<<

SEXISM, STEREOTYPING, AND THE GENDER WAGE GAP

A Thesis

Presented to

The Faculty of Graduate Studies

O t-

The University of Guelph

In partial fulfilment of requirements

for the degree of

Doctor of

June, 2000

@Christine Alksnis, 2000 National Library Bibliothéque nationale I*I of Canada du Canada Acquisitions and Acquisitions et Bibliographie Services services bibliographiques 395 Wellington Street 395. nie Wellington Ottawa ON K1A ON4 Ottawa ON K1A ON4 Canada Canada Your fila Vom nlemw

Our W Notre raterence

The author has granted a non- L'auteur a accordé une licence non exclusive licence allowing the exclusive permettant à la National Library of Canada to Bibliothèque nationale du Canada de reproduce, loan, distribute or sel1 reproduire, prêter, distribuer ou copies of this thesis in microform, vendre des copies de cette thèse sous paper or electronic formats. la forme de microfiche/fiim, de reproduction sur papier ou sur format électronique.

The author retains ownership of the L'auteur conserve la propriété du copyright in this thesis. Neither the droit d'auteur qui protège cette thèse. thesis nor substantial extracts f?om it Ni la thèse ni des extraits substantiels may be printed or otherwise de celle-ci ne doivent être imprimés reproduced without the author's ou autrement reproduits sans son permission. autorisation.

Canada ABSTRACT

SEXISM, STEREOTYPING, AND THE GENDER WAGE GAP

Christine Alksnis Advisor: University of Guelph, 2000 Dr. S. Desmarais

For decades, women's average earnings have been a fraction of men's earnings.

This dissertation is predicated on the assumption that gender stereotypes facilitate sexist behaviour by employers that in turn contributes, at least in part, to the persistence of this gender wage gap. The ways in which stereotypes lead women to be penalized in terms of hiring (access) and salary (value) was explored: the specific stereotypes examined related to segregation of the workforce and noms dictating that women assume more responsibility for childcare than men do.

Study I was primarily concerned with the question of whether women's work is valued as highly as that of men. This study checked whether undergraduates would assign different salaries to three types ofjobs wherein the actual responsibilities and duties camed out were the same, but the job was situated in either a traditionally masculine or ferninine domain.

Studies 2 and 3 examined whether and how women's expressed intentions to have a family hindered their access to jobs and affected the salaries they were assigned. In

Study 2, undergraduate participants were asked to listen to a mock interview and then to indicate whether they would hire the candidate and what saiary they would allocate. The

impact that candidate's family status. candidate's gender. and gender of panicipant had on these dependent variables was assessed. Four farnily status conditions were considered, wherein the candidate: (1) expressed an interest in starting a farnily; (2) revealed that s/he was unable to have children; (3) revealed that s/he did not want chiidren; or, (4) revealed that s/he was asthmatic [comparison condition].

In Study 3. a comrnunity sample comprised pnmanly of full-time workers and managers was asked to indicate whether a prospective employee's family status. living arrangements and comection to the community were advantageous or disadvantageous when applying for a job. Of primary interest was respondents' rating of four different types of familial circumstances, three of which were identical to those assessed in Study 2.

The ways in which the results fit with traditional sexist beliefs as well as with contemporary theonzing about "neosexism" and "modem sexism" are discussed. ACKNOWLEDGEMENTS

There are a lot of people to thank for their support dunng my Ph.D. career.

First, to Serge - I am so fortunate to have had the opportunity to work with you! Thank you for your enccuragement, your enthusiasm, and the many things you did to ease the

process. Thanks also to my advisory cornmittee members, Jim Curtis, Mary hnEvans

and Joanna Boehnert for their guidance, and special thanks to Jim for making it possible

for me to get involved with the KWMAS. I'm grateful to the extemal members of my

examination cornmittee, Francine Tougas and Karen Korabik for raising such interestin-

issues at my defense. and particularly to Francine Tougas for agreeing to corne to Guelph

while on sabbatical. Thanks also go to the chair. Brian Earn. for his contributions at the

defense and for creating such a Friendly atmosphere.

I've made some wonderfil fnendships in my time at Guelph. I'm especially grateful to Karen, Casey and Kem - thank you so much for being there for me. gooîïng

off with me and heiping me through. No one could ask for better Friends!

My appreciation goes to al1 the people who agreed to take part in my studies. and to those on campus who kindly allowed me to corne into their classes in order to recruit "just a few more people."

Thanks also go to my family for al1 the support that they've given me, not just through the Ph-D. but always. TABLE OF CONTENTS

THEGENDER WAGE GAP: A SUMMARY OF THE PROBLEM...... -7 SEXISM:AN OVERVIEW...... 6 "*New" Types of Srxisrn ...... -8 Relevant Lrgsfatioti ...... 10 GENDERSTEREOTYPES: BELEFS THAT HELP MAINTHE GENDER WAGE GAP ...... 12 Stereotypes rehtirig to oc~r~pationdsrgregatior? ...... 13 Stereotypes relating to fomiiy resy amih ili tirs ...... 15 Subtypes ...... 17 Thearefical approaches to sterrotypi~~g...... 18 MACRO-LEVELTKEOEUES ...... -20 Econornic theorirs ...... 21

STUDY HYPOTHESESFOR PWY &VALYSES ...... 37 METHOD...... -38 Powcr Calcufatioru ro L)ererrnitie S'ample Six ...... 38 Porticipunts ...... 39

Annual Salary ...... -42 Compensatory Dimensions ...... -42 Previrw of Analyses ...... 43 Manipulation Checks ...... -43 Pnmary Analyses ...... -44 Supplementary Analyses ...... -45 RESULTS ...... -46 Mrmipdation <'hecks ...... 46 Gender assigned to jobs by participants ...... 46 Evaluation of jobs on compensatory factors ...... 48 Primary Analysrs...... 50 Al1 Participants ...... 50 Participants who açree with stereotyped gender ofjob ...... 53

Maniplation C'hrcks ...... 56 Primaty & Szrppkmetrtaty At~aiyses...... 57 Limitations ...... 61

Full-Time Non-Managerial Workers ...... 144 Managers ...... 146 Drscuss~o~...... 149 Preihinary Anulysrs ...... 149 Primary Analyses ...... 1% Impact of Applicant 's Gender ...... 150 Impact of Respondent's Gender ...... 154 DEerences between Study 2 and 3 ...... 156 GENERAL DISCUSSION ...... 160 SUMMARY OF FINDIN(~S...... 160 LIMITATIONS...... 164 ~MPLICATIONS...... 166 Policies ariLiressir~gthe Geirder IVtrpi Gap ...... 166 Employer and Employee Reactions to Policies ...... 169 The Bottom Line ...... 172 Individual-Lr vel Factors ...... 172 The Gender Gap and Potrrrtial Itfoihrrhood ...... 175 CONCLUSIONS...... 178 REFERENCES ...... 180

APPENDIX A: STUDY t QUESTIONNAIRE ...... 199 APPENDK B: STUDY 1 TABLES OF MEANS AND SUMMARY TABLES..... 208 APPENDIX C: SWDY 2 iMOCK INTERVIEW SCRIPT ...... 216 APPENDIX D: STUDY 2 QUESTIONNAIRE ...... 221 APPENDIX E: STUDY 2 TABLES OF MEANS AND SUMMARY TABLES ..... 234 APPENDM F: STUDY 3 QUESTIONS ...... 257

APPENDIX G: STUDY 3 TABLES OF MEANS AND SUMMARY TABLES .... 259 List of Tables

Table 1 Background for Study 1 Panicipants ...... 40

Table 2 Gender Assigned to Target Job by Gender of Respondent ...... 48

Table 3 Mean Salary Estimates for 3 Types of Jobs by Gender Assigned to Target Job for Male and Female Participants: MI Participants ...... 5 1

Table 4 Mean Salary Estimates for 3 Types of Jobs by Stereotypes Gender of Job for Male and Female Participants: Participants who agree with Stereotypes Job Gender ...... -...... 54

Table 5 Background Information for Study 2 Participants ...... 76

Table 6 Mean Ratings of Commo~ess/~ppropriateness/Re~evanceof Interview Questions ...... 93

Table 7 Mean Ratings of Male and Female Applicant's Dedication by Pmicipant Gender and Familykiealth Status ...... 97

Table 8 Mean Ratings of How Well Male and Female Canadiates Would "Fit in" by Participant Gender and Family/Health Status...... 98

Table 9 Mean Estimates of Number of Years Male and Female Candidates will Stay with Hiring Company by Participant Gender and Family/Hedth Status ...... 99

Table 10 Mean Ratings of Male and Female Applicants' Likelihood to be Hired by Participant Gender and FamilyRlealth Status...... 105

Table 1 1 Mean Starting Salary Assiçned to Male and Female Applicants by Participant Gender and FamilyMealth Status...... 108

Table 12 Summary of Significant Effects for Primary Analyses of Study 2...... 1 14

Table 13 Background Information for Participants in Study 3 ...... 129

Table 14 Mean Ratings of -\dvantageousness of Employee Farnily/Living Charactenstics by Gender of Applicant and Gender of Participant: Ail Respondents ...... -...... 137

Table 15 Mean Ratings of .-\dvantageousnessof Employee Farnily Charactenstics by Gender of Applicant and Gender of Participants: Full-Time Non-Managers

Only ...... -...... -...... *...... *. 145 Table 16 Mean Ratings of Advantageousness of Employee Family Charactenstics by Gender of Applicant and Gender of Participant: Managers Only ...... 147

Table 17 Sumrnary of Significant Effects for Primary Analyses of Study 3 ...... 151

Table B-1 Factorial MANOVA on 9 Compensatory Factors ...... 209

Table B-2 Compensatory Factors .. Means and Standard Deviations by Job Lncome Level ...... -210

Table 8-3 Factorial ANOVA on Salary for Al1 Participants ...... ZII

Table 8-4 Factorial ANOVA on Salary: Participants who agree with Stereotyped Gender of Job ...... 212

Table B-5 Correlations between Potential Covanates and Salary by Stereotyped Gender of Job and Job Income LeveI ...... 213

Table B-6 Factorial WCOVA on Salary with Length of Tirne in Workforce and Ratings on Compensatory Factors Covaried ...... 214

Table 8-7 Mean Salary Estimates For 3 Types oFJobs by Stereotyped Gender of Target Job for Male and Female Participants: Covanate Analyses ...... 715

Table E- 1 Principal Component Analysis of Items Related to Candidate's Suitability as an Employee ...... 235

Table E-2 Mean Ratings of Compensatory Factors ...... 236

Table E-3 Factorial MANOVA on Ratings of Interviewer ...... 137

Table E-4 Mean Ratinçs of interviewer's Performance ...... 237

Table E-5 Factorial MANOVA on Ratings of Candidate ...... 238

Table E-6 ANOVA on Ratinçs of Candidate .. Confident ...... 239

Table E-7 ANOVA on Ratings of Candidate .. Hesitant ...... 240

Table E-8 ANOVA on Ratinçs of Candidate .. Good impression ...... -241

Table E-9 ANOVA on Ratinçs of Candidate .. Spoke Clearly ...... 242

Table E-IO ANOVA on Ratings of Candidate .. Handled Inte~ewWeil ...... 243 Table E-1 1 ANOVA on Ratings of Candidate .. Similar to Actual Candidate ...... 244

Table E- 12 Ratings of Candidate's Performance by Gender of Participant. Gender of Target ...... 245

Table E-13 Factorial WOVAon FamiIyRlealth Status Query: Commo~ess. Appropnateness. Relevance ...... -246

Table E- 14 ANOVA on FamilyNealth Status Query: Commo~ess...... 247

Table E-15 MOVA on FarnilykIealth Status Query: Appropnateness...... 248

Table E- 16 ANOVA on FarnilyWealth Status Query: Relevance ...... 249

Table E-17 Family/Health Status Queries: Mean Ratings of their Relevance by Gender of Applicant and Gender of Participant ...... 250

Table E- 18 Factorial MANOVA on Applicant Dedication . "Fitting in" and Turnover Potentiai ...... 251

Table E-19 Factorial ANOVA on Candidate's Dedication as an Employee ...... 252

Table E-20 Factorial ANOVA on How WelI Candidate Would "Fit In" ...... 253

Table E-2 1 Factorial ANOVA on Number of Years Candidate Will Stay with Company ...... 254

Table E-22 Factorial MOVA on Candidate's Likelihood of Being Hired ...... 255

Table E-23 Factorial ANOVA on Starting Salary Assigned to Applicant ...... -256

Table G-l Mixed-Design ANOVA on 9 Employee Characteristics: Al1 Respondents...... -260

Table G-2 Mixed-Design 3-way ANOVA on 4 Family Status Items: AI1 Respondents ...... 261

Table G-3 Mixed-Design 3-way ANOVA on 4 Family Status Items: Full-Time Non-Managers Only ...... 261

Table G-4 Mixed-Design 3-way ANOVA on 4 Family Status Items: Managers Only ...... 263

Table G-5 Intercorrelations Arnong Potential Covariates and Dependent Variables by Gender of Tarçet and Participant Gender for Al1 Respondents ...... 264 List of Fiçures

Figure 1 Mean Ratings of Compensatory Factors by Job: MI Participants...... 49

Figure 2 Mean Salary by Job. Gender of Job and Participant Gender: Ail Participants .. 52

Figure 3 Mean Salary by Job. Gender of Job and Participant Gender for Participants who Agree with Stereotyped Job Gender ...... 52

Figure 4 Mean Ratings of Compensatory Factors for Assistant Bank Manager: Ai1 Participants ...... 88

Figure 5 Mean Ratings of Interviewer's Performance on Four Dimensions: Al1 Participants ...... ,...... 89

Figure 6 Mean Ratings of Candidate's Performance on Six Dimensions ...... 91

Figure 7 Mean Ratings of FamilyEiealth Question on 3 Dimensions by Candidate's FamilyMealth Status and Gender ...... 95

Figure 8 Mean Dedication Ratings by FamilyRIealth Status: Separately by Applicant Gender and Overalt ...... 10 1

Figure 9 Predicted Number of Years Candidate Will Stay with Company by Familyl Health Status .+...... 103

Figure 10 Mean Ratings of Likelihood of Hirinç Candidate by FamilyMealth Status . . . 106

Figure 1 I Mean Salary Assiçned to Candidate by FarnilyMealth Status ...... 109

Figure 12 Interaction between FamilylLiving Arrangements and Applicant Gender: Al1 Respondents...... 139

Figure 13 Interaction between Respondent Gender and Family Status ...... 142

Figure 14 Interaction between Applicant Gender and Family Status ...... 143 Sexism Stereotyping, and the Gender Wage Gap

Introduction

Surveys of workers engaged in a wide variety of occupations have shown that for decades. women's average eaniings have been a fraction of men's eamings (Statistics

Canada, 1999a). This gendered wage gap has also been denonstrated within occupational domains: corporate (Frieze, Olson. & Good. 1990; Jacobs. 1992). acadernia (Bellas.

1994), and law (Kay & Hagan. 1995) to name a few. Social scientists who seek to explain the gender wage gap rnust answer two questions: (1) is there differential accrss to jobs on the basis of gender; and (2) is women's work perceived to have less value than comparable work done by men? (Drolet. 1999). If the answer to either query is in the affirmative then sexism. broadly defined as " and against women as

members of a social category" (Unger & Saundra, 1993; p. 14 1). still contributes to the wage gap. This dissertation explores some of the ways in which stereotypes influence women's workplace outcomes such that women are penalized in terms of hinng (access) and saiq (value).

Stereotyprs involve the association of personality traits. , physicai attnbutes and occupations with social groups (Deaux & Lewis. 1984). Stereotyping is the act of using these stereotypes when fonning an impression of an individual (Brewer, 1996). Of course, it is not possible for of the gender-linked stereotypes that could possibly have an impact on how employers treat women to receive in-depth examination in this dissertation. The ones explored here relate to two wage gap-related phenornena that have

proven to be remarkably resistant to change over the years: gender segregation of the labour force and cultural noms stipulating that women shoulder more domestic responsibilities, particularly in the realm of childcare. than men are required to undertake

(Green & Russo, 1993; Shapiro & Stelcner. 1987). Using both expenmental and survey

methods, the three studies that comprise this thesis examine stereotypes about women's

occupational and familial roles in an effort to show that regardless of whether a job itself is

considered female-typed or the candidate for a job is a wornan, negative appraisals often

ensue. The ways in which the obtained results fit with traditional sexist beliefs as well as

with contemporary theorizing about "neosexism" and "modern sexism" will be explicated

(Swim, Aikin, Hall. & Hunter. 1995; Tougas. Brown. Beaton. & Joly, 1995).

The Gender Waee Gao: A Sumrnary of the Probiem

When Statistics Canada first started reponing on the gender wage gap in 1967.

women who worked full-time year-round eamed approximately 58% of what men did. By

1997, the most recent year for which data is available. the gap had been reduced. but the

fernalemale pay difference was still substantial at 72.5% (Statistics Canada, 1999a).

Segregation of the labour force along gender lines is frequently cited as one of the major

reasons for the wage gap between male and female full-time workers (hstrong &

Armstrong, 1994; England, 1992; McDermon, 1996; Treiman & Hartmann, 198 1 ).

Women and men tend to be found in different occupations. with men being represented in

far more types of occupations than women are (Benokraitis & Feagin 1995; Unger &

Saundra, 1993). Furthemore. many of the occupations that pay higher wages are

currently male-dominated (e.g., engineering, finance), accounting for at Ieast some of the

gap. It has been estimated that just over 50% of women would need to change their

occupations in order to achieve a distribution that is similar to men's (Reskin & Padavic,

1994). Men and women also tend to hold different types of jobs within occupations. Jobs are even more highiy segregated than occupations, with over three-quarters of women needing to change jobs before women and men are equally distnbuted (Tomaskovic-

Devey, 1995). Segregation Frequently occurs at the level of firms, departments and shifts as well (Bielby & Baron, 1986). The upshot of these different divisions along gender lines is that the vast majority of employees (approximately 90%) cany out t heir day-to-day work in settings composed primanly (at least 80%) of same-sex workers (Bielby & Baron.

1986).

In addition, anaiysis of wage data collected throughout the years has yielded a consistent pattern of a srna11 gap for cohons entering the labour force and a considerablv wider gap for older cohons (England, 1992). This increase in the female-male pay difference that occurs over the life cycle is at least partially explicable by the fact that women are more likely than men to make choices that involve forgoing work experience for farnily reasons, either by tuming down ovenime, promotions and/or transfers or by

leaving the labour force to bear and raise children (CARNET. 1992. cited in Lero &

Johnson, 1994; Marshall, 1994; Moen. 1985). The desire to avoid shortchanging one's

family may aIso explain some women's decisions to refrain from obtaining the educational training required to enter well-paying yet time-intensive professional careers. thus contributing fùnher to the gendered wage gap (Eccles. 1994).

Attempts to quanti@ the contribution of vanous factors to the gap indicate that

even after statistically controlling for occupation and a wide range of other work and

educational history variables (e-g., years of schooling, job tenure, organizational tenure.

su pervi sory responsibilities, etc.). over 50% of the wage gap between male and female full-time workers remains unexplained (Christofides & Swindinsky. 1994; Drolet. 1 999;

Jagacinski, LeBold, & Linden, 1987; Kilboume. England. & Beron, 1994; Wellington.

1994). Psychologists and sociologists often make the inference that this unaccounted for variance in salary is due to gender-based discrimination (Wilson, 1996) whereas econornists are more likely to attribute the remaining variance to market factors (e.g.. an oversupply of workers which leads to lower waçes), or to unmeasured worker characteristics (Fuchs, 1 988; Perlman & Pike. 1 994).

The fact that unexplained gender differences in average wages have been detected even at career entry, a time when there is not yet differentiation in experiential factors, suggests that discrimination against wornen truly & a problem in today's labour market.

For example, a female-rnale wage ratio of 84% was found in a recent American study of young workers in a variety of occupations (Marini & Fan, 1997). Gendered pay differentials are also present among highly educated entrants to the workforce. Davies.

Mosher and O'Grady ( 1996) looked at three cohorts of Canadian university graduates and found that their pay ratios were between 83% and 9 1%. As with older full-time workers. a substantial portion of the gap cannot be explained; Davies et al. (1996) found that the proponion of gap accounted for ranged from a low of 26% to a high of 52%. These male-fernale differences, although srnaller than those found among Full-time workers. are not trivial - any pay disadvantage that occurs at career outset can be greatly compounded over the years given that pay raises, when granted. usually consist of a percentage of one's current wage (Bamum,Liden, & DiTomaso, 1995).

Research on how men and women corne to inhabit separate spheres in the world of work and why women tend to have lower average salaries has been a multidisciplinary endeavour. Investigators working From varied perspectives tend to disagree about the that sexisrn plays in maintaining the wage gap. To summarize the major theoretical frameworks briefly. economists usually claim that market forces act to rninimize. or

perhaps even eradicate. sexism at work - most tend to emphasize women's choices when explaining segregation and pay differentials (Chnstofides & Swidinsky, 1994; Fuchs.

1988; Gunderson, 1998; Hakim, 1995). Psychologists and sociologists are more inclined to assert that sexism accounts for at Ieast sorne of the prevailing gender disparities. though there are differences of opinion about the mechanisms underlying the inequities currently observed. Some contend that segregation and the pay gap are the by-product of dischinatory methods that were built into organizational systerns over the years and that

today's employers no longer actively seek to hinder women (Canadian Human Riçhts

Commission. 1998a). Others argue that employers still hold beliefs that are stereotypical and sexist. and that these beliefs can lead to discriminatory conduct that is sometimes

intentional and sornetimes not (Benokraitis & Feagin, 1995; Nonhcraft & Gutek. 1993).

Finaily, some claim that women's continued disadvantaged position in the workplace is rnostly attributable to active efforts on the pan of employers and advantaged employees to

keep wornen out of better-paying jobs. thus maintaining the status quo (Reskin. 1988;

Tomaskovic-Devey, 1995).

Whether or not it will ever be possible to close the gap clearly depends on which of these proposed expianations is closest to the tmth. If segregation and pay variations are due to the existence of outmoded discriminatory systems that remain in place due to some fonof organizational inertia (Baron. 199 1). or are attributable to unintentional acts by individuals that coincidentally have deleterious effects for women, then perhaps the will to create a fairer system exists and it may eventually be possible to eradicate the gender wage gap via interventions. If empioyers' conduct is intentionally discriminatory and/or there is collusion to keep women at a disadvantage. then the outlook for bridging the gap is far less optirnistic. This dissertation focuses on the implications that individuals' stereotyping has for maintaining the gender wage gap and attempts to speak to the issue of whether and how this cognitive process can be transcended. First however, it is necessary to @ the term "sexism."

Sexism: An Overview

Sex dominates race, age and occupation as a basis for the social classification of others (Brewer & Lui. 1989; Fiske, 1998; Fiske. Haslam. & Fiske. 199 1 : Fiske & Stevens.

1993; Stangor, Lynch. Duan, & Glass. 1992). The preeminence of this means of classification is evidenced by the fact that people find it very difficult to interact with someone when they are unsure of that person's sex. whereas social relations are not necessarily hindered due to ignorance of another's age or racial background (Ridgeway.

1997). Before continuing, it should be noted that the terms "sex" and "sender" are not always used consistently in the . In . many researchers and theorists tend to use the word "sex" to refer to distinctions that are biological in origin and

"gender" to refer to psychological and social constmctions of what it means to be male or fernale (Deaux, 1985; Unger, 1990). Given that stereotypes are psychoiogical and social constructions, the term "gender" is used throughout this dissertation. 1

I However. the fact dut psychologists use the tenn .'sesism" rather tlwn "gendensm" seems to fly in the face of the rule just aniculated. Regardless. the term "sesism" is used in ihis investigation. The impact of gender in our day to day lives is both insidious and pervasive

(Deaux & LaFrance. 1998; West & Zimmerman. 1987). Unger and Saundra (1993) have made similar statements about sexism and delineate how it may occur at individual. social. institutional and cultural levels. Briefly. individual sexism entails negative cognitive against women in the form of in-group/out-group categorization. biases in evaluation/attnbutions. and stereotypes. It should be noted that the latter have been descnbed as both descriptive and prescriptive. and that the prescriptive nature of stereotypes is thought to be more potent in the realm of gender than it is in other domains

(Fiske & Stevens, 1993). As we shall see. this panicular feature of gender stereotypes may contnbute to wornen's chronic disadvantage in the workplace by restricting the work roles that are "appropriate" for women. Social sexisrn refers to interactions between individuais that involve negative behaviours directed at targets as a result of their gender

(e.g., less attending, more criticizing, etc.). lnstitutional sexism involves the differential impact of institutional structures on men and women (e.g.. the strength of the relationship arnong educatioh expenence and income difFers depending on one's gender. such that men reap greater benefit). Cultural sexism encompasses societal assumptions and beliefs about the value of gender categories (e.g.. that "male" characteristics are better than

"fernale" ones) as well as the general social practices that maintain gender discrimination

(e.g.. legislation).

At each of these four levels. the expression of sexism may differ on dimensions such as visibility, intent, degree of hminfiicted on recipient. ease of documentation, and availability of rernedial measures. such that sexism can be manifested in blatant, subtle or coven ways (Benokraitis & Feagin, 1995). Moreover, these four levels interact with each other, reflecting and reinforcing each other (Unger & Saundra. 1993). The curent investigation focuses on just such an interaction in the realm of employment, that is. the gender gap in salaries as an exarnple of institutional sexism that is fostered by individuals'

(i.e., employers') sexist and behaviours, which are in tum reflective of cultural views,

"New" Tmes of Sexism

With respect to individual level sexism, a distinction has been made between -'old- fashioned" sexism, which entails endorsement of traditional gender roles accompanied by a host of attitudes and behaviours that presuppose wornen's inferionty (see Carneron,

1977), and a newer form. which is characterized by denial of ongoing gender-based discrimination in society and its institutions. opposition to women's politicai and economic demands. and resentment of gains that women obtain through special programs such as (Swim et al.. 1995; Tougas et al.. 1995). This recent phenornenon is referred to as "neosexism" by Tougas and her associates whereas Swim and her colleaçues use the tenn "modem sexism."

Efforts to assess individual differences on these constmcts have revealed that while they are correlated with old-fashioned sexism (Swim et al. report 4's between .47and .54 and Tougas et al. report a zero-order correlation of 34). the newer constnicts are more strongly related to individuals' notions regarding women's employment. To illustrate.

Tougas et al. (1995) found that neosexism was inversely related to evaluations of women's potentiai as workers and that it was significantly negatively correlated with support for affirmative action whereas old-fashioned sexism was not. Swim et al. ( 1995) determined that modem sexists were more likely than old-fashioned sexists to underestimate the extent of gender segregation in the workforce, and less likely to attribute segregation to discrimination ador prejudice against women.

These contemporary forms of sexism are thought to be increasingly prevalent in today's "politicaily correct" climate as they allow people to support policies and views that consistently disadvantage women without publicly endorsing blatantly negative attitudes about them (Tougas et al., 1995). Indeed. people are far less likely ro express flagrantiy rnisogynistic notions than they have been in the past (Fleming, 1988; Kahn &

Crosby. 1985: Mason & Lu. 1988). In fact. attitudinal surveys imply that there is widespread acceptance of and support for in access to employment opponunities (Baine, 1999) and pay (Crosby, Farrell. & Cameron. 1994). Yet these professed egalitarian views stand in stark contrast to the inequities expenenced by women in the workforce (Kahn & Crosby, 1985). hinting at the possibiiity that these attitudes are not translated into action, at least not by empioyers. Of course. the attitude-behaviour consistency issue is not a new one in ; there is a large body of research documenting such discrepancies.

This dissertation was conducted with an eye toward ascertaining whether the responses of study participants reflected the persona1 inconsistency alluded to in Touças et al.'s (1995) definition of neosexism: "a manifestation of a between egalitanan values and residual negative feelings toward women" (p. 843). Sirnilar aims have guided past research. For example, the demonstration of professed attitudes toward women that are positive when asked directly, but gender biased when participants were allowed to answer fiom someone else's perspective has been taken as evidence of the operation of modem sexism (e-g.,Harvie, Marshall-McCaskey. & Johnston, 1998). The approach taken in this investigation was to check whether individuals would rate women and men as equals in some respects, but nevertheless make decisions about access and value that indirectly hindered women. The question of whether or not participants were aware of any incongmity was not addressed, however the theones as articulated suggest that individuals are not conscious of their lack of consistency (Eberhardt & Fiske, 1996).

Relevant Leeislation

In recent years, policies and legislation that are actually predicated on the assumption that sexism (as well as other types of discrimination) occur in the workplace have been enacted. /anti-discrimination legislation, which is designed to prevent the unequai treatment of individuals in employment and a variety of other settings

(e.g., housing, provision of services2),exists at both the federal and provincial levels

(Canadian Human Rights Act. 1976; Ontario Human Rights Code. IWO). Federal legislation protects only those individuals whose employment falls within federal junsdiction (Le.. federal public servants. individuals working in the communications and transportation industries) whereas provincial legislation. at least in Ontario, is more Far- reaching and thus more pertinent to most citizens.

There are a oumber of "protected grounds" outlined in both the federai and provincial human rights codes - a partial list includes age. gender, . marital status. family status. race. ancestry. place of origin, and

Basically, this legislation prohibits employers from engaging in recniitment. hiring, promotion or wage-setting practices that unfairly disadvantage individuals on the basis of

The areas covered depend on the panicular piece of legislaiion. one or more of the protected grounds. There are some exceptions to this general nile where restncting the applicant pool is deemed justifiable (e.g., for jobs that involve the provision of seMces to special groups like battered women. or as part of employment equity/affirmative action programs designed to rectifi unjust . selection and/or promotion practices of the past; Ontario Human Rights Code, 1990). This dissertation focuses on the question of whether individuals are disadvantaged dunng hiring or wage setting on the grounds of gender or family status.

Ln addition to human rights legislation. three specific policies that aim to reduce pay disparities between groups have been enacted at the federal level as well as in some provincial jurisdictions. Each of these policies endeavours to correct a different type of gender imbalance in employment: ernployment equitylaffirmative action3 initiatives are designed to facilitate the entry of women into jobs traditionally held by men; equal pay poiicies equalize wages between men and women in those jobs that are relatively well integrated with respect to gender: and. comparable worth (a.k.a. pay equity or "equal pay for work of equal value") is aimed at increasing the wages of women who remain in jobs that are traditionally held by women by ensuring that they are paid similarly to "male" jobs of equal difficulty and equal education. responsibility and skiIl requirements (Blum. 1987;

England, 1992; England & Markowitz. 1996).

The of these corrective policies indicates that there has been a continuing problem with sexism in employment settings. The question remains however. are the cunent gender dispanties simply the result of outdated seist institutionai practices that

In Canada. the term -empIoyment equity" is used wliereas in ihe . the term used is "affinnative action." Both phesdescribe the same type of initiative. persist because of organizational inertia (Baron, 199 1). or do individual-level phenornena, like adherence to sexist stereotypes by employers, continue to play a role in maintaining unfair practices?

Gender Stereoty~es:Beliefs that Hel~Maintain the Gender Wace Gap

Many social scientists. regardless of their theoretical . will concede that stereotypes have contnbuted to gender inequities in the past. It is well-documented that it was cornmon practice for employers to separate male and female workers. and to permit women entry only to selected occupations. Job advenisements were often directed specifically at one gender or the other (Bem & Bem. 1973). Furthermore. Levinson's

( 1975) field study demonstrated that this form of was usudly accompanied by attempts to deter prospective applicants if they were of the "wrong" gender. Women calling to inquire about male-typed jobs were advised that the jobs were too hard. required irregular hours, or called for too much physical strength. Men who called to inquire about advenisements for female-typed jobs were also discouraged fiom pursuing those positions by employers; the jobs were frequently described as too easy. boring, or low-paying for men and on occasion, male callers were actually informed about a better-paying job that was avaiiable.

Moreover. few researchers will dispute that when women first entered the paid workforce, they were paid less than men even when doing the same job because it was assumed that women's monetary contributions to their households' income were merely

"extras." Essentially, ernployers justified paying wornen a wage that could not provide adequate subsiaence for a farnily because it was generally accepted that men should work outside the home to support their families while women should concentrate their efforts within the home (Kessler-Harris. 1987). The curent investigation aims to show that stereotypes about occupational and familial roles still intluence women's employment outcornes.

Stereotyes relatine to occupational semegation

There exists a stereotype that wornen are generally less competent than men are

(Eagly & Mladanic. 1994; Fiske. 1998) Women's competence must be proved while men's is taken for granted (Lockheed & Hall. 1976). Research suggests that this alleged inadequacy causes women to be viewed as infenor workers and to experience hiring discrimination (Foschi. Lai. & Sigerson. 1994: Foschi. Sigerson, & Lembesis. 1995).

Notions about competence may combine with stereotypes about women's cognitive. physicai and/or personality charactenstics to facilitate by intluencing employers' and employees' ideas about the type of jobs that women are capable of doing (Bielby & Baron, 1986). Job characteristics that coincide with stereotypes about male and female traits make it such that those jobs are viewed as

"women's" or "men's" work (Martin. 1992).

Not oniy are women's characteristics perceived to render them unsuitable for certain jobs, jobs are seen as requinng the characteristics of the group that already dominates it. In other words, it appears that individuais engaged in hiring tasks ofien

(mistakenly) assume that because men are more likely than women to hold a particular job, it is necessary to have "men's" traits in order to do that job (Cejka & Eagly. 1999). The circularity of reasoning that accompanies gender stereotypes about traits and cornpetencies can certainly help explain why women have had difficulty entenng the ranks of many male-dominated jobs. By way of example. it has been suggested that the deanh of women in the highest levels of management is attributable in pan to the fact that they are perceived as lacking the "male" traits (e-g., assertiveness, toughness, independence) needed (Chusmir. 1984; Eagly & Mladanic, 1994). Paradoxically, it has been shown that even wornen who have the requisite "male" characteristics may find themselves barred

From rising up through the ranks because they are no longer sufficiently ''ferninine" - a testament to the strong presctiptive component of gender stereotypes (Fiske. BersoK

Borgida, Deaux. & Heilman, 199 1 ).

Stereotypes about occupations. which incorporate information about the

"appropriate" job-holder sex. grew out of the repeated pairing of gender with particular jobs (Beggs & Doolittle. 1993; Glick, Zion. & Nelson. 1988). These stereotypes have implications for gender-differentiated access to jobs, as demonstrated by Glick et al's

(1988) finding that even if the previous work expenences of hypotheticai male and female job candidates are more applicable to an "opposite-sex" job. raters are nevertheless more likely to interview women for jobs that are typically female-dominated and men for male- dominated jobs. These occupational stereotypes also have ramifications for the valuing of occupations; when data are aggregated across different occupations. people demonstrate their understanding that salary and prestige levels are lower for female- than for male- dorninated occupations (Glick Wilk, & Perreault. 1995). Study 1 of this dissertation

focuses on this phenornenon of differential valuing - it aims to make the case that the devaluation of women's work continues to occur by illustrating that women's jobs are

consistently assigned less remuneration even when the "male" and "female" jobs being

compared are identical with respect to their requirements. Stereotges relating to familv responsibilities

It is assumed that people, especially women, are naturally inclined toward having children (Mïall, 1985; Thompson & Walker. 1989). There is evidence to suggest that people who violate this nom are perceived in a negative light; specifically, they are seen as psychologically maladjusted, lonely, unhappy, emotionally immature, materialistic, selfish, and misguided in their decision to remain childless (Mid. 1986; Peterson, 1983;

Veevers, 1980). More recently. Lampman and Dowling-Guyer ( 1995) found that individuals who are voluntarily childless are generally seen as less driven and less canng than those who are involuntanly childless or who have children.

The expectation that most people will eventually have a family, together with internalization of the notions that men should be "breadwinners" within families while women should be responsible for dornestic chores and childcare (Bernard. 198 1 ; Daniels.

1987; Szinovac, 1984). could quite conceivably comibute to gendered pay differentials.

As in the past, employers confronted with women of childbearing age likely assume that these women will start a family at some point in the future if they have not done so already. Furthemore, the stereotype that these women are the caregivers and not the breadwimers in their farnilies likely permits employers to jus@ paying women less t han men (England, 1992).

Stereotypes about female domesticity may also have negative repercussions for women's access to jobs, particularly in professional domains (Konrad & Cannings. 1997;

Northcraft & Gutek, 1993; Swiss & Walker. 1993). Essentially. women are seen as

fulfilling their role by placing their farnilies above paid work (Daniels & Weingarten. 1988;

Pleck, 1987). Even when married women do work outside the home. and the majonty (6 1.5%) of them do (Statistics Canada, 1 999b). gender stereotypes prescribe that wornen's primary role "should" be that of family caregiver. As such. women are believed to be less committed to staying in the workforce than men are (Parasuraman & Greenhaus.

1993). Unfominately, there is little in the way of research that directly examines employers' attitudes regarding gender-linked family roles (Reskin. 1993) or how employers ascertain employees' level of commitment (Allen. Russell, & Rush 1994).

However. anecdotal evidence certainly supports the notion that women's familial role results in restricted access to certain jobs. For instance. Benokraitis and Feagin (1 995) relate comments overheard in an academic setting that illustrate how women are eliminated as serious candidates for promotion on the basis of familial responsibilities.

Examination of these remarks reveals that women at al1 stages of the life cycle were efKective!y excluded: younger women who were getting mamed were discounted because they were probably "too busy" planning their wedding, women with young children were judged to be too occupied with child-rearing, whereas women who had grown children were perceived to be wanting to enjoy their empty nest and thus unlikely to want to cake on the burdens of a more senior position.

Although the curent investigation explores gender in hiring rather than in promotion, it is not much of a stretch to suggest that the sentiments described by

Benokraitis and Feagin (1995) are influential in hiring situations as well. Furthemore. it probably is not necessary for employers to know with ce~aintywhether a woman has a farnily in order to make stereotypical assessments about her. People can and do engage in stereotyping on the bais of appearance alone (Brewer & Lui, 1988) - one can usually estimate roughly how old a person is and rnake a corresponding inference about the type of family demands they might have. At the very least. one is able to deduce that a woman is of child-bearing age; as such. there is always the chance that she will not be reliablv available to continue working for her employer in the long term. Studies 2 and 3 were conducted in an effort to provide a direct test of whether and how wornen's expressed intentions to have a farnily hindered their access to jobs and affected the salaries they were assigned.

Subtypes

There may be circumstances under which broad stereotypes about women and domesticity are recognized as inapplicable. As Deaux and Kite (1993) explain: "in . each gender stereotype refers to approximately half of the world's population. Alrnost by definition, such general categories lack discriminatory power: as a consequence. people develop beiiefs about particular kinds of women and men that may or may not share features with the niore general categories" (p. 1 15). Encountering women who have traits that contradict general stereotypes does not usually result in major modification of the stereotype as a whole. Rather. such women are seen as "special cases" (Weber &

Crocker, 1983). and a stereotype that describes this subgroup is adopted while al1 other women continue to be slotted into the already established general stereotype (Hewstone.

1994; Rothbart & John. 1985). It should be noted that disconfirming evidence rnust be presented on several occasions in order for a subtype to evoive; it is easy to dismiss the occasional occurrence of contrary evidence.

A number of subtypes have been identified within each gender - categories for wornen include bimbo, . sexy woman. athlete and businesdcareer woman

(Deaux, Winton, Crowley, & Lewis, 1985; Eckes. 1994; Noseworthy & Lott, 1984). The last one is of partiçular interest for the purposes of this dissertation. Many of the traits associated with this subtype (e.g.. ambitious. hard working. independent) are inconsistent with the more generic female stereotype, which most closely resembles the sexy wornan or housewife subtype (Fiske, 1998). In addition to possessing traits that connote cornmitment to work, a career woman is probably perceived as likely to be childless. It has been frequently observed that women who want to break through the "" must sacrifice a great deal in order to achieve that goal - having a family is ofien one of those things (Burke & McKeen. 1993: Cooney & Uhlenberg, 199 1 : Parasuraman &

Greenhaus, 1993). Studies 2 and 3 explore whether women who are childless and inrend to remain so are judged any more likely to be hired than women who are family-oriented.

Study 2 also checks whether such women are better cornpensated than women with familial aspirations or responsibilities.

Theoretical approaches to stereot~ping

Just as there are different kinds of gender stereotypes. there are different approaches to the study of stereotypes. The three main theoretical perspectives that underlie the investigation of stereotypes are motivational. sociocultural, and cognitive

(Deaux & Kite. 1993). The motivational framework asserts that people exhibit stereotypical and prejudicial thinking because it serves to bolster one's persona1 identity

(e.g., via enhanced self-esteem). The sociocultural stance is that stereotypes are shared systems and that the personal beiiefs held by an individual reflect the outlook. The cognitive approach emphasizes the fact that human beings' information- processing capacities are limited - gender stereotypes are construed as merely one manifestation of a more generd tendency to categorize the huge volumes of information encountered on a daily basis in order to render it manageable and rneaningfid. Theorking

in this tradition has evolved such that the stereotype went from being conceptualized as a

largely inaccurate and negative method of classifying people (AUport, 1954; Lippman.

1922) to being appreciated as a usefbl tool that sometimes does and sometimes does not

lead to faulty conclusions (Tajfel, 1969).

The majority of psychological research on gender stereotyping has been

cognitively oriented @eaux & Kite. 1993; Unger & Saundra, 1993). although the

socioculturai notion that stereotypes are shared by individual rnembers of a social group is

implicit in rnost investigations. It is not surprising that a great deal of research effort is

expended on identifjmg the cognitive underpinnings of stereotyping; ultimateiy. any

intervention that is aimed at recti@ing discriminatory behaviour and/or changing cultural

belief systems needs to have some effect on individuals' cognitions.

Cognitive research indicates that the activation of one's stereotypes can occur

almost automatically upon encountering information that relates to a stereotyped group

(see Fiske, 1998 for a review). Activation is virtually guaranteed if the content of the

stereotype is perceived to be relevant to the situation at hand (Gilbert & Hixon, 199 1).

