Measuring Social Preferences on Amazon Mechanical Turk

Social preferences are receiving increased attention in the social sciences, especially in behavioral economics and social psychology. From this arises the need to measure individuals’ social preferences in both the laboratory and in surveys of the broader population. The recently proposed SVO slider measure (Murphy et al. 2011) is supposed to be feasible for laboratory as well as for survey research. Our aim is to evaluate this measure using an online survey distributed on AmazonMechanical Turk (MTurk). We compare the elicited social preferences on MTurk to those found in laboratory settings, look at sociodemographic variation in measured social preferences and evaluate the measure’s test-retest reliability. In addition, we investigate how the standard dictator game performs as an alternative (and shorter) measure of prosocial preferences. Finally, we explore the correlation of these two incentivized measures with established survey items on self-reported prosocial behavior. Results show that social preferences elicited with the SVO-Slider on MTurk have a similar distribution to those found in laboratory settings. Also, the SVO slider turns out to have a high test-retest reliability (Pearson’s r = 0.79). However, the SVO measure correlates only weakly with self-reported prosocial behavior items but, interestingly, considerably with the survey response time.