Accountability Traps ∗

We introduce a concept of accountability traps. Two conditions characterize an accountability trap: (i) a society is caught in a self-reinforcing pattern of behavior with low accountability and (ii) within the same set of institutions, there is another selfreinforcing pattern of behavior with greater accountability and higher voter welfare. We show that a canonical model of elections is consistent with accountability traps. In the model, the source of accountability traps is “bad expectations” about governance. Specifically, if a society finds itself in an accountability trap, it is playing an equilibrium in which the voter expects politicians to shirk and politicians expect the voter to apply low standards for reelection. The results have implications for the role of institutional reform in development. ∗For helpful comments and feedback, we are indebted to John Duggan, Rafael Hortala-Vallve, Dimitri Landa, Adam Meirowitz, David Myatt, Andrew Rehfeld, Rainer Schwabe, Dan Silverman, and seminar audiences at University of Arizona, Arizona State, Berkeley, Chicago, Maryland, Rochester, Virginia, the European Political Science Association Meetings, the Annual Meetings of the Midwest Political Science Association, the Society for Advancement of Economic Theory, and the World Congress of the Game Theory Society. This research was supported by NSF grants SES-0819152 (Bueno de Mesquita) and SES-0819200 (Friedenberg). In addition, Friedenberg thanks New York University and University College London for their hospitality. †Harris School of Public Policy Studies, University of Chicago, email: [email protected] ‡Harris School of Public Policy Studies, University of Chicago, email: [email protected]. §W.P. Carey School of Business, Arizona State University, email: [email protected] Electoral accountability is an important feature of democratic societies. The idea is that giving citizens formal accountability—i.e., the formal right to reward or punish policy makers—will lead to policy choices that are better for citizen welfare. This linkage is premised on the claim that formal accountability does, in fact, affect decisions. Two distinct empirical literatures have examined this issue. The first asks if voters actually make use of the channel of formal accountability—that is, if, in practice, they vote in a way that provides incentives (Canes-Wrone, Brady and Cogan, 2002; Wolfers, 2007; Berry and Howell, 2007; Ansolabehere and Jones, 2010). The second asks if formal accountability does, in fact, increase effective accountability—i.e., if incumbent choices are in fact shaped by the incentives voters provide (Carey et al., 2006; Besley and Case, 1995, 2003; Alt, Bueno de Mesquita and Rose, 2011; de Janvry, Finan and Sadoulet, 2012; Finan and Ferraz, 2011; Gagliarducci and Nannicini, Forthcoming). Of course, these two issues are deeply interwoven. An incumbent’s actions are shaped by how she expects voters to respond, and a voter’s electoral response to outcomes is conditioned on what actions he believes the incumbent has taken. This raises the possibility that different societies—with identical institutions—might differ in terms of how well they reap the welfare benefits of formal accountability: They may differ in the incumbent’s expectations of the voters’ behavior and the voters’ expectations of the incumbent’s behavior. As a consequence, they may differ in the actions chosen by the incumbent and so may differ in the voters’ welfare. If societies with identical institutions can differ in their level of effective accountability, it raises the possibility that there is scope for improving democratic performance and voter welfare without changing institutions. To this end, we introduce the concept of an accountability trap. A model of electoral accountability is simply a game that describes a fixed set of institutions. The model is consistent with an accountability trap if two conditions hold. First, there are (at least) two equilibria of the model that differ in terms of their level of effective accountability. Second, voter welfare is higher in the equilibrium with a higher level of effective accountability. A society is caught in an accountability trap if it is in fact playing an equilibrium with both a lower level of effective accountability and a lower level of voter welfare than some other equilibrium. We study a canonical model of the principal-agent relationship between a voter and a politician. Within the context of this model, we provide necessary and sufficient conditions for the existence of an accountability trap. We split the analysis into two steps. First, we provide a necessary and sufficient condition for the existence of multiple equilibria that differ in their level of effective accountability. Second, we provide necessary and sufficient