According to Unger and Saundra ( 1993). the highly gender-segregated nature of the

workplace means that the issue of gender lies just below the surface in vinually al1

employment contexts. As a result, it seems plausible to suggest that gender is routinely

relevant in employers' decisions about access (hiring) and value (saiary).

Still, the possession and activation of stereotypes does not necessarily lead to their

use in (Devine, 1989). In social situations. people cm successfully

avoid amving at stereotyped impressions by paying attention to and processing information about the unique individual with whom they are interacting (Fiske. 1989.

1998). But reliance on readily available stereotypes in impression formation likely under circumstances where there is limited information on which to base expectations for the future (Heilman, Manell, & Simon, 1988; Fiske & Taylor. 199 1). Hiring involves generating expectations about an aspinng employee's future performance on the basis of a relatively brief encounter, limiting the amount of information that can be obtained. thus it stands to reason that the potential for gender stereotypes to enter into this process is quite high (Gerhart, 1990).

This investigation can be characterized as having a strong sociocultural flavour. as it is primarily concemed with making explicit how culturally shared beliefs about women and paid work find expression in individual participants' views and behaviours in hinng and wage-setting scenarios. There are cognitive elements as well. For example. one assumption underlying the investigation is that the content of the gender stereotypes explored here is conceived by participants as being relevant to hinng and wage setting and as such those stereotypes wili be activated fairly automatically. In addition. the findings

(principdly those of Studies 2 and 3) may have implications for our understanding of the circumstances that facilitate employers' acting upon these activated stereotypes. particularly in hinng situations.

Macro-level

Although the current research focuses on how the individual-level process of stereotyping contributes to gender-linked workplace inequities. a more complete understanding of women's relatively lower position in the workforce requires a consideration of the macro-level theoretical frameworks from which the issue of gendered pay differentiais has been approached. What follows is a selective oveMew of theories that social scientists frorn other disciplines have drawn upon in order to explain - and in some cases, to justie - gender segregation and the gender gap in wages. Where applicable, linkages to stereotyping will be pointed out.

Economic theories

Human capital theory, an econornic approach, is the most fiequently used framework for the exploration of the gender gap in wages (Marini & Fan, 1997). Human capital theory States that certain worker characteristics (e.g.. education, work experience. on the job training, also known as "human capital factors") are linked with higher wages.

Both gender segregation and the wage gap are seen as resulting From fi-eely made. rational choices by female workers to invest less in obtaining ernployment-relevant experience and skills than in family life. andor to select jobs that can provide rewards early in the career trajectory (i.e., in the years pnor to childbearing) and have more flexibility with respect to scheduling than those held by men (Becker. 1980; Filer. 1985; Fuchs. 1988; Polachek,

1978). According to human capital theorists, these choices ensure that women find themselves in jobs that pay less over the course of a lifetime and are occupied predominantly by other women.

Theonsts working from this stance acknowledge that in the course of pursuing maximum profits. which constitutes most employers' main goal, employers mq make choices that lead to segregation and lower pay for women. To wit, employers recognize that rnany female workers will withdraw or reduce their involvement in the labour force at some point in their careers (Marshall, 1994; Moen. 1985). Uncenainty about whether any given female employee will eventually request a leave of absence (or perhaps leave her job entirely) makes it rational and defensible from an employer's perspective to choose either to rehin from hiring women or to slot women into lower paying jobs which entail lower turnover costs. An employer who exercises either of these choices is engaging in

"statistical discrimination" (Smith, 1987).

Human capital theorists do not acknowledge that employers may deny women access to jobs for reasons other than reducing turnover costs. For example. they argue that ernployers would not exclude women simply because they or other employees (who are likely to be men) do not want or like to have wornen in the workplace, a phenornenon known as "taste" discrimination (England. 1992). Economists daim that whatever employers' tastes may be. they are unlikely to lead to sexist hiring practices in a cornpetitive market because female workers tend to have lower wages and excluding wornen from participation in the labour force would actually prevent rnaximization of profits. Proponents of this theory do acknowledge that sexisrn has any role to play in reducing women's pay rates; they contend that wages are determined solely by market forces (Jacobs & Steinberg, 1995). In essence. economists tend to explain segregation as resulting largely From women's choices (Le.. "supply" side factors) and the wage gap as the natural outgrowth of this self-segregation (Jacobs, 1995). There is some acknowledgement that employers' discrimination (i.e., "demand" side factors) could contribute to the maintenance of segregation and the wage gap by denying women access to certain jobs. but there is no admission that employers devalue the work done by women.

The studies conducted for this dissertation test some of the daims made by economists. Specifically, Study 1 is designed to examine whether identical jobs that are in stereotypical "male" and "female" domains in fact differentially valued, contributing to gendered wage differences. Studies 2 and 3 are aimed at showing how the provision of information about a prospective female ernployee's childbeaing intentions influences hiring decisions about that woman. The results are expected to show that statistical discrimination is not merely a theoretical possibility but is in fact highly likely when women apply for the same jobs that men do. such that women and men have differential access to jobs.

Sociolo~icaltheories

While sociologists customarily analyze the same data that econornists examine. their theories are broader in scope than economic ones because they also tend to consider the impact of organizationd. cultural and political factors (Jacobs & Steinberg, 1995).

For instance. while human capital theonsts' contention that some women make educational and occupational choices based on their prioritizing of family rather than workplace achievements has merit (Eccles. 1994). the degree to which these choices are influenced by firmiy entrenched social noms tends to be overlooked by economists. In fact. research suggests that despite professing attitudes regarding the desirability of sharing domestic responsibilities equally. men are resistant to increasing their involvement in the home. particularly in the area of childcare. in order to match women's increased involvement in the paid-labour force (for reviews. see Crosby. Farrell. & Carneron. 1994.

Crosby & Jaskar, 1993; Thompson & Walker. 1989). Thus the choices that women make are perhaps less "free" than econornists imagine.

Three schools of thought drive socioIogica1 research on gendered pay differentials: status anainment theory. the stmcniralist school. and the occupationalist school (Auster. 1989). It should be noted that these theories differ with respect to whether they assume

that men and women can compete on equal footing for the same jobs or whether there

exist separate labour markets for men and women. each comprising distinct jobs and

ladders of progression (Moul, 1999). Status attainment theory would fa11 into the first

category whereas the structuralist and occupationalist approaches are in the latter group.

However, status attainment theory will not be described in detail here because it posits

many of the same arguments as human capital theory. which has already been described.

The structuralist school looks for an explanation for gendered wage differences in

organizational factors. ExempliQing this approach, Baron and Newman ( 1990) discuss

how variables such as unionization, the length of time that a job has existed, whether the job's parameters are well or vaguely defined. and whether there is a sole incumbent or

many people occupying the same position. contnbute to relatively higher or lower pay

rates. Researchers working from this perspective have ascenained that many of the jobs

occupied by women are characterized by lower starting salaries than those held by men

(Baron & Newman, 1990) and fewer opponunities for advancement or raises (Jacobs.

1995). These structural factors likely influence how employees and employers corne to

think about work. For instance. they might help explain the lower salary expectations

observed on the part of female workers (Brief, Rose. & Aldag, 1977; Lefkowitz. 1994;

Markham, South, Bonjean, & Corder. 1985: Stevens. Bavetta, & Gist, 1993) and

employers' tendency to evaluate male and female jobs differently (Auster. 1989). In other

words, the structure of the workplace is seen as impacting on psychological phenomena.

The occupationalist approach is founded upon the assumption that cultural sexism

exists, and that it pemeates employment settings. The idea that women's activities are less valued than those of men in contemporary society is seen as central to explaining why women's jobs and occupations are accorded less prestige4and are paid less than men's

(Croxton, van Rensselaer, Dutton, & Ellis, 1989; Glick et al.. 1995; Unger & Crawford.

1992). Tomaskovic-Devey's (1993, 1995) theorizing on the phenomena of status composition and status closure is illustrative of this perspective. He argues that jobs occupied by women tend to be devaiued (status composition) because wornen have lower status in employment realrns as well as more generally. In addition. Tomaskovic-Devey claims that employers conspire with advantaged employees to maintain their privilege by preventing women from gaining access to better jobs (status closure/discrimination).

Tougas et al. (1995) would appear to agree with this position. arguing that considerations of collective interest underlie both "traditionai" sexisrn and neosexism. In a related vein,

Baron and Pfeffer (1 994) assert that wage gaps and gender segregation persist because the highly paid in-group (Le., white men) do not want to start competing with a group (i.e.. women) that they never had to compete with before. NI of these conceptualizations are rerniniscent of motivationally-based theorizing about stereotyping in that they posit that individuals' use of gender stereotypes brings personai gain. In this case however. the gain is instrumental (i.e., continued access to good jobs) rather than psychological (i.e.. increased self-esteem).

Although status closure and statistical discrimination can bring about the same outcornes (Le., women are denied certain jobs), there is an important distinction to be

' It should be notai that over tirne. jobs or occupations cm cllange from being '.male" to king -Temale." ft has knsuggested that an occupation's prestige is already on the decline before women are aIlowed access to it (Reskin & Roos. IWO: Strober. i Y 84). Prestige then contini~esto decrease as more women enter the occupatiodjob due to cultural devaiuing of the work done by women (Tornaskovic-Devey. 1995). made between the two concepts. In the former case, the aim of the employer is to retain advantage for in-group members whereas in the latter case, the goal is simply to maintain profits at an acceptable level by excluding women whose long-term availability for work is questionable. Study 2 may shed some light on which of these might underlie hiring decisions, as explained below.

Of the sociological perspectives descnbed. this dissertation has the most in common with an occupationalist approach. Study 1 attempts to show, consistent with

Tomaskovic-Devey's status composition formulation. that stereotypically female jobs are devdued simply because they are done by women. Studies 2 and 3 check whether participants acting in the role of ernployers will be more likely to deny &lwomen access to jobs. as Tomaskovic-Devey claims occurs in status closure, or whether they are more likely to exclude oniy those women with family aspirations, which would support econornists' claims that any differential access to jobs by women is based on "rational" considerations relating to their continued availability for work in the long nin.

Description of Studies

The current investigation assumes that culturdly shared gender stereotypes facilitate sexist behaviour on the part of employers, which in mm contributes to the persistence of the gender wage gap. The specific stereotypes examined relate to segregation of the workforce and noms dictating that women assume more responsibility for chiidcare than men do: these two phenomena were chosen for study because they have repeatedly been cited as major contributors to the gender wage gap (see Armstrong &

Armstrong, 1994; England 1992; Joshi, 199 1 ; Shapiro & Stelcner, 1987). The studies conducted show that activation of occupational stereotypes about the gender composition of jobs (Study 1) or stereotypes about how women's family responsibilities impact on their jobs (Studies 2 and 3). leads individuals to penalize women in tems of hiring (access) and wage setting (value). Ail three studies have points of contact with a variety of theoretical fiarneworks, as outlined in the preceding sections.

This dissertation avoids pitfails associated with early work on sexism, some of which utilized designs involving ody the manipulation of target gender followed by of any differences in ratings of the targets to gender-based discrimination - a demonstration of circular reasoning at its finest (Martinko & Gardner. 1983). In al1 three studies reported here, the gender manipulation is but one among others. This strategy allows for greater confidence that significant effects involving gender. whether they be main effects or interactions. actually reflect sexism rather than the tautological reasoninç that Martinko and Gardner warn against. To illustrate. Study 1 tests whether a job's gender-type is a significant consideration under three different conditions: low, medium or high income level. Studies 2 and 3 examine whether a job applicant's gender interacts with familial responsibilities, contributing to less favourable evaluations of women who have or want children.

More specifically. Study 1 focuses on gender segregation and its implications for the salaries assigned to rnde- and female-typed jobs. This study is primarily concerned with the question of whether women's work is valued as highly as that of men.

Experimental methods and a between-subjects design were used to examine whether participants would assign different pay to three types of jobs wherein the actual responsibilities and duties carried out by men and women were the sarne, but the job was situated in either a traditionally masculine or ferninine domain. The discovery of pay differentials between such jobs would suggest that gender-based discrimination. arising fiom occupational stereotyping and the devaiuation of the work typically done by women. influences salary allocation.

Study 2 is concemed with stereotypes about women's familial responsibilities and their impact on hiring and salary decisions, as well as judgments about the rnerits ofjob applicants. In other words, this study addresses the issue of women's access to jobs as well as the issue of whether work done by women is equally valued. Expenmental methods were used once again, but this time only one job was considered - that of assistant bank manager. Gender of applicant and candidates' family-reiated intentions were each manipulated creating eight possible combinations; male and female participants

were exposed to only one of these eight conditions. The impact of the gender of

participant, gender of target and family status independent variables, alone as well as in concert, was measured with respect to perceptions of candidate suitability, turnover potential and ability to "fit in," as well as hiring (access) and salary ailotment (value) decisions. A finding of less favourable ratings on these dependent variables for female candidates who indicated that they would be taking on family responsibilities would illustrate that stereotypes about women's domestic vs. career commitments contribute to

sexist treatment. This study also checks whether presenting information associated with a

career-onented female subtype leads to a more favourable pattern of results.

ui Snidy 3, the question of whether women with family responsibilities are

perceived as less desirable employees than childless women andor men is approached in a different way - via surveying a non-student sample and using a mixed design. This study

focuses exclusively on the issue of differential access to jobs by men and women. More precisely, it checks whether stereotypes about wornen's orientation to family women's access to jobs. By polling a community sample. it was hoped that it wouid be possible to overcome one of the limitations that fiequently plagues psychologicai research: the reliance on relatively homogenous sarnples of introductory psychology students. the generalizability of whose responses is often assumed but rarely demonstrated. Once again. the job considered by respondents was that of assistant bank manager. As in Study 2, gender of applicant was a between-subjects variable, however this tirne. respondents rated the male or female applicant under nine different conditions (within-subjects component).

Briefly. participants were asked to indicate whether a prospective employee's family status. living arrangements and comection to the community were advantageous or disadvantageous when applying for a job. The living arrangements and community comection conditions were included primarily to provide standards against which to compare the perceived impact of familial circumstances.

Two 6nal points about the design of the studies are in order before embarking upon presentation of the studies themselves. Namely. Studies 1 and 2 were conducted with student samples and al1 three studies involved role-playing of employer behaviour.

Neither of these design features are seen as threats to the validity of the findings. There is a long history of research involving employment situation simulations, usually conducted with undergraduate panicipants, and at least with respect to hiring decisions, meta- anaiysis reveals that there is little difference in the results obtained with student and professional samples (Olian, Schwab. & Haberfeld. 1988). Nevertheless, steps were taken to increase confidence that the results obtained in the curent investigation refleaed as accurately as possible the perceptions of individuals who influence hiring and pay decisions in real-We. By sheer good fortune. the community sample surveyed in Study 3 included a sufficient nurnber of individuals who worked as managers to test the hypotheses advanced in this investigation; the responses of this specialized sub-sarnple were examined in isolation and contrasted with those of the remaining respondents in an effort to ascenain comparability. Study 1

This study addresses economists' daim that sexism has linle to do with the gendered pay differentials that accompany segregation. Since "male" and "fernale" jobs

can difEer on task dimensions as well as gender composition, ascertaining whether some

portion of the wage gap is attributable to sexism rather than to dissimilar job duties

requires that the male and femaie jobs considered consist of identical content. Ifjobs in

ferninine domains are still devaiued under these circumstances. it would suggest that

sexism contnbutes to maintenance of the gender wage gap.

It has already been established that in actual work settings. male-dominated and

fernale-dorninated positions involving comparable tasks have frequently had disparate rates

of pay, customady to men's advantage (England. 1992; Ruble, Cohen, & Ruble. 1984.

Tomaskovic-Devey, 1995). Sociologicai research indicates that when men and women are

segregated From each other. it is easier for employers to implement unequal pay schernes.

The mechanism by which this can occur within organizations is through a proliferation of job titles for jobs that have sirnilar tasks (Martin, 1992; Tomaskovic-Devey. 1995). Once

categorized under different titles. men and women avail themselves of different standards

of comparison in order to judge the adequacy of the pay that they receive - thus women

end up cornparhg themselves to other women working at the same job. If their pay is as

good as that received by other women. then they regard the pay as fair. in spite of the fact

that it is generally lower than men's5 (Bylsma & Major, 1992, 1994; Byisma Major. &

' There is aiso evidence that in some cirtumst;uices. the stmdard of comparison could be personal instead of social. In other words. nther than refemng to similar othen. wornen use their past pay to gauge what they deserve to be paid in subsequent situations (Desmamis & Curtis. 1997). Nevertheles. both the social and personai standard are infiuenced by insti tu tional sexism. Cozzarelli, 1995). Both female employees and their ernployers then start perceiving the existing workplace order as "the way it should bey'such that gendered wage differentiais in segregated jobs are perpetuated. By ascertaining whether people tend to devalue jobs in ferninine domains, this study speaks to the question of whether it is reasonable to expect that the trend toward underpaying women working at female-typed jobs will continue.

Participants in this study were randornly assigned to one of six possible Gender of

Job x Income of Job combinations. The income categories were low (department store clerk), medium (teacher) and high (magazine editor). The inclusion of income level is important in that it allows one to see whether gender bias, if it exists, operates in a consistent marner across socio-economic levels. Job titles and descriptions were taken from the National Occupation Classification (1 992) and were identical for the "male" and

"female" jobs within each income level. As such. the men's and women's jobs were exactly the same with respect to educational, ski11 and responsibility requiremenrs and work conditions. and would have similar turnover costs. Only the labels that appeared next to the job titles were varied in an effort to reflect traditionally masculine and ferninine domains: hardware vs. china/crystal for the department store clerk job, industrial vs. home econornics for the teacher and automotive vs. gourmet food for the magazine editor.

The dependent variable in the current study is the annual salary that participants assign to a position. It appears that psychological explorations of the gender wage gap rarely use salary allotment as a dependent variable despite its seerningly obvious utility as an indicator of behavioural intentions and more generally as a measure of worth. A few studies were Iocated (e.g., Glick, 199 1; Momson, Bell, Momson, Murray, & O'Connor,

1994). however, methodological Limitations such as demand characteristics and ambiguous operationalkation of the sdary measure were identified, leaving the door open for Further investigation of this variable.

One possible explanation for why safary appraisals are infiequently sought could be t hat researchers doubt people's ability to provide accu rate estimates. Generally speaking, people are reluctant to discuss openly their personal income; this reticence rnay render many unclear about how much compensation is associated with different types of jobs. As a result, estimates of salary are likely to exhibit a fair degree of variability. However. as long as this variability does not differ greatly in magnitude across experimental cells

(Tabachnick & Fidell, 1989). any average differences uncovered arnong jobs are meaningful. Thus it is rnaintained that participants' allotment of lower pay to female- typed jobs in this study would indicate that jobs located in a ferninine domain m devalued simply because they are done by women.

The dependent variable that most psychological investigations of the gender wage gap have used instead of salary is the number of job evaluation points assigned to a position. Job evaluation systems assess the requirements of various jobs against four main critena: responsibilities entailed, amount of education, skills required to do the job. and difficulty of working conditions. The four criteria act as "compensatory factors" in that higher ratings ofjobs on these factors result in higher pay (Steinberg, 1995). Within each of these four dimensions are several subscales on which the job is rated; the number of points that the job receives on each subscale is then summed to give a total job evaluation score. It should be noted that comparable worth poiicies/legisiation, one of three initiatives undertaken by some governments and corporations in an effort to equalize men's and women's pay. relies on job evaluation systems to rate the tasks involved in male- and female-dominated jobs. Male and female jobs that receive the same number of points are supposed to receive equivalent pay. In view of the fact that job evaluation points are intimately linked with salary, the findings of studies that utilize job evaluation points as a measure ofjob worth are relevant to the present inquiry - a selective overview follows.

Some studies have investigated how gender impacts on perceived job worth by manipulating the gender of the job incumbent in a particular position and measuring how provision of this information affects the number of job points assigned. In many cases. such a manipulation results in no significant gender of incumbent effects (Grams &

Schwab, 1985; McArthur & Obrant. 1986; McShane. 1990; Schwab & Grams, 1985). although the jobs occupied by men were nevenheless rated as having "higher monetary wonh" in one study (McArthur & Obrant. 1986). Unfortunately, demand characteristics make the results of studies that explicitly refer to gender in this way difficult to interpret. tncumbent gender is never included in the job descriptions that are utilized in real-life work seaings - it may be that specimg it renders the hypotheses of these types of investigations too obvious. such that individuals may be inclined to provide "politically correct" responses.

In their discussion of the limitations of job evaluation research, Mount and Ellis

(1989) suggest that extemal validity would be rnaximized by haMng people appraise jobs that &I actuaily d8er with respect to gender composition rather than artificially manipulating the gender variable. Of the studies that are designed so that gender of the job-holder can be inferred from the position's Me. many have required that participants allocate job evaluation points to male and fernale jobs that were quite different with respect to content (e.g., nurse versus electrician). While it is telling that women's jobs are assigned fewer job evaluation points in these studies. (see Homsby, Benson, & Smith,

1987; Mahoney & Blake, 1987; Rynes. Weber. & Milkovich. 1989). the case for the

presence of continued sexism in the workplace is best illustrated by asking respondents to make direct cornparisons ofjobs on gender-linked job titles while job descriptions are held

constant. This level of experimental control is seen in only a few studies.

For instance, Mount and EUis ( 1987) asked comparable worth specialists to rate

male and female jobs at two different income levels (nurse aide vs. otderly and YMCA

director vs. YWCA director). Counter to expectations, the female jobs actually received

marginally higher ratings. However. this unexpected finding might be attributable to the

investigators' utilkation of a within-subjects design on the gender of job independent

variable. wherein each person rated a male job for one of the income levels and a female job for the other. As with studies that speciQ gender of incumbent, this approach

probably made the hypotheses of the study obvious to the participants. Moreover. a

sample comprised of compensation specialists is likely to be exquisitely sensitive to issues

of pay discrimination and they may have overcompensated women in this experimental

task.

Naughton ( 1988) improved on Mount and Ellis' work by using a between-subjects

manipulation on the gender ofjob variable. He also recruited participants who had the

potential to attain manageriai positions in the future (Le., business students); this choice of

sample likely increased the extemal validity of the study given that in most workplaces.

managers are more likely to make salary decisions than are compensation specialists. He

provided his pmicipants with either a male or female job title at two income levels (Le., executive secretary vs. assistant to the vice president for administration and associate of nursing vs. associate professor of engineering). Once again. the job description remained constant within each income level. This time, male jobs received

better ratings on various sub-components of the job evaluation scheme as well as higher total ratings.

Study 1 updates and expands upon the findings of research conducted in the rnid- to late 1980's. when investigations ofjob evaluation were most common. The fact that

there has been linle research on job evaluation systems and how they might be used to

decrease the wage gap in the last decade is perhaps not surprising, as the prevailing

political climate in the last few years has emphasized fiscal restraint. In Canada, existing

federal and provincial legislation appear to endorse pay equity and comparable worth in

the abstract. but the relevant laws do not encourage employers to be proactive in pursuing the goal of Fair pay for ail workers (McDermott, 1996). Rather. the legislation requires

that workers who feel that they are underpaid. usually women. must bring a complaint

against their employers to a human nghts tribunal. Such a course of action generally

necessitates strong union support for the wfiers' claim and requires that employees be

willing to wait out a potentiaily lengthy adversanal process (Brenner, 1987) - indeed

Canadian cases involving Bell Canada workers and federal public servants have seen many years of litigation.

Employers' reluctance to incur the costs of btinging salaries in predominantly

fernale jobs up to the level dready enjoyed by those holding "male" jobs is evidenced by

their readiness to appeal tribunal decisions that favour female cornplainants (Canadian

Human Rights Commission, 1998b). Ironically. the federal govemment's numerous attempts to avoid complying with rulings awarding female public servants equal pay for their work (LeBlanc, 1999) is in direct opposition to their professed stance on the issue of gender equality (Status of Women Canada, 1995). The negative press that occurs every time a human rights tribunal or court niles in favour of underpaid workers suggests that many people share employers' views that fair pay policies are not "affordable" (see

Simpson 1999; Wente, 1997). However, a recent Environics poll indicates that the

Canadian public strongly supports equal pay policies. which leads one to consider the possibility that unfavourable press coverage may be attributable to a vocal minonty and is nor an accurate reflection of society's attitudes (Baine, 1999).

Regardless of whether comparable wonh policies are wady received by employers andor the general public. the question of whether the discrepant pay associated with male- and female-typed jobs reflects gender bias ments attention. This study seeks to ascenain whether the gender of the typical job incumbent is related to a job's wonh by examining how participants perceive the worth of positions that are alike with respect to tasks, but are located in a stereotypicaliy masculine or ferninine domain. The means by which "worth" is measured is through the salaries that participants assign. The findings can help us determine whether devaluation of wornen's jobs is a plausible explanation for the differentid pay that occurs in gender-segregated jobs.

Hyotheses for Prima? Analvses

The primary analyses involved assessing the impact that gender of the target job, income level ofjob and gender of participant had on the salaries assigned to six different jobs. A main effect for income level of the job being rated was expected to explain the lion's share of the vanance in salary. The gender of the iarget job was aiso expected to explain unique variance in assigned salary above and beyond that explained by income level, yet it was difficult to know whether the magnitude of the difference between "male" and "female" job salaries would be consistent across income level, or whether the two main effects just described would be qualified by a significant interaction between gender of target job and income level. As a result, no specific predictions were made regarding an interaction effect.

It was also unclear whether gender of participant would yield any significant effects. On the one hand, it seemed plausible that men and women would have shared ideas about how much rnoney should be assigned to male and femaie jobs such that no gender of rater eEect would emerge (Glick. 199 1; Naughton, 1988). On the other hand. previous research designed to investigate gender differences in deservingness shows that wornen will pay themselves less than men in experimental tasks (Bylsma & Major. 1992:

Kahn, Nelson, & Gaedden. 1980; Major. McFarlin, & Gagnon. 1984). By extension. when paying others. female participants might pay workers less than males do (Jackson &

Grabski, 1988). hother uncenainty was whether women would consistently assign lower pay than men did to al1 jobs. resulting in a main effect for participant gender, or whether the gender type and income level of the job would also make a diference to raters. resulting in some type of two-way or even a three-way interaction involving gender of participant. Again. no specific predictions were made on this point.

Met hod

Power Calculations to Determine Sample Size

It was difficult to obtain an estirnate of expected efEect size Ekom the literature given that no comparable studies using dollar amount of salary as a dependent variable could be located. but it was anticipated that the effects for the primary analyses were most likely to be either "small" or "medium" in size. Unfortunately, the sample size required to find a smail effect, even at Iow power, was prohibitively large (e-g., a sample size of 600 is required to find a smdl effect at power of .3). Therefore, the decision was made to obtain a sample size that would yield adequate power for a medium-sized effect. An a priori power analysis, conducted using the GPOWER statistical package, indicated that a sample of 240 furnished power of approximately .7 if a medium-sized effect was being sought.

Partici~ants

The total sample consisted of 260 students. Most of the participants in this study

(n = 22 1, 85%) were recniited through the introductory psychology pool and received course credit for their participation. The remainder (n = 39. 15%) were recmited corn undergraduate courses in the sciences and social sciences. Preliminary analyses indicated no group differences on the dependent variable of interest that could be traced to differences in recruitment strategy. Information about the charactenstics of the current sarnple appears in Table 1.

Procedure

The sign-up sheets for this study informed panicipants that the purpose of the investigation was to determine the factors that people considered most imponant when evaluating jobs. They were told that they would be asked to rate jobs on numerous dimensions, as well as to provide salary estimates. The University of Guelph has a strong program in IndustriaVOrganizational Psychology, thus it seemed reasonabie to assume that participants wouid that the study's focus tmly was on the job evaluation process. Table 1 Background Information for Studv 1 Participants

Gender Femaie MaIe

Current Employment Status Ernployed Not Employed

Major Sciences Psychology Applied Sciences Sociai Sciences Arts B.Comm./Marketing

Age M [sd] 19.98 [2.05]

Note: Not al1 respondents provided demographic information. percentages reflect valid responses. Furthemore, the between-subjects manipulation of job gender utilized ensured that our interest in gendered pay differentials was not transparent.

Students took part in the study either individually or in small groups; al1 of the participants were tested by a female researcher. Participants were asked to read one of the six job desct-iptions and rate the job on each of nine scales. These scales tapped into the four major compensatory factors that influence pay rates for jobs and fom the basis ofjob evaiuation: required skills, education, amount and type of responsibilities associated with the job. and working conditions. Participants were then asked "In your estirnate. what annuai salary should be associated with this job?" Despite the fact that salary estimates were requested in annual dollars eamed. some people ignored this instmction and instead provided an hourly rate. Such responses were transformed into a yearly amount by multiplying by 40 hourdweek and 52 weekdyear.

Mer these tasks were completed. the sheets containing the job description. the participants' responses on the rating scales and salary estimates were collected.

Participants were then instructed to think about the job that they had just assessed and the charactenstics of the typical person who did this job. This strategy of obtaining information about the "typical" job holder has been used in previous research in order to elicit occupational stereotypes (Glick 199 1). On a second questionnaire, they were asked to rate some of the characteristics that this "typical" job holder would have, including age. how long the person had been in the workforce and. most imponantly for our purposes, whether the person was male or female. Participants were aiso provided with space to write down ideas about any other characteristics that they thought the typical job-holder had. if they wanted to do so. A brief questionnaire designed to gather some demographic and work history information corn participants was adrninistered at the same time (copies of these questionnaires appear in Appendix A).

Measures

Annual Salarv

As outlined in the hypotheses section. the primary dependent variable of interest in

Study 1 consisted of the annual salary assiyned to the target job. In large-scale surveys exarnining the gender wage gap arnong full-time workers engaged in a vxiety of occupations. investigators often use the natural log transformation on salary data in order to make it more normal for parametric analyses (e.g.. Marini & Fan, 1997; Wellington.

1994). Unforninately, this transformation tends to diston data to such an extent that differences between men's and women's salary are minirnized (Tomaskovic-Devey. 1995)

Given that the range of dollar values that was expected for these six jobs was likely to be smaller and more normally distnbuted than the range of salaries found among full-time employees working at diverse occupations, and that ANOVA is fairiy robust to violations of normality (Tabachnick & Fidell, 1989). this study used actual dollar values assigned by participants as the dependent variable.

Compensatory Dimensions

The nhe rating scales that tapped into underlying compensatory dimensions were subjected to internai reliabiiity analyses in an effort to determine whether they could be combined to forrn a sinde scale. There were three items related to the educational and training requirements for the job (Le., shortfiong, generaVspeciaiized and easy/difficult), two tapping required skills (i. e., p hysicdmental and simple/complex), two related to responsibilities for materials, equipment and people (Le., slightkonsiderable and specifidbroad) and two assessing working conditions (Le.. undemanding/demanding and variedhepetitive [this last item was reverse-coded]). These nine items consisted of opposing adjective pairs, each of which was rated on a 7-point scale where the first adjective in the pair served as an anchor for " 1" while the second adjective anchored "7 "

Although not labeled as such, "4" was the implicit neutral point. Taken together, these nine items served as the dependent variables for one of the preliminary manipulation checks.

In addition, the results for the sample as a whole indicated that these nine items displayed sufficient intemal consistency to jus@ combining them into a single scale

(Cronbach a = .84). The newly constructed scale ranged fiom a minimum of 9 to a maximum of 63, with higher scores indicating the presence of job-related characteristics that are traditionally associated with higher pay. This total scale was incorporated into the supplemental analyses described in the next section.

Preview of Analyses

Mani~ulationChecks

The first manipulation check involved ascertaining whether the label reflecting the masculine or ferninine domain of the job was perceived by participants in the intended way

(e.g., chindcrystal store clerk = "fernale". etc.). As data collection progressed. it became apparent that a smail minority of students (22 out of 260) were reluctant to assign a gender to the "typicai" job-holder. Furthemore, some disagreed with the researchers' ideas about the gender domain of the taget job. as evidenced by their labeling a job that had been conceived of as fernale gender-typed as male or vice versa. In order to ensure reasonable power for ANOVA the assignment of participants to conditions proceeded in a rnanner that ensured that there were at least 15 participants per ceil who agreed with the stereotyped gender of the target job.

Next, a 3 x 2 x 2 (Income Level of Job x Gender of Participant x Gender of Target

Job) between-subjects design hierarchical MANOVA was conducted on the 9 rating scales

in an effort to detemine how the six jobs differed on these compensatory factors. Of course it was expected that job income level (low. medium, high) wouid yield a significant

main effect; this analysis served primarily as a check on the comparability of the male and female jobs within each incorne category. The decision to use the nine separate items as

the dependent variables rather than conducting an ANOVA on the single total score obtained by combining al1 nine items came from a desire to be able to identify the specific

dimensions where differences lay in the event of a significant overall test. A hierarchical

mode1 was used in order to determine whether gender of target job (as assigned by

participant) explained unique variance beyond that explained by the income level of the job and gender of respondent. A finding of no gender of target job effects would support the

idea that the stereotypically male and female jobs at each income level were viewed by

panicipants as comparable.

Prima- Analyses

To test the hypothesis that stereotypically female jobs would be paid less than male jobs. two 3 x 2 x 7 (Income Levei of Job x Gender of Participant x Gender of Target Job) between-subjects hierarchicai ANOVAs were conducted using salary as the dependent variable. For the first ANOVA the "gender of target job" independent variable consisted of the gender that participants assigned to the job rather than the gender that had been conceived by the researcher as fitting the job. For the second ANOVA. the sarne 3 x 7 x 2 analysis was conducted on the sub-sample that agreed with the researcher's notions about the stereotyped gender of the target job. As in the analysis involving compensatory factors. hierarchical models were used in order to determine whether gender of target job explained unique variance in salary above and beyond that explained by the income level of the job and gender of participant.

Supplementaxy Analyses

In supplementary analyses, correlations were nin in an effort to identiQ additional participant- and target job-related variables that influenced salary allotments. This method of identifjmg covariates was adopted in the interests of maintaining a reasonabie sample size for the analyses. If al1 the potential covariates had simply been entered in an

ANCOVA without doing this preliminary step. the amount of missing data on the covariates would have reduced the sample size by about 40 participants.

Ideally, one runs correlations between the dependent variable and potential covariates for each experimentai ce11 separately so as to ensure that no covariate interacts with any of the independent variables to affect the dependent variable (Tabachnick &

Fidell, 1989). Variables that appear to be consistently correlated with dependent variable are then added as covariates to the existing ANOVA model. Unfortunately, the ceIl sizes here were too srnall to run meaningful correlations for al1 12 experimental cells. The fact that the study's main focus was on the job gender-type and how it acted in concert with job incorne level to influence salary, together with results of the prirnary analyses. led to the decision to run correlations only for the 6 income leveVjob gender combinations. The correiations assessed the impact of three participant background variables and two target job-related variables on salary: participants' self-reported weekly wages at their most recent job, participants' current employment status (employed vs. not), mothers' current engagement in paid work outside the home (yes or no), number of years that participants thought the "typical" target job-holder had spent in the workforce. and total scores on the scale created by combining participants' ratings of the target job on the nine compensatory factors. Variables that yielded substantial correlations with salary across the six experirnental conditions were then included in a 3x2~2ANCOVA.

Results

Mani~ulationChecks

Gender assiened to iobs by ~artici~ants

The gender of the "typical" job-holder. as identified by the respondent. was compared against the researchers' ideas about the stereotyped gender of the target job. A relatively small number of participants (n = 7) failed to indicate the gender of the "typical" job-holder; these people were dropped from tùnher analyses. Of the valid responses, a relatively small minority (6%) said that the typical person could be "both" or "either" sex.

This equivocal response was especially likely to be given for the highest income categoiy-

1 1% of those who were asked to rate the editor gave this response (n = 10) while only 3% asked to rate the clerk (n = 2) and 3% of those asked to rate the teacher position (n=3) gave such a response. To facilitate the analyses. al1 respondents who gave an "either" or

"both" response were also dropped fiom further analyses.

Of the remaining participants (n=238), the vast majority (82%) gave answers that agreed with the researcher's preconceptions regarding the gender domain of the target jobs. The other 18% of respondents reversed the job gender (e.g.. they labeled a job that

had been conceived of as female gender-typed as male or vice versa). Table 2 illustrates

that reversals were fairly ifiequent for the department store clerk target jobs whereas

they were more cummon for the teacher and editor target jobs. Although the pattern of

agreement appeared tu differ for male and female panicipants. a senes of two-sample z-

tests reveaied only one statistically significant difference (see Table 2). Specifically. male

respondents were more likely than female respondenis to indicate that the "typical" editor

of an automotive magazine was male, reflecting greater agreement with the stereotyped

gender of that job (93.75% agreement for men vs. 60% agreement for women; -2.93.

p<.Ol). The fact that some panicipants made unexpected gender domain assignments to

the target jobs led to the decision to perform the prirnary analyses on the salq dependent

variable twice - once on participants who assigned a gender to the job. regardless of

whether the assignent agreed with the researcher's preconceived notions of the

stereotyped gender of the job. and then once again with the sub-sample that did agree with

the a priori ideas regarding the gender of the target job. Table 2 Gender Assiened to Target Job bv Gender of Respondent

Gender Assigned to Job % Agreeing with Stereotyped z Maie Female Gender of Job Participant Sex Male Femaie Male Female Mate Femaie N N N N % 96 JO-b Clerk: ChinaiCrystal - 1 15 20 100.00 95.24 1 .O2

Clerk: Hardware 16 19 1 2 94.12 90.48 0.42

Teacher: Home Ec. 3 - 15 20 83.33 100.00 1.90

Teacher: Indust. Arts 15 19 3 9 83.33 67.86 1.24

Editor: Food 8 4 16 15 66.67 78.95 0.92 Editor: Automotive 15 15 1 10 93.75 00.00 2.93'.. .e<.O 1

Evaluation ofjobs on compensatory factors

The second manipulation check involved conducting a 3 x 2 x 2 (Income Level of lob x Gender of Participant x Gender of Target Job) between-subjects design MANOVA on the 9 items tapping compensatory factors. The MANOVA sumrnary.table for this analysis appean in Table B-1 of Appendix B whereas the means and standard deviations by job income level are listed in Table 8-2 of Appendix B; means are also depicted in

Figure 1. Although there was a significant main effect for income level of job (E, ls.rn, =

13.97; ec.00 i), none ofthe other main effects or interactions achieved statistical significance. Given that the purpose of this analysis was simply to demonstrate a Lkof effect for gender of job. no follow-up analyses were conducted in an effort to identify which of the Nne items displayed a significant effect for job income level. Figure 1 Mean Ratinps of Corn~ensatorvFactors bv Job: Ail Panicipants

Mental Complex Long Specific Oifficutt Considerable Broad Oemanding Repetrtive

Physical Simple Short General Easy Slight Specific Undemand Varted Compensatory Factors lm Clerk UTeacher H Editor ( Primarv Analvses

Ail Partici~ants

The first hierarchical ANOVA conducted on the salary dependent variable included ail participants (i.e., those who agreed with the researcher's ideas regarding the gender domain of the target job as well as those who did not; n=238). This analysis revealed significant main effects for two of the three independent variables. The ANOVA table for this analysis is presented in Table B-3 of Appendix B whereas means and standard deviations are presented in Table 3 and the main effects are depicted in Figure 2.

Examination of the standard deviations reveais that the variability of the sdary estimates was acceptable across the 4 cells within each of the three income levels. although variability of estimates seemed to increase quite a bit with the incorne level of the job.

Still. at no time did différences between any two cells in the model exceed the 20: 1 ratio deemed permissible by Tabachnick & Fidell ( 1989). making the data suitable for analysis using ANOVA.

Not surprisingly, salaries assigned to the low. medium and high income levels were significantly different from each other (E11226)= 109.72.pC.00 1; $ = .49). Tukey's HSD post-hoc analyses showed that salary assigned to the editor was significantly higher than that assigned to the teacher. which in turn was significantly higher than that assigned to the department store clerk (&iior = $48.070.24. MiaAa= $47.256.25. _nicrcr=

$2 1.9 17.68; p<.Oj). Also. as expected, participants assigned higher salanes to the jobs Table 3 Mean Salarv Estimates for 3 Twes of Jobs bv Gender Assiened to Tareet lob for Male and Female Participants: Al1 Participants

Gender Assigned to Job Male Fernale +O* Main Effect of Participant Sex Male Fernale Male Female Job lncome Level -M -M M M -M (sd) (d) (sd) (4) (sa Job Income Level Store Clerk 23,550 23,250 22,53 5 19,070 21,915 (Lw (5,535) (8,966) (6,556) (4,664) (6,775) n= 16 ne20 n=16 n=22

Teac her 43.61 1 41,133 43.528 41,207 42.256 (Medium) (8,452) (10,986) (11,620) (8,981) (9,789) n= 18 n=15 n= 18 n=39

Magazine Editor 5 1,544 5 1.160 47.47 1 42,934 48,070 (Wh) (13,498) (17,759) (13,369) (16.938) (1 5,737) n=23 n= 19 n=17 n=25

Main Effect of 39,800 Participant Sex ( 15,334)

'~ainEffect of 3 9,652 36,537 Gender of Job ( 16.566) ( 15383) '~<.05,**O p<.OO i Figure 2 Mean Salarv bv Job. Gender of Job and Participant Gender for Al1 Participants

50000 2- m L ;33MxIo 21. P 2ooclo

10000

O

Job lncome Level (pc.001) Jab Gender (pc 05) Participant Sex (n.s.) Main Effects

Figure 3 Mean Salav b~ Job. Gender of Job and Participant Gender for Panicioants who aoree .- - with Stereotvped Job Gender

CI. Tch-Ed. M F M F Job lncome Level (pc.001) Job Gender (p=.052) Main Effects that they perceived to be "male" jobs than to jobs they perceived to be "female" (& =

$39.65 1.71. & = $36.527.3 1; EL,Z26)= 4.29. p<.OS: = -02). No main effect was detected for gender of respondent. nor were there any statisticaily significant interactions.

