Despite substantial advances in women’s educational attainment, significant gender gaps persist in labor market outcomes. One candidate mechanism is that career choices are guided by inaccurate beliefs about what potential partners value. For instance, if men (women) overestimate the opposite gender's preference for traditional gender roles, they may overinvest (underinvest) in their careers. We measure the accuracy of these beliefs in a large-scale field experiment, and employ a novel methodology to incentivize preferences about potential partners by collaborating with a Swiss dating app with over 28,000 active users. We find substantial and systematic misperceptions: individuals overestimate the importance that the opposite gender places on traditional gender roles. These gaps are large and robust across specifications, suggesting that women’s career-limiting choices may be based on inaccurate perceptions of

what men actually value, potentially perpetuating gender stereotypes and contributing to persistent labor market inequality.

We investigate the emergence of meritocratic and egalitarian narratives. Views about the fairness of social outcomes and attitudes towards redistribution are influenced by our interactions with others. We study such interactions in an online experiment by allowing participants to discuss fairness criteria with another subject before making a redistribution decision (allocation in a dictator game). Our goal is to investigate whether interacting with a partner of a similar “type” facilitates the emergence of self-serving narratives and the implementation of more selfish allocations. We also study whether subjects self-select into homogeneous communication environments that facilitate self-persuasion. Our research aims to provide evidence on the effect of social influence on the formation of self-serving beliefs.