the sport with which they are associated. Hughes & Shank, 2005, p214 These characteristics, individually, are less emotive and less subjective than ‘scandal’. They provide a means to identify and measure sources of risk to sponsorship objectives.

Using Hughes and Shank’s (2005) scandal characteristics we should expect that sponsorship objectives are threatened in a variety of ways. Both illegal and unethical activities might be expected to threaten sponsorships, such activities should be relatively easy to recognise. The involvement of ‘multiple parties’ suggests collusion or conspiracy, and a systemic, possibly organizational, risk to the sponsor. Behaviour that is sustained over a period of time suggests deliberate, rather than accidental breach. The final characteristic identifies the culmination of the previous points,

27 which is an impact on the sport, which, by extension, might include the sponsor. Each of these points has some merit; ‘scandal’ remains problematic because has been used to aggregate a number of independent characteristics that may correlate, but should not be required to co-exist. A less evocative term, used by Aaker, Fournier and Brasel (2004); Wilson, Stavros and Westberg (2008); and, Sassenberg and Johnson Morgan (2010), is ‘transgression’.

A transgression, according to Aaker, Fournier and Brasel, is “a violation of the implicit or explicit rules guiding relationship performance and evaluation” (2004, p100). This definition is less salacious than the intrigue-filled ‘scandal’. A difficulty with this term is that its application requires an understanding of the context to which it is applied. For an individual to anticipate violating an implicit rule they would have to have grounding in the values of the group. In relation sports, and sport sponsorship, this is problematic. Employment contracts and sponsorship contracts for athletes provide explicit rules for behaviour. Implicit rules must then be either social norms of ‘good behaviour’, legal standards of ‘duty of care’ or some ephemeral code of conduct that an individual might reasonably contend they knew nothing of. The ‘implicit’ problem, relating to norms or duty of care, may refer to moral problems.

Moral hazards, Morals clauses and Moral risks

There are several ways to approach moral problems. ‘Moral hazard’ is a term used most frequently in the insurance industry and is explained as a risk-related problem within principal-agent relationships (Baker, 1996-97; Caves, 2003; Holstrom, 1979, 1982; Stiroh, 2007). The moral hazard argues that when the burden of risk is borne by the principal, dependent upon “unobservable” behaviour of the agent, moral hazard exists as the temptation for the agent to maximise their individual utility (Baker, 1996-1997; Holstrom, 1982). In the sponsorship context, moral hazard might suggest that teams and players that receive sponsorship dollars are enabled to continue their activities, effectively ‘insured’ against commercial loss. Sponsors, however, while trying to achieve corporate image objectives of their own, are exposed to risk through the same association.

Baker (1996-1997) provides that the use of ‘moral hazard’ is increasingly universalised beyond the scope of insurance contexts, although retaining the idea of insurance in its application. The most common solutions to moral hazards are incentives and punishments which reduce the attractiveness of cheating, or increase the appeal of contract compliance for the agent (Arrow, 1963; Baker, 1996-1997; Cairns, Jennett & Sloane, 1986; Holstrom, 1979; Mirrlees, 1999).

Use of the term ‘moral hazard’ in sponsorship contracts is problematic. Although the sponsoring company may be regarded as an insurer of the commercial success of the sponsee (as principal); it may also be regarded as a seeker of insurance as it attempts to develop brand image or other corporate objectives through the sponsee. It is when the sponsor takes the position of insurance-seeker that the moral hazard application to the sponsorship context falters. When surveying off-field behaviours of sportspeople, it is not the behaviour of the sponsor that threatens the principal-agent relationship, but the insurer or player-team-related organisation.

Although moral hazard is an inappropriate label to the risk faced by sponsors, morality remains an important theme. Literature on sports law refers to ‘morals clauses’ included in contracts for athletes, actors and entertainers contracted for a variety, (e.g., advertising, endorsement and sponsorship) of purposes (Auerbach, 2005-06; Kressler, 2005-06; George, 2009). Although this literature refers to endorsement, the application of morals clauses has a much wider scope. Augustine- Schlossinger, for example, demonstrates that terminology is used loosely; “These sponsorships, generally called endorsements…” (Augustine-Schlossinger, 2003, p281). Kressler notes the use of morals clauses in such areas as endorsement, advertising, and employment, relating to such individuals as actors, athletes, musicians, news readers, and a railway conductor (see: Kressler, 2005-06, p247). This literature identifies the increasing popularity of morals clauses, particularly for high risk populations, which appear mostly comprised of young men earning large salaries (see: Auerbach, 2005-06; Shilts, Jett & Desiato, 2007-08).

The purpose of morals clauses is closely aligned with the maintenance of corporate strategy. Kressler (2005-06) establishes that use of endorsements is driven by a desire to differentiate and that companies “turn to celebrities as a relatively cheap,

29 easy and familiar way to manipulate consumer perception of a product” (Kressler, 2005-06, p240). Auerbach also uses common marketing objectives to explain the importance of morals clauses, including that contracts are determined by “the company’s intended image…” (2005-06, p3); out of a “…growing concern for corporate reputations” (2005-06, p15). Morals clauses allow a company: …to cancel the agreement in the event the athlete does something to tarnish his or her image, and consequently, the image of the endorsee or its products. Auerbach, 2005-06, p3. The purpose of morals clauses, then, serves to protect the company that is paying to achieve image or reputation-based objectives. The types of behaviours that morals clauses seek to control are broad, and several classes of contract might exist for athletes posing different levels of risk, or exerting different levels of star power (Auerbach, 2005-06; Kressler, 2005-06).

George (2009) and others (see: Auerbach, 2005-06; Kressler, 2005-06; Reed et al, 2010) relate that companies employing morals clauses aim to avoid disrepute or poor public opinion. Specific behaviours may not be stipulated because of their ability to limit the scope of athlete responsibility. Instead, clauses may stipulate effects of behaviour as defining conditions: …more than just a requirement to obey the law, but also to refrain from behaviour that tends to ‘shock, insult and offend the community and public morals and decency’, bring the artist into ‘public disrepute, contempt, scorn and ridicule’, or hurt or prejudice the interests of, lower the public prestige of, or reflect unfavourably upon, the artist’s employer or the industry in general. Kressler, 2005-06, p245 These terms allow that companies may seek remedies for broad contractual breaches. The advantage of the legally regulated morals clause, over a definition of transgression, is that there is a basis from which to infer the ‘implicit’ rules.

In the Australian context, George (2009) explains the test for disreputable behaviour, as outlined by the Court of Arbitration for Sport (CAS) is: …conduct bringing a person into disrepute must lower the reputation of the person in the eyes of ordinary members of the public to a significant extent… George, 2009, p30.

This test is a measure of public perception; it does not consider residual effects on the athlete’s sporting code, their team or sponsors. George provides also that, “Aggression, public drunkenness and other antisocial behaviour, even if not criminal, may be disreputable” (2009, p30); all of these examples are commonly reported in Australian newspapers (George, 2009).

Concluding on terms

Each of the available terms provides an opportunity for a particular type of research. Scandal argues attention to the conditions contributing to, and outcomes of, acts judged scandalous; transgression, attention to how implicit and explicit rules are communicated, and their influences. Moral hazard, in the sponsorship context, allows consideration of the sponsor as agent and sponsor as client. Morals clauses imply that identifiable populations are likely to threaten the objectives of the principal.

In the cases of scandal and transgression a ‘wrong’ exists prima facie. In the cases of morals clauses and moral hazard, the law is used to assert legal rights and prevent ‘wrongs’. Each of the available terms includes a notion of obligation and duty and ideas of right and wrong. The moral perspective is considered in this thesis, (see: Chapter 5 for the pretest of a cue diagnosticity measure). Morality is rejected as the guiding ethic for this research because of its emphasis on right/wrong and good/bad which are regarded to have polarizing effects on judgments. Athlete off-field behaviours will be referred to throughout this thesis as ‘off-field behaviours’, ‘behaviours’ or ‘acts’ with the aim to avoid cueing judgments.

In the quagmire of terms, a marketing question exists. An individual athlete or team member, whose behaviour is disreputable, may lose their sponsorship if their sponsor objects to the behaviour. These contract cancellations occur because sponsoring companies expect that disrepute will be associated with the sponsor or the sponsor’s brand. The question for this thesis is not, why sponsors believe this image transfer might occur, but to question whether a risk, in the form of negative consumer perceptions, is present in tests of consumers’ perceptions. The consumer, their perception of, and response to off-field behaviours are the focus of this research.

Removing the judgment terms allows that potentially positive acts, as well as potentially negative acts, can be considered in this research.

Research Agenda and Research Questions

The question for this research asks: Are consumers’ evaluations of an athlete, the athlete’s team and the team’s sponsor influenced by news-media reports detailing the athlete’s off-field behaviours?

The sport investigated in this research is the Australian National Rugby League (NRL) competition. This limitation considers the location of the researcher (Queensland) and the characteristics of the local population. While Australia is known for its love of sport (ABS, 2009a); not all sports are universally popular across all states and territories (ABS, 2009a). Statistics reveal that Australian Rules football (the Australian Football League: AFL) is the most popular sport in Australia by attendance at sporting events (ABS, 2009a). In Queensland, rates of attendance at AFL games are below average at 7.4% (ABS, 2009a). In comparison, attendance rates at Rugby League are 16% overall (or 20% for men only) (ABS, 2009a). These statistics make Rugby League the most popular sport in Queensland by attendance, which supports that the local population is more likely to be knowledgable, or interested, in the purported behaviours of the local team.

In developing the research question several preliminary questions are addressed. They are outlined here according to their answering by literature review or answering through primary research.

Questions addressed in the literature review are: a. What characteristics of sports make them desirable sponsorship properties? b. What objectives drive sponsorship? c. What theories explain how sponsorship ‘works’ in the minds of consumers to achieve sponsorship objectives?

The review of literature summarises three groups of sponsorship objectives: communications (exposure and awareness), consumer evaluations (attitudes and image), and behavioural. The literature review further identifies important theories that explain how sponsorship ‘works’. Social identity theory (Tajfel, 1974, 1978) provides an understanding of how consumer evaluations of in- and out-groups (liked sports teams, and disliked teams) will vary according to the individual’s level of identification. Heider’s (1958) balance theory is also used to consider the individual’s desire for consistent judgments of linked entities (sponsor, team and athlete). These theories provide the foundation for the research reported and have contributed to the overall research questions for this project.

Elements within the research question, and issues raised in the literature review prompt several research questions that must be answered prior to the address of the research question. These preliminary questions upon which the overall research is contingent include:

RQ1. What types of athlete off-field behaviours are reported? RQ2. Removed from news-media and sports contexts, how do individuals evaluate the off-field behaviours reported? RQ3. If group (team) identification has the power to influence consumer evaluations, how should team identification be measured for this research?

When the research question is directly addressed in Study 2, it acquires two theoretical sub-questions. These are shown below. RQ4. Are consumers’ evaluations of an athlete, the athlete’s team and the team’s sponsor influenced by news-media reports detailing the athlete’s off-field behaviours? a. Does team identification moderate evaluations? b. Are evaluations balanced?

A content-by-chapter table, on the following page, summarises the research questions and provides an introduction to the studies.

Table 1.2: Research questions & goals Research question Research Method Reporting goals Chapter design Literature  Background to the sponsorship industry 2 Review  Sponsorship objectives  Important characteristics of sports  Theories that explain sponsorship outcomes Conceptual Framework 3 RQ 1: What types of athlete off-field behaviours are Exploratory Study 1:  List of athlete off-field behaviours 4 reported? Descriptive Content  Category frequencies & comparisons Analysis RQ 2: Removed from news-media formats and the Descriptive Pretest 1:  Evaluation of behaviours using cue 5 sports context, how do individuals evaluate the off-field Survey diagnosticity measure behaviours reported?  Evaluation of behaviours using attitude toward the act measure  Establishing the the range of evaluative valence for behaviours tested RQ 3: How can team identification be measured for this Descriptive Pretest 2:  Test of identification scale items 5 research? Survey  Development of a reduced identification measure RQ 4: Are consumers’ evaluations of an athlete, the Explanatory Study 2:  Test of news’ article stimuli influence on 6 athlete’s team and the team’s sponsor influenced by Experiment evaluation of sponsor, team and athlete news-media reports detailing the athlete’s off-field variables behaviours?  Establishing the impact of team identification on evaluations  Establishing the extent of evaluative balance Conclusions, Limitations and Future Research 7

Delimitations

The effects of news reports of off-field behaviours of sportspeople on consumer attitudes and perceptions, presents a broad field for research. Three conditions that delimit this project are presented below.

The first relates to the disciplinary focus of the research. This project is not grounded in the discipline of economics; nor is the purpose of this project to query the theory of ‘moral hazards’. Further, while this study has implications to public relations literature, it does not take a public relations perspective. The public relations literature on crisis management is well developed; the view taken in this research is that companies have adequate advice as to how to respond to potential sporting ‘scandals’ as ‘crises’. The question for this research remains a sponsorship question focused on whether the achievement of sponsorship objectives, as measured through consumer evaluations, is threatened by off-field behaviours of sportspeople.

The second delimitation to this project entails a focus on the behaviours of sportspeople that present as moral risks to sponsors. The behaviours of team officials, managers, umpires, and event organisers are excluded from investigation. The justification for this lies in the bases of consumer-event relationships as driven by affect for the event, sport or team. The behaviours of officials, umpires and managers are considered secondary to consumer responses to behaviours of sportspeople.

The final delimitation is one of practicality. This research focuses on the Australian National Rugby League (NRL) competition. This delimitation considers the location of the researcher and the characteristics of the local population; it has previously been noted in the Research Questions section.

Conclusions

This chapter has provided an introduction to the research context presented as the effects of off-field behaviours of athletes on sponsor’s goals. Scandal, transgression, moral hazard and moral risk have been rejected as frames for the research. Off-field behaviours, including both positively and negatively perceived behaviours are investigated in this research.

The research questions have developed from the growing literature that suggests sponsors, and sports sponsors in particular, are becoming increasingly wary of risks to corporate objectives through negative associations. This thesis aims to develop an understanding of consumer perceptions of off-field behaviours in order to establish the extent of risk to sponsors objectives when the news-media reports athlete off- field behaviours in concert with sponsor’s brand names.

The following chapter provides a literature review which provides background information about sponsorship from a marketing perspective. Sponsorship is defined and the size of the global sponsorship industry is discussed. The fears of sponsors in being associated with athlete’s negative off-field behaviours are presented as fears for the achievement of corporate objectives; correspondingly, common sponsorship objectives are reviewed. Finally, the literature review will expand upon the issue of how sponsorship ‘works’ to achieve its objectives. This section presents the information processing model, social identification theory, and balance theory to explain processes of consumer associative memory which has the potential to link negative athletes with sponsor’s brands.

Chapter 2: Review of Literatures

The potential threats posed by the off-field behaviours of players derive from the expectations and objectives that sponsors conceive for their sponsorship efforts. The logic here is that off-field behaviour can only be regarded as sufficiently perilous to justify contract cancellation if it can be linked with potentially negative outcomes for the sponsor, or the failure to achieve the positive rewards expected from investment in sponsorship. The structure of this review of literature aims to answer three questions. These are: a. What characteristics of sports make them desirable sponsorship properties? b. What objectives drive sponsorship? c. What theories explain how sponsorship ‘works’ in the minds of consumers to achieve sponsorship objectives?

The structure of this review first provides a background to sponsorship research; providing a definition of sponsorship and discussion of global sponsorship spending. Sports sponsorships are next introduced as a specific application within the sponsorship market.The value of sports sponsorships, over alternate properties, is argued from the perspectives of audience reach (attendance and mediexposure) and affect reach. These sections are followed by a review of the objectives of corporate sponsorships. These objectives are specifically linked to the hierarchy of effects, which is followed by a more detailed evaluation of sponsorship research that discusses the means of measuring achievement of objectives.

The final section is devoted to key theories that explain how sponsorship objectives are achieved. This section identifies mere exposure, cognitive consistency, and paired learning paradigms as answering ‘how’ sponsorship works in the mind of consumers. The information processing model is central to each of these theories and is provided as a foundation.

Theories that explain the achievement of sponsorship objectives are identified as:

37 mere exposure; learning theories of classical conditioning and associative memory; consistency theories of congruence and balance theory; and, social identity theory. Heider’s balance theory (1958) and the social identity theory of identification (Tajfel, 1974, 1978) are identified as key theories which explain evaluations of athlete off- field behaviour and associated evaluations of their team and sponsors.

A Background on Sponsorship

Sponsorship defined

There are many definitions of sponsorship; elements common to most include: the use of a third party to host a corporate brand; the idea of a reciprocal benefit acquired by that host (financial or in kind); and the purpose of achieving corporate objectives which may include access to a consumer market. A definition of sponsorship provided by Javalgi, Taylor, Gross and Lampman, says that sponsorship is “the underwriting of a special event to support corporate objectives by enhancing corporate image, increasing awareness of brands, or directly stimulating sales of products and services” (Javalgi, Traylor, Gross & Lampman, 1994, p48). This definition takes into account the fee paid to the sponsee (the recipient of corporate support), but does not clarify the purpose of that fee, which may otherwise take the form of a salary or provision of resources or in-kind support. It is also too narrow in scope; in identifying sponsorship as event-related it does not consider individual or team sponsorships which may operate beyond an event or series of events.

An extended definition of sponsorship is provided by Pope: Sponsorship is the provision of resources (e.g., money, people, equipment) by an organization (the sponsor) directly to an individual, authority, or body (the sponsee), to enable the latter to pursue some activity in return for benefits contemplated in terms of the sponsor’s promotion strategy… which can be expressed in terms of corporate, marketing or media objectives. Pope, 1998 This definition has elements in common with those provided by Meenaghan (1983) and Javalgi et al (1994). Its value is in providing additional detail. Pope’s definition

38 includes sponsorship of individuals, organisations or events; and, it identifies the purpose of sponsorships as contributing to the operation of the business or role of that individual, organization or event. This allows that sponsorship payments or contracts may be distinguished from endorsements which also serve corporate objectives and may involve fees to individuals or event-organisations, but generally do not act to directly contribute to an individual’s ability to participate in an event, or an event’s financial viability.

The Sponsorship Model

Figure 2.1: The Sponsorship Model

Sponsorship-linked marketing

Consumers/ Sponsee Sponsor Audiences

Event participants

Event characteristics

Event type

Sources: Cornwell & Maignan, 1998; Cornwell, Weeks & Roy, 2005; Gwinner, 1997; Meenaghan, 2001

Definitions of sponsorship generally have not included direct reference to sponsee audiences or consumers. The link between the sponsor and sponsee consumers in sponsorship definitions has been implicit, using reference to corporate objectives to indicate the importance of audiences.

When sponsorship relationships are modeled (see Figure 2.1 above), there are three principal entities. The first is the company providing a sponsorship fee or other form of support (sponsors); the second are the event-related recipients of financial or other support (sponsees); finally, there are consumers or audiences of the sponsees. What

39 the model makes clear is that direct relationships or interactions exist between sponsor and sponsee, and between sponsee and audience, however, relationships between the sponsor and audiences are indirect, and mediated.

Sponsorship fees paid to sponsees are often invisible to consumers. When sponsorships are visible, as in the case of team or event naming rights, or the badging of shirts, sport strips, equipment or event infrastructure, they have the potential to generate consumer awareness or attention. The research of Bennett (1999) and Pitts and Slatterly (2004) identifies that in tests of event-sponsor recall, event attendees’ exhibit false recall and yea-saying when asked to name event sponsors. Bennett’s (1999) research establishes further that recall of sponsorship signage differs according to level of event involvement (measured according to frequency of event attendance). Together, these papers suggest that sponsorship used alone is not well suited to generate strong associations or memory for sponsor-sponsee links.

Other sponsorship researchers have indicated that sponsorship is a low elaboration medium, or subject to low levels of consumer information processing (Cornwell, Weeks & Roy, 2005; Lardinoit & Derbaix, 2001; Pitts & Slatterly, 2004). These views endorse Cornwell’s view that ‘sponsorship-linked marketing’ (see: Cornwell, 1995; Cornwell & Maignan, 1998; Cornwell, Roy & Steinard, 2001) is required to raise awareness for sponsorship programs. Sponsorship-linked marketing, ideally, makes visible a sponsorship relationship to consumers. It is defined as, “…the orchestration and implementation of marketing activities for the purpose of building and communicating an association to a sponsorship” (Cornwell, 1995, p15).

This notion of sponsorship-linked marketing generally includes the roles of sponsorship leverage and activation. Kinney and McDaniel (1996) provided one of the earliest explanations of leverage, saying that it is the practice of using supporting funds on other forms of marketing communications, to raise the profile of a sponsorship. They go so far as to say that this spending can be dollar for dollar matching of sponsorship and support funds (Kinney & McDaniel, 1996, p250). In 2009, Cameron explains that activation costs associated with sponsorships can be a ratio as high as 3:1 (Cameron, 2009); specifically, “…for every US$1 spent on rights fees, US$3 is spent on marketing support” (Cameron, 2009, p133).

If any confusion in terms appears to exist here, it is because leverage and activation are sometimes used interchangeably (DeGaris, West & Dodds, 2009; Weeks, Cornwell & Drennan, 2008). In their correct use, leverage most often refers to funding (DeGaris, West & Dodds, 2009). Activation otherwise refers to the creation of awareness of the sponsorship through other forms of marketing communications (DeGaris, West & Dodds, 2008), or, refers to situations where, “…the potential exists for audiences to interact or in some way become involved with the sponsor (Weeks, Cornwell & Drennan, 2008, 638).

Articulation of sponsorship relationships (see: Coppetti, Wentzel, Tomczak & Henkel, 2009; Weeks, Cornwell & Drennan, 2008) and the development of consumer brand/sponsorship experiences and engagement are advocated means to increase consumer elaboration and memory for sponsorships. There remains, however, actors and issues outside of the sponsorship contract that influence consumer attention to, and memory for sponsorships. Borland and Macdonald (2003) note the influence of weather, competition scheduling, availability of television broadcast and stadium atmospherics as contributing to quality of event viewing; each of these also has the potential to influence attention to sponsorship messages. Competing sponsorship messages also produce ‘noise’ in the sponsorship environment (Crompton, 2004; Preuss, Gemeinder & Seguin, 2008).

The effects of competitive messages should not be underestimated when considering message noise. To provide a brief example, the National Rugby League (NRL) competition, in 2012, has 18 sponsors. These include: Bigpond, NSW Destination, Harvey Norman, Home Timber & Hardware, VB (beer), Bundaberg Rum, tab sportsbet, keno, Toyota, CocaCola, AAMI, nib, Sydney Olympic Park, Powerade, Mother, Holiday Inn, and OPSM. The, local, Brisbane Broncos, have 38 sponsors, including: XXXX (beer), Nike, sportingbet, arrow energy, CocaCola Zero, Powerade, The Bronco’s League Club, WOW Sight’n’Sound, and many others. For any NRL match the Broncos participate in, the number of sponsors is likely to approach one hundred. Not all of those sponsors has a large match presence, but the message noise and message competition (especially where sponsors compete in the same product category) is large.

The potential for sponsorship messages being confounded by the news-media is also present. The news they provide to event/sport consumers is thought to have the potential to influence consumer perceptions of sport events, teams, and sponsors.

The sponsorship model illustrates some of the assumptions present in the sponsorship environment. These include an indirect path to consumer influence. The model also highlights that the attention of sponsors and consumers may not be aligned; for example, an individual may support a local venue, whereas a sponsor may support a performance company that has travelled to that venue. Finally, the model generally ignores factors that may negatively influence the achievement of sponsorship objectives, such as weather, broadcast scheduling, competing messages, and the news-media.

In simplifying the sponsorship environment, sponsorship models make the achievement of consumer-based sponsorship objectives appear straightforward. The following sections provide an overview of sponsorship industry size and growth. These sections provide more evidence that the prevailing approach to sponsorship is one characterised by positive expectations.

Size of the Sponsorship Market

Corporate spending on sponsorship is significant. Meenaghan reported that worldwide sponsorship spending saw an increase from $2 billion to roughly $23 billion during the period 1984 to 1999 (Meenaghan, 2001a). More recently, Dardis reported that the global sponsorship industry was expected to see investment to the value of “…$37.7 billion US by the end of 2007” (Dardis, 2009, p37) on the basis of figures provided by the International Events Group (IEG). In 2009, influenced by global economic conditions, sponsorship spending was expected to grow more slowly than in previous years. The revised growth rate estimated by IEG was predicted to be 3.1%, taking global sponsorship spending within the region of $44 billion (IEG, 2009a).

The size of the sponsorship market, or the money available for this medium, is very large. However, as the notion of sponsorship-linked marketing suggests, the relationship of sponsorship activities to overall corporate advertising and promotional activities should also be considered.

Spending on Marketing Communications

Sponsorship spending is significant relative to spending on other forms of marketing communications. In recent decades, sponsorship spending has grown faster than spending on traditional media (Coppetti, Wentzel, Tomczak & Henkel, 2009; Cornwell, 2008; Dardis, 2009). North American data on the relative growth rates of sponsorship, sales promotion and advertising indicate that sponsorship grew an average of 9.89% yearly for the period 1997 to 2007, whereas sales promotions grew at a rate of 5%, and advertising 5.22% (Dardis, 2009).

Figure 2.2: Annual change in spending on marketing communications

40 35 % year-on-year change in 30 spending 25 20

Advertising 15

Sales 10 Promotion 5 Sponsorship

1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009(f) Year

Data sourced from: IEG Inc. (2003); IEGSR (2007, 2009)

According to IEG figures (see Appendix 2.1) sponsorship spending has a higher baseline level of annual growth compared to that of advertising and sales promotion spending. For the ten years 1999-2009 advertising has grown at 2.95% annually, sales promotion 3.22 %, and sponsorship 8.63% annually (IEG Inc., 2003; IEGSR,

2007, 2009). These growth rates are similar to those found by Dardis (2009).

The average change in spending above the baseline average is equivalent year-on- year, to 1.10% for advertising spending, 0.97% for sales promotion, and sponsorship 1.09%. These figures support the notion that sponsorship spending is growing faster than traditional media, but they also suggest that growth across the forms has an underlying stability.

Perry’s (1998) list of reasons by which research may be justified, includes a consideration of size of the marketplace. The sponsorship marketplace is, worldwide, worth a large amount of money. Further, growth has been relatively stable over time. One justification of this research project is that research on sponsorship is less prevalent, and less well developed than research on advertising, despite equivalent growth rates. Research on sponsorship that considers the economic impact of that activity, the value to sponsoring companies, or the issues that influence the achievement of consumer-based objectives is justified according to the value of this market.

Superlative Sport

Within the sponsorship industry, whatever the level of spending or overall reliability of statistics, the dominance of sport as the property receiving the mass of sponsorship dollars is clear. For the North American market, IEG reports that sport sponsorships are projected to receive 68% of sponsorship dollars in 2009 (IEG, 2009a). Other authors suggest that the level of sponsorship spending on sport is consistently higher than this, equivalent to 84% in 2007, and not likely to be challenged by other properties (Fenton, 2009). Of the other properties, arts are thought to receive 5%; causes, 9%; festivals, fairs and events, 5%; tours and attractions, 10% and associations and memberships, 3% of sponsorship spending (IEG, 2009a).

The value of the sponsorship industry and rates of growth are perhaps responsible for the ways that practitioners and academics refer to sponsorship. The statistics reveal stable growth, the dominance of sport, and not much else. Researchers, however, are

44 effusive in their praise for sponsorship.

Berkes, Nyerges and Vaczi suggest that sponsorship is “a positive medium”, and that sponsorship spending is “...continuing to grow at an extraordinary rate” (Berkes, Nyerges, & Vaczi, 2009, p36). Papadimitriou, Apostolopoulou and Dounis (2008) call the increase in sponsorship spending 'dramatic'. Alexander believes that sponsorship “offers considerable opportunity”, and “is a potentially powerful method” (Alexander, 2009, p347, 348). Coppetti et. al., argue that “...the benefits of sponsoring congruent events are manifold” (2009, p19). Finally, Cunningham, Cornwell and Coote believe that the rapid expansion of sponsorship makes “identifying any large scale or public event sans sponsorship... virtually impossible” (Cunningham, Cornwell & Coote, 2009, p65). 'Virtually impossible' is quite a claim. Each of the above testimonials encourages the development of a notion that sponsorship is an extraordinary marketing tool. The language is positively superlative, this in itself, needs review.

The use of superlative adjectives to describe sponsorship also raises the question, ‘why?’ It is not sufficient to suppose that because sports are the largest recipients of sponsorship dollars they are necessarily the most successful properties for achieving sponsorship objectives.

The dominance of sporting events as recipients of sponsorship dollars is attributed to the visibility and popularity of the property, or decision-making processes of managers (Cornwell, Roy & Steinard, 2001; Meenaghan, 1991; Sandler & Shani, 1993; Sylvestre & Moutinho, 2008). Sports are both mass market (Slater & Lloyd, 2004), appealing to many; and, objects of intense affect or identification (Cornwell & Coote, 2005; Gwinner & Swanson, 2003; Mahony, Madrigal & Howard, 2000), imbued with cultural or ritual meaning for some (Moutinho, Dionisio & Leal, 2007; Neale, Mizerski & Lee, 2008). Because of these characteristics, sports are natural targets for organisations that wish to communicate with specific groups of people, and the means to reach a mass market. These characteristics also provide a specific drive for sponsors to be aware of both audience metrics, and sponsee-based affect.

The Appeal of Sports for Sponsorship

This section answers question a for the literature review: What characteristics of sports make them desirable sponsorship properties?

Here, the superlative appreciation of sport is questioned. Using Australian examples, the goal is to discuss the ‘rewards’ of sports sponsorship, including sponsorship metrics of audience size. Sponsorship research is used to discuss the affective reach of sports. It follows that if sports properties are used to achieve sponsorship objectives, those properties must offer some competitive advantage over alternative properties. As noted in the previous section, sports are popular sponsorship properties because they provide access to mass markets (Slater & Lloyd, 2004), as well as providing access to highly targeted ‘loyal’ or ‘committed’ fans (Cornwell & Coote, 2005; Gwinner & Swanson, 2003; Mahony, Madrigal & Howard, 2000).

Audience Reach

The overwhelming preference for sports as sponsorship properties creates an expectation that all audience metrics endorse sports over other properties. Recent global statistics suggest that sports account for somewhere between 68% (IEG, June 8, 2009), and 84% (Fenton, 2009) of all sponsorship spending. Older Australian statistics indicate that sports were the recipients of roughly 60% of sponsorship funding in 1996-1997; other recipients include: education programs (8%), arts and cultural programs (6.25%), trade shows and conferences (11%), and other (14.5%), (ABS, 1999). These statistics suggest that the global preference for sports sponsorship is replicated in Australia.

Attendance provides the first series of numbers considered. Sports, in Australia, are not the recipients of the largest in-person audiences. The performing arts, instead, appear to receive the largest in-person attendance figures (see: ABS, 2010a, 2010b). In 2005-2006 performing arts in Australia achieved 11.4 million paid attendances in Australia, equivalent to 50% of the adult Australian population (ABS, 2008). In contrast, sports achieved 7.1 million attendances during the same period (ABS,

2010b). By 2009-2010, sports attendances had increased to 7.6 million attendances (ABS, 2011a). These figures show that the proportion of adult population attending sports is stable over time, 44% in 2005-2006, and 43% in 2009-2010 (ABS, 2010b; 2011a). This stability suggests that sports are a long way behind performing arts attendance figures. Attendance figures for performing arts and sports are provided in Appendix 2.2. The 11.4 million performing arts attendances in 2005-06 dwarf sports 7.1 million for the same period, and does not suggest the value of sports as sponsorship properties.

There are also important regional differences in attendance. Table 2.3 provides attendance rates for the five most popular sports in Australia, and Queensland. The ABS data of 2009-2010 show that the Australian Football League (AFL) is the most popular code in Australia by attendance, with approximately 16.2% of the population attending at least one game during the year (ABS, 2010b). The number one sport in Queensland, in contrast, is Rugby League, which sees approximately 17.2% of the Queensland population attending at least one game during the year. In Queensland, the AFL achieves an attendance rate of only 6.1% of the adult population. The research reported here aims to accommodate the local population; the regional differences highlighted justify attention to the National Rugby League (NRL) competition in research conducted in Queensland.

Table 2.3: Top 5 Popular sports in Australia & Queensland, 2009-10 Australia Queensland Number Attendance Rank Number Attendance Rank (‘000) Rate (%) (‘000) Rate (%) AFL 2831.8 16.2 1 212.7 6.1 4 Horse racing 1940.3 11.1 2 364.9 10.5 2 Rugby League 1563.8 8.9 3 598.0 17.2 1 Motor sports 1423.0 8.1 4 353.4 10.2 3 Soccer 938.8 5.4 5 144.7 4.2 5 Data source: ABS, 2010: catalogue 4174.0

Regional differences in attendance, while pertinent to this research, do not explain the dominence of sports as sponsorship recipients. Television ratings, rather than attendance figures explain sponsor interest in sports. A variety of bodies provide aggregated OZTAM data which measures free-to-air and subscription television viewing across Australian markets. Screen Australia, the Australian Federal Government funding body for Australian screen productions, provides the top-20 47 rating programs in Australia each year, from 2001 (see: screenaustralia.gov.au, n.d). Figure 2.3 provides a categorised representation of Screen Australia data on top rating programs from 2001-2009. This data establishes that from 2005, 50% of the top-rating programs in Australia each year are sports programs. Prior to 2005, sports achieved an approximate 25% of top-rating programs. Other high-ranking programs fall within the reality-television genre, including programs such as MasterChef, Big Brother, and Survivor.

Figure 2.3: Top 20 rating television programs by genre: Australia 2001-09 20

10 other comedy 5 news

0 tv drama reality tv sports

Data source: Screen Australia, nd. (www.screenaustralia.gov.au/research/statistics/wftvtopprog.asp)

Performing arts on television do not achieve high ratings. The ABS attendance data distinguishes among performances of classical music, popular music, drama, musicals and opera, dance and circuses (ABS, 2010a). The Screen Australia data indicates interest in comedy, and made-for-television drama productions, but no sign of popular or classical music, or any other category identified by the ABS. These data support the use of sponsorship or product placement in reality television programs, and sports.

A closer look at the sporting programs that attract large television audiences is also warranted (see Table 2.4). During the 2001-2009 period 28 sports programs achieved top-20 ratings. Two of these programs achieved a top-20 rating every year of that period; these are the AFL Grand Final program, and the Rugby League Final. The screening of the Melbourne Cup achieved top-20 ratings for every year except 2001.

These sports, (AFL, rugby league and horse racing) are the top three sports in Australia by attendance (see: ABS, 2010b), which suggests a relationship between attendance and television viewing for sports in Australia.

Audience viewing behaviour and sport format are also important for television audiences. Television viewing behaviour can be seen in the spillover of ratings for the AFL Grand Final and Australian Open Tennis. The networks hosting these programs create ‘event days’ with pre-match, post-match and trophy presentation programs attached to competition programs. These additional offerings have achieved top-20 ratings for a variety of reasons, including turning the television on early, leaving it on post-match, or genuine interest in commentary and trophy ceremonies.

The importance of sport format is evidenced by the arrival of cricket in the top-20 ratings with the 20/20 (short) format. Traditional cricket formats do not feature in television ratings despite cricket being ranked sixth among sports by attendance in Australia (ABS, 2010b). Nationalism, as a relation of state-based rivalry also influences ratings for sports programs. High-rating 20/20 cricket matches, and World Cup soccer matches are those that include Australia as a competitor. A nationalism- effect may also extend to sports events hosted by Australia; the Melbourne Commonwealth Games produced more high-rating programs than the Manchester Commonwealth Games in 2002.

Many of the explanations for high ratings of sports programs reflect the underlying popularity of the sport. Some suggest clever business strategy and marketing. Others hint at the confluence of marketing and affect. Overall, television ratings provide a greater justification for sponsor interest in sports, relative to attendance figures. The role of affect, however, remains important.

Table 2.4: Sports programs ranked by audience size (000’s), 2001-09 2001 2002 2003 2004 2005 2006 2007 2008* 2009 Rank 000’s Rank 000’s Rank 000’s Rank 000’s Rank 000’s Rank 000’s Rank 000’s Rank 000’s Rank 000’s AFL Grand Final- Pre-Match 4 2593 13 2271 13 1913 AFL Grand Final 3 2604 3 2626 4 2966 3 2796 2 3386 2 3145 1 2563 1 2491 3 2878 AFL Grand Final- Post-Match 3 2981 9 2501 4 2387 11 2009 7 2448 AFL Grand Final- Presentation 2 2468 Australian Open Tennis– Pre-Match 7 2486 Australian Open Tennis– Men’s Final 1 4043 4 2748 2 2442 3 2447 10 2246 Australian Open Tennis– Presentation 5 2344 5 2207 Australian Open Tennis– Other matches 11 2297 9 2316 Wimbledon 1 3036 20/20 Cricket 19 2179 6 2306 9 2077 15 2123 20/20 Cricket 17 2039 One-day Cricket 19 2082 16 1864 One-day Cricket 19 2036 18 1846 Cricket World Cup 5 2465 Melbourne Cup 4 2503 10 2244 5 2471 6 2506 12 2272 8 2191 4 2272 4 2673 Rugby League- State of Origin- Game 1 11 2020 7 2092 8 2322 Rugby League- State of Origin- Game 2 12 1961 8 2084 13 2134 Rugby League- State of Origin- Game 3 15 1890 6 2145 18 1907 Rugby League- Final 17 2097 10 2177 8 2352 13 2107 5 2563 7 2553 3 2422 10 2051 6 2528 World Cup - Soccer 2 2702 8 2484 11 2297 World Cup - Soccer 11 2221 16 2199 Melbourne- Commonwealth Games 1 3561 Melbourne- Commonwealth Games 5 2736 Melbourne- Commonwealth Games 15 2229 Manchester- Commonwealth Games 11 2156 Athens - Olympics 6 2304 Rugby World Cup 1 4016 World Championship Swimming 5 2507 Data source: Screen Australia, nd. (www.screenaustralia.gov.au/research/statistics/wftvtopprog.asp) *Data for 2008 is the Screen Australia data excluding Beijing Olympics programs which would otherwise indicate that sports achieved all top-20 ratings programs in 2008.

The Many Affects

This section refers back, briefly to attendance figures for both performing arts and sports, to provide an illustration of affect in attendance. Following that brief foray, the theories that explain sponsor interest in customer affect, (loyalty, commitment, and identification) are reviewed.

Attendance figures from the ABS (ABS, 2010a, 2010b) show the mass appeal of sports and performing arts, as well as the dedication of more involved event consumers. This argument is best represented visually (see Figures 2.4, and 2.5). The data sets are represented using different frequency scales because aggregation of the performing arts attendance data obscures the high frequency bump, and the sports data is available with only three categories. For performing arts and sports, large proportions of the Australia population attend an event once a year. Both also see frequency of attendance drops in the 3 to 5 attendances range; and a resurgence in the ‘6 or more’ attendances category.

Figure 2.4: Sports attendance by frequency of attendance, 2009-10 80

60 AFL 50 Horses

40 Motor Sports Rugby League 30 Rugby Union 20 Soccer 10

0 1 or 2 3 to 5 6 or more

Data Source: ABS, 2010b

Figure 2.5: Performing arts attendance by frequency of attendance, 2005-06 70

50 classical popular 40 theatre 30 dance

20 musical & opera other 10

0 1 2 3 4 5 6 or more

Data Source: ABS, 2010a

This data most probably reflects the impact of ‘season’ tickets, which encourage behavioural loyalty. It also suggests the weakness of ‘novelty’ productions for commitment; as the categories of sports and performing arts that achieve the highest rates of ‘once a year’ attendance, also appear the least likely to experience the ‘6 or more’ resurgence. The events that experience the ‘6 or more’ bump also appear to have flatter attendance curves (see: AFL, Soccer, Rugby League, and Popular Music), which argues more robust demand overall.

The data also provides that the proportions of sports and performing arts attendees who might be called ‘behaviourally loyal’ are very different. For the performing arts, the ‘6 or more’ attendances group represents 8.2% of attendances on average. There are large differences within the performing arts sector; only 3.8% of performing arts audiences patronise the ‘musical & opera’ category ‘6 or more’ times each year, whereas, this number is 13.7% for ‘popular music’ and 12.1% for ‘classical music’ audiences. There are also notable differences among sports in their ability to attract ‘6 or more’ attendance audiences. On average, 22.2% of sports attendees attend ‘6 or more’ times each year. Horse racing has the lowest rate of high frequency attendees among the sports, equivalent to 7.9% of horse racing attendees. Soccer achieves the highest rate of high frequency attendees, equivalent to 32.7% of attendees, followed by AFL (31.2%), and Rugby League (27.2%). This data provides a strong incentive to focus on sports for potential sponsors aiming to capitalise upon property-based

52 affect rather than novelty or occasional attendance.

A variety of theories are used to explain multiple event attendances. Among these are loyalty, involvement, affect, commitment, and identification; several are discussed here. An important assumption that drives sponsorship is the potential that consumer- event, or property-loyalty, creates the opportunity for affect transfer from the event or property, to the sponsor or sponsor’s brand (Burnett, Menon & Smart, 1993; Mahony, Madrigal & Howard, 2000; Neale & Funk, 2006; Quester & Farrelly, 1998; Shannon, 1999). Sponsorship research on this subject provides mixed support for the idea of loyalty transfer (see: Dionisio, Leal & Moutinho, 2008; Moutinho, Dionisio & Leal, 2007). These results will be discussed, following a brief overview of the terms and theories.

Loyalty is most often conceived with reference to the work of Jacoby and Kyner (1973) or Jacoby and Chestnut (1978) (see: Bristow & Sebastian, 2003; Funk, 2002; Mahony, Madrigal & Howard, 2000; Pritchard, Havitz & Howard, 1999; Neale & Funk, 2006; Theysohn, Hinz, Nosworthy & Kirchner, 2009). Defined: …brand loyalty is (1) the biased (i.e., non-random), (2) behavioural response (i.e., purchase), (3) expressed over time, (4) by some decision-making unit, (5) with respect to one or more alternative brands out of a set of such brands, and (6) is a function of psychological (decision-making, evaluative) processes. Jacoby & Kyner, 1973, p2 Jacoby and Kyner (1973) argue that all behaviour is not equal, and that true loyalty should involve not only repeated behaviour, but also the evaluative (or attitudinal) reasons underlying purchase. These principal conditions determine that loyalty is most often explained as having two components, which are commitment (or attitudinal loyalty) and behavioural consistency (Bristow & Sebastian, 2003; Funk, 2002; Mahony & Madrigal, 1999; Mahony, Madrigal & Howard, 2000).

Commitment, which features in definition of loyalty as a form of psychological attachment (Mahony, Madrigal & Howard, 2000), also features in social identity conceptualisations of identification as a form of affective attachment (Ellemers, Kortekaas & Ouwerkerk, 1999). Opinion is divided, however, as to whether

53 commitment is a component of identification or a consequence of it (see: Ashforth & Mael, 1989; Ellemers, Kortekaas & Ouwerkerk, 1999).

According to Ashforth and Mael (1989) commitment and loyalty are both consequences of identification. However, they also provide that their objection to the inclusion of commitment in a notion of identification is based on the observation that “Commitment scales consistently feature generalised usage of the terms goals and values…” (Ashforth & Mael, 1989, p23). This feature provides a direct challenge to the four principles of identification developed by Ashforth and Mael (1989). The third of these principles is used to argue: … acceptance of the category as a definition of self does not necessarily mean acceptance of those values and attitudes. Ashforth & Mael, 1989, p22 Thus, identification is a personal-perceptual construct, one that is robust to group failures, and one in which the identified individual retains a level of independence in their values and attitudes. That identification is robust to successes and failures should suggest a commitment of some endurance.

Ellemers, Kortekass and Ouwerkerk (1999) use the definition of social identity provided by Tajfel (1978) which says it is: …that part of an individual’s self-concept which derives from his knowledge of his membership of a social group (or groups) together with the value and emotional significance attached to that membership. Tajfel, 1978, p63 Many social identity theorists support the tripartite definition of identification which recognises cognitive, affective, and evaluative components (see: Dimmock, Grove & Eklund, 2005; Ellemers, Kortekaas & Ouwerkerk, 1999; Heere & James, 2007; Jackson, 2002). Ellemers, Kortekaas and Ouwerkerk (1999), maintain that the affective (emotional significance) component of identification is a form of commitment.

Reviewing both loyalty and identification reveals that each of these constructs contains a substantial affective element. The term ‘commitment’ occupies a middle- ground in loyalty and identification theories which provides that theorists refer to it

54 as a form of psychological attachment (Ellemers, Kortekaas & Ouwerkerk, 1999; Mahony, Madrigal & Howard, 2000). Loyalty also considers the evidence of affect, in repeated or consistent behaviours (Jacoby & Kyner, 1973). This behavioural element is excluded from many conceptualisations of identification (Ashforth & Mael, 1989; Ellemers, Kortekaas & Ouwerkerk, 1999). These theories provide different means to assess whether event-based affect transfers to sponsor’s brands. Measures of loyalty should consider behaviours, whereas measures of identification should be restricted to cognitive, evaluative and affective approaches.

Sponsorship research provides mixed support for the idea that property-loyalty can transfer to the sponsor or sponsor’s brand. Several authors suggest the possibility of affect transfer in uncertain terms. Burnett, Menon and Smart, write that “Ideally, this allegiance to the team is transferred to the products advertised during sports events” (1993, p22). Quester and Farrelly propose “It also seems reasonable to assume that involvement in, and loyalty to, an event can transfer into brand loyalty toward a sponsor’s products…” (1998, p544). More convincingly, Mahony, Madrigal and Howard conclude their research; “The stronger the attitude typically the greater the likelihood of congruent behaviour” (2000, p22). None of these authors specifically tested the proposition that event/property-loyalty is transferable, nor have other authors who expressed similar sentiments (see: Donahay & Rosenberger, 2007; Hunt, Bristol & Bashaw, 1999; Shannon, 1999). The reasons for this are probably related to loyalty theory and conceptual redundancy.

In the first instance, the characteristics of loyalty and the nature of sponsorship present difficulties for research design. In order to study whether loyalty transfers, it must first be established that event/property loyalty exists (in terms of, at least, both repeated behaviour and affect), and then evaluate the same toward the sponsor or sponsor’s brand, and any contribution of property loyalty to the sponsor’s brand. A confound, of pre-existing experience with the sponsor would be difficult to control, as would experience with the sponsor occurring after a sponsorship stimulus (but before a loyalty measure). That loyalty assumes longevity is not problematic in relation to the property, but is problematic for measures of sponsor-based loyalty as naturalistic research would be confounded, and experimental longitudinal research, difficult to implement sufficient to establish ‘loyalty’.

The possibility of loyalty transfer, however, cannot be dismissed. It is approached quantitatively by Sirgy, Lee, Johar and Tidwell (2008), and in qualitative research Moutinho, Dionisio and Leal (2007) and Dionisio, Leal and Moutinho (2008). In research conducted in Spain on soccer fans, Dionisio, Leal and Moutinho (2008) found that loyalty transfer was unlikely because fans regarded sponsors as an out- group to soccer-supporting groups. Specifically, they note: …resistance against sponsor brands in both the ‘organised supporting group’ and the ‘devoted fans’ group because they do not consider these companies as having a real emotional link with the club. Dionisio, Leal & Moutinho, 2008, p33 Other research conducted by the same authors, however, produces a different conclusion.

Among surfers, and surfing-fans (non-surfers) in Portugal, reactions differed towards sponsors (Moutinho, Dionisio & Leal, 2007). Surfers demonstrated a marked ability to “distinguish between ‘real’ and loosely connected brands” (Moutinho, Dionisio & Leal, 2007, p682). In consequence, surfers were more likely to prefer the sponsor’s brand and to regard sponsors as “legitimate supporters” (Moutinho, Dionisio & Leal, 2007, p683). In contrast, fans of surfing did not distinguish between sponsor’s brands and (especially) clothing brands that were not necessarily suited to the purpose of surfing (Moutinho, Dionisio & Leal, 2007). These projects suggest that group relations are nuanced; sponsor loyalty may differ across sporting contexts and according to group membership. The contrast in attitudes of ‘participants’ and ‘fans’ is an especially useful one, as it identifies the out-group rejection that research on consumer perceptions of altruism expects to exist (see: Dardis, 2009; Rifon, Choi, Trimble & Li, 2004).

The quantitative research conducted by Sirgy et al (2008) studies self-congruity, as it contributes to brand loyalty. The authors refer broadly to the social identity theory of identification and question whether; …customers of a particular product who can identify with the people attending the sponsorship event are likely to feel more loyal to the brand, especially when they are emotionally involved with the sponsorship event

and aware that the firm is sponsoring the event. Sirgy et al, 2008, p1092 The results of the research provide that self-congruity with the sponsored event, in three out of five samples, (and in pooled data from all samples), contributes to loyalty towards the sponsor’s brand (Sirgy et al, 2008). The authors also find that event involvement and awareness of the event sponsorship contribute to the self- congruity effect on brand loyalty, but only for subjects categorised as highly involved or highly aware (Sirgy et al, 2008). This research provides support for the influence of self-congruity on loyalty towards the sponsor’s brand; future research on this subject, however, would benefit from a more tightly controlled method.

The affective appeal of sports and the means to study those (i.e., fan research, loyalty, commitment, identification, and involvement) strongly suggest the potential of sports for sponsorship. The affective appeal to audiences of other potential sponsorship properties, no doubt exists equally strongly. However, as a commercial sponsorship venture, sports offer stronger evidence, in consistent high-frequency attendance rates and broadcast message reach, for investors concerned with the achievement of objectives. Audience metrics, combined with affective involvement make sports a more valuable opportunity than other sponsorship properties.

Sponsorship Objectives

This section answers question b for the literature review: What objectives drive sponsorship?

Definitions of sponsorship emphasise the importance of the objectives driving sponsorship contracts (Javalgi et al, 1994; Meenaghan, 1983; Pope, 1998; Walliser, 2003). Sponsorship contracts may be driven by one, or a combination of objectives. Occasionally contracts appear not to be linked with objectives, as in cases where decision-making is driven by the “chairman’s choice” (Pearsall, 2010; Sylvestre & Moutinho, 2008), although there is evidence this approach is becoming less popular. Industry literature, particularly, indicates the increasing importance of quantitatively

57 measurable outcomes of sponsorships (Brewer & Pedersen, 2010; Cameron, 2009; Maestas, 2009; Priest, 2010). All objectives say something about the goals of sponsorship. Sponsorship is conceived as a communications device, as a means to influence consumer and stakeholder attitudes, and as a device to influence sales.

Pope and Voges have summarised key types of sponsorship objectives: … (a) media objectives (e.g., achieving cost effectiveness or reaching target markets), (b) marketing objectives (such as brand promotion, generating purchase intention, or generating sales increase), and (c) broader corporate objectives (which are principally image based). Pope & Voges, 2000, p96 The summary provided by Pope and Voges is explains a corporate perspective. Objectives may be identified according to the function or unit responsible for task implementation or measuring achievement.

Another approach to sponsorship objectives is to consider the consumer movement towards purchase (Lavidge & Steiner, 1961), also called the hierarchy of effects (Palda, 1966). The hierarchy of effects model is reviewed prior to specific discussion of sponsorship objectives. Table 2.5 provides an overview of consumer behavioural dimensions, the hierarchy of effects and specific sponsorship objectives. This table also notes the activities undertaken by sponsors to achieve objectives, and some of the information processing theories that are used to link objectives, activities and consumer behavioural outcomes.

Information in the table is drawn from a number of sources. Lavidge and Steiner (1961) provide the early model of consumer behavioural dimensions and consumer movement towards purchase. Palda (1966) renamed the movement the ‘hierarchy of effects’. Smith and Swinyard (1982) provide critical analysis of the model and introduce the possibility of alternative paths to purchase. These authors also provide a link to Fishbein and Ajzen’s (1975) expectancy-value model, (the earliest of the Attitude toward the Object formulations), which stresses consumer cognitive and evaluative processes contributing to purchase (see: Smith & swinyard, 1982). Where earlier theory development focused almost exclusively on advertising, Cornwell, Weeks and Roy (2005) provide an overview of objectives and consumer outcomes in

58 the sponsorship context. Their contribution to Table 2.5 is in the matching of objectives with consumer information processing theories. Many other authors have contributed to the literature on objectives and consumer responses, the authors noted in Table 2.5 are identified because they represent an advance from, or the most informative summary of what was previously understood.

The Hierarchy of Effects

The term ‘hierarchy of effects’ was coined by Palda (1966) to describe the progress of the consumer through steps of awareness, knowledge, liking, preference, conviction and purchase outlined by Lavidge and Steiner in 1961. For Lavidge and Steiner, the purpose was to explain advertising effects; “…if something is to happen in the long run, something must be happening in the short run, something that will ultimately lead to eventual sales results” (1961, p59). The ‘movement toward purchase’ by Lavidge and Steiner (1961) was presented as process identifying how the consumer process might culminate in the purchase of a product following advertising exposure.

When reviewing the ‘movement toward purchase’, Palda (1966) acknowledged that the idea was not new. The ‘hierarchy of effects’ can be seen in several models that explain purchase as resulting from awareness of advertising; including the AIDA (Attention-Interest-Desire-Action) model and similar processes (see: Kohlman, 1960; Webster, 1968). What Lavidge and Steiner had done that was novel was to explicitly and concisely link their process with the ‘three functions of advertising (Lavidge & Steiner, 1961, p60), or ‘related behavioural dimensions’ (Lavidge & Steiner, 1961, p61) of cognition, affect and conation (Smith & Swinyard, 1982).

The cognition-affect-conation approach is, itself, not new. It can be traced to18th century philosophers who “…divided psychology’s subject matter into three distinct faculties: cognition, affect, and conation” (Forgas, 2008, p94) which were later researched as distinct, rather than interdependent functions (Forgas, 2008). The relationship among cognition, affect and conation is preserved in the marketing literature (Feldman & Lynch, 1988) as belief-affect-behaviour (see: Fishbein &

Ajzen, 1975; Smith & Swinyard, 1982). It explains that knowledge of an object will determine affective response to it, and subsequent actions taken (i.e., approach, avoidance, etc.). The sequencing of these functions is not without controversy.

The progress of a consumer from awareness to purchase, or from cognition to behaviour should not be regarded as inevitable (see: Hunt, 1983; Palda, 1966; Smith & Swinyard, 1982). Palda introduced uncertainty about the hierarchy in 1966: …while “significantly” large numbers of respondents moved “up the hierarchical ladder” of awareness, liking of the advertising and of attitudes, performance on the last “rung” is difficult to assess. Add to this the uncertainty of the causal direction … and we end with an unsatisfactory feeling. Palda, 1966, p23

In subsequent years, the hierarchy of effects and the cognition-affect-conation process achieved wide theoretical support and some empirical support (see: Assael & Day, 1968; Hunt, 1983; O’Brien, 1971; Smith & Swinyard, 1982). However, there were also a “multitude of studies reporting low correlations between measures of attitudes and measures of behaviour” (Smith & Swinyard, 1982, p82). The cases, where consumers did not progress through cognition-affect-conation prompted the development of low involvement models of consumer behaviour.

The low involvement models of consumer behaviour suggest that consumers follow a cognition-conation-affect process in some purchase situations (Hunt, 1983; Smith & Swinyard, 1982). This process explains ‘trivial’ purchases in which affect develops after use (Hunt, 1983). That the low involvement process appears more common than the learning hierarchy relates to the quality of cognition, according to Smith and Swinyard (1982). Experience is more likely to provide higher order beliefs, which in turn leads to higher order affect, and more likely commitment to purchase (Smith & Swinyard, 1982).

The successes of advertising are achieved through the creation of awareness and development of product beliefs. Lower order beliefs (Smith & Swinyard, 1982) provide impetus for product trial and the development of higher order beliefs. The

Table 2.5: The Hierarchy of Effects and Sponsorship Objectives Related behavioural Movement towards Objectives Activities undertaken by the Information processing dimensions* purchase* Sponsor: assumptions & theory (Hierarchy of Effects)^

Purchase Sales CONATIVE Share value -the realm of motives ROI Conviction

Preference AFFECTIVE Attitudes & intentions Linking the sponsor & sponsee Associative networks -the realm of emotions Image development Experiences & Engagement Congruence Liking Image transfer Balance theory Classical conditioning

Knowledge Awareness Media exposure Mere exposure COGNITIVE Recall Target market reach -the realm of thoughts Recognition Awareness

Adapted from: Cornwell, Weeks & Roy, 2005; Lavidge & Steiner, 1961*; Palda, 1966; Smith & Swinyard, 1982. * ‘Related behavioural dimensions’ and the ‘Movement towards purchase’ are drawn directly from Lavidge & Steiner, 1961. ^ ‘Hierarchy of Effects’ as a term to describe the movements toward purchase was first used by Palda, 1966.

61 importance of the hierarchy of effects and cognition-affect-conation models is in their organisation of consumer-marketing processes. The weaknesses in these models have implications for the development of sponsorship objectives.

Sport Sponsorship: Objectives & Research outcomes

The hierarchy of effects provides the structure for the following review of sponsorship objectives. Three groups of objectives are reviewed. These groups of objectives will be referred to as ‘communications objectives’, ‘consumer-based objectives’, and ‘behavioural objectives’. The groups are distinguished by the point at which achievement of the objective is most commonly measured; communications objectives are commonly measured according to media exposure, consumers provide data on affect, evaluations, and image perception, and behaviours are measured at the point of purchase, or following some quantitative change in share value.

Communications objectives

Target market reach, consumer awareness, recall and recognition are among the most commonly stated sponsorship objectives (see: Alexander, 2009; Calderon-Martinez et al, 2005; Cameron, 2009; Cornwell, 2008; Felten, 2009; Fortunato, 2009; Grohs, Wagner & Vsetecka, 2004; Gwinner & Bennett, 2008; Maestas, 2009; McDonald, 1991; Speed & Thompson, 2000; Vale, Serra, Vale & Vieira, 2009). These objectives rely on the effective distribution of the marketing or sponsorship message, and consumer exposure to the same. An expectation underlying these objectives is that exposure to messages will promote subsequent consumer-based effects.

Where consumer effects (awareness, recall, recognition) are not measured to support claims of sponsorship effectiveness; measures of media exposure are still used as proxy of sponsorship achievement. Common measures are counts of logo exposures in television coverage, duration of air-time, and event audience numbers (see: Cameron, 2009; Calderon-Martinez et al, 2005; Felton, 2009; Harvey, 2001; Maestas, 2009). Each of these measures provides sponsors with an indication of how

62 much media exposure their brand has received. The value of exposure is calculated as: …the amount of advertising exposure (based on current advertising rates) received from a sponsorship via various media outlets … If the total value of the exposure is greater than the sponsorship investment, then the sponsorship effectively yields a discount on the effective media buy. Maestas, 2009, p99 This calculation yields only a measure of accounting success, there is no explicit consideration of audience attention.

Each of the communications objectives sought through the use of media exposure is essentially mute. Logo exposure in naturalistic environments does not provide a guaranteed link from attention to broadcast to sales behaviour. Furthermore, while discount media purchasing is sensible, it is also an activity that does not link to consumer attitudes or consumer behaviour (McDonald, 1991). These measures are also criticised as being ‘post hoc’ (Calderon-Martinez, 2005; Miyazaki & Morgan, 2001; Speed & Thompson, 2000) in that their effectiveness cannot be readily predicted. This renders them less reliable in terms of pre-planning or objectives- setting and quiet on the subject of purchase intentions or sales.

Consumer-based objectives

For the purposes of this review, consumer attitudes, intentions, image development, and image transfer objectives are grouped to form the category of consumer-based objectives. There are three reasons for this grouping: 1) all of the constructs in this group are measures of self-report, 2) where media exposure objectives assume limited or automated processing of the stimulus; attitude measures, image assessment and image matching measures all require greater levels of information processing, and 3) although these objectives require more consideration on the part of the consumer, they are not measures of behaviour, as are sales or share value.

Purchase intention is commonly regarded as a behavioural or conative objective (see: Breckler, 1984; Lavidge & Steiner, 1961; Cornwell, Weeks & Roy, 2005) because it

63 is regarded as a measure of consumer commitment to act. It exists as a useful proxy for actual behaviour in survey and experimental research settings. In this review, purchase intention is regarded as an attitudinal objective. It is included in this category because it is an “abstract” measure of behaviour (Pope, Voges & Brown, 2009, p6); and is, overall, “troublesome” with regard to predictive ability (Smith, Graetz & Westerbeek, 2008, p388). Research by Jamieson and Bass (1989) on the accuracy of purchase intention measures found that “…accurate predictions of purchase probabilities vary considerably across weighting schemes and products” (Jamieson & Bass, 1989, p344). These judgments and findings suggest that although purchase intention measures “…possess predictive usefulness” (Brown, Pope & Voges, 2003, p1669), they remain proto-behavioural, or attitudinal.

Attitudes and Purchase Intentions

The development of consumer attitudes are regarded as an important sponsorship objective by many (see: Cameron, 2009; Irwin, Lachowetz, Cornwell & Clark, 2003; Kinney & McDaniel, 1996; Lee, Sandler & Shani, 1997; Speed & Thompson, 2000; Walliser, 2003). The development of consumer intention to purchase the sponsored product is similarly popular as a sponsorship objective (see: Dardis, 2009; Dees, Bennett & Villegas, 2008; Kinney & McDaniel, 1996; Madrigal, 2000; Pope & Voges, 2000). These objectives have attracted a significant amount of attention directed towards explaining their effects and correlates, and also underlying psychological processes.

Sponsorship, and marketing, researchers, expectation that exposure to sponsorship stimuli, and attitudes, predict purchase intention (Harvey, 2001; Walliser, 2003). These expectations are the product earlier assumptions, including that cognition, affect, and conation are thought to correlate (Breckler, 1984; Tesser & Shaffer, 1990). The reasons for this include that stimuli prompting attitudinal responses usually activate all attitude components simultaneously, (Eagly & Chaiken, 2007). Another explanation is that individuals may exert effort to maintain a level of cognitive consistency in their evaluations (Folkes, 1988; Harris, Todorov & Fiske, 2005; Heider, 1958; Kelley, 1973; Tesser & Shaffer, 1990).

There is a substantial body of research that supports the influence of sponsorship on purchase intention (see: Dardis, 2009; Gwinner & Bennett, 2008; Lee & Cho, 2009; McDaniel, 1999; Pope & Voges, 2000). As a dependent condition, purchase intention is explained by a variety of antecedent attitudes or sponsorship functions, some of these are noted in the following review of attitude research.

Measures of attitudes, themselves, take a variety of forms. Attitudes towards sponsorship (as an activity) are studied; most commonly when researchers believe the consumer’s beliefs about the altruistic (as opposed to commercial) motivations of sponsors will influence other attitudes, purchase intention or future behaviour (see: d’Astous & Bitz, 1995; Dardis, 2009; Dean, 2002; Rifon et al, 2004). Attitudes towards sponsorship are also measured in research that argues they contribute to consumer perceived corporate image (Calderon-Martinez, Mas-Ruis & Nicolau- Gonzalbez, 2005; Stipp & Schiavone, 1996).

Perhaps the most common attitude measure used in sponsorship research is ‘attitude toward the sponsors’ brand’. The premise underlying its’ use is that favourable attitudes toward the sponsee will influence attitudes toward the sponsors’ brand (see: DeGaris, West & Dodds, 2009; Farrelly, Quester & Burton, 2006; Javalgi, Traylor, Gross & Lampman, 1994; Kelly & Whiteman, 2010; Lee, Sandler & Shani, 1997). By this reasoning, attitude toward the brand is one of the penultimate objectives.

Empirical results relating to this expectation are substantial. Coppetti, Wentzel, Tomczak and Henkel (2009) find that in experimental settings both articulation of the event sponsorship, and consumer participation in sponsorship experiences facilitate development of attitude towards the sponsors brand, and image transfer. Research conducted in a naturalistic setting) found that articulation of the sponsorship reproduced improved attitudes toward the brand and image transfer, although the successful application of participation in sponsored experiences was not replicated. Ferreira, Hall and Bennett (2008), in their study of self-reported Mountain Dew soft drink consumption at the Dew Action Sports Tour, found that sponsorship explains self-reported brand use. The authors also, however, found regional and demographic differences. These research projects introduce the importance of

65 experiential characteristics as they might influence achievement of sponsorship objectives.

Dardis (2009), testing a mere exposure hypothesis found that when individuals are provided with an incongruent sponsor/sponsee stimulus but multiple sponsorship messages (repetition), perceptions of congruence, sponsor credibility, community relations and purchase intention increase. Attitudes towards the sponsor’s brand are also positively influenced (Dardis, 2009). Gwinner and Bennett (2008), however, found that more closely aligned sponsors and sponsees (congruent sponsorships) produced positive attitudes towards the sponsors’ brand, which in turn predicts purchase intention. Lee and Cho (2009) have also studied congruent sponsorships and attitude toward the sponsors’ brand. They find that attitude toward the brand is associated with congruence, prior brand experience and purchase intention (Lee & Cho, 2009). These results indicate the potential for successful event sponsorships given congruent or incongruent sponsor/sponsee pairings, and the importance of additional strategic consideration when incongruent sponsorships are used.

Attitude and purchase intention research illustrates the complex interrelationships among sponsorship objectives. When consumer attitudes toward one object (the sponsee) are expected to govern evaluations of a second object (the sponsor or sponsor’s brand), the role of sponsorship in linking the objects is the facilitation of an ‘affect’ (McDaniel, 1999) or ‘image’ transfer (Gwinner, 1997). Further assumptions, including that of increased sponsorship success when the congruence of sponsor/sponsee pairings is considered, are also introduced. Although overlapping objectives are problematic in any situation, theoretical discussion is further complicated in the case of image objectives when ‘image’ is used both in reference to ‘corporate image’ (see: Javalgi et al, 1994; Nguyen & LeBlanc, 2001) and also to a form of brand image more closely related to ‘personality’ (see: Gwinner, 1997; Gwinner & Eaton, 1999; McDonald, 1991; Meenaghan, 2001).

Image: Brand Image and Corporate Image

Sponsorship research that uses the term ‘corporate image’ refers to images and brand associations held by consumers about the sponsoring organisation (see: Bennett,

1999; d’Astous & Bitz, 1995; Javalgi et al, 1994; Pope & Voges, 2000; Pope, Voges & Brown, 2004). Measures of corporate image are concerned with consumer perceptions of corporate engagement in community, treatment of employees, product/service quality and organisational policies or decision-making (see: Dowling, 1986; Javalgi et al, 1994; Brown & Dacin, 1997). Dowling (1986), in particular, developed a list of corporate image dimensions, including such items as: competent management, equal opportunity employee, quality products, sound financial condition, socially responsible, reliable, modern, technological leadership, sound financial investment, protects jobs of local workers, develops many new products, spends money on R&D…, cares about the local community, helps charities, and makes products that conserve energy Dowling, 1986, p112 This list has much in common with the six-item corporate image scale later developed by Javalgi et al (1994) following in-depth interviews with company managers. The Javalgi et al (1994) scale refers to product/ service quality, management, financial goals, community involvement, customer needs, and employee satisfaction.

Sponsorship research uses corporate image as a dependent variable influenced by consumer attitudes or involvement with the sponsee (see: Bennett, 1999; Brown & Dacin, 1997; Cornwell, Roy & Steinard, 2001; Javalgi et al, 1994; Pope & Voges, 2000; Pope, Voges & Brown, 2004). Pope and Voges (2000), among other results, found that when individuals believe an organisation is an event sponsor, purchase intention is positively impacted; they also found that corporate image, as a covariate, contributes to purchase intention, sponsorship awareness and brand name awareness. D’Astous and Bitz (1995) found that philanthropic sponsorships have a greater positive impact on corporate image than do commercial sponsorships; and, that consumer interest in an event also improves corporate image measures for sponsors.

Research conducted by Javalgi et al (1994) found that although awareness (aided and unaided) of sponsorship does not enhance perception of all dimensions of corporate image, when individuals are specifically made aware of the sponsorship effort, public image of the organisation improves. Javalgi et al (1994) also raise the possibility that corporate image may be negatively influenced, or not improved, by sponsorship

67 activities as sponsorship participation will increase the likelihood of publicity for the firm (positive or negative), and that sponsorship, on its own cannot “reverse prior negative perceptions” (Javalgi et al, 1994, p57).

The development of brand image as a sponsorship objective is used very differently to corporate image, despite initial similarities in definitions. According to Gwinner (1997; Gwinner & Eaton, 1999), citing Keller, brand image refers to “perceptions about a brand as reflected by the brand associations held in memory” (1993, p3). The common elements of corporate image and brand image definitions are the ideas of meaning and associations’ consumers hold for the object evaluated. Where the concepts diverge are in areas of more detailed conceptualisation and operationalization.

When outlining his conceptualisation of brand image Gwinner (1997) refers to McCracken’s (1989) model of the endorsement process. That model suggests that the celebrity endorser is imbued with cultural meanings acquired through their various roles, exposures, and associations (McCracken, 1989). Those meanings are transferred to products through the pairing of endorser and product. For example, McCracken claims James Garner is a successful endorser of Mazda because “he represents a bundle of meanings about maturity, American-ness, confidence, masculinity, intelligence, and good humour” (1989, p312). This example illustrates that the form of associations in brand image measures differs significantly from those used in the corporate image measures. Where corporate image deals with perceptions of product quality and managerial competence; brand image uses qualitative, personality-based characteristics to explain consumer associations.

Gwinner and Eaton (1999) went on to develop the process by which brand image transfer could be measured in the sponsorship environment. This process uses Aaker’s (1997) brand personality scale, and pre- and post-testing procedures to establish existing brand meaning, sponsee (event) meaning, and post-sponsorship changes in consumer perceived images. The brand image transfer process has been used by a number of sponsorship researchers (see: Cliffe & Motion, 2005; Coppetti, Wentzel, Tomczak & Henkel, 2009; Gwinner & Bennett, 2008; Gwinner & Eaton, 1999; McDaniel, 1999; Rifon, et al, 2004). Empirical research argues that transfer of

68 image from the sponsored event to the sponsor’s brand is supported (Gwinner & Bennett, 2008; Gwinner & Eaton, 1999; Rifon et al, 2004). The natural extension to this research is investigation of differentiated sponsorship success attributable to congruent sponsor-sponsee image partnerships.

