Online reputational loss aversion: Empirical evidence from StackOverflow.com

摘要

Do people care equally about reputational gains and losses? We study this question by analyzing users' contribution to online knowledge repositories, with data from StackOverflow.com, a website for user-curated technical questions and answers. Based on an analysis of about 1.2 million data points (22% of all relevant activity during a six-month period on this website) focusing on consistent, regular contributors, we find that users post fewer answers after receiving upvotes as well as downvotes. This is surprising and not consistent with rational behavior, because it is in the best interest of a user to post more answers after a downvote. Interestingly, users who receive downvotes post significantly fewer answers compared to users who receive upvotes. This suggests that reputational losses (downvotes) elicit a stronger reaction in users than reputational gains (upvotes), suggesting the presence of reputational loss aversion, after controlling for alternative explanations. Overall, users' reaction is about twice as strong for reputational losses as for gains. Loss aversion as indicated by relative reaction to downvotes compared to upvotes is convex in reputation – it is stronger among users in the higher reputation segments compared to the lower and middle ranges. Our paper is novel in extending the study of loss aversion to reputational settings, and studying ex post loss aversion, as compared to ex ante loss aversion in prior literature.