بیش از حد هوشمند برای خودخواهی؟ اندازه گیری توانایی شناختی، ترجیحات اجتماعی و ثبات

Abstract Although there is an increasing interest in examining the relationship between cognitive ability and economic behavior, less is known about the relationship between cognitive ability and social preferences. We investigate the relationship between consequential measures of cognitive ability and measures of social preferences. We have data on a series of small-stakes dictator-type decisions, known as Social Value Orientation (SVO), in addition to choices in a larger-stakes dictator game. We also have access to the grade point averages (GPA) and SAT (formerly referred to as the Scholastic Aptitude Test) outcomes of our subjects. We find that subjects who perform better on the Math portion of the SAT are more generous in both the dictator game and the SVO measure. By contrast we find that subjects with a higher GPA are more selfish in the dictator game and more generous according to the SVO. We also find some evidence that the subjects with higher GPA and higher SAT outcomes offer more consistent responses. Our results involving GPA and social preferences complement previous work which employ measures of cognitive ability which are sensitive to the intrinsic motivation of the subject. Our results involving SAT scores are without precedent in the literature and suggest that measures of cognitive ability, which are less sensitive to the intrinsic motivation of the subject, are positively related to generosity.

1. Introduction Researchers have made improvements in understanding behavior by conceptualizing choice as originating from a brain which is heterogenous across subjects and influenced by external factors. For instance, these successes include cognitive hierarchy models (Camerer et al., 2004, Nagel, 1995 and Costa-Gomes et al., 2001), the discovery of a relationship between play in games and the working memory capacity of the subject (Devetag and Warglien, 2003), the finding that subjects apply similar strategies across fundamentally different games which are played in parallel (Bednar et al., 2012 and Savikhin and Sheremeta, forthcoming), and a relationship between strategic sophistication and access to sleep (Dickinson and McElroy, 2010).1 The benefits of this conceptualization also offer an explanation of the subject-specific heterogeneity which is often found in economics experiments: subjects differ in their cognitive ability.2 As an implication of this, researchers have sought to identify a relationship between measures of cognitive ability and economic behavior in the laboratory. Specifically, experiments have found that measures of cognitive ability are related to performance on a dynamic savings problem (Ballinger et al., 2011), learning optimal behavior in a decision problem (Palacios-Huerta, 2003), mistakes on a forecasting task (Rydval, 2011), the complexity of the strategies implemented in the repeated prisoner's dilemma game (Jones, 2011), outcomes in the repeated prisoner's dilemma game (Jones, 2008), and choice in a beauty contest game (Burnham et al., 2009).3 While these papers examine the relationship between cognitive ability and outcomes in economics experiments, less is known about the relationship between cognitive ability and social preferences. Clarifying the relationship between cognitive ability and social preferences would seem to be useful in the interpretation of these experiments. Here we hope to shed new light on the relationship by analyzing dictator-type allocations decisions and measures of cognitive ability. Our measures of cognitive ability include data on grade point averages (hereafter GPA) and the national rank on the SAT.4 Upon completion of a college course, the instructor gives the student a grade which summarizes their performance in the course. GPA is a numerical representation of the average of the grades received. The SAT is an entrance examination for admission as a freshman to universities in the United States. The SAT has a verbal portion and a math portion, where a separate score is given on both portions. We refer to the SAT and GPA measures as consequential because they can have a large effect on the subsequent life outcomes of the subject. In our experiment, subjects make a choice in a dictator game in which it is possible to keep $10. Our subjects also complete a nine item Social Value Orientation (hereafter SVO) measure for smaller monetary stakes. Each of the nine items has an individualistic response, a prosocial response, and a competitive response. The individualistic response is the one in which the material payoffs accruing to oneself are the largest. In other words, selecting the individualistic choice suggests that the subject neither positively nor negatively values the material payoffs accruing to the other subject. The prosocial response is the one in which the sum of the material payoffs accruing to both the subject and the other subject are the largest. In other words, selecting the prosocial response suggests that the subject positively values the material payoffs accruing to the other subject. The competitive response is the one in which the difference between the material payoffs accruing to the subject and the other subject are the largest. In other words, selecting the competitive choice suggests that the subject negatively values the material payoffs accruing to the other subject. By observing choice in the dictator game and choice in the SVO we have two measures of the social preferences of the subject. We consider both measures because they are incentivized differently: the choice in the dictator game involves larger stakes and the SVO involves smaller stakes. We compare our measures of cognitive ability with our measures of social preferences. We find that higher GPA subjects are more selfish in the dictator game than are lower GPA subjects. We also find that subjects who performed better on the Math portion of the SAT are more generous in the dictator game than subjects who performed worse. We do not find a relationship between the Verbal portion of the SAT and choice in the dictator game. There is also evidence of a positive relationship between generosity in the SVO and each of our three measures of cognitive ability. Each of the nine items contained in the SVO are nearly identical.5 As such, the coherence of the choices on these items offers a measure of the consistency of a subject. We find some evidence that GPA, and outcomes on both portions of the SAT are each related to the consistency of SVO choices. 