The notion of ‘fit’ or congruence among the image of the sponsor or sponsor’s brand, and the sponsee (usually the sponsored event) suggests that greater success is likely to occur where there is greater perceived similarity among the sponsor and sponsee (Coppetti et al, 2009; Gwinner & Eaton, 1999; McDaniel, 1999; Rifon et al, 2004; Rodgers, 2003). McDaniel (1999) found only partial support for his hypotheses, which suggested greater purchase intent develops based on congruent or “match-up” manipulations of media vehicle, event and sponsor. Whereas Gwinner and Eaton (1999) found that image transfer from event to sponsor was stronger where functional, and image similarities existed among sponsor and sponsee. Rifon et al (2004) found that sponsor/sponsee congruence contributed positively to perceptions of sponsor altruism, sponsor credibility, and positive attitudes. Coppetti et al (2009) found that congruent sponsorships led to more positive evaluations of the sponsorship, improved attitudes toward the brand, and positive image transfer.

Behavioural objectives

Behavioural objectives, in this review, include such measures as changes in share value, return on investment (ROI), or sales (see: Cornwell, Pruitt & Van Ness, 2001; Hutchinson & Bouchet, 2010; Maestas, 2009). Each of these measures argues the importance of quantifiable measures of change that can be attributed to sponsorship activity.

Sales, as an outcome of sponsorship have not been reported in any sponsorship study; nor are sales responses likely to be reported, due to the “highly proprietary nature of individual product sales data” (Cornwell, Pruitt & Van Ness, 2001, p17). Research into ROI is similarly hampered by the need for accurate data, the lack of baseline measures and difficulty in identifying the myriad contributions to company profit (Maestas, 2009). Perhaps because of confidentiality problems, academic

69 research on ROI tends to take the form of qualitative interviews in which marketing executives are asked, generally, about sponsorship success (see: Hutchinson & Bouchet, 2010).

Changes in share value, however, have been studied by a number of sponsorship researchers who use the event study methodology to study abnormal stock market returns (see: Calderon-Martinez, Mas-Ruiz & Nicolau-Gonzalbez, 2005; Clark, Cornwell & Pruitt, 2002; Cornwell, Pruitt & Van Ness, 2001; Farrell & Frame, 1997; Johnston, 2010; Miyazaki & Morgan, 2001; Spais & Filis, 2008). This form of research aims to identify specific market events that cause ‘shocks’ or unexpected positive or negative share price movements relative to normal, or expected returns in average trading (see: Cornwell, Pruitt & Van Ness, 2001; Johnston, 2010; Miyazaki & Morgan, 2001). The events studied by sponsorship researchers are announcements of a sponsorship deal/ contract (Calderon-Martinez, Mas-Ruiz & Nicolau-Gonzalbez, 2005; Clarke, Cornwell & Pruitt, 2002, 2009; Cornwell, Pruitt & Clark, 2005; Farrell & Frame, 1997; Johnston, 2010; Miyazaki & Morgan, 2001; Spais & Filis, 2008), or the sponsor’s team winning an event (Cornwell, Pruitt & Van Ness, 2001). Where these events produce abnormal returns to sponsoring companies, the inference made is that the event has ‘signalled’ (Johnston, 2010; Miyazaki & Morgan, 2001) a strategically important decision or direction for the company.

Results of this research show the difficulties that exist in accounting for all variables that contribute to share price changes. Clark, Cornwell and Pruitt found, in 2002, that large, “…1.39 per cent” (Clark, Cornwell & Pruitt, 2002, p23) positive abnormal returns could be achieved by sponsoring companies. Variables contributing to this result include data on team wins, contract length, technology of the firm, and the whether the sponsor is local to the stadium sponsored (Clark, Cornwell & Pruitt, 2002). The same authors found, “…no evidence of market reaction to sponsorships” (Clark, Cornwell & Pruitt, 2009, p175) several years later. Their findings suggest that some abnormal positive returns are associated with specific sporting events (and not others), and that there appears to be an effect of new, versus existing or contract renewal, announcements (Clark, Cornwell & Pruitt, 2009). Research conducted by Cornwell, Pruitt and Clark (2005) also found that some sporting events were more likely than others to produce abnormal returns on sponsorship announcements for the

70 sponsor; and that sponsor-sponsee congruence influenced returns.

Studying philanthropic sponsorship, Calderon-Martinez, Mas-Ruis and Nicolau- Gonzalbez (2005) assess whether the purpose of the sponsorship venture had an impact on share prices. The authors found no significant abnormal returns for philanthropic sponsorships, but “…0.75% on average” (Calderon-Martinez, Mas- Ruis & Nicolau-Gonzalbez, 2005, p88) significant returns attributable to commercial sponsorships. Commercially motivated sponsorships are distinguished from philanthropic sponsorships dependent upon likely impact on “…(1) its awareness and sales, or (0) its corporate image and social recognition” (Calderon-Martinez, Mas- Ruis & Nicolau-Gonzalbez, 2005, p87).

Johnston (2010), in a more tightly controlled study, excluded announcements from the sample if the sponsor had made others (e.g., mergers; see, Johnston, 2010, p166) during the sampling period; announcements were also excluded if they related to existing contracts. This research found small (0.31 per cent; see, Johnston, 2010, p168), significant abnormal returns immediately following announcements, which then “dissipated quickly over the following days” (Johnston, 2010, p168). Hypotheses testing the effects of the value and duration of the sponsorship contract, indicated that value information did not contribute to share price fluctuation, but that duration of the contract had a significant effect on returns (Johnston, 2010). Johnston argues that in the Australian market, shorter sponsorship contracts (less than two years) are perceived as more positive than longer contracts (Johnston, 2010).

A final group of research does not support sponsorship announcements as producing positive sponsorship outcomes. This group includes Farrell and Frame (1997), Spais and Filis (2008) and Miyazaki and Morgan (2001). The research of Miyazaki and Morgan’s (2001) event study provides weak support for the idea that sponsorship announcements influence share prices. Only one measure of significant positive abnormal returns was found; and this occurred prior to the announcement day (Miyazaki & Morgan, 2001). The authors summarise, “…a lack of significant negative abnormal returns, along with one significant positive return, provides evidence that the purchase of sponsorship rights for the Olympic Games is a justifiable expense…” (Miyazaki & Morgan, 2001, p13).

Farrell and Frame (1997), studying the Atlanta Olympic Games found several significant negative abnormal returns to sponsors, but no significant positive abnormal returns. In trying to account for the effects of repeat/new sponsorships, sponsor asset value, sponsorship group and ownership, they further found no difference in returns for new versus repeat sponsors (Farrell & Frame, 1997). A single significant variable contributing to abnormal returns is ‘outside ownership’, this variable reflected large stock ownership by single individuals or organisations, which might be expected to influence corporate decisions (Farrell & Frame, 1997). The implication of this finding is that sponsorship decisions are not necessarily strategy driven (Farrell & Frame, 1997). The authors conclude that Olympic sponsorship “…may not be value-enhancing” (Farrell & Frame, 1997, p181). This conclusion is similar to Johnston; “…sponsorship expenditure in Australia is more or less value neutral” (2010, p156, 173).

Overall, research on behavioural outcomes of sponsorship reflects the need to consider the entire sponsorship environment. Naturalistic research, generally, is hampered by confounds. Event studies evaluating movement in the share prices of sponsors should be able to account for the impact of sponsorship announcements, however, it seems likely that specific events or competitions produce varying results (Clark, Cornwell & Pruitt, 2009; Cornwell, Pruitt & Clark, 2005). Other variables affecting abnormal share price movements are less consistent in their effects, such as new versus renewal contracts (Clark, Cornwell & Pruitt, 2009; Farrell & Frame, 1997; Johnston, 2010); and winning (Cornwell, Pruitt & Van Ness, 2001; Clark, Cornwell & Pruitt, 2002). Johnston (2010) acknowledged the problem of confounds when trying to eliminate announcements from her sample, however, only Spais and Filis (2008) considered fluctuations in share values of the sponsee in tandem with consideration of the sponsor share price. It would seem that the opportunity for further research in this area is vast.

Conclusions

The objectives identified here have been comprehensively studied and categorised in

72 a variety of ways in the sponsorship literature (see: Cornwell, Weeks & Roy, 2005; Meenaghan, 1998; Pope, 1998; Pope & Voges, 2000). The purpose of this section has been to identify the means by which the impact of off-field behaviours of sportspeople could be evaluated. The hierarchy of effects argues that sponsorship activities either immediately, or cumulatively, work to achieve sales. To evaluate the impact of the off-field behaviours of sportspeople for sponsors, understanding corporate objectives is important.

Media exposure, as a sponsorship objective, does not provide a means to assess the effects of off-field behaviours. Measures of consumer awareness, similarly, will not provide a means to consider the effects of off-field behaviours because they cannot explain the valence of information. The adage that ‘all publicity is good publicity’ appeals to some. Wilson, Stavros and Westberg (2008), write: We posit that instances such as John McEnroe’s tantrums were possibly positive for the Nike brand. More research is encouraged to understand if transgressions can have a positive impact and how practitioners may take advantage of this. Wilson, Stavros & Westberg, 2008, p106 Measures of exposure and awareness cannot evaluate whether negative behaviours can have positive impacts, only whether people know of them.

Measures of behaviour, such as share price variations, can indicate valence in the perception of off-field behaviours. This area of research, however, has yet to provide a detailed discussion of the variety of sponsorship markets evaluating sponsorship information. One might argue that share prices are determined by individual or corporate investors, brokers, and investment funds; the perceptions of people engaged in trading may vary considerably from brand end-users or event consumers.

Risks to share value are easy to anticipate, Knittel and Stango (2010) also provide evidence. These authors evaluated the effects of the news of Tiger Woods’ extramarital affairs on his sponsors’ share prices for the 10- and 15-day windows following the car accident that prompted Woods’ affairs being announced, they found: The top five sponsors (Accenture, Nike, Gillette, Electronic Arts and

Gatorade) lost 2-3 per cent of their aggregate market value after the accident, and his core three sponsors EA, Nike and PepsiCo (Gatorade) lost over four per cent. Knittel & Stango, 2010, p i. These results are interesting. An extension of this research should include a comparison of share values, and any rebound, for companies that maintained their sponsorship of Woods relative to those that cancelled their contracts. Given the research of Knittel and Stango (2010), it appears possible that the effects of large off- field scandals on sponsors’ share prices can be measured. It is questionable, however, whether the off-field behaviours of sportspeople in team sports, or whose personal profile is less than Woods’ would produce a measurable impact on share values.

The difficulties associated with media exposure objectives, and behavioural measures of sponsorship success, render the appropriate focus for this research, the category of consumer-based objectives. These objectives consider the valence of off- field information presented (generally negative), and provide the means to evaluate impact of any transfer of consumer perception to corporate image or brand attitudes. Occupying the intermediate rung on the ‘hierarchy of effects’, consumer-based objectives assume consumer awareness or exposure, and anticipate a behavioural impact equivalent (correlated with) the valence of attitudinal evaluations.

This section has identified the goals of corporate sponsors and the means of measuring the successful achievement of sponsorship objectives by sponsorship researchers. There remains for this literature review the question of how sponsorship works in the mind of consumers to achieve sponsorship objectives.

Theories that explain consumer responses to sponsorship

This section addresses question c for the literature review: What theories explain how sponsorship ‘works’ in the minds of consumers to achieve sponsorship objectives?

Sponsorship researchers continue to query how sponsorship works, or how sponsorship messages work in the mind of the consumer (see: Cornwell, Weeks & Roy, 2005; Dees, Bennett & Villegas, 2008; Gwinner, 1997; Javalgi et al, 1994; Lee, Sandler & Shani, 1997; Madrigal, 2000; Pope, 1998), to achieve sponsorship objectives. This is rendered in a variety of ways, but typical examples include: …an understanding of how sponsorship “works” has yet to be developed… Gwinner, 1997, p146); and,

…ignored the underlying theories and conceptual foundations that explain how sponsorship operates. Madrigal, 2000, p14 Cornwell, Weeks and Roy suggest that the basis of understanding how sponsorship ‘works’ can be found in information processing theories (2005, p21). These theories explain the response of individuals to information; how it is processed, remembered, recalled, and how it influences interpretation of new information. An assumption of this thesis is that information processing is at the foundation of all consumer responses to sponsorship. Impression formation theories are also reviewed to demonstrate how information about people is processed as a subset of information processing theory.

A variety of theories is currently used to explain sponsorship activities and intended outcomes; most derive either explicitly or implicitly from information processing models. These include, as shown in Table 2.5, mere exposure, cognitive consistency theories of balance and congruence, the social identity theory of identification, classical conditioning and associative networks (Cornwell, Weeks & Roy, 2005). This review addresses balance theory and the social identity theory of identification in addition to the information processing model.

Balance theory and identification are reviewed here because of the nature of sports and sports sponsorship. Sports, as discussed earlier, are subjects of intense interest, audience commitment and affect (Burnett, Menon & Smart, 1993; Mahony, Madrigal & Howard, 2000; Neale & Funk, 2006; Quester & Farrelly, 1998; Shannon, 1999). When corporate sponsors use sports to achieve such corporate objectives as brand

75 awareness, corporate image or brand attitude development, they are aiming to capitalize upon that property-based affect. Social identity theory (Tajfel, 1974, 1978) is reviewed here to explain the relationship of the individual to sports teams because those group memberships or group relationship affect perceptual valence.

Balance theory (Heider, 1958), as a cognitive consistency theory, argues that the individual prefers that linked entities are regarded equivalently (of similar valence). The individual, when reviewing the relationships of their friends, is inclined to prefer or approve of their friends’ friends (Von Hecker, 2004). In the sports sponsorship context, balance theory argues that the sports team, in a positive relationship with the individual, endorses the team’s sponsor as a friend to the individual.

The information processing model is implicitly useful to this research as it identifies the myriad ways human attention and memory can be developed. Balance theory and social identity theory, alternatively, are explicitly used, later in the thesis, to develop hypotheses and evaluation consumer evaluations of news’ reports.

The Information Processing Model

Within the field of consumer learning, frequent objects of study include: type of information (Erdem et al, 1999; Moschis, 1981, West, Brown & Hoch, 1996); information acquisition and processing (Biehal & Chakravarti, 1982; Heckler & Childers, 1992; Hutchinson & Alba, 1991; Van Osselaer & Alba, 2000); and retrieval and transfer issues (Dick, Chakravarti & Biehal, 1990; Gregan-Paxton & John, 1997; Hoch & Deighton, 1989; Hoch & Ha, 1986). These areas span straightforward learning subjects (learning contexts and learning skills; and, types of information); to the theoretical (how can the learning process be improved or blocked at different stages), and the very complex (how consumers elaborate on and use previously acquired information). Each of these areas addresses some element of the memory process.

The information processing model of learning is characterised by three key processes: acquisition, retention, and retrieval (Crowder, 1976; Hastie & Carlston,

1980). Memory itself is often presented as divided into long-term memory (LTM) and short-term memory (STM); a further category, working memory is sometimes used to explain the current processing of information (Bettman, 1979; Hastie & Carlston, 1980; Tulving, 1995). Each of these processes is reviewed here.

The information processing model of learning suggests that individuals presented with sensory information will process that information and retain it in memory according to their individual characteristics or situational influences (Almeida, 2006; Biehal & Chakravarti, 1982; Bettman, 1970, 1979; Johar, Maheshwaran & Peracchio, 2006; White & Carlston, 1983). Central to the model is the idea that processing of information is serial, parallel, and independent (Moscovitch, 2007; Tulving, 1995). It further suggests that processing follows the sequence of information presentation, and that encoding occurs following the successful processing of sensory information. The storage of encoded information may occur simultaneously in parallel systems (as, for example, both an episodic/event based memory, and also a general/ semantic principle), however, information is retrieved (independently) from a single system only (Tulving, 1995).

Acquisition

Acquisition is the first stage of the information processing model, and contain two sub-stages: the first of these is attention to target information (sensory stimuli), and the second, the encoding process (Hastie & Carlston, 1980). Target information includes “actions, appearances, conversation” (Hastie & Carlston, 1980, p4) in the case of impressions of people, and includes more diverse sensory stimuli in other contexts. In the marketing context, target information is sourced from advertising, point-of-purchase sources, salespeople and others (Alba & Hutchinson, 1987; Hirschman & Wallendorf, 1982; Keller, 1987), and may consist of product attribute and brand information (Alba & Hutchinson, 1987). Effectively, any ‘thing’ that might draw the attention of an individual and might contribute eventually to a product choice may be considered target information in the information processing model.

Encoding is the process that “intervenes between the perception of an event and the creation of the corresponding trace, a process ... that converts the stimulus energy into mnemonic information” (Tulving & Thomson, 1973, p354). Encoding works to develop “neural representations” of stimuli in memory (Craik, 2007, p129). However, not all stimuli are processed equally which means that the encoding process is made more or less efficient according to a number of factors (Craik, 2007).

The encoding of information is influenced by several characteristics of the individual, including: level of experience of the individual with the stimuli set (Alba & Hasher, 1983; Alba & Hutchinson, 1987; Craik, 2007; Murray, 1991); goals related to information use (Biehal & Chakravarti, 1982; Craik, 2007, and decision risks (Murray, 1991; Murray & Schlacter, 1990; Hirschman & Wallendorf, 1982); motivation of the individual to attend to stimuli (Craik, 2007; Hirschman & Wallendorf, 1982; Park & Hastak, 1994), and to engage in cognitive processing (Lynch & Srull, 1982). Encoding and storage is also influenced by such environmental issues as learning context (Craik, 2007; Keller, 1987); and task structure (Biehal & Chakravarti, 1982; Keller, 1987). Each of these influences exists as a potential limitation to the measurement of learning and information processing.

Retention

Memory is a term that is used in a variety of ways. Tulving (2000) listed these, suggesting that memory is, alternatively: the capacity to encode; the 'compartment' that stores information; the information that is stored; or the individuals' process of recalling something remembered. These senses of memory were later refined by Schacter (2007) who rejected “the individual's phenomenal awareness of remembering” as a form of memory on the basis that it cannot account for structural elements of memory. Common elements to a definition of memory suggest that memory is a “persisting change in the nervous system” (Schacter, 2007, p25) that produces a mental representation of events or an individuals’ experiences (Moscovitch, 2007; Schacter, 2007).

Memory is divided into short-term and long-term memory stores, occasionally

78 including distinct sensory stores, and/or working memory (Bettman, 1970, 1979; Atkinson & Shiffrin, 1968; Hastie & Carlston, 1980; Lynch & Srull, 1982). There is some debate as to whether a split between short-term and long-term memory can be justified (Keller, 1987). The distinction is retained, however, on the basis that it provides researchers with the means to explain different levels of processing and the cases where stimuli are not retained accessibly in long-term memory (Bettman, 1979; Horton & Mills, 1984; Keller, 1987).

Where the distinction between short-term and long-term memory is retained, short- term memory is referred to as “the locus of current processing activity” (Bettman, 1979, p37). A characteristic consistently used to distinguish short-term memory is its limited capacity, noted as “six or seven symbols” (Hastie & Carlston, 1980, p10), or “chunks” of information (Alba & Hutchinson, 1987; Bettman, 1979). Long-term memory alternatively is the enduring storage system, from which information is drawn into short-term memory for processing, in order to interpret “incoming information” (Bettman, 1979, p38).

Information stored within memory takes three forms: semantic, episodic/event, and procedural. Semantic memory was conceptualised originally as consistent with its name. Tulving provided that semantic memory is a: ...mental thesaurus, organised knowledge a person possess about words and other verbal symbols, their meaning and referents, about relations among them, and about rules, formulas, and algorithms for the manipulation of these symbols, concepts, and relations... (Tulving, 1972, p386) However, it is now more common to see semantic memory referred to as “general knowledge” (Almeida, 2006; Schacter, 2007; Tulving, 1995), or “factual information in the broadest sense” (Tulving, 1995, p841).

Alternatively, episodic memory stores information about personal experiences that are tied to both a temporal and spatial location (Almeida, 2006; Hastie & Carlston, 1980; Schacter, 2007; Tulving, 1972; Tulving & Thomson, 1973). Hastie and Carlston (1980) refer to 'event memory' (episodic memory), and conceptual memory stores instead of episodic and semantic stores. There is little to differentiate the characteristics of the stores. However, a particular point of difference is the notion

79 that conceptual (semantic) memory for these authors, contains both; “general categories for the description of people, social actions, and events” which might be consistent with the notion of 'general knowledge', and also “reasoning procedures and strategies”, and (Hastie & Carlston, 1980, p9). These reasoning procedures and skills sets are, for Schacter (2007), the characteristics of procedural memory, and distinct from semantic memory.

Retrieval

Memory is said to have a network structure and the principle elements are concepts (Dudai, Roediger & Tulving, 2007), also referred to as “cognitive elements” (Ostrom, Lingle, Pryor & Geva, 1980, p57). In long-term memory the network of concepts (represented as nodes) are joined by links that represent the relationships between concepts (Bettman, 1970, 1979; Keller, 1987). Concepts, or cognitive elements are “mental representations that encode sets of attributes that describe real or imaginary classes of items, processes or relationships”, they are “always linked to other concepts” (Dudai, Roediger & Tulving, 2007, p2). Where several concepts are closely related in memory, these concepts may be associated in memory in a theme (Ostrom, Lingle, Pryor & Geva, 1980), or schema (Alba & Hasher, 1983; White & Carlston, 1983). It is also noted that concepts are not “the terms in language that are used to express them” (Dudai, Roediger & Tulving, 2007, p2), which is why the concept “Salman Rushdie” represents neither the person, nor a complete or necessarily accurate representation of the man or his works.

Retrieval of information from memory, according to the standard memory model, is contingent upon the decoding process and the individuals' behaviours related to the information processing task, e.g., “recollection, evaluation, information seeking” (Hastie & Carlston, 1980, p4). This model suggests that retrieval of information from memory occurs according to the need of the individual (as in the case of recollection), and in response to cues or prompts that activate concepts in memory (see: Keller, 1987; Lynch & Srull, 1982; Tulving & Thomson, 1973). Retrieval is also dependent upon factors related to the encoding process (Lynch & Srull, 1982; Tulving & Pearlstone, 1966; Tulving & Thomson, 1973).

The encoding specificity principle suggests that similar cues and context at the point of retrieval as existed at the point of encoding will promote successful information retrieval (Higham, 2002; Kardes, Kalyanaram, Chandrashekaran & Dornoff, 1993; Tavassoli & Fitzsimons, 2006; Tulving & Thomson, 1973). Essentially this principle argues that it is easier to recall something in a similar situation to that you learned it in because there will be more to remind you of your initial exposure. Tulving and Thomson suggest also that retrieval of information “depends on how it was stored” (Tulving & Thomson, 1973, p359), a point that refers to the difference between availability and accessibility. Potentially, all information stored in memory is available to be recalled, however, all concepts in memory are not necessarily accessible all of the time (Lynch & Srull, 1982). This idea, that memory traces are not consistently available, relates also to the theory of spreading activation which will be discussed later.

Carlston and Skowronski (2005) also support the notion that the encoding process influences ability to recall information. These authors suggest that information that is encoded via an attributional process is more likely to be accessible than that encoded via an associative pathway. This theory differs significantly from Craik's 'levels of processing model' (Craik & Lockhart, 1972; Craik, 2002) that suggests that processing is contingent upon individual characteristics. The focus of Carlston & Skowronski's theory is instead that processing will enable, or eventually hinder, retrieval on the basis that concepts are explicitly linked during processing (2005). Associative processing, according to these authors, is “characterised as a relatively shallow activity that yields generic, unlabeled linkages in memory” (Carlston & Skowronski, 2005, p844). Attributional processing, alternatively, is an “…elaborative mental activity… - …resulting in the formation of labeled associative linkages that designate one construct as the property of another” (Carlston & Skowronski, 2005, p844). Recall, then, depends on the strength of the schema and relationships between concepts in the schema.

The proposal that associative and attributional processes differently effect recall (Carlston & Skowronski, 2005) is consistent with the notion of 'spreading activation' (Alba & Hasher, 1983; Burke & Srull, 1988; Keller, 1987). If concepts are

81 accessible, then retrieval occurs when a cue activates a concept. Spreading activation refers to the retrieval of information from long-term memory where “a particular node in LTM [long-term memory] is activated by a retrieval cue, and activation spreads from that node to other linked nodes in LTM” (Keller, 1987, p317).

Impression formation processes

Impression formation theories are concerned with the study of person perception and person memory (Hastie & Carlston, 1980), in particular, the initial or early perceptions of an individual. Impression formation is reviewed here because individuals’ perceptions of athlete behaviours off-field are central to the research reported.

Asch provides an insight into the study of impression formation: “We look at a person and immediately a certain impression of his character forms itself in us” (Asch, 1946, p258). The very first time we meet a person we begin to develop an idea about them, who they are, what they represent, and how they might behave (Jenkins, 2004). If we have a continuing relationship with that person we re-evaluate our impression of them each time we learn something new, dependent upon our interest in that person or motivation to attend to new information (Fiske & Neuberg, 1990). On this basis, an impression or person impression is “a heterogeneous collection of [cognitive] elements” (Ostrom, Lingle, Pryor & Geva, 1980, p56), where each element is distinguishable from the others. These elements include the behaviours of the target individual, their speech, mannerisms, traits or appearance.

Hastie and Park (1986) provide the most thorough review of the models of impression formation processes. These authors suggest that impression formation processes exist in three categories: independence, memory-causes-judgment, and judgment-causes-memory.

Independence

Independence models of impression formation suggest that memory for information is not stored with reference to judgment information (Hastie & Park, 1986). When presented with an information stimulus, we encode and store information in long- term memory so that it may be later recalled and used in judgment. The SPI (serial, parallel, independence) assumption (Tulving, 1995) is endorsed. This means that information is processed and stored in parallel stores (event-based and semantic memory), and can be recalled independently from a single store.

The 'independence' models best explain processes of inquiry. Inherent within them is the primary goal of organising or processing information. The formation of a judgment remains a secondary goal in these models.

Memory-causes-judgment

The memory-causes-judgment category suggests that information that has been processed may later be recalled in order to assist in judgment making (Hastie & Park, 1986). In a judgment situation, information in memory is used to produce the judgment, a key assumption being that there is a temporal separation of memory storage and judgment. The most common example of this type of model includes the 'availability' or 'accessibility' models (Hastie & Park, 1986), such as the Adaptive Networks Model (Carlston & Skowronski, 1986; Wyer & Carlston, 1979), or the Human Associative Memory (Anderson & Bower, 1973.

This group of models is not associated with notions of bias or faults in the encoding or retrieval of information (Hastie & Park, 1986). However, the notion of availability suggests that information that is recalled from memory in order to contribute to a judgment is implicitly biased, in that it is recalled. This 'bias' exists because other, perhaps equally relevant information has not been recalled.

Judgment-causes-memory

The judgment-causes-memory models suggest violations of the SPI assumption. 83

These models propose that when an information stimulus is presented and encoded into working memory, it is used immediately to form a judgment; a memory of the initial judgment and/or the information, is then stored in long-term memory (Hastie & Park, 1986). Later judgment situations are then influenced by earlier judgment- based processing. The influences of judgments on encoding, retrieval and processing are reflected in three sub-categories: biased retrieval; biased encoding; and incongruity-biased encoding.

'Biased retrieval' theories argue previous judgments influence later judgments by recalling information most closely related to the initial judgment (Hastie & Park, 1986). An example of the biased retrieval perspective is Burke and Srull's (1988) research on competitive interference; it argues that recall of earlier judgments is deflected through exposure to later stimulus information.

'Biased encoding' processes suggest that an initial judgment influences later judgment processes by filtering “subsequent stimulus by guiding search, encoding & comprehension” (Hastie & Park, 1986, p260). Many examples of this type of processing exist. Hutchinson and Alba (1991), for example, suggest that judgment influences later learning, where the simplifying role of categorisation may inhibit the storage of complex product information. Other research in this area includes self- referent encoding (Klein, Loftus & Burton, 1989; Sedikides & Skowronski, 1993); and paired-learning (Carlston & Mae, 2007; Carlston & Skowronski, 2005).

The final type of judgment-causes-memory impression formation process is the 'incongruity-biased encoding' model. Hastie and Park (1986) suggest that initial judgments influence later judgment situations where later, incongruent stimuli receive extended processing. The incongruence of the information, relative to that already held in long-term memory, is thought to enhance memorability. Examples of research using this logic include: Heise and Smith-Lovin's (1981) studies of context- congruent behaviours; Neuberg's (1989) argument that expectancies influence perceptions, and the work of Sedikides and Skowronski (1993) on behavioural consistency.

Mixed Models

Much of the research in impression formation exists within a mixed-model framework. Sherman & Klein (1994) suggest explicitly that impression formation processes include the abstract models, exemplar models, and mixed models. Fiske and Neuberg's continuum model was developed because existing processes were regarded as too static, or rigid in their approach. Their argument is that people “...people do not use just one strategy to understand other people; they use a variety of strategies” (Fiske & Neuberg, 1990, p1), and impression formation processes should reflect these imperfect human behaviours.

Conclusions on Information Processing &Impression Formation

The information processing model and theories of impression formation provide insight into how individuals process, store and recall information for use. These theories are implicitly important for this thesis because, without studying specific mechanisms involved in the judgment of sportspeople, teams and sponsors, it is recognised here that several viable theories exist to explain those judgments. The research context for this thesis is provided by an existing sporting code and team, which argues that information processing and judgments cannot be assessed with certainty without neuro-imaging technologies. Instead, the position taken here is to endorse Fiske and Neuberg’s view that people “...use a variety of strategies” (Fiske & Neuberg, 1990, p1) to process information and form judgments, and that those strategies are governed by individual differences.

Balance theory

Cognitive consistency is the “postulate that people tend to maintain logical consistency among their cognitions (and even between cognitions and more gross behaviour)” (McGuire, 1960, p345). In person perception this has the effect of biasing expectations and judgments in the pursuit of cognitive simplicity or economy (Peeters, 1971). The desire to maintain consistency in judgments has a further effect of reducing dissonance as the individual strives to avoid the discomfort that might

85 arise from engaging in counter-attitudinal behaviours (Gawronski & Strack, 2004).

Balance theory is one theory within the family of cognitive consistency theories. Another popular consistency theory used in sponsorship research are the congruence theories which argue both that congruence promotes recall, and also that incongruence is memorable (Dardis, 2009; Rifon et al, 2004).

Balance theory in sponsorship research is used to explain the motivation of an individual to develop a positive attitude towards a sponsor (see: Dalakas & Levin, 2005; Dean, 2002; Van Heerden, Kuiper & Saar, 2008; Reisinger, Grohs & Eder, 2006). The theory was developed by Heider (1958) to explain the desire of individuals to maintain cognitive balance, or, a consistency across evaluations of linked or related individuals. Cognitive balance is regarded as attractive because “maintaining inconsistency requires cognitive ability and some extra effort” (Crandall, Silvia, N'Gbala,Tsang, & Dawson, 2007, p13). Balance, then, provides individuals with the means to integrate person information with existing judgments on the basis of relationship cues.

A general approach to interpersonal relationships is explained using such sayings as “my enemy's enemy is my friend”, and “my friend's friend is also my friend” (Von Hecker, 2004, p28). In this sense, balance theory explains that individuals' maintain cognitive balance through the simplification of their approach to interpersonal relationships. The justification for liking a person derives from an assumed, if unknown, similarity when we know that person is liked by a friend. In the same way, disliking someone is justified on the basis that they are dissimilar to a cognitive representation of the self.

Heider (1958) explained relationships among dyads and triads, and for unit relations and sentiment relations; some examples and the general rules to recognise balance are provided here. In triads involving two people and an object; p is the perceiver; o represents a person to whom p is connected in some way; and x is used to denote an object, idea or impersonal entity (Heider, 1958; Srull, 1981; Zajonc, 1960). Sentiment relations are represented by Heider (1958) as L (like) or DL (dislike), which are differentiated from unit relations (U). Unit relations reflect the idea of a

86 connection between persons, or persons and objects, that are not strictly sentimental, these include family units, co-workers, similarity, ownership etc., (Heider, 1958). According to Heider, “U denotes the cognitive unit between two entities, and notU the fact the two entities are segregated” (1958, p201). This system provides the means to identify two forms of positive relation (L and U), and two forms of negative or avoidance relation (DL and notU) (Heider, 1958).

A further consideration is role of p and the direction of sentiment. The perceiver, p, is generally taken to be the center of focus, or the ego under investigation. Unit relations are considered to be reciprocal and both parties to the unit are expected to recognise the relationship. Sentiment, however, may be unreciprocated, and therefore directional.

Balance, for a triad, can be assessed according to three basic structures. Balance is seen when all relations among the persons or entities are positive, or when two negative and one positive relation exist (Heider, 1958). When two positive relations and one negative relation are seen, the result is cognitive imbalance (Heider, 1958). An imbalanced triad, according to Heider (1958) and others (see: Dalakas & Levin, 2005; Dardis, 2009; Dean, 2002; Osgood, 1960) is evidence of cognitive stress. This stress creates motivation for the perceiver to engage in a process of attitude or affect change toward a balanced state. Diagrams of balanced (triads A and B) and unbalanced (triad C) relations are provided in Figure 2.4.

Figure 2.6: Balanced and unbalanced triads

A) B) C) p p p

U o DL x o L x o x o x o x

Triads A and B in Figure 2.0.4 depict balanced relationships; triad C is unbalanced.

Triad A provides the example of three positive relations in which p likes both o and x, and the relationship between o and x is a (positive) unit relation. In triad B, p dislikes o and has a relationship or association with x; although o dislikes x, the relationship is balanced because p’s dislike of o is not inconsistent with o’s dislike of x. In each of these examples, the balance achieved reflects a situation of cognitive consistency for p. In the final example, p may be motivated to re-evaluate his/her relationship with o or x because the triad is unbalanced.

Sponsorship research using balance theory expects that when the potential consumer of the sponsor’s brand is a fan of x (the event/sport), cognitive stress or the desire for balance will motivate p to like o (the sponsor’s brand). This configuration of expected sponsorship effects is endorsed, and empirically tested by numerous researchers (see: Dalakis & Levin, 2005; Dardis, 2009; Dean, 2002; Reisinger, Grohs & Eder, 2006; Van Heerden, Kuiper & Saar, 2008).

Three sponsorship papers present similar hypotheses and consistent results endorsing balance theory, these are Dalakis and Levin (2005), Dean (2002) and Reisinger, Grohs and Eder (2006). Dean’s research finds that sponsorship of a liked charity event contributes positively to perceptions of corporate community relations (Dean, 2002). Dalakis and Levin (2009) explain that fans of a particular NASCAR driver will like the sponsor of that driver, and further, will not like the sponsor of rival drivers. Results of their research support the hypotheses. Finally, Reisinger, Grohs and Eder hypothesise that “Fans of team A should have a more positive attitude towards sponsors of team A than neutral spectators” (2006, p2), and that spectators who do not support a team will also not support that team’s sponsor; these hypotheses are also empirically supported.

The question of how balance works to achieve sponsorship objectives remains to be addressed. The mechanisms underlying balance theory have been tested by Picek, Sherman and Shiffrin (1975), Hummert, Crockett and Kemper (1990) and Gawronski, Walther and Blank (2005). These authors suggest that when an individual is given the task of evaluating relationships, cognitive balance and consistency drive inferences more often than alternative methods and that serial encoding of information explains its’ success (Gawronski, Walther & Blank, 2005;

Hummert, Crockett & Kemper, 1990; Picek, Sherman & Shiffrin, 1975).

Balance theory tests of recall argue that a balanced network of social relations is preferred by the individual (Heider, 1958, cited by Picek, Sherman & Shiffrin, 1975). Methodologically, balance preferences are expected to drive easier and therefore faster recall (Picek, Sherman & Shiffrin, 1975). Further, when the stimulus is an imbalanced network, the individual will be motivated to change relationship valences to create balance, which leads to an increased likelihood in recall tests of errors in imbalanced networks as the individual tries to recall a balanced network (Picek, Sherman & Shiffrin, 1975).

To test balance theory propositions Picek, Sherman and Shiffrin (1975) presented respondents with sets of four-person social structures. Four structures were designed, the characteristics of these structures were: a) complete and balanced; b) complete and imbalanced; c) incomplete and balanceable; and d) incomplete and nonbalanceable. A structure was complete when all social relationships were defined, and otherwise incomplete. A balanceable structure was one which, through inference, an individual might construct a balanced structure; a non-balanceable structure, even through a process of inference, cannot be balanced. Results for this research established that recall is best for “balanced or balanceable structures” (Picek, Sherman & Shiffrin, 1975, p766). This result is further explained by the notion of serial processing.

The balance theory proposition that argues better recall of relationships due to preference for balance was supported in both balanced and un-balanceable structures in this research because the first two (of four) relations presented in suggest the story is balanceable. Recall for later relationships declines when the third relationship presented suggests an unbalanced structure. Picek, Sherman and Shiffrin explain: The difference between balance(d) (able) and nonbalance(d) (able) structures is negligible until presentation position three (the first relationship indicating imbalance), at which time performance drops for the nonbalance(d) (able) structures. This finding is explicable in terms of a serial encoding model… Picek, Sherman & Shiffrin, 1976, p766 The argument endorsed by this result is that the individual develops an integrated

89 cognitive representation as they are presented with each new piece of information; but integration falters when new information has the effect of unbalancing the social structure. Given a preference for balanced information, unbalancing information is more likely to be incorrectly encoded or forgotten.

The authors go on to study the pattern of mistaken inferences. Mistaken inferences are referred to as ‘intrusions’. A balance theory proposition would argue that all ‘intrusions’ have the intention of balancing relations in a social structure, this result, however, was not found. Intrusions instead reflect the composition of the social structure when missing relationship information is encountered (Picek, Sherman & Shiffrin, 1975). Effectively: For balanceable stories, subjects made responses to nonstated links in the direction of balance; for nonbalanceable stories they made responses to nonstated links in the direction of imbalance. Picek, Sherman & Shiffrin, 1975, p765 This finding is one that argues strongly in favour of both serial encoding and cognitive consistency. It suggests that when the encoding of information encounters an unstated relationship, the most likely inference is one consistent with the pattern of relationships evident at that point in time.

Hummert, Crockett and Kemper (1990) conducted their experiments on memory for interpersonal relationships working from the premise that individuals use a balance approach when it is useful, but abandon it when it is not. In three experiments they compare individuals’ use of a balanced approach and a ‘propositional model’. The authors argue that the balance approach, and serial encoding, enables quick recall of information from an integrated memory structure once social information has been learned (Hummert, Crockett & Kemper, 1990). The propositional model, alternatively, says that pieces of information about a social structure are encoded and stored separately (Hummert, Crockett & Kemper, 1990). This will provide fast recall when a specific piece of information is cued and recalled, but that response time “…should increase monotonically” (Hummert, Crockett & Kemper, 1990, p8) when the individual is required to infer relations, and according to the complexity of that inference.

Results of the experiments conducted by Hummert, Crockett and Kemper (1990) argue the balance theory approach to memory of social structures is used predominantly. In each experiment, when response times were expected to increase due to the complexity of inference required, “…in no case was such an increase observed” (Hummert, Crockett & Kemper, 1990, p20). Consequently, it appears that the propositional model does not explain how individuals’ process information related to the structure of social relationships.

Gawronski, Walther and Blank (2005) also challenged the serial encoding thesis by providing that balancing social structures may occur during a judgment task. In order to test this proposition, Gawronski, Walther and Blank (2005) tested respondents’ explicit and implicit attitudes toward ‘target individuals’ after they had developed a positive or negative attitude towards a source individual, and the source’s attitude toward the target. The logic underlying this method argues that if balance developed during encoding, attitudes toward the target individual would be evident in testing of both implicit and explicit attitudes; whereas, if balance was developed during a judgment task it should affect only explicit attitudes (Gawronski, Walther & Blank, 2005).

The research established, across three experiments, that both explicit and implicit attitudes towards a target individual are influenced by the two-way interaction of source valence and observed sentiment (source’s attitude toward the target). Results endorse the serial encoding view of balance theory because information presented in the learning stages of the research was integrated prior to testing, determining that both explicit and implicit attitudes were later available for testing (Gawronski, Walther & Blank, 2005).

The sponsorship research, thus far, has failed to explain the cognitive mechanisms underlying balance theory beyond reference to cognitive consistency. Evidence from sponsorship and psychological research establishes the consistency of balance approaches to social relationships. Further, psychological research explains how balance works in the individual. As a process of cognitive economy, serial processing provides an integrated memory trace that can be accessed explicitly or implicitly.

Where congruence research is best suited to the study of the objective qualities of the sponsors’ brand and the sponsee in terms of image matching; balance theory deals with valenced interpersonal (or inter-object) relations, and the motivational stress to re-evaluate relations when sentiment relations are imbalanced. Consequently, balance theory is the more appropriate cognitive consistency theory to use when sentiment relations exist between sponsorship entities.

Identification

This review of identification is connected to the ‘how it works’ section because of the importance of the phenomenon in sports marketing and sponsorship research. The scholarship relating to identification is extensive, and there are a variety of means available to review it. This review will have six components: 1) A summary of Social Identity Theory principles (Tajfel, 1974, 1982) 2) Antecedents & Consequences of Identification 3) Types of Groups 4) Characteristics of Psychological Group Identification (Ashforth & Mael, 1989) 5) Identification Dimensionality 6) Sports, Sponsorship and Marketing Research in Identification The purpose of this section is to extend the discussion of identification as introduced in the earlier section: The Many Affects.

Identification, rather than any other property-affect measure, is reviewed because it explains the role of self-esteem in individual and group evaluations (Ashforth & Mael, 1989; Ellemers, Kortekaas & Ouwerkerk, 1999; Hogg & Terry, 2000; Parker, 2007; Wann, 2006)

Identification is a construct derived from social identity theory. Social identity theory suggests that the individual has two forms of identity (personal and social), which contribute to the individual's self-concept (Abrams & Hogg, 1988; Ashforth & Mael, 1989; Van Leeuwen, Quick & Daniel, 2002). Personal identity accounts for “what

92 one thinks or feels about oneself”, whereas the individual's social identity comprises “a number of group classifications with which one identifies” (Van Leeuwen, Quick & Daniel, 2002, p108). By this reasoning, identification explains the orientation of the individual toward groups as enacting identities that are social, rather than personal or private.

Principles of Social Identity Theory

The foundations of social identity theory are explained by Tajfel (1974, 1978, 1982), and Tajfel and Turner (1979), and summarised by Ellemers (1993) and others. According to Tajfel, the desire to explain intergroup behaviour originated in the aim to explain “…attitudes and behaviour towards outgroups” (Tajfel, 1974, p66). The focus on outgroups is a complex one as social identity does not focus on individual psychological processes, but those social psychological processes that explain how the individual acts as a member of a group. Underlying all social identity theory investigation is the situation from which a series of questions arise: …in order for the members of an ingroup to be able to hate or dislike an outgroup, or discriminate against it, they must first have acquired a sense of belonging to a group which is clearly distinct from the one they hate, dislike or discriminate against. Tajfel, 1974, p66 Although little social identity theory research deals with the idea of ‘hate’, for Tajfel (1974) identifies the four principles of social identity theory. Those principles of social identity, for Tajfel (1974), are: categorisation, social identity, social comparison, and psychological distinctiveness.

The categorisation process is used by the individual as a means to “systematise and simplify his environment…,” (Tajfel, 1974, p69). By aiding the individual to recognise social divisions, similarities or differences, the individual’s approach to a group might be determined (see: Ellemers, de Gilder & Haslam, 2004; Fielding, Hogg & Annandale, 2006; Tajfel, 1974). Self-categorisation says that: …in many situations people organise social information by categorising individuals into groups. This enables them to focus on collective properties

that are relevant to the situation at hand (e.g., students versus teachers), while neglecting the ‘noise’ of other variations (e.g., differences in age or clothing style) that occur among individuals within the same group. Ellemers, de Gilder & Haslam, 2004, p462 These approaches to social interaction suggest that adoption of a social identity is purposeful. Categorisation suggests what the role of the individual might be, and drives inferences which differentiate groups.

Categorisation also has implications to perception of group members. In the process of identifying the self as a group member, one recognises the out-group, and develops expectations about characteristics of in- and out-group members. Hogg and Abrams (1999) explain this as the depersonalisation effect of categorisation in which individual differences are deferred. Categorisation directs attention to prototypical characteristics which aid self-perception and representation, as well as interactions with out-group members (Hogg, 2000, 2009; Hogg & Abrams, 1999). The use of prototypes is also strongly associated with homogeneity effects in perception.

As a result of the inclination to simplify perception during the process of categorisation, various authors argue that perception of group members tends towards homogeneity (see: Brewer, 1993; Haslam, Oakes, Reynolds & Turner, 1999). This theoretical development is gradually becoming more complex as the homogeneity effect has been revised to consider that ingroup members may be regarded as heterogeneous (Brewer, 1993; Hutchison, Jetten, Christian & Haycraft, 2006); whereas outgroups are more likely to be perceived as homogeneous (Haslam, Rothschild & Ernst, 2000; Hogg, 2000).

The four principles, although not advocated as strongly in later works (see: Tajfel, 1982) exist in a sequence. This sequence runs “social categorisation – social identity – social comparison – psychological distinctiveness” (Tajfel, 1974, p76); it is argued to be “causal” (Tajfel, 1974, p76), suggesting necessary conditions and consequences. Social identity is defined:

…that part of an individual’s self-concept which derives from his knowledge of his membership of a social group (or groups) together with the emotional significance attached to that membership. Tajfel, 1974, p69 Later versions of this definition include ‘knowledge’ of group membership, the ‘affective’ emotional significance of that membership, and ‘value’ ascribed to the group (Tajfel, 1978, 1982). Following Tajfel’s (1974) sequence, the individual first recognises divisions (groups) in the social world, after which, they necessarily recognise their own place in it.

Once the individual has recognised their place (or, places) in the social world, it remains, to consider what the implications are, of group membership. Social comparison provides the means to consider one’s own group membership relative to alternative groups or group memberships. Social comparison argues that in addition to simplifying social interactions through processes of categorisation; “…the (relative) valence of group characteristics” (Ellemers, 1993, p29) also influences intergroup behaviour. Individuals are generally thought to seek positive social categorisations (Ellemers, 1993; Ellemers, de Gilder & Haslam, 2004; Grieve & Hogg, 1999) as providing positive effects to self-esteem (Hogg & Mullin, 1999).

The final principle of social identity theory is psychological distinctiveness. It argues that some of the value of group membership derives from a “… need for differentiation” (Tajfel, 1974, p74). This principle is also linked to the notion of self- esteem enhancement (see: Hogg, 2000; Hogg & Mullin, 1999; Tajfel, 1982). The closeness of social comparison and psychological distinctiveness is demonstrated by Hogg: Social identity rests on intergroup social comparisons that seek to confirm or to establish ingroup-favouring evaluative distinctiveness… motivated by an underlying need for self-esteem… Hogg, 2000, p123. Categorisation and self-definition, in comparison to these two later principles, appear to be the objective principles. The purposes of social comparison and distinctiveness, in comparison, are overtly related to the potential for bias, intergroup conflict, and at the individual level group-switching or mobility related goals.

The review of social identity theory thus far has yielded four principles (Tajfel, 1974), and a popular definition of social identity (Tajfel, 1974, 1978, 1982). The majority of this theorising derives from the work of Henri Tajfel. Another explanation of identification developed by Ashforth and Mael (1989) provides an alternative perspective.

Ashforth and Mael (1989) have also developed four principles of identification. Some principles are similar; others provide divergent insights and strong arguments to direct identification measurement. In order to distinguish between Tajfel’s (1974, 1982) principles and those developed by Ashforth and Mael (1989); the principles developed by Ashforth and Mael will be referred to as ‘characteristics’ in this review. The identification characteristics developed by Ashforth and Mael (1989) are: 1. Identification is a cognitive and perceptual construct 2. The individual experiences the fate of the group 3. Identification is distinguishable from internalisation 4. Group identification is relational (Ashforth & Mael, 1989, p21-22)

The first characteristic, that identification is cognitive and perceptual, explains that identification is based on self-perception and definition (Ashforth & Mael, 1989). According to Ashforth and Mael, identification is “not necessarily associated with any specific behaviours or affective states” (1989, p21). This characteristic might be interpreted as precluding the judgment of group membership on the basis of observable behaviour. The cognitive characteristic, in effect, is no different to a combination of Tajfel’s (1974, 1982) categorisation and social identity principles.

The second characteristic developed by Ashforth and Mael (1989) is problematic. The identified individual “…is seen as personally experiencing the successes and failures of the group” (Ashforth & Mael, 1989, p21). How this might be exempt from an affect-based interpretation is not clear. The authors go on to explain that identification is often “maintained in situations of great loss or suffering” (Ashforth & Mael, 1989, p21). The expectation prompted by this characteristic is that identified individuals do not leave, or attempt to leave a group when the group is disappointed

(e.g., a sporting team losing many games during a season). Commitment to a group, might be seen as loyalty; however, Ashforth and Mael specifically provide that loyalty is a construct distinct from identification (1989, p21).

The third characteristic, identification is not internalisation, provides further reiteration of the idea that identification is a cognitive construct. Stated: Whereas identification refers to the self in terms of social categories (I am), internalisation refers to the incorporation of values, attitudes, and so forth within the self as guiding principles (I believe). Ashforth & Mael, 1989, p21-22 The individual, for Ashforth and Mael (1989), retains their independence and is not subsumed within the category and its prototype. The internalisation characteristic is another that requires further debate. Brewer and Gardner (1996), for instance, argue that identification involves internalisation. Specifically, the notion of a collective (social) self “…reflects internalisations of the norms and characteristics of important reference groups…,” (Brewer & Gardner, 1996, p84).

Social identity theory recognises that individuals maintain multiple social identities (Brewer, 2001; Hogg, Terry & White, 1995; Turner, Oakes, Haslam & McGarty, 1994). These identities are enacted according to their salience (Turner et al, 1994). Where Ashforth and Mael (1989) refer to values, Brewer and Gardner (1996) refer only to norms, which remain far easier to adopt and discard contextually. It should be considered that a group which emphasises the importance of a particular system of values, to which the individual is susceptible, is likely to be a group of enduring importance to the individual. The internalisation of values therefore may overestimate the importance of the group in Ashforth and Mael’s (1989) characteristic.

The final characteristic identified by Ashforth and Mael (1989) is that group identifications are relational in the same way that personal identification is relational. Examples provided by Ashforth and Mael include the relationships between husband and wife, or doctor and patient (1989, p22). This characteristic, again, is similar to Tajfel’s reasoning that for self-categorisation and bias to occur, there must first be groups (see: Tajfel, 1974, p66 cited above).

What Ashforth and Mael (1989) add to social identity theory is the reiteration of some of Tajfel’s (1974) principles, and the opening of debate in areas where perhaps more is warranted. Of particular importance is the contribution of Ashforth and Mael (1989) to the literature on identification measurement which Tajfel does not participate in. In contributing to this literature Ashforth and Mael (1989) draw attention to the importance of construct definition and delineating antecedents and consequences. Each of the principles of social identity theory recognised by Tajfel (1974, 1982) has implications to antecedents and consequences. The diversity of social identity research following Tajfel has, however, produced the effect that antecedents and consequences can be minutely explained with reference to specific research projects, or broadly explained with sweeping statement (e.g., heightened category salience increases outgroup bias). Some of the commonly recognised antecedents and consequences of social identification are provided in the following section.

Antecedents & Consequences

In this section the antecedents and consequences of identification will be reviewed. This summary is not a census of extent identification research; instead it aims to provide an overview of the agreement (or nuance) among researchers on the subject of identification antecedents and consequences.

Antecedents

An overview of identification antecedents and consequences is provided by Ashforth and Mael (1989). These authors provide that a) group distinctiveness, b) group prestige, c) the importance of out-groups, and d) “the set of factors traditionally associated with group formation (interpersonal interaction, similarity, proximity, shared goals or threat, common history and so forth”, (Ashforth & Mael, 1989, p25) all exist as antecedents of identification. These factors, generally, argue that the value of the group as an entity (evidenced in distinctiveness, prestige, and alternative group

98 membership), and perception of group members contribute to identification.

Group prestige is recognised by a variety of researchers as an antecedent of identification, albeit in a variety of ways (see: Ahearne, Bhattacharya & Gruen, 2005; Cornwell & Coote, 2005; Ellemers, 1993; Gwinner & Swanson, 2003; Smidts, Pruyn & Van Riel, 2001). In management and sponsorship research, Cornwell and Coote (2005), Gwinner and Swanson (2003), and Smidts, Pruyn and van Riel (2001) each use a specific measure of organisation prestige. In each case, perceptions of prestige have statistically significant positive effects on identification.

Alternative measures of prestige are ‘construed external image’ (Ahearne, Bhattacharya & Gruen, 2005), and group status (Ellemers, 1993). Results for these constructs have been mixed. For Ahearne, Bhattacharya and Gruen (2005) ‘construed external image’ in a structural equation model, does not contribute directly to identification; instead it contributes to ‘perceived salesperson characteristics’ and ‘perceived company characteristics’ and each of those variables contributes directly to level of identification. The context of the research (medical practitioners/ pharmaceutical prescriptions) argues in favour of the model re-specification.

Group status produces stronger identification results because of its operationalization. Group status as discussed by Ellemers (1993) and used by Ellemers, Van Knippenberg, de Vries & Wilke (1988), and Barreto & Ellemers (2001) is experimentally manipulated. Ellemers et al (1988) found that level of group identification varied according to group status (high status = higher identification), and the difference was statistically significant. In the context of status-improvement, Barreto and Ellemers (2001) find that highly identified group members will work to improve group status whether improvement appears likely or not. Low identification, however, has the effect that members when anonymous will work for individual goals, but group goals when accountable (Barreto & Ellemers, 2001). Each of these variables, prestige, group status, and construed external image, rely on the motivation of the individual for self-enhancement. Association with a high-status group, or a group perceived by peers as having a positive image, encourages the individual in self-esteem goals (Hogg, 2000; Hogg & Mullin, 1999; Tajfel, 1982).

Prestige is not the only antecedent to acquire alternative names during research. Proximity, distinctiveness and similarity all appear as experimentally manipulated conditions, or as proxy measures. Cornwell and Coote, hypothesise, in the context of non-profit organisation sponsorship, “The number of events (races or similar events) in which the participant participates annually will be negatively related to their identification with the organization” (2005, p270). This hypothesis may be seen as a measure of domain involvement, and was not confirmed. This result may suggest that identification is developed for specific objects rather than in a category-based way. It may also suggest that respondents participating in numerous charity- sponsored athletic events do so for sporting/athletic reasons primarily, with altruistic motivations being secondary. Gwinner and Swanson (2003) also test a measure of domain involvement (generally; how much do you like football?) as an antecedent of identification (generally: how much to you identify with _____ team?). This research found that domain involvement contributed significantly to identification. Each of these domain-based measures suggests the importance of similarity as an antecedent of identification.

Gwinner and Swanson (2003) also use a measure of ‘association’ as an antecedent of identification. This variable provides a measure of relationship strength similar to those found in social network research (see: Granovetter, 1973; Gilbert & Karahalios, 2009; Kossinets & Watts, 2006). The number of ties respondents’ have with the university football team or university, i.e., “…student, alumni, employee…,” (Gwinner & Swanson, 2003, p282) contributes significantly to team identification.

A final antecedent recognised here is ‘categorisation’. According to the minimal group paradigm, allocation to a group (knowing than one is a group member) is sufficient to produce intergroup bias (Brewer, 1979; Ellemers et al, 1999; Tajfel, 1974). It is also sufficient to influence levels of identification (Ellemers et al, 1988; Grieve & Hogg, 1999; Hogg & Abrams, 1999; Pinter & Greenwald, 2011). It is on the basis that categorisation produces identification consequences that Ashforth and Mael (1989) argue cognitive awareness of group membership is the only necessary dimension to assert identification, although others may be sufficient.

100

A number of minimal group methods are used, including random allocation to groups (Ellemers et al, 1988; Grieve & Hogg, 1999; Vaughan, Tajfel & Williams, 1981), allocation based on false test scores (Billig & Tajfel, 1973; Ellemers et al, 1999; Tajfel & Billig, 1974), and others (see: Pinter & Greenwald, 2011 for a review). Each of these methods has the effect of influencing the allocation of resources, and levels of in-group identification. Pinter and Greenwald (2011) in two studies, test the efficacy of group allocation methods for producing both implicit and explicit attitudes and group identification. These studies reveal that all minimal group allocation procedures produce attitude and identification effects, however, the strength of these results varies (i.e., false test scores produce strong explicit but are not strong implicit scores; memorisation of member names produces strong implicit results, and average explicit scores). This methodology establishes that categorisation is the only necessary antecedent of behavioural consequences of group membership and identification.

Consequences

Social identity theory begins with the premise that categorisation is responsible of intergroup bias (Tajfel, 1974, 1982). Perhaps because the minimal group paradigm is so well established it is not surprising to find that “Research into the antecedents… have been fewer in number, have occasionally been difficult to interpret, and have sometimes been atheoretical in nature…,” (Dimmock & Gucciardi, 2005, p285). Research on identification consequences, however, is prolific. This review recognises three categories: individual psychological responses; behavioural responses; and, intergroup perceptual responses. These categories do not constitute a typology. Arguably, any particular consequence, or publication, might be allocated to more than one category.

101

Table 2.6: Consequences of Identification Category Characteristics Examples Individual Consequences experienced at the individual Effort, motivation, psychological level, and not in conjunction with group moods and responses members or in group-interaction situations emotions Behavioural Consequences about which inferences must Resource responses be made in order to develop an explanation. allocation, Outcomes that can be ascertained by absenteeism, and observation, or self-report measures. employee turnover Group Perceptual and evaluative results of intra- or Attitudes, perceptual inter-group stimuli attributed to group-based stereotype use, and responses processes. attributions.

Individual psychological responses

This category of consequences establishes the effects of group-related activities or stimuli on the individual. The majority of this work in the sporting/team context has been conducted by Wann and co-authors; Wann (2006) provides an overview of this work. Effects of team identification on the individual include impacts on sense of belonging, levels of loneliness, depression, negative emotions, openness, conscientiousness, extraversion and fatigue, among others (Wann, 2006).

Research that studies individual psychological responses establishes the effect of level of identification on moods and emotions. Specifically, Wann and Branscombe (1992) find that higher levels of identification produce more positive moods when the individual is exposed to positive information about their team (news reports of wins), and more negative moods when exposed to information about losses or team criticism. Wann, Dimmock and Grove (2003) find that identification with a local sporting team is associated with lower levels of loneliness and higher collective self- esteem, relative to identification with a distant sporting team. More evidence of the effects of identification is provided by Grieve, Shoenfelt, Wann & Zapalac (2009), whose study of a sport competition cancellation found that highly identified individuals experienced stronger emotional responses than less identified individuals.

Research conducted in organisational settings has found that higher levels of identification contribute to the individual’s willingness to work. Van Knippenberg and Van Schie (2000) find that employees who identify with their work-group are

102 more likely to experience job satisfaction, motivation and are also more likely to invest more effort toward their job. Work-group identification, in this research, is compared with organizational identification, which argues that proximity and familiarity at the group-level has stronger effects than more abstract levels of identification (Van Knippenberg & Van Schie, 2000). Van Knippenberg and Sleebos (2006), in survey research, compare the effects of commitment and identification on individual psychological responses and behavioural intentions. This research reveals that when controlling for identification, commitment contributes to job satisfaction and motivation (Van Knippenberg & Sleebos, 2006).

A final example in this category is the research conducted by Doosje, Branscombe, Spears and Manstead (2006) on group-based guilt. The authors find that when presented with stimuli detailing events of national significance (e.g., Dutch colonisation of Indonesia), high national identification results in self-esteem protecting responses (Doosje et al, 2006). Specifically, individuals with lower levels of national identification have higher levels of guilt for negatively presented events, where high identification individuals present less guilt (Doosje et al, 2006). Higher identification is also associated with greater concerns about information credibility, and lesser likelihood of judging financial compensation an appropriate remedy (Doosje et al, 2006).

The research conducted by Doosje et al (2006) is an example of work that might be categorised in more than one way. Guilt might generally be understood as an individual-level construct. However, the activity judged (colonisation) is enacted at the group-level, and results are presented according to group identification.

Behavioural Responses

The first of the behavioural consequences of identification is resource allocation. This form of in-group favouritism and out-group discrimination is the favoured dependent measure used by Tajfel (see: Billig and Tajfel 1973; Tajfel & Billig, 1974; Vaughan, Tajfel & Williams, 1981). Used in research, respondents allocate ‘money’ to in- or out-group members (Billig & Tajfel, 1973; Tajfel & Billig, 1974). Support

103 for the notion of in-group favouritism in resource allocation is strong. The resource allocation method has also been used by Grieve and Hogg (1999), Turner, Sachdev & Hogg (1983), Hogg and Turner (1985); in each case favouritism of in-group members is found.

In research that does not use Tajfel’s choice matrices, results for in-group favouritism through resource allocation are mixed. Mael and Ashforth (1992) find that higher levels of identification are associated with contributions (donations) to the organisation and other measures of organisation support (e.g., event participation, recommending the organisation to others). In other tests of resource allocations: Ahearne, Bhattacharya and Gruen (2005) and Kuenzel and Halliday (2008) find that identification contributes significantly to word-of-mouth and to purchase intention; Theodorakis, Koustelios, Robinson and Barlas (2009) confirm the role of identification in repurchase in a services marketing context; and, Gwinner and Swanson (2003) find significant support for sponsor patronage. In contrast, the research of Bhattacharya, Rao and Glynn (1995) supports the notion that identification is correlated with repeat visits in the museum context, but does not find a correlation among identification and donations.

In organisational research, patronage and repurchase objectives are studied as employee turnover and absenteeism. In this context Abrams, Ando and Hinkle (1998) establish cross-culturally, that higher levels of identification are responsible for low turnover intention among employees. They further demonstrate that identification effects are distinct from social norms, conceptually and empirically (Abrams, Ando & Hinkle, 1998). Mael & Ashforth (1995), in a longitudinal attrition study of US Army recruits, find that organisational identification explains attrition and retention in combination with bio-data factors. Van Knippenberg and Van Schie (2000) also argue identification explains employee retention; in particular, work- group identification is more important than organisational identification. These results remain contentious despite numerous statements in the literature supporting the negative association between identification and turnover intention (see: Ahearne, Bhattacharya & Gruen, 2005; Ashforth & Mael, 1989; Bhattacharya, Rao & Glynn, 1995; Dutton, Dukerich & Harquail, 1994; Mael & Ashforth, 1995; Van Knippenberg & Van Schie, 2000; Van Knippenberg & Sleebos, 2006).

104

The research that challenges the relationship between identification and turnover is provided by Foreman and Whetten (2002), and Van Knippenberg and Sleebos (2006). Foreman and Whetten (2002) find that identification does not significantly contribute to ‘continuance commitment’, instead retention is better explained by “…member’s social and economic dependencies” (2002, p627). Identification, does however contribute significantly to affective commitment (Foreman and Whetten, 2002). Van Knippenberg and Sleebos (2006) also find that commitment, controlling for identification, better explains absenteeism and turnover intent. The issue that these studies raise is one of conceptual clarity. Mael and Tetrick (1992) use confirmatory factor analyses to determine that commitment is empirically distinct from identification. Some argue that identification is an antecedent of commitment and loyalty (Ashforth and Mael, 1989). Others argue that commitment is a component of identification (Ellemers et al, 1999). Overall, research on turnover intention suggests that further research is required to fully articulate the relationship of commitment and identification and the effects of these constructs on employee retention and turnover intention.

A final behavioural consequence of identification is ‘BIRGing’ (Basking in Reflected Glory: BIRG). BIRGing has been discussed in an earlier section (BIRGing and CORFing) and will not be re-addressed here. In summary: to BIRG is to demonstrate one’s affiliation, pride, or identification with an association (e.g., sporting team, nationality, political group, activist group, etc.). This behaviour is principally measured according to individuals’ adoption of merchandise, clothing or other paraphernalia that is branded or labelled (see: Cialdini et al, 1976; Cialdini & de Nicholas, 1989; Hirt et al, 1992; Hunt, Bristol & Bashaw, 1999; Wann & Branscombe, 1990).

Group Perceptual Responses

The group perceptual responses category is comprised of a diverse range of variables and measurement techniques. Results of research are nuanced, and often dependent upon research design. This summary provides a general overview, but does not

105 provide an in-depth review of all experimental conditions and all results. Three (very broad) types of group perceptual responses are recognised here: attributions, similarity effects (including: stereotypes, prototypicality, polarisation, consensus, homogeneity), and attitudes (self-, and group-attitudes, or evaluations).

Attribution, as a dependent measure, establishes the extent to which group members determine the responsibility of group outcomes according to internal or external means (Fielding, Hogg & Annandale, 2006; Hewstone, Jaspars & Lalljee, 1982; Turner, Hogg, Turner & Smith, 1984; Wann, 2006). That is, whether success is attributed to group effort or skill, and failure is attributed to luck, a stronger competitor, or environmental factors (Eccleston & Major, 2006; Wann, 2006; Wann & Branscombe, 1992; Wann & Dolan, 1994).

In research conducted on attributions related to success and failure, identification influences both the type of attribution (responsibility/denial) and the number of attributions. Wann and Dolan (1994), studying sport spectators, find that highly identified fans make more attributions about game outcomes than less identified spectators. Furthermore, when their team loses, high identifiers are more likely to make more external attributions (i.e., the umpires are biased), and fewer internal attributions (i.e., we could have played better). In other sport-based research, Sherman et al (2007) studied athletes (study 1) and fans (study 2) to determine whether attributions could be mitigated. The authors found that athletes and highly identified fans showed greater bias in their attributions; they also showed that prompting respondents to consider group responsibility decreased overall levels of bias (Sherman et al, 2007).

Eccleston and Major (2006) studied attributions to ethnic identity among an ethnic minority sample using ambiguous information stimuli. The information stimuli presented diverse situations (e.g., speeding fine, job interview, dating) to determine whether failure or punishment in these settings would be attributed to ethnicity (Eccleston & Major, 2006). The authors find a positive correlation among discriminatory attributions and group identification, but little impact on member self- esteem (Eccleston & Major, 2006).

106

In an academic context, Hewstone, Jaspars and Lalljee (1982) studied public and comprehensive schoolboys’ evaluations of group ‘A’ levels results. The most common, significant, predictor of attributions was outcome (success/failure); however stereotypes also explained some differences. For example: Public schoolboys differentiate themselves from the CS [comprehensive school] boys by means of ability and effort attributions. They ascribe their own group’s failure less to (lack of) ability and, especially, more to (lack of) effort, than they do that of the outgroup… Hewstone, Jaspars, & Lalljee, 1982, p256 The differences in attributions comply with self- and out-group stereotypes (Hewstone, Jaspars & Lalljee, 1982).

Stereotypes and prototype-based depersonalisation are both consequences of categorisation (Hains, Hogg & Duck, 1997; Hogg & Turner, 1987). Although stereotypes and prototypes are consequences of identification, alone they are qualitative measures. Their true power is in explaining further consequences, such as social attraction, liking, and self-evaluative measures.

In the research conducted by Hewstone, Jaspars and Lalljee (1982), stereotypes were developed through content analyses of schoolboys’ essays on the topic of the similarities and differences among students at public and comprehensive schools. There was substantial agreement across groups as to the content of stereotypes (Hewstone, Jaspars & Lalljee, 1982).

Hogg and Turner (1987) studied the effects of gender stereotypes on in-group member self-perceptions. This research found that respondents “…tended to self- stereotype more on positive than neutral than negative items” (Hogg & Turner, 1987, p333), which supports the role of stereotypes in maintaining self-esteem. The research also found that subjects were more likely to self-stereotype in single-sex treatment groups than in mixed groups (Hogg & Turner, 1987), which argues a level of in-group conformity. Haslam et al (1999) support the results of Hogg and Turner (1987) in research on stereotype consensus. They find stereotypes are used more often by subjects in group conditions than individual conditions, arguing that group contexts trigger stereotype use and drive consensus (Haslam et al, 1999); this result

107 is particularly strong when group identification salience is manipulated.

In related literature, prototypes are; “…fuzzy sets that capture the context-dependent features of group membership, often in the form of representations of exemplary members…,” (Hogg & Terry, 2000, p123). As self-categorisation involves the simplification of social relations, it enables the individual to ‘depersonalise perceptions’ (Hogg, 2009; Hogg & Terry, 2000). Some of the effects of prototype use include: social attraction to in-group members (Hogg & Hains, 1996); liking of prototypical individuals (Hogg, CooperShaw & Holzworth, 1993); and, studying both stereotypical and prototypical leadership, Hains, Hogg and Duck (1997) find that respondents expect leaders with those traits to be more effective than leaders without them.

Another consequence of identification is the homogeneity bias. The homogeneity bias argues that due to the effects of categorisation, and consistent with depersonalisation, individuals assume similarity among group members – or, homogeneity of characteristics (Brewer, 1993; Hutchison, et al, 2006; Nadler, Harpaz-Gorodeisky & Ben-David, 2009). The homogeneity effect is nuanced, empirically and conceptually.

Brewer (1993) asserts that in-group members are more likely to perceive in-group members as relatively heterogeneous while viewing out-group members as being homogeneous. Nadler, Harpaz-Gorodeisky and Ben-David, (2009) find that perceived in-group homogeneity is higher among highly identified subjects relative to low identification subjects. Whereas Hutchison et al (2006) find that heterogeneity perception varies according to experimental conditions. Specifically, when group variability (heterogeneous/ homogeneous) interacts with identity threat (presence of strong alternative group), the result will be that highly identified subjects will confirm the manipulated group variability perception (group as heterogeneous if heterogeneous prior to threat and vice versa). The result is, dependent upon experimental conditions, that in-groups may perceive themselves as homogeneous or heterogeneous.

