Agendas, Strategic Voting, and Signaling with Incomplete Information

The literature on agendas with sincere and strategic voting represents an important contribution to our understanding of committees, of institutions, and of the opportunities to manipulate outcomes by the manipulation of institutions. That literature, though, imposes an assumption that may be unrealistic in many situations; namely, that everyone knows the preferences of everyone else. In this essay we apply Bayesian equilibrium analysis to show that the properties of agendas that others derive by assuming complete information do not hold necessarily under incomplete information. First, a Condorcet winner need not be selected, even if nearly everyone on the committee most prefers it. Second, the "two-step theorem," that any outcome reachable in n voting stages via some amendment agenda is reachable in two stages under sophisticated voting, need not hold. Third, nonbinding votes, such as straw polls, can critically affect final outcomes.