1.1. Our measures of cognitive ability: SAT and GPA In order to interpret the contributions of our results, it is essential to have an understanding of the literature on the SAT and GPA measures. Although to our knowledge, there does not exist a detailed examination of the differences between the SAT and GPA measures, there does exist helpful research. Research shows that SAT outcomes are strongly related to incentivized measures of general intelligence. For instance, Frey and Detterman (2004) find a positive relationship between SAT scores and scores on the Armed Services Vocational Aptitude Battery.6 The literature also finds a close relationship between SAT scores and subsequent GPA in college.7 Despite this close relationship between SAT and GPA measures, there remain variations in this relationship which can only be explained by factors other than those related to cognitive ability. For instance, personality (Noftle and Robins, 2007 and Kappe and van der Flier, 2012), patience (Kirby et al., 2005), and self-discipline in adolescents (Duckworth and Seligman, 2005) have been found to vary with GPA. In summary, both the SAT and GPA outcomes provide a measure of cognitive ability, however GPA in particular seems to be affected by factors other than cognitive ability. Finally, we note the research on the effects of rewards in cognitive tests. Research finds that intrinsic motivation and cognitive ability are separate components to the outcomes of tests which require cognitive effort.8 In particular, Segal (2012) finds evidence that the heterogeneously distributed intrinsic motivation to perform on tests which require cognitive effort, affects their outcomes. However, intrinsic motivation is not related to self-reported SAT scores. In light of this literature, our study makes the following contributions. First, to our knowledge, we are the first paper to examine the relationship between social preferences and measures of cognitive ability as consequential as GPA and SAT outcomes. Second, we note that one of our measures of cognitive ability, GPA, similar to the commonly employed unincentivized measures of cognitive ability, is sensitive to the intrinsic motivation of the subject. Also similar to the existing literature, we find that outcomes on measures of cognitive ability, which are affected by intrinsic motivation, are associated with less generous behavior in the dictator game and more generous behavior in the SVO. We therefore view our work as complementary to the existing literature. Third, to the extent that SAT scores are relatively unaffected by the intrinsic motivation of the subjects, our results regarding the SAT outcomes appear to be without precedent in the literature. These results suggest that when the intrinsic motivation of the subject is removed from the measure of cognitive ability, higher cognitive ability subjects are more generous in both the dictator game and the SVO. In other words, it seems that the differences between our results involving GPA and those involving SAT outcomes are due to the differences in the sensitivity to the intrinsic motivation of the subjects. 1.2. Related literature: cognitive ability and economic behavior There exists a literature which examines the relationship between measures of cognitive ability and economic preferences. However, much of the literature focuses on a different set of preferences, such as time preferences or preferences toward risk. For instance, Frederick (2005) reports that subjects who perform better on an IQ-type test exhibit more patience with respect to payments over time and exhibit less risk aversion over small-stakes gambles.9 By contrast, we examine the link between social preferences and measures of cognitive ability.10 There is also a literature which examines the relationship between the consistency of answers and measures of cognitive ability. For instance, Burks et al. (2009) finds that IQ-type test results are related to the consistency of choices made on questions involving time or risk preferences. Eckel (1999) finds that the GPA of the student subjects is related to the consistency of choices made on questions involving risk preferences. We perform a similar exercise and find some evidence that our measures of cognitive ability are related to consistency. Researchers have sought to understand the relationship between different personality features and social preferences. For instance, Van Lange et al. (1997) find that age, childhood experiences, and family structure are all related to social preferences. Also, Swope et al. (2008) find a weak relationship between the personality traits of United States Naval Academy students and behavior in the dictator game, ultimatum game, trust game, and prisoner's dilemma game. To our knowledge, there are only a few other papers which examine the relationship between measures of cognitive ability and social preferences. Brandstätter and Güth (2002) report a negative relationship between giving in a dictator game and performance on cognitive tests.11Ben-Ner et al. (2004) find a negative relationship between giving in a dictator game and performance on the Wonderlic test of cognitive ability. Further, the authors find that this relationship is stronger for women than for men.12Benjamin et