The final similarity consequence of identification is polarisation. Polarisation is the

108 effect of ‘group think’. Turner, Wetherell and Hogg explain: …individuals’ attitudes and opinions tend to shift, following group discussion, in the direction already favoured by the group so that the post- discussion consensus is more extreme than the mean of the individual pre-test responses… Turner, Wetherell & Hogg, 1989, p135. This effect recognises the motivation of the individual to comply, or to be seen to comply, with in-group attitudes or norms. The polarisation effect is established empirically by Turner, Wetherell and Hogg (1989), Abrams, Wetherell, Cochrane, Hogg and Turner (1990), and Hogg, Turner and Davidson (1990). The polarisation effect disappears in the dark, or when the individual is given the opportunity to provide judgments anonymously (Abrams et al, 1990).

The last category of perceptual measures is that of attitudes resulting from identification. Researchers have studied how group identification influences the evaluations of groups, (Dimmock, Grove & Eklund, 2005; Hogg, Turner and Smith, 1984; Hogg & Turner, 1987; Jackson, 2002); they have also studied how identification influences self-attitudes (see: Fielding, Hogg & Annandale, 2006; Gwinner & Swanson, 2003; Turner, Hogg, Turner & Smith, 1984). Higher levels of group identification consistently result in more positive evaluations of the in-group, and the self (Fielding, Hogg & Annandale, 2006; Turner, Hogg, Turner & Smith, 1984; Pinter & Greenwald, 2011). Further, group identification can produce third- party effects. Gwinner and Swanson (2003) find that identification with a sporting team contributes to positive attitudes towards the teams’ sponsor, satisfaction with the sponsor, and sponsor recognition.

Conclusions

The antecedents and consequences of identification are well documented. Antecedents include group distinctiveness, prestige, similarity, proximity and categorisation. Consequences of identification are myriad, nuanced and contingent upon experimental conditions. They are here categorised as individual psychological responses, behavioural measures, and evaluative measures. Overall; categorisation

109 remains the sole required condition for identification to be recognised. Predictable consequences of identification are that the behaviours and judgements of highly identified group members are likely to be in-group favouring in terms of resource allocation, and more positive in evaluations of the group.

Types of Groups

Social identity theory recognises that an individual has multiple interacting identities (Brewer, 2001; Hogg, Terry & White, 1995; Turner, Oakes, Haslam & McGarty, 1994). For example, an individual may regard themselves as a family member, an employee, a sporting team member, a volunteer, regionally identified, and many more. These identity orientations are personally meaningful, temporally based, or activated according to contextual salience. According to Tajfel, “...the term ‘group’ denotes a cognitive entity that is meaningful to the subject at a particular point in time...,” (Tajfel, 1974, p69). By this reasoning, social identities are developed according to the individual’s recognition of social divisions and their self-allocation to a salient entity.

The variations in identity salience influences research in several ways, two of these are contribution to the research context, and consideration of group type. Social identity studies have tried to explain social interactions and identification in work group or employment situations (Brown & Williams, 1984; Brown, Condor, Mathews, Wade & Williams, 1986; Mael & Tetrick, 1992); secondary and tertiary education environments (Cassidy & Trew, 2001; Mael & Ashforth, 1992); sporting or social contexts (Dimmock, Grove & Eklund, 2005; Gwinner & Swanson, 2003; Heere & James, 2007; Hirt, Zillman, Erickson & Kennedy, 1992), and in relation to ascribed characteristics (Ashmore, Deaux & McLaughlin-Volpe, 2004; Phinney, 1992).

Researchers concerned with social identities recognise numerous types of groups. Groups have been referred to as assigned, achieved, ascribed, ad hoc, natural, face to face, holographic, psychological, voluntary, real, or majority/minority groups (Ellemers et al, 1999; Hinkle et al, 1989; Jackson, 2002; Luhtanen & Crocker, 1992;

110

Mael & Ashforth, 1992; Mael & Tetrick, 1992; Phinney, 1992). Many of these are used interchangeably, or in ways that suggest they are not mutually exclusive.

The minimal group paradigm suggests that when an individual is assigned to a group, whether on the basis of real or fictitious characteristics or unexplained allocation, the individual accepts the label of ‘group member’ to some extent (Ellemers et al, 1999). On the basis of even this weak categorisation, individuals have been found to “behave in terms of their group membership” (Ellemers et al, 1999, p372), which is to demonstrate in-group favouritism or out-group differentiation. Minimal, or assigned, group membership has not been used in sport/team identification research.

Jackson uses the term ‘real groups’ (2002, p19) to distinguish groups that are objectively recognised. Real groups, according to Jackson, are recognised according to ascribed or achieved characteristics of the individual or face-to-face interaction among members (Jackson, 2002). This characterisation of group types is coherent, if not mutually exclusive. Although Jackson (2002) does not identify the alternative to ‘real groups’, they might readily be contrasted with psychological groups (Ashforth & Mael, 1989; Turner, 1984). The psychological group is defined perceptually as people who “define themselves in terms of the same social category” (Turner, 1984, p530). According to Ashforth and Mael, the psychological group member “…does not need to interact with or like other members, or be liked or accepted by them…,” (Ashforth & Mael, 1989, p24). In this sense, a psychological group is an abstract group of personal definition, rather than one that may be objectively recognised (i.e., employee groups, or health club membership groups).

The ascribed, achieved and face-to-face groups recognised by Jackson (2002) are categories also used by other authors. Achieved groups, according to Ellemers et al (1999), Jackson (2002) and Phiney (1992), are those characterised by the self- assignment of the individual to the group. This self-assignment does not reflect goal- directed action and achievement; instead, it is used as a direct contrast to the assigned group. That is, achieved groups are those that the individual believes they are a member of, not a group that they have been allocated to, and not necessarily one for which they have strived to gain membership. For Jackson, the achieved group label includes “fans of sports teams or musical groups” and “political interests” (Jackson,

111

2002, p20). Ashmore et al (2004) suggest that occupation and political party memberships are achieved groups. Finally, in relation to team identification, Dimmock et al (2005) suggests that team identification is a voluntary group membership, which may also be considered a form of achieved group.

Membership of an ascribed group is neither readily assigned, nor chosen. Jackson (2002) suggests that ethnicity and gender are ascribed characteristics. Gender and ethnicity are also recognised as natural groups (Ellemers et al., 1999, p375). However, the natural group label is problematic. Natural groups for Ellemers et al are those “generally more likely to involve self-selected group memberships” (1999, p375). This makes the natural group the equivalent of an achieved group. Ellemers et al (1999) use problem-solving style (deductive or inductive) as the self-selected and assigned groups in an experimental design; participants were ‘assessed’and assigned to a problem-solving style group or they asserted (self-selected) their own problem- solving style. The group allocation method used by Ellemers et al (1999) establishes the importance of understanding achieved groups as self-selected groups, as opposed to measures of goal-directed achievement.

In summary, four principal forms of group membership are recognised here. The assigned (minimal) group contains people who have been allocated to a particular treatment by researchers. Ascribed groups are those determined by characteristics of individuals that are generally not chosen by the individual (Jackson, 2002) and which may be observable or asserted by the individual or others. Achieved group memberships are self-selected (Ellemers et al, 1999), without necessarily implying goal-achievement. Finally, psychological groups (Ashforth & Mael, 1989; Turner, 1984) explain those groups that do not necessarily have a formal membership or face-to-face interaction, but are groups of sentiment. This review of groups fails a test of mutual exclusivity but highlights key differences in group membership. According to these characteristics, identification with a sporting team is most likely to evidence characteristics of achieved or psychological groups.

112

Defining Identification & Developing Dimensions

Identification, as previously noted, is often explained with reference to the definition provided by Tajfel (1978). This definition says that an individual’s social identification is: …that part of an individual’s self-concept which derives from his knowledge of his membership of a social group (or groups) together with the value and emotional significance attached to that membership. Tajfel, 1978, p63 This definition is cited by many researchers (see: Brown et al, 1986; Dimmock et al, 2005; Ellemers et al, 1999; Gurin & Townsend, 1986; Heere & James, 2007; Henry et al, 1999; Hinkle et al, 1989; Jackson, 2002; Karasawa, 1991; Luhtanen & Crocker, 1992; Phinney, 1992). It also contributes to scale development efforts in guiding researchers views about the dimensionality of the identification construct. Seven of the scales reviewed have used Tajfel’s (1978) definition to develop the conceptual structure of their measure of identification.

Tajfel’s definition is used to justify a tripartite structure of identification. This tripartite structure includes: a) knowledge of group membership (cognitive component), b) value of membership (evaluative component), and c) emotions related to group membership (affect). From Table 2.7, the popularity of the cognitive/affective/evaluative structure of identification is clear.

The cognitive dimension of identification reflects the self-categorisation principle of social identity theory (Ellemers et al, 1999; Henry et al, 1999; Jackson, 2002). This cognitive awareness of group membership is the minimum condition for identification (see: minimal group paradigm, for example: Tajfel, 1981). It is with regard for the minimal group paradigm that some authors regard the cognitive element of identification as the only required condition (Ashforth & Mael, 1989). However, it is common for items measuring cognitive identification to address subjects beyond knowledge of group membership. Deaux (1996), for example, argues that awareness of group membership includes self-labelling, and knowledge and beliefs associated with the group.

113

The ‘emotional significance’ component of group membership is usually referred to as the ‘affective’ dimension; it has also been rendered as ‘regard’ (Sellers et al., 1997), ‘interdependence’ and ‘interconnection of self’ (Heere & James, 2007) and ‘commitment’ (Ellemers et al., 1999). This dimension accounts for the feelings of the individual toward the group or group members. Henry et al (1999) regard the ‘emotion’ dimension as one that explains group cohesion by processes of interpersonal attraction. For Jackson, affect explains the satisfaction of the individual with their group membership and as well as the “sense of commitment to the group or belongingness” (2002, p16). Commitment is also used by Ellemers et al (1999) to explain the “emotional involvement” (1999, p372) of the individual with the group.

The final construct drawn from Tajfel’s (1978) definition is the evaluative component, which, according to Tajfel (1978) explains the ‘value of group membership’. As with any theoretical reference to evaluation, social identification ‘evaluation’ implies an attitudinal measure. Ellemers et al., following Tajfel (1981) define the evaluative dimension, “…a positive or negative value connotation attached to this group membership…” (Ellemers et al, 1999, p372). However, evaluation for social identification is also closely associated with the notion of self-esteem in the sense that group membership should be self-esteem enhancing, or that individuals generally strive for positive self-esteem (see; Brown et al., 1986; Ellemers et al., 1999).

Where researchers deviate from Tajfel’s definition in developing their structure of identification, the reasons may be found in history, theory and research context. These deviations have produced such dimensions of identification as: affirmation and denial, identity centrality, perception of shared- or common fate, behaviour and behavioural involvement (see: Driedger, 1976; Heere & James, 2007; Henry et al., 1999; Kelly, 1988; Phinney, 1992; Sellers et al., 1997).

114

Table 2.7: Identification Scale Dimensions Year Authors Context Cite* Use* Dimensions Reliability 1976 Driedger Ethnicity N N Affirmation/ Denial - 1986 Brown et al Organizational Y Y Awareness/ Evaluation/ Affect .71 1986 Gurin & Townsend Gender Y N Similarities/ Common Fate/ Centrality - 1988 Kelly Political N N Affirmation/ Denial .79 1989 Hinkle et al Group Y Y Cognitive/ Emotion/ Individual-Group Opposition .85-.94*** 1991 Karasawa Group Y Y Identification with Group (Cognitive/ Affective)/ Identification with Group Members - 1992 Luhtanen & Crocker Collective Y N Membership/ Private/ Public/ Identity .73-.85** 1992 Mael & Ashforth Organizational N N Uni-dimensional .87 1992 Mael & Tetrick Psychological group N N Shared experiences/ Shared characteristics .76 1992 Phinney Ethnicity Y N Affirmation/ Achievements/ Behaviours .81-.90*** 1993 Wann & Branscombe Sport N N Uni-dimensional .91-.93*** 1997 Sellers et al Ethnicity N N Centrality/ Private Regard/ Assimilation/ Humanist/ Minority/ Nationalist .60-.79** 1999 Ellemers et al Social Y Y Self-categorisation/ Commitment/ Group self-esteem .82 1999 Henry et al Group Y Y Cognitive/ Affective/ Behavioural .76-.89** 2002 Jackson Group Y Y Cognitive/ Affective/ Evaluative .84-.92** 2005 Dimmock et al Team Y Y Cognitive/ Affective/ Evaluative - 2007 Heere & James Team Y N Public evaluation/ Private Evaluation/ Interconnection/ Interdependence/ Behaviour/ .73-.86** Cognitive

*Cite and Use columns refer to Tajfel’s tripartite definition of social identity as including, ‘knowledge’, ‘value’, and ‘emotional significance’ elements **Range of Cronbach’s Alpha scores measure alpha’s for dimensions/sub-scales ***Range of Cronbach’s Alpha scores refer to measures from multiple studies or samples

115

The first pair of dimensions used in identification scales not driven by Tajfel’s (1978) definition is the affirmation/ denial twin. Affirmation and denial are used by Driedger (1976), Kelly (1988) and Phinney (1992). Each of these scales is developed to discuss contexts which may be associated with personal or interpersonal conflict (ethnicity and politics). The history of publication rather than research contexts argues the affirmation/denial dimensions pre-date widespread knowledge of Tajfel’s (1978) definition. The dimensions may also derive from methodological bases, such as the split-half method, especially when the items are considered, see below:

Affirmation items Denial items Feels strong bonds toward ingroup Feels restricted by ethnic identity Participates in ethnic activities Tries to hide ingroup background Considers ingroup culture rich/precious Feels inferior to others in outgroups Wishes to remember ethnic heritage Is critical of ingroup customs Sees relevance of ethnic differentiations Makes excuses for ethnic ingroup Contributes to ingroup class discussions Afraid to express feelings about ingroup Is proud to identify with ingroup Is annoyed to reveal ethnic identity Driedger, 1976, p136

Although the pairs are not mirror-opposites, in meaning they reflect the alternative sentiment on each particular issue. The affirmation/denial dimensions remain methodologically useful. Historically, however, they appear to become less popular at about the same time Tajfel’s definition gains in popularity.

An example of the coinciding of less- and more- popularity is the case of Brown et al.’s (1986) scale. Brown et al (1986) note two contributions to their scale; the first is Tajfel’s definition (1978), and the second, Driedger (1976). Specifically, Brown et al (1986) note: In this definition it is possible to pick out three facets: awareness of group membership (which contributes to self-definition); evaluation (which relates to self-esteem) and affect. And, We therefore developed a new instrument, based on an original scale of ethnic identity devised by Driedger (1976). … The items also attempt to tap the three aspects of identity referred to above. Brown et al, 1986, p275

116

Driedger’s (1976) scale is shown above. Brown et al’s (1986) scale resembles it in having both 5 positive and 5 negative items. The language for several items is also similar. Instead of ‘strong bonds’ (Driedger, 1976) there are ‘strong ties’ (Brown et al, 1986); instead of being ‘proud’ to identify (Driedger, 1976), one might be ‘glad’ to identify (Brown et al, 1986); being ‘annoyed’, ‘trying to hide’, and being ‘critical’ (see: Driedger, 1976; Brown et al, 1986) are also similar adaptations.

When reviewing Brown et al’s scale, the affirmation-denial dimensions appear a better fit to the items than the tripartite explanation, which is not justified with any conceptual development: (Awareness is tapped by items 2 and 5; evaluation by items 1, 6, 7 and 10; affect by items 3, 4, 8 and 9). Brown et al, 1986, p275 Brown et al’s (1986) scale is the first to introduce a dimensional structure of identification that is argued to be conceptual rather than probably methodological. Brown et al’s (1986) scale is also a development from Driedger’s (1976) in reducing its reliance on behavioural items.

This research takes the view (following Ashforth & Mael, 1989) that a behavioural component of identification should be regarded as either antecedent or consequence to identification. The desire to recognise behaviours associated with group identification is not uncommon. Henry et al (1999), and Heere and James (2007) each include behavioural dimensions in their identification scales. Henry et al provide that their behavioural dimension, “…focuses on the group-level construct of cooperative interdependence” (1999, p561). They also provide that “…interdependent outcomes will evoke group identification” (Henry et al.,1999, p566), and: …we broaden the behavioural source of group identification to include outcome and behavioural interdependence – the need to coordinate actions among members in pursuit of group objectives. Henry et al., 1999, p567 The overall impression of this approach to behaviours must be caution. Ashforth and Mael (1989) recognise that the conditions contributing to group formation may also be antecedents of identification. The behavioural perspective of Henry et al (1999)

117

reflects this idea to the extent that it seems antecedent (evoking) of identification.

A further cause for thought is the context of behaviour; in this case, the group is necessarily a face-to-face group which engages in cooperative behaviours and the “pursuit of group objectives” (Henry et al, 1999, p567). Although Henry et al (1999) state their scale is developed for the purpose of explaining interacting groups, it is particularly in relation to behaviour that their scale cannot be used more generally, for psychological groups. There remains, finally, the conceptual problem of group objectives; following Ashforth and Mael (1989), this research takes the perspective that for group members to have shared values and shared goals is not a necessary condition of identification.

The ‘behavioural involvement’ dimension developed by Heere and James (2007) differs from that developed by Henry et al (1999). Instead of trying to capture a sense of interdependence, which for Heere and James (2007) is an affective dimension, behavioural involvement refers directly to participation in group-based activities (Heere & James, 2007, p70). The issue of whether behaviours may be antecedents or consequences of identification remains salient. The authors also explain, “behaviours may be important for gaining access and membership in a group…” (Heere & James, 2007, p70); this suggests a goal-directed antecedent. Behavioural measures of identification overall, remain problematic. They seem likely to remain problematic in cases where researchers do not articulate contextual antecedents and consequences, or the purposes and delimitation of their research in considerable depth.

‘Common fate’ is an identification dimension used by Gurin and Townsend (1986). For these authors common fate, “…is defined as perception of commonalities in the way group members are treated in society” (1986, p140). The outcomes of such beliefs are effects on intergroup comparisons, collective discontent (where a group is negatively stereotyped), and individual commitment to the group (Gurin & Townsend, 1986, p142-43). Each of the hypothesised consequences of perceived common fate is consistent with Tajfel’s (1974) consequences of group membership. For example, where group membership fails to be self-enhancing for the individual, the individual may re-evaluate their perception of group-status, attempt to leave the group, or “engage in social action which would lead to desirable changes in the

118

situation” (Tajfel, 1974, p70). The importance of common fate to identification, is then, clear. Common fate is also quite complicated.

Tajfel, discussing common fate, notes: A group becomes a group in the sense of being perceived as having common characteristics or a common fate only because other groups are present in the environment. Tajfel, 1974, p72 ‘Common fate’, then, begins with processes of self-group-other categorisations, but it is not merely recognition of similarity among in-group members. It is a concept involved in the multiple (and related) debates on self-categorisation and depersonalisation (Hogg, 2009; Hogg & Hains, 1996; Hogg & Terry, 2000; Turner & Bourhis, 1996); in-group and out-group homogeneity and heterogeneity (Brewer, 1993; Hutchison, Jetten, Christian & Haycraft, 2006) and the social identity theory principle of ‘social comparison’ (Tajfel, 1974, 1982; Hogg & Terry, 2000).

To operationalize their measure of common fate, Gurin and Townsend (1986) use three questions. These are: ‘Do you think that what happens to women generally in this country will have something to do with what happens in your life?’ Those who answered ‘yes’ were asked how much it would affect them. All respondents were then asked: ‘Do you think that the movement for women’s rights has affected you personally?’ Gurin & Townsend, 1986, p142 This measure, broadly, suffers from a lack of conceptual development and methodological looseness. The first two items have the potential to measure common fate. The third item speaks more to a related concept, that of interdependence. The eliding of common fate and interdependence is not uncommon. Henry et al, for example, state that “the common fate literature emphasises the behavioural aspect [of identification] by pointing to the importance of interdependence…” (1999, p560). Turner and Bourhis, alternatively, argue that while common fate is “… ‘being in the same boat’… Interdependence of outcomes is a functional relationship in which one’s own outcomes depend instrumentally on the actions of the other…” (1996, p38). The common fate dimension has, at present, no valuable representative.

119

The final dimension for review is that of ‘centrality’; it has been used by Gurin and Townsend (1986) and Sellers et al (1997). According to Sellers et al, centrality “refers to the extent to which a person normatively defines him or herself with regard to race” (1997, p806). For Gurin and Townsend, centrality “…provides an anchor for evaluating, comparing and reacting to the out-group…” (Gurin & Townsend, 1986, p141). These perspectives on centrality suggest that it may be a substitute for an holistic approach to identification. It appears to include both a cognitive awareness of group membership, and, a degree of affective response to the same.

Another point for consideration is that for Sellers et al (1997) centrality is not ‘salience’, which is conceived as a distinct dimension, although the two are related. Salience is sufficiently important for Gurin and Townsend (1986) to also refer to identity salience in their discussion of centrality. The relationship of ‘salience’ to this dimension further argues centrality’s ill-fit in an identification scale. As recognised by Gurin and Townsend, “categories and group boundaries can be made salient by features of the immediate environment” (1986, p141). These points argue that salience and centrality are contextual elements, or to be experimentally manipulated.

Overall, the debate on the dimensional structure of identification seems likely to continue as having unresolved structural arguments, opportunities for further contextual specification, and areas in need of further refinement. The tripartite structure of identification driven by Tajfel’s (1978) definition has attracted a lot of support. Whether the structure of identification should be regarded as tripartite and fixed is another question.

It is worth considering, given the importance of the definition, whether Tajfel intended his definition (Tajfel, 1978) to guide research into the dimensionality of identification. Tajfel was almost certainly satisfied with his definition of identification. An earlier version, (Tajfel, 1974, p69) provides only that social identification requires ‘knowledge’ of group membership and ‘emotional significance’, without requiring the ‘valuation’ of group membership. The definition as provided here (above) is reprinted or restated in several of his own, and co-written works (see: Tajfel, 1974, 1978, 1981, 1982; Tajfel & Turner, 1986). The definition

120

has also been cited or restated in later works by co-authors working in new areas (see: Reynolds, Turner & Haslam, 2000; Turner & Bourhis, 1996; Turner & Turner, 2001). It is noteworthy, however, that neither Tajfel, nor any of his co-authors, have engaged in the dimensionality debate or scale development research.

Tajfel’s own research was dominated by minimal group experiments (see: Billig & Tajfel, 1973; Tajfel 1981; Tajfel & Billig, 1974; Vaughan, Tajfel & Williams, 1981); measures of degree of identification in such contexts would be irrelevant. He has further provided no record of intending researchers outside of the minimal group setting to be guided by the tripartite definition when developing scale measures. To argue such, one might refer to Tajfel (1982), who refers to earlier work (Tajfel, 1981) when he argues: The value and cognitive functions of social accentuation provide a basis for the understanding of the structure and direction of biases in intergroup attitudes and stereotypes. Tajfel, 1982, p22 The above does not however, refer to a scale measure of identification or the use of value and cognitive functions as dimensions. Tajfel (1981) discussed the principles of social perception and presented ‘value-based’ research assessed in relation to size/dimensions (mostly of coins), and in relation to individual differences (pre- existing prejudice or learning in the case of familiarity with coins). Among his premises is the logic: “…judgments are made in dimensions in which scaling in magnitude is simultaneously a scaling in value” (Tajfel, 1981, p70). The link to social perception, the linking of the perceived value of coins (known and unknown), and the value of facial characteristics (to those prejudiced and those not) is superb. It is not an endorsement of identification measures; nor does the use of a measure of prejudice suggest any measure of group identification unless as a dissociative attitude. Similarly, the discussion of the cognitive function yields neither mention of scale measurement, nor any means to structure an approach to cognitive social identity (see: Tajfel, 1981, pp127-141). It cannot be said that Tajfel in developing a tripartite definition of identification intended such to guide scale development.

John C.Turner was Tajfel’s research partner and co-author in several academic works (see: Tajfel & Turner, 1979, 1986, 2004). This review takes the (nominal) view that

121

where no identification scale measure can be found within Tajfel’s own works, his views may be reflected in the later works of research partners. When discussing identification dimensionality Jackson justifies a multidimensional approach on the basis that Turner (1999) argues that identification is “likely not unitary” (Jackson, 2002, p17). Turner, in fact, is far more critical of identification scale measurement than such a mild comment would indicate. The criticisms include:  Social identity theory does not postulate a direct causal relationship between identification and intergroup bias, but a relationship mediated by context-specific conditions (Turner, 1999, p20).  Social identification is ideally tested using experimental designs; most commonly it is used as an individual difference variable and subject to all the flaws of other confounding individual difference variables (Turner, 1999, p21).  Identification measures, generally, appear measures of personal identity rather than social identity, which would involve ‘we’ statements more than ‘I’ statements (Turner, 1999, p21).  Without properly understanding the groups they try to explain, identification researchers have failed to identify legitimate out-groups and appropriate dimensions for measurement (Turner, 1999, p22). These criticisms are consistent with theory, and so useful. However, practice is also worth considering.

Turner’s (1999) criticisms of existing identification scales explain his lack of engagement in the scale development debate. Turner has, however, published extensively with another prominent social identity theory researcher, Michael A Hogg (see: Abrams, Wetherell, Cochrane, Hogg & Turner, 1990; Hogg & Turner, 1985a, 1985b, 1987; Turner, Hogg, Turner & Smith, 1984; Turner, Sachdev & Hogg, 1983; Turner, Wetherell & Hogg, 1989). In the cited publications, the minimal group paradigm is responsible for group allocation, and testing occurs within experimental designs; these characteristics are consistent with Turner’s (1999) theoretical understanding of how social identity research should function. Identification scale measurement is not used.

122

Turner, with Hogg and Davidson (see: Hogg, Turner & Davidson, 1990), have used a measure of identification. This (dependent) measure used in a minimal group experimental design is a single item, context-dependent question: How far do you feel you are a similar kind of person to the people in your group in terms of the issues you are going to be discussing? Hogg, Turner & Davidson, 1990, p85 This item, with two others (motivation and involvement questions), comprise the authors’ social identity measurement. These questions are not equivalent a scale measure, and so do not warrant an item-by-item comparison with the criticisms offered by Turner (1999) of identification scale measures.

Measures used by Hogg and co-authors (without Turner) do warrant consideration. The identification measures used include an 8-item scale (Fielding, Hogg & Annandale, 2006), a 9-item scale (Hogg & Hains, 1996), a ten-item scale (Grieve & Hogg, 1999), and an 11-item scale (Hains, Hogg & Duck, 1997). The earliest scale referred to is provided in Hogg and Hardie (1991) with others appearing in Hogg and Hardie (1992), and Hogg, CooperShaw and Holzworth (1993). Hogg and Hains (1996) provide the first explanation (and list) of items in providing that their scale has drawn items from Brown et al (1986) and others (see: Hogg, CooperShaw & Holzworth, 1993; Hogg & Hardie, 1991).

The authors have not provided any dimensional or conceptual discussion; indeed, the purposes of their research have not been scale development. Perhaps because of this, reporting scale-related information has not been a priority. Where factor analyses are discussed, the authors have favoured a uni-dimensional factor structure, for example: Factor analysis (with orthogonal varimax rotation) of these items produced one major factor (eigenvalue of 4.87, 44.2% variance accounted for) that stood out clearly from the rest of the field (eigenvalues of 1.14, 10.3. 0.77, etc). For this reason, we decided to specify a single-factor solution, which produced factor loadings ranging from .80 to .35 around a mean of .65. We computed a scale from the weighted average of the 11 items (weighted by factor loadings). The reliability of this scale was high (α = .87), and deletion of items did not improve reliability. Hains, Hogg & Duck, 1997, p1094

123

The Cronbach’s alpha for the 11-item scale (Hains, Hogg & Duck, 1997) is acceptable; the same alpha is reported for the 10-item scale by Grieve and Hogg (1999); which suggests that the scale is reliable across projects. In no other research, however, are reported alphas as strong, in some they are not reported at all (see: Fielding, Hogg & Annandale, 2006; Hogg, CooperShaw & Holzworth, 1993; Hogg & Hains, 1996; Hogg & Hardie, 1991).The value of these scales, if the authors’ theory-based credentials were not established, would appear almost coincidental.

To summarise; this overview of social identification measurement and its dimensions has shown the prevailing popularity of Tajfel’s (1978) definition as guiding theoretical development of construct dimensionality and scale development. The most commonly used dimensions in identification scales are cognitive, affective and evaluative components because of Tajfel’s definition. Despite this popularity, alternative dimensions, some theoretically viable and others not, have developed. Preeminent scholars in social identity theory (Tajfel, Turner, and Hogg) have not fully participated in the dimensionality debate. Turner (1999), in particular, appears to regard the individual difference approach to identification measurement as having a tenuous theoretical basis. The failure of leading social identity researchers to develop multi-dimensional identification measures (see: Fielding, Hogg & Annandale, 2006; Hogg & Hains, 1996) despite Turner (1999) suggesting there is some basis for that approach leaves unanswered the question of identification dimensionality. Although there are numerous extant identification scales, this review recognises little consensus as to a ‘correct’ approach to measurement, and much opportunity for refinement and re-development.

Summary of Theory Literatures

Numerous sponsorship researchers have argued the problem and opportunity of the ‘how sponsorship works’ (see: Cornwell, Weeks & Roy, 2005; Dees, Bennet & Villegas, 2008; Gwinner, 1997; Javalgi et al, 1994; Lee, Sandler & Shani, 1997; Madrigal, 2000; Pope, 1998). This might otherwise be written: ‘how can sponsorship achieve its objectives given: a) generally low levels of involvement, or low level

124

information processing (Cornwell, Weeks & Roy, 2005; Lardinoit & Derbaix, 2001; Pitts & Slatterly, 2004), or b) its need for direct activation (see: Cornwell, Roy & Steinard, 2001; Cornwell, Weeks & Roy, 2005; Lardinoit & Derbaix, 2001; Lee, Sandler & Shani, 1997; Madrigal, 2000; McDaniel, 1999).

The means by which sponsorship might achieve its objectives, from this review, are identified as information processing and individual difference processes. Other theoretical explanations abound, including mere exposure, classical conditioning, congruence, paired-associate learning and others (see: Cornwell, Weeks & Roy, 2005). Many of the information processing theories used by sponsorship researchers are adopted from the psychological or social psychological disciplines and have been adapted for use in marketing and sponsorship contexts. The authors who query how sponsorship works are right to do so as theory, methods and results are often contradictory.

There is, however, an extent to which the ‘how it works’ question is disingenuous. The underlying theories and processes borrowed and adapted from psychological sciences are well-developed; the processes by which they ‘work’ are reasonably well-understood, and becoming ever more clear given neuro-imaging opportunities and other technological introductions (see, for example: Harris, Todorov & Fiske, 2005). The problem remains that sponsorship contexts do not, and often cannot conform, (if they are to speak about their intended subjects), to the purity of research methods used in psychological research. Knowing ‘how’ sponsorship works will always need to be prefaced by an extended description of the research context and method.

The contribution of the balance theory and social identity theory literatures to this research is in their explanation of psychological processes that influence information processing and judgment-making. Each of these theories contributes to an understanding of the mediated nature of sponsorship in recognising the role of third- parties in interpersonal or, intergroup judgments.

Balance theory and the social identity theory of identification each explain likely perceptual valence in situations where an individual judges another person or people.

125

There are two key issues that illustrate the differences and similarities among the theories, these are: a) the mechanism used to explain perceptual outcomes, and b) and the form of judgment consequences which are explained by the level of focus.

The mechanisms used to explain perceptual balance are cognitive consistency (Gawronski & Strack, 2004; Gawronski, Walther & Blank, 2005), and memory models (Hummert, Crockett & Kemper, 1990; Picek, Sherman & Shiffrin, 1975). Testing balance propositions has involved the use of memory-for-judgment tests (Hummert, Crockett & Kemper, 1990), and implicit and explicit attitude measures (Gawronski & Strack, 2004; Gawronski, Walther & Blank, 2005). The research that has investigated the mechanisms underlying balance provides evidence that individuals recall or infer the valence of relationships existing among entities with a preference for consistency.

The principles of social identity include categorization, self-definition, social comparison and psychological distinctiveness (Tajfel, 1974, 1978, 1982). The effects of personal attraction and liking, self-esteem and other more fundamental concepts on inter- and intragroup perceptions have also been assessed as processes underlying social comparison or psychological distinctiveness (see: Hogg & Hains, 1996; Hogg & Turner, 1985a; Turner, Sachdev & Hogg, 1983). However, the most consistent experimentally manipulated condition remains simple categorization using the minimal group paradigm (see: Billig & Tajfel, 1973; Ellemers et al, 1988; Grieve & Hogg, 1999; Hogg & Abrams, 1999; Tajfel & Billig, 1974). Knowledge of group membership (self-categorisation) remains for Ashforth & Mael (1989), the only necessary condition required to assess identification.

The mechanism explaining identification effects, then, are less clear than those provided by balance theory research. Identification has, however, produced consistent research results over time. The opportunity to further explore and explain the precise mechanisms underlying social identification effects exists as an opportunity for further research. The opportunity to regard self-categorisation as a form of self-referencing, self-congruity (see: Sirgy et al, 2008), or consistency, also exists.

126

The second key issue differentiating the theories is the form of construct consequences and the level of focus. The level of focus or, level of interpersonal interaction, the theories explain is substantially different. The central actor in balance theory explanations is ‘p’, the perceiver. ‘P’ reviews or judges the relationships among individuals or entities in a systematic (even, syllogistic) way. When ‘p’ likes ‘A’, and ‘A’ likes ‘B’, then a balanced perceptual system for ‘p’ argues that ‘p’ will like ‘B’. Alternatively, when ‘p’ likes ‘A’, but ‘A’ does not like ‘B’, perceptual balance for ‘p’ can be achieved if p’s relationship with ‘B’ is dislike. These balance theory networks are ego networks that explain a single mind, and relationships among individuals in that mind.

Social identity explanations of interpersonal judgments are governed by a macro- level perspective which begins with the group. When judgments are explained using social identity theory, the principal questions are about group membership: which group does the individual (perceiver) identify with; which group is the perceived thought to be a member? Once group membership is known, perceptual valence and other consequences can be anticipated.

Together, the level of focus and underlying mechanisms determine the form of consequences expected. Balance theory, concerned with the individual, argues its consequences as effects on the individual. Heider (1958) suggested that imbalanced systems of judgments would result in cognitive stress for the individual. That stress created motivation for the individual to re-evaluate their judgments to create a balanced system (Heider, 1958). For balance researchers investigating the role of memory, consequences of imbalance include forgetting, and false inferences of (Picek, Sherman & Shiffrin, 1975). Memory for balanced networks’ relationships is evidenced by faster and more accurate recall of relationship valences (Hummert, Crockett & Kemper, 1990; Picek, Sherman & Shiffrin, 1975).

The consequences of identification include effects on the individual such as mood, motivation and effort (see: Wann & Branscombe, 1992; Grieve et al, 2009; Van Knippenberg & Van Schie, 2000). Consequences also include effects on attitudes, attributions, stereotype use (see: Fielding, Hogg & Annandale, 2006; Turner & Hogg, 1987; Turner et al, 1984), and behavioural consequences such as resource

127

allocation, absenteeism and turnover (see: Billig & Tajfel, 1973; Grieve & Hogg, 1999; Hogg & Turner, 1985; Vaughan, Tajfel & Williams, 1981). All of these are achieved with consideration of the group.

In effect, the difference among the consequences found in balance theory research and research conducted to study identification is disciplinary. Social identity theory conceptualisations of identification argue social-psychological consequences, rather than the more purely psychological effects explained by balance theory.

These differences, in mechanism and level of focus, argue the different abilities of balance and identification to explain sponsorship effects. Specifically, they suggest that balance theory can explain the desires and abilities of the individual related to processing information, where ‘balanced’ information is easier to recall and more likely to be generated in inferences. Social identification, alternatively, explains the individual as a group member whose judgments are governed by their relationships.

In the research for this thesis, balance and identification exist as sources of judgment bias or error. Balance theory argues that the individual receiving news of athlete’s off-field behaviours will balance their perceptions of related entities (sponsors, or the athlete’s team) with evaluations equivalent to the valence of the athlete’s off-field behavior, or feel stress to do such. Identification alternatively, argues that individuals, according to their self-defined group membership as a team supporter (or not), will have biased perceptions of the severity or positivity of off-field behaviours dependent upon knowledge of the actor’s group membership.

As both theories bias evaluations, a key methodological concern for this research is in identifying the source of bias and its contribution to variable variances. Identification, as shown in the review of dimensions, is frequently measured as an individual difference variable (Turner, 1999). Balance, alternatively, is the product of analysis of judgments and memory tests. This final difference between the theories argues the opportunity to identify and measure (and exclude) variance attributable to identification, and to discuss balance post hoc of measurement. These opportunities are discussed further in the following chapter which provides the conceptual framework for this research.

128

Chapter 3: Conceptual Framework

This research asks whether off-field behaviours of athletes have the potential to affect the achievement of sports’ sponsorship objectives. The research principally addresses negative off-field behaviours. The problem arises because of the many incidents of sponsorship contract cancellation following the reporting of athlete off- field behaviours (see: Byrnes & Phelps, 2009; Maiden, 2005; Zelkovich, 2010).

In this chapter, the conceptual framework for the research is presented. The research is developed across an initial content analysis (Study 1) and two pretests; it culminates in a final experiment (Study 2). The progress of the thesis is tabled in Chapter 1 (Introduction); it is presented in this chapter diagrammed as Figure 3.7.

There are four research questions addressed in this thesis. Each of these questions is answered by a study or pretest. The questions for this thesis are:

Table 3.8: Research Questions Research Questions Study/ Pretest RQ1. What types of athlete off-field behaviours are reported? Study 1: Content Analysis RQ2. Removed from news-media and sports contexts, how do Pretest: Valence of individuals evaluate the off-field behaviours reported? Behaviours RQ3. If group (team) identification has the power to influence Pretest: Identification consumer evaluations, how should team identification be measured scale for this research? RQ4. Are consumers’ evaluations of an athlete, the athlete’s team Study 2: Experiment and the team’s sponsor influenced by news-media reports detailing the athlete’s off-field behaviours? a. Does team identification moderate evaluations? b. Are evaluations balanced?

The research is progressive in the sense that the design of Study 2 is dependent upon the findings of earlier research. Figure 3.7 diagrams the research processes. This chapter provides an introduction to each phase of the research; a summary of the literature which has informed the research; and, research considerations which have implications to research design, analysis, and the generalizability of results (principally, sampling and respondent issues).

129

Figure 3.7: The research process

Problem field: Do negatives in the sponsorship environment affect the achievement of sponsorship objectives?

Limitation of scope: Effects of athlete’s Summary of Sponsorship Literature off-field behaviours on sponsorship Key objectives: Theories explaining the objectives Corporate Image mediation of Brand Attitude sponsorship effects: (evaluations) Identification Balance theory RQ1. What off-field behaviours of athletes are reported? Study 1: Content Analysis RQ3. If group (team) identification has the power to influence consumer evaluations, how should team identification be RQ2. Removed from measured for this research? news-media and sports Pretest: Identification scale contexts, how do individuals evaluate the off-field behaviours reported? Pretest: Valence of RQ4. Are consumer evaluations behaviours of an athlete, the athlete’s team and the team’s sponsor influenced by news-media reports detailing the athlete’s off-field behaviours? Study 2: Experiment

IdentificationRQ4a. Does team identification moderate evaluations?

RQ4b. Are evaluations balanced?

130

Conclusions from the Introduction & Literature Review

Chapter 1 introduced the problem field as negatives within the sponsorship environment. Several areas of risk to sponsors’ objectives were identified as the role of the news-media in agenda-setting and the use of news-values; consumer behaviours such as BIRGing and CORFing; and, the off-field behaviours of athletes. Many other risks to sponsorship contracts and sponsored event success exist (i.e., weather, insufficient event funding, etc), however, each of the potential negatives identified in Chapter 1 exist as knowable phenomena that are outside of the direct control of sponsors and event organisers.

BIRGing and CORFing (see: Cialdini et al, 1976; Snyder, Lassegard & Ford, 1986) and the role and effects of the news-media (see: Cohen, 1963; Fortunato, 2008; Galtung & Ruge, 1965) are areas of considerable research. Off-field behaviours of athletes, as scandals, are areas of emerging interest for sponsorship researchers. The majority of emerging research, however, aims to identify manager’s perspectives of such problems (see: Hughes & Shank, 2005, 2008; Kahuni, Rowley & Binsardi, 2009; Wilson, Stavros & Westberg, 2008) without addressing the consumer perspective. This oversight of consumer evaluations of off-field behaviours exists as an opportunity for the current research.

The literature review begins by identifying the sponsorship environment as one that is, primarily, mediated. The sponsorship contract exists between a sponsor and sponsee; effects of sponsorship, however, are often sought for consumers or audiences of the sponsee, where no direct relationship exists between the sponsor and consumer or audience. This mediated nature of sponsorship requires that sponsors understand the nature of the relationship between consumer-audiences and the sponsee, especially where sponsorship objectives are dependent upon those relationships.

The review of common sponsorship objectives argues that three categories of objectives exist: communications, consumer-based, and behavioural objectives. As this research is concerned with the consumer perspective, communications objectives which focus on media exposure, frequency of brand name display and audience

131

awareness of signage (see: Calderon-Martinez et al, 2005; Cameron, 2009; Cornwell, 2008; Felton, 2009) are regarded as inappropriate foci for this research. Behavioural objectives of sales and share price changes (see: Cornwell, Pruitt & Van Ness, 2001; Johnson, 2010; Maestas, 2009) are also inappropriate. Each of those categories of objectives assumes a consumer evaluation; for example, as a consequence of media exposure, or antecedent of share purchases or sales.

Attitude towards the brand (see: Coppetti et al, 2009; Dardis, 2009; DeGaris, West & Dodds, 2009; Farrelly, Quester & Burton, 2006; Gwinner & Bennett, 2008), and perceived corporate image (see: Bennett, 1999; d’Astous & Bitz, 1995; Javalgi et al, 1994; Pope & Voges, 2000; Pope, Voges & Brown, 2004, 2009) are frequently cited sponsorship objectives. They exist within the category of consumer-based objectives which includes variables directly assessing consumer responses to sponsorship stimuli. It is the popularity and direct assessment of these objectives that makes them appropriate measures in this research, which asks how consumers perceive the sponsor in a mediated context.

Theories that explain consumers’ relationships with sponsees, as they affect consumer relationships with sponsors are also identified in the literature review. These theories are the social identity theory of identification (Tajfel, 1978, 1982) and Heider’s (1958) balance theory. Each of these theories explains how an individual’s perceptions may be biased according to either their beliefs about their own group membership (Ashforth & Mael, 1989; Tajfel, 1982), or a desire for cognitive consistency (Gawronski & Strack, 2004; Gawronski, Walther & Blank, 2005). The differences between these theories have implications for construct measurement and analysis which are important for Study 2.

This preliminary work contributes to the design of the primary research for this thesis in two ways. The first, it highlights that the off-field behaviours discussed as problematic are currently unknown. The second contribution is the identification of dependent variables (brand attitudes and corporate image) and a mediating variable (identification) for the experimental study. In the following sections, key research considerations are discussed, the goals of the studies and key variables of interest are also identified.

132

Research Considerations: Athlete off-field behaviours, respondents & sampling

This research investigates negative, and positive, athlete off-field behaviours and related perceptions of the athlete, team, and sponsors. The decision has been made to study off-field behaviours in the specific context of the National Rugby League (NRL) competition for reasons provided previously (see: page 32). In studying the NRL and off-field behaviours, the research is exposed to several specific forms of criticism that other projects readily avoid. These relate to sport audiences and gender, and student samples. That these issues occur in tandem in this research might further raise the question of the effects of their confluence. Each of these issues is considered here, with consideration of their effects the generalisation of findings.

The issues are considered here, in the Conceptual Framework because they are not unique to a single study or pre-test. In discussing them here, the purpose is to avoid reproducing the same arguments across several chapters.

(Male) Sports & Sport Audiences

Professional sports may not be a purely masculine marketplace; media presentations of sports, however, rarely distribute women’s sports or sporting achievements as widely. This bias is evident in television ratings (see: page 50, Table 2.4), and also in content analyses of news-media television, newspaper and magazine reports.

Table 2.4, on page 50, entitled ‘Sports programs ranked by audience size (000’s), 2001-2009’ identifies 28 televised sporting events which achieved audiences ranging in size from approximately 1.8 million people to over 4 million. Nineteen of those programs are extremely likely to be male sports (AFL, Rugby League, World Cup Soccer, Australian Open male finals, One-day Cricket, 20/20 Cricket, and Cricket World Cup). The remaining programs televised include Commonwealth Games events, World Cup Swimming, Wimbledon, or other Australian Open Tennis matches. Within this list, there are no programs achieving large audiences that are exclusively broadcasting a women’s sporting event.

133

Content analyses of sports’ reporting argue the over-representation of male athletes and men’s sports in the media. This conclusion has been reached in many different countries: Britain (Biscomb & Griggs, 2012; King, 2007); Italy (Capranica & Aversa, 2002); U.S.A (Christopherson, Janning & McConnell, 2002; Kinnick, 1998; Pedersen, 2002); Canada (Crossman, Hyslop & Guthrie, 1994); Norway (Fasting & Tangen, 1983); Holland (Knoppers & Elling, 2004); and cross-country studies (Crossman, Vincent & Speed, 2007; Lee, 1992; Valgeirsson & Snyder, 1986; Vincent, Imwold, Masemann & Johnson, 2002; Yu, 2009). There can be little doubt that quantitatively, men’s sports or male athletes feature in more news stories than women athletes or women’s sports across media markets.

Sports audiences are also thought to be predominantly male. Capranica and Aversa (2002), in a study of Italian airtime and audience data during the Sydney Olympic Games found that the broadcast of women’s sports in Italy comprised 26% of all sports broadcast; women represented 38% of Italian athletes. Women comprised 40% of the total mean television audience (Capranica & Aversa, 2002).

Attendance at Australian sports events is skewed male. The Australian Bureau of Statistics provides that for the top 20 sports by attendance in 2009-10, women constitute 45.4% of spectators (ABS, 2010b). Average data, of course, obscures important differences. Women provide the majority of event spectators for tennis, netball, hockey, lawn bowls, equestrian and touch football events (ABS, 2010b). Some of those ‘majorities’ are extremely small (lawn bowls: 50.9%; hockey: 51.4%); others are extremely large (equestrian: 73.6%; touch football: 73.7%; netball: 69.6%) (ABS, 2010b). The only male-dominated sports with extremely large majorities are cricket (outdoor: 71.9% male), and boxing (87.0% male).

The number of men across Australia who attended a rugby league match in 2009-10 is 969 100 (62% of total attendances) (ABS 2010b). The number of Queenslanders (male and female) who attended a rugby match in 2009-10 is 598 000 (or, 38.24% of all rugby league attendances) (ABS, 2010b).

Further perusal of ABS sports attendance figures show that for the 18-24 age group, 107 000 males attended a rugby league match in 2009-10; 107 400 18-24 year-old

134

women attended a rugby league match in the same year (ABS 2010b). This data could represent an anomaly. It otherwise provides evidence that including young women in a study of news’ reports discussing NRL is not likely to create bias derived from an interaction of sport and gender.

Student Samples

Student samples, or student respondents, have provided data for Pretests 1 and 2, and Study 2. The use of student samples is justified for this research by the conditions associated with experimental designs (Study 2). Experimental designs utilise strict controls, including: random assignment to groups; sample homogeneity; the constancy of conditions; and, manipulation of the independent variable (Singleton & Straits, 2005). These controls are implemented to minimise threats to internal validity (Hooghe, Stolle, Maheo & Vissers, 2010; Singleton & Straits, 2005; Thomas, 2011).

One effect of the attempts to manage internal validity is that it prompts, in the case of student samples, questions about the generalizability of findings. University students have been thought to represent populations of privilege, alienated by their level of education and life experiences, from broader populations (see: Gordon, Slade & Schmitt, 1986; Hooghe et al, 2010; Singleton & Straits, 2005). The generalizability of experimental findings, however, should always be regarded as limited (Singleton & Straits, 2005) because of the controls instituted in the design which shield data from naturalistic sources of interference.

The use of student samples is appropriate, specifically, when the research design provides a “valid test of a hypothesis”, (Singleton & Straits, 2005, p161). Student samples might also be justified when the “purpose of the research is to examine general principles (i.e., can a phenomena occur), or if the question can be persuasively examined among students” (Shen, Kiger, Davies, Rasch, Simon and Ones, 2011, p1062). The research conducted for this thesis is a test of a general principle. Generalisation beyond the student population should be sought, according to Singleton and Straits (2005), through studies which seek to replicate research design and findings.

135

Study 1: Content Analysis of Local News

The research question for this study is: RQ1. What types of athlete off-field behaviours are reported?

The method used is a quantitative content analysis. The purpose is exploratory, and no hypotheses have been developed for this study. The method, summary tables and discussion of findings are provided in Chapter 4.

The goal of this study is to identify the range of athlete off-field behaviours commonly reported in the news-media, specifically in local newspapers. It does not intend to identify only behaviours that have been responsible for sponsorship contract cancellation. Instead, it is intended to identify a spectrum of behaviours that may be used in a balanced design for Study 2.

Study 1 reports: a) A list of athlete off-field behaviours b) Category frequencies and comparisons c) Attempted categorisation of behaviours d) Discussion of findings

Pretest: Valence of Behaviours

The research question for this pretest is: RQ2. Removed from news-media and sports contexts, how do individuals evaluate the off-field behaviours reported?

The method used for this pretest is a survey; results are descriptive. This pretest has two purposes. These are 1) to establish whether the behaviours identified in Study 1 are regarded as positive, negative or neutral when removed from sporting and news- media cues, and 2) to identify a measure of behaviour valence for use in Study 2.

136

Two measures are tested in this pretest. These are an Attitude toward the Act measure and a Moral Diagnositicity measure. The measures derive from different disciplines and are considered to have collinear potential.

The pretest of behaviour valence reports: a) Evaluation of behaviours using a moral diagnosticity measure b) Evaluation of behaviours using an attitude toward the act measure c) The range of evaluative valence of the behaviours tested (positive to negative)

Ethics approval for the materials and procedures developed for this study were granted for data collection at two universities.

Application title: Identification & Impressions of Personality Traits: scale reductions Institution Approval Reference Dates QUT 1000000476 Clearance to 21/04/2013 Griffith University MKT/31/10/HREC Approved: 22 September 2010

Pretest: Identification scale

The research question for this pretest is: RQ3. If group (team) identification has the power to influence consumer evaluations, how should team identification be measured for this research?

The data collection method used for this pretest is a survey. Data is analysed using exploratory and confirmatory factor analyses (EFA, and CFA). The purpose of this pretest is to identify a team identification scale measure that can be used in Study 2. The literature reveals a large number of group identification scales. The existing sports team identification scales, however, are those in need of refinement and replication. This study identifies conceptually strong items from existing scales, queries the dimensional nature of identification, and produces an identification scale for use in the current research.

137

The pretest of identification scale measures reports: a) Tests of identification scale items b) Development of an identification scale measure

Ethics approval for the materials and procedures developed for this study were granted for data collection at two universities.

Study 2: Experimental test of news’ stimuli on consumer evaluations

The research question for this study is: RQ4. Are consumers’ evaluations of an athlete, the athlete’s team and the team’s sponsor influenced by news-media reports detailing the athlete’s off-field behaviours?

Sub-questions for the study are: a. Does team identification moderate evaluations? b. Are evaluations balanced?

The method used is an experimental design. The purpose is to explain the relationships among consumer evaluations of an athlete, the athlete’s off-field behaviours, their team and corporate sponsors. The method, tables of findings and discussion are provided in Chapter 6. The variables used in the research, hypotheses and research models are presented here.

The Study 2 experiment reports: a) The effects of news article stimuli on athlete, team and sponsor evaluations b) The effects of identification on athlete, team and sponsor evaluations c) The extent of evaluative balance across evaluations

138

Ethics approval for the materials and procedures developed for this study were granted for data collection at two universities.

Application title: Tests of news reporting on person perception, sponsorship and branding

Institution Approval Reference Dates QUT 1000000736 Clearance to 7/09/2013 Griffith University MKT/29/10/HREC Approved: 22 September 2010

Variables used in Study 2

A description of variables used in this research is provided here. The majority of measures are operationalised as either 7-point Likert-type, or semantic differential scales. The independent (treatment) variables in this research are a variety of off- field behaviours identified in the Study 1 content analysis.

Three categories of dependent variables have been used in this research: person impression variables (attitude toward the act; and person likeability); team variables (attitude toward the brand; and corporate image); and, sponsor variables (attitude toward the brand; and corporate image).

Treatment Groups: Off-field incidents reported

Study 1 produced a list of commonly reported off-field behaviours. Three of the behaviours identified in Study 1 and pretested in Pre-test 1 were used in the Study 2 experiment as information-based stimuli. Those behaviours are: assault, drink driving, and fundraising for charity.

Covariate: Team identification

Identification with the sporting team is conceptualized as a self-defined attachment to

139

the sporting team. According to social identity theory, identification with a group or organisation has the potential to influence or regulate individuals’ responses activities within or related to their group (Ashforth & Mael, 1989; Ellemers et al, 1999).

This variable is measured using a scale developed in Pre-test 2. The scale items for this measure are: 1. I have a strong sense of belonging to the team 2. I think of the team as part of who I am 3. I am like other fans of the team 4. Others describe me as a typical fan of the team 5. I have a strong attachment to the team

Dependent Variables

Four scales are used in the measurement of dependent variables in Study 2 experiment. These measures include a measure of ‘person impression’, ‘attitude toward the act’, ‘attitude toward the brand’, and ‘corporate image’.

Person impression: person likability

The person impression measure derives from the social-psychological literature that asks respondents to evaluate a person following information stimuli. In previous research, this variable has been operationalised as a single-item measure. Respondents asked to provide a person impression are told to indicate how “likeable” a person is, or the warmth/coldness of their feelings toward an individual (Fiske, 1980; Klein, 1996).

The measure used in this research is a modification of Fiske’s (1980) person impression measure. Respondents were asked, “How generally likable is the person described in the newspaper article?” The Likert-type statement consists of responses ranging from (1) not likeable at all, to (7) very likeable.

140

Attitude toward the Act

The Attitude toward the Act construct is usually used to measure a “subject’s evaluation of a purchase activity” (Bruner and Hensel, 1994, p68). In this research, the construct is used to assess respondents’ evaluations of off-field behaviours. Use of the Attitude toward the Act scale in this research is justified on the basis of its use in social marketing research, where behaviour (and behaviour change) is the ‘purchase activity’ evaluated (Bagozzi, 1982).

Bruner and Hensel (1994) reviewed fifteen sources using a semantic differential measure for an attitude to the act variable and found a total of sixteen items that had been used. Research using the combination of items used in this research has not been found. However, Bruner and Hensel report that of the fifteen sources they reviewed, “no reason is generally given to justify the particular sets of adjectives used in particular studies” (1994, p68). Despite this, research cited by Bruner and Hensel reports Cronbach’s alpha’s ranging from .81 to .95 (1994, p69).

Use of an Attitude toward the Act measure was trialed in Pre-test 1. A four-item semantic differential scale was tested; items used include: Good/ Bad; Harmful/ Beneficial; Pleasant/ Unpleasant; and Safe/ Unsafe. These items were selected for their suitability in assessment of off-field behaviours. Cronbach’s alphas’ revealed the scale was improved, across several scenarios, with the exclusion of the Pleasant/Unpleasant item. The Study 2 used a 3-item Attitude toward the Act scale.

Attitude toward the Brand

The Attitude toward the Brand measure is utilised twice in this research; to measure attitude toward the sponsors’ brand, and also attitude toward the sporting team as a brand. The Attitude toward the Brand measure shares the uncertain parentage, and judgment-driven item selection of the attitude toward the act measure (Bruner & Hensel, 1994).

141

Items used in this research are: Bad/ Good; Unfavourable/ Favourable; and, Negative/ Positive. These items are commonly used to measure attitude toward the brand, and have been used by Grossbart, Muehling and Kangun (1986), Kinney and McDaniel (1996), Muehling and Laczniak (1988), Roy and Cornwell (2003), and Sujan and Bettman (1989). A further justification of the use of these items is that the reliability of this scale is almost guaranteed; the sources cited report Cronbach’s alphas ranging from .93 to .97 across six research projects.

Corporate Image

The corporate image variable is also utilised twice in this research, to measure perceptions of the corporate sponsor, and the sporting team (as the organisation responsible for the sporting team as brand).

The corporate image construct “…captures the subjective perceptions of the company” (Javalgi et al, 1994, p50). This measure, Javalgi et al (1994) suggest, is comprised of six components, although the reliability and validity are not reported. Pope and Voges (1999) argue for the refinement of the corporate image construct to a five-item scale on the basis that Cronbach’s alpha for the scale improves, from .67 to .76 (Pope & Voges, 1999, p22). Pope, Voges and Brown (2009) report equally high alpha’s, ranging from .72 to .82 in two studies, each across four testing periods.

Demographic Variables

Demographic information collected included details of age, gender and first- language spoken. Age is not expected to influence results.

Gender and First-language spoken are collected, and considered potential confounds to analysis. Gender is not expected to influence results; however, as rugby league is played at professional level by men only. First-language spoken is sought because the population of undergraduate business school enrolments typically contains large numbers of international students. Any persistent difference in interpretation of the 142

information stimulus may be attributable to language proficiency (or cultural differences).

Hypotheses & Research Models

There are three groups of hypotheses. They are reflect expected effects of treatment group, identification with the sporting team, and evaluative balance.

Two research models are developed. The first research model reflects the effects of treatment group and identification (hypotheses 1 and 2). It is the experimental model showing treatment groups, the covariate (identification) and dependent variables.

Treatment Group

If negative news’ reports of athlete’s off-field behaviours constitute a threat to the sponsor’s objectives, the valence of off-field behaviours reported should influence evaluations of athletes, their team, and sponsors.

Therefore, effects of treatment group are hypothesised: H1: Valence of news’ will be reflected in evaluations of the athlete, their team and the team’s sponsor.

H1a: Valence of news’ will be reflected in evaluation of the athlete’s behaviour (Attitude toward the Act). H1b: Valence of news’ will be reflected in evaluation of the athlete (Person likability). H1c: Valence of news’ will be reflected in evaluation of the team (Attitude toward the Team).

H1d: Valence of news’ will be reflected in evaluation of the image of the team (Team Corporate Image). H1e: Valence of news’ will be reflected in evaluation of the sponsor (Attitude

143

toward the Sponsor). H1f: Valence of news’ will be reflected in evaluation of the image of the sponsor (Sponsor Corporate Image).

Identification

Social identity theory explains that identification with a social group influences perceptions of that group and associated out-groups (Hogg, 2000, 2009; Hogg & Abrams, 1999; Tajfel, 1974, 1982). High levels of group identification are associated with in-group favouritism (Fielding, Hogg & Annandale, 2006; Hogg, Turner & Smith, 1984). Further, when negative group information is presented, highly identified subjects maintain positive self-esteem in evaluative contexts (Doosje et al, 2006; Wann & Branscombe, 1992).

Therefore: H2: Level of Team Identification will moderate evaluations of group members (athlete, team, and sponsor)

H2a: Team Identification will moderate evaluations of the athlete’s behaviour (Attitude toward the Act). H2b: Team Identification will moderate evaluations of the athlete (Person likability). H2c: Team Identification will moderate evaluations of the team (Attitude toward the Team). H2d: Team Identification will moderate evaluations of the image of the team (Team Corporate Image). H2e: Team Identification will moderate evaluations of the sponsor (Attitude toward the Sponsor). H2f: Team Identification will moderate evaluations of the image of the sponsor (Sponsor Corporate Image).

144

Research Model A: Treatment and Identification

Model A shows that news’ article stimuli comprise the treatment groups for the experiment, and that Team Identification is tested as a covariate variable. Hypotheses 1 and 2 are depicted in this model.

Figure 3.8: Model A: Hypotheses 1 & 2

Person Variables: Person Likability Attitude toward the Act

H1 News’ Team Variables: article Corporate Image stimuli .00 Attitude toward the Brand 0) Team H2 Sponsor Variables: Identification Corporate Image Attitude toward the Brand

Balance

Balance theory explains the desire of the individual to maintain consistency across evaluations of linked or related individuals and objects (Heider, 1958). The linked individuals and entities in this research are the athlete, the athlete’s team, and the team’s sponsor. Therefore: H3: The valence of the reported behaviour (positive/ negative) will be maintained across evaluations of the athlete, their team, and the sponsor.

145

Research Model B: Balance

Model B, depicting Study 2 hypotheses, is shown in two formats. These models represent the balance theory hypothesis. ‘P’ in these models represents the perceiver. The models assume three principal Unit Relations: an employment contract (Athlete- Team); a sponsorship contract (Team-Sponsor); and the association between team employee and sponsor (Sponsor-Athlete). Unit Relations are positively valenced.

Figure 3.9: Model B: Hypothesis 3

U Athlete Team

DL / L U DL/ L

DL/ L p Sponsor

Notation for the model: DL: Dislike (negative sentiment relation); L: Like (positive sentiment relation); U: Unit Relation (positive/ structural tie); p: the perceiver

Balance in triads may be determined in three ways. A triad is balanced when all three of the relations are positive or when two of the relations are negative and one is positive. Imbalance occurs when two of the relations are positive and one is negative. Heider, 1958, p202. For example, the Athlete-Team-perceiver triad is balanced positive where the perceiver (p) likes the team, and likes the athlete, and the athlete and team have a unit relation.

146

The majority of balance theory literature refers to relationship triads (see: Crandall et al., 2007; Dalakis & Levin, 2005; Dean, 2002; Langer, et all., 2009; Reisinger, Grohs & Eder, 2006; Van Heerden, Kuiper & Saar, 2008). Hummert, Crocker and Kemper (1990) identify two means of achieving balance in a group of four. These are: “…either every member of the group will like everyone else… or the group will be split into two mutually hostile cliques (Hummert, Crocker & Kemper, 1990, p6).

Figure 3.10: Model B (ii): Hypothesis 3

Athlete Athlete

U DL/ L U DL/ L U U

DL/ L DL/ L Team p Team

U DL/ L DL/ L U

Sponsor Sponsor

Effects of treatment: Group Effects of covariate: Identification with Team

This model assumes that responses to each of the pairs of variables will be consistent, i.e., that the respondent will respondent positively to both Team Corporate Image and Team Attitude toward the Brand, or negatively toward both person variables (likability and Attitude toward the Act).

147

Chapter 4: Study 1: Content analysis of local news

The current study is a content analysis of local news-media reporting of the Australian National Rugby League (NRL) competition. This limitation is chosen on the basis of the location of the researcher and the characteristics of the local population. Australia is a country known for its love of sport (ABS, 2009a). Popular sports include football (Australian Rules), soccer, swimming, netball, tennis, cricket and rugby, among others. The popularity of various sports is not equally distributed across all states and territories (ABS, 2009a).

Australian statistics reveal that Australian Rules football is the most popular sport for attendance at sporting events (excluding junior sport events) (ABS, 2009a). In Queensland, however, annual rates of attendance at Australian Rules games are 7.4% (ABS, 2009a). In comparison, attendance rates at Rugby League are 16% overall (or 20% for men only) (ABS, 2009a). These statistics make Rugby League the most popular sport in Queensland, by attendance.

The research question for this study is: RQ1. What types of athlete off-field behaviours are reported?

This study is exploratory and descriptive. It aims to produce a list of the off-field behaviours of NRL players that are commonly reported in the local newspaper.

Aims

The goal of this research, overall, is to establish whether negative off-field behaviours of athletes affect the achievement of their sponsor’s objectives; producing a list of athlete off-field behaviours (the goal for this study) is a small part of the overall goal. A key assumption of the research is that where sponsorship contracts

148

are cancelled following the news-media reporting of negative off-field behaviours, sponsors have anticipated a threat to the achievement of their objectives. Many examples of sponsorship contract cancellation exist. They include the following:

1. Michael Phelps, USA swimmer, was punished with a three-month ban from competition after being filmed smoking cannabis. His sponsor, Kellogg’s, shortly thereafter released a statement that they would not renew his contract given that his behaviour was “not consistent with the image of Kellogg” (BBC SPORT, 2009).

2. Stephanie Rice, Australian swimmer, had her sponsorship contract with Jaguar Australia revoked following a homophobic tweet (Zelkovich, 2010).

3. LG Electronics Australia did not to renew their sponsorship of Australian rugby league team, the Cronulla Sharks, following a number of off-field events. These events included the positive drug test of player Reni Maitua, and allegations of a group sex scandal. LG released a statement saying; While the recent controversies around the NRL (National Rugby League), and the Sharks in particular, were certainly a significant element in our decision we also considered the direction of our company and where we want to take the LG brand. Byrnes & Phelps, 2009 LG was not the only Cronulla Sharks sponsor to cancel their contract.

4. The Transport Accident Commission (TAC), a compulsory third-party insurance company in Victoria has been involved twice in contract cancellations. The TAC, as well as providing drivers with insurance, is also a promoter of road safety; in particular anti-speeding and anti-drink-driving campaigns (Broad, 2009; Maiden, 2005; Roberts, 2008). The Richmond Tigers, and the Collingwood Magpies, both Australia Football League teams, have lost deals with the TAC because of players caught drink-driving (Broad, 2009; Maiden, 2005; Roberts, 2008). According to Maiden, “the promotional contracts it [TAC] writes have penalty clauses in them to cover undesirable behaviour” (2005).

149

5. Finally, the UK Court of Appeal has ruled in favour Etihad Airways and Aldar Properties, whose contractual rights in sponsorship from Formula One team, Force India, were breached. The bases of lawsuits brought by Etihad and Aldar were clauses identifying Etihad’s right to be the only airline sponsoring Force India, and the desire of both companies to avoid associations with tobacco, pornography, gambling or alcohol related ventures (EWCA Civ 1051, 2010). When Force India accepted sponsorship from Kingfisher Airlines and Kingfisher beer, the team materially breached their agreements with Etihad Airways and Aldar Properties.

Each of these examples provides insight into the corporate objectives of sponsors. Corporate image, brand awareness and access to target markets all feature prominently. The examples also demonstrate that contract cancellations occur because of activities undertaken at the team, club management or athlete level, in the course of competition and outside of it. Many more examples of contract cancellations can be found by searching news’ databases; the results of such searches, however, are unsystematic and subject to the biases of the researcher.

This content analysis aims to provide a systematic picture of the reporting of off-field behaviours of the athlete’s from one sport in a local area. Several limitations are implicit in this aim. These limits are applied to locality, sporting code and media. Specifically, the local daily newspaper in Queensland is surveyed for information regarding the National Rugby League competition. Further details on the content analysis method and coding are provided in the following section.

Method

A review of articles referring to the Australian National Rugby League (NRL) competition was conducted in order to develop a list of off-field player behaviours reported in the news-media. The process for conducting content analyses developed by Neuendorf (2002, p50-51) was used to structure the study. The model is used in the following headings.

150

Theory & Rationale

Agenda-setting theory suggests that producers of news-media have the potential to influence not only what news-consumers think about, but also “…how they think about those issues” (Fortunato, 2008, p118). Local newspaper articles referring to the NRL competition are the subjects of this content analysis as they represent a mass circulation media format with the potential to influence local opinions.

The rationale for the study is to produce a list of off-behaviours of rugby league players that might influence consumer perceptions of sponsorships, or sponsor’s approaches to sponsorship contracts. Hypotheses have not been developed for this study as the purpose is exploratory rather than relationship testing.

Conceptualisation

Behaviours are the variables studied in this content analysis. A further condition is evidence of individual will. Behaviours that appear directed by a collective (sport team, club or competition authority) are not regarded as the off-field behaviours of individuals.

Further items coded in this study include references to injuries, on-field incidents, match play and match preparation (training and team selection). On-field incidents are those that involve dangerous or unacceptable conduct on-field during the course of competition.

Operationalisation & Coding

The unit of data collection is the newspaper article; multiple behaviours may be coded from each article. Six variables were coded for each article. These variables are tabled below.

151

Table 4.9: Variable Descriptions & Coding Variable Description Coding name Article type Articles were coded for: on-field and competition-related Nominal: issues (1); off-field incidents (2); reports of both on- and numbers 1-4 off-field issues (3); and extraneous reports (4). The extraneous reports category included articles recording share prices, eulogies, consumer comments, references to other sports, and others Off-field Off-field behaviours were coded if they conformed to the Binary: 0 = behaviours appearance of individually directed/ motivated behaviours absent; 1 = provided in the above section. present

Off-field behaviour is recognised through individuating information. Four pieces of information: the ‘who’, ‘what’, ‘when’, and, ‘where’ of reporting are used to fully identify a specific event. An incident is coded if at least two of the four pieces of information are present in a newspaper article. Number of A count of off-field incidents referred to in each article. Frequency off-field behaviours per article Match play This category coded training, team selection, match play Binary: 0 = & match and post-match evaluation of competition. absent; 1 = preparation present Injuries Articles in this category reflect whether injuries have been Binary: 0 = sustained by players absent; 1 = present On-field Reference unsporting or dangerous play during Binary: 0 = incidents competition; also includes mentions of the sport judiciary absent; 1 = body and suspensions or fines allocated as punishment for present on-field incidents.

The majority of variables were coded as binary categorical variables, 0= absent, 1= present. Only the ‘article code’ variable and ‘number of off-field behaviours’ variables did not conform to this system. The ‘article code’ variable comprised four categories, and the ‘number of off-field behaviours’ category is a raw count.

