Trading Votes for Votes . ∗ A Decentralized Matching

We study vote trading—trading of votes for votes—through the tools of the matching literature. Decentralized vote trading is modeled via an algorithm that executes a sequence of binding, myopic, strictly improving pair-wise trades. We show that with binary alternatives, an odd number of voters, and separable preferences, the sequence always converges to a stable vote allocation—a distribution of votes such that no further pairwise improving vote trade exists. As a result, the algorithm has predictive power on the voting outcomes arising from vote trading. In special cases (with three voters and any number of issues, or with two issues and any number of voters) the outcome corresponding to a stable vote allocation is guaranteed to be the Condorcet winner, if one exists. In general, however, no such regularity holds. Allowing for coalition trades, as opposed to pair-wise trading only, rules out Pareto-dominated outcomes, but does not improve the results otherwise. When the algorithm is extended to coalition trading, stability is no longer guaranteed and voting cycles are possible. In addition, when a stable outcome exists, it need not coincide with the Condorcet winner, when the latter exists. JEL Classification : C62, C72, D70, D72, P16