Participants who aeree with stereotyped gender ofjob

A second 3 x 2 x 2 hierarchical ANOVA was conducted including only the sub- sample that agreed with the stereotyped gender of the target jobs (n = 196). The ANOU summary table for this sub-group appears in Table B-4 of Appendix B while means and standard deviations are presented in Table 4 and main effects are depicted in Figure 3.

This analysis revealed a pattern of results similar to that obtained with the sample as a whole. Once again. salaries assigned to the low. medium and high income levels were significantly different From each other (&L184)= 1 12.80, p<.OO i: = 35). Tukey's HSD showed that as before. the salary assigned to the department store clerk was significantly lower than that assigned to the teacher or the editor and the salaries assigned to the teacher and the editor were significantly different from each other (Mc,&= $2 1.95 5.83.

Miuehsr = $4 1,66 1.54. &dito, = $49.544.26; gc.05). Funhenore. although participants still assigned higher salaries to "male" jobs than to "female" jobs. the effect fell just short of statistical significance (Mn, = $38.46 1.47. & = $35.774.93; E(i.iari= 3.82. p=.052: r12 =

.Oz). As in the earlier analysis. no statistically significant two-way or three-way interactions were found, nor was there a main effect of gender of respondent. Table 4

Mean Salary Estimates for 3 T-yes of Jobs bv Stereotvped Gender of Job for Male and Female Participants: Panici ants who agree with Stereotyped Job Gender

Stereotyped Gender of Target Job

Male Female- - **+ Main Effect of Participant Sex Male Female Male Female Job Incorne Level -M -M -M -M -M

Job Income ~eve~ Store Clerk

Teac her

Magazine Editor

Main Effect of Participant Sex

'Main Effect of Gender of Job Supplementary Analyses

These analyses were conducted both for the sample as a whole as well as for the subsample who agreed with the researcher's ideas about the jobs' stereotyped gender. In an effort to avoid tedious repetition, only the results for the subsample are presented here. although it should be noted that the analyses conducted with the entire sample yielded comparable results.

Of these five potential covanates assessed via correlatioos, two were fairly consistently related to the dependent variable of salary across the 6 experimental conditions: length of time that "typical" job-holder had been in the workforce. and the total scale comprised of the 9 cornpensatory factors (the correlation table appears in Table

B-5 of Appendix B). The remaining three variables yielded some positive and some negative correlations across the six cells. indicating the presence of interactions with the independent variables. These interactions render the variables inappropriate for use as covariates. thus they were not included in the ANCOVA (Tabachnick & Fidell. 1989).

The oumber of years that the "typical" job-holder had spent in the workforce was a significant predictor of salary across 4 of the 6 cells. Workers who were construed as

having more experience were assigned higher salaries (r's ranged from .16 to 48).

Similarly, scores on the scale created by combining the 9 compensatory factors were

significantly predictive of salary assignment in 4 of the 6 cells. Higher ratings on these

compensatory factors led to higher assigned salaries (r's ranged frorn -18 to .60).

Although the zero-order correlations for the workforce and compensatory factor scales

were not always statistically significant given the ce11 sizes. the correlations were in a

positive direction in al1 cells, thus the decision was made to go ahead with the ANCOVA analysis. The sumrnary table for the ANCOVA appears in Table B-6 of Appendix B whereas the means and standard deviations appear in Table B-7 of Appendix B.

When entered as covariates on the first step in the hierarchicd ANCOVA model, both variables were highly significant: number of years in workfbrce (&1.174) = 71.51, pC.00 1; ri' = .29), and compensatory factors (F(1.174) = 164.53. g<.00 1; $ = .49).

Nevertheless. even after inclusion of these variables in the model. the income level of job

(LL17J) = 40.50, pC.00 1; q2 = .32) and the perceived gender domain of the target job

(Ec1.174) = 4.33. ~<.05;q' = .02) were significant. Basically, addition of these variables to the model increased the total amount of variance explained. but the main independent variables of interest retained their predictive capacity. Tukey's HSD tests revealed that the job of editor was still paid more than that of teacher, which in tum was paid more than the store clerk (Nid = $2 1,789.67, = $4 1.862.90, = $49.409.49; pC.05). whereas male jobs were paid more than femaie jobs overall (Mt,= $3 8,464.44,& =

$35.803.76). The same ANCOVA conducted for the overall sample yielded a similar pattern of results.

Discussion

Mani~ulationChecks

The fact that 93% of participants were willing to assign a gender to the "typical" job holder for the job that they rated (rather than responding witb "both" or "either") suggests that people are willing and able to engage in the occupational stereotyping described in pnor research (Beggs & Doolittle. 1993; Glick et al.. 1988). It is impossible to know whether people who provided an equivocal response were attempting to be "politically correct." In the future, it would be interesting to explore the ways in which people who provide the "either" or "both" response differ From those who stereotype.

Of the participants who did assign a gender to the job presented, some categorized the job in an unexpected way. Still. their responses were usually understandable from the perspective of gender stereotyping - these participants may have focused on gender stereotypic traits contained in the job descriptions rather than on the gendered job title.

Consider the jobs t hat received the most "reversed" gender responses, that is, the two dif5erent editor positions. Classifjing the automotive editor job as female may indicate that a participant was responding to the verbal skills component of the job. proficiency in language being traditionally associated with fernales. By the same token. labeling the gourmet food editor job as male could imply that a participant was reacting to the supervisory and decision-making duties connected with the position, these tasks senerally being more common in men's rather than women's jobs (please refer to the questionnaires in Appendix A for full descriptions of the tasks associated with the editor job). The important point is that the jobs that participants identified as female-typed were assigned

less pay than the jobs they identified as male-typed. which suggests that participants'

responses were influenced by sexism.

Primarv & Suoplementa-/ Analyses

The finding that "male" and "female" jobs were not perceived by participants as

significantly different on the 9 job-related compensatory factors but were nevertheless paid

diEerently implies that the same job is viewed as less valuable when it is camed out in a

ferninine domain. This result is illustrative of what has been called "occupationa1

discrimination" (Tomaskovic-Devey, 1995). which seems to fit best under the nibric of old-fashioned sexism. Nevertheless. the discrepancy in the results obtained for the compensatory factor ratings and the sdary allotments is rerniniscent of modem sexism in that the "gender blindness" with which participants rated the jobs' responsibilities. skills, working conditions and educational requirements disappeared when they were asked to evaluate the jobs in a rnonetary way.

Alternative economically based explanations for the gender of job effect were considered but did not seem persuasive. For example. it does not seem plausible to suggest that participants believed there was either an oversupply of workers for the

"female" jobs or an unusually high demand for workers to fil1 the "male" jobs assessed here, such that more pay is required to entice workers into the "male" domains."or does it appear that job gender was confounded with any other factors that couid legitimately impact on wages, like quality of performance or joblorganizational tenure. Even though higher pay may be allotted to workers on the basis of ment or work experience (England.

1992; England & Markowitz, 1996). participants in this study were not provided with any information about the quality of any individual job holders' performance whereas the work experience variable was statistically controlled in supplemental analyses with no effect on the outcome. In shon. it appears that panicipants have engaged in wage discrimination based on the gender-typing of the job.

The fact that the inclusion of covanates in the analytical mode1 did not alter the predictive capacity of job gender bolsters confidence in the conclusion that sexism infiuences perceptions of a job's worth. Thus even if participants do not consider both

é Howmr. it appears that for envy or midde-level positions. ihis course of action is more often proposed as a solution for Iabour shortages in "malen jobs than in "fexnde" jobs (Buttner & Rosen. 1987). jobs within a particuiar income level as strictly equivalent with respect to content (e.g., the hardware store clerk job may connote the need for physical strength or an ability to engage in manual labour that perhaps is not present in the corresponding fernale job of chindcrystal clerk), the ANCOVA statistically "equalized" the jobs on this and other potentially confounding compensatory factors by including the nine item scale as a covariate. Nevertheless, in order to nile out alternatives to the sexism explmation offered here, it might be valuable to inquire about a wider range of compensatory factors than examined in this study andor to ascertain whether the "male" jobs might be viewed by participants as involving additional tasks beyond those delineated in the job description.

While the goal of the study was not to determine how precisely participants' salary estimates would approach , one might wonder how the responses compared to the actual salanes associated with clerk. teacher and editor jobs. In fact, cornparison of the estimates contained in Tables 3 and 4 with recent figures provided by the federal governrnent show that most participants' salary assignments fell well within the range of actual salanes associated with these jobs (i. e.. estimates provided by the entire sample. shown in Table 3. were = $2 1.9 1 7.68. Mluîhn= $42.25625, Mcdifm= $48.07024).

Actual full-time annual salaries for retail sales clerks ranged between $1 1.100 and

$52.800, while the average was $29.900; for secondary school teachers the range was

$30.100 to $6 1.800. with an average of $46.000, and for editors the range was $70.100 to

S61.300. with an average of $39.700 (Human Resources Development Canada, 1998 - note that salaries broken down by gendered domain are not available for these positions).

This similarity neutralizes the potential critique that the current study's value is comprornised by students' unfamiliarity with the task of assigning annual salaries to jobs. Depending on whether one Iooks at the results of the analyses for the entire sample or for the subsarnple tested, the main effect for job gender shows that on average. female- typed jobs were paid between $2.600 and $3,100 less than male-typed jobs. The interaction between income level and gender of target job was significant in either analysis, suggesting that the devaiuing of women's work occurs regardless of income level. Nevertheless. the smaller discrepancy in pay assigned to the male versus female teacher positions ($1.000 or less) relative to the gender differences observed for the other

IWO positions (about $3.000 for the clerk and $5.000 for the editor) merits comment. At the time the study was conducted, teachers were receiving a great deal of press in Ontario: there had been provincial cuts in tùnding for education. numerous schools were threatened with cIosure, a teachers' strike occurred. and there was much discussion in the about how difficult it would be for teachers to continue to do their jobs effectively in lighr of the goverment-irnposed changes to the education system. It may be that the intense media coverage surrounding these issues resulted in a situation where the gender manipulation presented in the study was less salient to participants than the plight of teachers generally, resulting in similar evaluations of the job regardless of its gender domain.

The lack of significant interaction effects involving the gender of participant variable implies that men and women process gender-linked job information in similar ways. a finding that is consistent with prior research examining the content of occupational stereotypes (Glick et al.. 1995). The failure to find a main effect for participant gender suggests that men and women occupying power positions in the work world would be equally tikely to underpay women. A more cautious interpretation is that male and female respondents are equally likely to enter the ranks of permanent workers with an understanding of the nom that "women's" jobs do and/or should get paid less than comparable "men's" jobs.

It should be noted that aithough male and female participants demonstrate similarity of responses, the underlying their responses could conceivably differ.

Glick and Fiske (1996) point out that for women. stereotyping and discrimination against females is cuituraliy lemed whereas for men, sexist acts could also reflect a desire to maintain group advantage, whether conscious or unconscious. Clearly, it is vinually impossible to design a study that could ascenain which of these motives drive participants' behaviour and it is acknowledged that this study cannot speak to these issues.

Limitations

The possibility that participants' responses might have been infiuenced by the researcher's gender (Le.. female) was considered. in light of pnor research indicating that such effects can occur (see Unger & Saundra 1993 for a review). Increased of gender during the rating task could have a variety of consequences. For male participants. the effect of dealing with a female researcher could have been to reduce the probability of uncovering sexism since men may be inclined to give higher ratings to stimulus persons of the same sex as the expenmenter (Etaugh. Houtler. & Ptasnik 1988). On the other hand, men have dso been show to rate women unfavourably when in groups with other men

(Starer & Denrnark, 1974). Because there were no controls imposed on group composition in this investigation, it sometirnes happened that a group was comprised pnmarily or exclusively of men; this occurrence may have increased the Iikelihood of finding sexism. More generaily, there could have been a tendency for participants to behave in socially desirable andfor cornpliant ways regardless of their gender or any reaction they might have had to the researcher's gender. In other words. respondents may have attempted to appear non-discriminatory by paying a fernale-typed job more than they would otherwise, or to acquiesce to perceived exyenmenter demand by penalizing female- typed jobs. It is aiso possible that despite the carefùl wording of the primary question of interest (i-e., "what annual salary should be associated with this job?"). participants' responses reflect their cognizance of actual pay differences between "male" and "female" jobs rather than their opinions regarding what the jobs "should" be paid.

In the end, it appears that there were just as many processes pulling for underestimation of the role of sexism in wage-setting for gender segregated jobs as for overestimation. No would have permitted us to establish with cenainty which if any, of these processes infiuenced participants' responses in this study. but the experimental design utilized at lest allowed us to be reasonably confident that these opposing forces were balanced out across conditions and participants. leaving us with a final picture that is fairly accurate.

Conclusions

Considered in isolation, participants' responses on the salary allotment question appear to be most consistent with old-fashioned sexism in that jobs located in female domains were devalued relative to identical jobs in male domains. However. the discrepancy between participants' ratings of the jobs on cornpensatory factors and the actual compensation (value) they assigned is rerniniscent of the contradictions inherent in contemporary sexism, whereby people claim to believe in gender equality but sirnultaneously endorse policies and views that disadvantage women. In this study. the participants seemed to hold the male- and female-typed jobs' responsibilities. skills. working conditions and educational requirements in equal regard, but this parity was not evident when they were asked to evaluate the jobs in a monetary way.

The results of this study irnply that contrary to economists' assenions. sexist notions about the worth of male- and female-typed jobs likely continue to affect women's employment outcomes, specifically salary. Admittedly. the effect size associated with the job gender main effect (q2=.02)was relatively small compared to that obtained for job income level (q2=.30).It is. however. sirnilar in magnitude to that observed in non-

experimental studies examining gender differences in income afler other potentially

relevant factors are controlled (e-g.. Melamed. 1995). tt should be noted that investigators engaged in research on gender issues have cautioned against dismissing effect sizes that are deemed "small" according to statistical ; such effects have been show to have important implications for women's outcomes (Eagly, 1995;

Ridgeway, 1997). Study 2

Hawig (re)established in Study 1 that male- and femaie-dominated positions are likely to receive dserential pay, the next obvious question is: why do women accept these lower paying jobs? Could it be that women are simply less concemed about money than men are? Are there advantages associated with female-dominated jobs that offset the lower pay that seerns to be part and parce1 of doing "women's" work? This section addresses these questions. and then turns to a consideration of whether employers make decisions that result in less favourable employment outcornes for women in terms of access and value. thus contributing to the wage gap.

Early research on work values suggested that women placed less emphasis on remuneration for work than did men (Banol, 1976; Bnef & Aldag, 1975; Schuler. 1975).

However, recent surveys no longer support the contention that men's and women's valuing of income is different. Rowe and Snizek (1995) analyzed data from 12 national

(U.S.) samples of full-time workers collected in the years from 1973 to 1990 and found that women and men ranked the importance of income and four other values (Le.. feeling of accomplishment, chance for advancement. job secunty and availability of short working hours) in the exact same order. Tolbert and Moen (1998) used the same national data in order to make cross-age cornparisons on these same five work values. Regression analyses reveaied that employee gender predicted ratings of the importance of promotions and security (valued more by men) as well as feelings of accomplishrnent (more valued by women), but not ratings of income or short working hours. The gender differences that did exist were widest among younger workers. decreasing greatly with age, but al1 in all. differences in preferences were relatively srna11 or nonexistent over the 20 year period. Thus it seerns that the wage gap cannot be attnbuted to women being less interested in compensation issues than men are.

Nor does it appear that women activeiy seek out female-dominated jobs for the reason that such jobs will more readily accommodate farnily responsibilities. Indeed jobs that are traditionally occupied by women are no more flexible or "family friendly" than are the jobs occupied by men (England, 1992; Glass & Camangg, 1992; Kessler-Hams. 1987;

LaRoss% 1988; Moen, 1985; Weiss. 1985). In addition, female-dominated jobs do not have srnaller penalties for intermittence (Duncan. Pms, & Sandy, 1993; England. 1984). which constitutes a personal work history pattern that is more characteristic of women's career paths than men's. Put another way. it seems unlikely that there are so manv family- fnendly perks associated with these jobs that women are induced to self-segregate and to accept lower pay. However, a work history pattern that is likely to include one or more interruptions of substantid length may not be tembly appealing from the perspective of an employer. Thus it rnay be that noms surrounding women's greater responsibility for childcare and family life may result in them being prevented from obtaining some jobs. particuiarly higher-paying jobs that are currently occupied predominantly by men (Konrad

& Cannings, 1997).

Statistical discrimination is posited to be an expression of employers' uncenainty regarding how committed women are to continuing their employment in the long run when faced with the competing demands of their families. While it certainly appears that employers convey their doubts to women (Amato. 1992; Gibbons, 1992; Swiss & Walker,

1993), research that examines employers' attitudes regarding gender rotes (Reskin. 1993) or how employers ascertain which employees are committed and which ones are not (Allen et al., 1994) is lacking. Allen et al. (1994) made an attempt to assess the impact of family-related work interruptions - they asked business smdents to rate the organizational commitment of "high perforrning" male or female employees under one of three leave of absence conditions (i.e., parental. medical or no leave). They found that female employees who had taken a 3-month parental leave were rated as no less committed than other employees were. Though this finding appears encouraging for women. it should be borne in rnind that this investigation focused on how people explain past behaviour when competent performance was clearly indicated; it may not be reflective of how people generate expectations regarding friture events when they do not know much about the target person. In addition. a 3-month parental leave is considerably shorter than the length of time that employees are ordinarily permitted to take. Perhaps this piece of information was constnied as evidence of the target ernployee's cornmitment, thus confounding the results.

Generally speaking, it is likely that employers expect women will experience family-work contlict, and that on occasion, family concerns will take precedence over work demands (Northcrafl& Gutek, 1993; Pleck, 1977). Russell and Rush ( 1987) examined attributions for poor performance on a work assignrnent and found that mothers were least penalized of ail groups and their substandard performance was excused via an extemal attribution. Simiiarly, and Eskilson (1988) found that the attributions raters made regarding familylcareer conflict were extemal, variable and uncontrollable. resulting in a lower likeiihood of a punitive response when family life intmded on work. Again. at hstblush, these hdhgs appear to bode well for women. But it should be noted that

Wiley and Eskilson's (1988) respondents rated family-work confiict as more stable for female employees than for males and hinted that their response to future occurrences of famiiy responsibilities interfenng with work would be less sympathetic. In short, the ability to explain away iderior performance that occurs as a result of family demands and the corresponding lack of punitiveness (at least for initial occurrences) does not speak to the question of whether employers would avoid hinng women if given the opportunity to do so. The current study directly addresses this question of access.

Furihemore, Study 2 uses experimental methods to test whether stereotypes about women's family comrnitments affect the salaries that employers assign to women. This study complements the findings of large-scale surveys, which strongiy implicate marital and parental status as factors that influence men's and women's wages. The findings on marital status are actually somewhat vaned for women. with some studies showing that married women are compensated more poorly than singie women (Cooney & Uhlenberg,

199 1 : Joshi, 199 1; Waldfogel, 1W8), whereas other studies have shown no such disadvantage (Landau & Arthur, 1992; Marini & Fan. 1997). Mamed men, however, seem to be paid better than single men fairly consistently (Baker, Benjamin, Desaulniers. &

Grant. 1995; Cooney & Uhienberg, 199 1 : Denton & Hunter, 1987; Rosenfeld &

Kalleberg, 1990; Shapiro & Stelcner, 1987). With respect to parental status, having children is customarily associated with financial benefit for men (Blau & Beller. 1988;

Cooney & Uhlenberg, 199 1; Hersch 1985) but not for women, for whom parenthood confers either no advantage (Elersch, 1985) or a distinct disadvantage (Mau & Beller.

1988: Cooney & ühlenberg, 199 1). The fact that mamage and parenthood has been repeatedly associated with better salaries for men while the opposite pattern tends to emerge for women points to sexism; these findings are in line with the notion that men should be the breadwinners within a family unit whereas women do not need to make a

"living wage." Taken together, these findings suggest that a study focusing on both hiring

(access) and salary (value) is warranted.

In order to test the idea that farnily responsibilities are detnmental to women's employment outcomes. participants were asked to listen to a mock interview and then to indicate whether they would hire the candidate and what salary they would assign. In addition, they were asked to consider how well the target would "fit in" at the workplace and how they would rate the applicant on variables relating to predictions of applicants' future behaviour, including: turnover potential. perceptions of the prospective employee's ambition, job involvement, willingness to respond to job demands involving work outside of regular hours, and potential for the employee's persona1 life to interfere with work.'

These variables were included in an effort to clan9 the factors that impacted on participants' hiring and wage-setting decisions.

During the mock inte~ew.participants were provided with specific information about prospective employees' family-related choices. Each participant was presented with either a male or femaie candidate; this candidate's family or health characteristics were manipulated, resulting in four farnilyhealth conditions: (1) candidate who expressed an interest in starting a family; (2) candidate who revealed that she was unable to have children: (3) candidate who revealed that s/he did not want children; and, (4) candidate who revealed that s/he was asthmatic. The first three conditions test the impact of stereotypes - these candidates have characteristics that are either in line with generic

-- - - The last five variables in the list were actually combined via factor analysis - more details are provided in the method section. stereotypes about women (Le.. that they are farnily oriented) or not (i. e., no family wantedlnot possible). The disability condition was included as a cornparison condition that conceivably requires one or more leaves of absence, just as maternity could. It was not expected that this particular ailment would be viewed by participants as so severe that the acted appiicant would be perceived as unable to do the job. nor was it presumed that by making illness saiient, the findings of this study could speak to the potential for discrimination faced by people with . lnstead the goal was to include a condition wherein the candidate's continuing availability could not be taken for granted, just as in the case of potential motherhood. By claribng for participants whether a pa~ticularcandidate is likely to request accommodations such as parental or sick leave. the question of whether bias against ail women (statistical discrimination) might be replaced by bias against subgroups of wonien (Le.. family-onented women) is addressed.

It is acknowledged that information about reproductive intentions is unlikely to be available to employers in an actual interview situation. Indeed questions related to marital status, family status, or disability contravene the interviewhg guidelines out by the

Ontario Human Rights Commission (1997). But the idea was not to convince the participants that they were listening to a "real" interview. Rather. they were told that one of their tasks was to assess whether the questions comprising the interview made for an effective inte~ewingstrategy. Moreover, they were told that it was a re-enactment, primarily because it seemed unlikely that anyone would believe that a bank would agree to participate in a study that would involve tape-recording actual job applicants. In effect, the interview scenario is merely a vehicle by which to tap into stereotypes about men's and wornen's commitment to work. There was another benefit associated with conducting this investigation in the context of an inte~ew- this scenario permits a consideration of how employers are placed in a situation that virtually guarantees that they will engage in gender stereotyping and perhaps even facilitates the discriminatory treatment of women. Research illustrates that in interview situations. the default tendency is for interviewers to behave in ways that increase the likelihood that their stereotypically-denved hypotheses are confirmed

(Neuberg. 1989). Certain social goals on the pan of the interviewer can modify this tendency; for example, people who are motivated to obtain accurate impressions of the interviewee. who want to be liked, or who want to be perceived as competedfair wili be more likely to conduct themselves in ways that are conducive to obtaining accurate rather than stereotyped impressions of the target individual. On the other hand. inte~ewers who want to generate an impression rapidly or who want to be seen as intimidating are more likely tu behave in ways that increase the chances that their stereotypes will be confirmed (Neuberg, 1994).

It seems reasonable to suggest that in employment settings. interviewers frequently have pals that are at odds with obtaining stereotype-disconfirming information about their interviewees. First. when choosing among job candidates. the time constraint inherent in an interview requires fairly rapid formation of impressions about the applicants.

Second, while the desire to choose the best candidate from a pool of applicants may increase an interviewer's motivation to perceive accurately an inte~ewee'sunique characteristics, the incentive to save one's fim fiom incurring potentially avoidable costs

(Le.. replacing a femaie worker who has left. either temporarily or permanently, in order to have a child) could very wefl be equally or more compelling. The effect of these interviewer goals may be exacerbated by the impact of provincial human rights legislation, which, as just discussed, imposes limitations on the topics of questioning permined during an employment inte~ew.By preventing employers from inquiring about candidates' marital and/or family status dunng employment inte~ews.this legislation makes it even more likely that interviewers will

Fom impressions and expectancies of women that are rooted in gender stereotypes.

Specificaily, the stereotyped description and prescription that women act as the primary caregivers for their children is udikely to be put aside, thus disadvantaging fernale candidates during hiring by setting up negative expectations regarding their long-term availability for work.

It must be emphasized that the preceding comments regarding the shortcomings of the Ontario Human Rights Commission's policy on employment interviews is not indicative of the author's opposition to the goals of human rightdanti-discrimination legislation. On the contrary. it is vital that such Iaws exist in order to discourage the unequd treatment of people on the basis of their membership in a panicular group.

However. it is also important to examine thoroughly the unintended negative effects that the law and its associated policies rnay have for women. The findings of both Studies 2 and 3 have implications for Our understanding of how this legislation rnay contribute to a situation where the perception of women in stereotypical tems is encouraged.

The major contribution of this study is that it tests whether cenainty regarding farnily intentions decreases gender-based discrimination in hiring and wage setting decisions. If the femaie applicants who reveal inability or lack of desire to have children are rated more positively ador allotted higher starting salaries than the family-onented female target, it would suggest that people refiain From statistically discriminating against

-al1 women once they have information that speaks to the potential for long-term job continuity.

Hyotheses for Prima- Analyses

It was anticipated that the ANOVAs conducted on the five dependent variables would yield similar patterns of results. More precisely. it was expected that a main effect would emerge for farnilyhealth status such that applicants who expressed a desire to start a family would be rated less favourably than applicants who indicated that they did not want or were unabie to have children. As for the "illness" condition, it was unclear what pattern of results to expect. It may have been that asthmatic applicants would be rated as unfavourably as those who wanted children because they could require accommodations. such as sick leave, at inopponune times. On the other hand. it is possible that asthma is perceived as a relatively minor medical condition that would not require much in the way of accommodation on the part of the employer. In fact. even if the raters believed that the asthmatic applicants would be unpredictable with respect to their need for leaves, these candidates might still be rated relatively positively because their medicai needs were beyond their own control and it would not seem fair to penalize them (Colella, DeNisi. &

Varma, 1998).

Furthemore, whereas fernale jobs were rated lower than their male counterparts in

Study 1, it was not clear whether there would be a siçnificant main effect of applicant gender in Study 2. In other words, it may not be the case that female applicants for the position of assistant bank manager will be rated lower than male applicants overail - a pattern of results consistent with old-fashioned sexism. Rather, if people who make hiring decisions tend to engage in statistical discrimination in the absence of concrete information about women's reliability as employees in the short- and long-term, it seems plausible to suggest that once they have such information, only select groups of women will be

penaiized. In other words, the main eKect of farnily/heaith status should be qualified by an

interaction with gender of applicant.

When the predicted interaction is decomposed such that male and fernale applicants are considered separately, it is expected that male applicants in ail three family conditions will be rated similarly. On the other hand. female applicants who are infertile or

do not want a farnily will likely be rated more positively than the female applicant who wants a farnily. It may be that the infertile woman could receive the most favourable

ratings of any of the target women - unlike the female applicant who reveals that she does

not want children this woman is more predictable to her employer because her childless

status is due to medical reasons, which tend to remain stable, rather than to personal

choice. which can change. It is not clear whether the infertile female applicant will be

rated as highly as any of the target men, however. it seemed likely that a female

prospective employee who planned to start a family would be rated lower than a male

applicant who made the same revelation. Indeed. planning for a family rnight actually

enhance a male candidate's perceived desirability because it may feed into a "family man

as stable" stereotype (Valian, 1998).

As alluded to in the description of the family/health status main eEect, it was

difncult to predict how the male and female asthmatic applicants would be rated relative to

the other candidates. It is possible that any negative ratings that may be associated with

the "iiiness" condition wiii be exacerbated if the target is a woman (i.e., she is "doubly" unpredictable as a worker because she could potentially become il1 or she could potentially leave to parent a child). Aitematively, the nature of the illness may be perceived as insufficiently senous to impose a "double jeopardy" of disability and gender.

No specific predictions regarding gender of participant effects were made. It seemed possible that there would be no main effect of respondent gender because men and women of university age might have shared ideas about how much men and wornen should get paid andor whether they are good workers. It seemed equally plausible that participant gender rnight interact with one or both of the other independent variables. For example. female participants rnight be more attuned to the difficulties faced by workers trying to balance family and career and therefore less likely to discriminate against job applicants on the basis of family status, resulting in an interaction between familylhealth status and respondent sex. Or, female participants might feel that family-related aspirations are Iiabilities oniy for female candidates. resulting in a three-way interaction.

Method

Power Calculations to Determine Sample Size

As with Study 1. estirnates of expected efEect sizes for the pnmary analyses of

Study 2 could not be obtained using similar expenmental studies. However, large scale survey research suggests that the proportion of variance in salary attributable to gender after ali other variables are controlled is between .O 1 and .O2(e.g., Melamed, 1995). Also. the results of Study I provided some guidance with respect to the size of the gender of target effect that rnight be expected with the dependent variable of salary (a dependent variable that Study i and Study 2 had in common). The effect of gender of target job on salary was ~rnall(~*= .02). But the desirability of obtaining a sarnple large enough to ensure adequate power to find a small effect needed to be balanced with considerations of the feasibility of recmiting several hundred students. In the end, the decision was made to ensure that there were at least 15 participants per cell- if as recniitment proceeded it became evident that this minimum number could be improved upon, then more participants would be included in the interests of increasing the power of the analyses.

Participants

The total sample consisted of 3 15 students. which according to the GPOWER analyses, yielded power of. 15 for a small effect. Most of the participants (n = 257;

8 1.6%) were recniited through the introductory psychology pool and received course credit for their participation. The remainder (n = 58: 18.4%) were recmited fiom undergraduate courses in the social sciences. Preliminary analyses indicated no statistically significant differences in the dependent variables of interest that could be traced to recniitment strategy. The sample was comprised of participants of typical undergraduate age (i.e.,25 years of age and under). Thus this study captures the attitudes of those who are relatively new to the paid workforce. Table 5 provides information about the characteristics of this sample.

Procedure

The study was advertised as an investigation of the intewiew process, the goals of which were to obtain respondents' impressions of the candidate as well as to ascertain what constituted appropriate, effective questioning. In an attempt to encourage high rates of participation, a lottery draw for pnzes of % 100. $50 and $25 was held. Ballots for the Table 5 Background Information for Stud~2 Participants

-N (%) Sex Female 166 (52.70) Male 149 (47.30)

Marital status Single 307 (98.10) Mamed/living together 6 (1.90)

Current employment status Employed 85 (27.00) Not employed 230 (73.00)

Ever held supervisory position? Yes 159 (50.50) No 156 (49.50)

S tudent status Full-time Part-time

Major Sciences 69 (21.90) Social Sciences 64 (20.32) Psychology 62 (19.68) Arts 47 (14.92) Applied Sciences 38 (12.06) B.Comm./Marketing 35 (1 1.1 1)

Note: Not al1 respondents provided demographic information, percentages reflect valid responses. lottery were filled out and collected separately from the forms in an effort to reassure participants that their responses tmly were anonymous and that potential for winning the pnze was independent of the answers given on the questionnaire.

Participants took part in the study in groups consisting of two or three people.

Care was taken to set up the seating arrangements such that participants were unable to see or consult with each other as they worked on the expenmental task. In al1 conditions, the candidate to be considered was applying for the same job. that of assistant bank manager. A management job was selected because it was likely to be perceived as the type of higher-level position that women have had dificulty accessing in the past. management jobs having traditionally been associated with men (Notthcraft & Gutek.

1993; Parasuraman & Greenhaus. 1993). but that was no longer unattainable. In fact. the proportion of female managers in the financial and business services has risen considerably in Canada in the last 1O years (Human Resources Developrnent Canada, 1998).'

Participants were made aware that the portion of the inte~ewthey would hear was a re-enactment rather than a tape recording of an actual interview. The interview was presented over headphones in an effort to encourage participants to attend carefùlly to the information presented in the audio-tape. Panicipants were also provided with paper on which to take notes about the details of the interview: by encouraging note-taking, it was hoped that participants would be more likely to listen attentively to the tape. Before listening to the audio-taped interview, participants received a written job description for

3 Still, it should be noted that women's increased presence in management in ment years (Adler & Imeli. 1988. 1994; Davidson & Cooper. 1993) lias been îargely limited to loîv- and middle-level positions (Boyci. MulvihiII & Myles. 1995; Martin. 1993). the position of assistant bank manager. This description was sirnilar in form to those appearing in the National Occupation Classification ( 1997). At this point, no information about the gender of the candidate was provided, participants were simply asked to rate the target job on the same nine compensatory factors assessed in Study 1. A copy of the job description together with the nine rating scales appears together with the entire questionnaire in Appendix D.

Participants then listened to the brief audio-taped segment. A wide range of inte~ewquestions had been generated and incorporated into the script that actors read for the audio-tape. Some of these questions were appropnate while others were inappropriate to ask in a job interview - indeed some of the questions addressed topics that the Ontario Human Rights Commission advises strongly against inquirins about

(e-g..family/health status. history of sexual harassrnent. use of employee assistance prograrns. ethnic background). The impropriety of some of these questions likely did not come as a surprise to participants since they knew in advance that one of their tasks would be to evaluate al1 of the questions with respect to their aptness for a job interview.

Furthermore, it was expected that the inclusion of a variety of inappropriate questions concealed our specific interest in the impact of family status on the assessment of job candidates. The content and order of presentation of these interview questions was

constant across ail conditions.

The preamble to the interview provided information about the applicant's gender,

age and marital status. The latter two characteristics were constant across ail conditions -

the candidate was always 30 years of age and had been married for three years.

Information about the four farnilyhealth status categories was revealed near the end of the interview in the form of applicant's answers to quenes by the interviewer (Le., "Do you have a family?" or "Do you have any health issues that our insurance Company should know about?"). The entire interview scnpt is included in Appendix C. Mer listening to the tape, participants completed a questionnaire concerning their opinions about the candidate, the inte~ewitself. as well as their own personal background. The questions were compnsed predominantly of rating scales. although there was a section where participants were asked to provide any additional information that they wanted to share regarding their decision-making process. The questionnaire appears in Appendix D

Measures

Compensatory Dimensions

The nine cornpensatory factors that served as dependent variables for one of the preliminary analyses in Study 2 have already been descnbed in detail in Study 1. To review, there were three items related to the educational and training requirements for the job (Le., shodlong, generaVspecialized and easy/difficult). two tapping required skills

(i.e., physicaVmenta1 and simpIe/complex). two related to responsibilities for materials, equipment and people (i.e.. slight/considerable and specifichroad) and two assessing working conditions (i.e.. undernanding/demanding and varied/repetitive [this last item was reverse-coded]). Once again, these nine items consisted of opposing adjective pairs. each of which was rated on a 7-point scale where the first adjective in the pair served as an anchor for "1" while the second adjective anchored "7." The main purpose of including the rating task was to encourage the participants to artend to the requirements of the job. Interviewer and Candidate Performance

Additional dependent variables for the prelirninary analyses appeared in the first two sections of the questionnaire. Section one included items tapping general aspects of the interviewer's and the applicant's performance; these questions were included in order to make the study's focus on gender issues less obvious. as well as to ascertain whether respondents found the actors' performances similarly plausible across conditions. There were four questions related to the interviewer's behaviour (i.e., "the inte~ewerwas personable," "the interviewer spoke clearly," "the interviewer had a good interviewing technique." and "the interviewer's behaviour was similar to an actual interviewer's behaviour") and six related to the candidate's behaviour (Le.. "the candidate was confident." "the candidate was hesitant in giving her answers." "the candidate made a good impression," "the candidate spoke clearly." "the candidate handled the interview situation well," and "the candidate's behaviour was simliar to an actual candidate's behaviour"). Each of these items was rated on a 7-point scale ranging fkom 1 "strongly agree" to 7 "strongly disagree."

Interview Questions

The second section of the questionnaire involved presentation of al1 13 interview questions that had been asked of the applicant. Participants were asked to rate each inte~ewquestion with respect to whether it was perceived to be (i) cornrnonly asked in an interview situation (ii) appropriate to ask of applicants. and (iii) relevant to job performance. The goal of having participants revisit and evaluate the inte~ewquestions was to increase the saliency of the familylhealth status manipulation pnor to rating the candidate on the primary dependent variables of interest (described in the upcoming subsections). As mentioned earlier, the family/health status manipulation was presented in the form of candidate's answers to the interviewer's Iast query

The first 12 questions. presented in the same order across al1 conditions, were as follows: "1 see from your application that you've worked in a bank setting before. Can you tell me a little bit about that?"; "What was your salas, at your previous job?": "What is your ethnic background?"; "What kind of skills do you bring with you that make you a good employee?"; "What personal qualities do you have that make you a good employee?"; "Have you ever been involved in a difficult situation with a fellow worker. for exampie, have you ever been sexually harassed by one of your coworkers?"; "At any of your previous jobs, have you ever used an Employee Assistance Program to deal with a persona1 problem?"; "What are some of the things you like to do when you're not working?"; "Why do you want to work for Our company?"; "How do you view your future with our company?"; "1 see that you've just recently moved to Guelph. Where are you from originally?"; "How long have you been mamed?". The finai question of the interview was one of the following: "Do you have a family?" or "Do you have any health issues that our insurance company should know about?". Al1 interview questions were rated on 7-point scales where 1 = "commonly asked." "inappropriate" and "relevant to job performance" whereas 7 = "rarely asked," "appropnate" and %relevant to job performance."

Predictions Regardin3 Candidate's Work Dedication

Sections three and four contained the dependent variables of interest for the pnmary analyses. In section three of the questionnaire, respondents' impressions of how good an employee the candidate would be were obtained via their ratings of the candidate on eighteen job-related questions. These questions were adapted fiorn previous studies of worker job satisfaction, cornmitment and turnover.

Specifically, there were five questions on the applicant's predictabilitv with respect

to availability for overtime on evenings and weekends (Le., "this candidate is unlikely to

stay late at work," "this candidate is likely to go to work eariy," "this candidate is unlikely

to b~gwork home and work dunng the evenings," "this candidate is likely to work on

weekends," and "this candidate is unlikely to travel for business purposes" (Rosen &

Korabik. 1995)), five questions related to ambition (Le., "it will not be important to the

candidate to keep advancing her career." "it will not be important to the candidate to be

seen as very successful at her job." "the candidate will want very badly to 'rnake her mark'

in her career," "the candidate will be prepared to work long and hard to gain prominence

in her career," and "the candidate's goal will be to reach the 'top' in her career" (Rosen &

Korabik, 1995)). four questions about the degree of job involvement the applicant could

be expected to exhibit (Le.. "the most important things that happen to the candidate will

involve her present job," "most of the candidate's interests wil1 be centred around her job," "the candidate will be very personally involved in her job," and "the candidate will

feel that her job is only a small part of who she is" (Kanungo. 1982))- and four questions

on the degree to which the candidate's persona1 life might be expected to interfere with

work (i.e.. "the candidate will often be too tired at work because of her life

circumstances," "demands related to the candidate's life circumstances will be so great

that they will take away from her work," "coworkers will dislike how often the candidate

is preoccupied with her personal issues while at work." and "the candidate's Me

circumstances are likely to take up time that she should spend at work (Burley, 1989; cited in Gutek, Searle. & Klepa, 1991)). These items were rated on a 7-point scale ranging frorn 1 "strongly disagree" to 7 "strongly agree."

Ratings of the candidate on these 18 items were sufficiently inter-correlated to ment conducting a principal component analysis (Le., there were several correlations that were greater than .3. as per Tabachnick and FidelI(1989)). The critenon for retaining items was a factor loading greater than 0.3. The optimal solution consisted of a sinsle component; factor loadings and communalities for these items are provided in Table E- I of Appendix E. Two of the 18 items failed to reach the criterion factor loading (Le.. one item fiom each of the dimensions of ambition. and degree of job involvement). The remaining 16 items compnsed a scale with a minimum possible value of 16 and a maximum of 112 that had acceptable interna1 reliability (Cronbach's a = .73. which is above the commonly accepted standard of .70 (Nunnally, 1978)). Higher scores on the overall scale indicated that the candidate was rated as a more dedicated prospective employee because s/he was more arnbitious. more involved in the job. more likely to work ovenime and/or less Iikely to have a personal life that interfered with w~rk;as a result. this newly created dependent variable was labelled "dedication."

Candidate's Ability to "Fit in." Predicted Leneth of Tenure. Hiring and Salary

In section four. questions about how well the respondent felt that the target would

"fit in" at the workplace (on a 7-point scale ranging frorn 1 "not at all" to 7 "very well"), number of years they predicted the applicant would stay with the bank, as well as how likely the participmt would be to hire the applicant (on a 7-point scale ranging from 1 "not at ali likely" to 7 "very likely") and the recommended starting salary were asked. Al1 of the questions in this section were prefaced with a brief paragraph indicating that the final decision about hinng an applicant is usually based on a composite of decisions by a hiring committee. Although the participants in this study did not themselves engage in a group discussion of the candidate's ment (each person simply provided written responses to the queries; see Foschi, Lai. & Sigerson, 1994), it was hoped that providing information about this practice would counteract any tendency for participants to respond to any perceived demand characteristics by agreeing to hire the candidate.

In order to circumvent the potential problem of participants being too unfamiliar with the task of assigning salaries to give a reasonable estimate for this position. they were provided with the actual range of salaries associated with the job of bank manaser

(obtained from Human Resources Development Canada 1998). It should also be noted that the "fit in" variable was included as a measure of taste discrimination, the phenornenon whereby employers and advantaged employees conspire to exclude an undesired group (England, 1992).

Preview of Analvses

Preiiminarv Analvses

The nine job-related compensatory factors were subjected to a MANOVA with gender of participant as the independent variable in order to ensure that both men and women had sirnilar understandings of the nature of the assistant bank manager position with respect to its responsibilities. educational requirements, working conditions and skik No significant participant gender effect was expected.

