Brown Bag Seminar Industrial Economics

Risky decisions in organizations may be distorted away from profit maximization if managers care about their reputation, i.e., public perceptions of their ability. Because reputation is hard to measure, documenting such distortions is difficult in most contexts. We provide evidence that the decision making of head coaches in the National Football League (NFL) is distorted by reputation concerns. Using sentiment analysis on text data from millions of coach-related posts on Twitter, we construct a high-frequency reputation measure and show that it is predictive of coach dismissals. We link this reputation measure to coaches' decisions whether to attempt 4th down conversions, a routinely made in-game choice involving risk. We show that coaches' decisions respond not only to expected win probabilities but also to expected reputation changes. The evidence is consistent with a model of career concerns in which coaches have private information about conversion probabilities, generating conservatism in decision making. Studying teams' staggered adoption of data analytics, we find that having an analytics division exacerbates the effect of decision outcomes on reputation and increases coaches' conservatism.

(with Paul Bose and Florian Schütt)