Working Papers

Optimal Stopping and Reputation [Online Appendix] Submitted

Abstract: An agent decides when to take an irreversible action while learning dynamically whether it will succeed. He is reputation-driven, seeking to signal his ability to learn. In equilibrium, the agent engages in contrarian behavior: he acts for a larger range of beliefs when the  prior probability of success is lower. Reputation induces systematic distortions in decision timing: the agent acts prematurely when acting is highly disadvantageous ex ante, and with delay when it is advantageous. Distortions can dynamically reverse under moderate priors, with premature action at the beginning of the research process and excess caution later.

Competition and Herding in Breaking News [Online Appendix] Submitted

Abstract: I present a dynamic model of breaking news. Firms are rewarded for preempting their competitors and for making credible reports. Errors occur when firms fake, reporting without evidence. In equilibrium, competition and observational learning exacerbate errors and give rise to rich dynamics in reporting. Errors propagate through the market via a copycat effect, where a new report triggers a surge in faking. The copycat effect is driven by observational learning but depends on the nature of competition, as it is more prevalent in settings where there are smaller benefits to being a first reporter. This behavior implies not only the propagation of errors but also herding on the timing of news.

Homophily in Social Media and News Polarization (with Luis Abreu and Doh-Shin Jeon)

Abstract: We consider an ad-financed media firm that chooses the ideological location of its news when consumers who directly receive the news can share it with their followers on social media. When the firm maximizes the breadth of news sharing, it tends to produce polarized news if the mean (the variance) of ideological locations of the followers of a direct consumer is a convex (concave) function of the latter’s location. This implies that it is the curvature, rather than the slope, of homophily that determines news polarization so that surprisingly, larger homophily at the center (extremes) can lead to (no) polarization.

Reputation: Speed vs. Accuracy  Submitted

Abstract: A sender dynamically learns about a state via conclusive signals and decides when to make a report. He seeks a reputation for having a high learning ability. In equilibrium, the sender faces an endogenous speed-accuracy tradeoff: accurate reports are better for reputation, but reputation suffers from delay. When the sender’s learning ability is low, he increasingly prioritizes accuracy over speed as time passes, yet later reports can be less informative. Uninformed reports largely confirm the prior-likely state, and thus reports that are a priori unlikely are more informative.