Sampling

A census of articles published in The Courier Mail was conducted for the period January to June 2009. This period includes the off-season, and the beginning of the competitive season. A random sample of articles was not deemed appropriate as a

152

census allowed observation of behaviours reported multiple times during the period.

The Courier Mail newspaper is the Queensland local daily newspaper. Use of The Courier Mail as the source publication is justified based on the average readership in 2009. The NewsSpace website (www.newsspace.com.au) cites Roy Morgan Research to suggest that average (Monday to Friday) metropolitan readership of The Courier Mail is 603,000; in regional areas of Queensland the Monday-Friday average is 193,000. The population of Queensland in 2009 was estimated at 4.43 million people (ABS, 2008-09a). Census data indicates that, on average, there are 3 people per household (ABS, 2010d: National Regional Profile). Combined, these numbers suggest that The Courier Mail reached roughly 53.9% of Queensland households (Monday to Friday) in 2009. These numbers, approximate though they are, provide an indication of the reach of The Courier Mail, and its potential influence on Queensland sport consumers.

A search of the ‘factiva’ database yielded a total of 1834 articles referring to the NRL during the six-month period January-June 2009. The search terms include competition name, colloquial name and all names of teams participating in the competition. These are shown below:

NRL OR rugby league OR Newcastle Knights OR North Queensland Cowboys OR Penrith Panthers OR New Zealand Warriors OR Gold Coast Titans OR Parramatta Eels OR Canberra Raiders OR Wests Tigers OR Brisbane Broncos OR St George Illawarra Dragons OR Cronulla Sharks OR South Sydney Rabbitohs OR Melbourne Storm OR Canterbury Bulldogs OR Sydney Roosters.

Training & Pilot reliability, Coding, & Final reliability

Multiple coders were not used for this study. The generation of a behavior list is the purpose of this study. The use of multiple coders was judged not necessary for a minimalist agenda.

153

Tabulation, Analysis & Reporting

The data in this content analysis are coded in three ways. One variable (article type) is coded purely nominally; numbers 1-4 depict categories of article content. Several variables (off-field; match play; injuries; and, on-field) are binary coded. The final variable (number of off-field behaviours reported) is a frequency or raw count of incidents reported.

The type of data collected influences the type of analyses conducted. In this research, numeric data are compared using raw frequencies, Cochran’s Q, and multiple t-tests. Qualitative information is also collected to describe the type of off-field behaviours reported, the actor, and other ‘who’, ‘what’, ‘when’, and ‘where’ information. The quantitative forms of analysis and related concerns (Bonferroni corrections) are discussed below.

Cochran Q: tests of difference for binary variables

The Cochran Q test is a nonparametric test of difference used when multiple variables or stimuli are compared (Conover, 1980; Sprent & Smeeton, 2000). It is specifically designed to analyse binary, or dichotomous, data (Conover, 1980; Sprent & Smeeton, 2000). In the current analysis, the Cochran Q test is justified on several bases, these are:

 Cochran Q is designed for multiple groups (StatSoft, 1994; Sprent & Smeeton, 2000). For two groups, McNemar’s Chi-square test produces an equivalent result (Conover, 1980; Sprent & Smeeton, 2000).  Where Cochran Q is not used for multiple groups, it is particularly suited to multiple tests of the same variable, making it appropriate for longitudinal data (Conover, 1980; StatSoft, 2000).  Cochran Q is appropriate for dependent samples (Sprent & Smeeton, 2000; StatSoft, 1994).  Although many nonparametric techniques are not suited to large sample sizes

154

(StatSoft, 1994); “for large samples Q has approximately a chi-squared distribution” (Sprent & Smeeton, 2000, p227), indicating that Cochran Q can process data from large sample sizes.  The Kruskal-Wallis analysis of ranks is the nonparametric equivalent to ANOVA. This test is appropriate for independent samples and ordinal levels of measurement (Sprent & Smeeton, 2000; StatSoft, 1994). In the current analysis, it would be fanciful to suggest that binary data is ordinal, or that the samples are independent.

Tests of Difference & Bonferroni

Multiple t-tests are used in this analysis to assess differences in the frequency of a code’s occurrence within a variable (i.e., does code 2 appear with equal frequency to codes 1, 3, and 4). The multiple t-test identifies differences between code pairs; it is used instead of a one-way anova which, when significant, requires a step-down procedure to identify specific pair differences.

The use of a multiple t-test raises the problem of a Bonferroni inequality, which argues the increased likelihood of a Type I error in multiple comparison analyses (Benjamini & Hochberg, 1995; Hair, Anderson, Tatham & Black, 1995; Simes, 1986). The Bonferroni correction makes significance testing more stringent by altering the acceptable p-value to reflect the change in risk level of a Type 1 error – mistakenly rejecting the null hypothesis (Simes, 1986). Specifically, “the classical

Bonferroni multiple test procedure is usually performed by rejecting H0 = {H1…, Hn} if any p-value is less than α/n” (Simes, 1986, p751).

Reporting

The reporting of results for this study is provided under the heading: Findings.

155

Findings

A search of the factiva database revealed that The Courier Mail newspaper published a total of 1834 articles referring to the NRL during the six-month period January- June 2009. In this section frequencies are provided for each variable followed by tests of difference to establish the basis of comparisons across variables, finally, a list of the off-field behaviours reported during the period is provided.

Frequencies for ‘Article type’ variable categories by month

Table 4.10: Frequencies: Article type by month 1 2 3 4

On-field & Off-field Both on- & Extraneous Row totals competition incidents only off-field reports only reports January 37 3 8 60 108 February 86 0 20 75 181 March 177 6 60 150 393 April 174 9 40 119 342 May 218 1 77 135 431 June 203 2 32 142 379 Column 895 21 237 681 1834 totals

Table 4.10 shows the frequencies for each ‘article type’ category by month. Row and group totals indicate no missing values. The ‘article type’ variable is coded using a nominal level of measurement. The four categories are mutually exclusive.

The NRL competition generally begins competitive play in March, following pre- season training and ‘friendly’ games. Diagram 4.11 depicts the Table 4.10 data visually.

156

Figure 4.11: Article type: NRL in the Courier Mail January-June 2009 250

200 competition-related 150 only off-field only 100 both on&off-field

50 extraneous

Tests of Difference: Multiple T-test comparison of ‘Article type’ categories

Table 4.11, below, shows the p-levels for a multiple t-test used to determine whether the ‘article type’ categories are reported at significantly different levels. As there are six direct comparison tests, the acceptable p-value with Bonferroni correction is adjusted to .008. This correction yields four significant comparisons between category frequencies (1-2; 1-3; 2-4; and 3-4).

Table 4.11: Multiple t-test comparison: Article type codes Code 1 Code 2 Code 3 Code 4

On-field & Off-field Both on- & off- Extraneous competition incidents only field reports reports only Code 1 Code 2 0.004 Code 3 0.003 0.017 Code 4 0.072 0.000 0.000

Two issues are suggested by the comparison of frequencies. 1. “Noise” (or, integration) in NRL reporting This issue is prompted by the observed high frequencies, and no significant difference, in the frequencies of codes 1 and 4, p>. 008. This might indicate a degree of integration of the NRL into many facets of social/ daily life.

157

2. The business of sport is sport Frequencies for codes 3 and 2 are significantly different to frequencies for code 1 (p<.008). This result suggests that the focus of news-media reporting of NRL is sport. There is no evidence in code frequencies of over-reporting of off-field behaviours, or “hype”.

Frequencies for binary-coded variables

Table 4.12, below, shows the month-by-month frequencies for the ‘match play & preparation’, ‘reports of off-field incidents’, ‘on-field incidents’, and ‘injuries’ variables. Each of these variables was coded using a nominal level of measurement; the binary codes used are 0 = absent; 1 = present. In Table 4.12, the frequencies recorded for each variable by month reflect the count of 1= present totals for that variable.

Table 4.12: Frequencies: Binary variables by Month Match play & Articles Injuries On-field preparation referring to off- incidents field incidents January 14 11 5 1 February 55 20 29 0 March 170 66 78 18 April 158 49 69 27 May 132 78 74 49 June 107 34 96 23 Column totals 636 258 351 118

Table 4.13, below, shows the result of a Cochran Q test of difference across the four binary variables. The p-value for this test indicates that the reporting of these variables is significantly different.

158

Table 4.13: Cochran Q test of difference: Binary variables N=1834 Degrees of Freedom =3 Q = 571.7312 p <.000 Sum Percent Percent

0’s 1’s Match play & preparation 636.0000 65.32170 34.67830 Off-field incidents 258.0000 85.93239 14.06761 Injuries 351.0000 80.86150 19.13850 On-field incidents 118.0000 93.56598 6.43402

The Cochran Q test is used here in order to establish whether differences between variables exist. The significant p-value (p<.000) reported in Table 4.13 indicates that the binary variables have different distributions, and so are reported differently.

A further test of difference, also using Cochran Q, is of month-by-month differences within each variable’s distribution. The content analysis conducted coded articles reported by The Courier Mail from January to June in 2009. The NRL season, as noted earlier, begins in March. As the content analysis includes the pre-season period, it is expected that month-by-month differences in NRL reporting will be evident. The results of the Cochran Q tests of the difference in reporting each variable on a month-by-month basis are reported in Table 4.14, below.

Table 4.14: Cochran Q test of month-by-month differences: Binary variables Variable Degrees of Q p Freedom Match play & preparation 5 41.81818 .000 Off-field incidents 5 16.52113 .005 Injuries 5 21.79612 .000 On-field incidents 5 46.94737 .000

Table 4.14 shows that the reporting of match play, off-field incidents, injuries and on-field incidents differ on a month-by-month basis. There are significant differences within each variable by the month of reporting. Inferences derived from these comparisons are prosaic, and best illustrated using a diagram, see Diagram 4.12.

Several inferences are possible. The first is that variables reported reflect the maturation of the competitive season (i.e., reports of injuries increase). The second

159

inference is the seeming absence of a quota or level of reporting. In relation to the two peaks in reporting of off-field incidents, and also in relation to the apparent decline in match-play reports, it might be concluded that ‘shocks’ or events within the environment provide impetus for greater or lesser amounts of reporting.

Figure 4.12: Competition: match play, injuries, on-field & off-field reports 180 160 140 Match play & 120 preparation 100 Articles referring to 80 off-field incidents 60 On-field incidents 40 Injuries 20 0

The decline in match play reports at the mid-season point is attributable to the State of Origin competition, a three match annual competition between two state-based teams (Queensland and New South Wales). The State of Origin runs concurrently with the NRL competitive season and matches are played in June and July; State of Origin articles were not coded as part of the NRL season in this content analysis.

A final consideration is the apparent difference in reporting of off-field incidents and reporting of on-field incidents. The distributions of these variables may be significantly different (See Table 4.14). On the basis of Diagram 4.12, it would appear that off-field incidents have been reported more than on-field incidents which reflect unsporting or dangerous conduct. This situation may warrant further investigation into socially sanctioned or accepted forums for potentially harmful behaviours.

160

Frequencies for articles reporting off-field incidents & numbers of incidents

Table 4.15 shows the frequencies for two variables: the ‘articles referring to off-field incidents’ variable (a binary variable), and the ‘number of incidents reported per article’ variable (ratio scaled).

Table 4.15: Frequencies: Number of Articles & Number of incidents Articles referring to Number of off-field Average: off-field incidents incidents reported incidents per article January 11 13 1.18 February 20 33 1.65 March 66 128 1.94 April 49 86 1.76 May 78 122 1.56 June 34 48 1.41 Column totals 258 430

Test of Difference: T-test of ‘Articles’ variable & ‘Number of incidents’ variable

A paired t-test was conducted to test the month-by-month differences in the number of articles and number of incidents reported. No directional hypothesis is formulated for this test, although it is clear that the numbers of incidents reported must be at least equal to the number of articles that refer to them.

Table 4.16: Test of Difference: Numbers of Articles & Incidents Mean Standard t Degrees of p deviation Freedom Articles 43.00 26.17 Number of 71.83 47.98 -3.08604 5 .0273 incidents

The t-test establishes that in the current sample, articles that report off-field incidents involving NRL players frequently report more than one off-field incident per article. This result is significant, p<.05. The result of this comparison suggests the need for further research in this area. At present, the difference between numbers of articles referring to off-field incidents, and the numbers of incidents reported can be explained in a variety of ways. These might include:

161

 Category-based memory processes through which a journalist will recall similarly categorised behaviours cued by characteristics of the immediate reporting context.  Socio-cultural practice, such as the providing of examples when discussing a specific incident (not inconsistent with the first point).  Deliberately created “hype” by the news-media.  A non-generalisable artefact of the NRL season reviewed.

Off-field incidents

The main purpose of this content analysis is the production of a list of off-field incidents, or behaviours, involving NRL players. The census of articles has yielded a list of 52 discrete events, seven of which involved multiple actors. Where individuals were identified as participating in a multi-actor event they were coded individually. The two exceptions to this rule are the incidents that involve players only, and no other members of the general public. In cases when individuals were not identified by name, the team name was recorded as a substitute.

A complete list of events by actors is provided in Table 4.17, below.

The list of events by actor shown in Table 4.17 is largely uninformative. Data reduction through categorisation of the incidents serves to simplify analysis; there are several means by which this might occur. These include: by actor; by year; by morality/ immorality; by severity; or according to criminality of the behaviour; or, some combination of types.

162

Table 4.17: NRL off-field incidents reported by the Courier Mail Event by actor Event Year Count Event by actor Event Year Count id id Barba & Idris: Players involved in a brawl 12001 2009 2 Laffranchi: charged with rape. 12028 2008 2 (assault) Bird: charged with assault of his girlfriend 12002 2008 27 Maitua: conviction for assault overturned 12029 2005 1 Boyd: alleged sexual assault of a woman 12021 2008 14 Maitua: DUI 12030 2006 1 Canterbury Bulldogs: alleged gang rape 12003 2004 6 Maitua: tested positive for performance 12031 2009 10 enhancing drugs. Carney: AVO ban from the town of Goulburn. 12004 2009 4 Mason: Drink: ban for attending training after 12032 2009 9 drinking. Carney: Drink: sacked by club for urinating on a 12005 2008 20 Mason: urinating in public. 12033 2009 3 man. Carney: DUI 12006 2008 4 Michaels: Drink: public drinking – no match ban 12024 2009 7 Carney: nude photos found on a rental phone. 12007 2009 1 Myles: Drink: ban for attending training after 12034 2009 11 drinking. Cherrington: assault of girlfriend 12008 2009 1 Nahi: tested positive for amphetamine use. 12035 2009 1 Clinton: fined for woman in hotel room before a 12009 2009 13 Naiqama: cleared of assault charges. 12036 2009 5 match Costigan: DUI 12010 2006 2 Newcastle Knights: alleged gang rape. 12037 2005 1 Crocker: convicted of affray. 12011 2005 16 Sa: charged with assault/ pub brawl 12038 2008 1 Cronulla Sharks: allegations of group sexual 12012 2002 57 Sailor: Two-year suspension for cocaine use. 12039 2006 5 assault Earl: charged with assault of a woman 12013 2009 1 Seymour: Drink: refused service at a nightclub 12040 2009 8 Eastwood: repeated speeding offences. 12014 2006 7 Slyney: public nuisance 12041 2009 1 Fai: saved his brother from drowning in a rip. 12015 2009 5 Stewart: AVO 12042 2009 2 Friend: charged with assault of a woman 12016 2009 2 Stewart: charged with sexual assault of a 17- 12043 2009 47 year old girl. Friend: DUI 12017 2009 9 Stewart: Drink: refused service 12044 2009 37 Hindmarsh: auction boots for cancer charity 12018 2009 1 Taumata: charged with assault after a pub brawl. 12045 2009 4 Hodges: Drink: late for public appearance. 12019 2009 1 Tautai: DUI 12046 2009 1 Holdsworth: charged with assault/ pub brawl 12020 2009 1 Thaiday: alleged sexual assault of a woman 12021 2008 16 Hunt: alleged sexual assault of a woman 12021 2008 19 Thaiday: promotes community health program. 12047 2009 3 Iosefa: charged with affray. 12022 2009 4 Thompson: assault of girlfriend 12048 2009 3 Jeffrey: DUI 12023 2009 3 Watmough: assault of a sponsor 12049 2009 12 Kenny: Drink: public drinking – no match ban 12024 2009 8 Watmough: AVO for assault of girlfriend 12050 2007 1 Kenny: raises money for brain tumour sufferer 12025 2009 3 Webb: DUI 12051 2008 2 Kenny: volunteer physiotherapy 12026 2009 1 Wests-Tigers: alleged gang rape. 12052 2009 2 Lacey & Te'Reo: involved in a brawl (assault). 12027 2007 2 Total incident reports 430

163

Categorising by event characteristics

Several forms of categorization are not pursued here because they either do not serve effectively to reduce the data, or, they neglect the primary purpose of the content analysis, which is to develop a list of behaviours. Means of categorising off-field incidents have been mentioned above. These included, by year, morality, severity, or according to criminality.

Categorising events according to their year of occurrence, does little to reduce the data in a meaningful way. Of the incidents reported in 2009, nearly 79% of those occurred in 2008 or 2009. The lifetime of ‘old’ incidents would constitute a productive area of research, as the characteristics which govern their reporting outside of their year of origin give insight into news-values. This is otherwise an uninformative form of categorization.

Table 4.18: Number of Incidents & Total Reports by Year 2009 2008 2007 2006 2005 2004 2003 2002 Row totals Number of incidents 34 7 2 4 3 1 0 1 52 Total reports 229 105 2 15 17 6 0 56 430 Average reports 6.74 15 1 3.75 5.67 6 0 56 per incident

The behaviours coded in this study are those thought to be driven by individual will (rather than club or sponsor). A morality-based approach to categorization asks whether acts contributed to the liberty or quality of life of another individual, or represented the limitation or abrogation of duty or care. Five incidents, using these characteristics, were regarded as moral or altruistic:  Hindmarsh auctioning boots for charity  Kenny raising money for a brain cancer sufferer  Kenny providing free physiotherapy for a spina bifida sufferer  Fai saving his brother from drowning  Thaiday promoting an Indigenous community health program

The size of this group does not allow quantitative analysis. As little could be achieved using only moral and immoral categories for the off-field incidents

164

reported; other means are examined. Judging the severity of incidents constitutes an inherently subjective task; it is avoided here.

The criminality of behaviours, alone, is not deemed an appropriate means of categorisation for this study on the basis that not all of the negative incidents reflect illegal acts. Many of the immoral behaviours represent breaches of duty (the breaking of Club rules for off-field behaviour).

The categorization scheme developed here combines characteristics of behaviours reported as they relate to the central themes of morality: harm, care, duty and justice and a legal schema. The Australian Standard Offence Classification (ASOC) (2008) (Australian Bureau of Statistics, 2008-09) was reviewed in the preparation of the categorisation framework to gain insight into the legal categorisation scheme.

Table 4.19: Categories of Behaviour Category Definition Anti-social Illegal behaviours that have the potential to result in injury to the behaviours individual or other members of the public. This category includes: traffic offences, drug use, driving under the influence of alcohol (DUI), public nuisance, affray and the distribution of nude photos. Violent Illegal behaviours and police actions relating to those behaviours. behaviours Behaviours are characterised by potential or actual harm to others. The category includes: Apprehended Violence Orders (AVO), and assault charges. Due to the following category, this category reflects violence against men. Violence Charges or allegations of sexual assault, rape, gang rape, assault against against women women, or AVOs resulting from violence against women. Breaches of This category includes behaviours that are not necessarily illegal; they Club rules represent breaches of duty as they may affect training or match performance. Behaviours include drinking, or fraternising with women. Positive or Incidents or behaviours of individuals that are not obviously organised by altruistic the athlete’s Club or the NRL competition which result in potential or behaviours actual benefit to individuals outside of the NRL competition.

Table 4.20 shows the behaviours coded, the ASOC classification that would deal with the illegal acts identified, consideration of actual versus potential harm, and consideration of public concern for women in relation to the NRL.

165

Table 4.20: Categorisation consideations: ASOC and Morality Incident description, behaviour Australian Standard Offence Actual or Actual Category or crime Classification: Division Potential harm to harm to any person women Traffic & vehicle regulatory offences Speeding offences Dangerous or negligent acts endangering Potential No Anti-social Driving under the influence of alcohol persons

Amphetamine or cocaine use Performance enhancing drug use * Illicit drug offences Potential No Anti-social

Public nuisance Public urination Public order offences Potential No Anti-social Public drinking Affray Violent incidents No Assault Acts intended to cause injury Actual Violence against Yes women Sexual assault Violence against Sexual assault and related offences Actual Yes Alleged rape women Alcohol:  refusal of service in public venues Public order offences  late for training or drunk at (no police involvement) training Potential No Breach of Club rules

 late or missed public appearance

 Woman in hotel room prior to a match Promotion of a public health campaign Auction memorabilia for charity Positive or altruistic Save family member from drowning Potential No behaviours Raise money for a cancer sufferer Provide voluntary physiotherapy treatment

166

Table 4.21 shows the number of reports per incident according to assigned category. The categorization task produced five groups; incidents are not evenly distributed across the groups. The difference in group sizes makes quantitative analysis difficult. Several analysis techniques have been considered, and rejected, these include:  Multiple t-tests are hindered by heterogeneous variance, differences in sample sizes and non-normal distribution (Hair et al, 1995; StatSoft, 1994, p1318).  Freidman’s non-parametric ANOVA assumes equal medians (StatSoft, 1994). This condition is not met by the sample.  The Kruskal-Wallis non-parametric ANOVA tests multiple groups, but assumes independent samples and equal medians (StatSoft, 1994), these conditions are not met.

Table 4.21: Categories, Incidents, & Incident reports Anti-social Violent incidents Violent incidents Breaches of Club Positive or incidents involving women rules altruistic incidents event id count event id count event id count event id count event id count 12006 4 12001 2 12002 27 12005 20 12015 5 12007 1 12004 4 12003 6 12009 13 12018 1 12010 2 12020 1 12008 1 12019 1 12025 3 12011 16 12027 2 12012 57 12024 15 12026 1 12014 7 12029 1 12013 1 12032 9 12047 3 12017 9 12036 5 12016 2 12034 11 12022 4 12038 1 12021 49 12040 8 12023 3 12042 2 12028 2 12044 37 12030 1 12045 4 12037 1 12031 10 12049 12 12043 47 12033 3 12048 3 12035 1 12050 1 12039 5 12052 2 12041 1 12046 1 12051 2 16 70 10 34 13 199 8 114 5 13 events total events total events total events total events Total reports reports reports reports reports average 4.4 3.4 15.3 14.2 2.6

167

The categorization of off-field behaviours has not contributed to data reduction or analysis. However, several further questions are raised as a result. These questions are:  Are greater numbers of off-field behaviours reported because they have implications to team selection and training, in the same way that injuries have implications to those team functions?  Would positive or altruistic behaviours more reasonably be categorised as ‘extraneous’ reports? That is, are positive off-field behaviours reported for the sake of public interest, rather than for their implications to the sport?

Discussion

The research question guiding this content analysis is: What types of athlete off-field behaviours are reported?

Table 4.17 provides the answer to the research question for this study.

During the six month period from January to June 2009, The Courier Mail reported 52 discrete off-field events in 258 news articles. In many news articles, more than one event was mentioned. The incidents reported were categorised according to common characteristics, yielding five categories. Categorisation did not contribute to the quantitative analysis. A large number of incidents reported involve actual harm or violence (n= 23); many are identified as anti-social, with the potential to cause harm to the individual or others (n=16); very few of the behaviours reported are positive (n= 5). The selectivity involved in reporting has implications for a balanced design of the Study 2 experiment; these will be addressed in the following pretest.

Three further issues were identified for discussion. These are the issues of socially sanctioned violence in sports, ‘noise’ in the news’ environment, and ‘hype’ and news-values in sports reporting.

168

Sanctioned Violence

The issue of violence in sports falls outside of the main interest of this thesis. The issue of acceptable violence was raised as a result of the comparison of rates of reporting of on-field incidents and off-field incidents where off-field incidents out- number on-field incidents reported. There is much research that addresses the issue of on-field violence in sport.

Competitive sport is seen to require of athletes a “level of aggression, intimidation, and even violence” (Kim & Parlow, 2009, p584) in order to compete (see also: Grange & Kerr, 2010; Raney &, Depalma, 2006). The discourse of sport is often characterised by conflict, war or battle metaphors (Benford, 2007; Dittmer, 2006; Heinegg, 2003); and this travels with notions of the development of ‘socially useful skills’ (Peabody, 2009; Standen, 2009). Sport is seen as a providing a suitable venue for social violence (Dittmer, 2006) in that it is represented as controlling, or redirecting aggression into socially approved forms (Lasch, 1988). Finally, approval of on-field violence is often discussed in the context of audience enjoyment of the same (Jones, Ferguson & Stewart, 1993; Raney & Depalma, 2006).

The relative under-representation of on-field incidents reported to off-field incidents in this study does no more than support previous research that acknowledges the notion of acceptable aggression; while still including notions of unacceptable violence. The social frames for sport, and violence in sport, are not the focus of this thesis, and so, are not discussed further.

Noise in the news’ environment

The issue of noise is raised to explain the large number of ‘extraneous’ reports found in the initial analysis of article codes. In percentage terms, extraneous articles accounted for approximately 37% of the total. This percentage is equivalent to that found in other content analyses of sporting news. For example, Lyytimaki and Tapio (2009) found 49.4% of articles in their sample not germane to their research; Rissel, Bonfiglioli, Emilsen and Smith (2010) excluded 36% of articles from their sample

169

for irrelevance. Although the percentage of extraneous reports appears large, it reflects the integration of sport in daily life and news.

‘Noise’ as a concept, derives from physical technological systems designed to transfer messages between senders and receivers in communications processes. Shannon (1948) (see: Doms & Morin, 2004) developed mathematical models to account for message transmission in telegraph systems. Noise, in these systems indicated a situation in which “the signal is perturbed”, which “means that the received signal is not necessarily the same as that sent out by the transmitter” (Shannon, 1948, p19). These accounts of noise were adapted to explain human communication and are used to explain ‘perturbations’ in the reception of advertising messages and research contexts (Wakefield, Becker-Olsen & Cornwell, 2007).

Noise in psychological explanations of memory and information processing carries the same meaning as noise in the above mentioned contexts. It is taken to “reduce the discriminability of the signal” (Eich, 1985, p9). Eich’s explanation of noise suggests that it is an outcome of associative processing (Eich, 1985). Specifically, when people associate two stimuli (ie., a sportsperson identified as a team member, and their off-field behaviour), a single memory trace results. This trace is a composite trace that is the convolution of: a) the sportsperson as team member, and b) the off- field behaviour of the person. Retrieval from memory occurs when some new stimuli (e.g., the team wins a premiership) cues, on the basis of correlation, the composite trace. Processing of the new stimuli (premiership information) occurs within the context of the composite trace. Expecting that ‘team’ information is the basis of the cued (correlated) recall in this example, ‘noise’ exists in the form of information related to the sportsperson, and their off-field behaviour. These theories of noise suggest that it is the context of the retrieval (or reception) process that distinguishes which piece of information is signal (message) from that which is noise (Eich, 1985; Shannon, 1948).

The danger here is in conflating the noise in individuals’ information and memory processing, with news-media environmental, or supply-side, noise. Noise exists in both contexts. Consumers of the news are exposed to many discrete pieces of information that interact with their composite memory traces. At the news-media

170

supply-side for news content creation, journalists and editors have to process various news events and make choices about “what is news?” In processing alternatives, some output is noise.

In the current context, the concept of ‘noise’ explains that the associative environment for news consumers of NRL reporting is busy. There are many possible associations that readers may hold about the NRL in memory within composite traces. The processing of news-media reports as new stimuli will interact with those composite traces with a lot of competition. While this issue does not contribute to the ‘list of behaviours’ mandate for this content analysis, it has implications for information processing in Study 2. It adds further weight to concern that the already mediated sponsorship environment is busy with competing messages.

Explaining over-representation: The Courts, Hype, and News-values

The final issue relates to the idea of ‘hype’ or sensationalism in reporting. The issue of hype in reporting is raised on the basis of two pieces of information generated by the content analysis. These are: 1. Tests of difference of ‘number of articles’ and ‘number of incidents reported’ – the number of incidents reported is statistically significantly higher than the number of articles referring to off-field incidents; where off-field incidents are reported, it is likely that more than one incident will be discussed. 2. Age of incidents reported – articles published in 2009 referred to incidents occurring in six of the seven previous years. It might be reasonable to expect incidents that occurred late in 2008 to be reported in early 2009; the reporting of older incidents have less obvious motivations.

New data is introduced here to consider whether the level of reporting evidenced in this content analysis can be assessed as ‘hype’ in reporting. Table 4.13 records the number of off-field incidents involving NRL players that could be categorised according to the Australian Standard Offence Classification (ASOC). The table also records the relative proportion of those offences tried in Magistrate’s courts that occur in the Australian adult male population.

171

Table 4.22: Representation: Court offences & NRL off-field incidents Criminal offences Magistrates’ Higher NRL incidents reported in (finalised): Courts Courts The Courier Mail 2008-2009 (male data only) (all data) All incidents 2009 only N=492,197 N=52 N=34 Traffic offences 209,146 15 8 3 . % of incidents 42.5% 15.4% 8.8% Public order offences 59,255 141 4 3 . % of incidents 12.0% 7.7% 8.8% Acts intended to cause 49,070 2,962 14 9 injury . % of incidents 9.9% 26.9% 26.5% Sexual Assault 4,265 1,928 7 2 . % of incidents 0.8% 13.5% 5.8% *data are drawn from the current content analysis and ABS (2008-09b) catalogue number 4513.0

What is clear is that both acts intended to cause injury (assault) and sexual assault offences appear in the NRL reports above the level expected according to criminal prosecutions. This occurs in both the column that records events occurring in all years, and the column recording 2009 events only. The relative over-representation of assault and sexual assault in news’ reports is possible evidence of ‘hype’. Although it is not tested here, it should not be assumed that NRL players are more likely than other men in the general population to engage in the reported behaviours. For that argument to be reasonable; it would also be reasonable also to expect that NRL players are less likely than the general population of adult men to engage in public order or traffic offences.

These over- and under-representations are apparently consistent with the logic of judicial processes, and probably also consistent with individuals’ perceptions of the behaviours involved. This contention is made on the basis of the role of different levels of court/ prosecution. Magistrate’s courts are “the lowest level of criminal court” (ABS, 2008-09b, p5). The ‘Higher Courts’ label in Table 4.13 records data from the Supreme Courts and Intermediate Courts. These courts deal with “the most serious criminal matters…These include offences such as murder, manslaughter and drug trafficking as well as serious sexual offences, robberies and assaults” (ABS, 2008-09b, p5). Calculating from the data in Table 4.13: the percentage of offences finalised by type of court reveal that in 2008-09 the Higher Courts ruled on: 0.007% of traffic offences; 0.2% of public order offences; 5.7% of acts intended to cause

172

injury; and, 31.1% of sexual assaults. The seriousness of the offence, shown in the delegation of cases to Higher Courts, is reflected in the magnification of those offences in news-media reporting.

Hype and news values explain discrepancies in levels of behaviour reviewed by the courts, and levels of reporting the same behaviours in the news-media. ‘Hype’ is used with largely pejorative implication in relation to news-media reporting, it is not commonly used within the discipline. Vasterman suggests that ‘hype’ is not a popular word because of its “implicit value judgments” (2005, p508). Hype is seen as creating misunderstandings among readers, including distorted perceptions of issue risk or prevalence (Blood & Holland, 2004; Petersen, Anderson, Allan & Wilkinson, 2009). It is also said to affect the quality, and perceptions of quality and ethical standards of reporting (Meijer, 2003). Alternatively, hype is thought to provide a competitive benefit to journalists trying to get stories published, and occasionally brings beneficial attention to neglected issues (Petersen, Anderson, Allan & Wilkinson, 2009).

Vasterman suggests that ‘media-hype’ is an appropriate term when discussing the amplification or magnification of a news issue (Vasterman, 2005, p511). Magnification might be understood as representing the situation where “the media go into one case in depth, reporting every detail” (Vasterman, 2005, p516). Amplification, on the other hand, occurs when “they widen the ground they cover by reporting all kinds of events under the umbrella of the same news theme” (Vasterman, 2005, p516). Hype, then, appears to serve journalistic purposes by providing larger or broader coverage consistent with audience characteristics.

Media-hype indicates the propensity of news-media industries to create news, rather than report it according to an agenda-setting program (see: Cohen 1963; Fortunato, 2008; Kiousis & Wu, 2008; McCombs et al, 1997). It does not answer the ‘why’ other than to imply market-forces or consumer demand. Nor does it suggest which issues are subject to hype. On these questions it is useful to refer to news values.

News-values, discussed in Chapter 1, were first argued by Galtung and Ruge (1965). They suggest that “since we cannot register everything, we have to select, and the

173

question is what will strike our attention” (Galtung & Ruge, 1965, p65). Galtung and Ruge (1965) provide a list of the values driving news’ story selection; modernized lists also exist (see: Harcup & O’Neill, 2001).

Beyond the nature of the sportsperson as celebrity, the off-field behaviours of NRL players fall within the categories of sex, crime and human interest. For example, the three most commonly reported behaviours in the current study are assault, of either men or women (n=13), sexual assault (n=7), and driving under the influence of alcohol (n=7). These behaviours are also easily categorised as ‘bad’ news. The positive off-field behaviours engaged in by NRL players (e.g., donations to charity, volunteering etc.) should be newsworthy on the basis of being “good” news. It is this conclusion that returns the discussion to the issue of hype. Assuming that the good and bad off-field behaviours of NRL players are of equal extremity; why then does it appear that ‘bad’ behaviours a) out-number the good, and b) are subject to processes of both magnification and amplification which are not obvious for good behaviours?

A possible explanation is that bad news is worse than good news is good. For example, Galtung and Ruge (1965) identify negativity as newsworthy but not positivity, which is included in modern lists. This would suggest that, for Galtung and Ruge (1965) negativity is always compelling, but good news is merely, sometimes newsworthy. The theory of the negativity effect supports this argument.

The negativity effect suggests that negatively valenced information is more influential in person perception tasks than positively valenced information of equal extremity (see: Birnbaum, 1972; Dreben, Fiske & Hastie, 1979; Martijn, et al, 1992; Skowronski & Carlston, 1987, 1989, 1992). In relation to perception and judgment of the NRL player occurring at the level of the news production, bad behaviour is likely to appear more newsworthy than good behaviour.

Galtung and Ruge’s (1965) unexpectedness and consonance criteria might also explain the newsworthiness of the off-field behaviours of NRL players. The unexpectedness criteria argues, “the more unexpected the signal, the more probable that it will be recorded as worth listening to” (1965, p65). The consonance criterion provides the alternative; that information that confirms the receivers’ expectations

174

will be regarded as newsworthy (Galtung & Ruge, 1965). This section argues that many possible explanations exist for the publication of reports on the off-field behaviours of NRL players. Hype, news-values, and consumer interest are all possible explanations of levels of reporting. That levels of reporting may objectively differ from levels of behaviour occurrence in the broader population should be the subject of future research.

Conclusions

This content analysis has identified off-field behaviours of NRL players reported in The Courier Mail. The majority of the behaviours identified are negative; with only five incidents providing positive or altruistic behaviours.

In the discussion; noise, hype and news values were used to explain why off-field behaviours of sportspeople, particularly negative off-field behaviours, are regarded as newsworthy. Data from the content analysis, with data from the ABS on court cases finalised (ABS, 2008-09b) suggests that over-reporting of negative incidents might occur. Opportunities for further research exist to explain whether true differences exist in levels of behaviour occurrence in professional athlete and the broader population.

In the following pretest, measures consumer perceptions assess the valence of off- field behaviours identified in this content analysis.

175

Chapter 5: Pretests of Variables

Two pretests are reported in this chapter. These pretests contribute to the final study by identifying an appropriate measure of behaviour valence, and a measure of identification with the sporting team.

The first pretest reports the testing of behaviour valence measures using behaviours identified in the content analysis of Study 1. This pretest addresses Research Question 2.

The second pretest reports testing of identification scale items and the development of an identification measure for use in Study 2. This pretest addresses Research Question 3.

The structure of reporting is outlined here: 1. Research question 2. Literature & Aims 3. Method 4. Findings 5. Discussion

176

Pretest 1: Valence of Behaviours

This pretest assesses individuals’ impressions of a selection of the off-field behaviours reported in the Study 1 content analysis. Two measures of behaviour valence are assessed. The first is an ‘Attitude toward the Act’ measure. The second is a cue diagnosticity measure.

The test contributes to Study 2 by providing information on the perceived valence of depicted behaviours and addressing the methodological question of scale measure suitability. It also has implications to the design of Study 2 in identifying scenarios that would enable a balanced design.

The research question for this pretest is: RQ2. Removed from news-media and sports contexts, how do individuals evaluate the off-field behaviours reported?

The pretest uses surveys to obtain ratings of behaviours. The output is descriptive.

Literature & Aims

Two measures are tested in this pre-test. These are a cue diagnosticity measure (Martijn et al, 1992; Skowronski & Carlston, 1987, 1989, 1992), and an ‘attitude toward the act’ measure (Ajzen & Fishbein, 1970; Farah & Newman, 2010).

Cue diagnosticity

In person impression studies, trait measures have been used for two purposes: methodological tests of information integration and information weighting (Birnbaum, 1973, 1974), and the related task of assessing negativity and positivity effects of behavioural information on overall person impressions (Martijn et al, 1992; Skowronski and Carlston, 1987, 1989, 1992).

177

Skowronski and Carlston (1989) argue the superiority of the cue diagnosticity approach to trait assessment on several bases. These, related, bases include that the method is probabilistic, in being probabilistic it indicates the degree to which information assessed is useful for recognising a specific trait, this in turn suggests the impact of the cue on overall impression formation (Skowronski & Carlston, 1989). The authors also argue that “…extreme and negative behaviours are generally perceived as more diagnostic than are moderate or positive behaviours” (Skowronski & Carlston, 1989, p137).

The cue diagnosticity measure assesses the degree to which a personality trait is cued by the stimulus information provided. Skowronksi and Carlston (1987, 1989) and Martijn et al (1992) have used the measure to identify behaviours associated with ability/skill traits, and morality traits. Morality is a close fit for the subject matter studied here, which determines the Skowronksi and Carlston’s (1987, 1989) cue diagnosticity measure can be implemented without alteration.

Attitude toward the Act

Measures of Attitude toward the Act are popular in marketing studies (see: Meirick, 2002; Fitzsimons, Nunes & Williams, 2007; Mitchell & Olson, 1981); particularly those that have adopted all or part of Fishbein and Ajzen’s Theory of Planned Behaviour or Theory of Reasoned Action (see: Chiou, Huang & Lee, 2005; Kulviwat, Bruner & Al-Shuridah, 2009; Taylor, Ishida & Wallace, 2009). These models argue that various antecedent conditions (attitudes, subjective norms, perceived behavioural control) contribute to behavioural intention and later, behaviour (Ajzen & Fishbein, 1970). Attitude toward the Act, or A-act, is the “attitude toward performing a given behaviour in a given situation” (Ajzen, 1970, p467).

Attitude toward the Act is also popular in social marketing research which anticipates the outcome of behaviour-reduction and behaviour-encouragement campaigns (i.e., quit smoking, or increase exercise programs) based on attitudes and

178

perceived norms (see: Bai et al, 2010; Chiou, Huang & Lee, 2005; Zhao & Pechman, 2007). In each case, research has focused on the individual’s evaluation of the behaviour presented as a personal choice.

The current research project is differs from much of Attitude toward the Act research in that it asks respondents to evaluate a behaviour that is not presented as an activity the individual will personally have to choose or reject. It is effectively a measure of an ‘attitude toward others’ acts’. Dean (2002), in research which investigated perceptions of sponsorships, utilised a form of the Attitude toward the Act variable for the same purpose. That measure asked respondents to rate the altruism, generosity, unselfishness, and kindness of sponsors (Dean, 2002).

Scale characteristics & measurement

Two key issues distinguish the scales. These are the objects of measurement and the form of measurent.

Attitude toward the Act is conceived as an attitude toward undertaking a behaviour (Ajzen, 1970). It has been measured as an attitude toward the object, including advertisements (Meirick, 2002); attitude toward behaviour (Farah & Newman, 2010; Fitzsimons, Nunes & Williams, 2007), behaviour adoption (Kulviwat, Buriner & Al- Shuridah, 2009), or purchase (Mitchell & Olson, 1981). Its use, then, is diverse and objects of study are not limited to ‘acts’.

Trait measures (Skowronski & Carlston, 1987, 1989) focus on the measurement of individual’s perceived attributes. They have been employed more broadly to study trait transfer or inference in the case of the superstitious banana (Brown & Bassili, 2002), dog owners and their pet’s traits (Mae, McMorris & Hendry, 2004), and people acquiring the character of symbols (Carlston & Mae, 2007). Trait measures then, focus on whether the perceiver believes an object to possess an intrinsic characteristic; values of the perceiver are implicit in the judgment task. Attitude measures, alternatively, ask whether the objects assessed are good or bad, or likely to motivate action by the perceiver; values of the perceiver, in these task, are explicit.

179

The form of measurement used for attitude and trait measures also differs. Attitude toward the Act is predominantly measured using semantic differential scales, although likert-type scales are not uncommon. The majority of scales use 3 or 4 items with 7-point increments (Bruner & Hensel, 1992, p72). An overall score is produced from the sum of item ratings. A large variety of items have been used to measure the construct. Bruner and Hensel (1992) identify the following items: Foolish-Wise Worthless-Valuable Non-existent- Existent Safe-Risky Good-Bad Impossible-Possible Harmful-Beneficial Unlikely-Likely Try product- Not try product Pleasant-Unpleasant Improbably-Probable Influential-Not influential Useful-Useless Punishing-Rewarding Certain- Uncertain

Measurement of cue diagnosticity involves two items. These items ask how characteristic a scenario (behaviour) would be of a person with the target trait, and of a person without the trait. Data processing produces a probability that the scenario depicts trait behaviour, and an opposite probability that the scenario depicts non-trait behaviour.

The process for calculating trait diagnosticities is developed by Skowronski and Carlston (1987). The process, as outlined by those authors is provided in an example: …if the probability that a kind man will kick his cat is .4, and the probability that an unkind man will kick his cat is .8, then the diagnosticity of the behaviour kicked his cat for the trait category kind would be .33 (.4/[.4+.8]), and the diagnosticity of this behaviour for the category unkind would be .67 (.8/[.4+.8]). Skowronksi & Carlston, 1987, p693 This process produces a normalised joint probability. It does not present a probability of a single event occurring (trait presence 0/1), but a probability of trait presence, and contingent probability of trait absence. A similar process is used by Martijn et al (1992), however they do not normalise their scores, producing probabilities, that if viewed in tandem (both trait presence and trait absence), would sum to more than one.

180

Aims

The purpose of this pretest is to determine the range of perceptual valences associated with the behaviours identified in Study 1. Both the Attitude toward the Act measure and the cue diagnosticity measures can produce such lists. The pretest, then, has a secondary purpose of identifying the most suitable behaviour valence measure.

Method

In the Study 1 content analysis, 52 discrete off-field incidents were identified. Several issues militate against testing all incidents identified in Study 1. The possibility of respondent fatigue and the effects of fatigue on data quality is the most important of these issues. Sample wear-out, or the over-sampling of the undergraduate student population is another consideration.

A survey was used to test 14 behaviours on two variables: cue diagnosticity and Attitude toward the Act. This structure kept the survey form short in order to reduce effects of fatigue and administration time. Forms were distributed randomly in an undergraduate business class. Demographic characteristics (age and gender) were collected.

Items

Fourteen behaviours are tested in the pretest. All of the off-field incidents categorised as ‘positive or altruistic’ in Study 1 were included. All of the breaches of club rules were excluded from the study. A variety of incidents were drawn from the other categories. The list of behaviours tested in this pre-test is provided in Table 5.23.

The actors in the scenarios developed for this pre-test were not depicted as NRL players or as sportspeople. Representing the behaviours as those undertaken by ordinary members of the public aims to provide this test with measures of attitudes 181

that are not influenced by identification with sportspeople or sporting competitions. Items, as much as possible, inferred no gender to the actor.

Table 5.23: Behaviour valence: Scenarios from Study 1 Items Category Promotes a community health care program. Positive/altruistic Volunteers as a physiotherapist to help an injured teenager. Positive/altruistic Helped to raise money for a stranger with cancer. Positive/altruistic Auctioned memorabilia to raise money for charity. Positive/altruistic Saved their brother from drowning. Positive/altruistic Was accused of sexual intercourse without consent. Violence/Women Participated in group sex with friends and a waitress. Violence/Women Punched their girlfriend or boyfriend. Violence Was arrested for assault after fracturing a man’s skull. Violence Was fined for driving under the influence of alcohol. Anti-social Uses performance enhancing drugs. Anti-social Urinated in public. Anti-social Lost their driver’s license for repeated speeding offences. Anti-social Used cocaine. Anti-social

Scale measures

The current study presented its trait cue diagnosticity measure in the following format:

For the following statements, please indicate whether the behaviour is one you would expect of a moral person on the left, and an immoral person on the right.

Moral person Immoral person

Very unlik Very unlikely Moderately unlikely Slightly Neutral likely Slightly likely Moderately likely Very unlikely Very unlikely Moderately unlikely Slightly Neutral likely Slightly likely Moderately likely Very

ely

1 2 3 4 5 6 7 promotes a 1 2 3 4 5 6 7 community health-care program

182

The study presented its Attitude toward the Act measure in the following format:

1. I believe that to promote a community health care program is: Bad Good a. 1 2 3 4 5 6 7 Harmful Beneficial b. 1 2 3 4 5 6 7 Pleasant Unpleasant c. 1 2 3 4 5 6 7 Safe Unsafe d. 1 2 3 4 5 6 7

The scale measures used for both the variables provide 7 points of variation for behaviour ratings. The cue diagnosticity measure uses a Likert-type scale used to measure responses, these are anchored 1=very unlikely, and 7=very likely to represent the trait. The Attitude toward the Act measure also uses a semantic differential scale with 7 points.

Findings

Fifty-one responses were collected. Demographic characteristics for the sample reveal that 52.9% of respondents are female and 37.2% are male (5 missing values). The average age of respondents is 21.9 with a standard deviation of 4.54 (6 missing values).

Findings are provided first for the cue diagnosticity measure, followed by reporting of the findings for the Attitude toward the Act measure.

Moral Cue Diagnosticity

Table 5.24 shows the moral scores generated by pretest data. Diagnosticities were calculated for each respondent; the average diagnosticity for the sample is provided in the table. The moral score is generated by multiplying the item moral diagnosticity by the scale anchor 7=very likely. Items are ranked according to their moral score.

183

Table 5.24: Scenario: Moral Cue Diagnosticities and Scores Diagnosticities Moral Item score Moral Immoral Promotes a community health care program. .771 .229 5.40 Helped to raise money for a stranger with cancer. .768 .232 5.37 Auctioned memorabilia to raise money for charity. .760 .240 5.34 Volunteers as a physiotherapist to aid an injured teenager. .743 .257 5.20 Saved their brother from drowning. .680 .320 4.76 Used cocaine. .310 .690 2.19 Was fined for driving under the influence of alcohol. .310 .690 2.17 Lost his driver’s license for repeated speeding offences. .290 .710 2.02 Uses performance enhancing drugs. .281 .719 1.97 Urinated in public. .270 .730 1.91 Participated in group sex with friends and a waitress. .270 .730 1.87 Was accused of sexual intercourse without consent. .260 .740 1.81 Was arrested for assault after fracturing a man’s skull. .235 .765 1.64 Punched their girlfriend or boyfriend. .200 .800 1.40

The results of the ranking show that the ‘positive or altruistic’ behaviours identified in Study 1 are regarded as most indicative of moral behaviour. Scenarios categorized as anti-social behaviours in Study 1, involving potential harm follow (i.e., DUI, speeding, cocaine use). Finally, scenarios that reflect actual harm and violent behaviour are ranked lowest. The scenario scores are polarised, with no scenario providing a neutral diagnosticity.

A repeated measures ANOVA was run to establish whether differences apparent in the normalised moral scores (Table 5.25), are significant.

Table 5.25: Repeated measures ANOVA: Differences in Diagnosticity Effect Value F Hypothesis Error df P df Moral trait Pillai’s Trace 0.930 21.513 13 21 .000 diagnosticites Wilks’ Lambda 0.070 21.513 13 21 .000 Hotelling’s Trace 13.317 21.513 13 21 .000 Roy’s Largest 13.317 21.513 13 21 .000 Root

The repeated measures ANOVA found significant differences in the test of the scenarios using diagnosticities. Pairwise comparisons are tabled (Table 5.26) below.

184

Table 5.26: Pairwise comparisons of Moral Cue Diagnosticities 1 2 3 4 5 6 7 8 9 10 11 12 13 Raise money for a stranger with cancer. (1) Promotes community health care program. (2) .480 Auctioned memorabilia for charity. (3) .714 .234 Saved their brother from drowning. (4) .042 .006 .019 Volunteers as a physiotherapist. (5) .841 .244 .796 .050 Participated in group sex. (6) .000 .000 .000 .000 .000 Uses performance enhancing drugs. (7) .000 .000 .000 .000 .000 .971 Repeated speeding offences. (8) .000 .000 .000 .000 .000 .118 .491 Urinated in public. (9) .000 .000 .000 .000 .000 .473 .767 .277 Used cocaine. (10) .000 .000 .000 .000 .000 .161 .383 .711 .346 Was fined for DUI . (11) .000 .000 .000 .000 .000 .560 .411 .966 .716 .786 Accused of intercourse without consent. (12) .000 .000 .000 .000 .000 .788 .921 .086 .428 .156 .417 Punched their girlfriend or boyfriend. (13) .000 .000 .000 .000 .000 .112 .065 .005 .029 .007 .001 .118 Was arrested for assault. (14) .000 .000 .000 .000 .000 .551 .496 .106 .330 .090 .072 .640 .014

185

Shaded areas have been used to distinguish between three groups of results apparent in the pairwise comparisons. The first group indicates that among the moral scenarios (1-5) there are few significant differences. The significant differences in this group all relate to the “saved their brother from drowning” scenario.

The second group (lower left) indicates that significant differences exist between the all of the moral scenarios and all of the less moral scenarios (6-14).

The final group (bottom right) indicates that the majority of pairwise comparisons among the less moral scenarios show no significant differences. Exceptions to this are provided by the “punched their girlfriend or boyfriend” scenario.

Two conclusions may be drawn from the pairwise comparisons. The first is that the scenarios (save a brother from drowning, and; punched a girlfriend or boyfriend) that deviated from expectations (showed significant differences within their cohort) share a common characteristic. These scenarios are the only scenarios that refer to a close interpersonal relationship. All of the other scenarios provide more abstract references to third parties, with no implication of social or familial bonds.

The second conclusion relates to the moral diagnosticity measure in relation to the scenarios tested. Three possibilities exist here; the first is that the moral diagnosticity measure does not provide a sensitive measure to differentiate within trait cohorts. The second is that respondents are sensitive to nuance; DUI (potential harm) is significantly different to ‘accused of sex without consent’ (actual harm, if a genuine accusation). The final possibility is that there really are no differences in the perceptions of the morality of these behaviours.

In the following section, results of the Attitude toward the Act measure are provided.

Attitude toward the Act

Results for the Attitude toward the Act measure are provided in two parts. The first provides the factor analyses that test the items used to measure the construct. The

186

second section provides the analysis of mean differences between the scenarios, using a repeated-measures ANOVA.

Attitude toward the Act: factor analyses

The exploratory factor analyses presented below were generated using a principal components analysis, the solutions are un-rotated. Using a varimax rotation only one change would be made among the factor solutions; the ‘repeated speeding offences’ scenario produces a two-factor solution when rotated.

The un-rotated solution is favoured in this situation as the summated data will be normalised in order to compare the Moral Trait Perception data with the Attitude toward the Act data on a common scale. These comparisons will occur both at the level of comparing ANOVA results, and when testing the possibility of collinearity. Normalising the summated scores after conducting the factor analyses prevents the possibility of normalising the data twice, while creating a data set measured on the same scale (probabilities) for comparison.

The result of the factor analyses suggests three groups of scenarios according to variations in their factor structures. Differing factor structures across the scenarios were not anticipated; this creates an issue for Study 2 design as survey design and analysis will require that scenarios should have similar factor structures.

The first group contains five scenarios which are structurally consistent. Each of the scenarios has an improved, or equivalent, Cronbach’s alpha if the pleasant/ unpleasant item is removed, although factor loadings are acceptable without removals. They all provide single factor solutions, and there is a scenario that represents each of the categories of off-field incidents. The factor analysis results for this group are provided in Table 5.27. The improved alphas are provided in brackets. On the basis of their similarities, this group is regarded as the most appropriate for use in Study 2.

187

Table 5.27: Behaviour EFA: Common Structue Scenarios improved by removing item: Pleasant/ The second group retains all four items tested. Unpleasant This group contains five of the 14 scenarios Factor 1 DUI: anti-social tested. These scenarios produce either single- Bad/Good -0.855 or two-factor solutions. Cronbach’s alpha’s are Harmful/Beneficial -0.851 Pleasant/Unpleasant -0.698 not improved if any of the items are removed, Safe/Unsafe -0.825 and overall, alpha’s are high (ranging from .70 Variance explained 0.656 Cronbach’s alpha .771 (.803) to .86). This group does not have a Cocaine: anti-social representative scenario from each of the off- Bad/Good -0.853 Harmful/Beneficial -0.864 field categories. It is also complicated by the Pleasant/Unpleasant -0.766 moral scenario which is the only scenario to Safe/Unsafe -0.920 Variance explained 0.727 produce a two-factor solution in this group. Cronbach’s alpha .793 (.849) The factor structures for this group are Assault: violence Bad/Good -0.738 provided in the appendix for this chapter Harmful/Beneficial -0.915 (Appendix 5.1: Attitude toward the Act factor Pleasant/Unpleasant -0.772 Safe/Unsafe -0.949 analyses). Variance explained 0.720

Cronbach’s alpha .834 (.849) Group sex: women The final group of scenarios contains four Bad/Good -0.966 relatively inconsistent item analyses. Each of Harmful/Beneficial -0.893 Pleasant/Unpleasant -0.890 the scenarios has an improved Cronbach’s Safe/Unsafe -0.928 alpha if one of the items is removed. The Variance explained 0.847 Cronbach’s alpha .935 (.935) ‘auctions memorabilia for charity’ scenario is Cancer improved if the harmful/beneficial item is fundraising: moral Bad/Good -0.880 removed. Harmful/Beneficial -0.878 Pleasant/Unpleasant -0.729 Safe/Unsafe -0.836 The remaining three scenarios (punching a Variance explained 0.694 girlfriend or boyfriend, providing volunteer Cronbach’s alpha .845 (.853) physiotherapy, and saving one’s brother from drowning) improve their Cronbach’s alpha if the bad/good item is removed. Both of the moral scenarios produce two-factor solutions with weak reliability scores. Removing the bad/good item from the moral scenarios is unlikely to simplify their

188

factor structures. This group does not contain a scenario from each category of off- field incidents, and the reliability results suggest that the moral scenarios are not well accounted for by the items used. The factor structures for this group are also provided in the appendix for this chapter.

Attitude toward the Act: mean differences

In this section descriptive statistics are provided for each of the scenarios. The factor structures identified in the previous section were applied to create the summated attitude scores. Table 5.28 below provides means, medians, standard deviation and mean normalised score for each scenario.

Table 5.28: Scenarios: A-act descriptive statistics & scores Mean Std. Item N Median Mean normalised dev. score Helped to raise money for a stranger with cancer. 49 6.33 5.98 1.15 .84 Promotes a community health care program. 48 6 5.95 0.94 .82 Auctioned memorabilia to raise money for 40 6 5.38 1.6 .71 charity. Saved their brother from drowning. 39 5 5.21 1.37 .70 Volunteers as a physiotherapist to help an injured 49 5 5.13 1.22 .70 teenager. Participated in group sex with friends and a 42 2.2 2.5 1.57 .36 waitress. Uses performance enhancing drugs. 49 2.5 2.35 1.16 .33 Lost their driver’s license for repeated speeding 43 2 2.19 1.24 .31 offences. Urinated in public. 43 2 2.1 1.32 .30 Used cocaine. 43 1 1.41 0.87 .20 Was fined for driving under the influence of 49 1 1.37 0.85 .19 alcohol. Was accused of sexual intercourse without 43 1 1.36 0.91 .19 consent. Punched their girlfriend or boyfriend. 49 1 1.33 1.04 .19 Was arrested for assault after fracturing a man’s 49 1 1.14 0.49 .16 skull.

The descriptive statistics reveal that the scenarios tested are polarised, with no neutral examples.

To test whether the differences in means across scenarios can be assumed, a repeated

189

measures ANOVA was conducted. The results of this test are shown in Table 5.29.

Table 5.29: Repeated measures ANOVA: differences in A-act Effect Value F Hypothesis Error df P df Attitude Pillai’s Trace .957 49.765 13 29 .000 toward Wilks’ Lambda .043 49.765 13 29 .000 the Act Hotelling’s Trace 22.308 49.765 13 29 .000 Roy’s Largest 22.308 49.765 13 29 .000 Root

The result of the repeated measures ANOVA suggest that significant differences were found between the scenario means. Pairwise comparisons among the scenarios are provided in Table 5.30 on the following page. The three shaded areas indicate three levels of means, i.e., mean values of 5/6, 2, and 1.

The first shaded group shows the comparison of the positive/ altruistic scenarios. No significant differences are found among scenarios 1, 2, and 3; significant differences are found between that group and scenarios 4 and 5. These differences are consistent with the mean (and median) differences shown in Table 5.28 which shows moral scenarios with means of 5 and those with means of 6. The differences might otherwise reflect the lower levels of item reliabilities found in the factor analyses (drowning = .544; physiotherapy = .601).

The second group of scenarios (scenarios 6-9); earlier shown to have mean and median scores of approximately 2, show no significant differences within the group.

The majority of pairwise comparisons in the third shaded group show no significant differences. The sole exception to this is the comparison of cocaine use and assault scenarios, which test as significantly different. By mean ranking cocaine use has the highest mean in this group; the assault scenario has the lowest rank. The difference between these scenarios provides support for the idea that mean gradients are distinguished by the ANOVA test.

190

Table 5.30: Pairwise comparisons of A-act scores 1 2 3 4 5 6 7 8 9 10 11 12 13 Raise money for a stranger with cancer. (1) Promotes community health care program. (2) .758 Auctioned memorabilia for charity. (3) .104 .139 Saved their brother from drowning. (4) .026 .039 .470 Volunteers as a physiotherapist. (5) .001 .002 .842 .669 Participated in group sex. (6) .000 .000 .000 .000 .000 Uses performance enhancing drugs. (7) .000 .000 .000 .000 .000 .461 Repeated speeding offences. (8) .000 .000 .000 .000 .000 .269 .976 Urinated in public. (9) .000 .000 .000 .000 .000 .085 .785 .581 Used cocaine. (10) .000 .000 .000 .000 .000 .000 .001 .000 .000 Was fined for DUI . (11) .000 .000 .000 .000 .000 .000 .000 .000 .001 .718 Accused of intercourse without consent. (12) .000 .000 .000 .000 .000 .000 .001 .000 .001 .755 .993 Punched their girlfriend or boyfriend. (13) .000 .000 .000 .000 .000 .001 .002 .001 .005 .663 .830 .829 Was arrested for assault. (14) .000 .000 .000 .000 .000 .000 .000 .000 .000 .036 .115 .093 .263

191

Discussion

The research question for this pretest asked: Removed from news-media formats (newspaper articles) and the sports context (no mention of athletes), how do individuals evaluate the off-field behaviours reported?

Before the research question is discussed, the implications of use of each of the measures in Study 2 are considered. The cue diagnosticity measure, in particular, is regarded as problematic.

Collinearity & Cueing

Underlying the tests of variables are sub-questions: 1. Might the measures be collinear? 2. Could both variables be used in research?

Moral theorists argue that it is possible to distinguish between between believing that an act is ‘right’ and desiring to undertake that act. The principles involved in this argument follow: (1) Moral judgements like ‘it is right that I Φ’ express beliefs. (2) There is a necessary connection between being in the state that they judgement ‘it is right that I Φ’ expresses and having a motivating reason to Φ. (3) Motivating reasons are constituted, inter alia, by desires. The Moral Problem is that the following principle is also plausible: (4) There is no necessary connection between believing ‘it is right that I Φ’ and desiring to Φ. Ferrari & Reaber, 2009, p1 Using the above logic, the moral cue diagnosticity can be seen as judgement of the form, ‘it is right that I Φ’. The Attitude toward the Act measure, as noted in the literature for this study, is more commonly used to measure respondents’ attitude toward behaviour adoption, or, the ‘motivating reason to Φ’. While there is a connection between the two ideas, they are distinguishable.

192

Although logic argues against collinearity of the measures, data may argue differently. Collinearity and singularity can be recognised according to a variety of related measures; simple correlation, squared multiple correlation (SMC), variance inflation factor (VIF) scores, tolerance scores, and comparison of variance proportions (see: Hair et al, 1995; Tabachnick & Fidell, 2007). The simplest measure of assessing collinearity among the measures is through a review of correlations. According to Tabachnick and Fidell (2007), correlations of .90 and above indicate collinearity likely to affect analyses.

Table 5.31: Pearson correlation: Cue Diagnosticity & A-act Pearson’s P n correlation Raise money for a stranger with cancer. .184 .220 46 Promotes community health care program. .183 .223 46 Auctioned memorabilia for charity. .174 .276 41 Saved their brother from drowning. .011 .944 41 Volunteers as a physiotherapist. .403 .006** 45 Participated in group sex. .253 .116 40 Uses performance enhancing drugs. .404 .005** 46 Repeated speeding offences. .079 .615 43 Urinated in public. .131 .402 43 Used cocaine. .312 .044* 42 Was fined for DUI. .289 .057 46 Accused of intercourse without consent. .070 .655 43 Punched their girlfriend or boyfriend. .017 .909 46 Was arrested for assault. .301 .042* 46 * Correlation is significant at the 0.05 level (2-tailed) ** Correlation is significant at the 0.01 level (2-tailed)

The .90 correlation threshold recommended by Tabachnick and Fidell (2007) has not been reached by any of the scenarios tested. Several scenarios show significant p- values, indicating a strong relationship between the measures. Overall, however, there is little indication of a collinearity problem, and no likelihood of singularity in the measures.

Cue: duty

As the measures are not collinear, the issue of whether both measures can or should be used becomes pertinent. The cue diagnosticity measure provided a different

193

overview of the scenarios to that provided by the Attitude toward the Act measure in the repeated measures ANOVAs. Where the Attitude toward the Act pairwise comparisons showed differentiation among three groups of scenarios, the cue diagnosticity measure identified only two groups clearly. These groups differentiate the behaviours that are obviously positive, from those that are clearly negative.

Two scenarios did not conform to the positive/negative polarization in the cue diagnosticity tests: the ‘saved their brother from drowning’ and the ‘punched a girlfriend or boyfriend’ scenarios. These were the only scenarios to suggest a close social or familial relationship. The results suggest that morality, as tested, is sensitive not only to the ideas of good and bad, but also to levels of obligation developed through relationships. This idea, while interesting, is not new. Morality is conceptualised as a social phenomenon (see: Chiu, Dweck, Tong, & Fu, 1997; Copp, 1997; Ellemers, Pagliaro, Barreto, & Leach, 2008; Haidt & Graham, 2007; Janoff- Bulman, Sheikh & Hepp, 2009; Gray & Wenger, 2009), and all moral theories deal with the obligations of the individual to others, or to the reasoned principles by which a person may live in society.

The cases involving a sibling, a girlfriend or boyfriend imply a ‘special relationship’. These, according to Yeager, create a “duty-to-rescue” (1993, p8). The author continues, “...special relationships... should alert both the potential rescuer and the imperiled to the imperiled’s right to be rescued by that rescuer” (Yeager, 1993, p10). This duty might equally be expected to involve refraining from causing harm. Such special relationships include those of parent and child, husband and wife, or master and servant; obligations to strangers are, by contrast, imperfect (Yeager, 1993). According to Jacquette, “usually there is a ranking of obligations according to a hierarchy that supports a resolution of conflicts of duty” (1991, p43), favouring family or in-group members supports these assumptions (Leach, Ellemers & Barreto, 2007).

Overall, a measure of moral cue diagnosticity, in this research context, raises additional questions. It suggests that not only a measure of cue diagnosticity should be used, but also a measure of expected duties, or normative obligation. Further, in its inclusion of the words, ‘moral’ and ‘immoral’, it raises the possibility of cued or

194

biased evaluations toward a level of social compliance. On its own, or even in combination with the Attitude toward the Act, cue diagnosticity is not capable of telling a ‘whole’ story. It is likely to be associated with high levels of unexplained variance.

The advantage of Attitude toward the Act measure is in its transparency. The components of the scale reflect conceptualisation; they also facilitate post hoc explanation. The measure does not ask the complicated and nuanced question, ‘it is right?’ (Ferrari & Reaber, 2009). It asks the easier question, ‘how do I feel about it?’ These differences establish Attitude toward the Act as a more manageable measure for use in Study 2.

The research question

The goal of this pretest is to establish the evaluative valence of behaviours identified in Study 1. Two answers to the question are provided: the first is the range of moral cue diagnosticities developed; the second, the range of Attitude toward the Act scores. The conclusion of the previous section argues that, for this research, moral cue diagnosticity is a complicated, and therefore inappropriate, measure. The range of A-act scores, then, provides the answer to the research question for this study.

The factor analyses conducted on A-act scales serve to reduce the list of behaviours tested, according to their utility in Study 2. Only five scenarios are testable using the same combination of items, with a similar level of reliability. This group includes the scenarios:  Raise money for a stranger with cancer  Participated in group sex  Used cocaine  Was fined for DUI  Was arrested for assault

The relationships between scenarios, according to the pairwise comparisons, are represented in Figure 5.13 as a Venn diagram. Their means, or evaluative valences,

195

which answer the research question for this pretest, are provided in the diagram.

Figure 5.13: A-act relationships between scenarios

Cancer Group Cocaine DUI Assault fund- sex use raising mean=2.5 mean=1.4 mean=1.3 mean=1.1 mean=5.9

The diagram shows, also, that no statistically significant difference exists among the cocaine use, DUI and assault scenarios according to the A-act measure. This result argues that a balanced experimental design on the basis of A-act valences is unlikely.

Conclusions

This pretest has raised and answered several questions. Among the conclusions are: 1. The moral cue diagnosticity produces quite different evaluative outcomes to the A-act measure tested. The moral diagnosticity measure itself, is problematic in the current context. 2. Not all scenarios identified, according to exploratory factor analyses conducted, are testabled using the same combination of A-act items. 3. Attitude toward the Act has identified five scenarios, which are testable using the same items. These scenarios provide an initial list from which development of Study 2 scenarios can begin. The pre-test has also identified the need for future conceptual work and research to confirm or elaborate on the results found in the testing of the moral cue diagnosticity measure.

196

Pretest: Team Identification

Identification explains the individuals’ sense of belonging to a group. It is also responsible for inter-group bias and intra-group favouritism. The issue, for this pretest is how identification as a construct for measurement can be studied.

The research question for this pretest is: RQ3. If group (team) identification has the power to influence consumer evaluations, how should team identification be measured for this research?

Literature & Aims

Identification has been widely studied and numerous scales have been developed in order to measure the construct. Eighteen identification scales, published between 1976 and 2007 have been identified. Figure 5.14 shows a genealogy of those scales. Only scale development publications are depicted in the figure.

In identifying the most appropriate scale, four key issues are considered. These relate to the conceptualisation of identification, hypothesised dimensional structure, items for measurement, and context of the research. Identification conceptualisation, and dimensional structure of identification have been discussed in the literature review of Chapter 2; those issues are summarised here briefly. Extant scale items are discussed here, followed by a review of existing sports team identification scales.

Conceptualisation

The social identity theory of identification is a social-psychological theory which explains the individual as a group member. Social identity is said to be a theory that explains intergroup conflict (Tajfel & Turner, 1979). It is otherwise used to explain the desire of the individual to develop or maintain a positive social identity (Ellemers et al, 1988).

197

Figure 5.14: Genealogy of Identification scales

Early research Rosenberg 1965 Driedger 1976

The boom Brown & Williams 1984

Brown et al Gurin & 1986 Townsend 1986 Kelly 1988

Ashforth & Hinkle et al 1989 Mael 1989 Crocker & Luhtanen 1990

Karasawa 1991 Mael & Mael & Luhtanen & Phinney Hirt et al Wann & Ashforth Tetrick Crocker 1992 1992 1992 1992 1992 Branscombe Ellemers 1993 1993

Sellers et al 1997 Henry et al Jackson & Ellemers et al 1999 1999 Smith, 1999 Cassidy & Underwood, Bond Trew 2001 & Baer, 2001 Jackson 2002 Ashmore et al 2004 Dimmock et al 2005 The period of Heere & refining measures James 2007

Social Identity Theory ______*bold text indicates scales Identity Theory ______*italic text indicates literature only Other theories ………………….

198

Antecedents of identification include group characteristics, such as distinctiveness and prestige (Ashforth & Mael, 1989). Other antecedents are the characteristics “…associated with group formation (interpersonal interaction, similarity, proximity, shared goals or threat, common history and so forth…” (Ashforth & Mael, 1989, p25). For the individual, the only required condition for identification to occur is categorisation (Ashforth & Mael, 1989). This is evidenced by minimal group research in which respondents, assigned to groups on the basis of real or fictitious characteristics, produce identification outcomes (see: Billig & Tajfel, 1973; Tajfel & Billig, 1974).