Then, in an effort to determine whether the interviewer-candidate conversation was perceived similarly in al1 conditions. participants' impressions of the interviewer's and candidate's respective performances were analyzed via two separate 2 x 4 x 2 (Gender of Participant x Family/Health Stanis x Gender of Applicant) between-subjects MANOVAS.

It seems reasonable to infer that the 8 experirnental conditions are comparable if the independent variables of Family/Health Status and applicant gender do not emerge as significant in these analyses, either alone or in concert.

Next, the mean ratings of the appropriateness. commonness and relevance of al1 thirteen interview questions were examined. Potential differences on the ratings on the family/health status question which preceded the experimental manipulation were of

particular interest (Le.. "Do you have a family?" or "Do you have any health issues that

Our insurance Company should know about?"). A 2 x 4 x 2 (Gender of Participant x

FamilylHealth Status x Gender of Applicant) MANOVA was conducted on these three

dimensions for the last interview question. tf statistical discrimination against wornen

occurs because of employer uncertainty regarding women's intentions to assume familial

obligations. which would presumably interfere with work responsibilities. then the

interview question regarding family status should be perceived as especially appropriate

andor relevant for fernaie applicants.

Primant Analyses

As outlined in the "Measures" section, ratings of the candidate on the 18 items dealing with predicted avaiiability for overtime. ambition. degree of job involvement. and degree to which tne candidate's personal life rnight be expected to interfere with work were factor analyzed. This analysis resulted in a single 16-item scale. labeled "dedication."

The employee dedication dependent variable that emerged from this analysis was entered

into a 2 x 4 x 2 (Gender of Participant x Family/Heaith Status x Gender of Applicant)

between-subjects MANOVA with the two other dependent variables that tapped participants' perceptions of the applicant's future behaviour: (1) the length of time they predicted the applicant would stay with the bank, and (2) how well the applicant would

"fit in" at the workplace. The cornmon thread among these three variables that justified their being analyzed together was that they al1 tapped different aspects of respondents' perceptions/predictions about the candidate.

Because it seemed likely that panici pant s' behavioural response to ward the candidate could be more prone to the iduence of social desirability than simple predictions of the applicant's behaviour. the remaining two dependent variables, likelihood of hiring and salary allocation, were analyzed separately from the first three. Moreover. it seemed prudent to examine the latter two variables separately in light of Auster's ( 1989) insight that bias could enter into workplace decision-making at either the stage where one judges whether an award should be given (e.g.. hiring promotion) OR at the stase where the magnitude of the award is being decided upon (e.g.. the salary allocation stage). Thus questions about how likely the participant would be to hire the applicant and the recomrnended starting sdary were analyzed individually using two separate 2 x 4 x 7

(Gender of Participant x FamilyiHealth Status x Gender of Appiicant) between-subjects

ANOVAs. Given that the hypotheses for al1 of these primary analyses involved interaction effects rather than main effects, al1 ANOVAs were conducted using the regession approach to panitioning sums of squares instead of the hierarchical approach used in

Study 1. Results

Com~ensatorvDimensions

Participants rated the target job (assistant bank manager) on the same nine compensatory Factors that were used to rate the six jobs in Study 1. This task was undertaken prior to disclosure of the candidate's gender and familyhealth status, therefore no comparisons across the different experimental conditions were made. A between- subjects MANOVA comparing male and female respondents' ratings on these nine items revealed no significant gender differences (FtrZssi= l.l3, -). Means and standard deviations For the overall sample on these items appear in Table E-2 of Appendix E and are illustrated in Figure 4. Interna1 reliability analyses revealed that the nine variables did not cohere particularly well (alpha = S9). therefore these ratings were not incorporated into any of the remaining prelirninary or primary analyses.

Interviewer and Candidate Performance

Two separate 2 x 4 x 2 (Gender of Participant x Family/Health Status x Gender of

Applicant) MANOVAs were conducted in order to ascertain whether the interviewer- candidate conversation was perceived similarly in al1 conditions. One MANOVA was conducted on the four questions tapping the interviewer's performance whereas the other

MANOVA dealt with the six questions assessing the candidates' performance.

For the interviewer's behaviour, there were no significant main effects or interactions detected on the four dependent variables. The summary table for this

MANOVA is in Table E-3 of Appendix E whereas the means are illustrated in Figure 5. Figure 4 Means Ratinas of Cornoensaton, Factors for Assistant Bank Manager: Ai1 Participants

Mental Cornplex Lonn Sdfic Difiicult Considerable 5mad Oemandinq Repelitive I

Physical Simple Short General Easy Slight Specific Undemand. Varied

Compensatory Factors Figure 5 Mean Ratines of Interviewer's Performance on 4 Dimensions: All Particioants

Strongly Disagiee

Strongly Personable Spoke Clearly Good Technique Similar ta Real Agree Interviewer's Performance Means for each of the four variables are presented for the sample as a whole given the lack of significant effects on any of these variables. Clearly. participants agreed that the interviewer spoke clearly and was personable. Average ratings for sirnilanty to an actual interviewer and possessing good inte~ewingtechnique were also on the "agree" side of the scale. but were doser to the midpoint and appeared to display slightly more variability than the first two items did (see rneans and standard deviations in Table E-4 of

Appendix E).

When the candidate's behaviour was considered, the MANOVA reveaied significant main effects for gender of participant (5b.2921= 2.84. eC.05; r12 = 06)and gender of target (F[6.~9~)= 4.76. ~<.001 ; = .09). These findings warranted conducting subsequent univanate analyses on each dependent variable separately. The summary table

for ths MANOVA appears in Table E-5 of Appendix E whereas the six follow-up univariate ANOVAs are in Tables E-6 through E-1 1 respectively in Appendix E. Means and standard deviations for ratings of the six candidate behaviours appear in Table E- 12 of

Appendix E; means are depicted by participant gender and by gender of target in Figure 6.

The follow-up tests were corrected for intlated alpha levels using the Bonferroni method such that the critical p-value for a significant univariate test was set at p<.008. These

analyses indicated that male participants rated candidates less favourably than did fernale participants on the dimensions of confidence (E, I.B7) = 12.84, p<.OO 1 ; q2 = .04), making a

good impression (F1lJa, = 7.88, g<.008;rlz = .03).and handling the interview well (F,

= 8.93. e<.OOS; = .03). The main effect of gender of target was significant on the

hesitation (F(1292)= 7.94. ~<.008;q2 = -03)and clear speaking dimensions (F(1.297)= 1 1.97, Figure 6 Mean Ratines of Candidate's Performance on 6 Dimensions

By Participant Gender

Stmngiy Confident Hesitant Good Impression Spoke Cleariy Handled Well Sinilar ta Real Agt- (ps.001) (ns.) (P<.OO~ (ns.) (PC.~) (ris.)

Candidate's Performance 10 ale HFemale 1

By Gender of Target Stmngly

S'm"gb Confident Heutant Good Impression Spoke Clearty Handled Well Similar ta Real Agr- (n.s.) (P<.W) (n.s.) (p<.w) (n.s.) {n.s.)

Candidate's Performance -1 p<.OOl; q2 = .04), with the fernale candidate rated siightly better than the male on both of these dimensions. It should be noted that these differences were not perceived to threaten the validity of the prirnary analyses - the implications of these findings are dealt with more fully in the discussion.

Interview Questions

Turning to the content of the interview, participants' perceptions of the commonness. appropriateness and relevance of the family/health status question that preceded the experimental manipulation were of special interest. Ali thineen interview questions were placed in order of their overall mean ratings on each of the three dimensions; a ranking of" 1" indicates that the question was ranked most common (or appropriate or relevant) in a real interview situation. Table 6 shows that on al1 three dimensions, the familylhealth query ranked roughly in the middle among the thirteen questions while within each dimension, the rnean ratings of this question feil quite close to the midpoint of the 7-point scale (although the means fell just to the "rare,"

"inappropriate," and "irrelevant" sides of the scales).

A 2 x 4 x 2 (Gender of Participant x FamilyfHeaIth Status x Gender of Applicant)

MANOVA was conducted on the commonness. appropriateness and relevance of the familylhealth question in order to ascertain whether it was viewed as especially pertinent for femaie applicants. This MANOVA appears in Table E- 13 of Appendix E. The analysis revealed a main effect for family/health status (&sss, = 2.62, gC.0 1; $ = -03) and gender of appiicant (F[329~)= 2.63, ~=.05; = .O;). as well as a significant interaction Table 6

Mean Ratinps of Comrnonness/Appropriateness/Relevance of Interview Ouestions

Common Appropriate Relevant Questions in order of Rank Rank Rank appearance /13 M(sd) /13 -M (d) /13 bl(sd)

Previous job

Previous salary

Ethnic background

S kills

Personal qualities

Ever sexually harassed

Ever used EAP

Hobbies

Why this Company

Future aspirations

Where are you from

How long married

Family/health status

Note: Al1 items were rated on 7-point scales where 1 = "commonly asked," "inappropriate" and "relevant to job performance" whereas 7 = "rarely asked," "appropriate" and "irrelevant to job performance" between gender of participant and gender of applicant (Ft3.m)= 2.85, ~c.05;'12 = 03).

These findings warranted conducting follow-up univanate analyses.

Once corrections for inflated alpha were made via the Bonferroni method such that the cntical p-value was reset at g<.0 17, significant effects emerged only for the

"relevance" dependent variable. The follow-up ANOVA source tables for the commonness. appropriateness and relevance variables are in Tables 8-14 through 8-16.

The means for al1 three dependent variables are presented by family/health status and by gender of applicant in Figure 7. The means and standard deviations for the "relevance" dependent variable are also presented in table form iii Table E-17 of Appendix E.

Specifically. there was a gender of applicant main effect wherein a family/health question was perceived as more relevant for female targets than males (-(1.~96)= 6.5 1. pc.0 17,

&.T~% = 4.34,KLTaq = 4.55). There was also a significant main effect for farnily/health condition (Euv296) = 4.76, pC.005); pairwise Tukey's tests revealed that health status

(having asthrna) was rated as more relevant to job performance than any of the family status conditions = 3.97 vs. = 4.64, Md,., ,,, = 1.83, and MW,.,= 4.95). As with the ratings of the candidate, these differences in the ratings of the inte~ewquestions across conditions were not perceived to threaten the validity of the primary analyses - more detail is provided in the discussion. Figure 7 Mean Ratines of Farnily/Health Ouestion on 3 Dimensions Bv Candidate's Family/Health S tatus and Gender

Rare Appropriat lrrelevant e

ppppp Family/Health Gender Farnily/Health Gender Family/Heal th Gender n s n s n s n s p< 005 pc 017

WCOA M F W C OA M F

Inappropnate Relevant Interview Questions Primary Analyses

Predictions of Ap~licant'sDedication, "Fitting In" and Length of Tenure

As indicated earlier, the applicant work dedication dependent variable that emerged from the factor analysis was entered into a 2 x 4 x 2 (Gender of Participant x

Family/Health Status x Gender of Applicant) between-subjects MANOVA with two other dependent variables: (1) predicted length of time the applicant would stay with the bank, and (2) how well the applicant wouid "fit in" at the workplace. The MANOVA source table for this analysis appears in Table E- 18 of Appendix E whereas the ce11 means and standard deviations for applicant dedication. applicant's ability to fit in and predicted time before turnover appear in Tables 7. 8 and 9 respectively. Significant main effects emerged in the IMANOVA for gender of participant (&3.211) = 4.10. e<.O 1 ; ri' = .04) and familylhealth status (i?(9.828) = 3.17. g<. O0 1 ; ri' = .O;). There was also a significant interaction between familylhealth status and applicant gender (E(9.1R8)= 2.32, gc.05; rf =

.03). These multivanate effects were followed up via ANOVAs conducted on each dependent variable separately; a more stringent cnterion of p<.0 17 was set in order to correct for inflated alpha levels.

The ANOVA for the applicant dedication variable appears in Table E- 19 of

Appendix E. A significanr main effect emerged for gender of participant (F( 1.276) = 10.81. gc.00 1; = .04) wherein female participants rated applicants more favourably than did male panicipants (MF.R- = 89.17. MILR-~= 85.93). The main effect of famiIy/heaIth status was also significant (F(3.216) = 3.57. pC.0 17: = .O4). Tukey's HSD tests showed that applicants who indicated that they did not want a family were rated as more suitable ernployees than those who expressed a desire for children or those who were asthmatic

96 Table 7 Mean Ratin~sof' Male and Female Applicant's Dedication b~ Participant Gender and FamilyHealth Status

Gender of Applicant Male Female '~ainEffect of Participant Sex Male Female Male Female Family/Health -M -M -M -M -M (sd)- (dl (.si!) (d) (sd)- Familv/HealthStatus 84.61 Want children (8.48) n= 18

Can't have children 84.19 (8.33) n=18

Don't want children 87.72 (6-64) n=18

Have asthma 85.4 1 (7.73) n=17

*O* Main Effect of 85.93 Participant Gender (9.15)

Main Effect of 87.22 88.0 1 Applicant Gender (8.3 1) (9.65) *w* 'gC.0 17, gc.00 1

Note: This scale had a minimum possible value of 16 and a maximum of 1 12, with iarger numbers indicating a more favourabie rating of the candidate. Also, cntical p-values for this analysis were adjusted using a Bonferroni correction. A minimum p-value of pc.0 17 was required to deem an effect statistically significant. Table 8 Mean Ratines of How Well Male and Female Candidates Would "Fit In" bv Particioant Gender and Family/Health Status

Gender of Applicant Maie Female Main Effect of Participant Sex Male Female Male Female Family/Health M M -M M M (sd)- (sd) (d) (Sc!) (sd) FamiIyMealth Status 5.67 5.88 6.00 Want children (O. 84) (1.02) (0.9 1) n=18 1146 n=18

Can7thave children 6.00 5.59 6.10 (1.28) (1.73) (O. 64) n= 18 n= 17 n=20

Don't want chiIdren 6.22 6.33 6.20 (0.65) (O. 84) (O. 70) n= 18 n=18 n=20

Have asthma 5.76 6.22 6.42 (1.15) (0.65) (0.61) n=17 n=18 n= 19

Main Effect of 5.96 6.18 Participant Gender (1 .OS) (0.67)

Main Effect of 6.05 Amiicant Gender (O.84)

Note: Participants were asked to rate on 7-point scale how well the candidate would "fit in" at the Company where 1 = "not at al]" and 7 = "very well". Aso, cntical p-values for this analysis were adjusted using a Bonferroni correction. A minimum p-value of p<.017 was required to deem an effect statistically significant. Table 9 Mean Estirnates of Number of Years Male and Female Candidates Will Stav with Hirinq Comoany bv Participant Gender and Familmealth Status

Gender of Applicant Male Female "~ainEffect of Participant Sex Male Fernale Male FemaIe Familykiealth -M -M -M -M -M (Sb) (.sa (sd) (A) (sd) FarnilykIealth Status 11.69 13.03 9.94 9.1 1 10.97 Want children

Can' t have children

Don't want children

Have asthrna

Main Effect of Participant Sex

Main Effect of Applicant Gender (8.36) (8.3 5) '~c.01 7, "e<.005

Note: Critical p-values for this analysis were adjusted using a Bonferroni correction. A minimum p-value of pC.0 17 was required to deem an effect statisticaily significant. ,, ,, = 89.68 versus Mfafii~86.0 1 and Mah= 86.03). These main effects were qualified by a significant three-way interaction (53,276) = 3.57. p<.O 17; = -04) and a family/health status by gender of applicant interaction (E0.276) = 3.76. p<,017;qZ = .04).

However, because the MANOVA did not produce a significant result on the three-way interaction, it is only statisticaily justifiable to follow up the two-way interaction between familylhealth and gender of applicant (Tabachnick &Fideil, 1989). This interaction. together with the main effect for farnily status. is depicted in Figure 8.

Decomposition of the two-way interaction revealed results in line with expectations. When male applicants were isolated. there were no differences among familyhealth status groups (E(3.27676)= 1.82. n.s.) whereas arnong female applicants. significant group differences ernerged (F@J~~)= 5.29. p<. O 17). Tukey ' s tests showed that female candidates who expressed a desire for children were rated as less suitable employees (M~MI~*.~.~= 83.82) than women who either did not want or could not have children (Md,, ,,Uq.~a9 = 89.82 and &n+t hvd~F.Taqj= 9 1 22 respectively). Looking within family/health status groups, gender of applicant differences emerged under two of the four conditions: female applicants who wanted children were rated as less suitable than sirnilar maie appiicants (ç1.276) = 4.33. P<.0 1 7; MF,TJ~~~~~~~~= 83.82and MhLTaremrrmiir

= 88.07) whereas female applicants who were unable to have children were rated more highly than comparable maie appiicants (F(1.276)= 6.46. pC.0 17; MF.Tar8(aun*tbve = 9 1.22 and

MM~ar~don'tbvc = 86.18). Figure 8 Mean Dedication Ratines bv Farnil~kiealthStatus: Seoaratelv by Applicant Gender and Overall

Want Children Can't Have Chitdren Don't Want Children

FarnilylHealth Status +Female Applicant - - * -Male Amlicant Univariate tests on participants' ratings of how well the candidate would "fit in" at the bank showed that fernale respondents were siightly more likely than males to indicate that the applicant would fit in (MF.Rae= 6.18. MMRctp= 5.96). However, the more stringent critenon for significance which was adopted in order to maintain the farnily-wise error rate at an acceptable level (Le.. pC.0 17) resulted in a nonsignificant test (Et1.276) =

4.67, p=.032). No other significant main effects or interactions emerged. Further examination of the ce11 means indicated that in al1 cases. the applicant was rated as quite likely to "fit in" at the bank (the lowesr mean rating was 5.5 on a 7-point scale). Refer back to Table 8 for ceIl means and standard deviations.

In the follow-up univariate tests on participants' predicted estimates of how long the applicant would stay at the bank once hired. only the family/health status main effect reached significance (&76) = 5.35. pC.00 l : q2 = .06). This main eEect is depicted in

Figure 9 whereas ceil rneans and standard deviations appear in Table 9. Tukey's tests showed that the applicant who declared that dhe wanted a family was expected to stay with the fim for a significantly shorter penod of time (10.97 years) than the applicant who could not have children (15.99 years) or who expressed a desire to remain chiidless ( 15.42 years). None of the other familyhealth status groups differed significantly fiom each other. Figure 9 Predicted Number of Years Candidate Will Sta~with Company BY FamilvhIealth Status

Want Children Have Dan't Want Asthma Children Children

FamilyIHealth Status Likelihood of Hiring

Participants' ratings of whether they would hire the applicant were analyzed via a 2 x 4 x 2 (Gender of Participant x FamilyRIealth Status x Gender of Applicant) ANOVA.

The ce11 means and standard deviations appear in Table 10 whereas the ANOVA source table for this analysis appears in Table E-20 of Appendix E. Examination of the cell rneans revealed that al1 of the average ratings were well above the midpoint of the scale. indicating that the applicant was quite likely to be hired in al1 conditions. Still, two of the three main effects were significant: gender of panicipant (EL1,2741= 29.26, p<.00 1; v2 =

.IO) and family/health status (E(3.274)= 2.73. pC.05; q' = .03).Although none of the pair- wise cornparisons for the latter effect reached statistical significance, ordenng of the means indicates that the asthmatic candidate received the highest mean rating (MUh=

6.16) while candidates who expressed a desire for a family received the lowest rating

(Mermi,=5.82) - see Figure 10 for depictjon of the family/health status main effect. With respect to the gender of respondent main effect, female participants were more likely than male participants to indicate that they would hire the applicant (MF.Rap= 6.26, Mhi.Rsp-

5.76). Furthermore, the main effect of gender of applicant approached significance (E, 1.2741

= 3.72, p=.055; q2 = .O1). such that participants overall expressed a slight preference for hiring the female candidate (MF~,,= 6.1 1. M&traro= 5.94). There were no significant two-way or three-way interactions. Table 10 Mean Ratings of Male and Female Applicants' Likelihood to be Hired bv Participant Gender and Familv/Health Status

Gender of Applicant Male Female '~ainEffect of Participant Sex Male Female Male Female FamilyMeaith -M -M -M -M -M

Family/Healt h Status 5.13 6.15 5.78 6.05 5.82 Want children ( 1.46) (0.49) (O.94) (0.86) (1.01) n=15 n=20 n=18 n=2 1

Can't have children 5.56 6.36 5.8 1 6.50 6.04 (1 .?O) (0.45) (0.91) (OS1) (0.89) n=18 n=19 n= 16 n=18

Don't want children 5.76 6.26 6.06 6.20 6.08 (1.09) (0.56) (0.68) (O.62) (O.76) n= 17 n= 19 n=16 n=20

Have asthrna 5 -97 6.30 5.94 6.50 6.16 (O.62) (O.62) (O. 64) (O.52) (0.63) n= 17 n=20 n= 18 n=18

*LI Main Effect of 5.76 6.26 Participant Sex (O.99) (0.60)

'Main Effect of A~~licantGender

Note: Participants were asked to rate on 7-point scale how likely they would be to hire the candidate where 1 = "not at al1 likely" and 7 = "very likely". Figure 10 Mean Ratines of Likelihood of Hirine Candidate by FamilvMealth Status

. . Want Children Have Want Asthma Nat al al1 Children Children

FamilyIHealth Status Startina sala?

Finally, starting salaries assigned to applicants were analyzed with a 2 x 4 x 2

(Gender of Participant x Farnily/Health Status x Gender of Applicant) ANOVA. The ce11

means and standard deviations appear in Table 1 1 whereas the ANOVA source table for this andysis appears in Table E-23 of Appendix E. This tirne. only the familylhealth status main effect reached significance = 3 .O 1. g<.05: $ = .Oj), al1 other main effects and interactions were nonsignificant. Tukey's tests showed that the mean salary oRered to the applicant who declared that she wanted a farnily (Memiiy= $40.839) was significantly lower than that offered to the applicant who indicated that s/he did not want children

(h/bon-t = $44,070). but that none of the other family/health status cornparisons were statisticdly significant; refer to Figure 1 1 for illustration of the means, Table 11 Mean Starting Salarv Assimied to Male and Female Applicant bv Participant Gender and FamiidHeait h Status

Gender of Applicant Male FemaIe %lain Effect of Participant Sex Male Fernale Male Female FamilyMealt h -M -M -M -M -M

FarnWHealth S tatus Want children

Can't have chiIdren

Don't want ctuldren

Have asthrna

Main Effect of Participant Sex

Main Effect of A~~iicantGender Figure 11 Mean Salary Assiened to Candidate by Familv/Health Status

Want Children Can't Have Don't Want Asthma Children Children

FamilylHealth Status Discussion

Preliminary Analvses

The preliminary analyses revealed the possibility that the male and female target conditions were not strictly equivalent; the female applicant was rated as speaking more clearly and with less hesitation dunng the interview than the maie candidate did. It was unclear whether these particular dimensions of intewiew performance would be viewed by participants as indicative of the applicants' cornpetence in the role of assistant bank manager. However, the fact that the slight differences in the actors' performances in the simulated interview favoured the female candidate rneant that the odds may have been somewhat against finding that sexism infiuenced ratings of the candidate on the primary dependent variables. Thus the finding that female candidates actually fared worse than male candidates in one of the primary analyses becomes even more striking (recall that the female candidate who adrnitted to wanting a family was rated as less dedicated to work than a similarly inclined man).

The preliminary analyses also uncovered a pattern of participant gender main effects that proved to be persistent in the primary analyses; male participants tended to be harsher judges of candidates, regardless of the candidate's gender. In the preliminary analyses. male participants rated the candidates less favourably than did female participants on the dimensions of confidence. making a good impression, and handling the interview well. However, no gender of participants differences emerged in ratings of the interviewer's perfomance or in evaluations of the work conditions. skill, responsibility and educationai requirements of the assistant bank manager job. The finding that maie respondents were generally harder on the applicants was not perceived to compromise the validity of the analyses conducted - more detail about this point is provided in an upcoming section describing the primaq analyses.

While the questionnaire items related to the prelirninary analyses were included primarily to deflect suspicion that we were interested in how participants reacted to the applicants' family status, they actually provided some pertinent information on the latter issue. albeit indirectly. To illustrate. numerous participants picked up on the fact that many of the questions in the audio-taped excerpt were not truly befitting an interview, as shown by their mean ratings of the questions on the "cornmon." "appropriate" and

"relevant" dimensions. Yet the magnitude of the standard deviations for these items suggests that there was a wider range of opinion about the commonness. relevance and appropriateness of the more "personal" questions (e.3..marital and farnily status. place of origin, use of employee assistance prograrns, involvement in incidents of sexual harassment) than for questions that were more clearly related to assessing job-related qualities (e.g., work skills, motivation for applying) and for the question regarding the applicant's ethnic background, which most people correctly identified as a prohibited topic in the context of ernployment interviews. Of al1 the questions. the family/health status query displayed the most variability on the three dimensions. Furthemore, about one- third of participants actually rated this query as "cornmon." "appropriate" and "relevant." in direct contravention of the anti-discrimination legislation currently in effect. It cannot be said with ce~aintywhether this pattern of responses reflects lack of awareness or lack of endorsement of the legislation that forbids such questions. Either way, the pnmary analyses show that participants take these personal characteristics into account when making their evaluations.

Before tuming to the main analyses. it is acknowledged that participants rated the family/health status question as more relevant for the asthmatic candidate than for any of the other candidates, suggesting that reproductive status is seen as Iess relevant to job performance than is health status. Neveriheless. this finding does not nile out the possibility that sexism influences women's employment outcornes - recall that there was also a main effect for applicant gender. wherein the familyhealth status question was rated as more relevant for female applicants than for males. This gender effect is consistent with the cultural nom that women are "supposed" to be responsible for childcare and with the expectation that women's childcare duties will impact on their work (for a review. see

Thompson & Walker ( 1989)).

The fact that the health status question was rated as more relevant to job performance than a question about family status may actually reflect the operation of social desirability factors. Participants may have felt that it was more socially acceptable to raise concems about whether a person's illness might impact on their job performance than it was to question whether one's family life is a potential hindrance. While both types of inquines are strongly discouraged in the policy on employment interviews outlined by the provincial Human Rights Commission ( 1997). the health status question may appear more justifiable, since it might be reframed by Iisteners as an expression of concem for another's physical well-being. Altematively. participants may not have been aware that it is not permissible for employers to pose health questions to candidates dunng an interview, even under conditions where employen pay for employees' health insurance coverage (as was the case here). On the other hand, participants did seem to recognize that it was not politically correct to rate the family status question as being "appropriate" or "relevant" in the context of a job interview; it was not uncommon for students to comment informal!y, upon completion of the expenment, that they knew that a prospective employee's farnily status was not permitted to be among an employer's considerations when evaluating candidates. Regardless, farnily status did in fact influence participants' ratings of the appiicants on 4 of the 5 principal dependent variables. It is these analyses that we tum to next.

Primarv Analyses

Table 12 surnrnarizes the statistically significant main effects and interactions

obtained for the five major dependent variables of interest. In this study, candidates who admitted to suffenng from asthma did not seem to bear the burden of discrimination; such

individuals were given middling ratings on the primary dependent variables whereas on the

hiring variable, the ordering of means suggested that they were marginally more likely to be hired than the candidate who wanted children (refer back to Table IO). It is unclear

whether the relatively positive ratings associated with this condition were due to the fact

that this particular illness was viewed as insuficiently serious to threaten Iong-term health

and availability for work or whether it reflects evaluators' for disabled persons,

which can sometimes prevail over more negative tendencies (e-g., underestimation of

abilities andor denial of the opponunity to prove cornpetence; see Colella, DeNisi, &

Varma, 1997 and Stone & Colella. 1996 for reviews). Either way, it is uniikely that

participants viewed the asthmatic woman simply as a woman whose reproductive

intentions were unknown because these medical issues had been made saiient; as a result, Table 12 Summaw of Sianificant Effects for Prirnary Analvses of Studv 2

Participant Sex p~.OOl p=.032 n.s. p<.001 M. F .LI. F .LI. F

FamiIfleal th pd17 ns. p<.001 pa5 pc.05 Status Don't Want .ilsltum Don? Wmt Wmt Ki& post hoc: ns Don'i Wyit .WUL liids Don'; Wyit..Wyit Ki& Cm't Have Wmi Ki&

Applicant Gender ns. ns. n.s. p=.055 F M

Familflealth x p<.O 17 ns. ns. A~~licantGender Wyit Lds: F- M Cyini HaW: F .hf interaction - Fanales: Wûnt Don'i Fd-: wmt- Cm't

Note: In order to prevent inflated alpha levels in the univariate analyses following the significant MANOVA on the "dedicated," "fit in" and "number of years" variables. a Bonferroni correction was used to make the significance levels more stringent (Le.. pC.0 17) on dependent variables # 1, #2 and #3. it is impossible to draw any fim conclusions about how people would rate individuais who do provide information about their intentions compared to those who do. A potentially miitful avenue for future research on statistical discrimination might be to include a condition where the applicant simply divulges that ihe does not yet have a family with no funher elaboration and to compare how targets in such a condition fare against the three family status conditions examined in this investigation.

With respect to the question of whether women were perceived less favourably than men were. only one of the dependent variables in the primary analyses yielded a

(marginally significant) main effect of applicant gender. This effect. on the "hiring" variable. was actually in the opposite direction to that expected; female candidates were slightly more likely to be hired than male candidates. This result nins counter to the findings of many earlier studies, which when subjected to a meta-analysis by Olian.

Schwab, and Haberfeld ( 1988), revealed a preference against women in hiring decisions.

While women as a group were not rated lower than men were on any of the major dependent variables. women who admitted that they wanted children were thought to be less dedicated to work than women who did not want or could not have children.

Furthemore. fernale applicants who wanted to have children were seen as less dedicated than male applicants who were planning to have a family. Conversely, femaie applicants who were unable to have children were viewed as more dedicated to work than similar male applicants. Thus not al1 women were being painted with the same brush - some women (i.e., those for whom there is no possibility or desire to bear children) were seen by participants as being every bit as dedicated to work as men. This differentid rating of prospective fernale employees on the dedication variable according to their child-bearing intentions suggests that participants may draw upon a "business woman" subtype. This subtype is characterized by high cornmitment to work, and a lack of personal and/or family obligations that could potentially interfere with career advancement (Deaux et al.. 198 5).

The pattern of results on the pnmary dependent variables do seem to suppon

Tornoskovic-Devey's (1993, 1995) proposition that employers strive to keep women out of workplaces simply because they are wornen. Indeed. there were no significant applicant gender or familylhealth status effects on the dependent measure (Le.. the "fit in" variable) that was included in order to assess taste discrimination. Nor is the lack of significant findings regarding target gender consistent with the sexist stereotype that women are generally less competent than men. Rather. the selective nature of the bias on the hiring variable (i.e., against candidates with familial obligations) suggests that pragmatic considerations, like maintaining profits by reducing turnover costs, motivate panicipants' decisions. Still, this is not to Say that in real-life setting, employers' decisions to exclude women from jobs are ody ever based on pragmatic factors related to predicted continuity of availability. Bielby and Baron (1986) have illustrated that employers also discriminate simply on the basis of old-fashioned sexist stereotypes (e.g.. they dismiss fernale applicants for male-dominated jobs without checking whether they can actuaiiy do the job).

At first blush. it seems accurate to Say that family Iife was perceived by panicipants as an impediment to being a good employee for men and women alike. In addition to the interaction between family status and applicant gender observed for the "dedication" variable, there were significant main effects for familyhealth status on the predicted length of organizational anachment, likelihood of hiring and saiary allocation. Basically, applicants who revealed dunng the interview that they wanted to start a family were rated less favourably than other candidates on the dependent variables measured. For instance, it was predicted that candidates who wanted children would stay with the hiring Company for a shorter amount of time than those who did not want or could not have children. On the hiring variable, none of the pair-wise post hoc comparisons achieved significance. however, the trend observed was that the applicant who expressed a desire for children had the lowest rating. Moreover, the mean salary offered to applicants who declared that they wanted a family was about $3.000 lower than that offered to the applicants who indicated that they did not want children.

The pattern of resuits obtained is in keeping with the indirect expression of gender bias that is characteristic of modem sexism. Instead of uniformly devaluing women across the dependent variables measured, a course of action that would be consonant with old- fashioned sexism, participants rated women as being as likely as men to "fit in" and to be hired. Thus consistent with modem sexism. there was no explicit expression of bias against women - they were evaluated as positively as men were. Yet at the same time. participants judged candidates who expressed an interest in the (stereotypically female) sphere of farnily life more harshly. Pnor research suggests that having a family enhances a man's image as an employee whereas the opposite is tme for wornen (Melamed. 1995;

Swiss & Walker, 1993). hence in actual workplace settings, the negative outcomes assigned to family-oriented candidates in this study would probably fa11 disproportionately upon female applicants who wanted families. This indirect means of disadvantaging women is the essence of modem sexism (Swim et al.. 1995; Tougas et ai.. 1995). A possible alternative explanation for the significant effect of family status on salary is provided by Rusbult. Lowery, Hubbard, Maravankin and Neises (1 988). They found that competent employees who were perceived as potentially mobile were compensated more generously than those who were perceived as unable to move as a result of obligations. The rationale that Rusbult et al. provided for this finding was that mobile employees might have been perceived as needing the extra monetary incentive in order to induce them to stay. Nevertheless. this interpretation is consistent with a modem sexism explanation wherein people endorse actions that "happen" to disadvantage women. given that most people understand that women with families are more Iikely than men with families to act as though they are entrapped (CARNET. 1992; cited in Lero & Johnson.

1994).

Moreover, the fact that information about candidates' reproductive intentions is rarely, if ever, available in job interviews may reduce &iwomen's chances of success when intewiewing for jobs similar to the one assessed in this investigation. As pointed out earlier, anti-discrimination legislation prohibits employers fiom probing candidates regarding farnily matters. This proscription could increase the likelihood that employers will discriminate against al1 women because it renders them unable to gather information that could counteract their stereotyped expectancies (the expectancies being that most women want to have a family. and that having a family will cause a femalr worker to be less dedicated. less productive, and eventually less committed to staying in her job). In effea, this legislation and its associated policies may make it more Iikely that employers will engage in statistical discrimination. As with the prelirninary analyses, some significant gender of participant effects emerged - men were less likely to rate candidates favourably with respect to dedication, and less likely to hire the applicant presented to thernm9Supplernentary analyses were conducted in order to explore the possibility that male participants were more likely than fernale participants to have had supe~soryexperience, and that this personal history of supervisory experience somehow influenced their evaluations on the dedication and hiring variables. But chi-square analyses revealed that male and fernale participants were equally likely to have acted in a supeMsory capacity at some point in their work history

(X:~l=0.075, m.),and while it was not feasible to incorporate supervisory expenence as an additional independent variable in the ANOVAs because ce11 sizes and statistical power would have been inadequate, covarying this variable pior to entering the independent variables did not account for a significant amount of variance in any of the dependent variables, nor did it change the pattern of results obtained.

The fact that main effects emerged for the gender of participant variable on some of the dependent variables was not seen as inconsistent with the interpretation that stereotypes about farnily responsibilities affect job candidates' fortunes with respect to hiring and salary. tnterpretation may have been more troublesome if there were significant interactions involving participant gender on either the hiring or salary dependent variables.

Under those circumstances. it would have been more difficult to tease out the impact that male participants' generally harsher evaluations of the candidate on the dimensions of

9 Recall that the Bonferroni correction employed rendered the gender of participant main effect for the "fit in" depetident variabIe non-signrficant, despite the fact that it obtained a p-value of -032 (new p-value had been set at p<.O17). confidence. making a good impression and handling the interview had on their salary and hiring judgments; further analyses would have been required. As the situation stands. there is little reason to conduct additional analyses on these data - the sexism interpretation offered appears sound.

OveraU, the findings obtained were in line with the expectation that sexism would influence the ratings of applicants. Although al1 applicants were rated above the midpoint on the various dependent variables regardless of their gender or family status, the fact that there were statistically significant diEerences among them indicates that participants did exhibit discriminatory behaviour. It should also be noted that participants were asked to rate the candidates on 7-point continuous variables: in the work world. the relevant outcomes are dichotomous (i.e.. one gets the job/raise/promotion one seeks or one does not). Therefore despite the relatively small eflect sizes observed for the gender of applicant and family/status effects on the continuous dependent variables in the primary analyses (ranging from q2=.~1 to q2=.06depending on which analysis is being considered). there is reason to believe that the findings obtained in Study 2 nevertheless

have meaningful implications in actuai hiring situations (Eagly. 1995; Ridgeway. 1997).

Essentially, the point here is that srnall advantages can be sufficient to result in success for one applicant and failure for the rest. Furthenore. the accumulation of little advantages and disadvantages can lead to huge differences in outcomes over the long term (Valian,

1998). as evidenced by the findings of a recent study on promotion practices (Martel].

Lane, & Emrich, 1996). Using a computer simulation, these investigators ascertained that

if one started with equal numbers of men and women at the Iowest tier of an ei@-Ievel

hierarchy and a bias in favour of promoting men that accounted for 1% of the variability in promotion, the highest level of the organization would eventually consist primady (65%) of men.

Limitations

It is acknowledged that the participants in this study were asked to rnake hiring decisions under conditions that were not identical to those encountered by employers.

First of alIl some participants indicated via their open-ended cornments that they would need to have information about dl of the candidates being considered for the job before they could make a decision about whether or not to hire the target person they had been presented with. lndeed the small effect sizes observed here could also be partiaily attributable to the fact that each participant rated only one candidate; Olian et al.'s ( 1988) rneta-analysis reveded that in hiring studies. larger gender effects emerged in investigations requinng subjects to assess the merits of several candidates than in studies that involved rating a sole candidate. Study 3 remedies this situation somewhat by allowing respondents to rate job applicants possessing a variety of different characteristics. thus it cm be constmed as having "multiple" candidates.

Second, there were no visual cues available to raters. Some of the participants indicated that they would also want to consider how the candidate presented hidherself with respect to attire, neatness, and so on before rnaking a final judgment regarding hiring; this point is well taken. Nevertheless, the omission of visual material was not perceived as threatening the vdidity of the finding that stereotypes influence employment decisions relating to access and value. Visual information would not have precluded the operation of the "family" stereotypes activated here. Rather, the presentation of visud information

has the potential to activate a wide array of additional stereotypes (e-g., attraaiveness. race, age etc.) that people use to classiq others according to personality traits and character types (Brewer & Lui, 1988). Thus it seems unlikely that the absence of visual stimuli left participants with so little information about the candidates that they had no other relevant knowledge except their internalized gender stereotypes to guide their decisions.

The differences between the studies conducted and real-life work situations can actually be refiamed as a strength of the investigation. These studies were conducted under circumstances where the participants were not going to be held responsible for the consequences of the salary and hiring decisions they made and were probably less invested than real employers would be in the task at hand; significant effects emerged nevertheless.

Thus these findings might in fact be underestirnating the problem of sexism.

Lastly, the sample used here was pnmarily cornposed of students enrolled in psychology courses - it could be argued that their background in psychology rendered them relatively sophisticated participants. They may have been motivated to try to figure out the study's hypotheses and possibly to comply with what they perceived to be the researcher's demands. Moreover, the individuals in this sample have achieved a fairly high level of education. Perhaps. given the tendency for education to be linked with more liberal attitudes (Kane. 1995). the responses obtained regarding hinng (access) and salary

(value) do not accurately capture the opinions or behavioural proclivities of the general population. In an effort to address these concems. Study 3 was conducted with a heterogeneous sample that was more representative of society at large and used a methodoiogy that minimized the potential for demand characteristics. Conclusions

The goal of this study was to explore whether impending family responsibilities impacted on participants' evaluations of job applicants with respect to dedication to work. predicted time before turnover, applicant's likelihood of fitting in at the hiring Company. hiring decision and salary allocation. The only way to address these issues was to present information to participants that is not generally available to prospective employers (at least not to those employers who follow the Ontario Human Rights Commission's (1997) guidelines for conducting interviews). Providing information about future family plans certainly infiuenced participants' ratings of the candidates, as evidenced by their less favourable evaiuations of candidates who wanted children as compared to thosr who did not want or could not have them on the dimensions of work dedication, predicted time until turnover and salary. Applicants who expressed a desire to stan a farnily were also slightly less likely to be hired. In short. it seemed that individuals drew upon stereotypes about family responsibilities when engaged in decision-making in the realms of hiring

(access) and wage setting (value).

The results of this study led to a consideration of how hurnan nghts legislation may have the paradoxical effect of rendering employers more likely to engage in statistical discrimination. By making it illegal to explicitly ask about marital and family status during the application, screening, and interview processes. this Iegislation may have the unintended effect of encouraging employers to rely on stereotypes about women's probable family aspirations and eventual lack of avaiiability for work in the long term. It is not much of a leap to suggest that these stereotypes could Iead employen to refrain from employing women, especially for high-paying professional jobs, thus contributing to the continued existence of the gender wage gap.

