Why would potentially intolerant majorities in a democracy protect the rights of unpopular groups or minorities? This paper postulates a dynamic agency model in which potentially tolerant legal standards emerge over time, despite all individuals' having intolerant views. Individuals in society make repeated choices which have social impact. A majority vote each period determines which of these activities are protected. Imperfect observability or interpretability of these activities necessitates that the dominant groups will not impose standards which are too intolerant, otherwise they may end up severely punishing members of their own group by mistake. We examine the Markov Perfect equilibria of a dynamic game in which there is potential turnover in the dominant group, and government improves with time in its ability to correctly observe and interpret citizens' activities. It is shown that societies with nonstationary population characteristics may be more amenable to stable and tolerant standards, while societies with stationary characteristics are more apt to choose more intolerant and unstable ones. Tolerant and stable standards tend to arise in response to a risk sharing motive between the different groups that tradeoff political power. Each group seeks to prevent auditing capabilities of government from improving too much over time in order to prevent future majorities from successfully enforcing more intolerant standards.

This paper examines the Faustian dynamics of policy and power. We posit a general class of dynamic games in which current policies affect the future distribution of political power, resulting in the following “Faustian trade-off”: if the current ruler chooses his preferred policy, he then sacrifices future political power; yet if he wants to preserve his future power, he must sacrifice his present policy objectives. The trade-off comes from the fact that the current political ruler/pivotal voter cannot uncouple the direct effect of his policy from its indirect effect on future power. A policy-endogenous (PE) equilibrium describes this endogenous transfer of power and the resulting evolution of policy and political power over time. We show that the Faustian trade-off in a PE equilibrium is decomposed into two basic rationales. The political preservation effect induces more tempered policy choices than if one's policy choice did not affect one's political fortunes. However, the reformation effect induces “more aggressive” policies in order to exploit the productivity gains from policies chosen by even more aggressive successors. We distinguish between political systems that give rise to monotone Faustian dynamics—political power that progressively evolves towards more fiscally liberal types of leaders—and cyclical Faustian dynamics—political power that oscillates between liberal and conservative types of leaders. In each case, we show that the Faustian trade-off moderates the choices of each type of leader.

We study strategic information transmission in an organization consisting of an infinite sequence of individual decision-makers. Each decision-maker chooses an action and receives an informative but imperfect signal of the once-and-for-all realization of an unobserved state. The state affects all individuals' preferences over present and future decisions. Decision-makers do not directly observe the realized signals or actions of their predecessors. Instead, they must rely on cheap-talk messages in order to accumulate information about the state. Each decision-maker is therefore both a receiver of information with respect to his decision and a sender with respect to all future decisions. We show that if preferences are not perfectly aligned, “full learning” equilibria—ones in which the individuals' posterior beliefs eventually place full weight on the true state—do not exist. This is so both in the case of private communication, in which each individual only hears the message of his immediate predecessor, and in the case of public communication, in which a decision-maker hears the message of all his predecessors. Surprisingly, in the latter case full learning may be impossible even in the limit as all members of the organization become perfectly patient. We also consider the case where all individuals have access to a mediator who can work across time periods arbitrarily far apart. In this case, full learning equilibria exist.