Consequences of identification are myriad. They include effects on the individual’s mood or emotions (Wann, 2006; Wann & Branscombe, 1992); and behavioural outcomes such as absenteeism, turnover, and resource allocation (Grieve & Hogg, 1999; Hogg & Turner, 1985; Turner, Sachdev & Hogg, 1983). Consequences of identification also include group perceptual outcomes such as attitudinal biases and favouritism, use of stereotypes and group-based attributions (see: Dimmock, Grove & Eklund, 2005; Eccleston & Major, 2006; Fielding, Hogg & Annandale, 2006; Hains, Hogg & Duck, 1997).

Identification is defined by its principal theorist, Henri Tajfel, as: …that part of an individual’s self-concept which derives from his knowledge of his membership of a social group (or groups) together with the value and emotional significance attached to that membership. Tajfel, 1978, p63 This definition is widely cited and used by identification researchers, and scale developers in particular (see: Brown et al, 1986; Dimmock et al, 2005; Ellemers et al, 1999; Gurin & Townsend, 1986; Heere & James, 2007; Henry et al, 1999; Hinkle et al, 1989; Karasawa, 1991; Jackson, 2002)

Dimensionality

Tajfel’s definition is important not just for its contribution to identification conceptualisation, but also its impact on the scale dimensionality literature. Of the

199

eighteen identification scales identified, seven state specifically that Tajfel’s (1978) definition determines the number and form of dimensions hypothesised in their scale development (see: Brown et al, 1986; Dimmock et al, 2005; Ellemers et al, 1999; Henry et al, 1999; Hinkle et al, 1989; Jackson, 2002; Karasawa, 1991).

The dimensions hypothesised are usually referred to as: a) knowledge of group membership (cognitive component); b) value of membership (evaluative component); and c) emotions related to group membership (affect).

Despite the popularity of these dimensions, there is no evidence Tajfel ever intended his definition to contribute to scale development or a dimensionality debate. Tajfel’s own research was predominantly experimental, using the minimal group approach (see: Billig & Tajfel, 1973; Tajfel & Billig, 1974). Further, key co-authors have been critical of identification scales in their use as individual difference variables and other failures of conceptualisation (see: Turner, 1999, p20-22).

Although there are numerous extant identification scales, this review recognises little consensus as to a ‘correct’ approach to measurement, and much opportunity for refinement and re-development.

Scale items

When all of the items developed for all of the identification scales are studied together there are roughly 130 unique items. Some of these items occur in as many as eight different scales, although the mode, and median, remains one. Approximately 35% of items have been used in more than one scale. This collection of numbers, in combination with the genealogy outlined in Figure 5.14 argues substantial conceptual agreement and commonality across scales.

The approach taken in this research is that scale items should reflect identification conceptualisation, explaining the knowledge and perceptions of the individual as a group member. Items and dimensions that specifically seek to measure behaviours are not regarded as appropriate identification scale items; this is not solely because

200

behaviours are conceived as identification consequences (Ashforth & Mael, 1989). Other identification consequences (emotions and attitudes) often contribute to item development. Behaviours are excluded here on the basis that the frequency or number of behaviours implies a longevity argument about group membership, where the identification of newer or less experienced group members should not necessarily be regarded as a lesser construct.

Scale context: Sports

Several team identification scales exist (see: Dimmock, Grove & Eklund, 2005; Heere & James, 2007; Hirt, Zillmann, Erickson & Kennedy, 1992; Wann & Branscombe, 1993) in a variety of forms.

Hirt et al (1992) use a five-item uni-dimensional measure of team identification that is not explicitly linked to theory or literature. The authors refer variously to “degree of interest in and fanship toward…” a team (Hirt et al, 1992, p727) when describing their measures, “identity importance” (Hirt et al, 1992, p729). No information is provided about scale validity or item reliabilities.

The Sport Spectator Identification Scale (SSIS) is produced by Wann & Branscombe (1993). SSIS items measure the importance of winning, degree of self-assessed fan level, degree to which friends see one as a fan, dislike of rival teams, display of merchandise and attention to broadcast media. It is a strong scale; the single factor accounts for more than 60% of total variance; and Cronbach's alpha for this scale is consistently above .90 (Wann & Branscombe, 1990, 1993, 1995).

The SSIS is uni-dimensional. It is also very popular (see: Grieve, Shoefelt, Wann & Zapalac, 2009; Theodorakis, Koustelios, Robinson & Barlas, 2009; Wann & Dolan, 1994; Wann, Dolan, McGeorge & Allison, 1994; Wann, Grieve, Waddill & Martin, 2008). The scale is linked to social identity theory, however the authors do not provide a substantial conceptual or scale development. Many of the items ask for behavioural indications of identification.

201

Dimmock et al (2005) and Heere and James (2007) have produced the two most recent team identification scales. The authors of these papers provide both theoretical background, and justification of item development or sourcing from other scales. Both of these scales are multi-dimensional.

Dimmock et al (2005) developed the Team Identification scale within the social identity framework. The conceptual development refers to Tajfel’s (1978) tripartite definition of identification, and scale items were drawn from early identification scales. In testing, the scale items produce a three-dimensional model which explains 51% of the variance Dimmock et al (2005). The factor structure does not conform completely to the affect-cognitive-evaluative model theorized, as affective and cognitive items load on the same factor. Dimmock et al (2005) argue that because team identification is a voluntary group membership, cognitive and affective components should be expected to load on the same factor on the basis of group solidarity. This argument, according to Heere and James (2007) illustrates a likely problem of discriminant validity among the dimensions.

The final team-based identification scale is the TEAM*ID scale developed by Heere and James (2007). The TEAM*ID scale is significantly different to the majority of identification scales. The differences derive from the diversity of sources used. Heere and James drawn from identity theory (see: Stryker & Burke, 2000), social identity theory (Tajfel, 1979), and ethnic identification (see: Ashmore, Deaux & McLaughlin-Volpe, 2004; Phinney, 1992; Sellers, Rowley, Chavous, Shelton & Smith, 1997). A 9-dimensional model is hypothesised. Testing reduced the scale to a 6-dimensional model comprised of 21 items. Heere and James recommend further testing of their scale, despite some positive results, as they feel that the measures “demonstrate considerable overlap” (Heere & James, 2007, p84).

Aims

Overall, the team identification scales vary in degree of theoretical development. Of the sports’ scales, two show substantial theoretical justification and testing (see: Dimmock et al, 2005; Heere & James, 2007), the remainder do not. Strong

202

theoretical development does not necessarily result in a cohesive scale. Nor does scale context make sense as a selection criterion when scale items do not reflect a unique understanding of ‘sports’ as a distinct example of identification.

The review of identification scale literature shows that the sports identification scales do not present an obvious, strong scale for use. It is also not clear that a tripartite dimensional structure is appropriate; and a specific dimensional structure of identification is not hypothesised at the beginning of this pretest. In determining an approach to identification measurement for this research, items from the most recent scales are selected for testing and scale development (see: Dimmock et al, 2005; Ellemers et al, 1999; Heere & James, 2007; Henry et al, 1997; Jackson, 2002; Sellers et al, 1997). Three scales from this period are judged unsuitable for refinement to a team identification measure; these are the scales of Sellers et al, (1997), Henry et al (1999), and Jackson (2002).

The Multidimensional Inventory of Black Identity (MIBI) scale of Sellers et al (1997) is not included in this research because the focus of that scale is specific to ethnicity. The ideology dimension of the MIBI, in particular, includes such constructs as the ‘oppressed minority’, ‘assimilation’, ‘nationalist’, and ‘humanist’ subscales (Sellers et al, 1997, p815), none of which can be easily adapted for use in a sport-team identification context.

The exclusions of the Henry et al (1999) and Jackson (2002) scales rare determined on the basis of the type of group the scales are designed to explain. Each of these scales measures the individual’s perception of direct group interaction, or a group- directed purpose that is not a requirement of psychological groups. These items and the theorising behind them are too restrictive to be applied to a psychological group. In the case of team identification, a person may feel like they have known a team since their childhood, or have watched many games and been invested in the outcomes of those games. However, that person need not attend home-games, link arms with a person wearing the same colours and chant the team song in order to establish their identification; it is sufficient that it identification is felt.

The remaining scales are those of Dimmock et al (2005), Ellemers et al (1999), and

203

Heere and James (2007). These scales have both strengths and weaknesses. The research of Dimmock et al (2005), Heere & James (2007), and Ellemers et al (1999) represents the research that has come before it through adoption of conceptual frameworks, and items. The approach recommended here is to use the items that have tested well in previous research (.70 factor loadings and above, following Nunnally & Bernstein, 1994) and to develop a strong model driven by the history of social identity theorising and subjects’ responses.

Method

Identification items were tested in a survey format administered to undergraduate student samples. Two tests of the identification items were conducted: the first test asked about identification with the Queensland team for the rugby league State of Origin competition; the second test asked about identification with the Brisbane Broncos team in the National Rugby League competition.

Items were measured using a 7-point Likert type scale. Values ranged from 7= strongly disagree to 1= strongly agree, with 4 as a neutral mid-point. Demographic characteristics (gender and age) were collected. Data were analysed using exploratory and confirmatory factor analyses.

Item selection

Identification items were drawn from three recent identification scales (see: Dimmock et al, 2005; Ellemers et al, 1999; Heere & James, 2007). Selected items achieved factor loadings above .70 in previous tests; 20 items met this criterion. These items are provided in Table 5.1, which shows the source publication, the dimension(s) the item has loaded on in the past, and the factor loading reported in each publication

204

Table 5.32: Identification: pretest items & origins Item Source Dimension Factor Loading The [team]’s successes are my successes. Dimmock et al, 2005 Affect .785 Heere & James, 2007 Interconnection of self .813 I have little respect for the [team]. Dimmock et al, 2005 Evaluation (personal) .765 Ellemers et al, 1999 Group self-esteem .77 Overall, the [team] are considered good by others. Dimmock et al, 2005 Evaluation (other) .796 Heere & James, 2007 Public Evaluation .800 I think of the [team] as part of who I am. Dimmock et al, 2005 Cognitive .693 I am proud to be a fan of the [team]. Heere & James, 2007 Private Evaluation .883 I think the [team] have little to be proud of. Dimmock et al, 2005 Evaluation (personal) .651 Ellemers et al, 1999 Group self-esteem .70 I identify with other fans of the [team]. Ellemers et al, 1999 Self-categorisation .80 I am like other fans of the [team]. Ellemers et al, 1999 Self-categorisation .82 I would like to continue support the [team]. Ellemers et al, 1999 Commitment to group .72 I dislike being a fan of the [team]. Ellemers et al, 1999 Commitment to group .76 I would rather be a fan of the [other team]. Ellemers et al, 1999 Commitment to group .88 Others describe me as a typical fan of the [team]. Dimmock et al, 2005 Cognitive .598 Heere & James, 2007 Self-categorisation .906 In general, I’m glad to be a fan of the [team]. Heere & James, 2007 Private evaluation .909 I have a strong sense of belonging to the [team]. Heere & James, 2007 Interconnection of self .829 I have a strong attachment to the [team]. Dimmock et al, 2005 Affect .566 Heere & James, 2007 Interconnection of self .874 When someone criticises the [team], it feels like a personal Heere & James, 2007 Interconnection of self .753 insult. I’m very interested in what others think about the [team]. Heere & James, 2007 Interconnection of self .822 I am aware of the tradition and history of the [team]. Heere & James, 2007 Cognitive awareness .817 I know the rituals that go with being a fan of the [team]. Heere & James, 2007 Cognitive awareness .835 I have knowledge of the successes and failures of the [team]. Heere & James, 2007 Cognitive awareness .866

205

The .70 level was selected as it provides a level of security when making judgments about the statistical significance of results. Spicer (2005) explains that a factor loading represents an item-to-factor correlation. Factor loadings of ± .30 are acceptable “mainly on the grounds that such a variable would be accounting for nearly 10% (.32) of the variance in a factor” (Spicer, 2005, p189). This minimum condition does not provide statistical significance for small samples. Whereas, factor loadings of .70 and above provide statistical significance with samples as small as 60 responses (Hair et al, 1995, p385).

Exploratory Factory Analyses

Exploratory factor analyses were conducted using principal components analysis with a varimax normalised rotation. Principal components analysis was used to ensure the extraction of uncorrelated factors (Spicer, 2005; Tabachnick & Fidell, 2007). The varimax rotation ensures the ease of interpretability through the “maximisation of variance within factors, across factors” (Tabachnick & Fidell, 2007, p638). These methods are regarded as most suitable for exploratory procedures using a relatively large number of variables.

Confirmatory Factory Analyses

Confirmatory factor analysis provides the means to test theorised relationships between observed and latent variables (Brown, 2000). It differs from exploratory factor analysis in that it makes “a priori specifications and restrictions” (Brown, 2000, p14) on the factor model.

The data are analysed using AMOS (Analysis of Moment Structures). Each of the observed variables is hypothesised to load on a single latent variable. Regression weights are constrained to one for a single observed variable per latent variable Regression weights for unique, or error, variances are also constrained to one.

206

Fit Indices

Several indices are used to evaluate model fit. The Chi-square statistic is used to evaluate a measure of absolute fit of the model to the data (Brown, 2000). The null hypothesis for this test is that the data matrix matches model specification (Brown, 2000). A significant Chi-square statistic will suggest that the data do not provide a good fit to the model. The Chi-square test, however, is often simultaneously thought to be too “stringent” (Brown, 2000, p81), and subject to producing inflated test statistics in the presence of larger sample sizes (>100), model misspecification, or non-normal data (Brown, 2000; Hu & Bentler, 1998).

The Root Mean Square Error of Approximation (RMSEA), Normed Fit Index (NFI), Comparative Fit Index (CFI), and Tucker-Lewis Index (TLI) are also reported in this research on the basis of providing less stringent measures of model fit.

The RMSEA (root mean square error of approximation) index, although based on the Chi-square statistic is grouped by Brown (2000) with other indices as an example of parsimony correction. The parsimony correction group of model fit indices, according to Brown (2000), favours those path models with fewer degrees of freedom. The RMSEA is also a population-based index, making it “less sensitive to distribution and sample size” (Hu & Bentler, 1998, p447). Although less stringent than the Chi-square test, several characteristics of this research militate against the RMSEA producing acceptable fit indices for this research.The number of observed variables in this research is larger than the norm; (median 9-10 observed variables) (Hu & Bentler, 1998, p431), and as the models are relatively unconstrained, the degrees of freedom are correspondingly large.

The RMSEA calculates a 90% confidence interval to estimate significance, although several methodologists recommend more stringent values (Brown, 2000). Hu and Bentler (1999) suggest that RMSEA of 0.0-0.05 indicates good fit, .06 or less is acceptable, but values approaching .10 should be used to reject the null hypothesis. Browne and Cudek (1993) suggest similar values; ≤ 0.05, acceptable, ≤ 0.08, adequate, and reject when RMSEA ≥ 0.10. Browne and Cudek (1993) also suggest the use of a closeness statistic (PCLOSE) which measures the significance of the

207

difference between the RMSEA value and .05; a significant statistic indicates that the RMSEA is not close to .05.

The final three fit indices that will be reported are the NFI, CFI and the TLI, all of which are comparative, or incremental fit indices (Brown, 2000). This type of model compares the specified path model with software-specified baseline models (Brown, 2000; Hu & Bentler, 1998) which suggest “no relationships among the variables” (Brown, 2000, p84). In practice, the baseline models used in AMOS specify that observed variables are uncorrelated, have equal correlations, and/or have means of zero (Arbuckle, 2007). Although comparative fit indices are sometimes regarded as the “things-could-be-much-worse philosophy of model evaluation” (Arbuckle, 2007, p595), the advantage of these models is that they provide an indication of “acceptable fit” (Brown, 2000, p84) where other models are significantly constrained.

Rejection of the null hypothesis using NFI, CFI and TLI is recommended when index values fall below .90 (Brown, 2000). Values between .90 and .95 indicate acceptable model fit; and values above .95 indicate good fit (Brown, 2000).

Residuals

Standardised residual scores measure the difference between observed and estimated correlations (Hair et al, 1995), and so provide an indication of how far the observed data varies from the hypothesised model. The threshold for assessing significant residuals is commonly held as ± 1.96, or 2.0 (Brown, 2006; Hair et al, 1995), although “Residual values greater than ± 2.58 are now to be considered statistically significant at a .05 level” (Hair et al, 1995, p644). Observed variables with high residual scores are reviewed in the analyses.

Parameter Estimates

Statistical significance alone, according to Brown (2006), is not sufficient to affirm 208

model fit. As confirmatory factor analysis is typically conducted with large sample sizes, statistical power is often achieved even for very small effects (Brown, 2006). Factor loadings, squared multiple correlations (r2 values), and latent factor correlations should all be considered in tandem with goodness of fit indices to determine model strength.

In congeneric models, factor loadings (standardised regression weights) are “interpreted as the correlation between the indicator and the latent factor” (Brown, 2006, p130-131). A common threshold used to identify “salient” standardised factor loadings is .30 or .40 (Brown, 2006). However, as the squared standardised factor loading, r2 represents the proportion of variance of the observed variable that is “explained by the latent factor” (Brown, 2006, p131), factor loadings of .70 and above are preferred (Hair et al, 1995).

Sample size

Sample size, and significance testing are important considerations for this pretest. Given the expectation that 5% of correlations may occur by chance, Hair et al recommend “at least five times as many observations as there are variables to be analysed, and the more acceptable range would be a ten-to-one ratio” (Hair et al, 1995, p373). Testing 20 items determines that the minimum acceptable sample size for the reported tests of identification items is one hundred.

Test contexts

The information provided to respondents for the first test of identification items is shown below: The State of Origin competition is a three-match competition played each year between the Queensland Maroons and the New South Wales Blues. Please indicate your agreement or disagreement with each of the statements below according to your feelings about the Queensland Maroons.

209

The State of Origin competition is an inter-state rugby league competition. Data collection for this pretest occurred in the final week of semester 1 classes in 2010 which falls in the middle of the State of Origin competition.

The information provided for the second test of identification items is shown below: The Brisbane Broncos are one of the Queensland representative teams in the Australian National Rugby League (NRL) competition. Please indicate your agreement or disagreement with each of the statements below according to your feelings about the Brisbane Broncos.

Data collection for this pretest occurred in the final week of semester 2 classes in 2010 (October 25-29). The NRL competitive season and finals had been completed on October 3rd; the Grand Final was played between St George Illawarra Dragons and the Sydney Roosters. The Brisbane Bronco’s finished the season10th of 16 teams.

EFA Findings: State of Origin data

Ninety-one responses were collected, from which 78 valid cases were processed. Respondent demographics reveal the sample is 38.5% male and 48.4% female, with 12 missing values. The average age of respondents is 21.3 years, with a standard deviation of 3.54; there were 13 missing values for this variable.

The factor solution is show in Table 5.33. Four factors were extracted with eigenvalues above .800. The fourth factor was included in the analysis after viewing the scree plot. Although Kaiser’s criterion argues the use of factors with eigenvalues equal to or greater than one, factors with eigenvalues less than one are still regarded as acceptable when those factors account for more than 10% of the variance (Cattell, 1966; Child, 1970). The four factor solution accounts for 74.5% of total variance. Eleven items achieved factor loadings of .70 and above.

210

Table 5.33: EFA: State of Origin solution Item 1 2 3 4 4 I think of the team as part of who I am 0.822 0.175 0.241 0.129 16 When someone criticises the team, it feels like a 0.806 0.088 0.233 0.002 personal insult. 15 I have a strong attachment to the team. 0.747 0.386 0.337 0.225 1 The team's successes are my successes. 0.702 0.266 0.177 0.180 14 I have a strong sense of belonging to the team. 0.643 0.444 0.342 0.319 17 I'm very interested in what others think about the 0.637 -0.079 0.414 0.114 team. 12 Others describe me as a typical fan of the team. 0.603 0.342 0.373 0.235 5 I am proud to be a fan of the team. 0.560 0.428 0.251 0.493 8 I am like other fans of the team. 0.551 0.334 0.221 0.455 7 I identify with other fans of the team. 0.547 0.357 0.291 0.561 13 In general, I'm glad to be a fan of the team. 0.539 0.477 0.137 0.525 2 I have little respect for the team. 0.202 0.835 0.190 -0.016 10 I dislike being a fan of the team. 0.170 0.826 0.179 0.173 11 I would rather be a fan of the other team. 0.195 0.783 -0.002 0.303 6 I think the team has little to be proud of. 0.007 0.668 0.084 0.191 9 I would like to continue to support the team. 0.479 0.492 0.256 0.551 18 I am aware of the tradition and history of the team. 0.218 0.134 0.871 0.132 19 I know the rituals that go with being a fan of the team. 0.382 0.231 0.723 0.309 20 I have knowledge of the successes and failures of the 0.313 0.171 0.758 0.222 team. 3 Overall, the team is considered good by others. -0.058 0.195 0.271 0.827

Eigenvalues 11.366 2.150 1.328 .825 % of Total Variance 28.6 19.3 14.1 12.5

Discussion: State of Origin data

The results of the pretest of identification items factor analysis can be summarised in two areas. These are: a) method-based issues and, b) dimension related issues.

In relation to method, the sample size for this pre-test is below the minimum 100 responses required. Acceptable factor-loadings were identified at the .70 level which usually allows statistical significance to be achieved with smaller sample sizes. This, however, does not mitigate the problem of small sample size and correlations occurring by chance (Hair et al, 1995). The outcome of this factor analysis suggests

211

that results should be regarded with some caution. Replication of this test with a larger sample or different population sample may not provide the same outcome

The items that load together show a degree of face validity. Factor one appears to group items that represent an affirmation or affective dimension. In previous studies three of the four items that load in the current analysis on factor one, tested as affective items.

The second factor reflects negative, denial or oppositional forms of identification. In previous studies, the three items loading on factor two were conceptualised as representing evaluative or affective dimensions (Dimmock et al, 2005; Ellemers et al, 1999). The existence of a negative dimension in an identification scale is not in itself problematic. The scale developed by Driedger (1976) was conceived of as having two dimensions; ‘affirmation’ of group membership and ‘denial’. This structure has been echoed in the work of Brown et al (1986), Kelly (1988) and Hinkle et al (1989).

The third factor contains three items from the Heere and James (2007) TEAM*ID scale. The items do not comply easily with any of the dimension definitions associated with the affective-cognitive-evaluative model, although Heere and James (2007) do not endorse that model exclusively. Similar items are found particularly in scales dealing with ethnic identification (see: Driedger, 1976; Phinney, 1992; Sellers et al, 1997). The argument could be made that these items reflect specific knowledge bases that may be consequences of identification rather than a dimension of the condition itself. The items, alternatively, may be regarded as representative of a value system that contributes to identification processes. The items will be retained for further testing.

The final factor is a single item, overall measure of identification. It accounts for more than 10% of the sample variance, and so is retained as a factor in this analysis despite its eigenvalue measuring below 1(see: Cattell, 1966; Child, 1970).

Overall, this exploratory study produced three strong factors and one factor with a marginal eigenvalue. The factors revealed might suggest that team identification suits an affective (affirmation), affective (denial), and value-based dimensional model of

212

identification. This idea diverges from recent models of identification which suggest affective-cognitive-evaluative models of identification. Further testing of this model is necessary for reasons of sample size and replication.

EFA Findings: Brisbane Broncos data

Four hundred and four responses were collected, from which 396 valid cases were processed. Respondent demographics reveal the sample is 40.8% male and 54.2% female, with 20 missing values. The average age of respondents is 21.68 years, with a standard deviation of 5.07 (range: 17-50 years); there were 19 missing values for this variable. The sample was comprised 75.5% English as a first language and 21.04% English as a second language, with 14 missing values.

An exploratory factor analysis was first conducted to evaluate whether the larger sample size, compared to the test of State of Origin data, substantially influenced factor structure. A three-factor solution was extracted with eigenvalues above 1.0; Table 5.34 shows the factor solution. The solution accounts for 67.7% of total variance. Eight items achieved factor loadings of .70 and above. This analysis suggests a weaker solution than that found in the initial exploratory factor analysis with factors two and three each explaining less than 10% of variance. The pattern of factor results, however, is relatively in-tact. There remains an affirmation factor, a denial factor, and a third that represents values-based identification.

The affirmation factor is the strongest and least ambiguous; it explains the majority of variance, and retains its integrity (with one exception). Item 19, in the State of Origin analysis, loaded with the values-based identification items; in the current analysis it cross-loads. The remaining two value-based items have maintained their grouping, suggesting a (weak) stable relationship.

The denial factor, in the initial exploratory factor analysis comprised three items loading above .70 (range .78 to .83), a forth item could be considered part of the group with a weaker loading of .66. In the current analysis three of the same four items load on this factor, but with much reduced factor loadings (range .63 to .69)

213

Table 5.34: EFA: Brisbane Broncos solution Item 1 2 3 When someone criticises the team, it feels like a personal 16 .860 .130 .108 insult. 4 I think of the team as part of who I am .841 .199 .165 14 I have a strong sense of belonging to the team. .803 .362 .247 12 Others describe me as a typical fan of the team. .785 .213 .103 15 I have a strong attachment to the team. .774 .340 .308 8 I am like other fans of the team. .774 .234 .229 1 The team's successes are my successes. .771 .163 .279 5 I am proud to be a fan of the team. .662 .382 .437 13 In general, I'm glad to be a fan of the team. .636 .405 .478 19 I know the rituals that go with being a fan of the team. .618 .528 .033 9 I would like to continue to support the team. .571 .405 .531 7 I identify with other fans of the team. .554 .551 .319 17 I'm very interested in what others think about the team. .554 .432 .058 20 I have knowledge of the successes and failures of the team. .227 .814 .070 18 I am aware of the tradition and history of the team. .406 .690 .041 2 I have little respect for the team. .358 .553 .516 3 Overall, the team is considered good by others. .186 .505 .423 6 I think the team has little to be proud of. .099 .199 .692 10 I dislike being a fan of the team. .196 .164 .647 11 I would rather be a fan of the other team. .097 -.327 .639

Eigenvalues 10.824 1.500 1.222 % of Total Variance 54.1 7.5 6.1

Discussion: Model Specifications

The exploratory factor analysis of Brisbane Broncos data provides some support for the three factor solution found in the initial test of State of Origin data. However, the Brisbane Bronocs test did not suffer from an insufficient sample. The test also showed that the second and third factors each failed to explain 10% of the variance. On these bases, a three-factor model as well as a single-factor model will be tested using confirmatory factor analyses.

The items to be tested and their relationships to latent variables are tabled below.

214

Model 1: three-variable model Latent variable Items Affirmation 1 The team's successes are my successes. 4 I think of the team as part of who I am 8 I am like other fans of the team. 12 Others describe me as a typical fan of the team. 14 I have a strong sense of belonging to the team. 15 I have a strong attachment to the team. 16 When someone criticises the team, it feels like a personal insult. Denial 6 I think the team has little to be proud of. 10 I dislike being a fan of the team. 11 I would rather be a fan of the other team. Values-based 18 I am aware of the tradition and history of the team. 19 I know the rituals that go with being a fan of the team. 20 I have knowledge of the successes and failures of the team.

Model 2: single-variable model Items 1 I have a strong sense of belonging to the team. 2 The team's successes are my successes. 3 I think of the team as part of who I am 4 I am like other fans of the team. 5 Others describe me as a typical fan of the team. 6 I have a strong attachment to the team. 7 When someone criticises the team, it feels like a personal insult.

CFA Findings: Brisbane Broncos data

Model 1: Fit 1

Forty-two parameters were estimated from 104 sample moments, producing 62 degrees of freedom. Fit indices for the initial confirmatory factor analysis indicate a weak fit for the model.

The p-value for the Chi-square test indicates that the model is not a perfect fit for the data. This result is not unexpected given the absolute nature of this test.

Chi-square statistic Df p 253.745 62 .000

The RMSEA result of .088 suggests that the model has received the minimum

215

amount of support. This model should not be rejected as it has not reached the .10 level (Brown, 2000; Hu & Bentler, 1999), but should be regarded with caution. Furthermore, the pCLOSE statistic indicates that the result achieved is significantly different from a p-value of .05.

Parsimony corrected fit index RMSEA Low90 High90 pCLOSE .088 .077 .100 .000 Comparative fit indices NFI CFI TLI .927 .944 .929

The comparative fit indices support the RMSEA result that the model provides a an acceptable, though weak fit with values below .95. The strength of the model suggests the need for a review of residuals, and modification indices to identify issues that would improve fit.

Residuals

The pattern of standardised residuals provides an initial point from which to suggest model refinement. Six residuals exceed ± 1.8; three of these exceed ± 2.0 (these are shown below in Table 5.35). Of the six large residuals, item 11 appears three times; and items 15 and 20 each occur twice. These patterns suggest that items 11, 15 and 20 are responsible for a degree of variation of the observed data matrix.

Table 5.35: Model 1: Large residual covariances Item Observed variable Item Observed variable Residual Covariance 11 Rather_ other 20 Success_ fail 3.009 11 Rather_ other 18 Tradition_ history 2.322 20 Success_ fail 16 Someone_ criticises 2.065 15 Strong_ attachment 10 Dislike_ being 1.901 11 Rather_ other 8 Am_ like 1.859 15 Strong_ attachment 19 Rituals 1.827

Appendix 5.2 provides the standardised residual covariances for Model 1.

216

Modification Indices

The AMOS modification output provides a total of 34 recommendations for introducing covariances that reached the modification index threshold 4.0 (Arbuckle, 2007). Fifteen of these recommendations are to allow latent variables to covary with error variances; these recommendations are rejected as they would violate to assumption of independence of latent and error variance (Tabachnick & Fidell, 2007).

Table 5.36 below provides a frequency of error variable presence in modification indices, the variance of the variable, and representation in high-residual pairs. Four items are highlighted in Table 5.36 as being problematic (Items: 11, 15, 16, and 20). Items 16 and 20 have error variables that are strongly represented in modification indices, with 6 and 7 appearances respectively. Error variance probably explains the presence of V1 in modification indices, and item 20 high-residual pairs. This item (20) is selected for removal from the model on the basis of its variance; Item 11 will be removed from the model for the same reason.

Table 5.36: Model 1: Error variable Modification Indices & Variances Item Observed variable Error M.I Variance High-residual variable Frequency pairs 4 Part _of A2 4 .611 No 14 Sense _belong A3 3 .469 No 12 Typical A4 1 .779 No 8 Am _like A5 3 .859 Once 1 Successes A7 2 1.121 No 18 Tradition _history V3 1 1.373 Once 19 Ritual V2 3 .702 Once 10 Dislike _being D2 3 1.726 Once 6 Little _proud D1 2 1.664 No 16 Someone _criticises A1 6 .801 Once 15 Strong _attachment A6 1 .674 More than once 20 Success _fail V1 7 2.269 More than once 11 Rather _other D3 2 2.286 More than once

Item 16, alternatively, cannot be explained with reference to error variance. In modification indices the A1 error variable is linked to A2, A5, A6, V1, D2, and D2 (see Appendix 5.3). The correlation of error variables, although not ideal, can be

217

justified on theoretical bases, in particular where an “exogenous common cause” (Brown, 2006, p181) can be identified.

For the A1 error variable, links to A2, A5, and A6 could be justified on the basis that all items refer to elements of identification that reflect attachment, internalisation, or the idea of a shared or common fate. This justification would be confounded by the direction of parameter change (see Appendix 5.3) which is not consistent across all variables. Additionally, the D1, D2, and V1 errors do not readily reflect ideas that might contribute to an understanding of internalisation or attachment. Item 16 is regarded as a potential confound to analysis, and removed from the model.

Item 15 is the final problematic observed variable; it is also selected for removal from the model. Item 15, which appeared in high-residual pairs twice, is not over- represented in covariance modification indices, and its level of error variance is acceptable. Furthermore, the high-residual pairs it contributed to were marginal in their excess (1.827, and 1.901). The modification indices for regression weights, however, suggest that this variable has some relationship with both the values and denial latent variables. Although the modification indices have exceeded the modification threshold of 4.0, introduction of these regression weights are regarded as potentially confounding of analysis.

The review of residuals, error variances and modification indices has concluded in the identification of four variables (Items: 11, 15, 16, and 20) that are regarded as unreliable or potentially confounding. These items are removed from the model; results for a second confirmatory factor analysis are provided below.

Covariance modification indices are also tabled in the Appendix for this study.

Model 1: Fit 2

Fit indices for the re-specified model show significant improvement over the initial model specification. Thirty parameters were estimated from 54 sample moments, producing 24 degrees of freedom.

218

The p-value for the Chi-square test indicates that the model is not a perfect fit for the data. It is significantly reduced from the previous analysis. Failure to achieve a perfect fit to the model, according to the chi-square test was not unexpected.

Chi-square statistic Df p 56.538 24 .000

In contrast to the chi-square p-value, the RMSEA result has improved from .080 to 0.059, which suggests that the model is now an acceptable fit for the data. The pCLOSE statistic reflects this improvement, and indicates that the RMSEA result is not significantly different to a p-value of .05. The comparative indices have correspondingly moved closer to 1.0; all now exceed 0.95, which provide further support of a good fit.

Parsimony corrected fit index RMSEA Low90 High90 pCLOSE .059 .039 .079 .219 Comparative fit indices NFI CFI TLI .973 .984 .976

Residuals & Modification Indices

The table outlining standardised residual scores (see Appendix 5.4) for Fit 2: Model 1 reveals no score has reached the minimum ±1.96 that would signal significant deviations from expected and observed correlations. The absence of large residuals supports a conclusion that the model is a good fit.

Modification indices also reveal a significant improvement in the model; 5 error covariances are suggested, although none of these will be introduced. Modification indices are tabled in the Appendix 5.5 to this study (Fit 2: Modification Indices: covariances). Two of the indices suggest covariance of error and latent variables (a2 - affirmation; and a4-values); to introduce these changes would violate independence rules. The remaining recommendations will not be introduced as no theoretical

219

reason exists to justify these. Each of the error covariances has a large modification index (range: 5.8 to 13.7), which signals a likely impact on size of the chi-square statistic. Modification indices do not recommend the introduction of any new regression weights.

Parameter estimates

The standardised factor model is provided in Figure 5.15 below. All parameter estimates (regression weights and covariances) have achieved statistical significance at the .001 level.

Table 5.37: Model 1, Fit 2: Regression estimates Standardised Observed Latent Standard Critical Estimate p Regression variable Variable Error Ratio Weight successes <--- affirmation .943 .043 21.727 *** .809 typical <--- affirmation .718 .036 20.118 *** .776 sense_belong <--- affirmation 1.000 .906 part_of <--- affirmation .871 .034 25.274 *** .871 am_like <--- affirmation .820 .039 21.234 *** .799 rituals <--- values 1.057 .068 15.479 *** .920 tradition_history <--- values 1.000 .755 little_proud <--- denial .827 .136 6.087 *** .548 dislike_being <--- denial 1.000 .641

*** p ≤ .001

All of the standardised regression weights have exceeded the .30 threshold; for affirmation and values variables all exceeded the .70 level. Standardised regression weights for the denial variable are uniformly low (.55 and .64). The denial factor is retained in the current model of identification for two reasons; the first is that factor loadings have reached a popular minimum standard. The second reason is that removal of this factor from the model actually decreases goodness of fit (Chi-square 41.407, df =13, p≤.000; RMSEA .074, pclose .052), rendering a borderline pclose result.

220

Figure 5.15: Standardised Factor Model: Model 1

.76 ..part of who I am. .87

.82

.91 Affirmation …sense of belonging

.60 .78 …a typical fan

.64 .80 I am like other fans… .81 .65 .62 The team’s successes

.85 .92

...know the rituals Values .78

.57 .75 ...aware of tradition

.46 .30 … little to be proud of .55

.41 Denial …dislike being a fan .64

Standardised Model

221

The squared multiple correlations, consistent with the standardised regression weights, provide that for most observed variables, the majority of variance is explained by the latent factor (Brown, 2006). The denial factors’ indicators provide the exception to this, as their smaller regression weights determine higher levels of unique or error variance (Brown, 2006). This might suggest the need to re-word or re-conceptualise the items in order to improve their relationship to the latent factor.

Table 5.38: Model 1, Fit2: Latent variable correlations Latent Standard Critical Estimate p Correlation variables Error Ratio affirmation <--> values 1.787 .185 9.639 *** .782 denial <--> values .748 .139 5.359 *** .462 denial <--> affirmation 1.011 .145 6.996 *** .616 *** p ≤ .001

The final issue to be addressed is the correlation between latent variables. The correlation between affirmation and values latent variables is large, and raises the question of discriminant validity. The threshold for a discriminant validity problem, according to Brown, is “.80 or .85” (2006, p131); this has not been reached in the current analysis. One explanation for the high correlation is the failure of confirmatory factor analysis to identify cross-loading items (Brown, 2006); the consequence of this is the inflation correlations between latent variables through a process of transitivity.

Model 2: A single factor

This model tests seven identification items that test strongly (factor loadings above .70; loading on a single factor) in the exploratory factor analyses of State of Origin and Brisbane Broncos data. The sample size for the confirmatory factor analysis is 397; 35 sample moments were analysed with 21 distinct parameters and 14 degrees of freedom.

The p-value for the Chi-square test indicates that the model is not a perfect fit for the data. This result is not unexpected given the absolute nature of this test.

222

Chi-square statistic Df p 68.215 14 .000

The RMSEA result of .099 suggests that the model is not a good fit for the data (Brown, 2000; Hu & Bentler, 1999). Furthermore, the pCLOSE statistic indicates that the result achieved is significantly different from a p-value of .05.

Parsimony corrected fit index RMSEA Low90 High90 pCLOSE .099 .076 .123 .000 Comparative fit indices NFI CFI TLI .971 .977 .966

The comparative fit indices show greater support for the uni-dimensional model, with values consistently above .95. The fit indices results suggest the need for a review of residuals, and modification indices.

Residuals & Modification Indices

The table outlining standardised residual scores (Appendix 5.6) for the current test of the model reveals no score has reached the minimum ±1.96 that would signal significant deviations from expected and observed correlations. The absence of large residuals supports that the model is a good fit.

Modification indices show covariance among error variables. The error variable for items 2 and 7 are each identified in three modifications. For each of these error variables one of the modifications is large (d2>d3 15.89; d7>d4 16.18); because these error variables are implicated in more than one error covariance; items 2 and 7 will be removed from the model. Modification indices are tabled in Appendix 5.7 for this study (Model 2: Fit 1: Modification Indices: covariances).

223

Model 2: Fit 2

With items 2 and 7 removed from the model due to large covariance among error variables; model fit is improved. The test has 20 discrete sample moments, and 15 parameters estimated; 5 degrees of freedom. The p-value for the Chi-square test indicates that the model is not an acceptable fit for the data.

Chi-square statistic Df p 12.457 5 .029

The RMSEA result of .061 suggests that the model has received an acceptable amount of support. Furthermore, the pCLOSE statistic indicates that the result achieved is not significantly different from a p-value of .05.

Parsimony corrected fit index RMSEA Low90 High90 pCLOSE .061 .018 .105 .279 Comparative fit indices NFI CFI TLI .992 .995 .990

The comparative fit indices show greater support for the re-specified uni-dimensional model, with values consistently above .95. The strength of the model suggests its soundness for use as a uni-dimensional identification scale.

Parameter estimates

The standardised factor model is provided in Table 5.39 below. All parameter estimates (regression weights and covariances) have achieved statistical significance at the .001 level. All factor loadings (standardised regression weights) are above the .70 threshold (Brown, 2006; Hair et al, 1995). The squared multiple correlations, consistent with the standardised regression weights, provide that the majority of variance is explained by the latent factor (Brown, 2006).

224

Table 5.39: Model 2, Fit 2: Standardised Model Standardised Squared Observed Latent Std Critical Estimate p Regression Multiple Variable Error Ratio Weight Correlation strong_attach.  Id. 1.017 .038 26.725 *** .881 .776 typical  Id. .717 .034 20.966 *** .785 .617 am_like  Id. .807 .038 21.516 *** .796 .633 part_of  Id. .839 .034 24.604 *** .849 .721 sense_belong.  Id. 1.000 .918 .843 *** p ≤ .001

Conclusions: Multidimensional & Unidimensional Identification

The confirmatory factor analyses have provided that an acceptable multi-dimensional identification scale can be formulated. Analyses also show that a relatively stronger uni-dimensional model, comprised mostly of ‘affirmation’ items, can be developed. In terms of developing a scale that is reliable, the multi-dimensional model achieves a Cronbach’s alpha of .891, which improves if either of the denial items is removed. The uni-dimensional model achieves a Cronbach’s alpha of .926; this does not improve if items are removed. The alphas for either model are acceptable.

The argument is not merely a conceptual one; multiple dimensions should not be necessary to measure identification. The issue of parsimony remains important. The uni-dimensional model of identification must be the preferred scale in research contexts where respondent fatigue is a significant concern. In research where hypotheses are tied to construct dimensions, a multi-dimensional model is appropriate. It is likely, that in regard to the current multi-dimensional model, further conceptual work in developing identification dimensions would seem prudent.

225

Chapter 6: Study 2: Test of news’ article stimuli effects on attitudes

The experiment reported in this chapter tests whether negative off-field behaviours of sportspeople influence consumer attitudes toward sponsors and teams. By extension, whether off-field behaviours have the power to influence corporate performance. Three scenarios (off-field behaviours) are tested; a single positive case is included to enable evaluation of potentially equal (if opposite) effects may accrue to sponsors and team when sportspeople engage in positive off-field behaviours.

The reporting of Study 2 begins with statement of research question and hypotheses. A method section follows which details experimental characteristics, administration of the experiment, and criteria for significance for this study. Sample data fit with analysis assumptions are reported. Potential confounds are assessed and the model is modified. The multivariate tests are provided followed by tests of hypotheses and comparisons of means. The chapter concludes with a discussion of results.

An experiment is used to evaluate the effects of reading a newspaper article reporting off-field incidents on attitudes. The experimental data is analysed using Multiple Analysis of Covariance (MANCOVA). Characteristics of the sample data do not allow for categorical use of the covariate (Identification with Team), which would enable a 3 (news stimuli) x 2 (identification: high/ low) factorial design for analysis using Multiple Analysis of Variance. Details of the variables are shown below.

226

Independent variables Covariate variable Dependent variables

Group Scenario Person variables A Assault Identification with Team Person Likeable B DUI Attitude toward the Act C Charity Team variables fundraising Attitude toward the Brand: Team Corporate Image: Team Sponsor variables Attitude toward the Brand: Sponsor Corporate Image: Sponsor

Hypotheses

Three groups of hypotheses have been developed; these account for the effects of treatment group and the covariate, and post hoc evaluation of the perceptual balance of aggregate evaluations. The model showing hypotheses 1 and 2 is provided below.

Figure 6.16: Model A: Hypotheses 1 & 2

Person Variables: Person Likability Attitude toward the Act

News’ H1 Team Variables: article Corporate Image stimuli Attitude toward the Brand H2 Sponsor Variables: Team Corporate Image Identification Attitude toward the Brand

A list of hypotheses follows: H1: Valence of news’ will be reflected in evaluations of the athlete, their team and the team’s sponsor. H1a: Valence of news’ will be reflected in evaluation of the athlete’s behaviour (Attitude toward the Act).

227

H1b: Valence of news’ will be reflected in evaluation of the athlete (Person likability). H1c: Valence of news’ will be reflected in evaluation of the team (Attitude toward the Team). H1d: Valence of news’ will be reflected in evaluation of the image of the team (Team Corporate Image). H1e: Valence of news’ will be reflected in evaluation of the sponsor (Attitude toward the Sponsor). H1f: Valence of news’ will be reflected in evaluation of the image of the sponsor (Sponsor Corporate Image). H2: Level of Team Identification will moderate evaluations of group members (athlete, team, and sponsor)

H2: Effects of Identification with the Team will be reflected in evaluations of the athlete, their team and the team’s sponsor. H2a: Team Identification will moderate evaluations of the athlete’s behaviour (Attitude toward the Act). H2b: Team Identification will moderate evaluations of the athlete (Person likability). H2c: Team Identification will moderate evaluations of the team (Attitude toward the Team). H2d: Team Identification will moderate evaluations of the image of the team (Team Corporate Image). H2e: Team Identification will moderate evaluations of the sponsor (Attitude toward the Sponsor). H2f: Team Identification will moderate evaluations of the image of the sponsor (Sponsor Corporate Image).

The balance theory hypothesis required that actors evaluated in the experiment (athlete, team and sponsor) are regarded as equally positive or negative. One balance hypothesis is developed; two models explain how balance will be assessed.

Hypothesis 3 will be evaluated using two forms of Model B.

H3: The valence of the reported behaviour (positive/ negative) will be maintained across evaluations of the athlete, their team, and the sponsor.

228

Model B (i), which anticipates equivalence in evaluations for all variables, is shown below.

Figure 6.17: Model B: Hypothesis 3

U Athlete Team

DL / L U DL/ L

DL/ L p Sponsor Notation for the model: DL: Dislike (negative sentiment relation); L: Like (positive sentiment relation); U: Unit Relation (positive/ structural tie); p: the perceiver

Model B, alternatively, anticipates that balanced models may be produced for effects of treatment group, distinct from a balance model of effects of the covariate. Model B is shown below.

Figure 6.18: Model B (ii): Hypothesis 3

Athlete Athlete

U DL/ L U DL/ L U U

DL/ L DL/ L Team p Team

U DL/ L DL/ L U

Sponsor Sponsor

Effects of treatment: Group Effects of covariate: Identification with Team

229

Method

The experimental design takes the forms of a post-test only control group experiment. This design is justified on the basis of parsimony as it reduces administration, sample attrition due to respondent drop-out, and respondent fatigue. The post-test only control group design maintains all of the minimum required features of an experiment (Singleton & Straits, 2005); these characteristics are discussed below according to their influence on the administration of this study. This discussion includes details of the administration and debriefing, including ethics, of the study.

Experiment characteristics

Manipulation of the independent variable

The independent variables in this experiment are represented as off-field behaviours in news articles (Groups A, B, and C). Each scenario explicitly provides, or allows inference on five key pieces of information. These are the ‘who’, ‘what’, ‘when’, ‘where’, and ‘why’ of situation description. The ‘who’ for all treatment scenarios is “Ben Howse”, this name is fictitious, and not developed to resemble the name of any current or previous NRL player. For each treatment scenario, two ‘what’ pieces of information are provided; the first ‘what’ describes the behaviour of the actor; the second ‘what’ describes the reason for the news article (e.g., a benefit event, an investigation, a legal appeal, etc). For each treatment scenario the Team, Sponsor, and Player are named; the control scenario does not provide a player name. The scenarios are provided in Appendix 6.1 for this study.

The behaviours tested in this experiment were controlled on a further two characteristics. The first characteristic was that behaviours should be measurable using the same scale on the Attitude toward the Act variable. For ease of analysis behaviours used in this experiment are measurable using a single Attitude toward the Act scale, comprised of three items.

The second characteristic is the intention to test a model balanced according to levels

230

of behaviour severity. Group C (charity fundraising) provides an example of positive off-field behaviours in contrast to Group A (assault causing injury) which is extremely negative. Group B (driving under the influence of alcohol) is expected to provide a neutral mid-point between groups A and C. This expectation is not driven by pretest results, which argue DUI should be regarded as extremely negative. Instead, this expectation is driven by the outcomes of the Study 1 content analysis which argue the mismatch between incident occurrence and reporting. On the basis that DUI is potentially under-reported in the news-media it is assumed to have either less news-value (lesser ability to shock), or a degree of normative acceptance not reflected in the pretest.

The behaviours are tabled below, which shows treatment group, category, and pretest statistics for the Attitude toward the Act variable. Means are represented as proportion of the scale value (likert-type scale: 1=extremely negative, 7=extremely positive). All scores reported from this point forward will use the same format.

Table 6.40: Pretest A-act scores Group Scenario Category Expected Pretest Pretest valence mean median A Assault Violence Negative .16 1 B DUI Anti-social Neutral .19 1 C Cancer Positive Positive .85 6.33

The comparison group

According to Singleton and Straits, true experiments have at least one comparison or control group (Singleton & Straits, 2005). Specifically, they provide that a true experiment has “…at least two groups, experimental and comparison” (Singleton & Straits, 2005, p159). Further support for the use of a comparison groups, as opposed to a ‘control’ group is provided by Boruch (1998). Boruch explains: In most applied social research, the “control” condition is not one in which any treatment is absent. Rather, the label usually denotes a condition in which conventional or customary treatment is delivered. Boruch, 1998, p177 The research reported here does not include a control group characterised by non- delivery of experimental stimulus. Instead, the ‘comparison group’ approach is used. 231

Specifically, the positive scenario (charity fundraising) will provide a contrast to the test of negative news-media reports.

Measurement of the dependent variable

Measurement of the dependent variable is another condition of true experiments (Singleton & Straits, 2005), specifically, that measurement follows exposure to the stimulus. Direction of influence also applies to the covariate variable. Measurement of the covariate precedes exposure to stimulus and dependent variables. This direction of influence allows the researcher to argue for causal relationships among the independent variable, covariates and dependent variables.

The direction of influence, in this experiment, is maintained through order of exposure. This order is: a) measurement of the covariate (Identification with Team), b) exposure to treatment, c) measurement of the dependent variables, and d) measurement of individual difference variables (age, gender, and language group). Appendix 6.2 provides the scales used to measure dependent variables.

With six dependent variables across three groups (person variables, team variables, and sponsor variables), and several items each, the risk of fatigue to respondents was considered. In order to address this concern, order of the dependent variables was manipulated. Two order treatments were used for the dependent variables: a) person variables, team variables, and sponsor variables, and b) sponsor variables, team variables, and person variables.

Random assignment to groups

Random assignment to groups is the condition of true that experiments that aims to reduce non-random sampling errors. By randomly assigning respondents to groups, the researcher avoids uneven distribution of individuals’ characteristics “or experiences that might confound the results” (Singleton & Straits, 2005, p157). Random assignment, in the current experiment, was ensured through sequencing the 232

treatment group booklets (ABC-ABC). Although data were collected across several undergraduate business classes, this sequence ensured that a single class did not, for example, receive only Group ‘A’ treatments.

Constancy of conditions

The constancy of research conditions reaffirms the principle behind random assignment to groups, which is the controlling of confounds. For this experiment a set of procedures were maintained across all data collection venues. These procedures included an introductory script identifying the research and asking for participation. This was followed by informed consent procedures, emphasising the voluntary nature of participation for respondents. Respondents were allocated 15 minutes to complete the exercise. Finally, collection of the survey forms was followed by respondent debriefing. Scripts for introduction and debriefing are included in Appendix 6.3.

Evaluation of test statistics

In this section the standards used to evaluate test outcomes in this research are discussed. The differences between Pillai’s Trace, Hotelling’s Trace, Wilks’ lambda and Roy’s Greatest Characteristic Root (referred to as Roy’s Largest Root in SPSS) as multivariate measures of significance are evaluated.

The test alpha, or significance level, and adjustments to alpha levels where multiple comparisons are used are discussed with reference to Bonferroni corrections. The statistical power and effect sizes sought to evaluate significant alphas are also provided.

Multivariate tests of significance

General Linear Model (GLM) tests in SPSS produce four tests of multivariate 233

significance: Wilks’ lambda, Pillai’s criterion, Hotelling’s Trace, and Roy’s Greatest Characteristic Root (renamed Roy’s largest root).

For this research, statistics generated by Pillai’s criterion, Hotelling’s Trace and Wilks’ lambda are the preferred measures of multivariate significance. Roy’s Greatest Characteristic Root (Roy’s gcr), while powerful, is not robust to violations of assumptions (Hair et al, 2006; Tabachnick & Fidell, 2007) of which there are several in this data set. Wilks’ lambda is also thought to be sensitive to violations of assumptions relative to Pillai’s criterion or Hotelling’s Trace (Hair et al, 2006; Tabachnick & Fidell, 2007). It is, however, regarded as more robust than Roy’s gcr for using multiple discriminant functions to assess significance, rather than simply the first discriminant function (Hair et al, 2006).

Alpha and adjustments

The significance level for this study is initially set at the .05 level. Adjustments to this level will be made in three situations. These are: when violation of assumptions requires a more stringent alpha; for post hoc tests where the family-wise error rate is inflated; and when the risk of Bonferroni inequalities arises due to multiple comparisons (Benjamini & Hochberg, 1995; Hair, Anderson, Tatham & Black, 1995; Simes, 1986; Tabachnick & Fidell, 2007). Two forms of adjustment are used, the Bonferroni correction, and Sidak’s procedure.

The Bonferroni correction makes significance testing more stringent by altering the acceptable p-value to reflect the change in risk level of a Type 1 error – mistakenly rejecting the null hypothesis (Simes, 1986). Specifically, “the classical Bonferroni multiple test procedure is usually performed by rejecting H0 = {H1…, Hn} if any p- value is less than α/n” (Simes, 1986, p751).

Sidak’s test is a modification of the Bonferroni correction that adjusts alpha using the 1/k formula: αs = 1- (1 - α) (Conagin & Barbin, 2006; Pizarro, Guerrera & Galindo, 2002) where k is the number of planned comparisons (Conagin & Barbin, 2006). Sidak’s test, along with Bonferroni and Scheffe’s tests are among the most

234

conservative corrections for alpha levels (Pizarro, Guerrera & Galindo, 2002). The advantage of Sidak’s test over that of Bonferroni is that it “provides tighter bounds than for the Bonferroni test” (Pizarro, Guerrera & Galindo, 2002, p161).

In this research, Bonferroni’s correction is applied when the significance level must be set manually in SPSS (i.e., t-tests), and Sidak’s test is used for multiple pairwise comparisons.

Statistical power & effect size

The statistical power of a test is a measure of the “probability that it will yield statistically significant results” (Cohen, 1988, p1). The desired level of statistical power is commonly set at .80 (Tabachnick & Fidell, 2007). Cohen explains statistical power and its relation to β, error, where a represents the significance level, and b is error: …with a =.05, power = .80, and hence b = 1-.80 =.20, the relative seriousness of Type I to Type II error is b/a = .20/.05 = 4 to 1; thus the mistaken rejection of the null hypothesis is considered four times as serious as mistaken acceptance. Cohen, 1988, p5. Where more stringent alphas are used (.001), with power of .80, the ratio of mistaken rejection to mistaken acceptance becomes less likely (200 to 1) the practicality of this however declines as the sample size needed to achieve .80 power increases. Statistical power is related not only to the test alpha, but also to the anticipated effect size of the treatment.

In order to achieve statistical power of .80 with alpha = .05, sample sizes range widely with anticipated effect sizes. Cohen provides sample size guidelines for ANOVA noting that for a design with six groups, the sample size increase from 14 per group for a large effect size, to 215 for a small effect size (Cohen, 1992, p158). These differences have a large impact on the resources required to conduct research. Matching power to sample size and test alpha’s are important not only for considering acceptance or rejection of the null hypothesis, but also in relation to the

235

cost of research.

Statistical power, in this research, is sought at the .80 level as “a materially smaller value than .80 would incur too great a risk of Type II error” (Cohen, 1992, p156). Cohen’s use of “materially” is extended also to discussion of power above the .80 level in saying “a materially larger value would result in a demand for N that is likely to exceed the investigator’s resources” (Cohen, 1992, p156). The implication, common to allocation of alpha levels, is that .80 is an arbitrary standard; a level materially like .80 power in this research, will be regarded as acceptable. The limit of material similarity will occur at the .75 level, at which point the ratio of false rejection to false acceptance increase from 4:1 to 5:1.

Effect size is a measure of the extent to which the null hypothesis is violated. The null hypothesis for analyses of variance and covariance, according to Cohen “states that the means or mean differences of specified (“fixed”) populations are equal, or, equivalently, that “effects” defined as linear functions of means are all zero” (1988, p273). An effect that causes rejection of the null hypothesis, then, produces an F statistic that indicates that group means have deviated from zero.

Establishing the significance of a small effect size, according to the above, requires sufficient statistical power to ensure the accurate rejection of the null hypothesis, and a correspondingly appropriate sample size. The risk associated with using large sample sizes occurs when the significance of a research finding is judged solely according to alphas’ achieved. Levine and Hullett note: …when sample sizes are large, even trivial effects can have impressive looking p-values. … p-values from null hypothesis significance tests reflect both the sample size and the magnitude of the effects studied. Levine & Hullett, 2002, p614. The size of the effect, therefore, needs to be measured not just in relation to its significance, but according to a measure of proportion of variance, or mean difference. There are several measures of effect size.

Partial eta squared is the measure of effect size provided by SPSS. Eta squared, partial eta squared, and omega squared are measures used to identify portions of

236

variance attributable to a treatment effect (Tabachnick & Fidell, 2007). Eta squared (η2) is calculated as the product of SS Effect over SS Total (Sink & Stroh, 2006; Snyder & Lawson, 1993; Tabachnick & Fidell, 2007). Partial eta squared is the product of SS Effect over SS Effect plus SS Error (Sink & Stroh, 2006; Tabachnick & Fidell, 2007); where eta squared sums to one, partial eta squared usually does not. The advantage of partial eta squared over eta squared is in the inclusion of the measure of SS Error which accounts for model-related variance, allowing greater generalisability of the result across studies (Cohen, 1973; Levine & Hullett, 2002; Tabachnick & Fidell, 2007).

Judging effect sizes, or the degree to which the level of variance accounted for is large, medium or small, is an area of substantial disagreement. For partial eta squared results, Sink and Stroh suggest that a small effect is equivalent to a proportion of .01, medium as .06, and a large effect size is approximately .14 (2006, p404). Tabachnick and Fidell (2007) suggest that Cohen (1988) argues for partial eta squared levels to signify effect sizes as .01 (small), .09 (medium), and .25 (large) (see: Tabachnick & Fidell, 2007, p55; Cohen, 1988, p532).

In this study the benchmarks recommended by Sink and Stroh (2006) and Tabachnick and Fidell (2007) are used to develop criteria for judging effect sizes. Values used previously as benchmarks are aggregated to provide ranges for acceptance effect sizes. These are:  Partial η2 ≤ .01 will denote a small effect size  Partial η2 values ranging from .06 to .09 will denote a medium effect, and  Partial η2 ≥ .25 will be regarded as large.

Overall, research results will be judged according to a tripartite system where p- value, effect size and observed power are viewed concurrently (Hair et al, 2006). The levels required for confidence in a statistically significant result are: a p-value less than, or equal to .05 (unless adjusted), partial η2 effect sizes of .01 and above, and observed power of .80, or one that is materially similar (.75 is an absolute minimum).

237

Test Metric

There is variety of procedures by which a multi-item scale might be combined for analysis. The most common of these are the scale sum and mean methods (see: Bunketorp, Carlsson, Kowlaski & Stener-Victorin, 2005; Tellegen, 1988). Median scores are the appropriate metric when data are highly skewed, to retain a measure of central tendency (see: Fiske, 1980), or when researchers aim to correctly analyse ordinal or categorical data (Bunketorp, Carlsson, Kowlaski & Stener-Victorin, 2005).

Stevens (1946) seminal work on ‘the Theory of Scales of Measurement’ identifies the appropriate forms of analysis given data produced by different scale types. According to these, Likert scales should be regarded as ordinal and restricted to analysis via median, percentile, mode, contingency correlation and number of cases (Stevens, 1946). Despite, this, means of Likert scale data are frequently used in marketing (see: Javalgi et al., 1994; McDaniel, 1999; Pope, Voges & Brown, 2004; Rifon, et al., 2004; Ruth & Simonin, 2003). In practice, Likert scale data are treated as interval.

Although the treatment of Likert scale data as interval violates Stevens’ (1946) rules for allowable transformations, the standard for others (Nunnally & Bernstein, 1994; Velleman &Wilkinson, 1993; Zinnes, 1969) is the retention of meaning and relationships within the data. Data invariance is threatened given any transformation; it is therefore worthwhile to sometimes consider Stevens’ rules “too strict to apply to real-world data” (Velleman & Wilkinson, 1993, p67), and further, that “Stevens’s proscriptions often lead to degrading data by rank ordering and unnecessarily resorting to nonparametric methods” (Velleman & Wilkinson, 1993, p67).

The scale metric used in this analysis is a scale ratio (i.e., sum of item scores over potential scale value). As this procedure has been applied to create all variables for analysis, the relationship among data and variables is maintained. The procedure also constitutes a linear transformation, which is allowable (Nunnally & Bernstein, 1994; Stevens, 1946) for interval data which is the status of summated data (Nunnally & Bernstein, 1994, p16).

238

Findings

The form of analysis used is the Multiple Analysis of Covariance (MANCOVA). MANCOVA is an extension of Analysis of Covariance (ANCOVA) that tests group- based mean differences, once differences in scores have been moderated to adjust for effects of the covariate variables (Tabachnick & Fidell, 2007). This occurs through the pooling of variance, or linear combination of dependent variables, which is then adjusted “to what would be obtained is all participants had the same scores on the covariates” (Tabachnick & Fidell, 2007, p264). Adjusting for the effect of covariates allows for the accounting of error variance attributable to the covariate; which in turn allows a clearer picture of the variance attributable to independent variables.

In this research, MANCOVA is justified because a factorial MANOVA is not possible. Covariate data has not revealed equivalent numbers of respondents who are high, and low team identifiers, which would allow categorical use of the data. On this basis, use of the Identification with Team data is treated as continuous, making MANCOVA the most appropriate form of analysis.

Data screening and Tests of Assumptions

The assumptions that must be tested prior to main analyses in MANCOVA are: 1. Equality of sample sizes, and no missing data attributable to ‘not at random’ causes 2. Multivariate normality 3. Lack of outliers 4. Linearity among dependent variables, and within the dependent variable- covariate variable pairs 5. Lack of multicollinearity or singularity 6. Reliability of covariate variables 7. Homogeneity of variance-covariance matrices 8. Homogeneity of regression Tabachnick and Fidell, 2007, p303

239

In this section, data screening and assumption tests are provided in two groups. Items 1-6 are reviewed prior to multivariate analysis, and used to remove or replace data and cases and to identify potential problems for the multivariate analysis. Items 7 and 8 are tested concurrent with the multivariate tests, although reported prior to those results. The format of the multivariate model is specified prior to tests of homogeneity of variance-covariance matrices and homogeneity of regression.

Names of variables are abbreviated in the following tables; a list of these abbreviations is provided below. The variables will be referred to in a variety of ways throughout the analysis. Team and Sponsor variables will be referred to as corporate variables; Likability and Attitude toward the Act, alternatively, as person variables.

Full name of variable Type Abbreviation Identification with Team Covariate variable ID Person Likability Dependent variable Likable Attitude toward the Act Dependent variable Att-ACT Team Corporate Image Dependent variable Team CI Team Attitude toward the Brand Dependent variable Team Att Sponsor Corporate Image Dependent variable Spons CI Sponsor Attitude toward the Brand Dependent variable Spons Att

Sample sizes and missing data

An initial review of group sample sizes and missing data reveals that group sizes are equivalent. Group sizes range from 63 to 69; specific values are provided in Appendix 6.4. A review of missing values shows that for all groups, missing data has occurred within acceptable limits.

At the group level, proportions of missing data for dependent variables are below the 5% level (Hair, Black, Babin & Anderson, 2010; Tabachnick & Fidell, 2007), and therefore acceptable. However, Hair et al (2010) and Tabachnick and Fidell (2007) also note the importance of patterns in missing data, as well as overall proportion.

240

Table 6.41: Assumptions: Missing data by case Group Case Likable Att- Team Team Spons Spons % Act CI Att CI Att missing DV data

Assault 43 1 1 0 1 1 1 16.6 Assault 89 0 0 1 1 1 1 33.3

N missing 1 1 1 0 0 0 Group % 1.6 1.6 1.6 0 0 0

DUI 149 1 1 0 1 1 1 16.6 DUI 178 1 1 1 1 0 1 16.6 DUI 180 0 0 0 0 1 1 66.6 DUI 183 0 0 0 0 0 0 100.0

N missing 2 2 3 2 2 1 Group % 2.9 2.9 4.4 2.9 2.9 1.5

Charity 83 0 0 0 0 0 0 100.0 Charity 159 1 0 0 0 0 0 83.3 Charity 163 1 0 1 1 1 1 16.6

N missing 1 3 2 2 2 2 Group % 1.4 4.3 2.9 2.9 2.9 2.9 *1 signifies a score recorded, 0 = missing value.

Missing values at the case (respondent) level are summarised in Table 6.41. Several cases have missing values above the 5% level; these cases demonstrate a ‘missing not at random’ (MNAR) pattern and are therefore problematic. Cases with more than 50% missing data (cases 83, 159, 180, and 183) are deleted from the sample. Case 89 is also deleted on the basis of not responding to person variables (likability and attitude toward the act), as this pattern is not random. For the remaining cases (43, 149, 163, 178), a group-level mean replacement is used for the missing variable value. Appendix 6.5 provides a summary of case deletions.

Normal distribution

Normal distribution has been assessed through evaluations of skewness and kurtosis, and the Kolmogorov-Smirnov statistic. Normal distribution is assessed at the group level for dependent variables. Several variables violate the assumption of normal distribution. Violations of normality in multivariate analyses of variance, according to Hair, Anderson, Tatham and Black, have “…little impact” (1992, p160) on results.

241

The single within-subjects variable (covariate: Identification with Team) is significantly skewed. This distribution suggests that despite ABS statistics suggesting that rugby league is the most popular sport in Queensland by match attendance (ABS, 2009a); there is an overall low level of identification with the local NRL team.

Variable N skewness kurtosis Kolmogorov- p Smirnov Z Identification 195 1.029 -.179 2.257 .000 with Team

The normality of dependent variables was tested at group level; these results also show significant deviations from normality. The patterns of skewness and kurtosis will be discussed at the group level, with accompanying tables of test results.

Group A: Pub brawl assault

This group exhibits the largest positive skew result for person variables, indicating that it is likely to be the most negatively valenced case. Large kurtosis values provide that there is little dispersion on the valence of the behaviour. Distribution of corporate evaluations (Team and Sponsor variables) is normal.

Table 6.42: Assumptions: Normal distribution: Group A Variable N skewness kurtosis Kolmogorov- p Smirnov Z Likable 62 .957 .977 1.613 .011 Att- ACT 62 .988 .513 1.495 .023 Team CI 62 -.106 -.355 .774 .587 Team Att 62 -.260 .153 .778 .581 Spons CI 62 .330 -.345 1.062 .210 Spons Att 62 -.145 1.015 1.157 .137

Group B: Driving Under the Influence of Alcohol

The Likable measure for this group is negatively skewed, indicating a degree of positivity in person perception; the negative kurtosis provides a flat curve and broader dispersion of results. Attitude toward the Act is an exaggerated example of the Likable pattern.

242

Table 6.43: Assumptions: Normal distribution: Group B Variable N skewness kurtosis Kolmogorov- p Smirnov Z Likable 66 -.323 -.711 1.640 .009 Att -ACT 66 -.110 -1.443 1.345 .054 Team CI 66 -.273 .673 1.185 .121 Team Att 66 .036 .687 1.304 .067 Spons CI 66 -.755 2.081 .981 .291 Spons Att 66 -.246 1.440 1.404 .039

The two corporate image measures have provided the normal distributions associated with this treatment; their skewness and kurtosis results are very similar.

Group C: Cancer charity fundraising

This group is the positively valenced example of off-field behaviour tested in this research. Three variables exhibit normal distribution. The person variables demonstrate negative skew through positive evaluations. Kurtosis on these variables indicates little dispersion. The corporate image variables exhibit similar results, with negative skew and positive kurtosis. Attitude variables are substantially less positive than corporate image results.

Table 6.44: Assumptions: Normal distribution: Group C Variable N skewness kurtosis Kolmogorov- p Smirnov Z Likable 67 -.800 .238 2.152 .000 Att-ACT 67 -.203 .196 1.336 .056 Team CI 67 -.639 .690 1.143 .147 Team Att 67 -.115 -.443 1.256 .085 Spons CI 67 -.744 1.793 1.521 .020 Spons Att 67 -.173 .472 1.290 .072

Discussion of normal distribution results

Tests of normal distribution have provided consistent results. The person likability variable is not normally distributed for any group. Team variables are normally distributed for all groups. Finally, tests of sponsor variables and attitude toward the act are normally distributed for some groups; there is no clear pattern to these results. The skewness and kurtosis for attitude toward the act across groups reflects the nature of the treatment information, much of which is both potentially emotive and

243

also strongly valenced. This outcome is not unexpected. Fiske (1980) and Skowronski and Carlston (1987, 1989) report skewed data in research on person impressions. Kanouse and Hanson (1972) also argued that “the tails of the psychological distribution are not symmetrical” (Fiske, 1980, p891).

Data transformation, usually recommended in cases where distributions violate the assumption of normality, is not an option for this data set, for two reasons. The first; transformation reduces the generalisability of results to the sample population (Tabachnick & Fidell, 2007). The second; transformations would need to be applied to the group-level subset of data that is not normal, i.e., Group B (DUI) data for Attitude toward the Sponsor, but not Group A or C data for the same variable. It also argues that the person likability, with distribution that is not normal across all groups, would face three separate transformations.

Outliers: Univariate & Multivariate

Outliers are assessed on both univariate and multivariate bases; tests are conducted at the group level. According to Hair et al (1992), multivariate analyses are extremely sensitive to outliers; on this basis, all outliers are deleted from further analyses.

The test for univariate outliers is an extreme standardised score (z score > 3.29, p<.001) (Tabachnick & Fidell, 2007). Three univariate outliers are identified (tabled below). These were deleted from the data set prior to the test for multivariate outliers.

The Mahalanobis distance test for multivariate outliers uses the χ2 distribution with degrees of freedom represented by the number of variables tested (Tabachnick & Fidell, 2007). Six dependent variables and the covariate variable were tested; the critical value for multivariate outliers, at alpha .001, is 24.322; no multivariate outliers are present.

244

Table 6.45: Assumptions: Outliers by case, group and variable Respondent Group*variable Z score Reason 149 DUI*Spons CI -3.57964 Extreme low score 64 Assault*Att-ACT 3.33229 Extreme high score 100 Assault*Spons Att -3.30877 Extreme low score Z- scores greater than 3.29 , p<.001 Mahalanobis critical value for 7 variables p<.001 = 24.322

Linear relationships among variables and covariates

Group-level matrices of bivariate scatterplots are provided in Appendix 6.6. Linearity of relationships between pairs of variables constitutes an important assumption of MANCOVA analyses; moderate violations of this assumption occur in the current research. Corporate variables have relationships that are linear. The relationship of person variables to all other variables is weaker. Identification shows a strong linear relationship with team-based variables.

