Counting votes right : strategic voters versus strategic parties

Abstract

This thesis, joint with F. Mezzanotti, provides a lower bound for the extent of strategic voting. Voters are strategic if they switch their vote from their favorite candidate to one of the main contenders in a tossup election. High levels of strategic voting are a concern for the representativity of democracy and the allocation efficiency of government goods and services. Recent work in economics has estimated that up to 80% of voters are strategic. We use a clean quasi experiment to highlight the shortcomings of previous identification strategies, which fail to fully account for the strategic behavior of parties. In an ideal experiment we would like to observe two identical votes with exogenous variation in the party victory probability. Among world parliamentary democracies 104 have a unique Chamber, 78 have two Chambers with different functions, and only one nation has two Chambers with the same identical functions: Italy. This allows us to observe two identical votes and therefore a valid counterfactual. In addition, the majority premia are calculated at the national level for the Congress ballot and at the regional level for the Senate ballot. This provides exogenous variation in the probability of victory. Because the two Chambers have identical functions, a sincere voter should vote for the same coalition in the two ballots. A strategic voter would instead respond to regions' specific victory probabilities. We combine this intuition with a geographical Regression Discontinuity approach, which allows us to compare voters across multiple Regional boundaries. We find much smaller estimates (5%) that we interpret as a lower bound but argue that it is a credible estimate. We also reconcile our result with the literature larger estimates (35% to 80%) showing how previous estimates could have confounded strategic parties and strategic voters due to the use of a non identical vote as counterfactual.