al. (forthcoming) find a weak relationship between cognitive ability and selfishness in the dictator game. These studies suggest that selfishness in the dictator game is increasing in their measures of cognitive ability. On the other hand, Millet and Dewitte (2007) find a positive relationship between the Raven Progressive Matrix test of cognitive ability and altruistic behavior. Their evidence comes from observations of choice in an expanded version of SVO. Whereas we closely follow the SVO format of Van Lange et al. (1997), which has three responses per item (competitive, individualistic, and prosocial), Millet and Dewitte also employ a fourth option, altruistic. The altruistic choice is distinguished from the prosocial choice in that, while both options yield identical amounts to the subject, the prosocial option sends an amount identical to that obtained by the subject, whereas the altruistic choice sends an even greater amount. In other words, the prosocial option is an even split and the altruistic option sends an even larger amount to the other subject, without reducing the subject's own allocation. The authors find evidence under rank order voting on hypothetical allocations that their measure of cognitive ability is positively related to preferences for altruism.13 However, the authors do not report such a relationship for the prosocial choices. In this paper, we find that the outcome on the Math portion of the SAT is associated with generosity on both measures of social preferences. We find that GPA is related to generosity on the SVO measure but related to selfishness in the dictator game. Finally, we find a relationship between the outcome of the Verbal portion of the SAT and generosity on the SVO measure, however we do not find a significant relationship involving the dictator game. How do our results relate to the literature examining social preferences and measures of cognitive ability? First, to the extent that GPA outcomes are affected by both cognitive ability and intrinsic motivation, as is the case for cognitive tests with low material incentives, then our results closely follow that found by Brandstätter and Güth (2002), Ben-Ner et al. (2004), and Benjamin et al. (forthcoming). Similar to these authors, we find that giving in dictator game is negatively related to such a measure of cognitive ability. Second, given reasonable assumptions about the preferences of the subjects, it would seem that the Millet and Dewitte subjects with a preference for either altruistic or prosocial preferences would be categorized as prosocial in our setting. Hence, similar to Millet and Dewitte (2007), we find a positive relationship between generosity according to the SVO measure and the outcome of a measure of cognitive ability which is relatively sensitive to intrinsic motivation. Third, to the extent that SAT scores are not significantly affected by the intrinsic motivation of the subjects, our results regarding the SAT outcomes seem to be without precedent in the literature. Our results suggest that measures of cognitive ability, which are relatively unaffected by intrinsic motivation, are positively associated with more generous behavior in both the dictator game and the SVO.

Discussion and conclusions Increasingly, researchers are interested in examining the relationship between cognitive ability and economic behavior. However, before researchers can make accurate inferences of such behavior given measures of cognitive ability, we must have a better understanding of other relevant correlates of cognitive ability. As such, in this paper we examine the relationship between consequential measures of cognitive ability and social preferences. We find that our measures of cognitive ability are related to social preferences. In particular, we find evidence of a negative relationship between performance on the Math portion of the SAT and selfishness in both the dictator game and the SVO measure. By contrast, we find a positive relationship between GPA and selfishness in the dictator game, but a negative relationship between GPA and selfishness on the SVO measure. Finally, we only find some evidence of a relationship between our measures of cognitive ability and the consistency of choices. To the extent that GPA is affected by both cognitive ability and intrinsic motivation, as evidence suggests that it is for unincentivized cognitive tests, then our results regarding GPA and generosity closely resemble that found in the literature (Brandstätter and Güth, 2002, Ben-Ner et al., 2004, Benjamin et al., forthcoming and Millet and Dewitte, 2007). In this sense, we view our results as offering a complementary view of the effects of cognitive tests which are affected by heterogenous intrinsic motivation. However, to our knowledge, our results regarding SAT outcomes are novel. We interpret our results involving SAT outcomes and social preferences as suggesting that higher measures of cognitive ability, when the measures are not significantly related to the intrinsic motivation of the subject, are associated with more generous behavior. We also note that we only find mixed evidence of a relationship between gender and social preferences. While we find that generosity in the SVO is related to the gender of the subject, we do not find such a relationship in the dictator allocations. Previous work has found a relationship between gender and social preferences,22 however our data only provides mixed evidence for this. While we are encouraged by our results, there is more to be explored. For instance, additional data is needed in order to better identify the relative merits of our measures of cognitive ability. We are also aware of the limitations of our measures of social preferences. One way to remedy this would be to conduct a thorough investigation of social preferences, ala Charness and Rabin (2002), when considering such consequential measures of cognitive ability.