Multicollinearity & Singularity

Multicollinearity and singularity may be assessed using a variety of methods, including correlations, condition indices, squared multiple correlations (SMCs), variance inflation factors (VIFs) and tolerance scores (Tabachnick & Fidell, 2007). Correlation coefficients are used for this research, shown below in Table 6.46.

Table 6.46: Assumptions: Pearson correlations: Dependent variables Likable Team Spons Att-ACT Team CI Att CI Pearson r .701** Att-ACT p .000 n 192 Pearson r .386** .221** Team CI p .000 .002 n 192 192 Pearson r .335** .161* .674** Team Att p .000 .026 .000 n 192 192 192 Pearson r .132** .016 .280** .279** Spons CI p .069 .825 .000 .000 n 192 192 192 192 Pearson r .157** .011 .264** .353** .744** Spons p .029 .876 .000 .000 .000 Att n 192 192 192 192 192 **. Correlation is significant at the 0.01 level (2-tailed). *. Correlation is significant at the 0.05 level (2-tailed).

245

Tabachnick and Fidell (2007) warn that bivariate correlations of .70 and above indicate considerable similarity; although, “The statistical problems created by singularity and multicollinearity occur at much higher correlations (.90 and above)” (Tabachnick & Fidell, 2007, p90). The correlations among variable pairs are acceptable.

Reliability of variables

Reliability of the covariate and dependent variables in MANCOVA analyses is required as unreliable measures “lead to a loss of power and a conservative statistical test due through underadjustment of the error term” (Tabachnick and Fidell, 2007, p203). Table 6.47 provides Cronbach’s alphas for each of the multi-item dependent variables and covariate.

Each of these variables exceeds the .80 level of reliability specified by Tabachnick and Fidell (2007), indicating they are sufficiently reliable for inclusion in analyses of experimental research.

Table 6.47: Assumptions: Scale reliabilities: Cronbach’s alpha Variable n Excluded Items Cronbach’s alpha ID 187 5 5 .932 Att-Act 191 1 3 .910 Team CI 189 3 5 .913 Team Att 191 1 3 .961 Spons CI 189 3 5 .884 Spons Att 192 0 3 .951

Conclusions

The tests of assumptions have resulted in a number of violations; some of these have been remedied, others are recognised for their implications to the multivariate analyses. A summary of tests conducted, violations and remedies is provided in Appendix 6.7; they are also discussed briefly here.

Sample sizes are equivalent across groups, and missing values were remedied to 246

maintain that equivalence. Univariate outliers were removed, also while maintaining sample size equivalence.

Correlations among dependent variables are moderately high within pairs of person, team and sponsor variables. Correlations have not reached or exceeded r ≥.90, the level at which high correlations cause problems in statistical analyses. All dependent variables are retained for multivariate analyses.

All multi-item measures are reliable according to tests of Cronbach’s alphas. All alphas exceed the .80 level required for covariates and dependent variables in multivariate analyses.

In relation to the normality of distributions and linearity among variables; each of the stimulus groups has two variables with non-normal distribution. These have not been remedied. Without transformation, linearity among pairs, viewed in scatterplots (Appendix 6.6), is moderate, but not strong. Transformation of scores, which would improve linear relationships would also hinder interpretation and limit generalisability of results (Tabachnick & Fidell, 2007). In the following sections confounds are assessed prior to final multivariate analysis.

Confound Model Specification

The MANCOVA model specified contains the following elements: Independent variables: Group/ Gender/ Language group Covariate: Identification with the Team Dependent variables: Person, Team and Sponsor variable pairs

Main effects are sought for each of the independent variables. Gender and Language group are assessed as potential confounds to the analysis only; there are no hypotheses associated with these variables. Interaction effects are sought to provide detail of the role of each potential confound to influence dependent variables with stimulus group and covariate variable.

247

The interactions sought are: Gender Language group Gender*Group Language*Group Gender*ID Language*ID

Tests of Confounds

Gender and language group are considered potential confounds to the analysis. The purpose of this test of main effects is to evaluate the influence of various categorical variables on the model specified, and the extent to which groups denoted by individual characteristics (gender/ language) constitute a distinct sample population.

The multivariate model is significant according to all multivariate test statistics. Pillai’s trace result (computed using alpha .025) provides the multivariate significance of the model as .000 (F=7.363, df = 12; 338.0). The result has sufficient observed power of 1.0, and partial η2 of .207.

The tests of main effects (Table 6.48) reveal no main effect of gender (F = 6.301, df = 6, p = .119). However, this result has insufficient observed power (.530) to disconfirm a main effect with confidence. Interaction effects of gender and group (F= 1.277, df = 12, p = .230) and gender and identification (F= 1.246, df = 6, p = .285) are not significant, and underpowered.

Language has produced a significant main effect (F = 2.175, df = 6, p = .048); however, this result has also failed to achieve sufficient observed power (achieved power = .664) for confidence. The interaction effect of language with group (F= 2.502, df = 12, p = .004) is significant and sufficiently powered (observed power = .949); partial η2 for this result is .082.

The interaction effect of language with identification is not significant and underpowered (F= 1.971, df = 12, p= .073). Between-subjects effects on dependent variables show that language differences are significant for person variables, but not corporate variables (see: Table 6.49 below). These differences are explored in three

248

further post-hoc tests.

Group*Language effects on Person variables: Post-hoc tests

Significant differences in dependent variable scores due to the interaction of language and stimulus group are found through group-level analysis. The results of these are tabled below (Table 6.50 Language*Group effects on Athlete variables).

The post-hoc comparison of person variable means shows significant differences in person likable and attitude toward the act scores for groups A and C that are associated with language groups. There is no significant difference between language group means for the DUI case.

Two interpretations of the means are possible. The first: that language groups differ in evaluation of extreme scenarios; of which, the assault case and the fundraising for charity case are representative. The second alternative argues that the ‘English as second language’ group has not differentiated among negative scenarios in scores on likability and attitude toward the act variables (means: .490, .445, .451, .445).

The comparison of language groups is hampered by unequal group sizes, which renders the results of the statistical tests uncertain. Overall, the results suggest meaningful group-based differences which deserve further investigation. In this research, the ‘English as a second language’ sub-group is excluded from further analyses because results suggest this group should be studied as a distinct population.

249

Table 6.48: Confounds: Tests of Main Effects of Gender & Language Effect Value F Hypothesis Error df p Partial η2 Noncentrality Observed df parameter powerb Pillai’s Trace .058 1.721a 6.000 168.000 .119 .058 10.327 .530 Gender Wilks’ lambda .942 1.721a 6.000 168.000 .119 .058 10.327 .530 Hotelling’s Trace .061 1.721a 6.000 168.000 .119 .058 10.327 .530 Roy’s Largest Root .061 1.721a 6.000 168.000 .119 .058 10.327 .530 Pillai’s Trace .087 1.277 12.000 338.000 .230 .043 15.323 .615 Gender* Wilks’ lambda .914 1.281a 12.000 336.000 .228 .044 15.368 .617 Group Hotelling’s Trace .092 1.284 12.000 334.000 .226 .044 15.411 .619 Roy’s Largest Root .075 2.118c 6.000 169.000 .054 .070 12.705 .649 Pillai’s Trace .043 1.246a 6.000 168.000 .285 .043 7.477 .368 Gender* Wilks’ lambda .957 1.246a 6.000 168.000 .285 .043 7.477 .368 ID Hotelling’s Trace .045 1.246a 6.000 168.000 .285 .043 7.477 .368 Roy’s Largest Root .045 1.246a 6.000 168.000 .285 .043 7.477 .368 Pillai’s Trace .072 2.175a 6.000 168.000 .048 .072 13.052 .664 Language Wilks’ lambda .928 2.175a 6.000 168.000 .048 .072 13.052 .664 Hotelling’s Trace .078 2.175a 6.000 168.000 .048 .072 13.052 .664 Roy’s Largest Root .078 2.175a 6.000 168.000 .048 .072 13.052 .664 Pillai’s Trace .163 2.502 12.000 338.000 .004 .082 30.019 .949 Language* Wilks’ lambda .842 2.516a 12.000 336.000 .004 .082 30.187 .951 Group Hotelling’s Trace .182 2.529 12.000 334.000 .003 .083 30.352 .952 Roy’s Largest Root .138 3.894c 6.000 169.000 .001 .121 23.366 .936 Pillai’s Trace .066 1.971a 6.000 168.000 .073 .066 11.823 .607 Language* Wilks’ lambda .934 1.971a 6.000 168.000 .073 .066 11.823 .607 ID Hotelling’s Trace .070 1.971a 6.000 168.000 .073 .066 11.823 .607 Roy’s Largest Root .070 1.971a 6.000 168.000 .073 .066 11.823 .607 a. Exact statistic b. Computed using alpha = .025

250

Table 6.49: Confounds: Between-subjects Effects of Language*Group Effect Dependent variable Type III Sum df Mean F p Partial η2 Noncentrality Observed of Squares Square parameter powera Likable .446 2 .223 7.269 .001 .078 14.537 .888 Language* Att-ACT .501 2 .251 8.066 .000 .085 16.132 .921 Group Team CI .029 2 .014 .773 .463 .009 1.545 .114 Team Att .004 2 .002 .070 .933 .001 .139 .031 Spons CI .005 2 .002 .132 .876 .002 .264 .038 Spons Att .011 2 .006 .213 .808 .002 .426 .046 a. Computed using alpha = .025

Table 6.50: Confounds: Comparison of Means: Athlete variables Group Variable Means English Std. Non- Std. t df Sig. Mean difference Std. Error Error English Error

Assaultb Likable .317a .026 .478a .052 -3.085 56 .003 -.16790 .05442 Att-ACT .272a .027 .434a .052 -3.027 56 .004 -16895 .05581 n 46 12

DUIc Likable .516a .026 .453a .042 1.147 63 .256 .06064 .05287 Att-ACT .347a .026 .445a .043 -1.826 63 .073 -.09137 .05003 n 48 17

Cancer Likable .756a .024 .627a .053 2.006 63 .049 .13256 .06609 Charityd Att-ACT .763a .024 .612a .054 2.621 63 .011 .14885 .05680 n 53 12 a. Covariates appearing in the model are evaluated at the following values: id_prop = .3375.

251

Final analysis: confounds removed

The model specified contains the following elements: Independent variable (Group); covariate (Identification with the Team); and, dependent variables: person, team and sponsor variable pairs.

Main effects are sought for treatment group (Group) and for the covariate (Identification with Team). The interaction between the Group independent variable and the covariate are sought to test the assumption of homogeneity of regression test, which would be violated if significant. The results of the homogeneity of regression test and tests of homogeneity of variance-covariance matrices are provided following sample characteristics, and prior to multivariate results.

Figure 6.19: Graph of variable means: Study 2

252

Summary of Sample Characteristics

During data screening and tests of assumptions several changes have been instituted. Characteristics of the sample are now summarised below. Changes instituted are summarised in Appendix 6.5: Summary of case deletions.

Categorical variables Label n % Total Male 66 43.7 Gender Female 79 52.3 Missing 6 4.0 151 Assault 48 31.8 Group DUI 48 31.8 Cancer charity 55 36.4 151

The ages of respondents range from 17-50 years. The sample mean age is 21.91 with a mode of 19 years; the standard deviation for this variable is 5.963. There were 5 missing values for the age variable.

Assumptions tested during multivariate analysis

Homogeneity of variance-covariance matrices

The multivariate test for homogeneity of variance-covariance matrices is Box’s M, which tests for the “equivalence of covariance matrices across the groups” (Hair, Black, Babin, Anderson & Tatham, 2006, p409). Box’s M is very sensitive to departures from normality, which has led to some authors recommending relaxation of alpha from .05 to .01 (Hair et al, 2006; Tabachnick & Fidell, 2007). Box’s M test for this data is shown below:

Table 6.51: Assumptions: Box’s M Box’s M F df1 df2 p 80.934 1.815 42 63116.339 .001

The univariate test for homogeneity of error variance is Levene’s test. Where Box’s M is not robust at .01, a review of the univariate test is useful to gauge the extent of the violation. Levene’s test of error variances is shown below:

253

Table 6.52: Assumptions: Levene’s test of error variances Variable F df1 df2 p Likable 6.038 2 148 .003 Att-ACT .568 2 148 .568 Team CI 2.874 2 148 .060 Team Att 1.856 2 148 .160 Spons CI 5.238 2 148 .006 Spons Att 1.692 2 148 .118

Levene’s test is not sensitive to departures from normal distribution, which explains the substantial difference in results from Box’s M. Levene’s test shows that equality of error variances across groups is found for all attitude measures. The null hypothesis is violated for the sponsor corporate image variable and the person likability measure.

Violation of the homogeneity of variance-covariance matrices assumption is not fatal to MANCOVA analysis. Tabachnick and Fidell (2007) recommend that when transformation is not used; applying a more stringent test alpha is appropriate. Specifically, they recommend use of alpha “.025 with moderate violation and .01 with severe violation” (Tabachnick & Fidell, 2007, p86). Box’s M for this analysis, and Levene’s univariate results suggest violations of the homogeneity of variance- covariance matrices assumption. As a consequence, the test alpha is modified.

Modified Test alpha

The test alpha used in this analysis is .0125. Violation of the assumption of homogeneity of variance-covariance matrices was expected given the violations of normal distribution found in group-level tests of assumptions.

Homogeneity of Regression

The assumption of homogeneity of regression is that the “groups do not differ in the regression of the DV [dependent variable] on Cov [covariate]” (Miller & Chapman, 2001, p43). A violation of this assumption threatens the premise that measurement of

254

the covariate is independent of the treatment. Homogeneity of regression is tested through significance testing of interaction effects among the covariate and treatment groups. The assumption of homogeneity of regression is not violated in this research.

Table 6.53: Assumptions: Homogeneity of Regression Dependent Type III df Mean F p variable Sum of Square Squares Likeable .056 2 .028 1.009 .367 Att-ACT .054 2 .027 .911 .405 Team CI .097 2 .048 2.426 .092 Group*ID Team Att .066 2 .033 1.128 .326 Spons CI .010 2 .005 .243 .785 Spons Att .048 2 .024 .862 .424

Omnibus F: the corrected model

One premise of this study is that if news-media reports of the off-field behaviours of sportspeople influence news’ consumer opinions, then effects of news’ will be distributed equally significantly across athlete, team, sponsor variables. The omnibus result for this study does not produce the minimum significant effect on each of the groups of variables. The result for the omnibus test is provided in Table 6.54.

Table 6.54: Omnibus F for the Corrected Model Dependent Type df Mean F p Partial Noncent. Observed variable III Sum Square η2 parameter powerb of Squares Likeable 5.950a 5 1.190 42.684 .000 .595 213.421 1.000 Att-ACT 7.104c 5 1.421 48.028 .000 .624 240.142 1.000 Team CI .894d 5 .179 8.944 .000 .236 44.719 .999 Team Att 2.609e 5 .522 17.723 .000 .379 88.614 1.000 Spons CI .026f 5 .005 .264 .932 .009 1.321 .038 Spons Att .081g 5 .016 .577 .717 .020 2.885 .085 a. R Squared = .595 (Adjusted R Squared = .581) b. Computed using alpha = .0125 c. R Squared = .624 (Adjusted R Squared = .611) d. R Squared = .236 (Adjusted R Squared = .209) e. R Squared = .379 (Adjusted R Squared = .358) f. R Squared = .009 (Adjusted R Squared = .025) g. R Squared = .020(Adjusted R Squared = .014)

Although the specified model has produced significant effects on the team and person variables, results argue against effects on sponsor variables.

The partial η2 indicates that the model has very small effects on sponsor variables.

255

The observed power for sponsor variable results is insufficient to declare no effect on sponsor variables with certainty. The combination of, p-value, effect size and observed power suggests that a very large sample size would be needed to judge the question of news-media effects on sponsor evaluations with certainty. On this basis, the results suggest that although the model has had significant effects on person and team variables; there is insufficient information to argue an effect on sponsor variables.

Multivariate effects: Significance of treatment effects

In this section the multivariate results are provided. These show effects on the dependent variables produced by treatment group and the covariate. Table 6.55 provides the multivariate tests.

The multivariate tests of the model suggest sufficient statistical power for the independent variable and covariate in tests of the model. The test results also show that in multivariate analyses, the effect sizes are large for the covariate (η2>.25), and medium for the treatment group (η2≤.09). The results also suggest that the independent variable and the covariate have produced significant effects on at least one of the dependent variables.

The interaction of Group and Identification is not significant (p>.0125), confirming that the homogeneity of regression assumption has not been violated.

Between-subjects effects & Tests of Hypotheses

Results of the between-subjects tests provide the tests of the hypotheses for this study. Three groups of hypotheses were developed. Hypothesis 1 assesses the effect of treatment group. Hypothesis 2 assesses effects of the covariate variable, and Hypothesis 3 assesses whether evaluations are balanced. The structure of this section will be to evaluate each of these hypotheses (and sub-hypotheses), in order.

256

Hypothesis 1: Effects of news’ on evaluations

Hypothesis 1 tests whether treatment group has influenced evaluations of athlete, team and sponsor variables; this hypothesis is partially supported. News’ article stimuli significantly influence perceptions of the athlete, but not team or sponsor variables.

H1a is accepted, (F=40.388, p<.0125), partial η2 suggests that treatment group has a large effect on Attitude toward the Act (partial η2 =.358). The observed power for this result is acceptable. H1b is also accepted (F=35.142, p<.0125). Treatment group produces a large effect size for the Person Likable variable, (partial η2= .326) which has acceptable power.

Hypotheses H1c-H1f are rejected. Treatment group has not influenced evaluations of team or sponsor variables in this study.

H1c: Valence of news’ will be reflected in evaluation of the team (Attitude toward the Team). H1d: Valence of news’ will be reflected in evaluation of the image of the team (Team Corporate Image). H1e: Valence of news’ will be reflected in evaluation of the sponsor (Attitude toward the Sponsor). H1f: Valence of news’ will be reflected in evaluation of the image of the sponsor (Sponsor Corporate Image).

257

Table 6.55: Multivariate Tests of Main Effects Effect Value F Hypothesis Error df p Partial η2 Noncentrality Observed df parameter powerb Pillai’s Trace .911 238.011a 6.000 140.000 .000 .911 1428.064 1.000 Intercept Wilks’ lambda .089 238.011a 6.000 140.000 .000 .911 1428.064 1.000 Hotelling’s Trace 10.200 238.011a 6.000 140.000 .000 .911 1428.064 1.000 Roy’s Largest Root 10.200 238.011a 6.000 140.000 .000 .911 1428.064 1.000 Pillai’s Trace .539 8.671 12.000 282.000 .000 .270 104.050 1.000 Group Wilks’ lambda .484 10.213a 12.000 280.000 .000 .304 122.555 1.000 Hotelling’s Trace 1.020 11.812 12.000 278.000 .000 .338 141.744 1.000 Roy’s Largest Root .971 22.821c 6.000 141.000 .000 .493 136.927 1.000 Pillai’s Trace .411 16.287a 6.000 140.000 .000 .411 97.724 1.000 ID Wilks’ lambda .589 16.287a 6.000 140.000 .000 .411 97.724 1.000 Hotelling’s Trace .698 16.287a 6.000 140.000 .000 .411 97.724 1.000 Roy’s Largest Root .698 16.287a 6.000 140.000 .000 .411 97.724 1.000 Pillai’s Trace .101 1.248 12.000 282.000 .250 .050 14.976 .495 Group*ID Wilks’ lambda .901 1.248a 12.000 280.000 .250 .051 14.978 .495 Hotelling’s Trace .108 1.248 12.000 278.000 .250 .051 14.979 .495 Roy’s Largest Root .082 1.939c 6.000 141.000 .079 .076 11.631 .489 a. Exact statistic b. Computed using alpha = .0125 c. The statistic is an upper bound on F that yields a lower bound on the significance level.

258

Table 6.56: Between-subjects effects of Group, Covariate & Interactions Effect Dependent Type III Sum df Mean F p Partial η2 Noncentrality Observed variable of Squares Square parameter powera Likable 7.233 1 7.233 259.457 .000 .641 259.457 1.000 Intercept Att-ACT 7.685 1 7.685 259.775 .000 .642 259.775 1.000 Team CI 13.047 1 13.047 652.966 .000 .818 652.966 1.000 Team Att 8.304 1 8.304 282.077 .000 .660 282.077 1.000 Spons CI 19.669 1 19.669 1002.215 .000 .874 1002.215 1.000 Spons Att 19.286 1 19.286 689.070 .000 .826 689.070 1.000 Likable 1.959 2 .980 35.142 .000 .326 70.284 1.000 Group Att-ACT 2.390 2 1.195 40.388 .000 .358 80.775 1.000 Team CI .045 2 .023 1.138 .323 .015 2.276 .109 Team Att .057 2 .029 .976 .379 .013 1.952 .091 Spons CI .020 2 .010 .514 .599 .007 1.027 .048 Spons Att .057 2 .029 1.026 .361 .014 2.051 .097 Likable .972 1 .972 34.879 .000 .194 34.879 1.000 Identification Att-ACT .145 1 .145 4.910 .028 .033 4.910 .380 with Team Team CI .662 1 .662 33.113 .000 .186 33.113 .999 Team Att 2.431 1 2.431 82.575 .000 .363 82.575 1.000 Spons CI .002 1 .002 .106 .746 .001 .106 .017 Spons Att .028 1 .028 .996 .320 .007 .996 .066 Likable .056 2 .028 1.009 .367 .014 2.017 .095 Group*ID Att-ACT .054 2 .027 .911 .405 .012 1.821 .085 Team CI .097 2 .048 2.426 .092 .032 4.853 .278 Team Att .066 2 .033 1.128 .326 .015 2.257 .108 Spons CI .010 2 .005 .243 .785 .003 .486 .027 Spons Att .048 2 .024 .862 .424 .012 1.724 .080 a. Computed using alpha = .0125

259

On the basis of significance levels alone, Team Corporate Image (F=1.138, p>.0125) and Attitude toward the Team (F=.976, p>.0125) are unaffected by treatment group. The partial η2 suggest that if effects of treatment group on team variables were to be found, they would be small (partial η2 ≤.01). Tests of these hypotheses are also insufficiently powered (observed power <.75), arguing that a much larger sample size is needed to establish a significant small effect of treatment group on team variables.

The sponsor variables (Corporate Image and Attitude toward the Sponsor) are not affected by treatment. The result for Sponsor Corporate Image (F=.514, p = .599) is not significant, and under-powered (observed power .048). Attitudes toward the Sponsor are not affected by treatment group (F=1.026, p>.0125) with non-significant p (.361), small effect size and limited observed power.

Hypothesis 2: Effects of Identification

Main effects of the covariate on evaluations of the athlete, team and sponsor are tested by Hypothesis 2. Results are mixed.

Results for Hypotheses 2a-2b, effects of the covariate on person variables, are mixed. Hypothesis 2a is rejected. Hypothesis 2b is supported.

H2a: Team Identification will moderate evaluations of the athlete’s behaviour (Attitude toward the Act). H2b: Team Identification will moderate evaluations of the athlete (Person likability).

Hypothesis 2a is rejected. The effect of Identification on Attitude toward the Act (H2a) is not significant (F=4.910, p>.0125); the result is under-powered (<.75). To accept the significant effect of Identification on Attitude toward the Act would risk a false rejection of the null hypothesis. The small partial η2 suggests that a larger sample size is needed to confirm effects of Identification on Attitude toward the Act.

260

Hypothesis 2b is supported. The effect of Identification on Person Likability (H2b) is significant (F=34.879, p<.0125). This result has a large amount of observed power (>.80) and a medium effect size (partial η2 = .194).

Hypotheses 2c and 2d, effects of Identification on team evaluations are accepted.

H2c: Team Identification will moderate evaluations of the team (Attitude toward the Team). H2d: Team Identification will moderate evaluations of the image of the team (Team Corporate Image).

Identification with the Team significantly affects both Team Corporate Image (F=33.113, p<.0125) and Attitude toward the Team (F=82.575, p<.0125). These results have sufficient observed power. The partial η2 results are .186 for Team Corporate Image, and .363 for Attitude toward the Team; these are medium and large effect sizes, respectively.

Identification does not have a significant effect on sponsor variables (Hypotheses 2e and 2f). Hypotheses 2e and 2f are rejected.

H2e: Team Identification will moderate evaluations of the sponsor (Attitude toward the Sponsor). H2f: Team Identification will moderate evaluations of the image of the sponsor (Sponsor Corporate Image).

The effect of the covariate on Attitude toward the Sponsor (Hypothesis 2e) is not significant (F=.996, p>.0125). The effect size of Identification with the Team, is very small (partial η2 <.007). Observed power for this result (.066) is insufficient for a confident endorsement of the alternative hypothesis.

The covariate has not produced a significant effect on Sponsor Corporate Image (F=.106, p>.0125); the partial η2 result suggests that the contribution of Identification to Sponsor Corporate Image variance is very small (partial η2 <.01). To confirm this

261

result would require a larger sample size. Observed power for this result, is correspondingly low (<.75).

Hypothesis 3: the balanced model

H3: The valence of the reported behaviour (positive/ negative) will be maintained across evaluations of the athlete, their team, and the sponsor.

Hypothesis 3 is receives limited support.

Perceptual balance among groups of four has been explained by Hummert, Crocker and Kemper (1990). The omnibus F result (Table 6.54) and tests of Hypotheses 1 and 2 (see hypotheses not supported: H1c-1f; H2a, H2e-f) indicate that the research model has not produced effects on sponsor variables. These results argue that neither research model B (i) nor B (ii) explain evaluations in this research.

The balance hypothesis is discussed in the following sections, according to significant results in multivariate testing.

Group-level means & tests of differences

In this section means of the dependent variables that have been significantly affected by treatment group and the covariate variable are assessed. Variables affected by treatment group (person variables) are reviewed first.

Effects of Group

These means (provided in Table 6.57) show that person likability is significantly affected by treatment group. The likability means for the scenarios are all significantly different from each other (p ≤ .000 for all comparisons). Two of these means (assault and charity fundraising) are significantly different from the scale mid-

262

Table 6.57: Pairwise comparison of means: Person variables Pairwise comparisons Difference from mid-point (0.57) Difference from other treatment means Est. Mean Mean Std. DV Marg. t df Sig. Group (I) Group (J) Diff. Sig.b diff. Err. Mean (I-J)

.322a -10.897 47 .000 -.25083 Assault DUI -.195* .034 .000 Charity -.430* .033 .000 Likable .518a -2.249 47 .029 -.05583 DUI Assault .195* .034 .000 Charity -.235* .033 .000 .753a 6.477 54 .000 .18636 Charity Assault .430* .033 .000 DUI .235* .033 .000

.287a -11.905 47 .000 -.28542 Assault DUI -.068 .035 .159 Charity -.472* .034 .000 Att-Act .354a -8.567 47 .000 -.21667 DUI Assault .068 .035 .159 Charity -.404* .034 .000 .758a 7.812 54 .000 .18964 Charity Assault .472* .034 .000 DUI .404* .034 .000 *. The mean difference is significant at the .0125 level. a. Covariates appearing in the model are evaluated at the following values: id_prop = .3386. b. Adjustment for multiple comparisons: Sidak.

263

point (.570), for each, p is ≤.000. The DUI mean is not significantly different from the scale mid-point (p > .0125). Tests of means establish that likability for assault is negative; likability for charity fundraising is positive, and DUI is neutral.

The means for the attitude toward the act variable show no significant difference between evaluations of assault and DUI (p > .0125). These means are also significantly different from the scale mid-point (p ≤ .000); mean differences show that each of these cases is negative in valence. The charity fundraising scenario mean is significantly different to both assault and DUI (p≤ .000); it is also significantly different from the scale mid-point, and therefore of positive valence.

Treatment group balance

Group stimuli significantly influence evaluations of the athlete only. Figure 6.20 below, shows the expected effects of treatment group on all entities. The figure also provides a general model of the effects of treatment group established by the analysis. Results argue consistency among evaluations of likability and attitude toward the act which establishes a balanced view of the person.

Figure 6.20: Balance: Hypothesised & Actual Effects of Group

Athlete: Likability Athlete (p≤ .000)

DL/ L DL/ L U U

U DL/ L p Team

DL/ L DL/ L U

Athlete: Attitude toward Sponsor the Act (p≤ .000)

Results: Significant effects of Group Hypothesised effects of Group (Person variables)

264

Group-level models are provided in Figure 6.21 which show the variable means by treatment group. The model for group A, (assault) is balanced with two negative evaluations and the unit relationship linking likability and attitude toward the act. The model for group C, charity fundraising, is a balanced model of positive valence.

Figure 6.21: Balance: Effects of Group on Evaluations of the Athlete

Likable mean: .322 Likable mean: .518 Likable mean: .753

D N L L U U U p p p

D DL L L

Att-ACT mean: .287 Att-ACT mean: .354 Att-ACT mean: .758

Assault DUI Charity

The DUI scenario is not balanced according to the criteria developed by Heider (1958). The person likability mean is equivalent to the scale mid-point. The measure of attitude toward the act is negatively valenced. Heider’s (1958) criteria do not refer to neutral evaluations; despite this, it should be considered that where models do not conform to criteria for balance, they are necessarily, not balanced. Imbalanced models, according to Heider (1958) and others (see: Dalakas & Levin, 2005; Dardis, 2009; Dean, 2002; Osgood, 1960) are evidence of cognitive stress, and provide pressure on the individual to alter their evaluations. At the level of the individual respondent, evaluations of the DUI case may well be balanced. At the aggregate level, however, this analysis argues that consensus in judgments of DUI is unlikely.

To provide further insight into the above models, the means of person likability and attitude toward the act are compared at the level of the group. Paired t-tests (test alpha = .0125) establish that there is no difference in the means of likability and attitude toward the act for groups A and C (assault and charity). This confirms that

265

both person means for the assault case are negative, and both means for the charity fundraising case are positive.

Table 6.58: Paired t-tests: Person variables by Group Person Attitude Sig. Mean Group Likability toward the Act t df (two- difference mean mean tailed) Assault .3192 .2846 .03458 1.569 47 .123 DUI .5142 .3533 .16083 5.398 47 .000 Charity .7564 .7596 -.00327 -.137 54 .892

Tests of mean difference for the DUI case, however, establish that the person likability mean is significantly different from the mean of DUI attitude toward the act. This result adds credence to the likelihood of cognitive stress in evaluations of DUI.

Effects of Identification

Three variables are affected by identification with the team; these are person likability, team corporate image, and attitude toward the team. Evaluations of the team are unaffected by treatment group, it is therefore unnecessary to establish differences between means at the group level. Because group-level differences have not occurred, team corporate image and attitude toward the team variable means are tested against the scale mid-point using variable-level means. These t-tests show that team variables are significantly different from the scale mid-point (p≤ .000); mean differences establish that evaluations of these variables are positive (see: Table 6.59: t-tests of team variables).

Table 6.59: Test of difference: Team variables Difference from mid-point (0.57) Sig. (two- DV Variable Mean t df Mean difference tailed)

Team CI .6497 6.158 150 .000 .07967

Team Att .6330 3.618 150 .000 .06306

266

Identification with the Team balance

Identification with the Team significantly influences person likability and team variables. Figure 6.22, shows the expected and actual effects of identification.The model of identification effects shows the potential for imbalance. Two positive evaluations exist: the athlete/team unit relation; and the positive evaluations of team variables. Evaluations of person likability are a source of model stress.

Figure 6.22: Balance: Hypothesised and Actual Effects of Identification

Athlete: Athlete ID *Likability (p≤ .000)

DL/ L DL/ L U U

DL/ L U p Team

L DL/ L U

Team: Corporate Image (p≤ .000) Sponsor Attitude toward the Brand (p≤

Results: Significant effects of Hypothesised effects of Identification

Figure 6.23, below, shows the group-level models for effects of identification with the team. Teamm-Att and team CI evaluations are constant across these models because treatment stimuli have not affected these variables. Group-level consideration is warranted because of the group effect on person likability.

Using Heider’s criteria, the model for group A (assault) has two positive relations (unit relation between athlete and team; and positive sentiment relation between perceiver and team) and one negative relation (perceiver evaluation of athlete likability). Tests of means have established that person likability for this scenario is negative and that team variable means are positive in valence. The model for the

267

assault scenario is imbalanced.

Figure 6.23: Balance: Effects of Identification on Likability & Team variables

Likable mean: .322 Likable mean: .518 Likable mean: .753

D N L L U U U p p p

L L L

Team CI mean: .649 Team CI mean: .649 Team CI mean: .649 Team Att. mean: .633 Team Att. mean: .633 Team Att. mean: .633

Assault DUI Charity

The DUI scenario is also imbalanced. The neutral person likability mean is not positive. Paired t-tests establish that person likability for the DUI scenario differs significantly from team variable means, which are positively valenced. This model should be regarded as imbalanced because the valence of each of the possible relationships is not the same.

Table 6.60: Tests of difference: DUI means Person Sig. Mean Group Likability Team means t df (two- difference mean tailed) Team CI DUI .5142 .6273 -.11312 -4.723 47 .000 Team Att. DUI .5142 .6304 -.11625 -4.669 47 .000

The model for the charity fundraising scenario is balanced. Person likability for this scenario is positive, as is the perception of team variables. The unit relation between athlete and team provides a third positive relationship and provides that the model conforms to one of Heider’s criteria for balance.

268

Balance models: summary

The results of this research show that news’ article stimuli produce effects on perception of the person discussed in the article, and evaluations of the team that athlete is associated with. There are no effects on sponsor variables.

Two alternative balance models were anticipated prior to analysis. These were: model A, arguing equivalent effects of treatment group and identification with the team on all dependent variables; and, model B specifying two categories of results (treatment group effects distinct from identification effects) that would have equivalent effects on all dependent variables. Neither of these models is supported by the results.

The group-level balance models produced are summarised in Table 6.61 below. These models provide that perception of the person is balanced when likability and attitude toward the act evaluations are equivalent; or, of equal valence. Balance is not maintained for the DUI case. Two explanations are possible: extremity of acts works to balance judgments by providing ‘diagnostic’ likability information; or, the normativity of acts fails to provide diagnostic likability information. The imbalanced DUI result suggests ‘stress’ and the potential for change in evaluations.

Table 6.61: Summary of Balance models Effects of Group Attitude toward Likability Person judgments Balance the Act Assault Negative Negative Unit relation Balance DUI Neutral Negative Unit relation Imbalance Charity Positive Positive Unit relation Balance

Effects of Identification Athlete-Team Likability Team variables Balance relationship Assault Negative Positive Unit relation Imbalance DUI Neutral Positive Unit relation Imbalance Charity Positive Positive Unit relation Balance

The group-level models of identification effects are imbalanced for the assault and DUI scenarios. This has occurred because ratings of the team variables are positive, and the athlete-team contract exists as a (positive) unit relation for balance evaluation. To balance models of identification effects using Heider’s (1958) criteria,

269

the third evaluation (person likability) must be positive. Only the charity fundraising scenario provides a positive person likability evaluation. These results suggest the likelihood of cognitive stress when evaluating the negative off-field behaviours of athletes in situations characterised by identification with the team. There is no cognitive stress in evaluation of positive off-field behaviour.

Discussion of Results

The aim of this research was to test the premise that news’ reports of athlete off-field behaviours have the potential to threaten the achievement of sponsors’ goals and objectives. A range of news’ articles were developed, based on real events and real news articles, to test their impact on respondents’ evaluations of the athlete, the team and the sponsor.

The research question for this study asked: RQ4. Are consumers’ evaluations of an athlete, the athlete’s team and the team’s sponsor influenced by news-media reports detailing the athlete’s off-field behaviours? a. Does team identification moderate evaluations? b. Are evaluations balanced?

The results of the research argue that, in an experimental test of the impact of news’ articles, several forms of consumer evaluations are influenced.

Consumer’s evlauations of athlete likability and attitude toward the act were influenced by treatment group. Athlete likability was also significantly influenced by identification with the sporting team; there was no effect of identification on attitude toward the act. These results argue that respondents in this research have maintained a level of impartiality in judgment of acts depicted in the news’ articles, in that behaviour valence does not vary according to level of team identification.

Evaluations of the sporting team are significantly influence by identification with the sporting team; they are not influenced by news’ article stimuli.

270

Evaluations of sponsor variables are not affected by treatment group; nor are they affected by identification with the sporting team. These results may be the product of a weak sponsorship stimulus. In stimulus materials, the sponsor’s brand name and sponsorship relationship is signposted, but not central to the description of off-field behaviours. However, the extremely small effect sizes produced in this research argue that extremely large sample sizes would be needed to confirm any effects on sponsor’s variables. Stronger sponsorship cues should also be considered in future research.

The main research question is answered: consumer evaluations of athlete, team and sponsor variables are differently affected by treatment group and identification with the sporting team.

The sub-question, ‘Does team identification moderate evaluations?’ is answered. Team identification moderates evaluations of team variables and athlete likability. These results have implications for sporting team and athlete management.

The balance theory sub-question is also answered; the results are nuanced. The effects of treatment group on athlete variables produced balanced results for scenarios defined by positive and negative valences (attitudes toward the act). Evaluations of the athlete were not balanced when attitude toward the act was neutral in valence, the DUI scenario. The result argues the need for future research into the normativity of the behaviour.

The effects of identification with the team on evaluative balance indicate likely situations of cognitive stress for perceivers. Team variables are positively valenced. In scenarios when athlete likability is negative or neutral; perceptual balance is not present. When athlete off-field behaviours are positive, and likability consequently positive; perceptual balance is achieved. These results also have implications for sporting team and athlete management.

271

Chapter 7: Conclusions & Future Research

This research developed to address the negatives in the sponsorship environment. News-media reports detailing athlete’s negative off-field behaviours, and subsequent cancellation of sponsorship contracts added impetus to this interest. In the course of ordinary news-media reporting of sports, we expect to read descriptions of match play, team selection, training, player injuries acquired during sports, and commentary on the state of the ‘competition’. It is less common to read the deliberate curtailing of a commercial agreement or sponsorship contract. While not a quotidian problem, it is not unheard of that sponsorship cancellations are attributed to the off-field, rather than on-field, activities of athletes or teams. When cancellations occur, the language used by sponsors approached for comment argues the importance of commercial sponsorship objectives. Some of these are shown below:

 Michael Phelps filmed smoking cannabis: sacked by Kelloggs for behaviour inconsistent “…with the image of Kellogg” (BBC SPORT, 2009).  Cronulla Sharks Rugby Club dumped by sponsor LG Electronics because of “…controversies around the NRL (National Rugby League), and the Sharks in particular” (Byrnes & Phelps, 2009). The statement released by LG Electronics proclaimed the importance of, “where we want to take the LG brand” (Byrnes & Phelps, 2009).  Revelations about Tiger Woods’ extramarital affairs led to the cancellation of his sponsorship contracts by Accenture and Gillette (NY Times, 2009); and reduced media exposure by Tag Heuer (The Canadian Press, 2009; UPI Energy Resources, 2009). Accenture, in the press release announcing the end of their relationship with Woods, noted that “…his achievements on the golf course have been a powerful metaphor for business success in Accenture’s advertising” (Accenture, 2009). Gillette, prior to cancelling their contract, chose to “limit Woods’s role in its marketing” (The NY Times, 2009).

272

The goals and objectives of commercial sponsorship include (positive) brand exposure in the media, audience reach, and the development of consumer attitudes, corporate or brand image, and purchase intentions (Cameron, 2009; Cornwell, Pruitt & Van Ness, 2001; Gwinner & Eaton, 1999). Negative off-field behaviour is conceived of as a threat to the objectives of commercial sponsorship. The effects of reports of negative off-field behaviours on such objectives as share values (see: Knittel & Stango, 2010) argues the validity of this approach.

The research reported in this thesis has identified the types of off-field behaviours reported in local news’ (Study 1: content analysis). It has assessed the evaluations of behaviours reported (pretest: valence of behaviours). It has also presented off-field behaviours to respondents and asked how the athlete, their team, and importantly, sponsor’s might be evaluated (Study 2: experiment).

This chapter summarises the findings of the research conducted. Limitations of the research and opportunities for future research are highlighted. Finally, the contributions of this research to sponsorship theory and practice are identified; implications for sporting teams are also considered.

Summary of Findings: Research Questions

RQ1. What types of athlete off-field behaviours are reported?

Study 1 provides a census of newspaper reports referring to the National Rugby League (NRL) published in The Courier Mail; it covers the period January to June 2009. The NRL, as the sporting context, and the Courier Mail, as news-provider, were targeted because of their roles in the region. NRL is the most popular sporting code in Queensland by attendance (ABS, 2009a), and on that basis, expected to be the subject of news’, and sports’ consumers interest. The Courier Mail has distribution in Queensland equivalent to approximately 50% of households (www.newsspace.com.au).

273

In the period January- June 2009, 1834 newspaper articles in The Courier Mail referred to the NRL competition. Of these articles: 895 referred to competition and match-reports only; 21 reported only off-field events; 237 reported both on- and off- field events; and 681 were extraneous to this study. The large number of extraneous reports (including such topics as share prices, eulogies, betting, and references to other sports) argues the NRL competition is embedded in many facets of Queensland life. The majority of articles, however, refer to the NRL competition and match-play. This early data suggests; the business of sport... is sport.

Of the articles that referred to off-field events (n=258), these provide 430 mentions of off-field events, for 52 discrete events. It was established that when an article reports any off-field behaviour; it will typically mention more than one off-field behaviour (p<.05). This finding is the first indication that news-media reporting may intensify interest in events beyond reporting the facts. In effect, it provides the possibility of a risk multiplier.

The types of events reported are varied. Very few positive off-field behaviours are reported during the period reviewed (n=5). The overwhelming majority are negative (n=47). An attempt to categorise the negative events produced four categories of negative incidents: anti-social incidents; violent incidents; violence involving women; and breaches of club rules. Breaches of club rules are not pursued in this research. The remaining four categories (including positive behaviours) provide the information stimuli for Study 2.

The events identified in the content analysis were further considered in relation to court statistics (ABS 2008-09b), and the notion of ‘news values’ (Galtung & Ruge, 1965; Harcup & O’Neill, 2001). What this data suggests is both an under-reporting and over-reporting of events according to their ‘newsworthiness’. The notion of news values argues that events ‘make the news’ when they meet certain criteria (Galtung & Ruge, 1965). This does not suggest that criteria are formal. Instead, reporting works in the same way as ‘perceptual psychology’ (Galtung & Ruge, 1965, p66); reporters “…have to select, and the question is what will strike our attention” (Galtung & Ruge, 1965, p65).

274

Study 1 identifies the types of off-field events that are reported. It also suggests the importance of the role of the media in developing issue salience and aiding audience interpretation of events.

RQ2. Removed from news-media and sports contexts, how do individuals evaluate the off-field behaviours reported?

The second research question asks how the off-field behaviours reported are evaluated. This question is answered in survey research conducted for Pretest 1. Tow measures of behaviour valence were used in this pretest: a moral cue diagnosticity measure (Skowronski & Carlston, 1987, 1989, 1992) and a measure of Attitude toward the Act.

The pretest showed that risky or potentially harmful behaviours (e.g., assault, DUI, drug use) would be assessed as negative by respondents on both morality and Att-Act measures. The pretest also suggested that altruistic behaviours would be perceived as morally good; respondents would hold positive attitudes towards these acts.

Despite the polarisation of evaluations, the pretest results suggested significant differences in measures of morality and Attitude toward the Act. Use of the morality measure was decided against for the Study 2 experiment for two reasons. The first was that morality scores provided limited within category differentiation. The second reason for excluding the morality measure was a based on the concern that ‘morality’ in the text of questions provided a cue to respondents to evaluate acts in more ‘black or white’ terms than they would otherwise. Study 2 results suggested the likelihood of spillover from morality-cues in A-act scores: see below.

Pretest mean Study 2 mean Assault .163 .287 DUI .196 .354 Charity fundraising .854 .758

It is probable that the difference in scores may be attributable to a) the small sample

275

size for the pretest, and/or b) response cueing. This result argues the importance of appropriate measures and methods.

RQ3. If group (team) identification has the power to influence consumer evaluations, how should team identification be measured for this research?

The third research question for this research asked how team identification could be measured. Pretest 2 provides a review of extant identification scales. In survey research 20 existing identification items were tested to develop an identification scale. Two scales were produced: one allows multi-dimensional scale measurement; the second provides a 5-item unidimensional scale. Although, in terms of model fit and conceptual strength the scales are equivalent; the unidimensional scale provided the more parsimonious measure for Study 2.

RQ4. Are consumers’ evaluations of an athlete, the athlete’s team and the team’s sponsor influenced by news-media reports detailing the athlete’s off-field behaviours?

Study 2 provides an experimental test of the influence of news’ articles on respondent evaluations of a psychological group. Social identity theory and balance theory contribute to the development of hypotheses for this study.

These theories argue that level of identification with the sporting team has the potential to bias evaluations of group members (Dimmock, Grove & Eklund, 2005; Hogg & Turner, 1987; Jackson, 2002; Turner, Hogg, Turner and Smith, 1984). They also argue that evaluations across group members should be consistent in order that the individual does not encounter cognitive stress (Heider, 1958).

Three stimuli were used in a between subjects test. Identification with the sporting team was measured as a covariate variable. Dependent variables included measures of Person Likability, Attitude toward the Act, Attitude toward the Team, Team Corporate Image, Attitude toward the Sponsor and Sponsor Corporate Image.

276

Results of the final analysis provide that treatment group produces main effects on evaluations of person variables (Attitude toward the Act, and Person Likability). Treatment group does not produce significant effects on team and sponsor variables. Identification, as covariate, produces main effects Person Likability, and Team variables (Corporate Image and Attitude toward the Team). Evaluations of sponsor variables are unaffected by treatment and identification with the team in this research.

This combination of results suggests that public opinion of the athlete is influenced by the athlete’s behaviour, and that there is no transfer of negative affect to the athlete’s team or the team’s sponsor. A balance theory approach confirms that perception of the individual is consistent according the valence of their behaviour.

Balance evaluations also establish that although the experimental stimuli have not produced significant effects on the team or sponsor, that contagion is possible. This contagion threat exists in the form of the Person Likability variable, which has variance explained by both treatment group and identification with the team. The dual source of variance creates the possibility of stress in evaluations where the athlete is grouped with team and sponsor. Overall, the effectiveness of social identity theory and balance theory in explaining evaluations of group members is confirmed.

The results of Study 2 provide that negative news’ about athlete’s off-field behaviours do not influence evaluations of sponsor variables. Sponsors’ objectives have not been subject to negative transfer from evaluations of the athlete in this research. However, the potential for evaluative stress, or contagion, is present. The findings of this research are good news for sponsors, but further research is needed to replicate results.

Limitations

The research reported here has several limitations. These include, principally: lack of development of questions raised by the content analysis; failure to address multiple

277

news’ sources in the content analysis; sample size; and effect sizes for sponsor variables. Many of these limitations could have been anticipated; in particular, the academic literature on sample size and effect size is easily sourced (see: Tabachnick & Fidell, 2007; Hair et al, 1995, 2006, 2010). One of the difficulties in undertaking Phd research is the too frequent encountering of one’s own ignorance and optimism. For example, given the relative dearth in reporting effect sizes, one doesn’t know what effect size to anticipate when one reads that sample size should be estimated in relation to anticipated effect sizes. One simply hopes that a small sample will be sufficient. Other problems derive from time and cost constraints which are presumably, present in all research contexts, and require rationalisation of the project according to the specific situation of the researcher.

The goal for Study 1 was to identify the types of off-field behaviours of sportspeople that were commonly reported. The study achieved that goal in relation to local news. A more developed representation of reporting would have considered multiple news- media sources and compared frequency of reports and well as qualitative characteristics.

The content analysis raised many questions. Although it fulfilled the goal of identifying types of behaviours; it also highlighted the need for a broader understanding of the function and ideologies of the news-media. The over- and under-reporting of off-field behaviours relative to the prevalence of similar behaviours reported as criminal cases, in particular, argues the need for marketing and public relations research. Research that determines the priorities of journalists and editors has the potential to inform, particularly, public relations practice.

Study 2 could be improved in many ways. Improvement to the sample size is the first means to improve the research of Study 2. Sample size is responsible for borderline significant results. The tripartite standard of requiring significant test alpha, observed power, and effect size to judge results is stringent and important. These standards provide confidence that results are not achieved by chance, and also that they are meaningful. The system, however, requires that the researcher is reasonably confident in anticipating the likely effect size associated with experimental conditions when designing the study and calculating sample size.

278

Study 2 produced one borderline result: the effect of Identification on Attitude toward the Act. The effect of Identification on Attitude toward the Act is ‘significant’ in conventional terms, with a p-value of .028 (test alpha =.0125). The result is unacceptable for being under powered, .380. The partial η2 value is small, .033. This argues the need for a larger sample size, possibly a much larger sample size, for confidence in the result. Relaxing the required degree of observed power would solve the problem of result acceptability. It would also, according to Cohen (1988), increase the risk of false rejection of the null hypothesis; which is especially critical when observing small effect sizes.

Effect sizes also signal how little sponsor variables have been influenced in this research. The experiment provides that the news’ can produce moderate-large effects on person variables, and identification produces moderate-very large effects on team variables and likability. None of the results are significant for sponsor variables, and the range of partial η2 (.001 - .014) for sponsor variables suggests that it is extremely unlikely, even with larger sample sizes, that sponsors will be affected by reports of off-field athlete behaviour.

Opportunities for Future Research

The current research has raised many questions incidental to its research goals. Further, the findings of research present a number of opportunities to improve and develop upon the reported research. The research might be extended into other contexts. The opportunity exists to investigate politicians and celebrities and the policies or products they endorse; these actors are subject to the same news processes and consideration within the scope of ‘news values’ as are athletes. Further opportunities also exist within the theoretical frames of attribution; news values; contagion; and, mere exposure, which are discussed further here.

279

Attribution

Attribution involves “the linking of an event with its underlying conditions” (Heider, 1958, p89). It is the search for explanations of behaviour. The current research hypothesised three means to explain how athlete off-field behaviours might influence the achievement of sponsors’ objectives: a) valence of behaviour, b) team identification, and c) perceptual balance.

The reported research did not utilise an explicit theoretical mechanism linking the athlete and sponsor. Instead, the measures used allow respondents to infer, or transfer affective or evaluative valence from one entity in the sponsorship model to another. This procedure is not dissimilar to real sponsorship environments, in which consumers are required to become aware, attend to information, self-motivate and process sponsorship information, often with little incentive or support.

The findings of the research present the possibility that team and sponsor ratings might have remained positive because no attribution process linked athlete behaviour with an origin or cause in the team or sponsorship relationship. Heider (1958) explains that attribution of causality to an actor requires the perceiver to identify the actor as capable of the act, or responsible for it. Specifically: Attribution to personal causality reduces the necessary conditions essentially to one, the person with intention, who, within a wide range of environmental vicissitudes, has control over the multitude of forces required to create the specific effect. Heider, 1958, p102. The results of this research, in particular the lack of significant effect of treatment group on team and sponsor variables, suggest that respondents regard the athlete as wholly responsible for the acts depicted. The team and the sponsor, alternatively, may not be regarded as having the intention to control or to contribute to the off-field behaviours depicted.

Introducing a measure of attribution into future research on off-field behaviours will serve to explicitly link the athlete and the sponsor. Attribution theory also provides the means of linking several areas of this research more closely. Kelley’s covariation

280

principle argues that the relationship between the “observed effect and its possible causes” (Kelley, 1973, p108) provides the perceiver with sufficient information to attribute event cause either to a person or to situation.

Three types of information are used to evaluate covariation and the likelihood that events are attributed to the actor, these are: distinctiveness, consensus, and consistency (Kelley, 1973). Consensus refers to whether other actors have behaved in the same manner in similar situations, effectively, agreement that the behavioural response is normative (Kelley, 1973). Consistency argues that a particular behaviour is characteristic of the person if they are seen to behave in the same way given the same stimulus over time (this characteristic most represents the notion of intent) (Kelley, 1973). Distinctiveness, finally, is whether the behaviour is a direct result of a stimulus, and can be attributed to that situation (Kelley, 1973). Harris, Todorov and Fiske summarise the results of their research: …describing behaviour that is low in consensus across actors, low in distinctiveness across entities, and high in consistency over time… uniquely elicits… person attributions… Harris, Todorov & Fisk, 2005, p763

Future studies of athlete off-field behaviours as they might influence sponsorship objectives should further consider the nature of the acts (normativity), behaviour contexts (distinctiveness), and of personal characteristics (consistency of acts).

Specifically, future research might include a 2 (behaviour valence: positive/ negative) x 2 (sponsor participation: low/ high) factorial test of attribution. Where scenarios present the sponsor as contributing to the behaviour undertaken by the athlete (e.g., providing an athlete with a car that is then later used by the athlete to drive dangerously above the speed-limit), attribution effects may occur.

Thresholds for Contagion

Contagion is the idea that an ‘essence’ can be imparted to an object through association or use (see: Argo, Dahl & Morales, 2008; Fernandez & Lastovicka, 2011;

281

Newman, Diesendruck & Bloom, 2011). In marketing research it is used to explain the effects of touch or contact on perceptions of an item, often in relation to the attractiveness of the person who has touched an item (Argo, Dahl & Morales, 2008; Kramer & Block, 2011). Perception may be differently influenced according to actual or perceived touch (Kramer & Block, 2011), known use or mere ownership (Newman, Diesendruck & Bloom, 2011), or even the resemblance of a ‘replica’ product to an ‘original’ (Fernandez & Lastovicka, 2011).

Contagion theory may be useful in explaining the level or point at which negative off-field behaviours transfer. The athlete’s involvement in negative off-field behaviours may be conceptualised according to ‘degree of touch’ or behaviour severity as providing a ‘threshold of contagion’ to the sponsor’s brand. Drawing from the potential the attribution scenario above, a project testing contagion might also be developed using a factorial design. To test threshold, increments denoting behaviour severity might be introduced. For example, three levels of speeding offence might be identified, ranging from 2-3 kilometers over the speed limit, to 40- 50 kilometers over the limit. Including an attribution measure within this research should distinguish acceptable transgressions from those that are non-normative or unacceptable.

Black sheep & Team culture

The ‘black sheep’ effect in social identity theory argues that positivity in evaluations of in-group members does not occur universally. When an in-group member is ‘deviant’ or ‘unlikable’ that in-group member is a ‘black sheep’ (see: Hogg, 2000; Hogg & Terry, 2000; Marques & Yzerbyt, 1988). The black sheep effect produces a case of “ingroup derogation” (Marques & Yzerbyt, 1988, p288) in which unlikable in-group members are evaluated as more negative than unlikable out-group members (Marques & Yzerbyt, 1988).

An agenda to study athlete off-field behaviours presents the possibility that prototypicality may influence consumer evaluations. This research may be improved through consideration of group-level prototypicality (Hogg, 2000; Hogg & Hains,

282

1996) and of ‘black sheep’. In particular: what are the prototypical characteristics of professional athletes? Given what characteristics are group members regarded as deviant, or black sheep? Or, as hypothesised by Marques and Yzerbyt (1988) is trait negativity sufficient?

Mere, or repeated, exposure

Mere exposure argues that repeated exposure to a message can improve affect or liking of the message (Janiszewski, 1993; Zajonc, 1980). The methodology has been used in sponsorship and marketing research to evaluate message attractiveness, memory and congruence perception (see: Bennett, 1999; Dardis, 2009; Olson & Thjomoe, 2003; Pitts & Slatterly, 2004). Results have been mixed, and suggest that any use of mere exposure should occur in conjunction with measures of involvement or identification.

Bennett (1999), for example, found that committed sports attendees had better memory for sports stadium signage and sponsorship messages than less committed or regular attendees. Pitts and Slatterly (2004), also in stadium research, found high levels of yea-saying and false identification of sponsors. Although called mere exposure, the level of exposure was not controlled in these studies.

In controlled research, Dardis (2009) finds that attitude toward the sponsor, perception of community relations, perceived congruence of sponsor-sponsee fit and purchase intentions improve given 4 rather than 1 exposure to a sponsorship stimulus. Olson & Thjomoe (2003) establish, with multiple new and existing brand stimuli, that repeated exposure improves attitude toward the advertisement for new brands. The effect does not occur when subjects might have existing brand knowledge. The results also vary according to information processing styles of the respondents; which again argues the importance of either involvement or ‘need for cognition’ measurement.

Mere exposure, in relation to negative off-field behaviours of sportspeople may produce several effects. Memory for events might improve. Evaluation of the person

283

may vary according to number of events presented. Severity of (or extremity of) attitude toward the act and person likability measures are likely to occur given repetition. It is also possible that repeated linking of the sponsor and the athlete in such contexts will increase the likelihood of a stimulus effect on sponsor evaluations.

What happens if you don’t cancel the contract?

The current research has measured evaluative outcomes as a direct consequence of negative information. There appears to be little likelihood of negative evaluation of the sponsor. However, the information stimuli used did not provide information about what happens after the off-field behaviour. Future research should consider whether evaluations of the sponsor alter according to corporate response.

Fink et al (2009) establish that ‘leadership response’ to off-field behaviours influence level of team identification (using a pre- and post-test methodology) for high identification respondents. Low identification respondents do not vary in their level of identification given strong (fast/ condemnatory) or weak (slow/ indecisive) corporate responses (Fink et al., 2009). Highly identified respondents, however, show declining identification scores in the face of weak leadership responses, and improved identification scores given strong responses (Fink et al., 2009). These results may be due to self-esteem protection and the goal of the individual to maintain a positive group image (see: Hogg, 2000; Hogg & Mullin, 1999).

Understanding of the current results would improve with knowledge about whether consumers regard cancelling contracts as positive and strong leadership responses; or, whether the cancellation of sponsorship contracts is regarded as in-group member disloyalty.

Contributions of the Research

The purpose of this research was to address the ‘negatives’ in the sports sponsorship environment. Within that broad research field, the impact of athletes’ behaviours off-

284

field as ‘scandals’ or ‘transgressions’ was presented as an emerging area of research (see: Hughes & Shank, 2005, 2008; Sassenberg & Morgan Johnson, 2010; Wilson, Stavros & Westberg, 2010).

Off-field behaviours of athletes have contributed to abnormal negative returns in share value (Knittel & Stango, 2010). The have also to contributed to sponsorship contract cancellations in cases where sponsors have sought to protect their brand image and marketing messages from taint (BBC Sport, 2009; Byrnes & Phelps, 2009; NY Times, 2009).

It is to the notion of taint, more specifically, the risks posed to achievement of sponsorship objectives, which this research contributes. The contributions of this research are to the areas of sports management and sponsorship policy development, and also sponsorship theory. These are discussed separately, below.

Contributions to Sponsorship Theory

This thesis has identified several means by which sponsorship objectives are thought to be achieved. Balance theory and social identity theory explain that relationships sports’ sponsors try to build with sports audiences are mediated by the relationships that audiences have with sponsees. Identification presents strong evidence that individuals favour in-group members in evaluative contexts and resource allocation (see: Fielding, Hogg & Annandale, 2006; Grieve & Hogg, 1999; Pinter & Greenwald, 2001; Turner et al, 1984; Turner, Sachdev & Hogg, 1983). Balance, alternatively, argues that cognitive consistency goals drive the evaluations of linked individuals (see: Crandall et al, 2007; Folkes, 1988; Harris, Todorov & Fiske, 2005).

Marketing research has also previously found that group prestige or distinctiveness explains consumer willingness to associate with an event (Cornwell and Coote, 2005); and, that identification explains third-party (mediated) effects such as attitude towards the sponsor, recognition and satisfaction with the sponsor (Gwinner & Swanson, 2003).

285

The results of Study 2 establish that consumer evaluations are influenced by presentations of news of athletes’ off-field behaviours. However, those influences are restricted to effects on evaluations of the athlete. Evaluation of the sponsor’s corporate image and attitude toward the sponsor’s brand are not influenced by information about assault, driving under the influence of alcohol, or raising money for charity. Attitude towards the team’s brand and corporate image are similarly unaffected by reports of athlete’s off-field behaviours. Overall, the effects of treatment group establish no risks to sponsors’ objectives, in a controlled context.

These results may appear controversial. They appear to challenge one of the fundamental premises of sponsorship activities; that sponsors can achieve corporate objectives through associations with sponsees. The research has endorsed the mechanisms of identification and balance; it found that identification influences evaluations of individuals linked to group. It also found that cognitive stress is likely to result in cases where the respondent likes a team but the team’s athlete is reported to have behaved negatively. On the basis of results achieved, and in the presence of null results (no endorsement of the mediated sponsorship relationship), this research does not argue that the achievement of sponsorship objectives through sponsees is a fundamentally flawed proposition that will never work. Instead, that it has not occurred in this research.

The theoretical contributions of this research are therefore twofold. The first theoretical contribution is support of the extant literature that says ‘beware the weak association’. A weak sponsorship association represents wasted money, and a null contribution to sponsorship objectives.

It should be noted that the sponsorship stimulus used in this research is not regarded as weak in the sense of being flawed. The stimulus used in the experimental research presented sponsorship relationship information and subsequently requested evaluations of the same. Relative to signage at sporting events, uniform sponsorship or other forms of sponsorship related communications, the stimulus used is ‘above the line’. Respondent attention to the sponsorship relationship is guaranteed through the question-response process. The stimulus may be regarded as weak only in the sense that the sponsor was not described as being a participant in the behaviours

286

described (therefore responsibility for the act should be perceived as limited), and the relationship cue was not presented multiple times.

There are several mechanisms which explain the lack of mediation in this research; many of these derive from the information processing model which has been discussed previously. Sponsorship researchers advocate the importance of leverage and activation (Kinney & McDaniel, 1996; Weeks, Cornwell & Drennan, 2008), of sponsor-sponsee congruence (Coppetti et al, 2009; Rifon et al, 2004; Ruth & Simonin, 2003), and message or image transfer or learning through repeated exposure (Bennett, 1999; Dardis, 2009; Olson & Thjomoe, 2003). Each of these mechanisms contributes fundamentally to strength of association. Implicitly, each argues sponsorship per se is insufficient to achieve sponsorship objectives; or, that the mediated relationship is naturally weak. All provide reasoning or empirical evidence that developed (strong) associations can contribute positively to sponsor brand attitudes and corporate image.

This research aids understanding of what does not contribute to evaluations. The off- field behaviour information stimuli used in this research produces no effect on sponsor variables. Further, the result is not influenced by positivity, negativity or extremity effects derived from the information stimulus. Although social psychological research argues that negative, and/or extreme information has greater weight in evaluations than neutral or moderate information (see: Fiske, 1980; Martijn et al, 1992; Skowronski & Carlston, 1987, 1989, 1992); this research found no group-level differences in effects on sponsor variables. The ‘failure’ of stimuli to significantly effect evaluations of sponsor variables is informative.

The results, overall, suggest that the sponsor has not been perceived as a group member in a context where relationships are important. That is, respondents highly identified with the athlete’s team have not evaluated the sponsor as if the sponsor was a group member. According to social identity research (Fielding, Hogg & Annandale, 2006; Turner et al, 1984), highly identified respondents, relative to less identified respondents, should evaluate the sponsor as more positive given a positive stimulus. When presented with a negative stimulus, highly identified respondents should also evaluate the sponsor more positively in comparison to low identification

287

respondents (Doosje et al, 2006; Fielding, Hogg & Annandale, 2006; Turner et al, 1984). This research found no effect of identification on evaluations of sponsor vairables. This suggests that the sponsor is not regarded as a group member by either identified or non-identified respondents, and therefore not subject to bias in the form of favouritism or derogation.

The second theoretical contribution is methodological. The multi-attribute (likability, and behaviour valence) approach to object evaluations recognises that judgments have multiple sources of variance, and that summated or single measures may fail to identify that variance. Essentially, the methodology asks: a) in principle is the object good or bad? and b) do you, personally, like it? The research identifies that person likability is not always predicted by behaviour valence (DUI is evaluated as a negative act; the person is not, however, unlikable). The finding is important because it belies the simplicity of attitude measures, and in so doing, it highlights the importance of the social in relation to the individual. Other sponsorship applications might find that respondents believe branded experiences are ‘good’ in the sense of being well-organised and creative, but not personally likable or likely to motivate action.

The results of this research identify that research methodologies need to adjust to the noisy and competitive sponsorship context in which attention may be fleeting or scarce. In principle, sponsorship stimuli used in research should reflect the reality of the corporate environment and respondent psychological processes. The results of this research also argue the importance of associative strength (developed relationships) in the mediated sponsorship environment.

Contributions to Sports Management & Sponsorship Practice

The research has implications beyond those purely theoretical. It provides sports sponsors and teams with means to assess athlete off-field behaviours as risks to objectives. It also provides team and athlete managers with an understanding of how consumers evaluate off-field behaviours and potential repercussions of such.

288

Sponsor objectives

Conventional wisdom in crisis management literature provides a variety of means to deal with events that threaten negative image development for the brand. These strategies take the form of denial, evasion of responsibility, challenging interpretation of the event, corrective action and apology (Brinson & Benoit, 1999; Fortunato, 2008). Each of these is an active strategy to attempt to address situations defined as crises.

Public relations strategies have much in common with social identity analyses of group membership. According to Tajfel (1974) one of the consequences of categorisation as a group member may be a negative evaluation of the group. If this occurs, the strategies available to the individual include: leave the group (unless impossible); “change one’s interpretation of the attributes” (Tajfel, 1974, p70); and, to engage in ‘activism’ to change perceptions or attributes (Tajfel, 1974, pp69-70). Effectively, both public relations’ crisis (or reputation) management and social identity approaches to group membership argue that you can support the ‘offender’ (or group), accept responsibility, or leave.

In the context of off-field behaviours of athletes and sponsor’s objectives; a more passive option exists. Nike, Gillette and Tag Heuer instituted passive strategies during the revelations about Tiger Woods (see: The Canadian Press, 2009; NY Times, 2009; Steel, 2010; UPI Energy Resources, 2009); these reduced media exposure linking the athlete and the brand. The research reported has established that consumers do not blame corporate sponsors for the off-field behaviours of sportspeople. A key contribution of this research is the support for sponsors adopting a ‘passive’ approach to athlete off-field behaviours.

Team objectives

Although the goal of this research has been to study sponsor’s objectives in the context of negative news, the implications for teams are equally interesting. The 289

results of the research show no effect of treatment group on evaluations of team variables. Evaluations of the team are, however, affected by level of respondent identification with the team; evaluations of athlete likability are similarly affected by the identification covariate.

The effects of identification on the athlete and the team indicate that the athlete and team are regarded as group members by respondents; relative to one’s own level of team identification, that group might be an in- or an out-group. The implications of the linking of athlete and team evaluations, according to balance theory (Heider, 1958), creates the potential for cognitive stress among respondents who evaluate the team and athlete as differently valenced.

The crisis management strategies available to sponsors, mentioned above, are clearly also available to teams. However, the research reported here, and that reported by Fink et al (2009) reveals a more tenuous public opinion position for teams relative to sponsors. The research conducted by Fink et al (2009), discussed previously, found that leadership responses to off-field behaviours affect posttest identification scores. Identification with the team is more resilient (amongst highly identified respondents) to a crisis when the leadership response is quick and decisive, condemning poor behaviour (Fink et al, 2009). In social identity terms, this strategy argues maintaining group self-esteem (see: Doosje et al, 2006; Hogg, 2000; Hogg & Mullin, 1999; Tajfel, 1982) by excising or neutralising a negative association.

Monitoring the effects of identification and the potential fallout from crises produces a fine like for team managers to walk. Condemning the behaviour, as advocated by Fink et al (2009), of ‘one of your own’ can be seen as disloyalty towards the group (Wann & Branscombe, 1992). The research conducted by Wann and Branscombe (1992) found that when presented with an information stimulus critical of the group, highly identified respondents were more upset when the author of that information was identified as a group member than when the author was described as being an out-group member.

In a different context altogether, Doosje et al (2006) found that when presented with evidence of negative behaviour of the in-group (i.e., colonisation) highly identified

290

individuals felt guilt when the source of the information was an in-group member. Levels of reported guilt were significantly lower when the information source was an out-group member (Doosje et al, 2006). In-group source credibility was also perceived as significantly higher among highly identified respondents (Doosje et al, 2006).

When reviewing the works of Wann and Branscombe (1992), Doosje et al (2006), and Fink et al (2009) in tandem, the practical implications appear straightforward. If an athlete behaves negatively off-field, then an in-group member (e.g., team official) should respond quickly to condemn the behaviour; the message will be regarded as more credible and therefore appropriate.

The research of Doosje et al (2006) raises a final point. That is: …high identifiers are only as willing as low identifiers to compensate when their group has acknowledged in some way their mistakes in the past, either by offering apologies or by presenting financial reparation. When the ingroup has not acknowledged their mistakes, identification was negatively related to recommendations for financial compensation… Doosje et al, 2006, p335 This finding presents the semantic problem of whether condemnation of a negative act is equivalent to apology, or the acknowledgement of consciously enacted fault (which may differ from condemning the foolish act which has an unfortunate result). It also introduces a temporal problem; does behaviour recency alter willingness to make reparations? These questions are currently unanswerable.

Thus far, the problems identified are framed as problems of the in-group that can be addressed through messages directed towards in-group members. Such a situation is unrealistic. This research found that team identification among the sample was not particularly high; 30 of 151 (19.9%) respondents could be categorised as highly identified when high identification is defined as a score at or above the scale mid- point. When message reception varies according to group identification it should be considered how a message ideal for 20% of the population will be received by less identified or out-group members. This problem will be addressed further in the following section.

