Rational and Naive Herding

In social-learning environments, we investigate implications of the assumption that people naively believe that each previous person’s action reflects solely that person’s private information, leading them to systematically imitate all predecessors even in the many circumstances where rational agents do not. Naive herders inadvertently over-weight early movers’ private signals by neglecting that interim herders’ actions also embed these signals. They herd with positive probability on incorrect actions across a broad array of richinformation settings where rational players never do, and–because they become fully confident even when wrong–can be harmed on average by observing others. JEL Classification: B49 Acknowledgments: This model originally appeared as part of a more general paper entitled “Inferential Naivety”, first presented in February 2006 and circulated in November 2006. On this or the earlier paper, we thank Asaf Plan and Zack Grossman for valuable research assistance and Ignacio Esponda, Guy Mayraz, Marco Ottaviani, Peter Sørensen, and Georg Weizsäcker for helpful conversations. We also thank seminar participants in 2006 at Berkeley, Birkbeck-UCL, Davis, ESSET Gerzensee, Harvard/MIT, and Lausanne; in 2007 at Bocconi, Brescia, ESEM 2007, Toulouse, and Wash U.; and in 2008 at Hebrew, Naples, Tel Aviv, UCL, UCSD and Yale. Rabin thanks the NSF for financial support, and the Trento, Italy Police Station for hospitality when some of the work on this paper was completed. Contact: [email protected]; [email protected]