Interestingly, when significant effects emerged in Study 2, they were more often main effects for farnily status rather than interactions between family status and applicant gender; Study 2 participants tended to punish anyone. man or woman, who said that they wanted children and to reward anyone who claimed that did not want or were unable to have them. This lack of significant candidate gender effects fits well with a modem sexism interpretation. As Eberhardt and Fiske ( 1996) point out. modem sexists are unlikely to consciously acknowledge their negative feelings toward women. For young people located in the relatively progressive environment found in academe, the pressure to perceive themselves as egalitarian andor to refrain from espousing blatantly sexist views is likely to be particularly strong. Study 3 was conducted with a cornmunity sample in order to check how similar these undergraduates' opinions were to those of the larger society. Study 3

Unlike Study 2, which looked at participants' tendencies to display sexism during both hiring and wage setting, Study 3 was restricted to exploration of the potentiai for

hiring discrimination. Nine questions related to a prospective employee' s farnily st atm. living arrangements and comection to the community were embedded within a large-scale

cornrnunity survey, the Kitchener-Waterloo Metropolitan Area Survey (KWMAS) in an effort to determine how these charactenstics might relate to hinng practices. This survey

was conducted by researchers at the University of Waterloo Survey Research Centre in

the summer and fa11 of 1998. although it is anticipated that KWMAS will become an

annual survey of households in the Kitchener-Waterloo-Cambridge area. 'O

As already discussed, this study serves as a check on the responses given by Study

2 participants. As in Study 2, each participant was presented with either a male or female

candidate. Although the sarnple in Study 3 was asked about their opinion regarding the

advantage or disadvantage conferred upon candidates under nine different circurnstances.

of prirnary interest was their ratings of four different types of familial circurnstances, three

of which were identicai to those assessed in Study 2. The four family conditions were:

(1) candidate who expressed an interest in starting a family; (2) candidate who revealed

that slhe was unable to have children; (3) candidate who revealed that s/he did not want

children; and, (4) candidate who had three children. Once again, these candidates have

characteristics that are either in line with genetic stereotypes about women (i-e., that they

'" Funding for the nuvey came from the University of WaterlooiSSHRC Small Grants Program. together with financial contributions from SSWgrants held by individual facuity members involved in the project. are family oriented) or not (i.e., no family wantedhot possible). Sirnilar to Study 2. this study seeks to determine whether women are rated lower than men regardless of their family aspirations or whether female job candidates who appear to fdl into a particular subtype by virtue of their childless status (i. e., business woman) might experience less discrimination than other women. Lt should be noted that a sub-sample of managers was isolated and examined separately for the express purpose of uncovering the attitudes and opinions of individuals who were most likely to have jobs that required them to make hiring decisions.

The fact that these nine questions about employment were interspersed among a variety of other topics addresses the potential criticism that the experimental situation in

Study 2 created demand characteristics that participants found too compelling to resist.

While it was still possible that Study 3 respondents would try to ascertain the hypotheses underlying the questions and endeavour to be cooperative, the between-subjects manipulation of candidate gender made it impossible for them to present themselves as non-sexist by rating men and women sirnilarly. Study 3 also contained a within-subjects cornponent that made it possible to discem how people feel that different employee characteristics compare to one another.

Hvpotheses for Prima9 Analvses

The expectation was that the pattern of results for this study would be similar to that originally hypothesized for Study 2. Despite the fact that the expected patterns did not ernerge in Study 2. the nature of the sample in Study 3 made the expectation of significant interaction effects seem plausible. The cornmunity sample, by virtue of either their age or their greater expenence in the workforce was thought to be more tikely to display overtly sexist attitudes or simply to be resigned to "the way things are" in the

workplace. Thus it was expected that a main effect would emerge for family/health status

such that applicants who already had children or who expressed a desire to start a family

would be rated less favourably tban applicants who indicated that they did not want or

were unable to have children. It was also anticipated that the main effect for farnily

cocdition would be qualified by an interaction between gender of target and farnily

condition such that female applicants would be rated lower than male applicants only

under some conditions; a female prospective employee who either had a family or planned

to start a family would be rated as king at more of a disadvantage than a similar male

applicant. Funhermore. it was expected that male targets in the four family conditions

assessed would be rated similarly whereas female targets who are infertile or do not want

children will likely be rated more positively than the female target who wants or already

has a family.

With respect to gender of respondent effects. it seemed plausible that the women

in this general sample might be more acutely aware of the difficulties that employees

encounter when attempting to balance work and family life than the undergraduates who

participated in Study 2. As such female respondents might be expected to give more

extreme ratings of the advantage or disadvantage that each family condition may confer on job applicants. It was unclear whether to expect a two-way interaction between gender of

participant and famiiy conditio~or whether a three-way interaction might emerge wherein

women respondents rated female applicants as particularly affected by family status

considerations. Method

Power Calculations

This study's within-subjects component reduces the size of the sample needed to

satisfactorily test a 4~2x2analytical model similar to the one that was used in Study 2. Of the seven hundred and fifty individuals selected. it was expected that there would be a

sufficient number of usable responses to conduct the planned analyses.

Partici~ants

Seven hundred and fi@ names were drawn fiom computerized tax assessrnent

records for the cities of Waterloo. Kitchener. Cambridge. as well as Woolwich and North

Dumfries Townships in southern Ontario. A payment manipulation, designed to detect the

eRect of monetary incentives on response rates, was expected to lead to a relatively high

number of surveys being retumed. In fact. three hundred and seventy-five usable surveys

were retumed. for a response rate of 50%. Funhermore. 49 respondents ( 14.0%) were

currently employed in a managerial capacity while an additional 10 (2.8%) held

professional positions in the business or finance sectors, allowing for an assessrnent of the

attitudes of people whose actual work experiences could be expected to be sirnilar to the

hiring scenario presented in the survey. General background information about this

sample is sumrnarized in Table 13. Table 13 Background Information for Partici~antsin Studv 3

-N (%) Sex Female Male Age: M [sd] Current marital status Married Not married Have Children Yes No Number of Children: M [sd] Highea Levei of Education Achieved Less than high school High School Graduate Post-Secondary, non-university Bachelor degree Graduate degree Employment Status Full-time Pan-time Retired Student Not employed Occupational Type Managerial Business, Finance, Administrative Natural and Applied Science Hedth Occupations Social Science, Education, Govenunent SeMce & Religion . , Recreation & Sport Sales and Service Trades. Transport, Equipment Primary Industry Processing, Manufacturing & Utilities

Note: Not ail respondents provided demographic information, percentages reflect valid responses. Procedure

Researchers at the University of Waterloo Survey Research Centre pre-tested the entire survey instrument in five face-to-face sessions. Following revisions that consisted mainly of ornitting questions or portions of questions. a pilot version of the survey was mailed out. From the single mail-out to 100 Kitchener addresses, 9 were retumed by the post office as non-deliverable. and 36 completed returns arrived. for a 40% response rate

(i.e.,3619 1). Athough some survey questions were slightly revised at this point, no substantive changes were made to the questions related to employee characteristics (a copy of the final version of the items used for Study 3 appears in Appendix F).

The topics covered in the KWMAS were quite diverse. The Nne items that form the basis of this study were embedded roughly in the rniddle of the survey, among questions related to faimess of income, feelings in the workplace, attitudes toward minorities and their implications for attitudes toward affirmative action programs. social identity. language. environmental issues, and items designed to identifjr reasons for non- response on the survey. The KWMAS also included an experiment with cash incentives.

Three random cells of 250 each were created: Cell 1. which was a control group receiving no incentives; CeIl 2. a "post-paid cell in which 165 was offered to those who sent back a compieted questionnaire; and Ce11 3. the "prepaid" cell. in which $5 was included in the

main mailing. For each cell, there was an appropriate cover letter descnbing the incentive

or making no such mention for Ce11 1.

The survey was publicized through an article in a local newspaper (the KW

Record, Saturday, November 7'. 1998) and an interview on the morning program of a

local radio station (CKGL on November 1 1' 1998). About a week before the surveys were mailed out, a pre-contact letter was sent to the 750 people who had been selected.

Six weeks after the first mail-out, a post card follow-up was sent. A second mailing of the questio~airewith a new cover letter was undertaken about two weeks after the post card follow-up. Three weeks afier the second mailing, a final reminder Ietter was sent.

Not al1 of the 750 surveys sent out were deliverable to their intended recipients.

The proportion of post office retums was 2 1%. if al1 reasons for retum (including a few deceased) are included. The percentage of retums because addressee had moved was

16%. Analysis of the pretest data showed that the "moved" and "no such address" post office retums were disproportionately Iower SES. So as not to allow this source of bias to go unaddressed, a 50% random sample of the 120 post office returns was selected for a follow-up mailing. These surveys were addressed to "The occupant(s)" rather than to a narned individual. These 60 rnailings resulted in 10 completed questionnaires. 9 post office retums (now re-classified to "no such address"). and the remainder not heard from.

These additional 10 questionnaires were included in the analyses reported here.

Manipulations of interest to other investigators participating in the KWMAS led to the creation of eight versions of the entire 500+ item suwey. however, the between- subjects manipulation of the job candidate's gender is the only manipulation considered for the purposes of this study. As in Study 2, the job that respondents considered was that of assistant bank manager. Half of the respondents were sent a survey that asked about a hypotheticd male applicant for this position while the other half were asked about a hypothetical fernale applicant. This strategy was adopted in an effort to reduce the

Iikelihood that respondents would suspect that the purpose of our nine questions was to ascertain gender-discnrninatory attitudes. Measures

The nine items that form the basis of this investigation were preceded by a brief opening statement designed to encourage respondents to place themselves in the role of an employer seeking to hire an assistant bank manager. The statement read as follows:

Suppose you were a boss or supervisor who needed CO fil1 a vacant position at the organization where you work. To make the task easier, let's use a specific example - imagine that you work for one of the major Canadian banks and you are evaluating applicants for the position of assistant bank manager. One of the candidates for this position, Jane Smith [John Smith]. is a rnamed 30 year old woman [man] who has worked in a bank setting for a year prior to applying for this job.

We would like to get your impressions of the persona1 characteristics that would make Jane [John] more or less desirable as an employee. We are interested in your assessrnent of whether Jane [John] would be at an advantage or at a disadvantage as a result of having the following characteristics. Using the scale provided. please circle the number that corresponds most closely to your opinion.

Based on this information. they were asked to rate the hypothetical job candidate on a list

of nine charactenstics with respect to whether they viewed each attribute as an advantage

or a disadvantage for the applicant. Each item was rated on a 7-point scale that went from

-3 ("this characteristic is a disadvantage when applying for a job") to +3 ("this

charactenstic is an advantage when applying for a job"). A score of O was represented as

a neutral mid-point on the scde, which presumably denoted no effect of the charactenstic

on potential for hiring.

The three family status conditions assessed using a between-subjects design in

Study 2 (i-e., wants children, cannot have children. does not want children) were rated by

-al1 respondents in Study 3. The health condition frorn Study 2 was replaced with an

additionai "family" condition (Le., "Jane and her husband are the parents of 3 young

children"). These four items comprised the main dependent variables of interest. The reason for omitting the health condition was twofold: first, the survey format of Study 3 might have made it difficult to frame a health-related item in a way that would not pull for a sociaiiy desirable response. and second, it seemed more fiuitfùl to follow up Study Z by concentrating specifically on farniiy- and -related issues.

Four more items relating to the applicant's living arrangements and connection to the community (Le., "Jane lives in another city and must spend an hour commuting to work" "Jane and her husband belong to various clubs and organizations in the community." "Jane and her husband have extended farnily that live nearby," and "Jane and her husband live 500 kilometres from their nearest relatives") and an item related to eldercare (Le.. "Jane and her husband are rcsponsible for the care of an elderly parent") were also presented. The rationaie for including these items is outlined in the upcoming section pertaining to the analyses conducted; these five items appeared in the preliminary analyses, as descnbed below, but not in the pnrnary analyses.

Preview of Anaivses

An omnibus test on al1 9 employee characteristics was conducted for the entire sample using a mixed design 9 x 2 x 2 (Employee Characteristic x Gender of Respondent x Gender of Applicant) ANOVA. The purpose of this analysis was to determine whether any of the employee attributes other than the four farniiy-related conditions of interest were differentially rated based on gender of applicant or gender of respondent. In addition, this analysis made it possible to ascertain how the four fdystatus items compared to other personal characteristics with respect to how advantageous they were. A routine check on the question of whether the data met dl the assumptions for with-subjects ANOVA revealed that the sphericity assumption had been violated, a common occurrence in within-subjects designs according to Howell(1992). As a result. the degrees of fieedom for within-subjects effects in a11 of the analyses conducted for

Study 3 were corrected using the Greenhouse-Geisser correction.

Prirnaq Analvses

Next. the four farnily items that comprise the main focus of this study (i.e.. wants children, cannot have children. do not want children and already has three children) were analyzed using a mixed design 4 x 2 x 2 (Family Status x Gender of Respondent x Gender of Applicant) ANOVA in order to pinpoint how the family-related responsibilities and intentions compared to each other. This same analysis was then repeated for selected subsarnples. narnely, ( I ) al1 full-time non-managerial workers (n= 143). and (2) al1 managerial workers (n=58). The purpose of looking at full-time workers separately was to isolate individuals who might be expected to have a more accurate perception of the relative advantages and disadvantages of the vanous employee attributes exarnined in t his study because they may themselves be attempting to balance the demands of both full-time work and family life or are habitually in contact with people who endeavor to do so. The aim of isolating managerial workers was to obtain the perceptions of individuals who could be expected to expenence situations similar to the hinng scenario dunng the course of their own jobs. By redoing the analysis with these two groups, it was hoped that the findings of Study 3 would be more generalizable to reai-life situations because the subsamples' attitudes might more closely reflect the attitudes of individuals who rnake hiring decisions. As in Study 2, al1 ANOVAs were conducted using the regression approach to partitioning variance.

Given that the KWMAS survey included a great deal of information on a wide range of persona1 history and attitudinal variables. supplementary analyses that included sorne of these variables as covanates were plamed. The goal of these analyses was to assess the explanatory power of the independent variables in the onginai 4 x 2 x 2 (Farnily

Status x Gender of Respondent x Gender of Applicant) ANOVA mode1 fier identifying and including relevant respondent variables (resulting in a 4 x 2 x 2 ANCOVA).

Unfortunately, there was too much rnissing data on the additional variables to carry out prelirninary correlational or ANCOVA analyses for full-tirne non-rnanagenal workers and managers separately. Therefore. the plan was revised such that these supplementary analyses would be conducted ody for the sample as a whole. not for either of the subsamples.

Potentiai dernographic and attitudinal covanates were assessed via zero-order correlations in order to see whether they might be significantly associated with the four dependent variables. These correlations assessed the impact of 4 demographic variables

(respondents' own parental status. respondents' rnean annual income, occupational ski11 level, and education level) and 6 attitudinal variables (attitudes toward affinnative action prograrns, beliefs that female workers are disadvantaged by negative stereotypes, beliefs that women experience barriers in the workplace as well as the importance of three different social identities - farnily, gender and work - to the respondent) on candidate ratings. Correlations were run for all 16 cells since the assumptions of ANCOVA require that only those variables that are consistently correlated with advantageousness across ail

16 cells can legitimately fùnction as covariates. Moreover. these covariates are not permitted to interact with the independent variables. since such interactions would preclude the use of ANCOVA (Tabachnick & Fidell. 1989).

Of these ten potential covanates. only the education vanable was related to the dependent variable of salary across more than one or two of the 16 conditions (see Table

G-5 in Appendix G). However. the direction of the relationship was positive in some cells and negative in others. As a result. it was not possible to conduct the planned

ANCOVAs.

Results

Preliminaw Analyses

The means and standard deviations for the mixed design 9 x 2 x 2 (Employee

Characteristic x Gender of Respondent x Gender of Applicant) ANOVA conducted on the entire sample (n=322 usable responses) appear in Table 14 while the ANOVA summary table appears in Table G-1 of Appendix G. This analysis revealed significant main effects for employee charactenstic (E[J.«,~~,= 138.02. g<.OO 1 ; $ = .30)and gender of applicant

(Gl.3is,= 5.66. p<.05; q2 = .OZ). but not for respondent gender (F(lJls,= 2.43, n.s.). The two significant main eRects were qualified by an emplo yee characteristic b y appli cant gender interaction (E{5.a.inq = 14.18, g<. O0 1 : q2 = .04); this interaction is depicted in

Figure 12. There were no other significant interactions. Table 14 Mean Ratings of Advanta~eousnessof Emolovee FamiIfliving Characteristics bv Gender of A~oiicantand Gender of Participant: Al1 Respondents

Gender of Applicant

Male Female

Participant Sex Male Female Male Female Overall

Emolovee Char. 1.43 1.45 1.45 1.44 Belongs to clubs (1.15) (1.03) (1. il) (1.10)

Extended family 0.42 0.44 0.49 0.49 nearby (O. 9 5) (0.87) (1.01) (O. 97)

Can't have children o. 10 0.55 0.60 0.3 5 (0.53) (0.99) (1.12) (0.9 1)

Want children OS1 -0.13 -0.1 1 0.27 (1 .O7) (1 28) ( 1 .O4) ( 1.30)

Have 3 children 0.46 O. 15 -0.26 0.27 (1.14) (1.17) ( 1.27) ( 1.28)

Don't want children 0.08 0.32 0.55 0.25 (0.52) (1.01) (1.02) (0.93)

Eldercare O. 18 O. 13 -0.07 O. I5 (0.99) (O. 72) (O.99) (O. 99)

Live 500 km fiom -0.02 -0.03 -0.08 -0.06 nearest relatives (0.50) (O. 70) (0.7 1) (O. 64)

Comrnutes 1 hour -0.83 -0.93 - 1.29 -1 .O2 (1.10) (1.15)

Note: These items were rated on a 7-point scale that went from -3 ("this characteristic is a disadvantage when applying for a job") to +3 ("this characteristic is an advantage when applying for a job"), with a score of O representing a neutral mid-point. Inspection of Figure 12 suggests that the significant multivariate interaction was due to the fact that the four types of childcare responsibilities were rated differently depending on whether the applicant was male or female. For the other five items. ratings of whether the characteristic was advantageous or disadvantageous appeared to be similar regardless of applicant gender. Follow-up 2 x 2 (Gender of Respondent x Gender of

Applicant) univariate analyses on each of the nine employee characteristics. adjusted for inflated alpha Ievels using a Bonferroni correction (revised p&.OOS), confirmed the conclusions drawn on the basis of scanning the means. There were significant gender of applicant effects on the wants children (E(1.3 igl = 33.71. g<.00 1 ; q2 = . IO), cannot have children (E, 18, = 19.30.p<.00 1 ; = .06),already has three children (E(1.3 = 23 94 g<.OOi; ri' = .07)and do not want children (&1318) = 14.83. g<.OOl;ri2 = .Os) items.

Female applicants who wanted children or already had three children were rated as being at more of a disadvantage than male applicants with similar family aspirations or responsibiiities (MwinUd~.rn~= -0.12 versus M~j~rd~~*~~.= 0.62 on the "wants children" item and M~~,,~..~epl.= -0.07 versus M~-(r,.~~ihl.~~l.= OS9 on the "have three children" item). On the other hand, female applicants who were infertile or who did not want children were rated as being at more of an advantage than their male counterparts

(Mcan.lle..spPl.= 0.57 versus &n'vdhl+plppl. = 0.14 on the "unable to have children" item and

MDan.e,wppL= 0.45 versus MDon*~~hL.wl. = 0.05on the "do not want children" item).

Clearly, the farnily status conditions were not the only ones that respondents perceived as relevant to hiring. Connection to the community. as sigmfied by the item indicating that the candidate and hidher spouse belonged to various clubs and Figure 12 Interaction between Familvkiving Arraneements and Apolicant Gender: Al1 Respondents

Famiiy/Living Amnge me nts organizations, was perceived as quite advantageous (Mkl,, = 1.44) whereas the candidate's need to comrnute for an hour was seen as disadvantageous (McOmu,, = - 1 .Oz).

The differences between these two items and the family status conditions are quite a bit larger than the dinerences found between men and women on any given "family" item

(between 1 and 1'/1 points on a Fpoint scale as opposed to 0.4 to 0.5 points); even in the absence of forma1 tests, it is obvious that these rating dispanties are significant. Similar overall patterns ernerged for the non-rnanagenal full-time workers and for the managers when these subsamples were considered separately. The implications of these findings are described more fully in the discussion section.

Primarv Analvses

Ail Respondents

A mixed design 4 x 2 x 2 (Family Status x Gender of Respondent x Gender of

Applicant) ANOVA was carried out in order to ascertain whether the family status by applicant gender interaction was statistically significant even when the family condition items were isoiated. The ANOVA summary table for this analysis appears in Table G-2 of

Appendix G - refer back to Table 14 for the means and standard deviations. A significant main effect emerged for gender of applicant (I~Jz,= 5.52, p<.05; q2 = .02) such that men were rated at more of an advantage than women when ratings were averaged across the four dependent variables (MbLAppl.= 0.35 VS. MF..~~,.= 0.20). This main effect was qualified by a Fdmily status by applicant gender interaction (E(1.74.562)= 33 -79,pc.00 1; r12 =

.IO). There was also a significant fardy status by respondent gender interaction (& t.74.J62) = 3.81, gc.05; q2 = .Ol). These two-way interactions are depicted in Figures 13 and 14

(for the sample as a whole as well as for each of the subsamples tested). None of the other main effects or interactions reached significance.

Depiction of the ce11 means for the interaction between gender of target and family condition revealed a zig-zag pattern such that the farnily circumstances that were rated as advantageous for male targets were disadvantageous for female applicants and vice versa.

Analysis of the between-subjects simple main effects" of applicant gender at each level of farnily condition indicated statistically significant differences between male and fernale applicants at al1 four levels. Inability or lack of desire to have children were perceived as more advantageous for female applicants than for males (MhL.%pl.idCJn.r= 0.1 4 VS. - M~,.~l.kKm'c= 0.58 (51-33) = 14-42,e<.o 1 ) and M~l.+~l.id~on'i= 0.07 VS. M~.~l.pl.l-n't -

0.44 (E(l,ju,= 10.63, p<.O 1)). whereas wanting or already having children were rated as more advantageous for men than for women (Muwl.rùwani = 0.62 vs. MF.wi.ewan,= -0.13

(F(1323) = 4 1.49,g<.O 1) and M&L~~.C~~= 0.58 vs. M~..~i.ravc = -0.09 (Gljrt, = 33.88. g<.O 1)). The within-subjects simple main effects of farnily status for male and female applicants considered separately were also significant (l?, 1.7.1.288.6 1) = 1 7.7 1. pc.00 1 ) and

(E(1.13311.3.1)= 16.41. pC.00 1 respectively). Games-Howell compansonsL2revealed that for both male and female applicants, those who wanted or had children were viewed significantly differently from those who could not have or did not want children; for men,

II As per Howell(1992). the error term for these analyses \vas connnicted by pooling across the enor terms for the sepante 2 (Respondent Ses) s 2 (Applicant Gender) ANOVAs conducted at each of the four levefs of family mtus. '' According to Howell (1992). this post-hoc technique is called for in situations where the sphericity assumption has been violated in tvithin-subjects designs. Figure 13 Interaction between Respondent Gender and Famiiy Status

Al1 Respondents

0.8 - 4 - Male Respondents 0.6 +Fernale Respondents ; 0.4 ,,--a- 0.2 O . . Want Children Cm't Have Have 3 Don't Want C hildren Chilùren Children Family Status

Full-Time Non-Managers 0.8 - 4 - iWeRespondents CI8 Gs 0.6 +Fernale Respondents 0.4 ----a.Lp 0s *s=w a, 0.2 Y c a O --

. 1 . 1 a a 4.2 1 x 1 1 . Want Chilcfren Crui't Have Have 3 Don't Want Chilûren Chilàren Children Family Stiitus

Managers O.8 Male Respondents Om - * - CIe 0.6 - - +Femaie Respondents

Wmt Children Can't Have Have 3 ChrIdren Don't Want Children Chldren Family Stvtus Figure 14 Interaction between A~obcantGender and Farnily Status

Ail Respondents

Want Children Cm't Have Have 3 Don't Want Ciuldren C hilùrrn Children

Full-Tirne Non-Managers

Wmt Chlh Can't Have Have 3 Clhildren Don? Want Children Chi ldren

Managers

Want Children Cari't Have Have 3 Children Dont Want Childreii Ciuldren Farnily Status the first two conditions were seen as significantly more advantageous than the latter two

((M~~l.(igwmt= 0.62. MM.~I.IÜW,~~= 0.58 as compared to MM~I.~C,.~= 0.14.

MhLAppl.wn*t= 0.07). whereas for women. the pattern was reversed (Mawi.aufant= -0.13.

ME~I.~~~~~= -0.09, as opposed to M~.Appl.id~an'l= 0.58. M~.~~.pi.ran~i= 0.44).

With respect to the interaction between gender of respondent and family condition. the within-subjects simple main effects of family status failed to reach significance when male and female respondents were considered separately. It appears that the interaction effect can be traçed to the discrepant ratings given to the applicant who has a family - this difference was about .35 on a 7-point scale while the other three differences ranged between .O2and -13. An F-test confirms that the simple effect of respondent gender at

"have 3 children" is significant (& 1.3~)= 10.38. ~<.005).Male respondents perceived that having a family was significantly more advantageous for a job applicant than female respondents did (M~LR-.(~== 0.45 vs. MF..*~I.~I(J~~= 0. IO). None of the other between- subjects simple effects achieved statistical significance.

Full-Time Non-Manaeenal Workers

The 4 x 2 x 2 analysis conducted on subsarnple of full-time worken in non- manageriai roles (n= 143) revealed results similar to those obtained for the sample as a whole (see Table 15 for means and Table G-3 in Appendix G for ANOVA summary table). There was still a significant interaction between gender of applicant and family condition (E(I.T).zJo)= 10.73. g<.OOi; ri2 = .07)and the form of this interaction was similar to that observed for the whole sample. As before. inability to have children was perceived as more beneficiai for female applicants than for male applicants (MCan*val.wl.= O-13 VS.

&I*vg.m~. = 0.55; 51.1~9)= 6.03, gc.0 1). whereas wanting or already having children Table 15 Mean Ratines of Advantatzeousness of Emplovee Familv Charactenstics by Gender of Apolicant and Gender of Participant: Full-Time non-managers onlv

Gender of Applicant

Male Femaf e

Participant Sex Male Female Male Fernale Overail -M M -M -M -M (sa (sa (4) (Sb) n=4 1 n=3 8 n=3 7 n=27 n= 143 Emolovee Char. O. 17 O. 08 0.59 0.48 0.3 1 Can't have children (0.86) (0.36) ( 1.07) (0.98) (O.87)

Want chikiren 0.85 0.32 -0.08 -0.04 0.30 (1.15) (1.07) (1.36) (0.76) (1.18)

Have 3 children 0.56 0.34 0.38 -0.36 0.30 (1.25) (1.10) (1.16) (0.94) (1.16)

Don't want children 0.12 0.08 0.32 0.44 0.23 (0.98) (0.36) (1.06) (0.85) (O.86)

Note: These items were rated on a 7-point scale that went fiom -3 ("this characteristic is a disadvantage when applying for a job") to +3 ("this characteristic is an advantage when applying for a job"). with a score of O representing a neutral mid-point. were rated as more beneficial for male applicants than for female applicants (Mw~nUài~~.~~~.

= 0.59 vs. MW~,,~~..~~.= -0.06; &1,139) = 15.96, p<.001.and MH3vda~l\ppl. = 0.46 vs.

Mhvd~,rnppl.= 0.1 1 ; 1, 139) = 4.8 1. fi<. 05). The difference between male and female applicants who did not want children failed to achieve statistical significance. The within- subjects simple main effects of family status for male and female applicants considered separately were also significant in this subsample (E(1.63.126.07) = 6.54. g<.005 and ~(1.78.11~.47)

= 4.34, pX.05 respectively). Post-hoc cornparisons made using the Games-Howell method showed that perceptions of male applicants who wanted children were more positive than perceptions of men who did not want children or were infertile (MWlnild\LAppl. = 0.59 as opposed to MDonlirdhLhppl.= 0.10. m~n.V~~~.,-vpl.= O. 13). Conversely, for female applicants. those who wanted a farnily were at a disadvantage relative to women who did not want or could not have children (MwJnus..9pl. = -0.06 vs. MDon.umF.:mI. = 0.38. Mc,n*umn.wi.= 0.55.

Women who had a family were also rated less favourably than women who were infertile

(M~LMZF..+~I.= O. 1 1 VS. MC*~*V*..+~I.= 0.5 5 ).

The interaction between gender of respondent and family condition was nonsignificant for this subsample (E(1.73.240.15)= 1.46, -) although a marginal main effect of gender of respondent emerged (Hi.139i = 3.72. g=.056; q' = -03). Male respondents' overall ratings tended to be somewhat higher than those of the females (MM.Rep.= 0.37 vs.

MF.~~.= 0-19). None of the remaining main or interaction effects proved to be. significant .

Managers

The sarne 4 x 2 x 2 analysis was conducted for the respondents who were managers. Means and standard deviations for this analysis are provided in Table 16. Table 16 Mean Ratines of Advantageousness of Employee Familv Characteristics bv Gender of A~plicantand Gender of Partici ant: Manasers onlv

Gender of Applicant

Male Femde

Participant Sex Male Female Male Fernde Overall -M M -M -M -lb t (sa (sol (sd) (sd) (sa n=17 n=12 n= 18 n=l 1 n=58 Emdovee Char. 0.24 0.33 0.28 O. 64 O. 34 Can't have children (0.75) (0.89) (0.67) (1.12) (0.83)

Want c hildren 0.65 0.33 -0.11 -0.5 5 O. 13 (1.58) (0.89) (0.76) (1 .04) (1.19)

Have 3 children 0.65 0.25 0.1 1 -0.18 0.34 (1.46) (0.97) (0.96) (1.17) (1.17)

Don't want children 0.12 O. 17 0.28 O.73 0.29 (0.70) (0.58) (0.67) (1.10) (O.77)

Note: These items were rated on a 7-point scale that went fiom -3 ("this characteristic is a disadvantage when applying for a job") to +3 ("this characteristic is an advantage when applying for a job"), with a score ofO representing a neutral mid-point. The ANOVA surnmary table appears in Appendix G (Table G-4). This time. the only significant effect was the interaction between gender of applicant and family condition

(&,. J7.84.65) = 4.62, ~c.05;q2 = .08). The fom of this interaction was comparable to those observed for full-tirne non-managerial workers and the sample as a whole, although the rnean ratings for the "wants children" condition appeared more polarized whereas the ratings for the "cannot have" condition were more sirnilar. Analysis of the between- subjects simple main eKects of appiicant gender at each level of family condition indicated only one statistically significant difference between male and fernale applicants: wanting children was rated as more beneficial for men than for women (MW.,,U~L~~.= 0.52 VS.

Mw~~*~.~~.= -0.28 (F(I.34) = 9.60, e<.Ol)). The within-subjects simple main effects of family status proved nonsignificant for men but it was significant for women (Ff159..1Z.w,-

3.93, eC.05). As before. comparisons were made using the Games-Howell technique; these analyses showed that female applicants who wanted children were seen to be at a significant disadvantage (MF..wI.aWJI= -0.28) relative to women who could not have

(MF.r\pel.@CJn*L= 0.41) or did not want children (ME.q,pl.~mm-i= 0.45).

The lack of a significant interaction effect for family status x participant gender among managers is Likely a power problem attributable to the small sample size and unequal numbers of male and female managers. Table 16 shows that there were 3 5 male and 23 female managers respectively, resulting in very small ce11 sizes of only 1 1 and 12 for female managers. Examination of Figure 13 suggests that the differences between male and female managers' ratinss of the various family constellations appear to be at least as large in magnitude as those provided by non-managerial full-time workers. In fact, if one focuses on the femaie respondents in particular. female managen' ratings actually appeared more extreme than the ratings given by female tùll-time non-managerial workers.

ANOVAs cornparhg managers to non-managerial full-time workers were conducted as planned but their results should be interpreted with caution given the smali sarnple size problem cited above. These ANOVAs, run separately for male and female respondents. revealed no significant differences between sub-sarnples.

Discussion

Prelirninary Analvses

Looking at the entire sample, the rnean ratings of the applicant on the scale measunng advantageousness were close to the neutral point for al1 four scenarios dealing with children. Means for the other family-related items (Le.. eldercare. distance fiom extended family) aiso hovered close to neutral. On the other hand. average ratings were distinctly positive for the "belongs to clubs" item and clearly negative for the "commutes for an hou? item. Whde it is easy to see how the cornmunity involvernent co~otedby club membership would be seen as advantageous From the perspective of a!employer. it is a bit pwzling that respondents evduated an applicant's residence in another city and resultant cornmuter status as so disadvantageous. Unfortunately, space limitations in the survey precluded probing for the reasons underlying the ratings assigned. Perhaps expressing doubts about a prospective employee because of hidher need to commute seemed less suggestive of discrimination to respondents than, for example. assening that family matters would impact on hiring. It is possible that respondents could justify giving lower ratings in the comrnuter situation by appealing to how inconvenient the emplo~ee might find it to spend so much time in transit. whereas they may have been reluctant to rate the family situations very negatively because they were aware that the latter stance would be inconsistent with workplace anti-discrimination policies/legislation.

Altematively they may simply have wanted to avoid appearing sexist andor "anti-family."

The fact that the ratings for the farnily intentions/constellation items were close to the midpoint and less disadvantageous than cornmuter status should not be interpreted as reflecting respondents' perception that farnily matters are irrelevant to hiring. Whether one considers the results for the sample as a whole or the results for either of the subsamples, a prospective employee's current or anticipated childcare responsibilities were seen as pertinent. and differentially so. depending on the applicant's gender, as we shall see in the next section.

Prim- Analvses

Impact of Applicant's Gender

Table 17 sumarizes the significant main effects and interactions obtained in Study

3. As anticipated, the four family-related conditions of interest (i.e.. have 3 children. want children, do not want children, cannot have children) were rated differently for male and fernale job candidates in both the full-time non-managerial sample and the sample as a whole. Furthemore, although small sample size rendered it difficult to obtain statistical significance for the analyses conducted with the manager subsarnple, the finding that male applicants who wanted a family were rated more highly than similar female applicants nevertheless emerged for this subgroup. These results are consistent with working women's claims that they are disadvantaged in hiring situations relative to men if they have family responsibilities or aspirations (Swiss & Walker. 1993). Table 17 Summarv of Simificant Effects for Primary Analyses of Studv 3

-- - - sarnple Al1 Respondents Full-Tirne Managers Effects (n=3 27) Non-managers (n=58)

Famrly Status x p<.m Participant Sex Have 3 iüdx Itf F Wyit WdS: &S. Interaction Cm't Have: as. Don't Wmt:- ns. Males: n.s. Fmlcs: ILS.

Family Status x p<.001 p~.OOi pc.05 Applicant Gender Havc 3 Ki&: XI F Have 3 Kids: 51 F Have 3 iüds: as. Wyit luds: Bi F Wmr fi&: hl F Wyit kds: 11 F interaction Cyi't Have: hl F Cmui'tHave: XI- F Cm't Have: as. Don? Wyit:- 51 F Don't Wmt:- hl F Don't WWI:- n.s. hldcs: W;~lt=Have Canlt=Don't Males: Wmt Can't=Don'i Malrs: as. Fcmala: WyitmHave Ca.n't=Don't Fmlcs: Wyii Cm't=Don't Fanala: Wmv C;in't=Don't Frnwla: Have Cm't Between- Subjects Participant Stx Marginal (pe. 10) M F

Applicant Gender Also as hypothesized, the entire sample and the full-time nonmanagerial workers rated female targets who had or wanted a family lower than those who were unable or lacked the desire to have a family. This result is reminiscent of one obtained in Study 2. where undergraduates rated the female candidate who indicated that she was planning to have a tàmily lower on dedication than female applicants who said that they did not want or could not have children. As alluded to in the discussion of Study 2 results, the relatively positive ratings assigned to nonrnothers may point to the drawing upon of a

"business woman" subtype (Deaux et al., 1985). These women are seen as devoted to their careers and as such are better risks when it cornes to hiring, at least corn an employer's perspective. In effect. Study 3 respondents' predictions of which types of women would (or would not) be in a position to gain access to managerial jobs matches exactly what one would expect if women's unpredictability rather than presumptions of their incornpetence or taste discrimination influences employers to refrain fiom hinng them. In other words, women who aspire to be mothers. or who already are mothers. are in a disadvantaged position.

The pattern was exactly reversed for male candidates - having or wanting a family resulted in higher ratings whereas it was perceived as less advantageous for men to not want or not be able to have a family. Again. these findings were as expected; it has fiequently been suggested that being a "family man" malces male candidates more appealing to employers as this orientation connotes a stable and responsible individual

(Melamed. 1995). In fact, it seems that having no children can actually be sornewhat of a liability for men. Lampman and Dowiing-Guyer ( 1995) discovered that raters saw an individual's Iack of desire for children as indicative of lower ambition (although inability to have children was not). Calhoun and Selby ( 1980) obtained slightly different results. finding that childless men were perceived unfavourably. regardless of whether this state was voluntary or involuntary.

hterestingly, even when men possessed charactenstics that are perceived as impediments. their impact was more ofken conceived of as "less advantageous" as opposed to distinctly "disadvantageous," as was the case for women. Figure 14 shows that across the (sub)samples, the lowest of the mean ratings for male candidates always remained above the midpoint of the scale whereas mean ratings for women who wanted or who had families Frequently fell below the rnidpoint. Taken together, these findings support the contention that wornen's greater responsibility for family matters constitutes a hindrance in the work world.

It seems safe to assume that respondents believe that women's access to certain jobs is restricted on the basis of their family responsibilities. Perhaps more imponantly, respondents who worked as managers also believed this to be the case - the strategy of examining the responses of a managerial subsample permits increased confidence that the findings accurately reflect the impact that çender stereotypes have in the workplace.

Indeed. the pattern of responding suggested that fernale managers' views in particular were more eareme than those of nonmanagerial workers. Unfortunately, srnall sample size prevented the drawing of firm conclusions about any differences between managers and others.

There seems to be little reason to doubt that the respondents in this study are tmtffilly reporting what they perceive to be the advantages and disadvantages of various candidate charactenstics. In real-life work settings, research suggests that people who do not themselves endorse discriminatory beliefs may still display discriminatory behaviours in order to advance their own position at their workplace andor to benefit their Company if they discem that the generic business nom or their specific workplace's nom is to be discriminatory (Smajkowski & Larwood. 199 1 ; Trentham & Larwood. 19%). However. this finding does not seem to account for the results observed here. No benefit accrues to the study respondents as a result of providing ratings that connote gender discrimination.

There is no boss in this situation to please by taking a sexist stance that they do not tnily endorse or to hold them accountable for any sexism-tinged responses. Also, given the hodgepodge of questions presented in the KWMAS survey as a whole. it is unlikely that the results of Study 3 can be explained away as due to demand characteristics - there were too many different topics addressed to be able to get an accurate handle on what the

"right" answers "should" be for the subset of nine questions that fomed the basis of Study

3. Furthemore, respondents probably have lower motivation to provide answers that correspond to what they think the researchers' hypotheses are in the context of opinion surveys than they (arguably) have in experimental situations.

lm~actof Resoondent's Gender

Significant effects involving respondent gender materialized, although they were less reliably uncovered than those involving applicant gender were. For the sample as a whole, there was a respondent gender by family condition interaction, albeit the only statistically significant simple main effect occurred for the "have children" item such that male respondents assigned higher ratings than female respondents did. This result might be attributable to male and fernale participants' actual expenences in the workplace. Men may have rated having a family as relatively advantageous because for them, it usually is - being a "family man" does not result in negative attributions about their work commitment

(in fact, quite the opposite is tme; Lamprnan & Dowling-Guyer. 1995) and often cames positive consequences for them in terms of salary (Blau & Beller, 1988; Cooney &

Uhlenberg, 199 1; Hersch, 1985). Female respondents may have rated this item as they did because they were sensitive to the fact that women's family responsibilities make them appear less committed to their employers (Parasuraman & Greenhaus, 1993).

Once full-time non-managerial workers were isolated, there was a tendency for female respondents to give lower ratings than men did for al1 conditions (Le., a marginaily significant participant gender main effect). Figure 13 shows that women's ratings of candidate's advantageous circumstances (i.e.. inability or lack of desire to have a family) were somewhat attenuated whereas disadvantageous circumstances (i.e.. having or wanting a family) were slightly exacerbated. Again. it is possible that participants' responses reflect their own expenences in the workplace. Overall. men are favoured in hiring (Olian et al., 1988) and reap greater financial advantages in the workplace regardless of their family situation (Drolet. 1999) whereas family life greatly influences how women are perceived at work (Swiss & Walker. 1993).

Finally, separate analyses for managers revealed no significant effects involving respondent gender despite the fact that depiction of the means in Figure 13 would suggest that a respondent gender by farnily status exists. It seems quite likely that this interaction failed to reach significance due to insufficient power - recall that the ce11 sizes were relatively small for these analyses (refer back to Table 16). The pattern of ratings by female managers actually looked more extreme in terms of its highs and lows than the pattern of results obtained for the whole sample or the Ml-time non-managerial workers. The female managers' opinions could quite conceivably reflect this group's acute understanding of how negatively employers view women's farnily responsibilities. especially at higher levels of work hierarchies. Such an understanding could be the result of managers' greater accuracy in judging how rnuch family life cm interfixe with work based on their observation of other employees struggling with these issues, or it could sirnply reflect their own personal experiences in attempting to stnke this balance (75% of the female managers had children). As already mentioned, analyses geared toward assessing differences between managers and non-managerial workers did not uncover statistically significant effects, probably due to the power issue noted above.

Differences between Study 2 and 3

As outlined in the introduction to Study 3. this investigation was designed to permit cornpansons with the results obtained in Study 2. Direct cornparison is only possible on the issue of access to jobs, aven that the survey format could not allow for the more extensive questioning (e.g.. on issues of value. dedication, turnover, etc.) that occurred in the experimental study. The first difference that should be noted is that the effect sizes for the significant findings obtained in Study 3 wete somewhat larger than those observed in the other two studies. The effects observed here accounted for between

7% and 10% of variance. whereas the effects in Study 2 explained between 1% and 6% of variance. As suggested in the discussion following Study 2. it is not unusual for between- subjects and within-subjects designs that address siinilar questions to yield differential effect sizes, with within-subjects designs generally producing effects that are larger in magnitude (Olian et al., 1988). The manner in which applicant gender made its influence felt certainiy differed in

Studies 2 and 3. The community sample appeared to draw upon more differentiated stereotypes than did the undergraduate participants in Study 2. Specifically, the high advantageousness ratings assigned to the women who did not want or could not have a fainiiy suggested that this sample drew upon the subtype businesslcareer woman when assigning their ratings. The Study 2 participants made evaluations t hat were consistent with this subtype on the dedication dependent variable. however, they were far more likely to punish both females and males who express family inclinations on the vanous dependent variables. This pattern was identified as in keeping with modem sexism. whereby the disadvantaging of women is indirect rather than blatant. The results obtained in Study 3 fdl more clearly within the rubric of old-fashioned sexism as they reflect the notion that women are primarily responsible for childcare within families whereas men are breadwi~ers.

tt is possible that the differences that were observed between Study 2 and 3 could reflect the differences in the task asked of the participants. Specifically. Study 2 asked participants how they would treat the candidate. based on the information they obtained over the course of the mock interview. Study 3 asked for respondents' opinions regarding how advantageous or detnmental particular family characteristics were without asking them to commit to a particular course of action one way or the other.

