Strategic Coordination Among American Voters

Cornell University, Ithaca NY

Investigators

Abstract

Recent analysis of National Election Studies (NES) data has produced surprisingly strong evidence of the existence of a large-scale noncooperative equilibrium among eligible voters (i.e., electors) in American presidential and House elections, with the equilibrium being supported by all electors having rational expectations about the election outcome (Mebane 2000a; Mebane and Sekhon 2002). Mebane and Sekhon refer to such a noncooperative equilibrium that is supported by all electors having rational expectations as "coordination." The analysis using survey data extends and refines the evidence in favor of such strategic coordination previously produced using aggregate district-level and time series data in connection with models of the American political economy (Alesina and Rosenthal 1989; 1995; Alesina, Londregan and Rosenthal 1993). The models of coordination purport to explain important political phenomena such as policy moderation, divided government and midterm loss. The emphasis on rational expectations and noncooperative equilibrium represents a substantial departure from the large related literature on ticket splitting. Mebane (2000a) and Mebane and Sekhon (2002) demonstrate that models of strategic coordination fit presidential year and midterm individual-level survey data significantly better than do nonstrategic models. Notwithstanding the results that strongly suggest the existence of strategic coordination, important questions remain concerning the validity of the models of coordination that have been tested so far. Mebane and Sekhon propose to extend the research they have already completed to explore and hopefully resolve some of these concerns. The theoretical and empirical models use common knowledge assumptions about individual voters that are unrealistic and unbelievable. In reality a coordinating voting equilibrium must depend on in- stitutions such as public opinion polls to aggregate and broadcast information. Mebane and Sekhon propose to weaken the information assumptions of the current models, especially by bringing in adaptive learning. Regarding the treatment of policy positions in current empirical models of coordination, questions arise regarding the models' assumption that all candidates of the same political party have the same positions and regarding the way the models treat the dimensionality of the policy space. Mebane and Sekhon propose to estimate models that use more flexible assumptions about policy positions. The policymaking institution assumed in Mebane (2000a) and Mebane and Sekhon (2002) ignores many features of the presidency and Congress that might affect electors' beliefs, strategies and behavior. Mebane and Sekhon propose to investigate the consequences of shifting from the current emphasis on the national share of the legislative votes (in particular House votes) each party receives to the share of seats, and in particular to the threshold of having a majority of seats. Another set of concerns center on the question: "are the effects of coordination large enough to be important?" Mebane and Sekhon (2002) supply an answer for the effect coordination may have on midterm cycles. They propose more generally to investigate the consequences of voters having equilibrium as opposed to unstructured beliefs about expected election outcomes. Mebane and Sekhon propose to address these concerns by completing two kinds of tasks: (a) develop- ing a suite of new coordinating model specifications that will be, as in Mebane (2000a) and Mebane and Sekhon (2002), at once equilibrium theoretical models and structural econometric models to be estimated using NES data; and (b) constructing simulation models that use NES data to let the simulations as closely as possible match the distributions of choices, preferences and information that exist in recent American electorates. Mebane and Sekhon propose to write software that will be freely available and easy to use, so that other researchers will be able to use the estimation technology of Mebane (2000a) and Mebane and Sekhon (2002). They also plan to make freely available the proposed simulation software.