Exploiting regression discontinuity designs in Brazilian, Indian, and Canadian first-past-the-post elections, we document that second-place candidates are substantially more likely than close third-place candidates to run in, and win, subsequent elections. Since both candidates lost the election and had similar electoral performance, this is the effect of being labeled the runner-up. Selection into candidacy is unlikely to explain the effect on winning subsequent elections, and we find no effect of finishing in third place versus fourth place. We develop a simple model of strategic coordination by voters that rationalizes the results and provides further predictions that are supported by the data.

Theories of multilateral bargaining and coalition formation applied to legislatures predict that parties’ seat shares determine their bargaining power. We present findings that are difficult to reconcile with this prediction, but consistent with a norm prescribing that “the most voted party should form the government”. We first present case studies from several countries and regression discontinuity design-based evidence from twenty-eight national European parliaments. We then focus on 2,898 Spanish municipal elections in which two parties tie in the number of seats. We find that the party with slightly more general election votes is substantially more likely to appoint the mayor. Since tied parties should (on average) have equal bargaining power, this identifies the effect of being labeled the most voted. This effect is comparable to that of obtaining an additional seat, and is also present when a right-wing party is the most voted and the second and third most voted parties are allied left-wing parties who can form a combined majority. A model where elections both aggregate information and discipline incumbents can rationalize our results and yields additional predictions we take to the data, such as voters punishing second most voted parties that appoint mayors.