Happy to Take Some Risk: Investigating the Dependence of Risk Preference on Mood Using Biometric Data
Total Page:16
File Type:pdf, Size:1020Kb
Happy to Take Some Risk: Investigating the Dependence of Risk Preference on Mood Using Biometric Data Bachir Kassas, Marco A. Palma, and Maria Porter Food and Resource Economics Department University of Florida [email protected] DRAFT Please do not circulate or cite without author permission Selected paper prepared for presentation at the 2019 Agricultural and Applied Economics Association Annual Meeting, Atlanta, Georgia, July 21-23, 2019 Copyright 2019 by Stephen Morgan and Bachir Kassas. Readers may make verbatim copies of this document for non-commercial purposes by any means, provided that this copyright notice appears on all such copies. 1. Introduction The importance of risk preferences in driving individual decisions under a wide range of economic settings has intensified the focus on understanding the main factors underlying individual behavior under risk. One of the key components found to affect risk-taking behavior is the mood state under which individuals make their decisions. In fact, a large body of work has documented the impact of mood on risk preferences (Cahlíková & Cingl, 2017; Drichoutis & Nayga Jr, 2013; Fehr-Duda, Epper, Bruhin, & Schubert, 2011; Kamstra, Kramer, & Levi, 2003; Kliger & Levy, 2003; Kuhnen & Knutson, 2011). However, despite the multitude of research on this topic, the literature carries contradictory results, implying that the mechanism through which mood influences how individuals make decisions under risk is still a subject of debate. This study is aimed at examining differences in the reported effect of incidental mood on risk attitudes arriving from key elements in the experimental protocol, namely the mood measurement manipulation check and risk preference elicitation mechanism used. To this end, facial expression analysis technology and pupil dilation measures are utilized as an alternative to the conventional survey-based mood measurement task when investigating the effect of positive and negative moods on risky decisions. The value of incorporating psychophysiological data in the analysis of the effect of mood on risk preferences, and individual behavior more generally, is underscored by highlighting a major drawback (mood dilution) from using surveys to elicit subjects’ moods. The results are compared across two popular risk preference instruments in the experimental economics literature, Holt- Laury (HL) and Eckel-Grossman (EG), to assess the stability of risk preferences, and treatment effects, under different elicitation mechanisms (Eckel & Grossman, 2002; Holt & Laury, 2002). The commonly adopted procedure for investigating the effect of incidental mood on risk preferences in the laboratory is to conduct a three-stage design, which is comprised of mood inducement, mood measurement, and risk preference elicitation (Bruyneel, Dewitte, Franses, & Dekimpe, 2009; Conte, Levati, & Nardi, 2018; Drichoutis & Nayga Jr, 2013; Fessler, Pillsworth, & Flamson, 2004; Kim & Kanfer, 2009; Treffers, Koellinger, & Picot, 2016). The mood measurement stage, which serves as a manipulation check to test the effectiveness of mood inducement, is usually conducted through self-reported surveys, predominantly the Positive and Negative Affect Schedule (PANAS) survey. We argue for a mood dilution that results from imposing an intermediate stage (taking the survey) between mood inducement and risk preference elicitation. We show that subjects’ induced mood is significantly diluted by the time they complete the mood measurement survey and are ready to reveal their preferences in the risk task. We demonstrate the bias in the results caused by this mood dilution, which can be resolved by using psychophysiological data instead of surveys for mood measurement. While testing the effect of positive and negative moods on risk preferences, we further divided each mood treatment (positive, negative, and neutral) into two groups. The first group, hereafter diluted group, followed the conventional three-stage design by sequentially completing mood inducement, followed by mood measurement using PANAS, and finally risk preference elicitation using HL and EG.1 Conversely, the second group, hereafter undiluted group, skipped the intermediate PANAS survey and went straight through from mood inducement to the risk preference elicitation stage. Facial expression analysis software and pupil dilation were used on both groups (diluted and undiluted) to measure subjects’ moods before and during mood inducement and right before they started the risk preference tasks. 