291

Evaluations of Behaviours

This section returns to public relations approaches to issues- or crisis management. Typologies of management strategies exist (see: Brinson & Benoit, 1999; Fortunato, 2008); these explain how an organisation might respond to situations which threaten the reputation of the organisation or the conduct of its business. Several process models for contingency planning and management also exist (see: Burnett, 1998; Chappelet, 2001; Coombs, 2007; Ritchie, 2004); these provide generalised procedures to guide organisations from pre-planning through to crisis recognition and toward situation resolution.

This research has shown that the off-field behaviours of athletes pose a risk to the corporate objectives of sporting teams. The risk of athlete off-field behaviours derives from the identification construct which explains variance in team variables and athlete likability. Athlete likability variance is also explained by treatment group which provides a gateway through which negative spillover might occur. A practical contribution of this research is the identification of classes of behaviour according to their in-principle evaluation (A-act) and according to the concomitant person likability measure.

The crisis management literature when mapping how to address situations begins with attempts to classify the situation. Burnett (1998) and Chappelet (2001) identify the probability of event occurrence as important; they and others identify time pressure, event severity or potential impact and responsibility (see: Ritchie, 2004; Coombs, 2007) as other important factors.

This research provides two forms of information that allow athlete or team managers to assess the potential risks derived from off-field behaviours. The first of these is provided by Study 1 which indicates a level of over-reporting of behaviours that are relatively uncommon in the broader population. Specifically, Study 1 identifies that sexual assaults and other acts intended to cause injury are reported inconsistently with their likely rates of occurrence. This information provides managers within the sports context to identify the probability of media hype occurring.

292

The second form of information relates to the likability and attitude toward the act measures. This research has demonstrated that behaviour valence does not always predict person likability. The result conforms to person impression theories which explain that information extremity is diagnostic for judgments, whereas moderate or neutral information is not (Skowronski & Carlston, 1987, 1989, 1992). Where A-act and likability scores differ significantly (i.e., DUI, in this research) the information stimulus is not, theoretically, extreme. The result of this reasoning is that athlete or team managers should consider the class of behaviour identified prior to instituting a strategic response.

Two classes of behaviour might be recognised. The first group represents extreme behaviours (positive or negative); the second, behaviours which might be evaluated as neutral or normative. Where the first group has the power to influence judgments of the person, and by extension, the team, the second group are less likely to do so when presented as a single instance. The strategic response to the reporting of off- field behaviours can therefore be determined by likely effects on judgment of the person.

Conclusions

This thesis began by recognising that sponsorship researchers have traditionally focused on means to improve the effectiveness of sponsorshiop activities (see: Chien, Cornwell & Stokes, 2005; Grohs, Wagner, & Vsetecka, 2004; Gwinner & Eaton, 1999). Negatives and risks in the sponsorship environment were identified as less developed, but increasingly important areas of research (see: Parker, 2007; Pope, Voges & Brown, 2009; Wilson, Stavros, and Westberg, 2008).

Within the emerging literature addressing negatives in the sponsorship environment, the perspectives of sponsoring organisations and marketing managers are well represented (see: Farrelly, Quester & Burton, 2006; Hughes & Shank, 2005, 2008; Wilson, Stavros & Westberg, 2008). Consumer evaluations, however, are under- represented. The research reported in this thesis assessed consumer evaluations of

293

one type of ‘negative’ in the sponsorship environment, the off-field behaviours of athletes which might influence the achievement of sponsor’s objectives.

Study 1 found that a wide range of athlete off-field behaviours are reported by the news-media including such things as: pub brawl assaults, sexual assaults, public urination, driving under the influence of alcohol, use of illicit and performance enhancing drugs, breaches of club rules, and positive or altruistic behaviours. Off- field behaviours are reported less frequently than match play and competition-related news, however, more frequently than on-field incidents reported to the sports judiciary. Off-field behaviours are also not reported at levels consistent with behaviour occurrence in the broader population; for example, acts intended to cause injury represent 9.9% of cases in Magistrates Courts, whereas they comprise 26.5% of the off-field behaviours reported in a local newspaper. This study identified the types of off-field behaviour reported; it also highlights the roles and potential impact of the news-media on public opinion.

The research models for Study 2 hypothesised that evaluations of the athlete, team and sponsor would be equally affected by news articles reporting off-field behaviour of an athlete. Equally valenced evaluations were expected because of the balance theory tenet which argues that the individual prefers to evaluate linked individuals or entities consistently (Heider, 1958). Equal effects of the stimuli were not found for the athlete, team and sponsor. Evaluations of the athlete were influenced by the stimulus, but not evaluations of the team or sponsor. Evaluations of the team and athlete likability were influenced by identification with the team. These results suggest greater risks to team objectives than sponsor objectives through their closer association with the athlete. The imbalance in several evaluative models recognises sources of cognitive stress for respondents, and the potential for negative spillover to team variables.

This research provides insight into one source of negative spillover to sponsor’s objectives, the off-field behaviours of athletes. It reveals that evaluations of attitude towards the sponsor’s brand and sponsor corporate image are unaffected by reports of negative off-field behaviours of sponsored athletes. It also demonstrates that sponsorship, per se, is regarded positively by respondents. Future research should

294

test the tolerance of Study 2 findings through use of repeated presentations of both negative stimuli and sponsorship relationship information. The findings of this research, overall, present an important step towards understanding the influence of athlete off-field behaviours and consumer evaluations on sponsors’ objectives.

295

Appendices

296

Chapter 1

Definition of Terms

Off-field behaviour: Behaviour of a sportsperson that occurs away from their primary workplace (the sports field, gymnasium or stadium) and role (athlete, team employee) that may have bearing on public perception of that person, and implications to their sporting code or team.

Moral/ Immoral: Across disciplines, morality is discussed in relation to the responsibility of the individual to prevent harm, exhibit care and duty and to act in ways that are just (see: Chiu, Dweck, Tong & Fu, 1997; Ellemers, Pagliaro, Barreto & Leach, 2008; Graham, Haidt & Nosek, 2009; Haidt & Graham, 2007). This research takes the perspective that moral acts demonstrate characteristics of care, duty or justice, and that immoral acts may involve harm (or potential harm), carelessness, abrogation of duty or injustice.

Moral risk: Occurs when the behaviour of a person is judged disreputable or immoral; the consequences of this behaviour has repercussions beyond the individual, to achievement of corporate (team or sponsor) objectives. The moral risk is conceived as a risk acquired by association.

Sponsee: The recipient of sponsorship resources.

Sponsor: The organization or individual that provides resources to a sponsee in order to assist the sponsee in their business (e.g., sport, entertainment, event-hosting).

Sponsorship: “…the provision of resources (e.g., money, people, equipment) by an organization (the sponsor) directly to an individual, authority, or body (the sponsee), to enable the latter to pursue some activity in return for benefits contemplated in terms of the sponsor’s promotion strategy… which can be expressed in terms of corporate, marketing or media objectives” (Pope, 1998).

Sponsorship property: Any team, event, performance, individual, or thing (i.e., location) that is offered to the sponsorship marketplace for sponsorship.

(Team) Identification: This research refers to social identity theory to define identification as the “psychological connection” (Dimmock, Grove & Eklund, 2005, p76), or the “the perception of oneness with or belongingness to some human aggregate” (Ashforth & Mael, 1989, p21). Team identification is a specific form of identification that an individual may feel about a specific sporting team as a representation of an important social group.

297

Chapter 2

2.1: Annual growth in spending: advertising, sales promotion & sponsorship, 1985-2009.

Year Advertising Sales Promotion Sponsorship Annual Annual Annual growth % Change growth % Change growth % Change 2009 (forecast) -4.2 -2.3 1.1 2008 0.6 4.8 2.0 4.3 11.4 10.3 2007 3.0 2.4 3.7 1.7 11.5 0.1 2006 2.9 -0.1 3.8 0.1 10.5 -1.0 2005 3.0 0.1 5.3 1.5 8.9 -1.6 2004 6.9 3.9 5.0 -0.3 8.7 -0.2 2003 5.2 -1.7 4.2 -0.8 6.2 -2.5 2002 2.6 -2.6 5.6 1.4 3.7 -2.5 2001 -4.1 -6.7 -5.6 -11.2 6.9 3.2 2000 9.8 13.9 6.3 11.9 14.0 7.1 1999 6.8 -3.0 7.4 1.1 12.0 -2.0 1998 7.1 0.3 4.2 -3.2 15.0 3.0 1997 6.6 -0.5 3.3 -0.9 9.0 -6.0 1996 7.6 1.0 4.6 1.3 15.0 6.0 1995 7.7 0.1 4.6 0.0 11.0 -4.0 1994 8.7 1.0 5.4 0.8 15.0 4.0 1993 5.2 -3.5 7.0 1.6 17.0 2.0 1992 4.0 -1.2 10.0 3.0 13.0 -4.0 1991 -1.5 -5.5 8.0 -2.0 11.0 -2.0 1990 5.0 6.5 6.0 -2.0 19.0 8.0 1989 6.0 1.0 6.0 0.0 22.0 3.0 1988 7.0 1.0 7.0 1.0 20.0 -2.0 1987 7.0 0.0 8.0 1.0 30.0 10.0 1986 7.0 0.0 9.0 1.0 35.0 5.0 1985 7.0 0.0 13.0 4.0 18.0 -17.0 Average annual growth 4.68 5.26 13.80 Average annual change 0.47 0.64 0.70 (2009-1999) 10-year average growth 2.95 3.22 8.63 10-year average change 1.10 0.97 1.09 Data sourced from: IEG Inc. (2003); IEGSR (2009); IEGSR (2007)

298

2.2: Frequency of Attendance at Performing Arts (2005-2006) and Sporting Events (2009-2010) in Australia

PERFORMING ARTS Number (‘000) Attendance Rate (%) 1 or 2 3 to 5 6 or more total 1 or 2 3 to 5 6 or more Classical 991.8 332.7 183.5 1508 65.8 22 12.1 Popular 2451.2 1032.5 552.3 4036 60.8 25.5 13.7 Theatre 1999.3 491.6 232.4 2723.3 73.4 18.1 8.5 Dance 1285 240.4 99.6 1625 79.1 14.8 6.1 Musical/Opera 2075.8 437.5 100.6 2613.9 79.4 16.7 3.8 Other 2216.8 303 135.1 2654.9 83.5 11.5 5.1

SPORTS Number (‘000) Attendance Rate (%) 1 or 2 3 to 5 6 or more total 1 or 2 3 to 5 6 or more AFL 1202.5 747.1 882.3 2831.9 42.5 26.4 31.2 Horse racing 1433.4 353.5 153.4 1940.3 73.9 18.2 7.9 Motor sports 966.7 256.8 199.5 1423 67.9 18 14 Rugby League 618.5 520 425.3 1563.8 39.6 33.3 27.2 Rugby Union 318.8 141.6 115.1 575.5 55.4 24.6 20 Soccer 373.7 258.5 306.7 938.9 39.8 27.5 32.7

Data sources: ABS, 2010: catalogue 41720; ABS, 2010: catalogue 41740

299

2.3: Impression formation processes Category Process* Research using process No-priority- 1. hypothesising Anderson, 1981 Fiske & Neuberg, 1990 independence 2. exposure Anderson & Hubert, 1963 Sherman & Klein, 1994 2-memory 3. encoding Devine, Hirt & Gehrke, 1990 Hoch & Ha, 1986 hypothesis 4. integration Dunning & Sherman, 1997 Park, DeKay & Kraus, 1994 Integration (Hoch & Deighton, 1989, p1) Erdem et al, 1999 Paunonen, 1989 – lens model Hypothesis testing Fiske & Neuberg, 1990 Kardes, Posavac, & Cronley, belief updating Hoch & Deighton, 1989 2004 Levin & Gaeth, 1988 Memory causes 1. stimulus information encoded into working memory – no judgment is made at this Biehal, & Chakravarti, 1983 Park, 1986 judgment time. Carlston & Skowronski, 2005 Paunonen, 1989 availability 2.encoding into long-term memory (LTM) DeCoster & Claypool, 2004 Pope, Voges & Brown, 2009 3.judgment situation (temporal separation) – retrieval of information from long-term Dick, Chakravarti, & Biehal, Sherman & Klein, 1994 memory 1990 Skowronski & Carlston, 1987 4.judgment occurs based on evidence retrieved Klein & Loftus, 1993 Tversky & Kahneman, 1973 5. Memory tests – retrieval process is repeated. Relationship between judgment & Locksley, Borgida, Brekke, & memory is produced. Hepburn, 1980 Judgment causes 1.stimulus information encoded into working memory Burke & Srull, 1988 Neuberg, 1989 memory 2.working memory: Leamer, 1974, 1975 Sedikides, & Skowronski, biased retrieval a)transfer to LTM Snyder & Uranowitz, 1978 1993 *selective recall b)information is used in judgment Srull, 1981 *confirmatory 3.prompted judgment explanation – subject considers LTM stored judgment or that memory in working memory (time dependent) *access-biased 4.memory test – judgment influences recall memory Judgment causes 1.stimulus information encoded into working memory & used for initial judgment Alba & Hasher, 1983 Klein, Loftus & Burton, 1989 memory 2.initial judgment filters subsequent stimulus by guiding search, encoding & Carlston & Mae, 2007 Gregan-Paxton, & John, biased encoding comprehension Carlston & Skowronski, 2005 1997 3.prompted judgment explanation - subject considers LTM stored judgment or that in Hastie, 1981 Lambert, 1995 working memory (time dependent) Hutchinson & Alba, 1991 Sedikides, & Skowronski, 4.memory test – search LTM for trace information – LTM storage is biased by initial 1993 judgment and recall reflects encoding bias Judgment causes 1. stimulus information encoded into working memory & used for initial judgment Hastie, 1980, 1984 memory 2.later information reviewed in context of initial judgment – incongruent extended Hastie & Kumar, 1979 incongruity-biased processing – result is enhanced memorability Heise & Smith-Lovin, 1981 encoding 3.prompted judgment explanation - subject considers LTM stored judgment or that in *incongruent recall working memory (time dependent) 4. memory test – search LTM for trace information – incongruent most likely to be recalled *processes are as articulated by Hastie & Park, 1986

300

Chapter 5

5.1: Attitude toward the Act factor analyses

Scenarios not improved by removing Scenarios improved by removing: items Harm/Beneficial Factor 1 Factor Factor Factor 2 1 2 Charity auction Community Health Bad/Good -0.894 Bad/Good -0.705 0.614 Harmful/Beneficial -0.743 Harmful/Beneficial -0.777 0.301 Pleasant/Unpleasant -0.839 Pleasant/Unpleasant -0.787 -0.287 Safe/Unsafe -0.901 Safe/Unsafe -0.651 -0.677 Variance explained 0.717 Variance explained 0.536 0.252 Cronbach’s alpha .858(.880) Cronbach’s alpha .702 Scenarios improved by removing: Performance drugs Bad/Good Bad/Good -0.796 Harmful/Beneficial -0.794 Punch friend Pleasant/Unpleasant -0.861 Bad/Good -0.585 Safe/Unsafe -0.887 Harmful/Beneficial -0.891 Variance explained 0.698 Pleasant/Unpleasant -0.965 Cronbach’s alpha .847 Safe/Unsafe -0.941 Public urination Variance explained 0.738 Bad/Good -0.887 Cronbach’s alpha .877(.932) Harmful/Beneficial -0.860 Volunteer therapy Pleasant/Unpleasant -0.776 Bad/Good -0.593 -0.673 Safe/Unsafe -0.833 Harmful/Beneficial -0.722 -0.518 Variance explained 0.706 Pleasant/Unpleasant -0.740 0.499 Cronbach’s alpha .850 Safe/Unsafe -0.637 0.634 Sex- no consent Variance explained 0.457 0.343 Bad/Good -0.857 Cronbach’s alpha .585(.601) Harmful/Beneficial -0.903 Save brother Pleasant/Unpleasant -0.929 Bad/Good -0.274 -0.851 Safe/Unsafe -0.764 Harmful/Beneficial -0.634 -0.337 Variance explained 0.750 Pleasant/Unpleasant -0.709 0.447 Cronbach’s alpha .863 Safe/Unsafe -0.788 0.164 Variance explained 0.400 0.266 Repeated speeding Cronbach’s alpha .493(.544) Bad/Good -0.855 Harmful/Beneficial -0.874 Pleasant/Unpleasant -0.779 Safe/Unsafe -0.788 Variance explained 0.681 Cronbach’s alpha .825

301

5.2: Model 1: Fit 1: Standardized Residual Covariances

Someone Rather Success Dislike Little Tradition Am Strong Sense Part Rituals Successes Typical _crit _other _fail _being _proud _history _like _attach _belong _of Someone .000 _crit Rather -1.381 .000 _other Success -2.065 -3.009 .000 _fail Dislike -.831 -.284 1.673 .000 _being Little -1.579 1.488 .644 -.539 .000 _proud Tradition -1.064 -2.322 1.726 -.010 .881 .000 _history Rituals -.057 -1.911 -.693 .399 .381 -.187 .000

Successes -.423 -.422 -1.154 1.405 .302 -1.320 -.043 .000

Am 1.173 -1.859 -.686 .790 .480 -.929 -.107 .048 .000 _like Strong -.466 -.500 -.224 1.901 .623 .202 1.827 .204 -.386 .000 _attach typical .235 -1.619 -1.205 .398 -1.182 -.800 1.350 -.642 .328 .232 .000

Sense .072 -1.606 .433 .697 .203 -.427 .576 -.221 -.119 .112 -.265 .000 _belong Part .437 -1.759 -1.202 .238 -.735 -.886 -.047 1.002 -.215 -.537 .178 .052 .000 _of

302

5.3: Model 1 Fit 1: Modification Indices: Covariances

Modification Parameter Error variables Index Change a5 <--> a1 20.985 .210 a4 <--> v1 4.729 -.159 a4 <--> v2 11.905 .170 a4 <--> a7 4.799 -.111 a3 <--> v1 15.079 .243 a2 <--> a6 10.971 -.124 a2 <--> a1 4.241 .081 a2 <--> a7 19.596 .205 a1 <--> a6 6.381 -.107 v3 <--> v1 28.213 .555 v2 <--> a6 20.483 .219 v2 <--> v1 8.467 -.240 v1 <--> a1 10.734 -.247 v1 <--> d3 6.088 -.309 d2 <--> a6 6.892 .178 d2 <--> a1 5.684 -.169 d2 <--> v1 6.390 .301 d1 <--> a1 5.688 -.164 d1 <--> d3 5.915 .275

Changes rejected due to violation of independence a7 <--> values 4.409 -.141 a6 <--> values 22.279 .257 a6 <--> affirmation 20.549 -.211 a6 <--> denial 13.103 .204 a2 <--> values 4.523 -.108 a2 <--> affirmation 4.370 .091 a1 <--> values 7.154 -.153 a1 <--> affirmation 9.482 .151 a1 <--> denial 12.942 -.212 d3 <--> values 5.610 -.225 v3 <--> values 4.697 .165 v3 <--> affirmation 7.045 -.180 v2 <--> values 7.166 -.154 v2 <--> affirmation 12.617 .187 v1 <--> affirmation 5.324 -.190

Model 1 Fit 1: Modification Indices: Regression Weights Observed Latent Modification Parameter variable variable Index Change strong_attach <--- values 5.591 .071 strong_attach <--- denial 5.582 .127 someone_crit <--- denial 5.947 -.136 rather_other <--- values 7.565 -.144

303

5.4: Model 1 Fit 2: Standardized Residual Covariances Am Dislike Little Tradition Sense Rituals Successes Typical Part _of _like _being _proud _history _belong Am _like .000

Dislike .209 .000 _being Little .636 .000 .000 _proud Tradition -.115 -.189 1.229 .000 _history Rituals -.347 -.379 .247 .000 .000

Successes .045 .729 .378 -.643 -.432 .000

Typical .451 -.181 -1.050 -.031 1.076 -.678 .000

Sense .065 .052 .383 .491 .318 -.215 -.123 .000 _belong Part _of -.294 -.512 -.692 -.226 -.525 .744 .064 -.023 .000

5.5: Model 1: Fit 2: Modification Indices: Covariances Modification Parameter Error variables Index Change a4 <--> v2 12.105 .175 a4 <--> a7 5.822 -.126 a2 <--> a7 13.704 .171

Changes rejected due to violation of independence a2 <--> affirmation 4.351 .092 a4 <--> values 6.259 .134

304

5.6: Model 2: Fit 1: Standardized Residual Covariances

Someone_Crit Successes Strong_Attach Typical Am_Like Part_of Sense_Belong someone_crit .000 successes -.597 .000 strong_attach -.434 .335 .000 typical .102 -.679 .403 .000 am_like .972 -.048 -.273 .272 .000 part_of .192 .855 -.462 .082 -.373 .000 sense_belong -.026 -.221 .350 -.217 -.137 -.003 .000

5.7: Model 2: Fit 1: Modification Indices: Covariances

Modification Parameter Error variables Index Change d2 <--> d7 5.965 -.125 d6 <--> d7 5.309 -.100 d5 <--> d2 5.576 -.120 d4 <--> d7 16.184 .179 d3 <--> d2 15.893 .181 d3 <--> d6 7.782 -.108 d1 <--> d6 7.305 .100

305

Chapter 6

6.1: Treatment scenarios

Group & scenario details News’ article

Who: Ben Howse Pub brawl appeal What 1: Assault When: 12 months ago Brisbane Broncos forward, Ben Howse, was back in court What 2: Appeal conviction yesterday to appeal against his conviction for assault.

Where: Court A Howse fractured a man’s skull in a pub brawl 12 months Why: To be inferred ago. He was convicted of assault and sentenced to 250 Team: Identified hours of community service. Sponsor: Identified Player: Identified The Brisbane Broncos are sponsored by WOW Sight & Sound.

Who: Ben Howse Pedal power recommended for Howse What 1: DUI When: After State of Brisbane Broncos enforcer, Ben Howse, was yesterday fined $750 and banned from driving after being caught Origin 2 drink-driving the morning after State of Origin 3. What : License suspended Where: Brisbane court Howse had attended an Origin series celebration party and B Why: To be inferred blew 0.21 – more than twice the legal alcohol limit – when Team: Identified breath-tested by police. Sponsor: Identified He pleaded guilty to the charge in a Brisbane court Player: Identified yesterday and was banned from driving for six months.

The Brisbane Broncos are sponsored by WOW Sight & Sound.

Who: Ben Howse Bronco to host charity benefit What 1: Cancer fundraising What 2: Gala benefit Tonight, Bronco Ben Howse will attend a gala benefit for Natalie Williams, a 20-year-old student who was last year When: Tonight diagnosed with an aggressive brain tumour. Where: Not provided C Why: To be inferred Howse has been instrumental in raising tens of thousands Team: Identified of dollars to cover medical expenses and a holiday for Sponsor: Identified Natalie and her mum. Player: Identified The Brisbane Broncos are sponsored by WOW Sight & Sound.

306

6.2: Covariate and Dependent variable scale items

There are six dependent variables. The ‘Person variables’ are: a) Person likability, and b) Attitude toward the Act. The ‘Corporate variables’ are: a) Corporate Image, and b) Attitude toward the Brand. Each of the ‘Corporate variables’ are measured once for the Team and once for the Sponsor.

Person Variables:

Person likeable Attitude toward the Act

Person Likeable

Instruction Please circle the number that best represents your feelings. Items: 1 How generally likable is the person described in the newspaper article? Scale size: 1-7 Likert-type scale, 1= very dislikeable, 7= very likeable, 4= neutral Sources Fiske, S.T. (1980). Attention and weight in person perception: the impact of negative and extreme behaviour. Journal of Personality and Social Psychology, 38 (6), pp889-906.

Klein, J.G. (1996). Negativity in impressions of presidential candidates revisited: the 1992 election. Personality and Social Psychology Bulletin, 22 (3), pp288-295.

Attitude toward the Act

Instruction Please circle the number that best represents your feelings. I believe the behaviour depicted in the newspaper article is: Items: 3 Bad-Good Harmful-Beneficial Safe-Unsafe Scale size: 1-7 Semantic differential Sources Bagozzi, R.P. (1982). A field of investigation of causal relations among cognitions, affect, intentions, and behavior. Journal of Marketing Research, 19 (November), pp562-584.

Oliver, R.L., & Bearden, W.O. (1985). Crossover effects in the theory of reasoned action: A moderating influence attempt. Journal of Consumer Research, 12 (December), pp324-340.

Raju, P.S., & Hastak, M. (1983). Pre-trial cognitive effects of cents-off coupons. Journal of Advertising, 12 (2), pp24-33.

307

Corporate Variables:

Team Corporate Image Attitude toward the Team (Attitude toward the Brand)

Sponsor Corporate Image Attitude toward the Sponsor (Attitude toward the Brand)

Corporate Image

Instruction Please circle the number that best represents your feelings. In your opinion, the [Team/Sponsor]… Items: 5 Has good products Is well managed Is involved in the community Responds to consumer needs Is a good company to work for Scale size: 1-7 Likert-type scale, 1=strongly disagree, 7=strongly agree, 4=neutral Sources Javalgi, R.G., Traylor, M.B., Gross, A.C., & Lampman, E. (1994). Awareness of sponsorship and corporate image: An empirical investigation. Journal of Advertising, 23 (4), pp47-58.

Pope, N.K.L., & Voges, K.E. (1999). Sponsorship and image: a replication and extension. Journal of Marketing Communications, 5 (1), pp17-28.

Attitude toward the Brand

Instruction My attitude toward the [Sponsor/Team] is: Items: 3 Bad-Good Unfavourable-Favourable Negative-Positive Scale size: 1-7 Semantic differential Sources Kinney, L., & McDaniel, S.R. (1996). Strategic implications of attitude-toward- the-Ad in leveraging event sponsorships. Journal of Sport Management, 10, pp250-261.

Muehling, D.D., & Laczniak, R.N. (1988). Advertising's immediate and delayed influence on brand attitudes: considerations across message-involvement levels. Journal of Advertising, 17 (4), pp23-34.

Covariate Variable:

Identification with the Team

Instruction Please indicate your agreement or disagreement with each of the statements below according to your feelings about the Brisbane Broncos. Items: 5 I have a strong sense of belonging to the team I think of the team as part of who I am I am like other fans of the team Others describe me as a typical fan of the team I have a strong attachment to the team Scale size: 1-7 Likert-type scale, 1= strongly disagree, 7= strongly agree, 4= neutral Sources Developed for this research, from pre-existing scales.

308

6.3: Introductory and debriefing scripts

Research introduction script

Good morning. I am doing research on sport sponsorship and news reporting.

This research requires a fairly large number of people to answer questions about how they feel about a particular rugby league team. I am here to ask if you would be willing to participate in my research and fill out the survey form.

Participation is completely voluntary. You are not required to participate as part of your enrolment in this course, and your participation (or non-participation) will have no effect on your grade for the course, your relationship with the course coordinator or tutors, or your relationship with QUT.

The survey is completely anonymous – it does not ask for your name. The survey also does not ask particularly private or sensitive questions – income, for example.

The survey form begins with questions asking you about how much you like a rugby team, there is then a short news article (2 paragraphs), and a series of questions follow. There are a total of 47 questions. The survey will probably take about 15 minutes to complete.

If you are interested in participating, I will be handing out the survey forms in the last 20 minutes of class time. If you do not want to participate, you will be free to leave, or to stay and study quietly.

Research debriefing script

Thank you for participating in this study. Your participation was very valuable to us. We know you are very busy and we appreciate the time you devoted to participating in the study.

There was some information about the study that I was not able to discuss with you prior to the study, because doing so probably would have impacted the decisions you made during the study. I would like to explain those things to you now.

In this study, we were interested in understanding the relationship between identification with a sporting team and response to different types of information (positive or negative) about a player from that team.

Based on prior research and information from the media, we would expect that people who identify with a team think positive actions are more positive than people who do not identify with the team. We would also expect that people who identify with a team do not rate negative behaviours as badly as people who do not identify with the team.

You were told that the behaviours reported were performed by members of a particular rugby league team. This was not true. In reality, the behaviours described, the activities have taken place. However, they were performed by various players from various teams.

The reason you were told the behaviours related to a single team was to make sure all participants were thinking about the same team when answering their questions. We hope this clarifies the purpose of the research, and the reason why deception was used.

Are there any questions or comments about the research? Thank you again for your time.

309

6.4: Group samples sizes and missing data

Group Likeable Att- Team – Att- Spons – Att- Act CI Team CI Spons

Assault N 62 62 62 63 63 63 Missing 1 1 1 0 0 0 Total 63 63 63 63 63 63

DUI N 66 66 65 66 66 67 Missing 2 2 3 2 2 1 Total 68 68 68 68 68 68

Cancer N 68 66 67 67 67 67 Missing 1 3 2 2 2 2 Total 69 69 69 69 69 69

Sample n = 200 4 6 6 4 4 3 % Missing by variable .02 .03 .03 .02 .02 .015

6.5: Summary of Case deletions

Group Case number/s Reason New Group n 89 Missing values Assault 64 Outlier 48 100 Outlier 1, 8, 14, 43, 143, Language confound 144, 145, 164, 166, 172, 196, 198 180 Missing values DUI 183 Missing values 48 149 Outlier 18, 21, 29, 53, 69, Language confound 73, 76, 103, 116, 118, 119, 128, 132, 133, 155, 175, 176 83 Missing values Charity fundraising 159 Missing values 55 25, 34, 35, 59, 78, Language confound 120, 123, 142, 161, 187, 189, 190

310

6.6: Matrices of scatterplots: tests of the linearity assumption

Assault

DUI

311

Charity

312

6.7: Summary of Tests of Assumptions

Assumption Test Violation Remedy or Effect on research Sample size equivalence Frequencies No Missing data Frequencies Yes Removal of cases with large amounts of missing data, and missing not at random data. Group-level Mean replacement for remaining missing values. Normality Skewness & Kurtosis Yes No transformation applied. Kolmogorov-Smirnov D Outliers Univariate: standardised z-scores Yes Cases removed ±3.29, p<.001 Multivariate: Mahalanobis distance No <24.322, p<.001 Linearity Scatterplots Yes nil Multicollinearity & Pearson’s r ≥.70 Yes Correlations within variable pairs are moderately high. Variables are Singularity conceptually distinct and correlated below the .90 level. All variables are retained. Variable reliability Cronbach’s alpha ≥.80 No Homogeneity of variance- Multivariate: Box’s M Yes Levene’s univariate test is reviewed. covariance matrices Univariate: Levene’s test Yes Data are not transformed. Test alpha is altered to .025 for test of confounds, and .0125 for final analysis. Homogeneity of regression Tests of Between-Subjects effects: No slopes Interaction of Group*Covariate

313

Appendix 9: Summary of tests of Hypotheses Hypothesis Accept/ p η2 Power Issue Reject 1 Valence of news’ will be reflected in evaluations of the athlete, their team and the team’s sponsor. 1a Valence of news’ will be reflected in Accept .000 .358 1.0 evaluation of the athlete’s behaviour (Attitude toward the Act). 1b Valence of news’ will be reflected in Accept .000 .326 1.0 evaluation of the athlete (Person likability). 1c Valence of news’ will be reflected in Reject .379 .013 .091 Insufficient evaluation of the team (Attitude toward power the Team). 1d Valence of news’ will be reflected in Reject .323 .015 .109 Insufficient evaluation of the image of the team power (Team Corporate Image). 1e Valence of news’ will be reflected in Reject .361 .014 .097 Insufficient evaluation of the sponsor (Attitude power toward the Sponsor). 1f Valence of news’ will be reflected in Reject .599 .007 .048 Insufficient evaluation of the image of the sponsor power (Sponsor Corporate Image).

2 Effects of Identification with the Team will be reflected in evaluations of the athlete, their team and the team’s sponsor 2a Team Identification will moderate Reject .028 .033 .380 Insufficient evaluations of the athlete’s behaviour power (Attitude toward the Act). 2b Team Identification will moderate Accept .000 .194 1.0 evaluations of the athlete (Person likability). 2c Team Identification will moderate Accept .000 .363 1.0 evaluations of the team (Attitude toward the Team). 2d Team Identification will moderate Accept .000 .186 .999 evaluations of the image of the team (Team Corporate Image). 2e Team Identification will moderate Reject .320 .007 .066 Insufficient evaluations of the sponsor (Attitude power toward the Sponsor). 2f Team Identification will moderate Reject .746 .001 .017 Insufficient evaluations of the image of the sponsor power (Sponsor Corporate Image).

3 The valence of the reported behaviour Mixed (positive/ negative) will be maintained across evaluations of the athlete, their team, and the sponsor.

314

315

References

Aaker, J. L. (1997). Dimensions of brand personality. Journal of Marketing Research, 34 (August), 347-56.

Aaker, J., Fournier, S., and Brasel, S.A. (2004). When Good Brands do Bad. Journal of Consumer Research, 31 (1), 1-16.

Abrams, D., Ando, K., & Hinkle, S. (1998). Psychological Attachment to the Group: Cross-Cultural Differences in Organizational Identification and Subjective Norms as Predictors of the Workers’ Turnover Intentions. Personality and Social Psychology Bulletin, 24 (10), 1027-1039.

Abrams, D., and Hogg, M.A. (1988). Comments on the motivational status of self- esteem in social identity and intergroup discrimination. European Journal of Social Psychology, 18, 317-34.

Abrams, D., Wetherell, M., Cochrane, S., Hogg, M.A., and Turner, J.C. (1990). Knowing what to think by knowing who you are: Self-categorization and the nature of norm formation, conformity and group polarization. British Journal of Social Psychology, 29 (2), 97-119.

Accenture. (2009, December 13). Accenture (NYSE: ACN) today announced that it will not continue its sponsorship agreement with Tiger Woods. http://newsroom.accenture.com/article_display.cfm?article_id=4915 Last accessed: 10 February 2012

Ahearne, M., Bhattacharya, C.B., and Gruen, T. (2005). Antecedents and Consequences of Customer-Company Identification: Expanding the Role of Relationship Marketing. Journal of Applied Psychology, 90 (3), 574-85.

Alba, J.W., and Hasher, L. (1983). Is memory schematic? Psychological Bulletin, 93 (2), 203-31.

Alba, J.W., & Hutchinson, J.W. (1987). Dimensions of consumer expertise. Journal of Consumer Research, 13 (4), 411-54.

Alexander, N. (2009). Defining brand values through sponsorship. International Journal of Retail & Distribution Management, 37 (4), 346-57.

Almeida, J. (2006). The semantic/episodic distinction: the case for social information processing. Journal of Experimental Social Psychology, 43, 842-49.

Anand, P., Holbrook, M.B. and Stephens, D. (1988). The Formation of Affective Judgments: The Cognitive-Affective Model Versus the Independence Hypothesis. Journal of Consumer Research, 15 (3), 386-91.

316

Anderson, N.H. (1981). Foundations of information integration theory. New York: Academic Press.

Anderson, J.R. (1983). The Architecture of Cognition. USA: Harvard University Press.

Anderson, J.R., and Bower, Gordon. H. (1973, 1974). Human associative memory. Washington: Winston.

Arbuckle, J.L. (2007). Amos 18 User’s Guide. USA: Amos Development Corporation.

Argo, J.J., Dahl, D.W., and Morales, A.C. (2008). Positive Consumer Contagion: Responses to Attractive Others in a Retail Context. Journal of Marketing Research, 45 (6), 690-701.

Arrow, K.J. (1963). Uncertainty and the welfare economics of medical care. The American Economic Review, 53 (5), 941-73.

Asch, S.E. (1946). Forming impressions of personality. Journal of Abnormal Social Psychology, 41, 258-90.

Ashcraft, R. (1968). Locke's state of nature: historical fact or moral fiction? The American Political Science Review, 62 (3), 898-915.

Ashforth, B.E., and Mael, F. (1989). Social identity theory and the organisation. The Academy of Management Review, 14 (1), 20-39.

Ashmore, R.D., Deaux, K., and McLaughlin-Volpe, T. (2004). An organizing framework for collective identity: Articulation and significance of multidimensionality. Psychological Bulletin, 130 (1), 80-114.

Assael, J. & Day, G.S. (1968). Attitudes and Awareness as Predictors of Market Share. Journal of Advertising Research (October), 3-12.

Atkinson, R.C., and Shiffrin, R.M. (1968). Human memory: a proposed system and its control processes, In The Psychology of Learning and Motivation: Advances in Research and Theory, vol. 2. K.W. Spence and J.T. Spence (Eds.). New York: Academic Press, 89-195.

Auerbach, D. (2005-06). Morals clauses as corporate protection in athlete endorsement contracts. DePaul Journal of Sports Law & Contemporary Problems, 1- 18.

Augustine-Schlossinger, L. (2003,). Legal considerations for sponsorship contracts of Olympic athletes. Villanova Sports & Entertainment Law Journal, 10, 281-96.

Australian Bureau of Statistics. (1999). 1996-97 Business Sponsorship: Australia. Catalogue: 41440.0. Australian Bureau of Statistics: Canberra.

317

Australian Bureau of Statistics. (2008). 2006-07 Performing Arts: Australia. Catalogue: 8697.0. Australian Bureau of Statistics: Canberra.

ABS (2008-09a). 2008-09 Regional Population Growth, Australia. Catalogue number: 3218.0. Australian Bureau of Statistics: Canberra.

ABS (2008-09b). Criminal Courts. Catalogue number: 4513.0. Australian Bureau of Statistics: Canberra.

Australian Bureau of Statistics. (2009a). 2009 Sports and Physical Recreation: A Statistical Overview, Australia. Catalogue: 4156.0. Australian Bureau of Statistics: Canberra.

Australian Bureau of Statistics. (2010a). Arts and Culture in Australia: A Statistical Overview: Australia. Catalogue: 4172.0. Australian Bureau of Statistics: Canberra.

Australian Bureau of Statistics. (2010b). 2009-10 Spectator Attendance at Sporting Events: Australia. Catalogue: 4174.0. Australian Bureau of Statistics: Canberra.

Australian Bureau of Statistics. (2010c). 2009-10 Attendance at Selected Cultural Venue and Events: Australia. Catalogue: 41440.0. Australian Bureau of Statistics: Canberra.

ABS (2010d). National Regional Profile: Queensland. Australian Bureau of Statistics: Canberra.

Australian Bureau of Statistics. (2011). Sports and Physical Recreation: A Statistical Overview: Australia. Catalogue: 4156.0. Australian Bureau of Statistics: Canberra.

Australian Bureau of Statistics. (2011). Arts and Culture in Australia: A Statistical Overview: Australia. Catalogue: 4172.0. Australian Bureau of Statistics: Canberra.

Ajzen, I., and Fishbein, M. (1970). The prediction of behaviour from attitudinal and normative variables. Journal of Experimental Social Psychology, 6, 466-87.

Bagozzi, R.P. (1982). A field investigation of causal relations among cognitions, affect, intentions, and behavior. Journal of Marketing Research, 19 (4), 562-83.

Bai, Y., Middlestadt, S.E., Peng, C-Y. J., and Fly, A.D. (2010). Predictors of continuation of exclusive breastfeeding for the first six months of life. Journal of Human Lactation, 26 (1), 26-34.

Baker, T. (1996-1997). On the genealogy of moral hazard. Texas Law Review, 75 (2), 237-92.

Bal, C., Quester, P., and Plewa, C. (2010). Emotions and sponsorship: A key to global effectiveness? A comparative study of Australia and France. Asia Pacific Journal of Marketing & Logistics, 22 (1), 40-54.

318

Barreto, M., and Ellemers, N. (2001). You can’t always do what you want: Social identity and self-presentational determinants of the choice to work for a low-status group. Personality and Social Psychology Bulletin, 26 (8), 891-906.

BBC SPORT (2009). Phelps accepts ‘fair’ punishment. http://news.bbc.co.uk/go/pr/fr/-/sport2/hi/olympic_games/7873669.stm. Posted: 2009/02/06 Last accessed: 11 February 2012

BBC SPORT (2010). Twitter rant costs Rice sponsor. http://news.bbc.co.uk/go/pr/fr/- /sport2/hi/other_sports/swimming/8976165.stm. Posted: 2010/09/07 Last accessed: 11 February 2012

Bhattacharya, C.B., Rao, H., and Glynn, M. (1995). Understanding the Bond of Identification: An Investigation of its Correlates among Art Museum Members. Journal of Marketing, 59 (4), 46-57.

Baum, M.A., and Potter, P.B.K. (2008). The relationships between mass media, public opinion, and foreign policy: toward a theoretic synthesis. Annual Review of Political Science, 11, 39-65.

Bek, M.G. (2004). Research note: Tabloidization of News Media: An Analysis of Television News in Turkey. European Journal of Communication, 19 (3), 371-86.

Benford, R.D. (2007). The college sports reform movement: reframing the “edutainment” industry. The Sociological Quarterly, 48, pp1-28.

Benjamini, Y., & Hochberg, Y. (1995). Controlling the false discovery rate: a practical and powerful approach to multiple testing. Journal of the Royal Statistical Society, B series, 57 (1), pp289-300.

Bennett, R. (1999). Sport sponsorship, spectator recall and false consensus. European Journal of Marketing, 33 (3/4), 291-313.

Berkes, P., Nyerges, M., and Vaczi, J. (2009). The changing of the sponsorship marketing: The case of Hungarian professional soccer club sponsors. Sportwissenschaft, 39, 35-44.

Bernache-Assolant, I., Lacassagne, M-F., and Braddock, J.H. (2007). Basking in reflected glory and blasting: differences in identity-management strategies between two groups of highly identified soccer fans. Journal of Language and Social Psychology, 26 (4), 381-88.

Bettman, J.R. (1970). Information processing models of consumer behaviour. Journal of Marketing Research, 7 (3), 370-76.

Bettman, J.R. (1979). Memory factors in consumer choice: a review. Journal of Marketing, 43 (Spring), 37-53.

319

Bhattacharya, C.B., Rao, Hayagreeva., and Glynn, Mary Ann. (1995). Understanding the Bond of Identification: An Investigation of its Correlates among Art Museum Members. Journal of Marketing, 59 (4), 46-57.

Biehal, G., and Chakravarti, D. (1982). Information-presentation format and learning goals as determinants of consumers' memory retrieval and choice processes. Journal of Consumer Research, 8 (4), 431-41.

Billig, M. and Tajfel, H. (1973). Social categorization and similarity in intergroup behaviour. European Journal of Social Psychology, 13 (1), 27-52.

Birnbaum, M.H. (1972). Morality judgments: tests of an averaging model. Journal of Experimental Psychology, 93, 35-42.

Biscomb, K., and Griggs, G. (2012). ‘A splendid effort!’: Print media reporting of England’s women’s performance in the 2009 Cricket World Cup. International Review for the Sociology of Sport, 1-13.

Biskup, C., and Pfister, G. (1999). I would like to be like her/him: Are athletes role- models for boys and girls? European Physical Education Review, 5 (3), 199-218.

Blood, W., and Holland, K. (2004). Risky news, madness and public crisis: a case study of the reporting and portrayal of mental health and illness in the Australian press. Journalism, 5 (3), 323-342.

Boen, F., Vanbeselaere, N., and Feys, J. (2002). Behavioural consequences of fluctuating group success: an internet study of soccer-team fans. The Journal of Social Psychology, 142 (6), 769-81.

Borland, J., and Macdonald, R. (2003). Demand for Sport. Oxford Review of Economic Policy, 19 (4), 478-502.

Bornstein, R.F. (1989). Exposure and Affect: Overview and Meta-Analysis of Research, 1968-1987. Psychological Bulletin, 106 (2), 265-89.

Boruch, R.F. (1998). Randomized Controlled Experiments for Evaluation and Planning. In, L. Bickman & D.J. Rog (Eds.), Handbook of applied social research methods (pp161-191). USA: SAGE Publications, Inc.

Bransford, J.D. and Franks, J.J. (1972). The abstraction of linguistic ideas: A review. Cognition, 1 (2-3), 211-49.

Brazeal, L. (2008). The image repair strategies of Terrell Owens. Public Relations Review, 34, 145-50.

Breckler, S.J. (1984). Empirical validation of affect, behaviour, and cognition as distinct components of attitude. Journal of Personality and Social Psychology, 47 (6), 1191-1205.

320

Brewer, M.B. (1979). In-group bias in the minimal intergroup situation: A cognitive- motivational analysis. Psychological Bulletin, 86, 307-24.

Brewer, M.B. (1993). Social identity, distinctiveness, and in-group homogeneity. Social Cognition, 11 (1), 150-64.

Brewer, M.B. (2001). The Many Faces of Social Identity: Implications for Political Psychology. Political Psychology, 22 (1), 115-25.

Brewer, M.B., and Gardner, W. (1996). Who is this “we”? Levels of collective identity and self representations. Journal of Personality and Social Psychology, 71, 83-93.

Brewer, R.M., and Pedersen, P.M. (2010). Franchises, value drivers and the application of valuation analysis to sports sponsorship. Journal of Sponsorship, 3 (2), 181-93.

Brinson, S.L., and Benoit, W.L. (1999). The tarnished star: Restoring Texaco’s damaged public image. Management Communication Quarterly, 12 (4), 483-510.

Bristow, D.N., and Sebastian, R.J. (2003). Holy cow! Wait ‘til next year! A closer look at the brand loyalty of Chicago Cubs baseball fans. Journal of Consumer Marketing, 18 (3), 256-75.

Brown, T.A. (2000). Confirmatory Factor Analysis for Applied Research. The Guildford Press: USA.

Brown, T.A. (2006). Confirmatory Factor Analysis for Applied Research. USA: The Guilford Press.

Brown, W.J., Basil, M.D., and Bocarnea, M.C. (2003). The influence of famous athletes on health beliefs and practices: Mark McGwire, child abuse prevention, and Androstenedione. Journal of Health Communication, 8, 41-57.

Brown, R.D., and Bassili, J.N. (2002). Spontaneous trait associations and the case of the superstitious banana. Journal of Experimental Social Psychology, 38, 87-92.

Brown,R., Condor, S., Mathews, A., Wade, G. and Williams, J. (1986). Explaining intergroup differentiation in an industrial organization. Journal of Occupational Psychology, 59, 273-86.

Brown, T.J., and Dacin, P.A. (1997). The company and the product: corporate associations and consumer product responses. Journal of Marketing, 61 (January), 68-84.

Brown, M., Pope, N., and Voges, K. (2003). An exploration of shopping orientations and online purchase intentions. European Journal of Marketing, 37 (11/12), 1666- 1684.

321

Brown, R. and Williams, J. (1984). Group identification: The same thing to all people? Human Relations, 37 (7), 547-64.

Browne, M.W., and Cudek, R. (1993). Alternative ways of assessing model fit. In, K.A. Bollen & J.S. Long (Eds.), Testing Structural Equation Models, (pp136-162). USA: Sage.

Bruner, G.C. II., and Hensel, P.J. (1994). Marketing Scales Handbook: A compilation of multi-item measures. Chicago: American Marketing Association.

Bruner, G.C., James, K., and Hensel, P.J. (2001). Marketing scales handbook: a compilation of multi-item measures. Volume 3. Chicago: American Marketing Association.

Bunketorp, L., Carlsson, J., Kowalski, J., and Stener-Victorin, E. (2005). Evaluating the reliability of multi-item scales: A non-parametric approach to the ordered categorical structure of data collected with the Swedish version of the Tampa Scale for Kenesiophobia and the Self-Efficacy Scale. Journal of Rehabilitation Medicine, 37, 330-34.

Burke, R.R., and Srull, T.K. (1988). Competitive interference and consumer memory for advertising. Journal of Consumer Research, 15 (1), 55-68.

Burnett, J.J. (1998). A Strategic Approach to Managing Crises. Public Relations Review, 24 (4), 475-88.

Burnett, J., Menon, A., and Smart, D.T. (1993). Sports Marketing: A New Ball Game with New Rules. Journal of Advertising Research (September/October), 21-35.

Burton, N., and Chadwick, S. (2009). Ambush marketing in sport: An analysis of sponsorship protection means and counter-measures. Journal of Sponsorship, 2 (4), 303-15.

Burke, R.R., and Srull, T.K. (1988). Competitive interference and consumer memory for advertising. Journal of Consumer Research, 15 (1), 55-68.

Byrnes, H., and Phelps, J. (2009, May 21). Major sponsor pulls plug on the Cronulla Sharks. The Daily Telegraph. http://www.dailytelegraph.com.au/sport/nrl/major- sponsor-pulls-plug-on-the-cronulla-sharks/story-e6frexrr-1225713946343

Cairns, J., Jennett, N., and Sloane, P.J. (1986). The economics of professional team sports: a survey of theory and evidence. Journal of Economic Studies, 13 (1), 3-80.

Calderon-Martinez, A., Mas-Ruis, F., and Nicolau-Gonzalbez, J.L. (2005). Commercial and philanthropic sponsorship: Direct and interaction effects on company performance. International Journal of Market Research, 47 (1), 75-99.

Cameron, N. (2009). Understanding sponsorship and its measurement implications. Journal of Sponsorship, 2 (2), 131-39.

322

Campbell, R.M., Aiken, D., and Kent, A. (2004). Beyond BIRGing and CORFing: continuing the exploration of fan behaviour. Sport Marketing Quarterly, 13, 151-57.

(The) Canadian Press (2009, December 19). GLF- Tiger- Woods-Tag- Heuer.

Capranica, L., and Aversa, F. (2002).Italian television sport coverage during the 2000 Sydney Olympic Games: A gender perspective. International Review for the Sociology of Sport, 37 (3-4), 337-49.

Carlston, D.E., and Mae, L. (2007). Posing with the flag: trait-specific effects of symbols on person perception. Journal of Experimental Social Psychology, 43, 241- 48.

Carlston, D.E., and Skowronski, J.J. (1986). Trait memory and behaviour memory: the effects of alternative pathways on impression judgment response times. Journal of Personality and Social Psychology, 50 (1), 5-13.

Carlston, D.E., and Skowronski, J.J. (2005). Linking versus thinking: evidence for the different associative and attributional bases of spontaneous trait transference and spontaneous trait inference. Journal of Personality and Social Psychology, 89 (6), 884-98.

Carrillat, F.A., Lafferty, B.A., and Harris, E.G. (2005). Investigating sponsorship effectiveness: Do less familiar brands have an advantage over more familiar brands in single and multiple sponsorship arrangements? Journal of Brand Management, 13 (1), 50-64.

Cassidy, C. and Trew, K. (2001). Assessing identity change: A longitudinal study of the transition from school to college. Group Processes and Intergroup Relations, 4 (1), 49-60.

Cattell, R.B. (1966). The scree test for the number of factors. Multivariate Behavioural Research, 1, 629-37.

Caves, R.E. (2003). Contracts between art and commerce. The Journal of Economic Perspectives, 17 (2), 73-84.

Chan, J., and Suen, W. (2009). Media as watchdogs: the role of news media in electoral competition. European Economic Review, 53, 799-814.

Chappelet, J-L. (2001). Risk Management for Large-Scale Events: The Case of the Olympic Winter Games. European Journal for Sport Management: Special Issue 2001, 6-21.

Chien, P-H.M., Cornwell, T.B., and Stokes, R. (2005). A Theoretical Framework for Analysis of Image Transfer in Multiple Sponsorships. In, S.Purchase (Ed.), Proceedings of ANZMAC: Broadening the Boundaries. Fremantle, WA: 5-7 December.

323

Child, D. (1970). The Essentials of Factor Analysis. London: Holt, Rinehart & Winston.

Chiou, J-S., Huang, C-Y., and Lee, H-H. (2005). The antecedents of music piracy attitudes and intentions. Journal of Business Ethics, 57, 161-74.

Chiu, C-Y., Dweck, C.S., Tong, J. Y-Y., and Fu, J. H-Y. (1997). Implicit theories and conceptions of morality. Journal of Personality and Social Psychology, 73 (5), 923-40.

Christopherson, N., Janning, M., and McConnell, E.D. (2002). Two kicks forward, One kick back: A content analysis of media discourses on the 1999 women’s World Cup Soccer Championship. Sociology of Sport Journal, 19, 170-88.

Cialdini, R.B., Borden, R.J., Thorne, A., Walker, M.R., Freeman, S., and Sloan, Ll.R. (1976). Basking in reflected glory: three (football) field studies. Journal of Personality and Social Psychology, 34 (3), 366-75.

Cialdini, R.B., and de Nicholas, M.E. (1989). Self-presentation by association. Journal of Personality and Social Psychology, 57 (4), 626-31.

Cialdini, R.B., and Richardson, K.D. (1980). Two indirect tactics of image management: basking and blasting. Journal of Personality and Social Psychology, 39 (3), 406-15.

Clark, J.M., Cornwell, T.B., and Pruitt, S.W. (2002). Corporate Stadium Sponsorships, Signaling Theory, Agency Conflicts, and Shareholder Wealth. Journal of Advertising Research, (November/December), 16-32.

Clark, J. M., Cornwell, T.B., and Pruitt, S.W. (2009). The Impact of Title Event Sponsorship Announcements on Shareholder Wealth. Marketing Letters, 20, 169-82.

Cliffe, S.J., and Motion, J. (2005). Building contemporary brands: a sponsorship- based strategy. Journal of Business Research, 58, 1068-1077.

Cohen, A.J. (2007). What the Liberal State Should Tolerate Within Its Borders. Canadian Journal of Philosophy, 37 (4), 479-514.

Cohen, B.C. (1963). The press and foreign policy. Princeton, NJ: Princeton University Press.

Cohen, J. (1988). Statistical power analysis for the behavioural sciences. New Jersey: Erlbaum.

Cohen, J. (1992). A power primer. Psychological Bulletin, 112, 155-59.

Cohen, J. (1994). The earth in round (p<.05). American Psychologist, December, 997-1003.

324

Collins, A.M., and Loftus, E.F. (1975). A Spreading-Activation Theory of Semantic Processing. Psychological Review, 82 (6), 407-28.

Conagin, A., and Barbin, D. (2006). Bonferroni’s and Sidak’s modified tests. Scientia Agricola, 63 (1), 70-76.

Conover, W.J. (1980). Practical Nonparametric Statistics: Second edition. USA: John Wiley & Sons, Inc.

Coombs, W.T. (2007). Attribution Theory as a guide for post-crisis communication research. Public Relations Review, 33, 135-39.

Copp, D. (1997). Does moral theory need the concept of society? Analyse & Kritik, 19, pp189-212.

Coppetti, C., Wentzel, D., Tomczak, T., and Henkel, S. (2009). Improving incongruent sponsorships through articulation of the sponsorship and audience participation. Journal of Marketing Communications, 15 (1), 17-34.

Cornwell, T.B. (1995). Sponsorship-linked marketing development. Sport Marketing Quarterly, 4 (4), 13-24.

Cornwell, T.B. (2008). State of the art and science in sponsorship-linked marketing.Journal of Advertising, 37 (3), 41-55.

Cornwell, T.B., and Coote, L.V. (2005). Corporate sponsorship of a cause: the role of identification in purchase intent. Journal of Business Research, 58, 268-76.

Cornwell, T. B., Humphreys, M S., Maguire, A.M., Weeks, C.S. and Tellegen, C.L. (2006). Sponsorship-linked marketing: The role of articulation in memory. Journal of Consumer Research, 33 (December), 312-21.

Cornwell, T. B., and Maignan, I. (1998). An international review of sponsorship research. Journal of Advertising, 27 (1), 1-21.

Cornwell, T. B., Pruitt, S.W., and Clark, J.M. (2005). The Relationship between Major-League Sports’ Official Sponsorship Announcements and the Stock Prices of Sponsoring Firms. Journal of the Academy of Marketing Science, 33 (4), 401-12.

Cornwell, T. B., Pruitt, S.W., and Van Ness, R. (2001). The Value of Winning in Motorsports: Sponsorship-linked Marketing. Journal of Advertising Research, (January/February), 17-31.

Cornwell, T. B., Roy, D.P., and Steinard, E.A. (2001). Exploring managers’ perceptions of the impact of sponsorship on brand equity. Journal of Advertising, 30 (2), 41-51.

Cornwell, T.B., Weeks, C.S., and Roy, D.P. (2005). Sponsorship-linked marketing: opening the black box. Journal of Advertising, 34 (2), 21-42.

325

Craik, F.I.M. (2002). Levels of processing: past, present... and future? Memory, 10 (5-6), 305-18.

Craik, F.I.M. (2007). Encoding: a cognitive perspective. In, H.L. Roediger III, Y. Crossman, J., Hyslop, P., and Guthrie, B. (1994). A content analysis of the sports section of Canada’s national newspaper with respect to gender and professional/ amateur status.. International Review for the Sociology of Sport, 29 (2), 123-32.

Crossman, J., Vincent, J., and Speed, H. (2007). ‘The Times they are a-changin’: Gender comparisons in three national newspapers of the 2004 Wimbledon Championships. International Review for the Sociology of Sport, 42 (1), 27-41.

Dudai & S.M. Fitzpatrick, Science of Memory: concepts. Great Britain: Oxford University Press, 129-35.

Craik, F.I.M., and Lockhart, R.S. (1972). Levels of processing: a framework for memory research. Journal of Verbal Learning and Verbal Behavior, 11, 671-84.

Crandall, C.S., Silvia, P.J., N'Gbala, A.N., Tsang, J., and Dawson, K. (2007). Balance theory, unit relations, and attribution: the underlying integrity of Heiderian Theory. Review of General Psychology, 11 (1), 12-30.

Crepeau, R.C. (1981). Sport, heroes and myth. Journal of Sport and Social Issues, 5, 23-31.

Crompton, J.L. (1994). Benefits and risks associated with sponsorship of major events. Festival Management & Event Tourism, 2, 65-74.

Crompton, J.L. (2004). Conceptualization and alternate operationalizations of the measurement of sponsorship effectiveness in sport. Leisure Studies, 23 (3), 267-81.

Crowder, R.G. (1976). Principles of learning and memory. NJ: Lawrence Erlbaum.

Cunningham, S., Cornwell, T.B., and Coote, L. (2009,). Expressing identity and shaping image: The relationship between corporate mission and corporate sponsorship. Journal of Sport Management, 23, 65-86.

Curran, J., Salovaar-Moring, I., Coen, S. and Iyengar, S. (2010). Crime, foreigners and hard news: A cross-national comparison of reporting and public perception. Journalism, 11(1), 3-19.

D’Astous, A., and Bitz, P. (1995). Consumer evaluations of sponsorship programmes. European Journal of Marketing, 29 (12), 6-22.

Dardis, F.E. (2009). Attenuating the negative effects of perceived incongruence in sponsorship: how message repetition can enhance evaluations of an “incongruent sponsor”. Journal of Promotion Management, 15 (1), 36-56.

326

Dalakas, V., and Levin, A.M. (2005). The Balance Theory Domino: How Sponsorship May Elicit Negative Consumer Attitudes. Advances in Consumer Research, 32, 91-97.

Dann, S. (2007) Lifestyle sponsorships and player lifestyle breaches: turning transgressions into opportunities for lifestyle message endorsements. Monash Business Review, 3 (2), 1-9.

Dean, D.H. (2002). Associating the Corporation with a Charitable Event through Sponsorship: Measuring the Effects on Corporate Community Relations. Journal of Advertising, 31 (4), 77-87.

Deaux, K. (1996). Social identification. In, E.T. Higgins & A.W.Kruglanski (Eds.), Social Psychology: Handbook of basic principles. New York: Guilford Press, pp777- 798.

DeCoster, J., and Claypool, H.M. (2004). A meta-analysis of priming effects on impression formation supporting a general model of information biases. Personality and Social Psychology Review, 8 (2), 2-27.

DeGaris, L., West, C., and Dodds, M. (2009). Leveraging and activating NASCAR with NASCAR-linked sales promotions. Journal of Sponsorship, 3 (1), 88-97.

Dees, W., Bennett, G., and Villegas, J. (2008). Measuring the Effectiveness of Sponsorship of an Elite Intercollegiate Football Program. Sport Marketing Quarterly, 17, 79-89.

Dellarosa, D., and Bourne, L.E. (1984). Decisions and Memory: Differential Retrievability of Consistent and Contradictory evidence. Journal of Verbal Learning and Verbal Behavior, 23, 669-82.

Denham, B.E. (2004). Hero or hypocrite? United States and international media portrayals of Carl Lewis amid revelations of a positive drug test. International Review for the Sociology of Sport, 39 (2), 167-85.

Dick, A., Chakravarti, D., and Biehal, G. (1990). Memory-based inferences during consumer choice. The Journal of Consumer Research, 17 (1), 82-93.

Dimitrov, R. (2008). Gender violence, fan activism and public relations in sport: The case of “Footy Fans against Sexual Assault”. Public Relations Review, 34, 90-98.

Dimmock, J.A., Grove, J. R., and Eklund, R.C. (2005). Reconceptualising team identification: new dimensions and their relationship to intergroup bias. Group Dynamics: Theory, Research, and Practice, 9 (2), 75-86.

Dimmock, J.A., and Gucciardi, D.F. (2008). The utility of modern theories of intergroup bias for research on antecedents of team identification. Psychology of Sport and Exercise, 9, 284-300.

327

Dionisio, P., Leal, C., and Moutinho, L. (2008). Fandom affiliation and tribal behaviour: a sports marketing application. Qualitative Market Research: An International Journal, 11 (1), 17-39.

Dittmer, R. (2006). The allegory of the stadium: sport and true education. Unpublished Masters’ thesis: Creighton University.

Doms, M. and Morin, N. (2004). Consumer sentiment, the economy, and the news media. Finance and Economics Discussion Series 2004- 51. Federal Reserve Board, Washington D.C.

Donahay, B., and Rosenberger, P.J. (2007). Using Brand Personality to Measure the Effectiveness of Image Transfer in Formula One Racing. Marketing Bulletin, 18, 1- 15.

Doosje, B.E.J.., Branscombe, N.R., Spears, R., and Manstead. A.S.R. (2006). Antecedents and Consequences of Group-Based Guilt: The Effects of Ingroup Identification. Group Processes & Intergroup Relations, 9 (3), 325-38.

Dowling, G.R. (1986). Managing your corporate images. Industrial Marketing Management, 15, 109-115.

Dreben, E.K., Fiske, S.T. and Hastie, R. (1979). The independence of evaluative and item information: impression and recall order effects in behaviour-based impression formation. Journal of Personality and Social Psychology, 37 (10), 1758-68.

Driedger, L. (1976). Ethnic self-identity: A comparison of ingroup evaluations. Sociometry, 39 (2), 131-41.

Driver, J. (2007). Ethics: the fundamentals. Singapore: Blackwell Publishing.

Dudai, Y., Roediger III, H.L., and Tulving, E. (2007). Memory concepts. In, H.L. Roediger III, Y. Dudai & S.M. Fitzpatrick, Science of Memory: concepts. Great Britain: Oxford University Press, 17-21.

Dueze, M. (2005). What is journalism? Professional identity and ideology of journalists reconsidered. Journalism, 6 (4), 442-64.

Dutton, J.E., Dukerich, J.M., and Harquail, C.V. (1994). Organizational Images and Member Identification. Administrative Science Quarterly, 39 (2), 239-63.

Eagly, A. H., and Chaiken, S. (2007). The advantages of an inclusive definition of attitude. Social Cognition, 25 (5), 582-602.

Eccleston, C.P., & Major, B.N. (2006). Attributions to discrimination and self- esteem: the role of group identification and appraisals. Group Processes & Intergroup Relations, 9 (2), pp147-162.

Edwards, K. (1990). The Interplay of Affect and Cognition in Attitude Formation and Change. Journal of Personality and Social Psychology, 59 (2), 202-16.

328

Egmond, F. (1995). The cock, the dog, the serpent, and the monkey. Reception and transmission of a Roman punishment, or historiography as history. International Journal of the Classical Tradition, 2 (2), 159-192.

Eich, J.M. (1985). Levels of Processing, Encoding Specificity, Elaboration, and CHARM. Psychological Review, 92 (1), 1-38.

Ellemers, N. (1993). The influence of socio-structural variables on identity management strategies. European Journal of Social Psychology, 4 (1), 27-57.

Ellemers, N., de Gilder, D., & Haslam, S.A. (2004). Motivating individuals and groups at work: a social identity perspective on leadership and group performance. Academy of Management Review, 29 (3), pp459-478.

Ellemers, N., Kortekaas, P., and Ouwerkerk, J.W. (1999).Self-categorisation, commitment to the group and group self-esteem as related by distinct aspects of social identity. European Journal of Social Psychology, 29, 371-89.

Ellemers, N., Pagliaro, S., Barreto, M., and Leach, C.W. (2008). Is it better to be moral than smart? The effects of morality and competence norms on the decision to work at group status improvement. Journal of Personality and Social Psychology, 95 (6), 1397-1410.

Ellemers, N., Van Knippenberg, A., de Vries, N., and Wilke, H. (1988). Social identification and permeability of group boundaries. European Journal of Social Psychology, 18, 497-513.

Entman, R.M. (1991) Framing U.S. Coverage of international news: contrasts in narratives of the KAL and Iran air incidents. Journal of Communication, 41 (4), 6-27.

Entman, R.M. (1993). Framing: towards clarification of a fractured paradigm. Journal of Communication, 43 (4), 51-58.

Erdem, T., Swait, J., Broniarczyk, S., Chakravarti, D., Kapferer, J-N., Keane, M., Roberts, J., Steenkamp, J-B.E.M., & Zettelmeyer, F. (1999). Brand equity, consumer learning and choice. Marketing Letters, 10 (3), 301-18.

EWCA Civ 1051 (2010). Case No: A2/2009/2493 Force India Formula One Team Limited v Etihad Airways PJSC and Aldar Properties PJSC.

Estes, W.K. (1960). Learning Theory and the New “Mental Chemistry”. Psychological Review, 67 (4), 207-23.

Farah, M.F., and Newman, A.J. (2010). Exploring consumer boycott intelligence using a socio-cognitive approach. Journal of Business Research, 63, 347-55.

Farrell, K. A., and Frame, W. S. (1997). The Value of Olympic Sponsorships: Who is Capturing the Gold? Journal of Market Focused Management, 2, 171-82.

329

Farrelly, F.J., and Quester, P.G. (2005). Examining important relationship quality constructs of the focal sponsorship exchange. Industrial Marketing Management, 34, 211-19.

Farrelly, F., Quester, P., and Burton, R. (2006). Changes in sponsorship value: Competencies and capabilities of successful sponsorship relationships. Industrial Marketing Management, 35, 1016-1026.

Farrelly, F., Quester, P., and Greyser, S.A. (2005). Defending the Co-Branding Benefits of Sponsorship B2B Partnerships: The Case of Ambush Marketing. Journal of Advertising Research, (September), 339-48.

Fasting, K., and Tangen, J. (1983). Gender and Sport in Norwegian Mass Media. International Review for the Sociology of Sport, 18, 61-70.

Feldman, J.M., and Lynch, J.G. (1988) Self-generated validity and other effects of measurement on belief, attitude, intention, and behaviour. Journal of Applied Psychology, 73 (3), 421-35.

Feldman, S.M., and Underwood, B.J. (1957). Stimulus recall following paired- associate learning. Journal of Experimental Psychology, 53 (1), 11-15.

Felten, J-B. (2009). Is sponsorship planning on a par with ‘classic’ media planning? Journal of Sponsorship, 2 (4), 379-86.

Fenton, W. (2009). The global sponsorship marketing. Journal of Sponsorship, 2 (2), 120-30.

Fernandez, K.V., and Lastovicka, J.L. (2011). Making Magic: Fetishes in Contemporary Consumption. Journal of Consumer Research, 38 (2), 278-99.

Ferrari, F., and Reaber, G. (2009). Dispositional Theories of Values: Michael Smith. Contextualism and Relativism Seminar, 20 October.

Ferreira, M., Hall, T.K and Bennett, G. (2008). Exploring brand positioning in a sponsorship context: a correspondence analysis of the Dew Action Sports Tour. Journal of Sport Management, 22, 734-61.

Fielding, K.S., Hogg, M.A., & Annandale, N. (2006). Reactions to positive deviance: social identity and attribution dimensions. Group Processes & Intergroup Relations, 9 (2), pp199-218.

Finch, D., O’Reilly, N., Varella, P., and Wolf, D. (2009). Return on trust: An empirical study of the role of sponsorship in stimulating consumer trust and loyalty. Journal of Sponsorship, 3 (1), 61-72.

Fink, J., Parker, H. M., Brett, M., and Higgins, J. (2009). Off-field Behavior of Athletes and Team Identification: Using Social Identity Theory and Balance Theory to Explain Fan Reactions. Journal of Sport Management, 23, 142-55.

330

Fishbein, M. and Ajzen, I. (1975). Belief, Attitude, Intention and Behavior. Reading, MA: Addison-Wesley.

Fiske, S.T. (1980). Attention and weight in person perception: the impact of negative and extreme behaviour. Journal of Personality and Social Psychology, 38 (6), 889- 906.

Fiske, S.T., and Neuberg, S.L. (1990). A continuum of impression formation, from category-based to individuating processes: influences of information and motivation on attention and interpretation. Advances in Experimental Social Psychology, 23, 1- 73.

Fitzsimons, G.J., Nunes, J.C., and Williams, P. (2007). License to Sin: The Liberating Role of Reporting Expectations. Journal of Consumer Research, 34 (June), 22-31.

Folkes, V.S. (1988). Recent Attribution Research in Consumer Behavior: A Review and New Directions. Journal of Consumer Research, 14 (4), 548-65.

Foreman, P., and Whetten, D.A. (2002). Members’ identification with multiple- identity organizations. Organization Science, 13 (6), 618-35.

Forgas, J.P. (2008). Affect and Cognition. Perspectives on Psychological Science, 3 (2), 94-101.

Fortunato, J.A. (2008). Restoring a reputation: the Duke University lacrosse scandal. Public Relations Review, 34, 116-23.

Franklin, B. (2008). The future of newspapers. Journalism Practice, 2 (3), 306-17.

Funk, D.C. (2002). Consumer-Based Marketing: The Use of Micro-Segmentation Strategies for Understanding Sport Consumption. International Journal of Sports Marketing & Sponsorship (September/October), 231-56.

Gagne, R.M. (1950). The effect of sequence of presentation of similar items on the learning of paired associates.

Galtung, J., and Ruge, M. (1965). The structure of foreign news: the presentation of the Congo, Cuba and Cyprus in four Norwegian newspapers. Journal of Peace Research, 2 (1), 64-91.

