Experimental Studies of Culture, Diversity and Crowdsourcing

Abstract

Using experimental and behavioral economics methods, my dissertation focuses on culture, diversity and crowdsourcing. The first chapter investigates the role of culture and institutional contexts on individual behavior and institutional performance. In our experiments, subjects play two distinct games simultaneously with different opponents. Introducing entropy as an empirical measure of behavioral variation in a normal form game, we find that the strategies used in games with low entropy are more likely to be transferred to games with high entropy and are less subject to influence by other games.

The second chapter studies how Asians and Caucasians coordinate and cooperate with each other differently when their various dimensions of natural identities are salient. In an experiment conducted at two large public universities in the United States, we manipulate the salience of participants' multidimensional natural identities and find that Asians, especially first-generation students, exhibit significantly more in-group favoritism and out-group discrimination when their ethnic identity is salient. However, this significant group bias disappears when their school identity is primed.

The third and fourth chapters evaluate the performance of all-pay auctions in crowdsourcing labor markets, both in the field and laboratory. In the field experiment conducted on Taskcn, we manipulate the size of the reward and the presence of a reserve in the form of the early entry of a high-quality submission. We find that a higher reward induces significantly more submissions and attracts higher quality users. However, unpredicted by theory, we find that high-quality users are significantly less likely to enter tasks where a high-quality solution has already been submitted, resulting in lower quality of subsequent submissions in such soft reserve treatments.

To directly compare the performance of simultaneous and sequential all-pay auctions both implemented in crowdsourcing sites, we conduct a laboratory experiment and find lower revenue in sequential than in simultaneous all-pay auctions. When the entry decisions are endogenous, individuals are more likely to enter an auction early when it has a favor-early tie-breaking rule compared to a favor-late tie-breaking rule. In addition, in simultaneous all-pay auctions, our data is better rationalized by a risk-aversion model instead of the risk-neutral model.