The different patterns observed in Studies 2 and 3 could also quite conceivably be chalked up to a cohon difference. AI1 of the Study 2 panicipants were aged 25 and younger whereas the average age of the Study 3 participants was aimost double that.

Also, most of the Study 2 participants were full-time students and unlikely to have children" whereas the majority of Study 3 participants were parents (74%); basically. their life experiences were quite different. But does the essence of the difference between the cohorts lie in their wiilingness to publicly endorse old-fashioned sexist ideas (resulting in tendencies toward modem sexism in the younger cohort) or. is it that the younger cohort of Study 2 is truly more egalitarian in its outlook. if naïve?

This latter possibility probably should not be disrnissed out of hand. To wit. perhaps Study 2 participants do not make the assurnption that women are the oniy ones whose family life can impact on their work. They may believe that family responsibilities are (or should be) more equally shared between men and women - indeed such egalitarian ideals have been expressed in pnor research assessing attitudes toward domestic duties

(see Crosby, Farrell. & Cameron. 1994 for a review). As a result. Study 2 participants could have seen family-work confiicts as equally likely for both male and female employees. which would explain their tendency to rate both women and men who aspired to family life less favourably on the hiring (access) variable, and to predict that family- oriented candidates will stay with the hiring Company for a shorter length of time. Study 3 respondents may have a different understanding of the potential impact of family life on work life by virtue of having to engage in this balance themselves ancilor observing women's patterns of workforce attachment firsthand. These older respondents' ratings could reflect their understanding that women end up assuming the bulk of childcare and general domestic responsibilities (Crosby & Jaskar. 1993; Duxbury & Higgins, 199 1;

Leslie. Anderson, & Branson. 1991). even if they do not necessarily believe that this

l3 Admittedly. they rvere not directiy asked if they were parents or not.

158 "should be the case. However, refiaming Study 2's younger cohori as "naïve egalitarians" does not fit well with the finding that those participants penalized the candidate who wanted a family with respect to starting salary on the order of thousands of dollars. In fùture research, the administration of sexism inventories of both the old- fashioned and modem variety could help clariQ respondents' underlying attitudinal base. GeneraI Discussion

Summarv of Findines

This dissertation demonstrated how stereotypes contribute to a state of affairs wherein female employees and women's work in general continue to be perceived negatively. The stereotypes exarnined were reflective of two phenornena that have repeatedly been cited as major contnbutors to the gender wage gap: segregation of the workforce by gender and childcare responsibilities that faIl predorninantly to women. The studies conducted tested whether stereotypes pertaining to the gender composition of occupations (Study 1) and to women's family responsibilities. specifically childcare

(Studies 2 and 3). promote sexism dunng both hiring (access) and wage-setting (value) that eventually gets manifested as a gender gap in wages.

While some investigators believe that gender segregation in the workplace will decrease in the corning years (Jacobsen, 1994). the fact remains that segregation at the level of jobs and occupations continues to be a prominent feature of the employment landscape (Drolet, 1999). Study 1 showed that male and female participants asked to evaluate jobs at low, medium or high income levels perceived stereotypically female jobs to be less valuable than comparable male jobs. as evidenced by their unequal salary allocations for jobs that entailed identical tasks but were located in different gender domains. These results are in keeping with what some theorists have called "old- fashioned" sexism - a stance that entails the explicit devaluation of women and female- typed activities (Cameron, 1977). Clearly. the association between women's work and

Iow worth continues to hold, making it fairly safe to predict that as long as the workforce is charactenzed by gender segregation, there will be a male-female wage gap. Studies 2 and 3 explored whether participants' evaluations of fernaie job candidates with stereotypical or non-stereotypical family aspirations reflected sexist of either the old-fashioned or modem variety. Study 2 revealed that women as a group were rated no less favourably than men on the five dependent variables assessed. indeed there was a slight advantage for female candidates in Study 2 with respect io hiring. These findings ied to the conclusion that women and men were viewed as being equaily capable of doing the job in question (Le.. assistant bank manager), and hinted that old-fashioned sexist ideas about women's infenor cornpetence are sornewhat less prevalent than they have been in the past. at least arnong undergraduate students.

Despite the fact that panicipants did not blatantly discriminate against women in their evaluations, they did display bias against candidates who were inclined towards the stereotypically ferninine realm of family. Specifically, female candidates who said that they wanted a family were seen as less dedicated to their work than similar men and less dedicated than women who either did not want or could not have children. In addition. the family status main effects uncovered on the predicted turnover. likelihood of hinng and salary allocation dependent variables showed that job candidates who expressed a desire for farnily Iife were consistently penalized. regardless of their gender. This pattern of results is rerniniscent of the indirect disadvantaging of wornen that is the hallmark of modem sexisrn; there was no blatant underrating of women. although the fact that domestic life and childcare is more typically associated with women in Nonh American society suggests that participants' lower hiring ratings and salary allotments will have more negative implications for female prospective employees than for males. Uniike the student sample in Study 2, the community sample surveyed in Study 3 made the direct co~ectionbetween gender and family life - according to these

respondents, women who had or wanted children were at a disadvantage in hinng

situations relative to similar men, consistent with stereotypes that women's fily

cornrnitrnents interfere with paid work whereas men's do not. Furthemore, female

candidates who did not want or could not have children were rated as being advantaged

relative to men with comparable characteristics as well as relative to women with family

responsibilities or inclinations. These results indicate that old-fashioned sexist ideas

relating to the gendered division of labour. wherein the man of the household assumes the

role of breadwimer while the woman assumes the role of child care provider. linger in

today's society. People apparently still see women's familial role as rendenng them

unsuitable for jobs that require a great deal of tirne commitment. thus it seems reasonable

to suggest that as a result. women will continue to be denied access to many higher paying jobs that are currently dorninated by men. Joshi (199 1) articulates this idea nicely: "in a

more general sense, the gender gap is probably the consequence of potential motherhood.

given the prevailing domestic division of labour" (p. 190).

Taken together, the results of these three studies suggest that women are

disadvantaged in the workplace in at least two ways: work that is perceived as being

female-typed is devalued; and. female job applicants are prone to be viewed as having

family interests or commitments that will interfere with their paid employment. In effect,

women are handicapped regardless of whether the job they seek is in a traditionally

"female" reaim or not. This dissertation also responded to the cal1 for research that delineates the conditions under which stereotypes are liable to lead to discriminatory behaviour (Fiske.

1998) by elucidating how provincial and federal anti-discrimination legislation can facilitate stereotypical thinking by individual employers. thus helping to maintain the gender wage gap. Studies 2 and 3 show that people adhere to the stereotype that family life and work are incompatible for fernale employees. as evidenced by less favourable evaluations for women who want children on the dedication dependent variable in Study 2 and the lower ratings çiven to women who want or have children in Study 3. However.

raters cm put aside this stereotype if they obtain individuating information suggesting that no work-farnily confiict will arise for a particular woman. More to the point, benefits

ensue for those who either cannot have or do not want children.

Investigators in the area have long recognized that the provision of sufficient

individuating information that runs counter to stereotyped expectations may offset the

tendency toward bias (Pratto & Bargh, 199 1 ; Seta & Seta, 1993). However, the

provincial Human Rights Code forbids employers from inquiring about a prospective

employee's marital or fhlystatus since this type of information has the potential to be

used inappropnately (e.g.. to screen out job applicants). Once again, it must be

emphasized that the goals of the current anti-discrimination legislation are sound: the

problem lies in the unintended side effects of the laws. It may be an unforninate

consequence of the legislation that & women are likely to be viewed by employers as

potentidly unreliable long-term workers and perhaps prone to discriminatory treatment

because of the prohibition on obtaining certain types of information. Limitations

It is acknowledged that the studies conducted for this dissertation did not comprise perfect simulations of hiring and wage setting situations. On one hand, the dependent variables used in these three studies (i.e., salary allotments. hiring-related appraisals) may be constmed as reflective of participants' behavioural intentions. As alluded to earlier. t here is a well-established tradition of research t hat relies upon simulations of decision- making scenarios - the assumption that the results obtained under these circurnstances mhor reality is fiequently made (e.g., MacDonald & Zanna, 1998). As such. the results can be viewed as illustrative of how individuals' sexist perceptions foster decisions that are eventually enshrined in sexist institutional practices (Auster, 1989). Altematively, as discussed in the context of Study 3. the results of these studies rnay demonstrate participants' understanding and echoing of business and/or societal noms but not

necessady support for said norms (Szwajkowski & Lanvood, 199 1; Trentham &

Lanvood, 1998). In future tesearch. it might be helpful to elucidate the underlying

motivation for discriminatory behaviour by administering a sexism inventory. Yet even if

the results obtained simply reflect participants' knowledge of sexist workplace norms

rather than their actual endorsement of sexist beliefs, the fact remains that there exist

sexist norms of which participants become aware. Basically. the discovery that the

hypothesized effects for gender of job. gender of candidate and applicant's family status

were statistically significant indicates that one way or another. sexism continues to malie

its presence felt in employment settings.

Still, this dissertation does not by any stretch of the imagination constinite the last

word on the role of sexism in the gender wage gap; the debate about how much of the gap can actuaiiy be attributed to individual-level sexism will likely continue to rage. The reason this is the case is because it is virtually impossible to operationalize and directly measure occurrences of gender-based discrimination in people's work lives. in effect, those who study the gender gap are in a no-win situation -regardless of the research methodology they use, it will never be possible to know with certainty exactly how employers make the hiring and wage-setting decisions that are suspected to contribute to the gap. On the one hand, large-scale investigations of income and hiring/promotion using employees' retrospective self-reports or corporate records cannot uncover specific events where sexism played a role. On the other hand, small-scale attempts to reenact employment situations while exercising experimental control are ofien critiqued For their lack of extemai validity . Funhermore. directly asking employers about their decision- making processes is not likeiy to be helpful because they have little inclination to admit to discriminatory acts and, according to theories of modem and neo-sexism, may not even be consciously aware that they behave in discriminatory ways (Eberhardt & Fiske, 1996;

Swim et al., 1995; Tougas et al., 1995).

Ultimately, there is no way to ascertain how closely the findings of this investigation rnirror workplace occurrences because discrimination is such a slippery phenornenon. However. this dissertation was built upon the foundation of a multi-method approach, which permits more confidence that the results obtained provide a fairly accurate picnire of reality. In other words. the research problem was addressed using diverse sarnples (Le.. undergraduate and community samples. as well as a managenal sub- sample) and distinct (Le.. experimental and survey) and convergence of findings still occurred. Moreover, certain design features actually rninimized the likelihood of detecting significant effects - participants in these three studies were unlikeiy to be as invested in the hiring and wage-setting simulations as employers in real-life settings wouid be since the participants were not in a position to incur any costs (e.g., payroil, training expenses associated with ernployee turnover). Yet significant effeas emerged nevenheless. Thus it seems reasonable to suggest that the pattern of results uncovered here constitutes a watered-down version of the sexism that can occur in employment settings. One can easily imagine that employers who are in a position to maintain employee continuity andfor profits at a high level would be much more Iikely to engage in the sexism-tinged decision-making demonstrated in these three studies.

The effects of individual sexism in employrnent practices can best be viewed in terms of a sequence of more and more narrowly gauged filters. Biases enter into the sequence at every level: recmitment of applicants in terms of the perceived sex-appropriateness of the occupation; responses to applicants; screening of candidates; rating of applicants who reach the intewiew stage; hiring andsalury dectsioru; and finally, promotional decisions once sorneone has been in the organization for a while [emphasis added]. (Unger & Saundra, 1993; p. 154)

Given that this investigation has answered in the affirmative the question of whether stereotypes can lead to gendered pay inequities. the question that logically follows is: what can be done to rectiS, these earnings dispanties? Some strategies for rnitigating the negative consequences of employers' stereotypes are available at both the institutional and individual levels. Unfortunately. the effectiveness of these solutions may be lirnited. as the following presentation aims to show

Policies addressine the Gender Waee Gao

As outtined in the introduction to this thesis, institutional-level remedies have been introduced and implemented in some organizations. To recapitulate, there is comparable wonk which relies upon job evaluation techniques to correct pay differentiak associated with stereotypically "male" and "femaie" jobs. Employment equityiaffinnative action involves increasing entry into jobs (and access to the accompanying higher salaries) for groups that have traditionally been disadvantaged, inciuding women, persons with disabilities and members of visible minorities. These initiatives are buttressed in some provincial and federal junsdictions by formal legislation (e.g.. Ontario's Pay Equity Act

(1990). Canada's Employment Equity Act (1 995)). This section will focus on comparable wonh and employrnent equity/affirmative action undertakings in particular, and examine features of these poiicies that may limit their impact on the gender wage gap. Some of these shortcomings are inherent in the policies themselves. whereas others have more to do with the manner in which the policies are implemented.

One major concem related to comparable worth is that it involves methods that some say are intrinsically gender biased (hey. 1986; DeCone. 1993). The compensatory factors that are used as the basis of salary assignment refer alrnost exclusively to characteristics associated with male-typed jobs while ignoring rnany of the tasks that are common to female-typed jobs (Steinberg, 1995). For example, responsibility for overseeing other employees' work, an activity that is more Frequently associated with "male" jobs. is given higher weight in job evaluation schemes than are unpleasant working conditions, which are often present in "female" jobs. In addition, the type of unpleasant working conditions that are routine in "female" jobs (e.g.. repeated exposure to obnoxious people that occurs in the service industry) often fail to be considered in job evaluation procedures (Treiman, 1 984). Because existing job evaluation instruments are able to pick up some gender inequalities that can in tum be corrected. many comparable worih proponents continue to advocate that this imperfect solution to the gender wage gap be adopted (England, 1992).

It is possible that improvements could be made to existing job evaluation systems in order to fùrther enhance women's outcornes (e-g., incorporating new compensatory factors;

McArthur. 1985). Even so. the situation in Ontario is such that women who are employed at small companies are unable to take advantage of "equal pay for work of equal value" legislation because provincial laws cover only those people who work in settings with ten or more employees (Pay Equity Act. 1990). In addition. there are numerous obstacles standing in the way of women who do qualifi for coverage under the law. As discussed in the context of Study 1. women who feel that they are underpaid need to have strong union support for their daim and must be reconciled with the fact that the process of obtaining fairer wages is adversarial and possibly quite lengthy (Brenner. 1987).

Employment equity/afirmative action programs do not appear to fare much better than comparable wonh at achieving their aims. Leck. St. Onge. and Lalancette (1995) analyzed data provided by al1 the Canadian companies that were subject to federal employment equity legislation and found limited evidence that this policy had been successfùl in decreasing the wage gap. Specifically. they examined how men and women fiom various groups (Le., people ivith disabilities. people from visible rninority groups. and

Aboriginals) fared relative to white men wit h respect to salary. They found that the gap between white men and the other grcups was in fact closing in the mid- to lower salary ranges and that white women reaped more benefits in terms of salary than the other target groups. Nevenheless. white men continued to dorninate in the highest incorne brackets, with the percentage of white men eaming $40,000 or better increasing more rapidly than any other group.

90th comparable worth and employrnent equitylaffirmative action initiatives appear to suffer from problems with implementation in Canada. There is limited monitoring of the effects of the policies or of employers' efforts to comply with them

(Leck et al., 1995). Also, it does not appear that the political wiil exists to enact stronger legislation. As it stands, current laws make their influence felt upon a relatively srna11 proportion of Canadian workers - much of the federal and provincial legislation appiies ody to govemment employees and those who work for agencies that receive contracts from the government. For instance. federal employment equity laws impact upon less than

10% of al1 Canadian workplaces (ieck et al.. 1995). Interestingly, Ontario was the oniy province ever to have passed progressive employment equity legislation that affected women empioyed outside of govemment settings and obligated employers in both the public and private sector to be proactive in achieving equity rather than operating on a complaints-based procedure (McDermott. 1996). This legisiation had origindly been enacted by the NDP govenunent in 1994. but was stmck down in 1998 upon the

Conservatives' rise to power in Ontario. Overall. the laws that remain in force do not affect the fortunes of the majority of working wornen one way or the other (McDermott,

1996).

Ern~lover and Emplovee Reactions to Policies

Moving beyond the consequences that institutional policies have for reducing the wage gap, the question of whether comparable worth and employment equity/affrmative action policies can help counteract sexism at the level of the individual must also be addressed. The extent to which employers' stereotypes and attitudes can be expected to

undergo transformation as a direct consequence of the implementation either of these

initiatives is not currently known. The readiness with which employers appeai tribunal

decisions that award better compensation to underpaid groups has already been alluded to

in Study 1 - this tendency suggests that conversion to a new, non-sexist way of thinking

certainly does not readily occur. Still, al1 may not be lost. Recent work on attitudes

toward employment equity shows that raising awareness of the goals and procedures

associated with these programs can give nse to greater openness to arguments about the justice of these initiatives (Tougas. Crosby. Joly. & Pelchat. 1993). which can facilitate

their acceptance.

Furthemore. the findings of a survey study involving respondents ffom five

countnes (Le., the United States, Canada. Australia, Sweden. and Nonvay) suggest that

the type of social policies in effect intluence attitudes independently of

individual-level factors (Baxter & Kane. 1995). Respondents living in countries that

stnved toward egalitarian policies (e-g.. Sweden, Nonvay) were thernselves more

egalitarian in their attitudes than respondents fiorn countries with less comprehensive

programs (e-g.,Canada, the US.). This result hints that irnprovement of the existing

gender gap-related policies like comparable worth and employment equity might be

accornpanied by a shift in attitudes of those in the workforce. including ernployers.

However. because Baxter and Kane's survey was cross-sectional in design, the direction

of the causal arrow cannot be conclusively detennined; it is possible that the progressive

gender role attitudes of the citizens led to the adoption of these policies rather than vice

versa. At the end of the day, perhaps the best that can be hoped for is that these policies will act to reduce employers' expression of sexism in hiring and wage-setting situations even if they cannot directly address the roots of discrimination.

Last but not least, it is important to consider the reactions of potentid beneficiaries and their CO-workersto these programs. It cannot be assumed that the programs will be ernbraced by ail, nor that they will have unconditionally positive effects for the intended targets. In faa, research has shown that both beneficianes of employment equity/ affirmative action and their peers rnay second-guess the cornpetence of the beneficiary to do the job for which s/he has been hired (Decter. 1980: Heilman, Simon. & Repper. 1987:

Roberts. 198 1). Clearly, steps need to be taken to rectiS, these perceptions if employment equity initiatives are to be successful.

The picture rnay not be as bleak as first imagined. at least for beneficiaries -

Tougas. Joly, Beaton. and St.-Pierre (1996) have demonstrated that beneficiaries'

assessrnent of their own qualifications. work motivation and work satisfaction are not

necessarily adversely affected by knowledçe that preferential selection methods are in

effiect at their workplace. Self-ratings on these variables depended on how fair

respondents thought their workplace employment equity/affirmative action prograrn was.

and the degree to which respondents were dissatisfied with their chances for success at

their workplace. In addition, experimental studies show that the reactions of

nonbeneficiaries to gender-based preferential selection are not necessarily negative, as long

as they believed that the outcome of the procedure was consistent with the beneficiary's

supenor qualifications (Heilman, McCullough. & Gilbert. 1996). Unfortunately. in real

workplace settings, individuals who are passed over as a result of preferential seiection do

not have access to such information, thus setting up a situation where they perceive that has occurred. Indeed the desire to avoid engaging in "reverse discrimination" has been cited by the provincial Conservatives as one of the reasons why

Ontario's employment equity legislation was stmck dom. More recent work suggests that clarity regarding the faa that ment considerations play a central role in employment equity/affirmative action programs can alleviate negative reactions to some extent

(Heilrnan, Battle, Keller, & Lee, 1998).

The Bottorn Line

Thus far, it appears that institutional-level policies like comparable wonh and employment equity have yielded limited success in closing the gender wage gap. The mediocre results observed can certainly be attributable in part to flaws in the design and implementation of these programs (Leck et al.. 1995; Steinberg, 1995). But just as no single factor or level of analysis is sufficient to explain gender discrimination @eaux &

LaFrance, 1998; Unger & Saundra, 1993). no simple explanations can be offered for the inability of institutional-level policies to eradicate the effects of such discrimination. As

Deaux and La France (1998) remind us. such events are multiply determined. For a better understanding of why sexism continues in the workplace. we need to turn to a consideration of individual-level factors.

Individuai-Level Factors

It has been suggested that our cognitive predilection to categorize individuals according to gender ensures that gendered inequalities will inevitably be built into workplace arrangements. Basicaily, the propensity to invoke a gender-based categorization is thought to Iead to activation of shared beliefs relating to gender, which subtly influence interactions even in situations where the individuals involved do not perceive gender as directly relevant to their exchange (Northcraft & Gutek, 1993;

Ridgeway, 1997). Ridgeway's ( 1997) discussion of this phenornenon focuses on beliefs related to the relative status of men and women wherein men are seen as diffisely more competent and generaily superior to women. The belief that men are superior to women certainly seems tu be reflected in the results of Study 1. where female-typed work was once again show to be devalued. Ridgeway (1997) maintains that these beliefs are well-

Iearned by individuals and exceedingly slow to change at the societal level, which helps explain why any attempt to transform existing methods of allocating work rewards is unlikely to result in a system that is greatly more egalitarian than currently exists.

Ridgeway's formulation certainly seems germane to the question of why women are disadvantaged in the workplace and why policies aimed at reducing the gender wage gap have produced less than spectacular effects.

Gutek's description of the impact of shared gender beliefs (in Nonhcraft & Gutek.

1993) covers different ground than Ridgeway. addressing beliefs and noms surrounding farnily role responsibilities. Gutek asserts that farnily responsibilities "may well be the single biggest irnpediment to attainment of a discrimination-free environment for women managers as well as other working wornen" (pp. 229-230). The gender role beliefs that she discusses include the notion that for men. fulfilling one's work role is the same as fulfilling one's family role of "good provider," as well as the understanding that women are expected (and permitted) to experience work-family confiict more than men do. It is not stretching very far to suggest that the culturally shared beliefs explored in Studies 2 and 3 of this dissertation - those relating women's proclivity to abandon work in favour of fdy,even if'temporariiy - are also prone to being automatically activated in the rninds of employers. As with the beliefs about gender and status discussed by Ridgeway ( 1997). the stereotypes regarding family responsibilities are learned early and learned well. whereas egalitarian beliefs are acquired later: because of this overleaming, sexisrn continues to have an advantage over egalitarianism (Fiske. 1998).

Despite the fact that individuals' (employers') gender-related beliefs are thought to be readily activated and acted upon. some investigators hold out hope that it is possible to counteract the negative effects that could ensue for women. Eberhardt and Fiske ( 1996) draw upon theoretical work on stereotyping to outline a number of strategies that potential targets of stereotyping and discrimination can exercise in order to discourage category-based thinking by perceivers/employers; a few of these are reviewed here. For example, Eberhardt and Fiske (1996) recornmend that women could try to get perceivers to look at them differently by increasing the salience of one's rnembership in categories that are not gender-based; in a job interview. this strategy might involve repeated emphasis on one's membership in the educational elite. The authors also suggest capitalizing on the interdependence of perceiver and target; again. for women attempting to access a desired job, this could consisr of highlighting how one's personal and work goals are in line with the hiring organization's own. A third strategy is to underscore the egalitarian views of some third party whose opinion is imponant to the perceiver. In hiring situations, this mi& entail job applicants making a point of mentioninç a company's employment equity/afErmative action program (if one is in effect) in an effort to remind the intemiewer that their organization has made a public deciaration of its cornmitment to women' s equality . The desired end result of utilizing these strategies is that targets will be perceived as individuais rather than as group members. While they make a valuable contribution toward the goal of fairer treatment for women, these recommendations are Iikely to be most successful for women who are already atvpical with respect to their farnily life aspirations. In other words, women who do not want or cannot have children stand the best chance of distancing themselves from the stereotypical female characteristics that ernployers seek to avoid (i.e.. questionable availability in the long terni).

The Gender Gap and Potential Motherhood

What about women who still plan to withdraw from the workforce in order to bear and raise children? Recall that Study 2's findings indicate that it is not far-fetched to suggest that individuals who want a family are treated inequitably. Job candidates who expressed an interest in having a family were rated less favourably than other types of applicants on most of the dependent variables tested. The findings of Study 3 are even more striking in that only women who are hmily-oriented are rated as being at a disadvantage during hiring. Are there any strategies that are available to these women to ensure that they will not be denied access to jobs?

Studies suggest that women continue to adhere to existing noms about which parent should take leave from work when a child is bom (i.e., the rnother). In a sarnple of undergraduate-aged women, Davey (1998) found that 50% expressed a preference for leaving work at the time of their first child's birth and not returning until their last child reached a certain age. This type of personal work history is obviously not ideal from an employer's perspective - it is understandable that some might be wary of hiring women, particularly young women, as a result. However, Davey's participants acknowledged that their preferred course of action would not necessady coincide with the choices that they would actually have to make; only 22% expected that they would be able to stop working outside the home for so long. In fact. Statistics Canada data (1999~)shows that 90% of women take matemity leaves that last less than a year: the average length of time away from work upon birth of a child was about six months. Nevenheless. this length of tirne is considerable from the perspective of an employer; they still need to find a way to have the person's duties met while she is away. Maternity leaves are inconvenient to employers because they require engaging in additional recmitrnent and hi ring (which cm be costly ). as well as to other employees. who may need to "covei' for the person who is on leave.

Again, it is perhaps not surprising that business owners might espouse human capital theory-tinged views and daim that statistical discrimination dunng the hiring process is rational.

Northcrafl (in Nonhcraft & Gutek. 1993) argued that alternative work arrangements such as flextime scheduling and telecommuting, which are being adopted by increasing numbers of companies, will help. He claimed that by vinue of the availability options, more men will take on household tasks that currently fa11 primarily to women: "as changes in U.S.business practices make it easier for men to shoulder a larger share of parental burdens, these changes in noms should also help corporationsforget parenting obligations as a consideration in the career advancement of women (much as currently they are probably forgetten for men)" (p. 223, emphasis in original). Other commentators have been similarly optimistic; Waldfogel ( 1 998) has suggested that policies allowing for longer maternity leave could help stuink the gender wage gap when combined with wage- related policies like comparable worth and employment equity. But could not this strategy of providing extended matemity leave backfire? Recently in Canada, the federal governrnent indicated that they were planning to change the existing matemity leave policies so that women could take longer leaves (for up to one year) paid for by the federal Employrnent Insurance program (McIlroy, 1999). Despite the fact that Canadian employers would not need to pay for these benefits. they do experience inconvenience when trying to replace workers who are on leave, as outlined above. Employers may choose to circumvent this whole procedure by engaging in statistical discrimination when faced with female job candidates.

It should also be noted that even though accommodations like leaves and flextime are technically avaiiable to &lemployees. managers appear to show greater willingness to grant alternative work arrangements such as working a reduced number of hours, job- sharing or taking an unpaid leave of absence to subordinates than to managers (Barham.

Goalieb, & Kelloway, 1998). In addition. while they are equally obliging of male and female subordinates, they are more likely to accommodate fernale managers than male managers (Barham. et al.. 1998). In another study, Allen and Russell ( 1999) found that male ernployees who take parental leave are more frowned upon than women are. One wonders if there are any solutions available that do not penalize women who want farnilies in some way. As it stands. women continue to be the ones who use these alternative work options far more fiequently than men do as well as to bear the burden of being perceived as lacking work commitment when they do so (Clay, 1998).

Equally pressing is the need to ensure that once access is achieved, women are not penalized in terms of pay. There is no excuse for hiring a woman and paying her less than a man who does the same job, yet Study 2 showed that this could potentidy occur; recall that job applicants who indicated that they wanted to have a family were assigned salaries that were up to $3.000 lower than other applicants. Similar findings have been reported in pnor research. Rosen and Mencle (1979) found that a strong affinnative action poiicy

appeared to reduce sex bias in selection but tended to result in lower salary

recommendations for the wornen who were hired. They interpreted this occurrence as

reflecting people's reaction against perceived hiring constraints. It is plausible that Study

2 participants felt similariy constrained, experiencing the hinng scenario as pulling for

them to agree to hire the candidate with which they had been presented regardless of

whether they tmly wanted to do so. As a result. they might have indicated their

disapproval of the candidate in a forum where they felt they had more latitude for

expression - in the starting salary. This finding. considered together with comparable

wonh initiatives that are themselves gender biased. anti-discrimination legislation that may actually prornote the use of stereotypes, and parental leave policies that actually penalize wornen, makes it clear that much work still needs to be done before we can Say that we

have generated solutions to the gender wage gap that do not have negative repercussions for the group that they are intended to help.

ConcIusions

If one's focus is on the magnitude of the wage gap rather than the fact that it exists. it is plain to see that the gap among tùll-time workers has actually shnink over the

last 25 years (Baker et al., 1995; Gundersoa 1998; Statistics Canada. 1999). The

explanation for this change is hotly contested. As a general rule. economists tend to argue that women currentiy enjoy greater advantages in the workf'orce than they did in previous years, in part because of a decrease in discrimination against women (Sorensen, 199 1). On the other hand, while many sociologists and psychologists agree that there have been

positive changes in terms of how women are perceived and in their productivity-related

characteristics (Drolet, 1999). they attribute the reduction in the gap largely to market

changes that have resulted in a decline in men's average wages (Fudge, 1996; lacobs,

1995; Mutari & Figart, 1997).

Obviously, this dissertation cannot speak with authority about of the

rnechanisrns that could potentially be responsible for changes in the size of the wage gap.

It is indisputable that women's lot in the workplace has improved in recent years, and

perhaps this shifi is due in part to changing views regarding women's cornpetence to

successfully engage in a wide variety of work roles, including roles that are associated

with better compensation. However. the three studies presented here indicate that

individual-level sexism Iikely still plays a role in maintaining a gendered wage gap.

Stereotypes about women's occupational and familial roles continue to be associated wiih

differentiai valuing of male- and female-typed jobs and differential access to jobs on the

basis of gender and gender roles. References

Adler. N.J., & Lnaeli, D.N. (1988). Women in Management Worldwide. London: Sharpe.

Adler, N.J., & lzraeli, D.N. (1994). Com~etitiveFrontiers: Wornen Managers in a Global Economy. Oxford: Blackwell.

Ailen, T.D., & Russell, J.E.A. (1999). Parental leave of absence: Sorne not so family-friendly implications. Journal of Applied Social Psycholog, 29, 166- 19 1.

Allen, T.D..Russell, J.E.A.. & Rush. M.C.( 1994). The effects of gender and leave of absence on attributions for high performance. perceived organizational cornmitment, and allocation of organizational rewards. Sex Roles. 3 1, 443 -464.

Allpon, G.W. ( 1954). The Nature of Preiudice. New York: Addison-Wesley

Amato, 1. (1992). Women have extra hoops to jump through. Science. 255, 1372-1373.

Armstrong, P.. & Armstrong, H. (1 994). The Double Ghetto (3rd Ed.). Toronto McClelland and Stewart.

Arvey, R.D.(1986). Sex bias in job evaluation procedures. Personnel Psvchologv. 39, 3 1 5-335.

hster. E.R.(1989). Task characteristics as a bridge between rnacro- and rnicrolevel research on salary inequality between men and women. Academy of Management Review. 14, 173- 1 93.

Baine, J. (1999). Workplace equality top priority. poll finds. The Kitchener- Waterloo Record, May 3, p. C 1.

Baker, M., Benjamin, D.. Desaulniers. A.. & Grant. M. (1995). The distribution of the maldfemale earnings differential, 1970- 1990. Canadian Journal of Economics. 28, 479-50 1.

Barham, L.J., Gottlieb, B.H..& Kelloway. E.K.(1998). Variables affecting managers' willingness to gant alternative work arrangements. The Journal of Social PSVC~O~O.W.138, 29 1-302.

Bamum P., Liden, R.C..& DiTomaso. N. (1995). Double jeopardy for women and minonties: Pay diEerences with age. Acadernv of Management Journal. 38, 863-880. Baron, J. (199 1). Organizational evidence of ascription in iabor markets. In R.R. Cornwall & P.V. Wumava (Eds.), New Approaches to Economic and Social Analvses of Discrimination. New York: Praeger.

Baron, J.N., & Newman, AE. (1990). For what it's worth: Organizations. occupations, and the value of work done by women and nonwhites. American Sociological Review. 55, 155- 175.

Baron, IN., & Pfeffer. J. (1 994). The social psychology of organizations and inequality. Social Psvcholow Ouarterlv. 57, 190-209.

Bartol. K.M. ( 1976). Relationship of sex and professional training area to job orientation. Journal of Applied Psvcholo~.6 1, 368-370.

Baxter, J., & Kane, E.W. (1995). Dependence and independence: A cross-national analysis of and gender attitudes. Gender and Society. 9, 193-2 15.

Becker. G,S. ( 1980). Human Capital: A Theoretical and Empirical Analvsis. with Soecial Reference to Education. Chicago: University of Chicago Press.

Beggs, J.M., & Doolittie. D.C. ( 1 993). Perceptions now and then of occupational sex typing: A replication of Shinar's 1975 study. Joumal of Amlied Social Ps~choloa -23, 1435-1453.

Bellas, M.L. (1994). Comparable worth in academia: The effects of faculty salaries of the sex composition and labor-market conditions of acadernic disciplines. Arnerican SocioIoeical Review. 59, 807-82 1.

Bem S .L.,& Bem. D.J. ( 1973). Does sex-biased advenising -'aid and abet" sex discrimination? Journal of Aoolied Social Psycholopv. 3, 6- 1 8.

Benokraitis, N.V., & Feagin, J.R.( 1995). Modem Sexism: Blatant. Subtle. and Covert Discrimination (2nd Ed.). Toronto: Prentice-Hall Canada Inc.

Bernard, J. (198 1). The good provider role: Its nse and fall. Amencan Psycholopist. 36, 1- 12.

Bielby, W.T., & Baron, J.N. (1986). Men and women at work: and statistical discrimination. Amencan Journal of Sociolow. 9 1, 759-799.

Blau. F.D..& Beller, A.H. ( 1988). Trend in earnings differentials by gender. 197 1 - 198 1. Industrial and Labor ReIations Review. 34, 563-577.

Blum L.M. (1987). Possibilities and Iimits of the comparable wonh movement. Gender & Society. 1,380-399. Boyd, M., Mulvihill, M.A., & Myles, J. (1995). Gender, power, and postindustrialism. In J.A. Jacobs (Ed.) Gender Inequality at Work. Newbury Park CA: Sage.

Brenner, J. (1987). Feminist political discourses: Radical versus liberal approaches to the feminization of and comparable worth. Gender & Society. 1, 447-465.

Brewer, M.B. (1996). When stereotypes lead to stereotyping: The use of stereotypes in person perception. In C.N.MacRae. C. Stangor & M. Hewstone (Eds.) Stereotypes and Stereotpinq. New York: The Guilford Press.

Brewer, M.B., & Lui. L.N. (1 989). The pnmacy of age and sex in the structure of person categories. Social Coenition. 7, 262-274.

Bnef, A.P., & Aldag, R. J. ( 1 975). Male-female differences in occupational attitudes within minonty groups. Journal of Vocational Behavior. 6, 305-3 14.

Brief. A.P., Rose, G.L., & Aldag, R.J. (1977). Sex differences in preferences off job attributes revisited. Journal of Apglied Pqcholow. 62, 645-646.

Burke, R.J., & McKeen. C.A. (1 993). Career priority patterns arnong managenal and professional women. A plied Psycholow: An International Review. 42, 34 1-3 52.

Bylsma, W.H., & Major, B. ( 1992). Two routes to eliminating gender differences in personal entitlement. Psychoiow of Women Quanerl~.16, 193-200.

Bylsma, W.H., & Major. B. (1994). Social comparkons and contentment: Explonng the psychological costs of the gender wage gap. Ps~cholowof Women Oumerly. 1 8, 24 1-249.

Bylsma, W.H., Major, B., & Couarelli. B. ( 1995). The influence of legitimacy appraisals on the determinants of entitlement beliefs. Basic and Applied Social Psvcholog-y, 17, 223-247.

Calhoun. L.G., & Selby, J. W. ( 1980). Voluntary childlessness. involuntary childlessness, and having children: A study of social perceptions. Farnily Relations. 29, 181-183.

Carneron, C. (1977). Sex-role attitudes. In S. Oskarnp (Ed.) Attitudes and Opinions. Englewood Cliffs. NJ: Prentice Hall.

Canadian Human Rights Act, Consolidated Statutes of Canada, Chapter H-6 (1976). Canadian Human Rights Commission (1998a). Equal Pay for Work of Eaual Value: Employs' Guide. Ottawa: Author.

Canadian Human Rights Commission ( 1998b). Federal Pay Equity Still in Limbo: CHRC Annual Report [On-line]. Available: http://www.chrc-ccdp-ca.

CARNET (1992). Work and Familv: The Survev. Unpublished data, Gerontology Research Centre, University of Guelph

Cejka, M. A., & Eagly, A. H. ( 1999). Gender-stereotypic images of occupations correspond to the sex segregation of employment. Personality and Social Psvchology Bulletin. 25, 413-423.

Christofides. L.N.,& Swindinsky. R. (1994). Wage determination by gender and visible minority statu: Evidence from the 1989 LMAS. Canadian Public Policv. 22, 34- 5 1.

Chusmir, L.H. ( 1984). Personnel administrators' perception of sex differences in motivation of managers: Research-based or stereotyped? International Journal of Women's Studies. 7, 1 7-23.

Clay, R.A. (1998). Many managers frown on use of flexible work options. MA Monitor. 29, 1 1.

Colella, A.. DeNisi, AS.. & Varma. A. ( 1997). Appraising the performance of employees with disabilities: A review and model. Human Resource Management Review. -7, 27-53.

Colella, A., DeNisi, A.S., & Varma, A. (1998). The impact of ratee's disability on performance judgments and choice as panner: The role of disability-job fit stereotypes and interdependence of rewards. Journal of Aoplied Ps~cholow.83, 102- 1 1 1.

Cooney, T.M.. & Uhlenberg, P. ( 199 1). Changes in work-family connections among highly educated men and women: 1970 to 1980. Journal of Familv Issues. 12, 69- 90.

Crosby, F.. Farrell, R., & Cameron. A.E. (1994). Changing sex-role expectations and men's concems with justice in the home. Ln M.J. Lemer & G. Mikula (Eds.) Entitlement and the Mectional Bond: Justice in Close Relationshi~s. New York: Plenum Press.

Crosby. F.I., & Jaskar. K.L. (1993). Women and men at home and at work: and illusions. In S. Oskamp & M. Costanzo (Eds.) Gender Issues in Contemporary Society. Newbury Park, Ch: Sage Publications Inc. Croxton, J.S., van Rensselaer, B.A.. Dutton, D.L., & Ellis, J.W. (1989). Mediating effects of prestige on occupational stereotypes. Psvcholo~icalReports. 64, 723 -73 2.

Daniels, A.K. (1987). Invisible work. Social Problerns. 34, 403-41 5.

Daniels, P., & Weingarten, K. (1988). The fatherhood click: The timing of parenthood in men's lives. In P. Bronstein & C.P.Cowan (Eds.)Fatherhood Today : Men's Chaneinp Role in the Family. New York: John Wiley and Sons.

Davey, E.H.( 1 998). Young wornen's expected and preferred patterns of employment and child care. Sex Roles. 3 8, 95- 102.

Davidson. M.J.. & Cooper. C.L.( 1993). Eurooean Women in Business and Management. London: Chapman.

Davies, S., Mosher. C.. & O'Grady, B. (1996). Educating women: Gender inequalities among Canadian university graduates. The Canadian Review of Sociolow and Anthro~oloq.33, 125- 142.

Deaux. K. (1985). Sex and gender. Annual Review of Psvcholow. 36'49-81.

Deaux, K.. & Kite, M. (1993). Gender stereotypes. In F.L.Denrnark & M.A. Paludi (Eds.) Psvcholoe;v of Women: A Handbook of Issues and Theories. Westpon. CT: Greenwood Press.

Deaux, K., & LaFrance. M. ( 1998). Gender. In D.T.Gilbert, S.T.Fiske & G. Lindzey (Eds.)The Handbook of Social Psvcholog (4" ed.. Vol. 2) Boston: McGraw- Hiil.

Deaux, K., & Lewis. L.L. (1984). The structure of gender stereotypes: Interrelationships among components and gender label. Journal of Personalit~and Social Psycholow. 46, 99 1- 1004.

Deaux, K., Winton, W., Crowley, M & Lewis, L.L. (1985). Level of categorization and content of gender stereotypes. Social Comition. 3, 145- 167.

DeCorte, W. (1993). Estimating sex-related bias in job evaluation. Journal of Occu~ationaland Or~anizationalPsvcholow, 66, 83-96.

Decter, M. ( 1980). Benign victimization. Policv Review. 13, 65-72.

Denton M.A, & Hunter, A.A. (1982). Economic Sectors and Gender Discrimination in Canada: A Critique and Test of Block and Walker.. .and Sorne New Evidence. Equaiity in the Workplace Series. Ottawa: Labour Canada, Women' s Bureau. Desmarais. S., & Curtis. J.E. (1997). Gender and perceived pay entitlement: Testing for effects of experience with income. Joumal of Personality and Social Psvcholo~~;v.72, 141-150.

Devine. P.G. (1989). Stereotypes and prejudice: Their automatic and controlled cornponents. Journal of Personalitv and Social Psycholow~56. 5- 18.

Drolet, M. (1999). The Persistent Gap: New Evidence on the Canadian Gender Wane Gap. (Catalogue No. 75 F0002MIE-99008). Ottawa: Statistics Canada, Incorne Statistics Division.