1 The order of the HL and EG tasks was randomized across subjects to account for any ordering effects. Furthermore, the stakes were normalized across the two tasks to get a more comparable measure of risk preferences that is not driven by differences in stakes. Since the HL task is prone to inconsistencies (subjects having multiple switches in the multiple price list), while the EG does not allow for inconsistent behavior, inconsistent subjects were dropped from the analysis to allow for a more direct comparison of the treatment effects across the two mechanisms. Our results point to a significant mood dilution among the diluted group, who went through the conventional three-stage experimental design. This was evident in the comparison of the mood measurement of those subjects during mood inducement and right before starting the risk task, which showed a substantial decay in the induced mood as a result of the intermediate PANAS survey. This mood dilution was further manifested in the treatment effects, where there was no significant change in the risk preferences of the diluted group across the mood treatments as opposed to a significant decrease in the risk aversion of both positive and negative mood treatments in the undiluted group. Based on the results obtained from the HL and EG tasks, the effect of mood on risk preferences did seem task dependent. Specifically, the treatment effects were more pronounced under the HL task, where the decrease in risk aversion was significant under both the positive and negative mood treatments. Conversely, the risk preferences obtained from the EG task showed no significant effect for the negative mood treatment and only a marginally significant effect for the positive mood treatment. The contribution of this study is threefold. First, it highlights some of the useful applications of psychophysiological data in behavioral and experimental economics research. Second, it underscores a major issue (mood dilution) in using conventional survey methods to elicit subjects’ moods, thus proposing a modification that can enhance the accuracy of experimental designs used to study the effect of incidental mood on individual preferences. Third, it demonstrates the dependency of the relationship between induced mood and risk preferences on the choice of risk preference elicitation mechanism, whereby highlighting the importance of the experimental task used and its influence on the observed treatment effects. Perhaps a broader benefit from this research is derived from the significant role that mood plays in shaping our behavior. The literature addressing the effect of mood on individual behavior is ever-growing across several disciplines including economics, psychology, sociology, marketing, and nutrition. For instance, positive mood has been linked to enhanced cognitive ability, productivity, patience, and healthy eating behavior (Erez & Isen, 2002; Fedorikhin & Patrick, 2010; Ifcher & Zarghamee, 2011; Isen, 2008). On the other hand, negative mood has been implicated with criminality, drug abuse, obesity, overspending, immorality, and antisocial behavior (Blumenthal, 2005; Burke Jr, Burke, & Rae, 1994; Cryder, Lerner, Gross, & Dahl, 2008; Ganem, 2010; Lerner, Small, & Loewenstein, 2004; Weiss, Griffin, & Mirin, 1992; Zeeck, Stelzer, Linster, Joos, & Hartmann, 2011). This highlights the importance of developing experimental protocols that can accurately gauge the effect of mood on individual preferences. The rest of the paper is organized as follows: Section 2 reviews the relevant literature, while section 3 describes the experimental design and data. Section 4 presents a simple theoretical model that was used to obtain point estimates of the coefficient of relative risk aversion under each treatment. Section 5 contains a discussion of the results and section 6 highlights the main findings and concludes. 2. Literature Review The economic perspective of the role of mood in decision-making under uncertainty has been conceptualized through changes in the probability weighing of different outcomes. This framework is evident in early models of anticipated emotions including regret aversion (Bell, 1982) and disappointment aversion (Gul, 1991; Loomes & Sugden, 1986). More importantly, it also seems to be the main theme in more recent work, which focused on studying the effect of mood experienced at the time of making the risky decision (Au, Chan, Wang, & Vertinsky, 2003; Bassi, Colacito, & Fulghieri, 2013; Campos-Vazquez & Cuilty, 2014; Fehr-Duda et al., 2011; Grable & Roszkowski, 2008; Kamstra et al., 2003; Kuhnen & Knutson, 2011). The general notion advanced by these studies is that positive (negative) mood drives people to become more optimistic (pessimistic), which in turn leads them to fixate more on positive (negative) outcomes and make riskier (safer) decisions. For instance, Fehr-Duda et al. (2011) conducted an experiment and found that pre-existing good mood causes individuals to adopt a more optimistic approach when weighing the