Garnsey, P. (1968). The criminal jurisdiction of governors. The Journal of Roman Studies, 58 (1 and 2), 51-59.

Gauthier, D. (1979). Thomas Hobbes: moral theorist. The Journal of Philosophy, 76 (10), 547-559.

Gauthier, D. (1986). Morals by agreement. Great Britain: Clarendon Press.

331

Gawronski, B., and Strack, F. (2004). On the propositional nature of cognitive consistency: Dissonance changes explicit, but not implicit attitudes. Journal of Experimental Social Psychology, 40, 535-42.

Gawronski, B., Walther, E., and Blank, H. (2005). Cognitive consistency and the formation of interpersonal attitudes: Cognitive balance affects the encoding of social information. Journal of Experimental Social Psychology, 41, 618-26.

Gellatly, A., Parker, A., Blurton, A., and Woods, C. (1994). Word Stem and Word Fragment Completion Following Semantic Activation and Elaboration. Journal of Experimental Psychology: Learning, Memory & Cognition, 20 (5), 1099-1107.

George, P. (2009). Sport in Disrepute. Australian and New Zealand Sports Law Journal, 4 (1), 24-54.

Giannoulakis, C., and Drayer, J. (2009). “Thugs” versus “Good Guys”: The impact of NBA cares on player image. European Sport Management Quarterly, 9 (4), 453- 468.

Gilbert, E., and Karahalios, K. (2009). Predicting ties strength with social media. Proceedings of the 27th International Conference on Human Factors in Computing Systems: CHI 2009. Boston, MA: April 4-9.

Glasser, T.L. (2009). Journalism and the second-person effect. Journalism, 10, 326- 28.

Gordon, M.E., Slade, L.A., and Schmitt, N. (1986). The “Science of the Sophomore” Revisited: From Conjecture to Empiricism. The Academy of Management Review, 11 (1), 191-207.

Graham, J., Haidt, J., and Nosek, B.A. (2009). Liberals and Conservatives rely on different sets of moral foundations. Journal of Personality and Social Psychology, 96 (5), 1029-1046.

Grange, P. & Kerr, J.H. (2010). Physical aggression in Australian football: a qualitative study of elite athletes. Psychology of Sport and Exercise, 11, pp36-43.

Granovetter, M.S. (1973). The strength of weak ties. American Journal of Sociology, 78 (6), 1360-80.

Gregan-Paxton, J., and John, D.R. (1997). Consumer learning by analogy: a model of internal knowledge transfer. The Journal of Consumer Research, 24 (3), 266-84.

Gluck, M. A., and Myers, C.M. (1995). Representation and Association in Memory: A Neurocomputational view of Hippocampal Functions. Current Directions in Psychological Science, 4, 23-29.

Graham, J., Haidt, J., and Nosek, B.A. (2009). Liberals and Conservatives rely on different sets of moral foundations. Journal of Personality and Social Psychology, 96 (5), 1029-1046.

332

Gray, K., & Wegner, D.M. (2009). Moral typecasting: divergent perceptions of moral agents and moral patients. Journal of Personality and Social Psychology, 96 (3), pp505-520.

Greenwald, A.G., and Banaji, M.R. (1995). Implicit Social Cognition: Attitudes, Self-Esteem, and Stereotypes. Psychological Review, 102 (1), 4-27.

Grieve, P.G., and Hogg, M.A. (1999). Subjective uncertainty and intergroup discrimination in the minimal group situation. Personality and Social Psychology Bulletin, 25 (8), 926-40.

Grieve, F.G., Shoenfelt, E.L., Wann, D.L., and Zapalac, R.K. (2009). The puck stops here: a brief report on National Hockey League fans’ reactions to the 2004-2005 lockout. International Journal of Sport Management and Marketing, 5 (1/2), 101-14.

Grohs, R., Wagner, U., and Vsetecka, S. (2004). Assessing the effectiveness of sport sponsorships – an empirical investigation. Schmalenbach Business Review, 56 (2), 119-38.

Grossbart, S., Muehling, D.D., and Kangun, N. (1986). Verbal and visual references to competition in comparative advertising. Journal of Advertising, 15 (1), 10-23.

Grunert, K.G. (1986). Cognitive Determinants of Attribute Information Usage. Journal of Economic Psychology, 7, 95-124.

Grunert, K.G. (1996). Automatic and Strategic Processes in Advertising Effects. Journal of Marketing, 60 (4), 88-101.

Gurin, P., and Townsend, A. (1986). Properties of gender identity and their implications for gender consciousness. British Journal of Social Psychology, 25, 139- 48.

Gwinner, K. (1997). A model of image creation and image transfer in event sponsorship. International Marketing Review, 14 (3), 145-58.

Gwinner, K., and Bennett, G. (2008).The impact of brand cohesiveness and sport identification on brand fit in a sponsorship context. Journal of Sport Management, 22, 410-26.

Gwinner, K.P., and Eaton, J. (1999). Building brand image through event sponsorship: the role of image transfer. Journal of Advertising, 28 (4), 47-57.

Gwinner, K., and Swanson, S.R. (2003). A model of fan identification: antecedents and sponsorship outcomes. Journal of Services Marketing, 17 (3), 275-94.

Haidt, J. (2008). Morality. Perspectives on Psychological Science, 3 (1), pp65-72.

333

Haidt, J., and Graham, J. (2007).When morality opposes justice: conservatives have moral i ntuitions that liberals may not recognise. Social Justice Research, 20 (1), 98- 116.

Hair, J.F., Anderson, R.E., Tatham,R.L., and Black, W.C. (1992). Multivariate data analysis: 3rd Edition. New York: MacMillan Publishing Company.

Hair, J.F., Anderson, R.E., Tatham. R.L., & Black, W.C. (1995). Multivariate Data Analysis. USA: Prentice-Hall International, Inc.

Hair, J.F., Black, W.C., Babin, B.J., Anderson, R.E., and Tatham, R.L. (2006). Multivariate Data Analysis: Sixth edition. New Jersey: Pearson Prentice Hall.

Hair, J.F., Black, W.C., Babin, B.J., and Anderson, R.E. (2010). Multivariate Data Analysis: A Global Perspective: Seventh Edition. USA: Pearson Prentice Hall.

Hains, S.C., Hogg, M.A., and Duck, J.M. (1997). Self-categorisation and leadership: effects of group prototypicality and leader stereotypicality. Personality and Social Psychology Bulletin, 23 (10), 1087-99.

Harcup, T., and O’Neill, D. (2001). What is News? Galtung and Ruge revisited. Journalism Studies, 2 (2), 261-80.

Harris, L.T., Todorov, A., and Fiske, S.T. (2005). Attributions on the brain: Neuro- imaging dispositional inferences, beyond theory of mind. NeuroImage, 28, 763-69.

Harvey, B. (2001). Measuring the effects of sponsorships. Journal of Advertising Research, (January), 59-65.

Haslam, S.A., Oakes, P.J., Reynolds, K.J., and Turner, J.C. (1999). Social identity salience and the emergence of stereotype consensus. Personality and Social Psychology Bulletin, 25 (7), 809-18.

Haslam, N., Rothschild, L., and Ernst, D. (2000). Essentialist beliefs about social categories. British Journal of Social Psychology, 39, 113-27.

Hastie, R. (1984). Causes and Effects of Causal Attribution. Journal of Personality and Social Psychology, 46 (1), 44-56.

Hastie, R., and Carlston, D. (1980). Theoretical issues in Person Memory. In, R. Hastie, T.M. Ostrom, E.B. Ebbesen, R.S. Wyer Jr., D.L. Hamilton, D.E. Carlston (Eds.) Person Memory: the cognitive basis of social perception. NJ: Lawrence Erlbaum Associates, Publishers, 1-53.

Hastie, R., and Kumar, P. A. (1979). Person Memory: Personality Traits as Organizing Principles in Memory for Behaviors. Journal of Personality and Social Psychology, 37 (1), 25-38.

334

Hastie, R., and Park., B. (1986). The relationship between memory and judgment depends on whether the judgment task is memory-based or on-line. Psychological Review, 93 (3), 258-68.

Heckler, S.E., and Childers, T.L. (1992). The role of expectancy and relevancy in memory for verbal and visual information: what is incongruency? Journal of Consumer Research, 18 (4), 475-92.

Heere, B., and James, J.D. (2007). Stepping outside the lines: developing a multi- dimensional team identity scale based on social identity theory. Sport Management Review, 10, 65-91.

Heider, F. (1958). The psychology of interpersonal relations. New York: John Wiley & Sons, Inc.

Heinegg, P. (2003). Philosopher in the Playground: Notes on the Meaning of Sport, In, Sports Ethics: An Anthology. (Ed) Jan Boxill. Oxford: Blackwell Publishing, pp53-55.

Heise, D. R., and Smith-Lovin, L. (1981). Impressions of goodness, powerfulness, and liveliness from discerned social events. Social Psychology Quarterly, 44 (2), 93- 106.

Henry, K.B., Arrow, H., and Carini, B. (1999). A tripartite model of group identification: theory and measurement. Small Group Research, 30 (5), 558-81.

Herbert, J. (2000). Journalism in the digital age: theory and practice for broadcast, print and on-line media. Oxford: Focal Press.

Hetherington, A. (1985). News, newspapers and television. London: Macmillan.

Hewstone, M., Jaspars, J., and Lalljee, M. (1982). Social representations, social attribution and social identity: the intergroup images of 'public' and 'comprehensive' schoolboys. European Journal of Social Psychology, 12, pp241-269.

Higham, P.A. (2002). Strong cues are not necessarily weak: Thomson and Tulving (1970) and the encoding specificity principle revisited. Memory & Cognition, 30 (1), 67-80.

Hinkle, S., Taylor, L.A., Fox-Cardamone, D.L., and Crook, K.F. (1989). Intragroup identification and intergroup differentiation: A multicomponent approach. British Journal of Social Psychology, 28, 305-17.

Hirt, E.R., Zillmann, D., Erickson, G.A., and Kennedy, C. (1992). Costs and benefits of allegiance: changes in fans' self-ascribed competencies after team victory versus defeat. Journal of Personality and Social Psychology, 63 (5), 724-38.

Hirschman, E.C., and Wallendorf, M. (1982). Motives underlying marketing information acquisition and knowledge transfer. Journal of Advertising, 11 (3), 25- 31.

335

Hoch, S.J., and Deighton, J. (1989). Managing what consumers learn from experience. Journal of Marketing, 53 (2), 1-20.

Hoch, S.J., and Ha, Y-W. (1986). Consumer learning: advertising and the ambiguity of product experience. The Journal of Consumer Research, 13 (2), 221-33.

Hogg, M.A. (1996). Intragroup Processes, Group Structure and Social Identity. In, P. Robinson (Ed.): Social Groups and Identities: Developing the Legacy of Henri Tajfel, (pp65-93). Great Britain: Butterworth-Heineman.

Hogg, M.A. (2000). Subjective uncertainty reduction through self-categorization: A motivational theory of social identity processes. European Review of Social Psychology, 11 (1), 223-255.

Hogg, M.A. (2009). Managing self-uncertainty through group identification. Psychological Inquiry, 20, 221-24.

Hogg, M.A., and Abrams, D. (1999). Social Identity and Social Cognition: Historical Background and Current Trends. In, M.A. Hogg & D. Abrams (Eds), Social Identity and Social Cognition, 1-25.

Hogg, M.A., CooperShaw, L., and Holzworth, D.W. (1993). Group prototypicality and depersonalised attraction in small interactive groups. Personality and Social Psychology Bulletin, 19 (4), 452-65.

Hogg, M.A., & Hains, S.C. (1996). Intergroup relations and group solidarity: effects of group identification and social beliefs on depersonalised attraction. Journal of Personality and Social Psychology, 70 (2), pp295-309.

Hogg, M.A., and Hardie, E.A. (1991). Social attraction, personal attraction, and self- categorization – a field study. Personality and Social Psychology, 17 (2), 175-80.

Hogg, M.A., and Mullin, B-A. (1999). Joining Groups to Reduce Uncertainty: Subjective Uncertainty Reduction and Group Identification. In, M.A. Hogg & D. Abrams (Eds), Social Identity and Social Cognition, 249-79.

Hogg, M.A., & Terry, D.J. (2000). Social identity and self-categorisation processes in organisational contexts. The Academy of Management Review, 25 (1), 121-40.

Hogg, M.A., Terry, D.J., and White, K.M. (1995). A Tale of Two Theories: A Critical Comparison of Identity Theory with Social Identity Theory. Social Psychology Quarterly, 58 (4), 255-69.

Hogg, M.A., and Turner, J.C. (1985a). When liking begets solidarity: An experiment on the role of interpersonal attraction in psychological group formation. British Journal of Social Psychology, 24 (4), 267-81.

336

Hogg, M.A., and Turner, J.C. (1985b). Interpersonal attraction, social identification and psychological group formation. European Journal of Social Psychology, 15, 51- 66.

Hogg, M.A., and Turner, J.C. (1987). Intergroup behaviour, self-stereotyping and the salience of social categories. British Journal of Social Psychology, 26 (4), 325-40.

Hogg, M.A., Turner, J.C., and Davidson, B. (1990). Polarized norms and social frames of references: A test of the self-categorization theory of group polarization. Basic and Applied Social Psychology, 11 (1), 77-100.

Holstrom, B. (1979). Moral hazard and observability. The Bell Journal of Economics, 10 (1), 74-91.

Holstrom, B. (1982). Moral hazard in teams. The Bell Journal of Economics, 13 (2), 324-40.

Hooghe, M., Stolle, D., Maheo, V-A., and Vissers, S. (2010). Why can’t a student be more like an average person?: Sampling and Attrition Effects in Social Science Field Experiments. The Annals of the American Academy of Political and Social Science, 628, 85-96.

Horowitz, T. (1998). Philosophical intuitions and psychological theory. Ethics, 108 (2), 367-85.

Horton, D.L., and Mills, C.B. (1984). Human learning and memory. Annual Review of Psychology, 35, 361-94.

Hu, L-T., & Bentler, P.M. (1998). Fit indices in Covariance Structure Modeling: Sensitivity to Underparameterized Model Misspecification. Psychological Methods, 3 (4), pp424-453.

Hu, L-T., & Bentler, P.M. (1999). Cutoff criteria for fit indexes in covariance structure analysis: Conventional criteria versus new alternatives. Structural Equation Modeling: A Multidisciplinary Journal, 6 (1), pp1-55.

Hughes, S., and Shank, M. (2005). Defining scandal in sports: media and corporate sponsor perspectives. Sport Marketing Quarterly, 14, 5, 207-16.

Hughes, S.F., and Shank, M.D. (2008). Assessing the impact of NCAA scandals: an exploratory analysis. International Journal of Sport Management and Marketing, 3 (1/2), 78-99.

Hummert, M., Crockett, W.H., and Kemper, S. (1990). Processing Mechanisms Underlying Use of the Balance Schema. Journal of Personality and Social Psychology, 58 (1), 5-21.

Hunt, K.A. Bristol, T., and Bashaw, R. E. (1999). A conceptual approach to classifying sports fans. Journal of Services Marketing, 13 (6), 439-52.

337

Hunt, S. D. (1983). General Theories and the Fundamental Explanada of Marketing. Journal of Marketing, 47 (4), 9-17.

Hunt, K.A., Bristol, T., and Bashaw, R.E. (1999). A conceptual approach to classifying sports fans. Journal of Services Marketing, 13 (6), 439-52.

Hutchinson, J.W., and Alba, J.W. (1991). Ignoring irrelevant information: situational determinants of consumer learning. Journal of Consumer Research, 18 (3), 325-45.

Hutchinson, M., and Bouchet, A. (2010). Sponsoring firms assess perceptions of sport property engagement and execution. Journal of Sponsorship, 4 (1), 59-71.

Hutchison, P., Jetten, J., Christian, J., and Haycraft, E. (2006). Protecting threatened identity: Sticking with the group by emphasizing ingroup heterogeneity. Personality and Social Psychology Bulletin, 32 (12), 1620-32.

IEG Inc. (2003). Marketers to forget sponsorship in 2004, December 29. IEG Sponsorship Report.

IEG. (2007). Mixed Bag: Sponsorship spending to surge in ’08, but not everyone will benefit, 24 December. http://www.sponsorship.com/iegsr/2007/12/24/Mixed-Bag-- Sponsorship-Spending-To-Surge-In--08,-B.aspx Last accessed: 11 February 2012

IEG. (2009a). IEG Sponsorship Report, 28 (11), June 8.

IEG. (2009b). IEG Sponsorship Report, 28 (16), 24 August.

IEG. (2012). IEG Sponsorship Report, January 3.

ING (2009, September 24). ING to terminate contract with Renault F1 with immediate effect. http://www.ing.com/Our-Company/Press-room/Press-release- archive/PressRelease/ING-to-terminate-contract-with-Renault-F1-with-immediate- effect.htm Last accessed: 11 February 2012

Irwin, R. L., Lachowetz, T., Cornwell, T.B., and Clark, J.S. (2003). Cause-related sport sponsorship: an assessment of spectator beliefs, attitudes, and behavioural intentions. Sport marketing Quarterly, 12 (3), 131-39.

ISPS. (n.d.) The International Sports Press Survey. www://playthegame.org/knowledgebank/theme-pages/the-international-sports-press- survey-2005.html

Jackson, J.W. (2002). Intergroup Attitudes as a Function of Different Dimensions of Group Identification and Perceived intergroup Conflict. Self and Identity, 1 (1), 11- 33.

Jacoby, J., and Kyner, D.B. (1973). Brand loyalty vs. repeat purchasing behaviour. Journal of Marketing Research, 10 (February), 1-9.

338

Jacoby, J., and Chestnut, R.W. (1978). Brand Loyalty Measurement and Management. New York: John Wiley & Sons.

Jacquette, D. (1991). Moral dilemmas, disjunctive obligations, and Kant’s principles that ‘ought’ implies ‘can’. Synthese, 88, 43-55.

Jamieson, L.F., and Bass, F.M. (1989). Adjusting Stated Intention Measures to Predict Trial Purchase of New Products: A Comparison of Models and Methods. Journal of Marketing Research, 26 (August), 336-45.

Janiszewski, C. (1993). Preattentive Mere Exposure Effects. Journal of Consumer Research, 20 (3), 376-92.

Janoff-Bulman, R., Sheikh, S., & Hepp, S. (2009). Proscriptive versus prescriptive morality: two faces of moral regulation. Journal of Personality and Social Psychology, 96 (3), pp521-537.

Javalgi, R.G., Traylor, M.B., Gross, A.C., and Lampman, E. (1994). Awareness of sponsorship and corporate image: an empirical investigation. Journal of Advertising, 23 (4), 47-58.

Jenkins, R., 2004. Social Identity: Second Edition. : Great Britain: Routledge: Taylor & Francis Group.

Jin, H-S. (2004). Compounding Consumer Interest: Effects of Advertising Campaign Publicity on the Ability to Recall Subsequent Advertisements. Journal of Advertising, 32 (4), 29-41.

Johansson, S. (2008). Gossip, sport and pretty girls: What does “trivial” journalism mean to tabloid newspaper readers? Journalism Practice, 2 (3), 402-13.

Johar, G.V., and Pham, M.T. (1999). Relatedness, Prominence, and Constructive Sponsor Identification. Journal of Marketing Research, 36 (3), 299-312.

Johar, G.V., Maheshwaran, D., and Peracchio, L.A. (2006). MAPping the frontiers: theoretical advances in consumer research on Memory, Affect, and Persuasion. Journal of Consumer Research, 33 (June), 139-49.

Johnston, M.A. (2010). The impact of sponsorship on shareholder wealth in Australia. Asia Pacific Journal of Marketing and Logistics, 22 (2), 156-78.

Johnson Morgan, M., Summers, J., and Sassenberg, A. (2008). The impact of negative publicity on: an individual sporting celebrity’s brand DNA; the brand DNA of their associated team and/or sport; and attitudes of their sponsors and partners. Presented at: Sport Marketing Association 6th Annual Conference: Bridging the Gap: Bringing the World Down Under, 17-19 July 2008, Gold Coast, Australia.

339

Jones, M.J., and Schumann, D.W. (2000). The Strategic Use of Celebrity Athlete Endorser in Sports Illustrated: An Historic Perspective. Sport Marketing Quarterly, 9 (2), 65-76.

Jones, J.C.H., Ferguson, D.G., & Stewart, K.G. (1993). Blood sports and cherry pie: some economics of violence in the National Hockey League. American Journal of Economics and Sociology, 52 (1), pp63-78.

Kahneman, D., and Tversky, A. (1979). Prospect theory: An analysis of decision under risk. Eocnometrica, 47, 313-27.

Kahuni, A.T., Rowley, J., and Binsardi, A. (2009). Guilty by association: image 'spill-over' in corporate co-branding. Corporate Reputation Review, 12, 1, 52-63.

Karasawa, M. (1991). Toward an assessment of social identity: the structure of group identification and its effects on in-group evaluations. British Journal of Social Psychology, 30, 293-07.

Kanouse, D.E., and Hanson, L.R. (1972). Negativity in evaluations. In, E.E. Jones et al (Eds.). Attribution: Perceiving the causes of behaviours. NJ: General Learning Press.

Kardes, F.R., Kalyanaram, G., Chandrashekaran, M., and Dornoff, R.J. (1993). Brand retrieval, consideration set composition, consumer choice, and the pioneering advantage. Journal of Consumer Research, 20 (1), 62-75.

Kaynak, E., Salman, G.G., and Tatoglu, E. (2008). An integrative framework linking brand associations and brand loyalty in professional sports. Journal of Brand Management, 15 (5), 336-57.

Keller, K.L. (1987). Memory factors in advertising: the effect of advertising retrieval cues on brand evaluations. Journal of Consumer Research, 14 (3), 316-33.

Keller, K.L. (1991). Memory and Evaluation Effects in Competitive Advertising Environments. Journal of Consumer Research, 17 (4), 463-76.

Keller, K.L. (1993). Conceptualising, Measuring, and Managing Customer-Based Brand Equity. Journal of Marketing, 57 (1), 1-22.

Kelley, H.H. (1973). The Processes of Causal Attribution. American Psychologist, (February), 107-28.

Kelley, H.H., and Michela, J.L. (1980). Attribution theory and research. Annual Review of Psychology, 31, 457-501.

Kelly, C. (1988). Intergroup differentiation in a political context. British Journal of Social Psychology, 27, 319-32.

Kelly, L., and Whiteman, C. (2010). Sports sponsorship as an IMC tool: An Australian sponsor’s perspective. Journal of Sponsorship, 4 (1), 26-37.

340

Kim, J., Lim, J-S., and Bhargava, M. (1998). The role of affect in attitude formation: A Classical Conditioning approach. Journal of the Academy of Marketing Science, 26 (2), 143-52.

Kim, J.Y., & Parlow, M.J. (2009). Off-court misbehaviour: sports leagues and private punishment. The Journal of Criminal Law & Criminology, 99 (3), pp573-597.

King, C. (2007). Media portrayals of male and female athletes: A text and picture analysis of British national newspaper coverage of the Olympic Games since 1948. International Review for the Sociology of Sport, 42 (2), 187-99.

Kinney, L., and McDaniel, S.R. (1996). Strategic Implications of Attitude-Toward- the-Ad in Leveraging Event Sponsorships. Journal of Sport Management, 10, 250- 61.

Kinnick, K.N. (1998). Gender bias in newspaper profiles of 1996 Olympic athletes: A content analysis of five major dailies. Women’s Studies in Communication, 21 (2), 212-37.

Kiousis, S., and Wu, X. (2008). International agenda-building and agenda-setting: exploring the influence of public relations counsel on US news media and public perceptions of foreign nations. The International Communication Gazette, 70 (1), 58- 75.

Klein, J.G. (1996). Negativity in impressions of presidential candidates revisited: the 1992 election. Personality and Social Psychology Bulletin, 22 (3), 288-95.

Klein, S.B., Loftus, J., and Burton, H.A. (1989). Two self-reference effects: the importance of distinguishing between self-descriptiveness judgments and autobiographical retrieval in self-referent encoding. Journal of Personality and Social Psychology, 56 (6), 853-65.

Knittel, C.R., and Stango, V. (2010). Celebrity Endorsements, Firm Value and Reputation Risk: Evidence from the Tiger Woods Scandal. Unpublished manuscript: http://faculty.gsm.ucdavis.edu/~vstango/. Last accessed: 17 October 2011.

Knoppers, A., and Elling, A. (2004). ‘We do not engage in promotional journalism’: Discursive strategies used by sport journalists to describe the selection process. International Review for the Sociology of Sport, 39 (1), 57-73.

Kohlman, C.W. (1960). IPSO in Industrial Marketing. Journal of Marketing, 24 (3), 55-57.

Korsgaard, C.M. (1996). Creating the kingdom of ends. USA: Cambridge University Press.

Kossinets, G., and Watts, D.J. (2006). Empirical analysis of an evolving social network. Science, 311 (January), 88-90.

341

Kramer, T., and Block, L. (2011). Nonconscious effects of peculiar beliefs on consumer psychology and choice. Journal of Consumer Psychology, 21, 101-11.

Kressler, N.B. (2005-06). Using the morals clause in talent agreements: A historical, legal and practical guide. Columbia Journal of Law & Arts, 29, 235-60.

Kuenzel, S., and Halliday, S.V. (2008). Investigating antecedents and consequences of brand identification. Journal of Product and Brand Management, 17 (5), 293-04.

Kulviwat, S., Bruner II, G.C., and Al-Shuridah, O. (2009). The role of social influence on adoption of high tech innovations: The moderating effect of public/private consumption. Journal of Business Research, 62, 706-12.

Lardinoit, T., and Derbaix, C. (2001). Sponsorship and recall of sponsors. Psychology & Marketing, 18 (2), 167-90.

Lasch, C. (1988). The degradation of sport, In, Philosophic Inquiry in Sport. (Ed.s) Klaus Meier & William Morgan. Illinois: Human Kinetics Publishers, pp403-418.

Lazarus, R.S. (1982). Thoughts on the Relations Between Emotion and Cognition. American Psychologist, 37 (9), 1019-1024.

Lazarus, R.S. (1984). On the Primacy of Cognition. American Psychologist, 39 (2), 124-29.

Leach, C.W., Ellemers, N., and Barreto, M. (2007). Group virtue: the importance of morality (vs. competence and sociability) in the positive evaluation of in-groups. Journal of Personality and Social Psychology, 93 (2), 234-49.

Lee, A.L. (2001). The Mere Exposure Effect: An Uncertainty Reduction Explanation Revisited. Personality and Social Psychology Bulletin, 27 (10), 1255-1266.

Lee, J. (1992). Media portrayals of male and female Olympic athletes: Analyses of newspaper accounts of the 1984 and the 1988 Summer Games. International Review for the Sociology of Sport, 27, 197-219.

Lee, H-S., and Choi, C-H. (2009). The matching effect of brand and sporting event personality: sponsorship implications. Journal of Sport Management, 23, 41-64.

Lee, M-S., Sandler, D.M., and Shani, D. (1997). Attitudinal constructs towards sponsorship: Scale development using three global sporting events. International Marketing Review, 14 (3), 159-69.

Levine, T.R., and Hullett, C.R. (2002). Eta squared, partial eta squared, and misreporting of effect size in communication research. Human Communication Research, 28 (4), 612-25.

342

Langer,T., Walther, E., Gawronski, B., and Blank, H. (2009). When linking is stronger than thinking: Associative transfer of valence disrupts the emergence of cognitive balance after attitude change. Journal of Experimental Social Psychology, 45, 1232-37.

Lavidge, R.J., and Steiner, G.A. (1961). A Model for Predictive Measurements of Advertising Effectiveness. Journal of Marketing, 25 (6), 59-62.

Leonard, D. (2007). Innocent until proven innocent: In defense of Duke Lacrosse and White Power (and against menacing black student-athletes, a black stripper, activists, and the Jewish media). Journal of Sport and Social Issues: 31 (1), 25-44.

Lockwood, P., Jordan, C.H., and Kunda, Z. (2002). Motivation by positive or negative role models: regulatory focus determines who will best inspire us. Journal of Personality and Social Psychology, 83 (4), 854-64.

Luhtanen, R., and Crocker, J. (1992). A collective self-esteem scale: self-evaluation of one’s social identity. Personality and Social Psychology Bulleting, 18 (3), 302- 318.

Lynch, J.G., and Srull, T.K. (1982). Memory and attentional factors in consumer choice: concepts and research methods. Journal of Consumer Research, 9 (1), 18-37.

Lyytimaki, J., and Tapio, P. (2009). Climate change as reported in the press of Finland: From screaming headlines to penetrating background noise. International Journal of Environmental Studies, 66 (6),723-35.

Madrigal, R. (1995). Cognitive and affective determinants of fan satisfactionwith sporting event attendance. Journal of Leisure Research, 27 (3), 205-27.

Madrigal, R. (2000). The Influence of Social Alliances with Sports Teams on Intentions to Purchase Corporate Sponsors’ Products. Journal of Advertising, 29 (4), 13-24.

Madrigal, R. (2001). Social identity effects in a belief-attitude-intentions hierarchy: implications for corporate sponsorship. Psychology & Marketing, 18 (2), 145-65.

MacKenzie, S. B., and Lutz, R.J. (1989). An empirical investigation of the structural antecedents of Attitude toward the Ad in an advertising pretesting context. Journal of Marketing, 53 (2), 48-65.

Mae, L., McMorris, L.E., and Hendry, J.L. (2004). Spontaneous trait transference from dogs to owners. Anthrozoos, 17 (3), 225-43.

Mael, F., & Ashforth, B.E (1992). Alumni and their alma mater: A partial test of the reformulated model of organizational identification. Journal of Organizational Behavior, 13, pp103-123.

Mael, F.A., and Ashforth, B.E. (1995). Loyal from day one: biodata, organizational identification, and turnover among newcomers. Personnel Psychology, 48, 309-33.

343

Mael, F.A., and Tetrick, L.E. (1992). Identifying organizational identification. Educational and Psychological Measurement, 52, 813-24.

Maestas, A.J. (2009). Guide to sponsorship return on investment. Journal of Sponsorship, 3 (1), 98-102.

Mahony, D.F., and Madrigal, R. (1999). The Effect of Individual Levels of Self- Monitoring on Loyalty to Professional Football Teams. International Journal of Sports Marketing & Sponsorship, 1 (2), 146-67.

Mahony, D.F., Madrigal, R., and Howard, D. (2000). Using the Psychological Commitment to Team (PCT) Scale to Segment Sport Consumers Based on Loyalty. Sport Marketing Quarterly, 9 (1), 15-25.

Maiden, M. (2005, April 11). Stakes high in sports sponsorship. Sydney Morning Herald. http://www.smh.com.au/news/Malcolm-Maiden/Stakes-high-in-sports- sponsorship/2005/04/10/1113071853973.html

Mandler, G. (1954). Response Factors in Human Learning. Psychological Review, 61 (4), 235-44.

Mandler, J.M., and Johnson, N.S. (1977). Remembrance of Things Parsed: Story Structure and Recall. Cognitive Psychology, 9, 111-51.

Martijn, C., Spears, R., Van der Pligt, J., and Jacobs, E. (1992). Negativity and positivity effects in person perception and inference: ability versus morality. European Journal of Social Psychology, 22, 453-63.

Marques, J.M., and Yzerbyt, V.Y. (1988). The black sheep effect: Judgmental extremity towards ingroup members in inter- and intra-group situations. European Journal of Social Psychology, 18, 287-92.

Martin, C.A., and Bush, A.J. (2000). Do role models influence teenagers’ purchase intentions and behaviour? Journal of Consumer Marketing, 17 (5), 441-54.

McCombs, M., Llamas, J.., Lopez-Escobar, E., and Rey, F. (1997). Candidate images in Spanish elections: second-level agenda-setting effects. Journalism and Mass Communication Quarterly, 74 (4), 703-17.

McCombs, M.E., and Shaw, D.L. (1972). The agenda-setting function of mass media. Public Opinion Quarterly, 36, 176-87.

McCracken, G. (1989). Who is the Celebrity Endorser? Cultural Foundations of the Endorsement Process. Journal of Consumer Research, 16 (3), 310-21.

McDaniel, S.R. (1999). An investigation of match-up effects in sport sponsorship advertising: the implications of consumer advertising schemas. Psychology & Marketing, 16 (2), 163-84.

344

McDaniel, M.A., Robinson-Riegler, B., and Einstein, G.O. (1998). Prospective remembering: Perceptually driven or conceptually driven processes? Memory & Cognition, 26 (1), 121-34.

McDonald, C. (1991). Sponsorship and the image of the sponsor. European Journal of Marketing, 25 (11), 31-38.

McGuire, W.J. (1960). Cognitive Consistency and Attitude Change. Journal of Abnormal and Social Psychology, 60 (3), 345-53.

McGuire, W.J. (1961). A Multiprocess Model for Paired-Associate Learning. Journal of Experimental Psychology, 62 (4), 335-47.

Meenaghan, J.A. (1983). Commercial sponsorship. European Journal of Marketing, 7 (7), 5-73.

Meenaghan, T. (1998). Current developments and future directions in commercial sponsorship. International Journal of Advertising, 17, 3-28.

Meenaghan, T. (2001). Sponsorship and Advertising: A Comparison of Consumer Perceptions. Psychology & Marketing, 18 (2), 191-215.

Meenaghan, T. (2001). Understanding sponsorship effects. Psychology & Marketing, 18 (2), 95-122.

Meenaghan, T. (1991). The role of sponsorship in the marketing communications mix. International Journal of Advertising, 10 (1), pp35-47.

Meijer, I.C. (2003). What is Quality Television News? A plea for extending the professional repertoire of newsmakers. Journalism Studies, 4 (1), 15-29.

Meirick, P. (2002). Cognitive Responses to Negative and Comparative Political Advertising. Journal of Advertising, 31 (1), 49-62.

Miller, C.B. (2009). Yes we did! Basking in reflected glory and cutting off reflected failure in the 2008 presidential election. Analyses of Social Issues and Public Policy, 9 (1), 283-96.

Miller, G.A., and Chapman, J.P. (2001). Misunderstanding analysis of covariance. Journal of Abnormal Psychology, 110 (1), 40-48.

Mirrlees, J.A. (1999). The theory of moral hazard and unobservable behaviour: Part I. The Review of Economic Studies, 66 (1), Special Issue: Contracts, 3-21.

Mitchell, A.A., and Olson, J.C. (1981). Are product attribute beliefs the only mediator of advertising effects on brand attitude? Journal of Marketing Research, 18 (3), 318-32.

345

Miyazaki, A.D., and Morgan, A.G. (2001). Assessing market value of event sponsoring: Corporate Olympic Sponsorships. Journal of Advertising Research, (January/February), 9-15.

Mizerski, R.W., Golden, L.L., and Kernan, J.B. (1979). The attribution process in consumer decision making. Journal of Consumer Research, 6 (2), 123-40.

Moore, B. (2001). Cruel and unusual punishment in the Roman Empire and dynastic China. International Journal of Politics, Culture and Society, 14 (4), 729-72.

Moschis, G.P. (1981). Patterns of Consumer Learning. Journal of the Academy of Marketing Science, 9 (2), 110-26.

Moscovitch, M. (2007). Memory: why the engram is elusive. In, H.L. Roediger III, Y. Dudai & S.M. Fitzpatrick, Science of Memory: concepts. Great Britain: Oxford University Press, 17-21.

Moutinho, L., Dionisio, P., and Leal, C. (2007). Surfing tribal behaviour: a sports marketing application. Marketing Intelligence & Planning, 25 (7), 668-90.

Muehling, D.D., and Laczniak, R.N. (1988). Advertising’s Immediate and Delayed Influence on Brand Attitudes: Considerations across Message-Involvement Levels. Journal of Advertising, 17 (4), 23-34.

Murray, K. B. (1991). A test of services marketing theory: consumer information acquisition activities. Journal of Marketing, 55 (January), 10-25.

Murray, K. B., and Schlacter, J.L. (1990). The impact of services goods on consumers' assessment of perceived risk and variability. Journal of the Academy of Marketing Science, 18 (1), 51-65.

Nadler, A., Harpaz-Gorodeisky, G., and Ben-David, Y. (2009). Defensive helping: Threat to group identity, ingroup identification, status stability, and common group identity as determinants of intergroup helping. Journal of Personality and Social Psychology, 97 (5), 823-34.

Neale, L., and Funk, D. (2006). Investigating motivation, attitudinal loyalty and attendance behaviour with fans of Australian Football. International Journal of Sports Marketing & Sponsorship, (July), 307-17.

Neale, L., Mizerski, R., and Lee, A. (2008). Measuring consumer rituals: a marketing application. In, Proceedings, Winter American Marketing Association Educators Conference: Marketing the Organization and its Products and Services, 15-19 February, Austin, Texas.

Neuberg, S.L. (1989). The goal of forming accurate impressions during social interactions. Attenuating the impact of negative expectancies. Journal of Personality and Social Psychology, 56 (3), 374-86.

Neuendorf, K.A (2002). The Content Analysis guidebook. USA: Sage Publications.

346

Newman, G.E., Diesendruck, G., and Bloom, P. (2011). Celebrity Contagion and the Value of Objects. Journal of Consumer Research, 38 (2), 215-228.

Nguyen, N., and Leblanc, G. (2001). Corporate image and corporate reputation in customers’ retention decisions in services. Journal of Retailing and Consumer Services, 8, 227-236.

Niblock, S., and Machin, D. (2007). News values for consumer groups: The case of Independent Radio News, London, UK. Journalism, 8 (2), 184-204.

Nunnally, J.C., and and Bernstein, I.H. (1994). Psychometric theory. New York: McGraw- Hill.

NY Times (2009, December 14). What they’re saying about Tiger Woods.

Obermiller, C. (1985). Varieties of Mere Exposure: The Effects of Processing Style and Repetition on Affective Response. Journal of Consumer Research, 12 (1), 17-30.

O’Brien, T. (1971). Stages of Consumer Decision Making. Journal of Marketing Research, 8 (August), 283-89.

OECD. (2009a) Dataset: Balance of Payments (MEI): Germany. Source: OECD.stat. Accessed: 21 August, 2009

OECD. (2009b) Dataset: Balance of Payments (MEI): Switzerland. Source: OECD.stat. Accessed: 21 August, 2009

OECD. (2009c) Dataset: Balance of Payments (MEI): Indonesia. Source: OECD.stat. Accessed: 21 August, 2009

Olson, M.A., and Fazio, R.H. (2001). Implicit attitude formation through classical conditioning. Psychological Science, 12 (5), 413-17.

Olson, E.L., and Thjᴓmᴓe, H.M. (2003). The effects of peripheral exposure to information on brand preference. European Journal of Marketing, 37 (1/2), 243-55.

Osgood, C.E. (1960). Cognitive Dynamics in the Conduct of Human Affairs. Public Opinion Quarterly, 24 (2), 341-65.

Ostrom, T.M., Lingle, J.H., Pryor, J.B and Geva, N. (1980). Cognitive organisation of person impressions, In, R. Hastie, T.M. Ostrom, E.B. Ebbesen, R.S. Wyer Jr., D.L. Hamilton, D.E. Carlston (Eds.), Person Memory: the cognitive basis of social perception. NJ: Lawrence Erlbaum Associates, Publishers, 55-88.

Paivio, A., Walsh, M., and Bons, T. (1994). Concreteness Effects on Memory: When and Why? Journal of Experimental Psychological: Learning, Memory, and Cognition, 20 (5), 1196-1204.

347

Palda, K.S. (1966). The hypothesis of a hierarchy of effects: A partial evaluation. Journal of Marketing Research, 3 (1), 13-24.

Papadimitriou, D., Apostolopoulou, A., and Dounis, T. (2008). Event sponsorship as a value creating strategy for brands. Journal of Product & Brand Management, 17 (4), 212-22.

Park, B. (1986). A method for studying the development of impressions of real people. Journal of Personality and Social Psychology, 51 (5), 907-17.

Park, J-W., and Hastak, M. (1994). Memory-based product judgments: effects of involvement at encoding and retrieval. Journal of Consumer Research, 21 (3), 534- 47.

Parker, H.M. (2007). The effect of negative sponsor information and team response on identification levels and consumer attitudes. Unpublished Doctoral thesis.

Paterson, B. (2007). A discourse analysis of the construction of mental illness in two UK newspapers from 1985-2000. Issues in Mental Health Nursing, 28, 1087-1103.

Paunonen, S.V. (1989). Consensus in personality judgments: moderating effects of target-rater acquaintanceship and behaviour observability. Journal of Personality and Social Psychology, 56 (5), 823-33.

Peabody, N. (2009).Disciplining the body, disciplining the body-politic: Physical culture and social violence among North Indian wrestlers. Comparative Studies in Society and History, 51 (2), pp372-400.

Pearsall, J. (2010). Sponsorship performance: What is the role of sponsorship metrics in proactively managing the sponsor-property relationship? Journal of Sponsorship, 3 (2), 115-23.

Pedersen, P.M. (2002). Examining equity in newspaper photographs: A content analysis of the print media photographic coverage of Interscholastic Athletics. International Review for the Sociology of Sport, 37 (3-4), 303-18.

Peeters, G. (1971). The positive-negative asymmetry: on cognitive consistency and positivity bias. European Journal of Social Psychology, I (4), 455-74.

Perry, C. (1998). A structured approach for presenting theses. Australasian Marketing Journal, 6 (1), pp63-85.

Petersen, A., Anderson, A., Allan, S., and Wilkinson, C. (2009). Opening the black box: scientists’ views on the role of the news media in the nanotechnology debate. Public Understanding of Science, 18 (5), 512-30.

Pham, M.T., and Johar, G.V. (2001). Market Prominence Biases in Sponsor Identification: Processes and Consequentiality. Psychology & Marketing, 18 (2), 123-43.

348

Pham, M.T., and Vanhuele, M. (1997). Analysing the memory impact of advertising. Marketing Letters, 8 (4), 407-17.

Phinney, J.S. (1992). The multigroup ethnic identity measure: a new scale for use with diverse groups. Journal of Adolescent Research, 7 (2), 156-76.

Picek, J.S., Sherman, S.J., and Shiffrin, R.M. (1975). Cognitive organization and coding of social structures. Journal of Personality and Social Psychology, 31 (4), 758-68.

Pinter, B., and Greenwald, A.G. (2011). A comparison of minimal group induction procedures. Group Processes & Intergroup Relations, 14 (1), 81-98.

Pitts, B.G. and Slatterly, J. (2004). An Examination of the Effects of Time on Sponsorship Awareness Levels. Sport Marketing Quarterly, 13 (1), 43-54.

Pizarro, J., Guerrero, E., and Galindo, P.L. (2002). Multiple comparison procedures applied to model selection. Neurocomputing, 48, 155-73.

Pope, N. (1998). Consumption values, sponsorship awareness, brand and product use. Journal of Product & Brand Management, 7 (2), 124-36.

Pope, N.K.Ll. & Voges, K.E. (1999). Sponsorship and image: a replication and extension. Journal of Marketing Communications, 5 (1),

Pope, N.K., and Voges, K.E. (2000). The impact of sport sponsorship activities, corporate image, and prior use on consumer purchase intention. Sport Marketing Quarterly, 9 (2), 96-102.

Pope, N.K.Ll., Voges, K.E., and Brown, M.R. (2004). The effect of provocation in the form of mild erotica on attitude to the ad and corporate image: Differences between cause-related and product-based advertising. Journal of Advertising, 33 (1), 69-82.

Pope, N., Voges, K.E., and Brown, M. (2009) Winning ways: Immediate and long- term effects of sponsorship on perceptions of brand quality and corporate image. Journal of Advertising, 38 (2), 5-20.

Pratkanis, A.R., and Greenwald, A.G. (1988). Recent Perspective on Unconscious Processing: Still No Marketing Applications. Psychology & Marketing, 5 (4), 337- 53.

Preuss, H., Gemeinder, K., and Seguin, B. (2008). Ambush Marketing in China: Counterbalancing Olympic Sponsorship Efforts. Asian Business & Management, 7, 243-63.

Pritchard, M.P., Havitz, M.E., and Howard, D.R. (1999). Analyzing the Commitment-Loyalty Link in Service Contexts. Journal of the Academy of Marketing Science, 27 (3), 333-48.

349

Quester, P., and Farrelly, F. (1998). Brand association and memory decay effects of sponsorship: the case of the Australian Formula One Grand Prix. Journal of Product & Brand Management, 7 (6), 539-56.

Quester, P.G., and Thompson, B. (2001). Advertising and promotion leverage on arts sponsorship effectiveness. Journal of Advertising Research, (January) 33-47.

Quillian, M.R. (1967). Word Concepts: A Theory and Simulation of Some Basic Semantic Capabilities. Behavioral Science, 12, 410-30.

Raney, A.A. & Depalma, A.J. (2006). The effect of viewing varying levels and contexts of violent sports programming on enjoyment, mood, and perceived violence. Mass Communication and Society, 9 (3), pp321-338.

Reed, M.H., Bhargava, M.N., Gordon, J., and Kjaer, M. (2010). Legal and regulatory update: Terminating a sponsorship relationship: conditions and clauses. Journal of Sponsorship, 4 (1), 79-92.

Reisinger, H., Grohs, R., and Eder, M. (2006). Adverse Effects of Sponsorship. Presented: 36th EMAC Conference, Reykjavik, Iceland.

Rifon, N,J., Choi, S.M., Trimble, C.S., and Li, H. (2004). Congruence effects in sponsorship: the mediating role of sponsor credibility and consumer attributions of sponsor motive. Journal of Advertising, 33 (1), 29-41.

Rissel, C., Bonfiglioli, C., Emilsen, A., and Smith, B.J. (2010). Representations of cycling in metropolitan newspapers – changes over time and differences between Sydney and Melbourne, Australia. BMC Public Health, 10, 371.

Ritchie, B.W. (2004). Chaos, crises and disasters: a strategic approach to crisis management in the tourism industry. Tourism Management, 25, 669-83.

Roberts, G. (2008, January 18). Vic: Sponsors wield their power. Australian Associated Press General News.

Roy, D.P., and Cornwell, T.B. (2003). Brand equity’s influence on responses to event sponsorship. Journal of Product & Brand Management, 12 (6), 377-93.

Ruth, J.A., and Simonin, B.L. (2003). “Brought to you by Brand A and Brand B”: Investigating multiple sponsors’ influence on consumers’ attitudes toward sponsored events. Journal of Advertising, 32 (3), 19-30.

Rodgers, S. (2003/2004). The effects of sponsorship relevance on consumer reactions to internet sponsorships. Journal of Advertising, 32 (4), 67-76.

Ross, K. (2007). The journalist, the housewife, the citizen and the press: women and men as sources in local news narratives. Journalism, 8 (4), 449-473.

350

Roy, D.P., and Cornwell, T.B. (2003). Brand equity’s influence on responses to event sponsorship. Journal of Product & Brand Management, 12 (6), 377-93.

Rowe, D. (2007). Sports Journalism: Still the ‘toy department’ of the news media? Journalism, 8 (4), 385-405.

Sandler, D.M., and Shani, D. (1993). Sponsorship and the Olympic Games: the consumer perspective. Sport Marketing Quarterly, II (3), 38-43.

Sassenberg, A., and Johnson Morgan, M. (2010). Scandals, sports and sponsors: what impact do sport celebrity transgressions have on consumer’s perceptions of the celebrity’s brand image and the brand image of their sponsors? Presented: 8th Annual Sport Marketing Association Conference, 26-29 October 2010, New Orleans, USA.

Schacter, D.L. (1987). Implicit Memory: History & Current Status. Journal of Experimental Psychology: Learning, Memory & Cognition, 13 (3), 501-18.

Schacter, D.L. (2007). Memory: delineating the core. In, H.L. Roediger III, Y. Dudai & S.M. Fitzpatrick, Science of Memory: concepts. Great Britain: Oxford University Press, 23-27.

Schultz-Jorgensen, S. (2005) ‘The World’s Best Advertising Agency: The Sports Press’, International Sports Press Survey 2005. Copenhagen: House of Monday Morning: Play the Game. Source: http://www.playthegame.org/knowledgebank/theme-pages/the-international-sports-press-survey-2005.html Last accessed: 12 February 2012

ScreenAustralia (n.d). AudioVisual Markets: Television. http://www.screenaustralia.gov.au/research/statistics/wftvtopprog.asp. Last Accessed: 30 July 2011.

ScreenAustralia (n.d). Archive: Television. www.screenaustralia.gov.au/research/statistics/wftvtopprog.asp. Last Accessed: 30 July 2011

Scheufele, D.A. (1999). Framing as a theory of media effects. Journal of Communication, 49 (1), 103-22.

Sedikides, C., and Skowronski, J.J. (1993). The self in impression formation: trait centrality and social perception. Journal of Experimental Social Psychology, 29, 347- 57.

351

Sellers, R.M., Rowley, S.A.J., Chavous, T,M., Shelton, J.N., and Smith, M.A. (1997). Multidimensional Inventory of Black Identity: A Preliminary Investigation of Reliability and Construct Validity. Journal of Personality and Social Psychology, 73 (4), 805-15.

Shannon, C.E. (1948). A Mathematical Theory of Communication. The Bell System Technical Journal, 27, 379-423.

Shannon, J.R. (1999). Sports marketing: an examination of academic marketing publication. Journal of Services Marketing, 13 (6), 517-34.

Shen, W., Kiger, T.B., Davies, S.E., Rasch, R.L., Simon, K.M., and Ones, D.S. (2011). Samples in Applied Psychology: Over a Decade of Research in Review. Journal of Applied Psychology, 96 (5), 1055-64.

Sherman, D.K., Kinias, Z., Major, B., Kim, H.S., and Prenovost, M. (2007). The group as a resource: reducing biased attributions for group success and failure via group affirmation. Personality and Social Psychology Bulletin, 33, 1100-1112.

Sherman, J.W., and Klein, S.B. (1994). Development and representation of personality impressions. Journal of Personality and Social Psychology, 67 (6), 972- 83.

Shilts, C., Jett, K., & Desiato, N. (2007-08). Making the Pitch: Player Endorsements in Professional Sports. Entertainment and Sports Lawyer, 25 (3), 2-7.

Simes, R.J. (1986). An improved Bonferroni procedure for multiple tests of significance. Biometrika, 73 (3), 751-754.

Singleton, R.A., & Straits, B.C. (2005). Approaches to social research. New York: Oxford University Press.

Sink, C.A., and Stroh, H.R. (2006). Practical Significance: The use of effect sizes in school counselling research. ASCA: Professional School Counseling, 401-411.

Sirgy, M.J., Lee, D-J., Johar, J.S., and Tidwell, J. (2008) Effect of self-congruity with sponsorship on brand loyalty. Journal of Business Research, 61, 1091-1097.

Skowronski, J.J., and Carlston, D.E. (1987). Social judgment and social memory: the role of cue diagnosticity in negativity, positivity, and extremity biases. Journal of Personality and Social Psychology, 52 (4), pp689-699.

Skowronski, J.J., & Carlston, D.E. (1989). Negativity and extremity biases in impression formation: a review of explanations. Psychological Bulletin, 105 (1), pp131-142.

Skowronski, J.J., & Carlston, D.E. (1992). Caught in the act: when impressions based on highly diagnostic behaviours are resistant to contradiction. European Journal of Social Psychology, 22, pp435-452.

352

Slater, J. and Lloyd, C. (2004). Chapter 9: It’s gotta be the shoes: Exploring the effects of relationship of Nike and Reebok sponsorship on two college athletic programs. In, (Eds) LR Kahle & C Riley, Sports Marketing and the Psychology of Marketing Communication, 191-210.

Smidts, A., Pruyn, A.Th.H., and Van Riel, C.B.M. (2001). The Impact of Employee Communication and Perceived External Prestige on Organizational Identification. The Academy of Management Journal, 44 (5), 1051-62.

Smith, A., Graetz, B., and Westerbeek, H. (2008). Sport sponsorship, team support and purchase intentions. Journal of Marketing Communications, 14 (5), 387-404.

Smith, R.E., and Swinyard, William. R. (1982). Information Response Models: An Integrated Approach. The Journal of Marketing, 46 (1), 81-93.

Smith, E.E., Shoben, E. J., & Rips, L.J. (1974). Structure and Process in Semantic Memory: A Feature Model for Semantic Decisions. Psychological Review, 81 (3), 214-41.

Snyder, C.R., Lassegard, M., and Ford, C.E. (1986). Distancing after group success and failure: basking in reflected glory and cutting off reflected failure. Journal of Personality and Social Psychology, 51, 382-88.

Snyder, P., and Lawson, S. (1993). Evaluating results using corrected and uncorrected effect size estimates. The Journal of Experimental Education, 61 (4), 334-49.

Soderman, S., and Dolles, H. (2010). Sponsoring the Beijing Olympic Games: Patterns of Sponsor Advertising. Asia Pacific Journal of Marketing & Logisitics, 22 (1), 8-24.

Spais, G.S., and Filis, G.N. (2008). Measuring stock market reaction to sponsorship announcements: The case of Fiat and Juventus. Journal of Targeting, Measurement and Analysis for Marketing, 16 (3), 169-80.

Spears, R., Doosje, B., and Ellemers, N. (1997). Self-stereotyping in the face of threats to group status and distinctiveness: the role of group identification. Personality and Social Psychology Bulletin, 23 (5), 538-53.

Speed, R., and Thompson, P. (2000). Determinants of sports sponsorship response. Journal of the Academy of Marketing Science, 28 (2), 226-38.

Spicer, J. (2005). Making sense of multivariate data analysis. Sage publications: U.S.A.

Sprent, P., & Smeeton, N.C. (2000). Applied nonparametric statistical methods: Third edition. Chapman & Hall/ CRC: USA.

353

Srull, T.K. (1981). Person Memory: Some Tests of Associative Storage and Retrieval Models. Journal of Experimental Psychology: Human Learning & Memory, 7 (6), 440-63.

Standen, J. (2009). The manly sports: the problematic use of criminal law to regulate sports violence. The Journal of Criminal Law & Criminology, 99 (3), pp619-642.

Staats, C.K., and Staats, A.W. (1957). Meaning Established by Classical Conditioning. Journal of Experimental Psychology, 54 (1), 74-80.

StatSoft (1994). Statistica: Volume I: General conventions & statistics I. StatSoft Inc.: USA.

StatSoft (1994). Statistica: Volume III: Statistics II. StatSoft Inc.: USA.

Steel, E. (2010, December 24). Gillette won’t renew sponsorship of Tiger Woods. The Wall Street Journal Online.

Stevens, S.S. (1946). On the Theory of Scales of Measurement. Science, 103 (2684), pp677-80.

Stipp, H. and Schiavone, N.P. (1996). Modeling the Immpact of Olympic Sponsorship on Corporate Image. Journal of Advertising Research, (July/ August), 22-28.

Stiroh, K.J. (2007). Playing for keeps: Pay and Performance in the NBA. Economic Inquiry, 45 (1), 145-61.

Stone, G., Joseph, M., and Jones, M. (2003). An exploratory study on the use of sports celebrities in advertising: A content analysis. Sport Marketing Quarterly, 12 (2), 94-102.

Stryker, S., and Burke, P.J. (2000). The past, present, and future of an identity theory. Social Psychology Quarterly, 63 (4), 284-97.

Stuart, E. W., Shimp, T., and Engle, R.W. (1987). Classical Conditioning of Consumer Attitudes: Four Experiments in an Advertising Context. Journal of Consumer Research, 14 (3), 334-49.

Sujan, M., and Bettman, J.R. (1989). The Effects of Brand Positioning Strategies on Consumer’s Brand and Category Perceptions: Some Insights from Schema Research. Journal of Marketing Research, 26 (4), 454-67.

Sylvestre, C.M., & Moutinho, L. (2008). Leveraging associations: the promotion of cultural sponsorships. Journal of Promotion Management, 13 (3), 281-303.

Tabachnick, B.G., & Fidell, L.S. (2007). Using Multivariate Statistics: fifth edition. Pearson Education, Inc.: USA

354

Tajfel, H. (1970). Aspects of national and ethnic loyalty. Social Science Information, 9 (3), 119-44.

Tajfel, H. (1974). Social identity and intergroup behaviour. Social Science Information, 13 (2), 65-93.

Tajfel, H. (1978). Social categorization, social identity and social comparison. In, H.Tajfel (Ed.), Differentiation between social groups: Studies in the social psychology of intergroup relations, 61-76. London: Academic Press.

Tajfel, H. (1979). Individuals and groups in social psychology. British Journal of Social and Clinical Psychology, 18, 183-190.

Tajfel, H. (1981). Human groups and social categories: Studies in social psychology. Cambridge: Cambridge University Press.

Tajfel, H. (1982). Social Psychology of Intergroup Relations. Annual Review of Psychology, 33, 1-39.

Tajfel, H., and Billig, M. (1974). Familiarity and Categorization in Intergroup Behavior. Journal of Experimental Social Psychology, 10, 159-70.

Tajfel, H., and Turner, J.C. (1979). An integrative theory of intergroup conflict. In, (Eds.) W.G. Austin & S.Worchel, The Social Psychology of Intergroup Relations. Wadsworth Inc: USA, 33-47.

Tajfel, H., and Turner, J.C. (1986). The Social Identity Theory of Intergroup Behaviour. In, (Eds.) S. Worchel & W.G. Austin, Psychology of Intergroup Relations. Chicago: Nelson-Hall.

Tajfel, H., and Turner, J.C. (2004). An Integrative Theory of Intergroup Conflict. In, (Eds.) M.J. Hatch & M. Schultz, Organizational Identity: A reader. New York: Oxford: Oxford University Press, 56-65.

Tankard, J., Hendrickson, L., Silberman, J., Bliss, K., and Ghanem, S. (1991). Media frames: approaches to conceptualisation and measurement. Paper presented at the annual convention of the Association for Education in Journalism and Mass Communication, Boston, MA.

Tavassoli, N.T., and Fitzsimons, G.J. (2006). Spoken and typed expressions of repeated attitudes: matching response modes leads to attitude retrieval versus construction. Journal of Consumer Research, 33 (September), 179-87.

Taylor, S.A., Ishida, C., and Wallace, D.W. (2009). Intention to Engage in Digital Piracy: A Conceptual Model and Empirical Test. Journal of Service Research, 11 (3), 246-62.

Tellegen, A. (1988). The Analysis of Consistency in Personality Assessment. Journal of Personality, 56 (3), 621-63.

355

Tesser, A., and Shaffer, D.R. (1990). Attitudes and Attitude Change. Annual Review of Psychology, 41, 479-523. themotorreport (2009). F1: Renault confirms F1 commitment, but loses sponsors. http://www.themotorreport.com.au/43376/f1-renault-confirms-f1-commitment-but- loses-sponsors Last accessed: 11 February 2012.

Theodorakis, N.D., Koustelios, A., Robinson, L., and Barlas, A. (2009). Moderating role of team identification on the relationship between service quality and repurchase intentions among spectators of professional sports. Managing Service Quality, 19 (4), 456-73.

Theysohn, S., Hinz, O., Nosworthy, S., and Kirchner, M. (2009). Official supporters clubs: the untapped potential of fan loyalty. International Journal of Sports Marketing & Sponsorship, (July), 302-24.

Thomas, R.W. (2011). When Student Samples Make Sense in Logistics Research. Journal of Business Logistics, 32 (3), 287-90.

Thorndyke, P.W. (1977). Cognitive Structures in Comprehension and Memory of Narrative Discourse. Cognitive Psychology, 9, 77-110.

Toffoletti, K. (2007). How is gender-based violence covered in the sporting news? An account of the Australian Football League sex scandal. Women's Studies International Forum, 30, pp427-438.

Trimble, C.S., and Rifon, N.J. (2006). Consumer perceptions of compatability in cause-related marketing messages. International Journal of Nonprofit and Voluntary Sector Marketing, 11, 29-47.

Tripodi, J.A. (2001). Sponsorship – A Confirmed Weapon in the Promotional Armoury. International Journal of Sports Sponsorship & Marketing, (March/April), 1-20.

Tsiotsou, R., and Alexandris, K. (2009). Delineating the outcomes of sponsorship: sponsor image, word of mouth, and purchase intentions. International Journal of Retail & Distribution Management, 37 (4), 358-69.

Tulving, E. (1972). Episodic and semantic memory. In, E.Tulving & W.Donaldson (Eds.), Organization of Memory. New York: Academic Press.

Tulving, E. (1995). Organization of memory: Quo vadis? In M. S. Gazzaniga (Ed.), The Cognitive Neurosciences. Cambridge, MA: MIT Press, 839-47.

Tulving, E. (2000). Concepts of memory. In, E.Tulving and F.I.M Craik (Eds.), The Oxford Handbook of Memory. New York: Oxford University Press, 33-44.

356

Tulving , E., and Pearlstone, Z. (1966). Availability versus Accessibility of Information in Memory for Words. Journal of Verbal Learning and Verbal Behavior, 5, 381-91.

Tulving, E., and Thomson, D.M. (1973). Encoding specificity and retrieval processes in episodic memory. Psychological Review, 80 (5), 352-73.

Turner, J.C. (1982) Towards a cognitive redefinition of the social group. In H. Tajfel (Ed.), Social Identity and Intergroup Relations (pp15-40). Cambridge: Cambridge University Press.

Turner, J.C. (1999). Some Current Issues in Research on Social Identity and Self- Categorisation Theories. In, N. Ellemers, R. Spears, and B. Doosje (Eds.), Social Identity: context, commitment, content (pp6-34). Great Britain: Blackwell Publishers Ltd.

Turner, J.C., and Bourhis, R.Y. (1996). Social Identity, Interdependence and the Social Group: A Reply to Rabbie et al. In, P. Robinson (Ed.): Social Groups and Identities: Developing the Legacy of Henri Tajfel, (pp25-63). Great Britain: Butterworth-Heineman.

Turner, J.C., and Haslam, S. A. (2001). Social Identity, Organizations, and Leadership. In, M. Turner (Ed.), Groups at Work: theory and research, (pp25-65). USA: Lawrence Erlbaum Associates.

Turner, J.C., Hogg, M.A., Turner P.J., and Smith, P.M. (1984). Failure and defeat as determinants of group cohesiveness. British Journal of Social Psychology, 23 (2), 97- 111.

Turner, J.C., Oakes, P.J., Haslam, S.A., and McGarty, C. (1994). Self and Collective: Cognition and Social Context. Personality and Social Psychology Bulletin, 20 (5), 454-63.

Turner, J.C., Sachdev, I., and Hogg, M.A. (1983). Social categorization, interpersonal attraction and group formation. British Journal of Social Psychology, 22 (3), 227-39.

Turner, J.C., Wetherell, M.S., and Hogg, M.A. (1989). Referent informational influence and group polarization. British Journal of Social Psychology, 28 (2), 135- 47.

Umemoto, T., and Hilgard, E.R. (1961). Paired-Associate Learning as a Function of Similarity: Common Stimulus and Response Items within the List. Journal of Experimental Psychology, 62 (2), 97-104.

UPI Energy Resources (2009, December 19). TAG Heuer to ‘downscale’ Tiger Woods.

Vale, J., Serra, E., Vale, V.T., and Vieira, J.C. (2009). The impact of sponsorship on a football team’s brand equity. Journal of Sponsorship, 2 (3), 267-80.

357

Valgeirsson, G., and Snyder, E.E. (1986). A cross-cultural comparison of newspaper sports sections. International Review for the Sociology of Sport, 21, 131-9.

Van Heerden, N., Kuiper, A., and Saar, H.M. (2008). South African Journal for Research in Sport, Physical Education and Recreation, 30 (2), 147-65.

Van Knippenberg, D., and Sleebos, E. (2006). Organizational identification versus organizational commitment: Self-definition, social exchange, and job attitudes. Journal of Organizational Behavior, 27, 571-84.

Van Knippenberg, D., and Van Schie, E.C.M. (2000). Foci and correlates of organizational identification. Journal of Occupational and Organizational Psychology, 73, 137-47.

Van Osselaer, S.M.J., and Alba, J.W. (2000). Consumer learning and brand equity. Journal of Consumer Research, 27 (1), 1-16.

Van Leeuwen, L., Quick, S., and Daniel, K. (2002). The sport spectator satisfaction model: a conceptual framework for understanding the satisfaction of spectators. Sport Management Review, 5, 99-128.

Van Reijmersdal, E.A., Neijens, P.C., and Smit, E.G. (2007). Effects of television brand placement on brand image. Psychology & Marketing, 24 (5), 403-20.

Vasterman, P.L.M. (2005). Media-hype: Self-reinforcing news waves, journalistic standards and the construction of social problems. European Journal of Communication, 20, 508-30.

Vaughan, G.M., Tajfel, H., and Williams, J. (1981). Bias in reward allocation in an intergroup and an interpersonal context. Social Psychology Quarterly, 44 (1), 37-42.

Velleman, P.F., and Wilkinson, L. (1993). Nominal, Ordinal, Interval, and Ratio Typologies Are Misleading. The American Statistician, 47 (1), 65-72.

Vincent, J., Imwold, V., and Johnson, J.T. (2002). A comparison of selected ‘serious’ and ‘popular’ British, Canadian, and United States newspaper coverage of female and male athletes competing in the Centennial Olympic Games: Did female athletes receive equitable coverage in the ‘Games of the Women’? International Review for the Sociology of Sport, 37 (3-4), 319-35.

Von Hecker, U. (2004). Disambiguating a Mental Model: Influence of Social Context. The Psychological Record, 54, 27-43.

Wakefield, K.L., Becker-Olsen, K., and Cornwell, T.B. (2007). I Spy a Sponsor: The Effects of Sponsorship Level, Prominence, Relatedness and Cueing on Recall Accuracy. European Advances in Consumer Research, 7, 136-140.

Walliser, B. (2003). An international review of sponsorship research: extension and update. International Journal of Advertising, 22, 5-40.

358

Wann, D.L. (2006). Understanding the positive social psychological benefits of sport team identification: the team identification-social psychological health model. Group Dynamics: Theory, Research, and Practice, 10 (4), 272-96.

Wann, D.L., and Branscombe, N.R. (1990). Die-hard and fair-weather fans: effects of identification on BIRGing and CORFing tendencies. Journal of Sport and Social Issues, 14 (2), 103-17.

Wann, D.L., and Branscombe, N.R. (1992). Emotional responses to the sports page. Journal of Sport and Social Issues, 16 (1), 49-64.

Wann, D.L., & Branscombe, N. (1993). Sports Fans: Measuring Degree of Identification with Their Team. International Journal of Sports Psychology, 24, 1-17.

Wann, Dimmock and Grove (2003). Generalizing the Team Identification – Psychological Health Model to a Different Sport and Culture: The Case of Australian Rules Football. Group Dynamics: Theory, Research, and Practice, 7 (4), 289-96.

Wann, D.L., and Dolan, T.J. (1994). Attributions of highly identified sports spectators. The Journal of Social Psychology, 134 (6), 783-92.

Wann, D.L., Dolan, T.J., McGeorge, K.K., and Allison, J.A. (1994). Relationships between spectator identification and spectators’ emotions, and competition outcome. Journal of Sport and Exercise Psychology, 16, 347-64.

Wann, D.L., Grieve, F.G., Waddill, P.J., and Martin, J. (2008). Use of retroactive pessimism as a method of coping with identity threat: the impact of group identification. Group Processes and Intergroup Relations, 11 (4), 439-50.

Wann, D.L., Hamlet, M.A., Wilson, T.M., and Hodges, J.A. (1995). Basking in reflected glory, cutting off reflected failure, and cutting off cuture failure: the importancve of group identification. Social Behaviour and Personality, 23 (4), 377- 88.

Wanta, W., Golan, G., and Lee, C. (2004). Agenda setting and international news: media influence on public perceptions of foreign nations. Journalism & Mass Communication Quarterly, 81 (2), 364-77.

Webster, F.E. (1968). Interpersonal Communication and Salesman Effectiveness. Journal of Marketing, 32 (3), 7-13.

Weaver, D.H. (2007). Thoughts on agenda setting, framing, and priming. Journal of Communication, 57, pp142-147.

Weeks, C.S, Cornwell, T.B., and Drennan, J.C. (2008). Leveraging Sponsorships on the Internet: Activation, Congruence, and Articulation. Psychology & Marketing, 25 (7), 637-54.

359

West, P.M., Brown, C.L., and Hoch, S.J. (1996). Consumption vocabulary and preference formation. Journal of Consumer Research, 23 (2), 120-35. Wyer, R.S., & Carlston, D.E. (1979). Social cognition, inference, and attribution. NJ: Erlbaum.

Westberg, D. (1994). Right practical reason: Aristotle, action, and prudence in Aquinas. Oxford: Clarendon Press.

White, J. D., and Carlston, D.E. (1983). Consequences of schemata for attention, impression, and recall in complex social interactions. Journal of Personality and Social Psychology, 45, 538-49.

Wilson, B., Stavros, C., and Westberg, K. (2008). Player transgressions and the management of the sport sponsor relationship. Public Relations Review, 34, 99-107.

Yeager, D.B. (1993). A radical community of aid: a rejoinder to opponents of affirmative duties to help strangers. Washington University Law Quarterly, 71 (1), 1- 58.

Yu, C-C. (2009). A content analysis of news coverage of Asian female Olympic athletes. International Review for the Sociology of Sport, 44 (2-3), 283-305.

Zajonc, R.B. (1980). Feeling and Thinking: Preferences Need No Inferences. American Psychologist, 35 (2), 151-75.

Zinnes, J.L. (1969). Scaling. Annual Review of Psychology, 20, 447-97. Zelkovich, C. (2010). Twitter reveals #sports character; Bosh, Ochocinco, Rice among perpetrators of puerile, pointless and rarely profound posts. The Toronto Star, 26 December.

Zhao, G.Z., and Pechman, C. (2007). The impact of regulatory focus on adolescents’ response to antismoking advertising campaigns. Journal of Marketing Research, November, 671-87.

Zhu, R. (2002). What if the father commits a crime? Journal of the History of Ideas, 63 (1), 1-17.

360