Duncan, K.C.. Pms, M.J.. & Sandy, J.G. ( 1993). Marital status, children and women's labor market choices. Journal of Socio-Economics, 22, 277-288.

Duxbury, L.E.,& Higgins, C.A. ( 199 1 ). Gender differences in work-family conflict. Journal of Apolied Psycholow. 76, 60-73.

Eagly, A.H. (1995). The science and politics of comparing women and men. Amencan Psvchologist. 50I31, 145- 158.

Eagly, AH.. & iWadinic. A. (1 994). kepeople prejudiced against women? Some answers Rom research on attitudes, gender stereotypes. and judgements of cornpetence. In W. Stroebe & M. Hewstone (Eds.) European Review of Social Psycholog (Vol. 5). Toronto: John Wiley & Sons Ltd..

Eberhardt. J.L.. & Fiske, S.T.(1996). Motivating individuals to change: What is a target to do? In C.N.Macrae. C. Stangor & M. Hewstone (Eds.) Stereotpes and Stereotypinq. New York: The Guilford Press.

Eccles, J. S. ( 1994). Understanding women's educational and occupational choices. Psvcholow of Women Ouarterl~.1 8, 585-609.

Eckes. T. (1994). Explorations in gender : Content and stnicture of female and male subtypes. Social Coenition. 12, 37-60.

Employrncnt Equity Act. Consolidated Statutes of Canada, c.44 (1995).

England, P. (1 984). Wage appreciation and depreciation: A test of neoclassical economic explmations of occupational sex segregation. Social Forces. 62, 726-749.

England. P. ( 1992). Comparable Worth: Theones and Evidence. New York: Aldine DeGniyter. England, P., & Markowitz, L. (1996). The moral dimension of feminist theones: Implications for pay equity. In D.Sciulli (Ed.) Macro Socio-Economics: From Theory to Activism. Armonk, NY: M.E. Sharpe, hc..

Etaugh, C., Houtler, B.D., & Ptasnik. P. ( 1988). Evaluating competence of women and men: Effects of expenmenter sender and group gender composition. Psvcholom of Women OuarterIv. 12, 19 1-200.

Filer, R.K.(1 985). Male-female wage differences: The importance of compensating differentials. Industrial and Labor Relations Review. 3 8, 426-43 7.

Fiske, S.T. (1989). Examining the role of intent: Toward understanding its role in stereotyping and prejudice. In J.S. Uleman & J.A. Bargh (Eds.), Unintended Thou~ht. New York: Guilford.

Fiske, S.T.(1 998). Stereotyping, prejudice, and discrimination. In D.T.Gilben. S.T. Fiske & G. Lindzey (Eds.) The Handbook of Social Ps~cholog~(4' ed.. Vol. 2) Boston: McGraw-Hill.

Fiske, S.T., Bersoff, D.N., Borgida, E.. Deaux, K., & Heilrnan, M.E. (199 1). Social science research on trial: Use of sex stereotyping research in Price Wuterhouse W. Hopkins. American Ps~choloeist.46, 1049- 1060.

Fiske, A.P., Haslarn. N.. & Fiske, S.T.( 199 1 ). Confushg one person with another: What errors reveal about the elementary forms of social relationship. Journal of Personalitv and Social Psyholow. 3 8, 889-906.

Fiske, S.T.,& Stevens. L.E. (1993). What's so special about sex? Gender stereotyping and discrimination. In S. Oskamp & M. Costanzo (Eds.) Gender Issues in Contemporary Society. Newbury Park, CA: Sage Publications.

Fiske, S.T..& Taylor, S.E.(1991). Social Cosnition (2" ed.). New York: McGraw-Hill.

Fleming, J.J. (1 988). Public opinion on change in women's rights and roles. In S.M. Dornbusch & M.H.Strobes (Eds.) . Children. and the New Families. New York: Guilford.

Foschi, M., Lai, L.. & Sigerson, K. (1994). Gender and double standards in the assessrnent of job applicants. Social Ps~cholowOuarterlv. 57, 326-339.

Foschi, M.. Sigerson, K., & Lembesis, M. ( 1995). Assessing job applicants: The relative effects of gender, academic record, and decision type. Small Group Research, 0131,328-352. Frieze, I.H., Olson, J.E., & Good. D.C. (1 990). Perceived and actual discrimination in the salaries of male and female managers. Journal of Applied Social Psychoog- 20, 46-67.

Fuchs, V.R. (1988). Women's Ouest for Economic Equalitv. Cambridge, MA: Harvard University Press.

Fudge, J. (1996). Fragmentation and feminization: The challenge of equity for labour-relations policy. In J. Brodie (Ed.) Women and Canadian Public Policy. Toronto: Harcourt Brace & Co..

Gerhart, B. ( 1990). Gender differences in current and starting salaries: The role of performance, coilege major, and job title. Industrial and Labor Relations Review. 43, 4 18.433.

Gibbons, A. (1992). Tenure: Does the old-boy network keep women from leaping over this cmcial career hurdle? Science. 255, 1 3 86- 1 3 87.

Gilbert, D.T., & Hixon, J.G. (1991). The trouble of thinking: Activation and application of stereotypic beliefs. Journal of Personality and Social Ps~chology.60, 509- C 17

Glass, J., & Carnarigg, V. (1992). Gender, parenthood. and job-famiiy compatibility. Amencan Joumal of Sociolow. 98, 13 1- 1 5 1.

Glick, P. (199 1). Trait-based and sex-based discrimination in occupational prestige, occupational salary, and hiring. Sex Roles. 25, 3 5 1-3 78.

Glick, P.. & Fiske. S.T.(1 996). The Inventory: DifFerentiating hostile and benevolent sexism. Joumal of Personalitv and Social Psvcholop. 70, 49 1-5 12.

Glick, P., Wilk, K., & Perreault, M. (1995). Images of occupations: Components of gender and status in occupational stereotypes. Sex Roles. 32, 565-582.

Glick, P., Zion. C..& Nelson, C. ( 1 988). What mediates sex discrimination in hiring decisions? Journal of Personality and Social Ps~cholow.5 5, 178- 186.

Grams, R., & Schwab, D.P.(1985). An investigation of systematic gender-related error in job evaluation. Academv of Management Journal. 28, 279-290.

Green, B.L..& Russo, N.F.(1 993). Work and family roles: Selected issues. In F.L. Denmark & M.A. Paludi (Eds.) Psvcholow of Women: A Handbook of Issues and Theories. Westpon, CT: Greenwood Press. Gundenon, M. (1998). Women and the Canadian Labour Market: Transition Towards the Future. Census Monograph Senes. (Catalogue No. 96-32 1-MPE No. 2). Ottawa: Statistics Canada.

Gutek, B.A., Searle, S., & Klepa, L. ( 199 1). Rational versus gender role explanations for work-family conflict. Journal of Amlied Psvcholow. 76, 560-568.

Hakim, C. ( 19%). Five ferninist myths about women's employment. British Joumal of Socioloerv. 46, 429-455.

Hanrie. K., Marshall-McCaskey, J ., & JO hnston, L. ( 1998). Gender-based biases in occupational hiring decisions. Joumal of Applied Social Psvcholow. 28, 1698- 17 1 1.

Heilman, M.E., Banle, W.S.,Keller. C.E., & Lee. RA (1998). Type of affirmative action policy: A determinant of reactions to sex-based preferential selection? Journal of Apolied Psvcholow. 83. 190-205.

Heilman, M.E.. McCullough, W.F.. & Gilben. D. ( 1996). The other side of affirmative action: Reactions of nonbeneticiaries to sex-based preferential selection. Joumal of A~~liedPsycholow. 8 1, 346-357.

Heilman, M.E.. Manell, R.F., & Simon, M.C. ( 1988). The vagaries of sex bias: Conditions regulating the undervaluation. equivaluation, and ovewaluation of female job applicants. Organizational Behavior and Human Decision Processes. 4 1, 98- 1 10.

Heilman, M.E., Simon, MC, & Repper. D.P. ( 1987). Intentionally favored. unintentionally harmed: The impact of gender-based preferential selection onf self- perceptions and self-evaluations. Journal of Applied Psvcholow. 72, 62-68.

Hersch, J. (1985). Effect of housework on earnings of husbands and wives: Evidence frorn full-time piece rate workers. Social Science Ouarterlv. 66, 2 10-2 1 7,

Hewstone, M. (1994). Revision and change of stereotypic beliefs: In search of the elusive subtyping model. In W. Stroebe & M. Hewstone (Eds.) European Review of Social Psvcholoev (Vol. 5). Toronto: John Wiley & Sons Ltd..

Hornsby, J.S., Benson. P.G.. & Smith, B.N. (1987). An investigation ofgender bias in the job evaluation process. Journal of Business and Pqcholow. 2, 150-162.

Howell. D.C.( 19%). Statistical Methods for Psychologv: 3" Edition. Boston: PWS-Kent Publishing Company.

Human Resources Development Canada ( 1998). Job Futures [On-line]. Available: http://mv.hrdc-drhc. ~c.ca/JobFutures/enqIish/volurne1 . Jackson, L.A., & Grabski, S.V.(1988). Perceptions of fair pay and the gender wage gap. Journal of A died Social Pqcholoev. 18. 606-625.

Jacobs, J.A. (1992). Women's entry into management: Trends in earnings, authority, and values among salaried managers. Administrative Science Ouarterl~.37, 282-3 O 1.

Jacobs, J.A. (1995). Introduction. In J.A. Jacobs (Ed.) Gender Inequalit~at Work. Thousand Oaks, CA: Sage Publications.

Jacobs, I.A.. & Steinberg, R. (1995). Further evidence on compensating differentials and the gender gap in wages. In J.A. Jacobs (Ed.) Gender Inequalit~at Work. Thousand Oaks. CA: Sage Publications.

Jacobsen, J. P. ( 1994). Sex segregation at work: Trends and predictions. Social Science Joumal, 3 1, 153- 169.

Jagacinski, C.M..LeBold. W.K., & Linden, K.W. (1987). The relative career advancement of men and women engineers in the United States. Work Stress. 1, 235-247.

Joshi, H. (199 1). Sex and motherhood as handicaps in the labour market. In M. MacLean & D. Groves (Eds.) Women's Issues in Social Policy. New York: Routledge.

Kahn. A.. Nelson, R.E.. & Gaeddert. W.P. (1980). Sex and sex composition of group as dererminants of reward allocation. Joumal of Personaii~and Social Pwcholo~ -3 8, 73 7-750.

Kahn, W.. & Crosby, F.J. (1985). Change and stasis: Discriminating between attitudes and discriminatory behavior. In L.Larwood. B.A. Gutek & A.H. Stromberg (Eds.) Women and Work: An Annual Review (Vol. 1). Beverly Hills. CA: Sage.

Kane, E.W. (1995). Education and beliefs about gender inequality. Social Problems, 42, 74-90.

Kanungo, D.N. (1987). Measurement ofjob and work involvement. Journal of Applied Psvcholow. 6713 1, 34 1-349.

Katz, D. (1987). Sex discrimination in hiring: The influence of organizational climate and need for approval on decision making behavior. Psvcholoev of Women Quarterly, 1 1, 1 1-20.

Kay. F.M..& Hagan, J. (1995). The persistent giass ceiling: Gendered inequalities in the eaniings of lawyers. British Joumal of Sociolow. 46(2), 279-3 10. Kessler-Hams, A. (1987). The debate over equality for women in the workplace: Recognizing differences. In N. Gerstel & H.E. Cross (Eds.) Families and Work. P hiladelphia: Temple University Press.

Kilbourne, B ., England, P., & Beron. K. ( 1994). Effects of individual, occupational and industriai characteristics on eamings: Intersections of race and gender. Social Forces. 72, 1 149- 1176.

Konrad, A.M., & Cannings, K. (1997). The effects of gender role congmence and statisticai discrimination on managerial advancement. Human Relations. 50, 1305- 1328.

Lampman, C., & Dowling-Guyer. S. (1995). Attitudes toward voluntary and involuntary childlessness. Basic and Aoplied Social Psvcholog. 17, 2 13-222.

Landau. !.. & Arthur, M.B. (1992). The relationship of marital status, spouse's career status, and gender to salary level. Sex Roles. 271 1 1/ 121, 665-68 1.

LaRossa, R. (1988). Fatherhod and social change. Familv Relations. 34, 45 1-47.

LeBlanc. D. (1999). Ottawa tries to cut deal on pay equity. The Globe and Mail, October 23, p. Al, A4.

Leck, J.D., St. Onge, S., & Laiancette, 1. (1 995). Wage gap changes among organizations subject to the Employmeiir Eqzriv Act. Canadian Public Policy. 2 1, 387- 400.

Lefkowitz, J. (1994). Sex-related differences in job attitudes and dispositional variables: Now you see them,. .. . Academy of Management Journal. 37, 323-349.

Lero, D.S..& Johnson, K.L.(1994). 1 10 Canadian Statistics on' Work & Famil?, Catalogue No. 94-E-204. Ottawa: Canadian Advisory Council on the Status of Women.

Leslie, L. A., Anderson. E. A., & Branson. M.P. ( 199 1). Responsibility for children: The role of gender and employment. Journal of Farnil~Issues. 12, 197-2 10.

Levinson, R.M. ( 1975). Sex discrimination and employment practices: An experiment with unconventional job inquines. Social Problerns. 22, 533-543.

Lippmann, W. (1 922). Public Opinion. New York: Harcourt, Brace.

Lockheed, M.E., & Hall, K.P.(1976). Conceptualizing sex as a status characteristic: Applications to training strategies. Journal of Social Issues. 32, 111-124. MacDonald, T.K.. & Zanna, M. P. ( 1 998). Cross-dimension ambivalence toward social groups: Can ambivalence affect intentions to hire feminists? Personality and Social Psvcholog~Bulletin. 24, 427-44 1.

Mahoney, T.A.. & Blake, R.H.(1987). Judgements of appropriate occupational pay as infiuenced by occupational charactenstics and sex characterizations. Applied Psvcholo~v:An International Review. 36( 1). 25-38.

Major, F.. McFarlin, D.. & Gagnon, D. ( 1984). Overworked and underpaid: On the nature of gender differences in persona1 entitlement. Journal of Personality and Social Psvcholow. 47, 1399- 14 12.

Marini, M.M.. & Fan. P.L. (1997). The gender gap in eanllngs at career entry American Sociological Review. 62, 5 88-692.

Markham. W.T.. South, S.J..Bonjean. C.M..& Corder. 1. (1985). Gender and opportunity in the federal bureaucre.cy. Arnerical Journal of Sociolow. 9 1, 129- 1 50.

Marshall. K. ( 1994). Balancing work and family responsibilities. Pers~ectives. (Catalogue 75-00 1E). Ottawa: Statistics Canada

Martell, R.F.. Lane. D.M..& Emrich, C. ( 1996). Male-female differences: A computer simulation. American Psycholoeist. 5 1. 1 57- 158.

Martin. P. Y. ( 1992). Gender. interaction. and inequaiity in organizations. In C.L. Ridgeway (Ed.) Gender. Interaction and Inecpality. New York: Spnnger-Verlag.

Martin, P. Y. ( 1993). Feminist practice in organizations: Implications for management. In E.A. Fagenson (Ed.) Women in Management: Trends. Issues. and Chailenoes in Manaeenal . Newbury Park. CA: Sage.

Martinko. M.J.. & Gardner. W.L.( 1983). A methodological review of sex-related access discrimination problems. Sex Roles. 9, 825-839.

Mason, K.O.. & Lu. Y. (1988). Attitudes toward women's familial roles: Changes in the United States. 197%1985. Gender & Society. 2. 39-57.

McArthur. L.Z.(1985). Social judgment biases in comparable worth analysis. In H.I. Hartmann (Ed.) Comparable Worth: New Directions for Research. Washington, DC: National Academy Press.

Mchhur. L.Z., & Obrant. S.W. ( 1986). Sex biases in comparable wonh analyses. hmal of Ao~liedSocial Psvcholoev. 16, 757-770. McDermott. P. (1996). Pay and employment equity: Why separate policies? In J. Brodie (Ed.) Women and Canadian Public Policy. Toronto: Harcourt Brace & Co..

McIlroy, A. ( 1 999). New parents to get more paid leave. The Globe and Mail. October 13. p. Al, A5.

McShane. S.L. ( 1990). Two tests of direct gender bias in job evaluation ratings. Journal of Occupational Psvcholo~.63, 129- 140.

Melamed, T. ( 1995). Barriers to wornen's career success: Human capital. career choices, structural determinants. or simply sex discrimination. Applied Psycholog: An International Review. 44, 295-3 14.

Miall, C.E. (1 985). Perceptions of informal sanctioning and the stigma of involuntary childlessness. Deviant Behavior. 6. 3 83-403.

Miall, C.E. ( 1 986). The stigma of involuntary childlessness. Social Pro blems. 3 3. 368-282.

Moen, P. (1985). Continuities and discontinuities in women's labor force activity. In G.H. Elder Jr. (Ed.) Life Course Dvnamics: Tra-iectories and Transitions. 1968- 1980. Ithaca, NY:Comell University Press.

Momson, T.G.,Bell. E.M.. Momson. M.A.. Murray. C.A.. & O'Connor. W. (1994). An examination of adolescents' salary expectations and gender-based occupational stereotyping. Youth & Society. 26, 1 78- 1 93.

Moul, G.M. (1 999). The Gendered Earnines G~Pin Canada: A Review of Literature. Unpublished manuscnpt. University of Waterloo.

Mount, M.K.. & Ellis. R.A. (1987). Investigation of bias in job evaluation ratings of comparable wmh study participants. Personnel Psycholow. 40, 85-96.

Mount. M.K.. & Ellis. R.A. (1989). Sources of bias in job evaluation: A review and critique of research. Journal of Social Issues. 45, 153- 167.

Mutari, E., & Figm. D.M.(1997). Markets, flexibility, and family: Evaluating the gendered discourse against pay equity. Joumal of Economic Issues. 3 1, 687-705-

National Occupation Classification ( 1992).

Naughton, T.J.(1988). Effect of female-linked job titles on job evaluation ratings. Journal of Management. 14, 567-578. Neuberg, S.L. ( 1989). The goal of forming accurate impressions dunng social interactions: Attenuating the impact of neçative expectancies. Journal of Personalitv and Social Psyholow. 56. 374-386.

Neuberg, S .L. ( 1994). Expectancy-confirmation process in stereotype-tinpd social encounters: The moderating role of social goals. In M.P. Zama & J.M. Olson (Eds.)The Psvcholom, of Pre-iudice: The Ontario Symposium (Vol. 7). Hillsdale, NJ: Erlbaum.

Northcrafi, G.B.. & Gutek, B.A. ( 1993). Point-counterpoint: Discrimination against women in management - Going. çoinç, gone or going but never gone? In E.A. Fagenson (Ed.) Women in Manaeement: Trends. Issues and Challenees in Manaeenal Diversitv.

Noseworthy, C.M., & Lott. A.J. ( 1984). The cognitive organization of gender- stereotypic categories. Personalitv and Social Psycholow Bulletin. 10, 474-48 1.

Numaily, J.C.(1 978). Psvchometric Theon). New York, NY:McGraw-Hill.

Olian, J.D.,Schwab, D.P & Haberfeld. Y .H.( 1988). The impact of applicant gender compared to qualifications on hiring recommendations: A meta-analysis of experimental studies. Organizational Behavior and Human Decision Processes. 4 1, 180- 195.

Ontario Human Rights Code, Revised Statutes of Ontario, Chapter H.19 ( 1990).

Ontario Human Rights Commission ( 1997). Em~ioynentA~oiication Foms and Interviews [On-line] . Available: http://www.ohrc. on.ca/english/publicat iondemp-app-forms-eng. km.

Parasuraman, S., & Greenhaus, J.H.( 1993). Persona1 portrait: The Iife-style of the woman manager. In E.A. Fagenson (Ed.) Women in Manaeement. Newbury Park, CA: Sage.

Pay Equity Act. Revised Statutes of Ontario, Chapter P.7 (1990).

Perlman, R.. & Pike, M. (1994). Sex Discrimination in the Labour Market: The Case for Comparable Worth. New York: St. Manin's Press, Inc.

Peterson, R.A. (1 983). Attitudes toward the childless spouse. Sex Roles. 9, 32 1- 33 1.

Pleck J. (1977). The work-family role system. Social Problems. 4,4 17-427 Pleck, J.H. (1987). Amencan fathering in historical perspective. In M.S. Kimmel (Ed.) Chaneine Men: New Directions in Research on Men and Masculinitv. Beverly Hills. CA: Sage.

Polachek, S. (1978). Sex differences in college major. Industrial and Labor Relations Review. 3 1, 498-508.

Pratto, F., & Bargh. J.A. ( 199 1 ). Stereotyping based on apparently individuating information: Trait and global components of sex stereotypes under attention overload. Joumal of Ex~enmentalSocial Psvcho~oev.27 26-47.

Reskin, B.F.(1988). Bringing the men back in: Sex differentiations and the devaluation of women's work. Gender and Societv. 3, 58-8 1.

Reskin, B.F. (1993). Sex segregation in the workplace. Annual Review of Sociolo~.19, 24 1-270.

Reskin, B.F.,& Padavic. 1. (1994). Women and Men at Work. Thousand Oaks. CA: Pine Forge Press.

Reskin, B.F.. & Roos, P. A. (1 990). Job Oueues. Gender Queues: Ex~laininq Women's Inroads into Male Occupations. Philadelphia. PA: Temple University Press.

Ridgeway. C. L. (1 997). Interaction and the conservation of gender inequality: Considering ernpioyment. Amencan Socioloeical Review. 62, 2 1 8-23 5.

Roberts, L.W. (198 1). Understanding affirmative action. In W.E. Block & M.A. Walker (Eds.) Discrimination. affirmative action and equal opportunitv. Vancouver. B.C.: The Fraser hstitute.

Rosen, B.. & Mende, M.F. ( 1979). Influence of strong versus weak fair employrnent policies and applicant's sex on selection decisions and salary recommendations in a management simulation. Journal of Applied Psvcholow. 64,435- 43 9.

Rosen, H.M.,& Korabik, K. ( 1995). The impact of children on women managers' career behavior and organizational commitment. Human Resource Management. 34, 5 13 - 528.

Rosenfeld, R.A., & Kalleberg, A.L. ( 199 1). Gender inequality in the labor market: A cross-national perspective. Acta Socioloeica. 34, 207-225.

Rothbart, M., & John, O.P.(1985). Social categorization and behavioral episodes: A cognitive analysis of the effects of intergroup contact. Journal of Social Issues. 4 1, 8 1- 104. Rowe, R., & Snizek, W.E. ( 1995). Gender differences in work values: Perpetuating the myth. Work and Occu~ations.22, 2 1 5-229.

Ruble. T.L.,Cohen, R., & Ruble. D.N. ( 1984). Sex stereotypes: Occupational bamers for women. Arnerican Behavioral Scientist. 27, 339-356.

Rusbult, C.E., Lower, D.. Hubbard, M.L., Maravankin. O.J.,& Neises, M. (1 988). Impact of employee mobility and employee performance on the allocation of rewards under conditions of contraint. Journal of Personality and Social Psycholow. 54, 605-6 15.

Russell, J.E.A., & Rush, M.C.(1987). The effects of sex and mantauparental status on performance evaluations and attributions. Sex Roles. 17. 22 1-236.

Rynes, S.L.. Weber. CL.. & Milkovich. G.T. (1989). Effects of market survey rates, job evaluation and job gender on job pay. Joumal of Applied Ps~cholow.74, 1 14- 123.

Schneer, J.A.. & Reitman, F. (1 995). The impact of gender as managerial careers unfold. Journal of Vocational Behavior. 47, 290-3 15.

Schuler, R.S. (1975). Sex, organizational level and outcome importance: Where the differences are. Personnel Psycholow. 28. 3 65-375.

Schwab, D.P., & Grams, R. (1985). Sex-related errors in job evaluation: A "real- world" test. Journal of A~pliedPs~cholow. 70, 533-539.

Seta, J.J., & Seta. C.E. (1 993). Stereotypes and the generation of compensatory and noncompensatory expectancies of group members. Personality and Social Psyholoey Bulletin. 19, 722-73 1.

Shapiro, D.M., & Stelcner, M. (1 987). The persistence of the male-female eamings gap in Canada, 1970-1980: The impact of equal pay laws and language policies. Canadian Public Policv. 13, 462-476.

Simpson, J. (1999). The politics of redress. The Globe and Mail, October 20. p A23.

Smith, J. (1987). Comparable worth, gender. and Human Capital Theory. In C. Bose & G. Spitze (Eds.) Ineredients for Women's Employnent Policv. Albany, NY: State University of New York Press.

Sorensen, E. (1991). Exdonne the Reasons Behind the Narrowing Gender Gap in Eaniin~s.Washington: The Urban Institute Press. Stangor. C.,Lynch L.. Duan, C.. & Glass. B. (1992). Categorization of individuals on the basis of multiple social features. Journal of Personalitv and Social Psvcholopv. 62, 207-2 1 8.

Starer, R.. & Denmark, F.L. (1974). Discrimination against aspinng wornen. International Journal of Group Tensions. 4, 65-7 1.

Statistics Canada. (1999a). Eamines of men and women. 1997 (Catalogue no. 13-2 1 7-XIB). Ottawa: Author.

Statistics Canada. ( 1999b). Labour Force Survev. 1 997 (Catalogue no. 7 1F0004- XCB). Ottawa: Author.

Statistics Canada (1 999c). Perspectives on Labour and Income (Catalogue no. 75- 000 1-XPE). Ottawa: Author.

Status of Women Canada (1995). Settine the Stase for the Next Century: The Federal Plan for Gender Eauality (Catalogue no. SW2 1 - 1 5/ 1995). Ottawa: Author.

Steinberg, R.J. (1995). Gendered instructions: Cultural lag and gender bias in the Hay system ofjob evaluation. In J.A. Jacobs (Ed.) Gender Inequalitv at Work. Thousand Oaks, CA: Sage Publications, Inc.

Stevens. C.K.. Bavetta, A.G.. & Gist, M.E. (1 993). Gender differences in the acquisition of sdary negotiations skills: The role of goals, self-efficacy, and perceived control. Joumal of Applied Ps~cholow.78. 723 -73 5.

Stone, D.L.,& Colella. A. (1 996). A framework for studying the effects of disability on work expenences. Academv of Management Review. 2 1, 352-401.

Strober. M.H.(1984) Toward a general theory of occupational sex segregation: The case of public school teaching. In B.F. Reskin (Ed.) Sex seareeation in the Workplace: Trands. Explanations. Remedies. Washington, DC: National Academy Press.

Swim, J.K.. Aikin, K.J., Hall, W.S..& Hunter. B.A. (1995). Sexism and : Old-fashioned and modem . Journal of Penonalitv and Social Ps~cholo~ 68(2), 199-2 14.

Swiss, D.J.,& Walker, J.P. ( 1993). Women and the WorkIFamilv Dilemma. New York, NY: John Wiley & Sons.

Szinovacz, M.E. (1984). Changing farnily roles and interactions. In B.B. Hess & M.B. Sussman (Eds.) Women and the Family: Two Decades of Change. New York: Haworth Press. Smajkowski, E., & Larwood, L. (1991). Rational decision processes and sex discrimination: testing 'rational' bias theoy. Journal of Oreanizational Behavior. 12, 507- 527.

Tabachnick, B.G.,& Fidell. L.S. ( i 989). Usinrr Muhivariate Statistics. Second Edition. New York: HarperCollins Publishers, Inc.

Tajfel, H. (1969). Cognitive aspects of prejudice. Journal of Social Issues. 35. 79- 97.

Thompson, L., & Walker, A. J. (1989). Gender in families: Women and men in marriage, work, and parenthood. Journal of Mamage and the Familv. 5 1, 845-87 1.

Tolben. P.S.. & Moen. P. (1998). Men's and women's definitions of"good" jobs: Similarities and differences by age and across tirne. Work and Occu~ations.25, 168- 194.

Tomaskovic-Devey. D. ( 1993). Gender and Racial Inequalitv at Work: The Sources and Consequences of Job Seereeation. Ithaca. NY: International Labor Relations Press.

Tomaskovic-Devey. D. ( 1995). Sex composition and gendered eamings inequality: A cornparison ofjob and occupational models. In J.A. Jacobs (Ed.) Gender Inequalit~at Work. Thousand Oaks, CA: Sage Publications.

Tougas, F., Brown. R., Beaton. A.M.. & Joly. S. (1995). Neosexism: Plus ca change, plus c'est pareil. Personalitv and Social Psvcholow Bulletin. 2 842-849

Tougas, F., Crosby, F., Joly, S., & Pelchat. D. (1993). Men's attitudes toward affirmative action: Justice and at the cross-roads. Social Justice Research. 8, 57-7 1.

Tougas. F., Joly, S ., Beaton, A.M.. & Si.-Pierre, L. ( 1996). Reactions of beneficiaries to preferential treatment: A realit check. Human Relations. 49, 453464.

Treiman. D.J.(1 984). Effect of choice of factor and factor weights in job evaluation. In Com~arableWorth and Wage Discrimination: Technical Possibiiities and Political Realities. Philadelphia, PA: Temple University Press.

Treiman, D.J.,& Hartmann, J.I. (1981). Women. Work and Waees: EquaI Pa? for Jobs of Eaual Value. Washington. D.C.:National Academy Press.

Trentham, S.. & Larwood. L. (1998). Gender discrimination and the workplace: An examination of rational bias theory. Sex Roles. 38, 1-28. Unger, R.K. (1990). Imperfect reflections of reality: Psychology constmcts gender. In R.T. Hare-Mustin & J. Maracek (Eds.) Makine a Difference: Psvcholo_w and the Construction of Gender. New Haven: Yale University Press.

Unger, R., & Crawford, M. (1992). Women and Gender: A Ferninist Psycholoq. New York, NY: Guilford.

Unger, R., & Saundra ( 1993). Sexism: An integrated perspective. in F. L Denrnark & M.A. Paludi (Eds.) Ps~choloeyof Women: A Handbook of Issues & Theories. Westpon, CT: Greenwood Press.

Valian, V. (1 998). Whv So Slow? Cambridge, MA: The MIT Press.

Veevers, J. ( 1 980). Childless bv Choice. Toronto: Buttenvorth.

Waldfogel. J. (1998). Understanding the "Family Gap" in pay for women with children. Journal of Economic Perspectives. 1 2, 137- 1 56.

Weber. R., & Crocker, J. (1983). Cognitive processes in the revision of stereotypic beliefs. Joumal of Personality and Social Psvcholow. 52, 700-709.

Weiss. R. S. (1 985). Men and the family. fa mil^ Process. 24, 49-58

Wellington, A.J. (1994). Accounting for the male/female wage gap among whites: 1976 and 1985. American Sociological Review, 59, 839-848.

Wente, M. (1997). Who pays for pay equity? The Globe and Mail, Septernber 27. p. D7.

West, C.,& Zimmerman, D.H. (1987). Doing gender. Gender & Society 1. 125- 151.

Wiley. M.G., & Eskilson A. ( 1988). Gender and familykareer confiict: Reactions of bosses. Sex Roles. 19, 445-466.

Wilson, S.J. ( 1996). Women. Families and Work (4th Ed.). Toronto: McGraw- Hill Ryerson Limited. Appendix A: Study 1 Questionnaire Pretend that you are an industnallorganizatioml p~chologistand we have hired you to carry oui a job evaluation. Job evaluation invoIves analyzing the content of \fririousjobs in order to establish a hierarchy withn an organization. You lme been provided witli a job description belorv. Ptease read ths description and carefully consider the difkent coiiiponents of the job wlien answenng the questions that

Job Title: Department store clerk - Hardware Greet custorners and discuss the type. quality ,and quantity of merchandise sought for purchase Advise customers on use and care of rnercliandisc Estimate or quote prices. credit terms. trade-in alloivances. wmties and delive- dates Prepare merchandise for purcliase Accept cash. cheque. credit card or automatic debit payment Assist in the display of merchandise Maintain sales records for inventory control

How would you rate tius job on the following diriiensions'? (Please circle the number corresponding to your amver)

Required skills and abilities: 1 7 3 1 5 6 7 physical mental

1 2 ? 4 5 6 7 simple comples

Required education and training: I 3 3 4 5 6 7 slio~ length~

1 7 3 4 5 6 7 general specialized

Responsibility for materials. 4 equipment. people: 1 2 -9 4 5 6 7 slight considerable

Job characteristics: 1 2 3 4 5 6 7 varied repetitive

1 -7 3 4 5 6 7 undemanding demanding

In your estimate. what annual saiary shodd be iissociated witli tiiis job? Pretend that you are an industnaYorganizational psychologist and we iwe hired you to carry out a job evaluation. Job evaiuation involves analyzing the content of various jobs in order to establish a hierarchy within an organization. You have been provided with a job description below. Please read this description and carefully consider the different coinponents of the job when answering the questions that foHow. - -- Job Title: Department store clerk - ChinaICrynal Greet custorners and discuss the type. qdity and quanti^ of merchandise sought for purchase Advise customers on use and care of it~erclraiidise Estimafe or quote prices, credit tenns. tradc-in nllowances. warranties and delivery dates Prepare merchandise for purchase Accept cash, cheque. credit card or automatic dcbii payment Assist in the display of merciundise Maintain sales records for inventon control

How would you rate this job on the following diiiiensions'? (Please circle the number corresponding to your answer)

Required skills and abilities: 1 2 3 4 5 6 7 p1iysical mental - 1 -3 3 4 5 6 t simple cornples

Required education and training: 1 2 3 4 5 6 7 short lengthy

1 2 3 4 5 6 7 general specialized

Responsibility for rnaterials. 4 equipment. people: 1 2 -3 1 5 6 7 slight considerable

1 2 3 4 5 6 7 specific broad

Job characteristics: 1 2 3 4 5 6 7 varied repeti tive

In your estimate. what annual salq shouid be associated with this job? Pretend that you are an industnaVorgani~tionalpsycliologist and we Imve hired you to cany out a job evaluation. Job evaluation involves analjzing the content of vanous jobs in order to establish a luerarchv within an orgtinization. You have ken provided witli a job description below. Please read this description and carefuily consider the different components of the job when answering the questions that follow.

Job Title: Teacher - Home economics

Prepare subject material for presentation to students according to ai approved curriculum Teach students using a systematic plan of lectures. discussions. audio-visual presentations Assign and correct homework Prepare. adrninister and correct tests Evaluate of students and discuss results with parents and school oEcials Participate in staff meeting. educational conferences and teaclier training workshops

How would you rate this job on the followving diineiisions? (Please circle the number corresponding to your answer)

Required skilIs and abilities: 1 2 3 4 5 6 7 p hysical rnentrtl

1 2 3 4 5 6 7 simple cornpies

Required education and training: 1 3 3 4 5 6 7 shon Iengthy

Responsibility for matenals. equipment. people: 1 2 3 4 5 6 7 siight considerable

1 2 3 4 5 6 7 specific broad lob characteristics: 1 2 3 4 5 6 7 varied repeti tive

In your estirnate, what annual salary should be associated witl~tius job? Pretend that you are an industriaVorganizationa1 psychologkt and we have hired you to carry out a job evaluation. Job evaluation involves analyzing the content of various jobs in order to establish a hierarchy within an organization. You have been provided witli a job description below. PIease read this description and carefiilly consider the different coriiponents of the job when mvering the questions that follow.

Job Titie: Teacher - Indusuial arts

Prepare subject materiai for presentation to students according to an approved curriculum Teach students using a qstematic plan of lectures. discussions. audio-visual presentations Assign and correct homework Prepare. administer and correct tests Evaluate progress of students and discuss results with parelits aiid school officiais Participate in staff meeting. educational conferences and tacher training workshops

How would you rate this job on the following dimensions'? (Please circle the number conesponding to your answer)

Required skills and abilities: 1 3 3 4 5 6 7 physical mentii

1 2 3 4 5 6 - simple comples

Required education and training: 1 2 3 4 5 6 7 short lengrhy

1 -7 3 4 5 6 7 gene ral specialized

Responsibility for materials equipment, people: 1 3 3 4 5 6 7 slight considerable

1 2 3 4 5 6 7 specific broad

Job characteristics:

In your estirnate, what annual salary shouid be associated with this job? Pretend that you are an industriaVorganizationa1 psycliologist and we have hired you to caq out a job evaluation. Job evaluation involves analyzing tlic content of wrious jobs in order to establish a hierarchy nithin an organization. You have bcen provided witli a job description below. Please read this description and carefùily consider the different coinponents of the job when answering the questions that follow. Job Title: Magazine Editor - Automotive Evaiuate suitabiiity of articles and recomrnend or inake changes in content. style and organization

Read and edit copy to be published to detect and correct erron i:i spelling. gnmmar. and syntas and shonen or lengthen copy as space or time rcquires 1 Confer with aurlion. staB witen. reporters and otliers rcgarding revisions to copy 1 Plan layout or fonnat of copy according to spacc allocations and significance of copy 1 Plan and coordinîte activities of stan and assure productioii deadlines are met 1. Plan covenge of upcoming events and assign work accordingly May negotiate royalties wih authors and initiate payments to freelance staff How wodd you rate this job on the foIlowîng dimcnsions'? (Plfiase circle the number corresponding to your answer)

Required skills and abilities: 1 2 3 4 5 6 7 physical mentaJ

1 3 J 4 5 6 7 simple cornples

Required education and uaining: 1 2 3 4 5 6 7 short length!

1 2 3 3 5 6 7 general speciaiized

Responsibility for materials, equipment, people: 1 2 3 4 5 6 7 siiglit considerable

1 2 3 4 5 6 7 specific broad

Job characteristics: 1 -9 3 4 5 6 7 varied repeti tive

1 2 3 4 5 6 7 undernanding demanding

In your estiniate. what annual saiaq should be associated with this job? Pretend that you are an indusviaYorganizational psycliologist and wc have hired you to cqout a job evaluation. Job evaluation involves rinalyzing tlie content of various jobs in order to establish a hrerarchy within an organization. You have been provided witli a job description below. Please rad ths description and carefully consider the diffiereni coniponents of the job ivlien annvering the questions that follow. Job Title: Magazine Editor -- Gounnet food Evduate suitability of articles and recornmend or ~iiakechanges in content. style and organization Read and edit copy to be published to detect aiid correct errors in spelling. grammar. and syntax and shorten or lengthen copy as space or time requires Confer with autliors. staff writers. reporters and otliers regarding revisions to copy Plan layout or format of copy according to space allocations and significance of copy Plan and coordinate activities of staff and assure production deadlines are met Plan coverage of upcoming events and assign work accordingly May ncgotiate royalties with autliors and initiale pnyiients to freelance staff

How wodd you rate this job on the following diinensions'? (Please circle the number corresponding to your answer)

Required skills and âbilities: 9 1 2 .? 4 5 6 7 pliysical mental

7 3 4 5 6 7 comples

Required education and training: 1 short

Responsibility for materials. equipment. people: 1 slight

2 3 4 5 6 7 broad

Job characteristics

1 undemanding

In your estimate. what annual salaq shouId be associated witli this job? Think again about the job htyou were asked to assess. Wliat are ihe cliaracteristics of the "tpical" person who does this job?

Do you think this person is: under 18 (circle only one)

Do you think tliis person is: Male

Do you think this person has ken in the workforce: less tiian 5 years 5-9 years 10- 14 years 15- 19 years 20+ years (circle only one)

Please provide us with ri description of any otlier "hpical"job holder characteristics that you think would be important. Background Information

How old are you'?

Are you: Male Female

What is your program of studylmajor'?

What is yow ethmc background?

Were yow parents born in Canada? - Fatlier: Yes - Motlier: Yes

Are you currently employed? Yes No

If so. wliat is your work?

What is your weekly saiaxy?

What was your last job (including surnrner jobs)'?

What was your weekly salary'?

Does your father work? Yes No

if so. what is his job?

What do you tfiink he earns annually'! $ Don't know

Does your mother work? Yes No

if so. what is her job'?

