Search results

In the postwar United States the president's party has always done worse in the midterm congressional elections than in the previous congressional election. Republican administrations exhibit below-average, and Democratic administrations above-average, economic growth in the first half of each term, whereas in the latter halves the two see equal growth. Our rational expectations model is consistent with these two regularities. In presidential elections, voters choose between two polarized candidates. They then use midterm elections to counterbalance the president's policies by strengthening the opposition in Congress. Since presidents of different parties are associated with different policies, our model predicts a (spurious) correlation between the state of the economy and elections. Our predictions contrast with those of retrospective voting models, in which voters reward the incumbent if the economy is doing well before the election. Our model performs empirically at least as well as, and often better than, alternative models.

In this study formal spatial models are applied to cross-sectional analysis of district results on the second ballot of French legislative elections. A model of probabilistic spatial voting better accounts for the data than either standard "ecological" models or a model of deterministic spatial voting. There are three substantive findings concerning voter behavior. First, the adjustment of voters to external information can be largely viewed as a shift in the spatial (Left-Right) distribution of voters. This shift, plus decisions by parties and candidates as to which districts parties will contest, determines the first ballot outcome. In arriving at second ballot choices, voters then appear to utilize decision rules that have a substantial degree of temporal stability. A second and related finding is that the second ballot can be reasonably accounted for by a single Left-Right dimension. Third, in those districts with three or more candidates on the second ballot, there may be substantial strategic voting with voters switching from candidates close to their ideal points but unlikely to win to more distant candidates who are more likely to win. The existence of strategic voting is suggested by the finding that models based solely on spatial preferences perform well for two-candidate districts, but less well for three- or four-candidate districts.

The 1951 elections for the French National Assembly permitted the formation of distinct formalized coalitions in each of 95 multimember districts. The resulting coalition outcomes are analyzed via a behavioral model that emphasizes real-time constraints that are largely ignored by formal game theory. Other major features of the model include: (1) coalitions are built incrementally and without defection until a coalition controls a majority of the votes; (2) offers to prospective partners are constrained by a graph that represents ideological and other constraints; (3) offers are made probabilistically, proportional to the votes controlled by the potential allies; (4) reciprocated offers are necessary and sufficient for a coalition to form. Estimated with data from districts with five or fewer coalescable parties or lists, the model is validated on districts with six parties or lists. The model has modest explanatory power, its major failure suggesting the influence of political forces at the national level that are beyond the scope of the district level model. Quite speculatively, these national forces may have been essential to preventing De Gaulle's coming to power in 1951. The model attests to both ideological and opportunistic facets of Fourth Republic politics.

Variations in second ballot abstention and blank and invalid ballot rates (over the cross-section of French election districts) are examined for all four legislative elections of the French Fifth Republic. Analysis was conducted primarily through a heuristic decision-making model and a spatial model developed from the theories of Riker, McKelvey, and Ordeshook, and Davis, Hinich, and Ordeshook.Abstentions appear to be primarily influenced by long-term factors and the competitiveness of the contest. Blank ballots appear to be primarily dependent upon short-term factors, especially nonvoting from the alienation that results when a candidate present on the first ballot is not present on the second. The alienation model and the heuristic model, though partly collinear, make independent contributions to the explanation of the blank ballot variance.