Mat do you think she earns annually'? S Don't know Appendix B: Study 1 Tables of Means and Surnmary Tables Table B- I

Factonal MANOVA on 9 Com~ensatorvFactors

Source Pillai's Hy~othesis Error -F B q2 Trace df -d f Job Incorne Level 0.75 18 422 13.97 .O00 .37"*

Participant Sex 0.04 9 210 0.86 -565 .O4

Target Job Gender 0.05 9 2 1O 1.33 221 .O5

Income x Participant O. 10 18 422 1.18 277 .O5

Income x Target 0.08 18 422 1.02 .436 .O4

Participant x Target O. 02 9 210 0.46 .899 .O2

Incorne x Part. x Targ. 0.05 18 422 0.57 -920 .O2 Table B-2

Com~ensatorvFactors -- Means and Standard Deviations by Job Income Level

Clerk Teacher Editor Overall M (sd) M (sd) M (sb) M (sd) Required skills and abilities P hysicaYmental 4.82 (0.96) 5.39 (0.86) 5.90 (0.83) 5.40 (0.98) Simple/complex 3.79 (1.18) 4.77 (0.93) 5.43 (0.96) 4.70 (1 22)

Required education and training Shortnong 2.96 (1.11) 5.08 (1.09) 4.84 (0.97) 4.34 (1.40) GeneraVspecialized 3.25(1.31) 4.84(1.23) 4.95(1.15) 4.39(1.44) Easy/difficult 3.00(1.03) 4.66(0.99) 4.89(0.98) 4.23(1.30)

Responsibility for matenals, equipment. people Slight/considerable 5.01 (1.29) 6.00 (1.01) 6.06 (0.89) 5.71 (1.16) Specifichroad 4.04 (1 .27) 4.36 (1.36) 4.09 (1.53) 4.17 (1.40)

Job charactenstics Undemandingdernanding 3.77 (1.10) 5.09 (1.23) 5.79 (0.90) 4.93 (1.36) Varied/repetitive(rev.coded) 4.44(1.41) 4.55(1.33) 3.71(1.62) 3.21(1.51)

Cronbach's a = 84 Note: Al1 items were rated on a 7-point scale. the first adjective in the pair served as an anchor for " 1" while the second adjective anchored "7". Table B-3

Factorial MOVA on Salary for Al1 Participants

Source Job Income Level

Participant Sex Target Job Gender

Income x Participant

Income x Target

Participant x Target

Inc. x Pan. x Target Error Total Table B-4

Factonal ANOVA on Sala-: Participants who agree with Stereotvped Gender of Job

Source Job Income Level

Participant Sex Target Job Gender

Income x Participant

Income x Target

Participant x Target

Inc. x Part. x Target Emor Total Table B-5

Correlations between Potential Covanates and Sala-/ bv Stereot~pedGender of Job and Job Income LeveI

Clerk's Salary Teacher's Salary Editor's Salary Assi.gned Gend: Male Female Male Female Male Female DEMOGR. Employed -.238 .O03 -. 130 -.298 -. 043 -.O3 1 Mom Works -. 156 -.O25 .?O9 -. 106 -. 3 20 -150 Recent Inc. .263 .15 1 ,172 -.213 .O35 -.O58 EMPLOYEE Time Working -361' .391° .364* -160 .220 .475** 9 Factors 377' .195 .177 ,389. .601m' S41" a 0 *p<.05.p<.O 1 Note: N's for these painvise correlations ranged from 26 to 35. Factonal ANCOVA on Salary with Leneth of Time in Workforce and Ratines on Compensatow Factors Covaried

Source Covariates Time in Workforce Compensatory Factors Independent Variables Job Income LeveI

Participant Sex Target Job Gender

Income x Participant

Income x Target

Participant x Target

Inc. x Pan. x Target Enor Total Table B-7

Mean Salam Estimates for 3 Types of Jobs by Stereotvoed Gender of Tareet Job for Male and Female Participants: Covariate Analyses

Gender Assigned to Job Male Female a** Main Efect of Participant Sex Male Fernale Male Female lob Income Level -M -M -M -M -M (sd)- (si!) (sd) (~çb) (sd) Job Income Level Store CIerk 23,550 33,235 22,397 18,727 2 1,789 (Lw (5,535) (9,554) (6,895) (4,693) (7,000) n= 16 n=17 n=14 n=20

Teac her 43,667 4 1,571 40,400 41,806 4 1,863 (Medium) (8,006) ( 1 1.264) (9,858) (8.6 1 1) (9,277) n=15 n= 14 n=15 n=l8

Magazine Editor 50,533 53.462 46,063 38,344 49,309 (Hkh) ( 14,086) ( 16.25 1 ) ( 12-438) (1 6,844) ( 14,779) n= 15 n=13 n= 16 n=15

Main Effect of 37,872 Participant Sex ( 14.5 8 5)

'~ainEffect of 3 8,464 35,804 Gender of Job ( 16,272) ( 15.50 1) & a*. p<.05, gc.00 1 Appendix C: Study 2 Mock lntem'ew Script Reurnble to interview - presented to participants via nudio-tape and on paper

[dl versions contained this introductory piece]

The goal of this study is to explore how impressions ofjob applicants are fomed during interviews. You will listen to an audio taped segment fiom an actual interview. The portion of the inteMew that you will hear is about five minutes long. Mer you have finished listening to the tape, we will ask you to evaluate both the interview itself and the suitability of the candidate for the job. Specifically, we are interested in: (1) your views on whether the questions asked dunng the inte~ewmake for an effective interviewing strategy, and (2) your assessrnent of whether the candidate would make a good employee based on the information obtained over the course of the interview -- we will ask you to evaluate the candidate on various dimensions, including a rating of how likely you would be to hire this particular candidate.

Please listen closelv to the audio taped segment as it will net be possible to replay the tape. Please feel free to make notes on the paper provided if you feel that it will help you to remember important elements. [spliced in one of these 2 versions] waie version] The segment you will hear occurs near the end of the interview -- the applicant, John Levishauskas. is 30 years old. holds an honours business degree. and he and his wife have recently moved to the Guelph area.

Fernale version] The segment you will hear occurs near the end of the interview - the applicant. Jane Levishauskas, is 30 years old, holds an honours business degree, and she and her husband have recently moved to the Guelph area. Mock Intetview -presented to parricipnn~~via audio-tape

[ail versions contain the following questions] Interviewer: Now let me tell you a bit about the job. You'd be assisting the branch manager in handling the bank's day to day administrative functions, including meeting with customers and dealing with their inquiries and concerns, preparing monthly financial statements and reports that go to head oifice, as well as getting involved in the branch's promotional activities and overseeing the work of the credit, Ioan and mortgage officen. 1 SEE FROM YOUR APPLICATION THAT YOU'VE WORIKED IN A BANK SETTING BEFORE. CAN YOU TELL ME A LITTLE BIT ABOUT TEIAT? Candidate: Before 1 applied for this position, I was working as a loan officer. 1 did that for a year, so I've had experience with al1 of the usual tasks associated with that job - interviewhg applicants for mortgages and other types of loans. evaluating references and credit ratings to determine whether an applicant is a good nsk, making recommendations for approval or rejection.

1: WHAT WAS YOUR SALARY AT YOUR PREVlOUS JOB? C: I was making about $32.000 a year.

1: OK... Just looking at your application again, 1 couldn't help but notice your last name - it's very unusual. WHAT'S YOUR ETHNIC BACKGROUND? C: People have made comments about my name before. 1 realize that it's a bit difEerent. My background is Lithuanian, but 1 was born in Canada.

1: Oh, 1 see. Well, back ta the task at hand,. . WEAT KIND OF SKILLS DO YOU BWGWITH YOU THAT MAKE YOU A GOOD EMPLOYEE? C: I think that in my capacity as a loans officer, 1 was able to Ieam a great deal about the requirements of smaller business owners as well as those of individual custorners who do their personal banking with us. 1 also believe that my business degree gives me a good understanding of the needs of a bank's corporate clients. 1 completed a management training prograrn while I was working as a loans officer which gave me the opporninity to sharpen my supe~soryskills.

1: Uh huh and WAAT PERSONAL QUALITES DO YOU BAVE THAT MAKE YOU A GOOD EMPLOYEE? C: I'm a very social person -- 1 like working with people. I've always chosen jobs that give me the opportunity to interact with different types of people. I've worked at a number of jobs that invoived a customer service component so I have a lot of experience in dealing with the public. I also seem to be good at handling difficult people. if someone is upset or dissatisfied. I always seem to be able to get them settled dom and work out a compromise, whether they're clients or coworkers. 1 think it's especially vaiuable to have good relationships with coworkers. 1: Right. And BAVE YOU EVER BEEN INVOLVED IN A DIFFTCULT SITUATION WITH A FELLOW WORKER, FOR EXAMPLE, EUVE YOU EVER BEEN SEXUALLY WSSEDBY ONE OF YOUR COWORKERS? C: I've been at a job where there was tension among some of the employees but I've only been involved as sort of a mediator. never as a direct party. 1 can recall one occurrence in particular where there were uncornfortable relationships between two people who worked together but it was over work-related issues, not persond or romantic involvernents. About the sema1 harassrnent in panicular. no, I've never had any experience with that.

1: How about managing difficult situations in your persona1 life... AT ANY OF YOUR PREVIOUS JOBS, BAVE YOU EVER USED AN EMPLOYEE ASSISTANCE PROGRAM TO DEAL WITH A PERSONAL PROBLEM? C: I think it's a great idea that companies have started to take steps to support their employees in that way but 1 can't say that 1 have used the prograrn myself.

1: Fair enough. WHAT ARE SOhiE OF THE THINGS YOU LIKE TO DO WEEN YOU'RE NOT WORKING? C: I'm a bit of an amateur photographer. 1 also like playing sports -- tennis and golf are two of my favourites. In the winter. I switch to cross-country skiing. 1 also enjoy reading in my spare time.

1: Sounds nice. So then ... WHY DO YOU WANT TO WORK FOR OUR COMPANY? C: 1 like the idea of working for a place that puts people first. this Company certainly seems to adhere to that philosophy. As you can see from my resume, 1 worked at one of your branches in Toronto just &er 1 graduated. At that time. there were a number of people who had more seniority than 1 did which limited my oppominities for advancement and 1 ended up moving out of the finance sector for a little while. 1 found that 1 missed it though, and I made the decision to return to banking. 1 liked the kind of environment that your cornpany provides and I'rn eager to get back on board with you.

1: And HOW DO YOU ViEW YOUR FUTURE WITH OUR COMPANY? C: 1 definitely have an eye toward upper management positions. I see myself taking every opportunity to develop my skills further so that in time. I'd be eligible to move from assistant manager status to branch manager and possibly beyond.

1: Well, we certainly are looking to hire someone who has the potential and desire to move ahead, although that sometimes requires reloeation. 1 SEE TEWT YOU'VE JUST RECENTLY MOVED TO GUELPH. WHERE ARE YOU FROM ORIGINALLY? C: I'rn from Toronto but I'm quite familiar with Guelph because 1 came here io go to school. Mer I finished my degree, I moved back to Toronto to find work. But 1 enjoyed my time in Guelph and &er giving it a lot of thought, my husbandlwife and 1 decided to move back here because we both really like this community. 1: HOW LONG HAVE YOU BEEN MARRIED? C: We've been mamed three years.

[splice one of these 4 versions] [Conditions 1,2 &31 1: That's great. You should know thrt the Company has a benefits package that includes coverage for spouses. DO YOU HAVE A FAMILY? [Condition 11 C: Well. we don't have any children yet but we are planning to have a farnily eventually. But for now, it's just the two of us. [condition 21 C: Well. actually. we're not able to have children. We gave the issue of adoption some thought but we've agreed that it'll just be the two of us. [condition 31 C: Well. actually, we have no plans to have any children. We've given the issue a lot of thought and we've both agreed that it'll just be the two of us.

[Condition JI 1: That's great. You should know that the company has a benefits package that includes coverage for spouses. DO YOU HAVE ANY BEALTH ISSUES TBAT OUR LNSURANCE COMPANY SHOULD KNOW ABOUT? [Condition 41 C: Well, 1 have asthma. It's usually controlled by medication but it sometimes happens that the condition is aggravated and 1 need to take a bit of time off work.

[dl versions contain this last piece] 1: OK. Well 1 think we've covered everything that 1 wanted to get through. It was nice to meet you. Before we finish up, are there nny other questions you have about the position or about anything else? C: Yes. just one -- when can we expect a decision about who got the job?

1: We're aiming to have al1 the interviews completed before the end of this week, so you should know by early next week. C: That's great. 1'11 look forward to hearing frorn you. I'm very excited about the possibility of working here, I think the company is very progressive and I'd like to have the opportunity to contribute. Thank you for meeting with me.

1: Thankyou. Appendix D: Study 2 Questionnaire nie inte~ewyou are about to hear is for tlie job presented below. Please read this job description and carefiilly consider the dinerent components of the job wlien annvering the questions that follow. lob Titie: Assistant Bank Manager Assia in the planning. organùing. direction and control of tlie branch operations of the bank Ennire the institution's policies and procedures are followed according to established guidelines and make recommendations for improvement Pmmote the mie of loan. invesunent and other banking senices to attract corponte and individual customers InteMew corporate and individual Ntomen and respond to customer enquiries Analne. review and approve or reject loan and credit applications Monitor processing of loan applications and credit investigations Oversee preparation of monthly financial and bnncli progress reports Recruit personnel and oversee tlieir training

How would you rate this job on the following dimensions'? (Please circle the nwnber corresponding to

Required skills and abilities: 1 2 3 4 5 6 7 physicai mental

1 2 3 4 5 6 7 simple comples

Required education and training: 1 short

2 3 J 5 6 7 dif3cul t

Responsibility for materials. equipment. people: I 2 - J 5 6 7 slight considerable

2 3 4 5 6 7 broad lob characteristics: 1 2 3 4 5 6 7 varied repe ti tive

1 undemanding The goal of this midy is to expIore how impressions ofjob applicants are formed during inte~eivs. You wiii listen to an audio taped re-enactment of the last segment of a job intewiew. The portion of the interview that you wiU hear is about five minutes long. Mer you have finished liaening to the tape. Ive wiii ask you to evaiuate both the intetview itself and the nlltability of the undidate for the job. Spedïcally. we are interesred in: (1) your views on whether the questions asked during the interview make for an effective interviewing strategy. and (2) pour assessrnent of whether the candidate would make a good employee baxd on the information obtained over the course of the intewiew - we will ask ?ou to evaluate the candidate on various dimensions. including a ratiiig of how (ikely you would be to hire this particular candidate.

Please listen closelv to the audio wped segiiient as it will abe possible to replay the tape. Please feel free to make notes on the paper provided if you fee4 that it will help you to remember important elements. The segment you will hear occm ntxr the end of the inteMew - the applicant. Jane Levishauskas. is 30 yean old. Iiolds an Iionours business degree. and slie and her husband have recently moved to the Guelph ara. SECTION 1: First. we'd like you to rate how weIl the interviewer and die candidate performed in the interview.

The interviewer was personable:

1 2 3 4 5 6 7 strongly agree suongly âisagree

The interviewer spoke clearly:

1 7 3 4 5 6 7 suongly agree strongly disagree

The interviewer had a good interviewing technique:

1 2 3 4 5 6 7 strongly rigree strongly disagree

The interviewer's behaviour wusimilar to an rictuül interviewer's behaviour:

1 7 3 4 5 6 7 strongly agree strongly disagree

The candidate wuconfident:

1 7 3 4 5 6 7 strongly agree strongly disagree

The candidate was hesitant in giving her answers:

1 2 3 4 5 6 7 stro ngl y agree suongiy disagree

The candidate made a good impression:

1 7 3 4 5 6 7 strongly agree strongly disagree

The candidate spoke clearly:

1 2 3 4 5 6 7 strongly agree strongly disagree

The candidate handfed the interview situation wefl:

1 2 3 4 5 6 7 strongly agree strongly disagree

The candidate's behaviour was similnr to ;ui actual candidate's behaviour:

1 7 3 J 5 6 7 strongiy agree strongly disagree SECTION 2: Ne.* we would like to get your iriipressions of the interview itself. We are intetesred in your ideas about what kind of questions make for an effective interview, generaily speahng. Below. we have presented the interview questions in the order tliat tliey appeared in the segment you have just heard. Please consider each question carewly and rate it on the sdes that follow. Circle the answer that corresponds most ciosely to your opinion.

1 see from your application that you've worked in ri bmk sctting before. Cm you tell me a little bit about that?

This type of question is:

1 -7 6 7 commonly asked rarely asked in interviews in inteniews

1 O9 6 7 inappropriate to appropriate to ask of applicants ask of appliczuits

1 2 6 7 relevant to job inelevant to job performance performance

What was your salary at your previous job?

This type of question is:

1 commoniy asked in interviews

I 6 7 inappropriate to appropriate to ask of applicants ask of applicants

1 6 7 relevant to job imlevant to job performance performance

What is your ethnic background?

This type of question is: 1 6 7 commonly asked rarely asked in interviews in interviews

1 6 7 inappropriate to appropriate to ask of applicants ask of applicants

I 6 7 relevant to job irrelevant to job performance performance What kind of skills do you bring with you that make you ii good employee?

This type of question is:

1 2 -1 4 5 6 7 coinmonly asked nrely asked in interviews in inte~ews

1 2 ? 4 5 6 7 inappropriate to appropriate to ask of applicants ask of applicants

1 2 I 4 5 6 7 relevant to job irrelevant to job performance performance

What personal qualities do you have that müke you a good employee?

This type of question is:

1 2 9 4 5 6 7 commonly asked nrely asked in intewiews in interviews

1 7 4 5 6 7 inappropriate to appropriate to ask of appliants risk of applimts

1 2 3 4 5 6 7 relevant to job irrelevant to job performance performance

Have you cver been involved in a diflicult ~ituatiunwith si feilow worker, for exmple, have you ever been sexually harasved by one of pur coworkcrs?

This type of question is:

1 6 7 commonly asked rarely asked in interviews in interviews

1 6 7 inappropriate to appropriate to ask of applimts ask of applicants

1 6 7 relevant to job irrelevant to job performance performance At any of your previous jobs, have you ever uscd an Ernployee Assistance Program to deal with a personal problem?

This type of question is:

1 2 3 4 5 6 7 comonly asked rarely asked in interviews in interviews

1 2 3 4 5 6 7 inappropriate to appropriate to ask of applicants ask of applicants

1 2 3 4 5 6 7 refevant to job irrelevant to job performance performance

What are some of the things you Iike to do when you're not working?

This type of question is:

1 6 7 commonly asked rareiy asked in interviews in intewiews

1 6 7 inappropriate to appropriate to ask of appticants ask of applicmts

L 6 7 relevant to job irrelevant to job perfomce performance

Why do you want to work for Our company?

This type of question is:

1 4 5 6 7 commonly asked rarely asked in interviews in interviews

1 4 5 6 7 inappropriate to appropriate CO ask of applicants ask of appliunts

1 4 5 6 7 relevant to job irrelevant to job performance performance How do you view your future with our company?

This type of question is:

1 2 3 4 5 6 7 commonly asked rare1y asked in interviews in inteMews

1 c.1 3 J 5 6 7 inappropriate to appropriate to ask of applimits ask of applicants

1 2 3 4 5 6 7 relevant to job irrelevant to job perfo mance performance

I set thrt you've just recently moved to Guelph. Where are ?ou from originally?

This type of question is:

1 6 7 commonl y asked mlyasked in interviews in interviews

1 6 7 inappropriate to appropriate to ask of applicants ask of applicants

1 6 7 reIevant to job irrelevant to job performance performance

How long have you been married?

This type of question is:

1 6 7 commonly asked rarely asked in interviews in interviews

1 6 7 inappropriate to appropriate to ask of applicants 3Sk of applicrints

1 6 7 relevant to job irrelevant to job performance performance [participants saw only of the following two questions. depending on their condition1

Do you have a family?

This type of question is:

1 -9 3 4 5 6 7 commonly asked rarely asked in interviews in interviews

i 1 2 -1 4 5 6 7 inappropriate to appropriate to ask of appticants ask of applicants

7 1 2 J 4 5 6 7 relevant to job imlevant to job performance performance

Do you have my hedtb issue3 tbat our invurance Company ~houldknow about?

This type of question is:

1 2 j 4 5 6 7 cornmoniy asked tarely asked in inteMcws in inte~ews

1 2 3 4 5 6 7 inappropriate to appropriate to ask of applicrints ask of applicants

i 1 2 .) J 5 6 7 relevant to job irrelevant to job performance performance SECTION 3: Now. we would like to get your impressions of the candidate. We are interested in your appraid of whether the candidate would make 11 good einployee in the long run based on the information obtained over the course of the interview. Plase rate the candidate on the following items. Using the scale providd please circle the answer that corresponds most closely to your opinion.

1 2 3 4 5 6 7 strongly ciiuiigree neither agree stroagly agree nor disagree

1. This candidate is uniikely to stay late at work.

2. The most important tiungs that happen to the candidate wilI involve her present job.

3. The candidate will often be too tired at work because of her life circumstances. 1 7 3 4 5 6 7

4. It will not be important to the candidate to keep advancing her career. 1 2 3 4 5 6 7

5. This candidate is likely to go to work erirly.

6. Most of the candidate's interests will be centred around lier job. 1 2 3 4 5 6 7

7. Demands reIated to the candidate's life circurnstances will be so great that they will take away from her work.

8. It will not be important to the candidate to be seen as very successful at her job.

9. This candidate is unlikely to bring work home and work during the evenings.

10. The candidate will be veq personally invoIved in her job.

II. Coworkers will dislike how often the candidate is preoccupied with her personai issues while at work.

12. The candidate will want vecy badiy to "make her mark" in lier ureer.

13. This candidate is likely to work on weekends. 1 2 3 4 5 6 7 strongly disrgree aeither agree strongly agree nor disagreg

14. The candidate will fml that her job is only a small part of who she is. 1 2 3 1 5 6 7

15. The candidate's life circumstances are likely to take up time that she should spend at work.

16. The candidate will be prepared to work long and hard to gain prominence in her axer.

17. This candidate is udikely to travel for business purposes.

18. The candidate's goal will be to reach the "top'' in lier career. SECTION 4: Hiring decisions are usually made by a selectioii panel comprised of severai people. In these situations, each penon's opinion is important althougli the final decision regarding whether or not a person is hired 1s based on liow the majority of panel iiieinbers vote. Assuming tiut you were part of such a panel, how would you vote'? in oii~erwords. Iiow likely woutd you be to hire this candidate'? Ptease circle the answer that corresponds most closely to your opinion.

1 2 3 4 5 6 7 not at al1 likely very tikely

We are interested in die factors that influenced your hiring decision. Please indicate whether the following considerations played a part in your decision.

1. (a) How well do you think this candidate would fit in witli the company?

1 2 3 4 5 6 7 not at al1 very well

(b) Did the issue of whether the candidate would "fit in" witli the Company influence your dccision? Yss No -

2. (a) ifths applicant were iiired. Iiow long (in years) would you predict thths candidate would star. with the company?

(b) Did the issue of liow long the candidate is likely to stay witli the cornpany influence your decision'? Yes No -

3. Please descnbe any other factors tht you considered wlieti inaking your decision about how likeiy 'ou would be to hire this candidate.

4. Assume that the majority of panel inembers voted to liire tliis candidate. The salaq for an assistant bank manager ranges behveen $20.700 and $68,900. Given wliat you have learned about the candidate during the job interview. wliat ,mual salm would ?ou recomiiiend for dus candidate'? SECTION 5: The questions in ttiis final section focus on mrpersonai background and family situation.

How old are you?

Are you: Male Fernale

Are you: Single Mamed/living witli a boflriend or girlfnend

What is your program of study/major'?

Are you a: Fdl-time student Part-tirne student

Are you currently employed? Yes No

If so, what is your work'?

What is your hourly wvage'?

How myhours do you usually work in a week?

Wiat was your Iast job (including summer jobs)'?

What was your hourly wage?

How many houn do you usudly work in ri week?

Have you ever held a job where you supervised oilier employees?

Yes No

if so, what was this job'?

Were your parents born in Canada'? - Fatlicr: Yes No - Mother: Yes No

Does your father have paid work? Yes No

If so. what is his job?

What do you think lie emsannually'? E Don't know

Does your mother have paid work'? Yes No

if so. wht is her job?

What do you thinic she eams amually'? $ Don't know Appendix E: Study 2 Tables of Means and Summary Tables Table E- 1

Principal Com~onentAnalvsis of Items Related to Candidate's Suitability as an Emolovee

Candidate characteristic/behaviour Loading h2 Availability for overtime Uniikely to stay late at work Likely to go to work early Unlikely to bring work home in the evening Likely to work on weekends Udikely to travel for business purposes

Job involvement Most important things will involve job Interests wiil be centred around job Very personally involved in job Job is only a small pan of who s/he is

Persona1 life interference with work Too tired at work because of life circumstances Life dernands so great they take away from work Co-workers dislike preoccupation with personal issues Life circumstances take up work time

Ambition Not imponant to keep advancing career Not important to be seen as successful at job Will want to "rnake a mark" in career Prepared to work hard to gain prominence Goal is to reach the "top" in career

Percentage of variance in variables accounted for 19.3 1% Note: Al1 items were rated on a 7-point scale ranging from 1 "strongly disagree" to 7 "strongly agree." Table E-2

Mean Ratings of Compensatow Factors

Compensatory factor M U)

Required skills and abilities P hysicaVmenta1 Simple/complex

Required education and training S hodlong GeneraVspecialized Easy/diffi cu 1t

Responsibility for rnatenals. equi pment . people Slightkonsiderabie Specifichroad

Job characteristics Undemanding/demandinç

Note: Al1 items were rated on a 7-point scale where the first adjective in the pair served as an anchor for "1" while the second adjective anchored "7". Table E-3

Factorial MANOVA on Ratines of Interviewer

Source Pillai's Hvpo t hesis Trace -d f Participant Sex 0.02 4 Farnily/Health S tatus 0.05

Applicant Gender 0.0 1

Participant x Family/Xealt h 0.03

Participant x Applicant 0,03

FamilylHealth x Applicant 0.03

Pan. x FamilyRIealth x Appl. 0.05

Table E-4

Mean Ratinas of Interviewer's Performance

- -- -- InteMewer ' s performance -M (sd)

Personable 2.68 (1.15)

Spoke clearly 1.96 (1.07)

Good interviewing technique 3.64 (1.47)

SimiIar to actual interviewer 3.29 (1 .52)

Note: Al1 items were rated on a 7-point scale where 1 = "strongly agree" and 7 = "strongly disagree". Table E-5

Factorial MANOVA on Ratines of Candidate

Source Pillai's Hypothesis Error F E 'I? Trace -df -df Participant Sex O.06 6 292 2.84 .O1 1 .06*

Family/Health Status 0.07 18 882 1.24 223 .O3

Applicant Gender 0.09 6 292 4.76 .O00 .09***

Participant x FamilykIealth 0.05 18 882 0.77 ,732 .OS

Participant x Applicant 0.01 6 292 0.43 .858 .O1

Family/Health x Applicant 0.06 18 882 0.98 .481 .O2

Part. x Family/Health s Appl. 0.08 18 882 i.37 .138 .O3 Tabie E-6

NOVA on Ratines of Candidate - Confident

Source Participant Sex

FamilyRIealt h S tatus 3.72 3 1.14 1.63 .182 .O2

Applicant Gender 5.29 1 5.39 6.97 .O09 .O2

Participant x FarnilylHealth O.29 3 0.10 O. 13 .943 .O0

Participant x Applicant O. 35 1 0.25 0.33 ,566 .O0

Family/Health x Applicant 1.81 3 0.60 0.79 .498 O1

Participant x FarniIyHealth x Applicant 3.75 3 1.25 1.65 A79 .O2

Total 249.83 313

am **a g<.008, p<.OO 1

Note: Critical p-values for this analysis were adjusted using a Bonferroni correction. A minimum p-value of p<.OO8 was required to deem a test statistically significant. Table E-7

ANOVA on Ratings of Candidate -- Hesitant

Source Participant Sex FamilyMealth Status Applicant Gender Participant x Family/Health

Participant x Applicant

Family/Health x Applicant

Participant x Family/Health x Applicant Error Total

Note: Critical p-values for this analysis were adjusted using a Bonferroni correction. A minimum p-value of gc.008 was required to deem a test statistically significant. Table E-8

ANOVA on Ratings of Candidate - Good Im~ression

Source Participant Sex

Applicant Gender

Participant x FamilyRiealth Participant x Applicant

FamilyNealth x Applicant

Participant x FarnilyRiealth x Applicant Error Total

Note: Critical p-values for this analysis were adjusted using a Bonferroni correction. A minimum p-value of p<.008 was required to deem a test statistically significant. Table E-9

ANOVA on Ratines of Candidate - Spoke Clearlv

Source Participant Sex

Applicant Gender 12.95 1 12.95 11.97 .O01 .04"'

Participant x FamilyMealth 0.58 3 0.20 0.18 .910 .O0

Participant x Applicant 0.0 1 1 0.01 0.01 .928 .O0

Family/Heaith x Applicant 2.7 1 3 0.90 0.84 ,476 .O1

Participant x FamilyRIealth x Applicant 2.79 3 0.76 0.71 30 .O1

Error 321.15 297 1.08 Total 345.69 313

Note: Criticai p-values for this analysis were adjusted using a Bonferroni correction. A minimum p-value of g<.OO8 was required to deem a test statistically significant. Table E- 10

ANOVA on Ratinrrs of Candidate - Handled Interview Well

Source Participant Sex Farnily/Health Sratus Applicant Gender

Participant x Family/Health

Participant x Applicant

FamilylHealth x Applicant

Participant x FarnilylHealth x Applicant Error Total

Note: Critical p-values for this analysis were adjusted using a Bonferroni correction. A minimum p-value of pc.008 was required to deem a test statistically significant. Tab te E- 1 1

ANOVA on Ratines of Candidate - Similar to Actuaf Candidate

Source Participant Sex FamilyMealth Status Applicant Gender

Participant x FamilyMealth

Participant x Applicanr

Family/Health x Applicant

Pariicipant x Family/E-Iealth x Applicant Error Total

Note: Cntical p-values for this analysis were adjusted using a Bonferroni correction. A minimum p-value of g<.OO8 was required to deem a test statistically significant. Table E- 12

Ratings of Candidate's Performance bv Gender of Participant. Gender of Tareet

Gender of Partici~ant Gender of Tareet Male Fernale Male Female Candidate's performance M (d) -M (sd) -M (sd) M (4)

Confident 1.72 (0.98) 1.37 (0.77)"' 1.66(1.03) 1.41(0.72)

Hesitant 5.97 (1.17) 6.16 (1.32) 5.88(1.33) 6.26 (1.14)"

Made a good impression 2.1 1 ( 1.14) 1 -78 (0.99)"' 2.04(1.15) l.gS(0.97)

Spoke clearly 1.86 ( 1.15) 1.59 (0.95) 1.92(1.13) 1.51 (0.94)"

Handled interview well 1.94 (1 .07) 1.59 (1.00)" 1.85 (1.15) 1.66(0.93)

Sirnilar to actual candidate 3.15 ( 1.33) 3.80 (1.19) 2.80 (1.26) 3.13 (1.26) 0. e<.008. .mm p< O0 1

Note: Al1 items were rated on a 7-point scale where I = "strongly agree" and 7 = "strongly disagree". Also. criticai p-values for this analysis were adjusted using a Bonferroni correction. A minimum p-value of gc.008 was required to deem a test statisticaily significant. Table E- 13

Factorial MANOVA on FarnilvMealth Status Ouerv: Commomess. Ap~ropnateness, Relevance

Source Pillai's E-&pothesis Error -F E r12 Trace -d f -Df Participant Sex 0.02 3 294 1.47 223 .O2

Family/Health Status O. 08 9 888 2.62 ,005 .03" Ap plicant Gender 0.03 3 294 2.63 050 .03'

Participant x FarnilykIealt h O. 03 9 888 1.07 .382 .O1

Participant x Applicant O.03 3 294 2.85 ,038 .03*

FamilyRIealth x Applicant 0.02 9 888 0.63 -774 .O1

Part. x Family/Health x Appl. O.02 9 888 0.72 .688 .O1 Table E-14

ANOVA on Farnilv/Health Status Ouerv: Commonness

Source Participant Sex FamiIy/Health Status Applicant Gender

Participant x FarnilyHealth

Participant x Applicant

FarnilyMealth x Applicant

Participant x FamilyRIealth x Applicant Error Total

Note: Critical p-values for this analysis were adjusted using a Bonferroni correction. A minimum p-value of g<. 0 1 7 was required to deem a test statistically significant. Table E- 15

ANOVA on FamilvMealth Status Ouery: A~orooriateness

Source

Participant Sex Family/Health Status Applicant Gender

Participant x FamilyfHealth

Participant x Applicant

Family/Health x Applicant

Participant x FarnilyMealth x Applicant Error Total

Note: Critical p-values for this analysis were adjusted using a Bonferroni correction. A minimum p-value of pc.0 17 was required to deem a test statistically significant. Table E- 16

ANOVA on Farnilv/Hedth Status Ouerv: Relevance

Source

Participant Sex FamilyMealth Status Applicant Gender Participant x FarnilyMealth

Participant x Applicant

FarnilyMealth x Applicant

Participant x FamilyRlealth x Applicant Error Tot al

Note: Critical p-values for this analysis were adjusted using a Bonferroni correction. A minimum p-value of p<.0 17 was required to deem a test statisticaily signifiant. Table E- 17

FamilvfHealth Status Ouenes: Mean Ratines of their Relevance bv Gender of A~dicant and Gender of Partici~ant

Gender of Applicant

Male FemaIe '*~ainEffea of Participant Sex Male Female Male Femaie FarnilyEIealth -M -M -M M -M

Familv/Health Status Want children 5.61 ( 1.33) n= 18

Can't have children 4.39 (1.79) n=I 8

Don't want children 5.16 (1.64) n=19

Has asthma

Main Effect of Participant Sex

'~ainEffect of 4.85 Applicant Gender (1.80)

Note: All items were rated on a 7-point scale. where 1 = "relevant to job performance" and 7 = "irrelevant to job performance." Aiso. critical p-values for this analysis were adjusted using a Bonferroni correction. A minimum p-value of pc.0 17 was required to deem a test statistically significant. Table E-18

Factorial MANOVA on A~dicantDedication. "Fittine In" and Turnover Potential

Source Pillai's HVDOthesis Error -F Q q2 Trace -Df -d f Participant Sex 0.04 3 274 4.10 .O07 .O4

Applicant Gender 0.00 3 274 0.41 .748 .O0

Participant x FarnilyMealth 0.03 9 828 0.94 .492 .O1

Participant x Applicant O. O0 3 274 0.13 .944 .O0

FamilyIHealt h x Applicant 0.07 9 828 2.32 .O14 .O3 Part. x FarnilyMealth x Appl. 0.05 9 828 1-64 .IO0 .O2 *** 'p<.05. "e<.~1. p<.OO I Table E- 19

Factorial ANOVA on Candidate's Dedication as an Employee

Source Participant Sex FarnilyIHealth Status Applicant Gender Participant x Family/Health

Participant x Applicant

~arnil~/'Healthx Applicant Part. x FamiIy/Health x Appl.

Total

Note: Critical p-values for this analysis were adjusted using a Bonferroni correction. A minimum p-value of pC.0 t 7 was required to deem an effe~statistically significant. Table E-20

Factorial ANOVA on How Weil Candidate Would "Fit In"

Source Participant Sex FamilylHealth Status Ap plicant Gender

Participant x FamilyMeaith

Participant x Applicant

FarnilyRIeaith x Applicant

Part. x FarnilyMealth x Appl.

Total

Note: Criticai p-values for this analysis were adjusted using a Bonferroni correction. A minimum p-value of pc.017 was required to deem an effect statistically significant. Table E-2 1

Factonal ANOVA on Number of Years Candidate WiII Stav with Com~anv

Source Participant Sex

Applicant Gender Participant x Farnily/Health Participant x Applicant

FarnilyMeaith x Applicant

Part. x FamilyIHealth x Appl Error Total

Note: Critical p-values for this analysis were adjusted using a Bonferroni correction. A minimum p-value of ex.0 1 7 was required to deem an effect statistically significant . Table E-22

Factorial ANOVA on Candidate's Likelihood of Beine Hired

Source -SS -Df MS -F P s' ParticiparatSex 18.49 1 18.49 29.26 .O00 -10

Farnily/Health Status 5.17 3 1.72 2.73 .O45 .O3

Applicant Gender 2.35 1 2.35 3.72 .O55 .O1

Participant x Family/HeaIth 1.76 3 0.59 0.93 .427 .O1

Participant x Target O. 73 1 0.73 1.15 285 .O0

Family/Hedth x Target 0.36 3 0.09 O. 14 .938 .O0

Part. x FarnilyMealth x Appl. 2.88 3 0.96 1.52 210 .O2

Total 204.06 259 Table E-23

Factorial ANOVA on Startine Salarv Assiened to Apolicant

Source Participant Sex

Applicant Gender

Participant x FarnilylHealth

Participant x Applicant

Family/Health x Applicant

Part. x FamilyfHealth x Appl. Error Total Appendix F: Study 3 Questions KWMAS questions

Suppose you were a boss or supervisor who needed to fil1 a vacant position at the organization where you work. To make the task easier. let's use a specific example - imagine that you work for one of the major Canadian banks and you are evaluating applicants for the position of assistant bank manager. One of the candidates for this position, Jane Smith. is a married 30 year old woman who has worked in a bank setting for a year pnor to applying for this job.

We would like to get your impressions of the persona1 characteristics that would make Jane more or less desirable as an employee. We are interested in your assessrnent of whether Jane would be at an advantage or at a disadvantage as a result of having the following charactenstics. Using the scale provided, please circle the number that corresponds most closely to your opinion.

- 3 -2 - 1 O + 1 +2 +3 This characteristic neutral This characteristic is a disadvantage is an advantage when applying for when applying for a job a job a) Jane lives in another city and must spend an hour commuting to work. b) Jane and her husband belong to vanous clubs and organizations in the cornmunity. c) Jane and her husband are planning to have a family in the tùture. d) Jane and her husband are unable to have children. e) Jane and her husband are the parents of 3 young children. f) Jane and her husband do not want to have children. g) Jane and her husband have extended farnily that live nearby. h) Jane and her husband are responsible for the care of an elderly parent. i) Jane and her husband live 500 kilometres from their nearest relatives. Appendix G: Study 3 Tables of Means and Summary Tables Table G- 1

Mixed-desien ANOVA on 9 Ernplo~eeC haracteristics: Al1 Respondents

Within-Subiects Effects Family/Living Arrangements

Family x Resp. Sex

Farnily x Applicant Gender

Family x Resp. x Appl. Error Between-Subiects Effects Respondent Gender Applicant Gender

Resp. x Appl. Error

Note: Degrees of freedom for within-subjects effects were corrected using the Greenhouse-Geisser correction. Table G-2

Mixed-design 3-wa~ANOVA on 4 fa mil^ S tatus Items: Al1 Respondents

Within-Subiects Effects -SS -d f -MS F -P q2 FamilylLiving Arrangements 2.4 1 1.74 0.80 0.79 SOO .O0

Family x Resp. Sex 1 1.66 i -74 6.70 3.81 .O28 .01*

Family x Applicant Gender 103.33 1.74 34.44 33.79 O00 .IO***

Family x Resp. x Appl. 1.66 1.74 0.55 0.54 ,557 .O0

Error 987.82 562.2 1 1 .O2 Between-Subiects Effects

Respondent Gender 2.94 1 2.94 2-17 ,142 .O1

Applicant Gender 7.48 I 7.48 5.52 .O19 .02*

Resp. x Appl. 1.01 1 1-01 0.75 ,388 O0

Error 43 7.86 3 23 1.36

Note: Degrees of freedom for within-subjects effects were corrected using the Greenhouse-Geisser correction. Table G-3

Mked design 3-wav ANOVA on 4 Family Status Items: Full-Time Non-Managers Odv

Within-Subiects Effects Family/Living Arrangements

Family x Resp. Sex

Farnily x Applicant Gender

Farnily x Resp. x Appl. Error Between-Subjects Effects Respondent Gender Applicant Gender

Resp. x Appl. Error Table G-4

Mked design 3-wav ANOVA on 4 Family Status Items: Managers Only

Within-Subiects Effects -SS -d f -MS F E ri2 Family/Living Arrangements i .60 1.57 1.02 0.61 ,509 .O1

Family x Resp. Sex 4.93 1.57 3.14 1.86 ,169 .O3

Farnily x Applicant Gender 12.20 1.57 7.78 4.62 -019 .08.

Family x Resp. x Appl. O.52 1.57 0.33 0.20 .768 .O0 Error 142.76 84.65 1.69 Between-Subiects Effects

Respondent Gender 0.21 1 0.21 O. 17 ,687 .O0

Applicant Gender 2.53 1 2.53 1.98 165 .O4

Resp. x Appl. 0.36 1 0.36 0.28 398 .O1

Error 68.95 54 1.28 *E<.05."P<.~ 1. mm. pc.001 Table G-5

Intercorrelations Among Potential Covariates and De~endentVariables bv Gender of Tarnet and Participant Gender for Al1 Respondent s

Want Cliildreri Can'i Have Childreri Ha\x 3 Children Don'i Want Children Torgcl Gend: Joliri Junt Johi JUIW Joliii Jsnr John 3 wic Pur(. Ses: M F M F M F M 1: M 1: M F M F M F DEMOGR. Edric. -.224 -.234. Incoriic -.274* -, 149 Occ'l I,t.vd .O52 ,025 Iiavc Kids -.O20 -.O28

A'l"l'1'llY Il< lui111 - 074 -. 1 1 8 -. 160 3 -.ON4 07 12 1 . 134 -.2i)l -. 103 -.24-1° -.O96 ,055 -.O08 .O87 .274** Gmkr 11i1p -. 150 .il39 -.O94 ,103 -.O78 ,020 - 146 -. 174 -.O63 -.O16 -.O45 ,088 - 8 .O05 - IO2 -. 166 Work Inil,. O 5 -.O66 .iW I ,096 - 074 -. 127 - 150 - l Il .O43 - 157 ,028 -.O26 -.O94 .O01 1 -.O27 AIT. Actioii .O07 01 6 1 90 122 -. 189 -,O07 026 . IO8 -.O74 1 .34 1 .* .O84 -.O82 -. 13 1 .O38 . 125 Fcm clisudv. -.285* ,O19 -. 103 -.O21 O45 180 1 3 -.Ili) .IO4 -.141 -.284*- -.O93 ,130 .010 .126 Barriers -.225 -. 145 .O92 ,175 .O50 ,098 -1 -137 .O02 -0-12 O 6 -065 ,043 -.O97 ,020 Note: TIic correlaiions reported arc painvise, the n's for ihcse relationships ranged from 55 to 10 1.

Educ. = higlusi level or cducaiion ( 1=less I han high scliool, 5=graduaie degrec) Inc = estimated pearly incomc Occ'l Ski11 = NOC skill lewt (l=nianageriaVprofcssional, .I=labouring/elen~eiiral) Have Kids = respondent is a parent (l-yes, O=no) Family Iinp. = Iiow iinponant are family relaiionships to your self-identity ( 1 =ver). iniportant, l=noi iriiportant ai all) Geiidcr Iiiip. = Low irnponani is king male or fernale to your self-ideniity (l=very iniponanc. J=not important at all) Work Iinp. = how important is work to pur self-identity (I=very importani, J=iioi iniponani ai all) Am. Action = 1 am in frivour of the lypical aTfiriiiative action prograin ( I =sirorigly disagree, 7=siroiigly agrec) Fein. Disiidv. = \\.oiiieri are disadvantagcd by negative gender sicrcoiypcs rcgiirding ilieir abilit~IO perîoriti u.ork ( I =strongly disagree, 7=sirongly agree) Barricrs = barricrs in ilic workplace rediice wonien's hiring and proiiioiioii ( 1-stroiigly disagree. 7=strongly agree)