Multidimensional Policy Experimentation

• In Event: Multi-Dimensionality in Policy Problems

Abstract

We study how voters’ ideological consistency across different issue areas influences policymaking. We build upon models of partisan identification (e.g., Fiorina 1978) that argue voters choose party affiliation based on their objective experiences under different partisan regimes. Such models maintain that voters consider their own wellbeing when choosing between candidates, but do not observe policy nor policy outcomes on specific issues. This simplification, although powerful, does not allow voters to draw cross-issue inferences about the effectiveness of multiple components of broader policy programs. Yet, today’s ideologically consistent, programmatic political parties encourage voters to think holistically across policy areas. We incorporate this feature of partisan politics into a game-theoretic model in which voters and politicians are uncertain about which policies are better for voters. An incumbent enacts a multidimensional policy and voters learn by observing the policy’s outcome on each issue. In equilibrium, the incumbent’s policy choice endogenously influences how much voters learn. Crucially, in our model each voter’s ideal policy configuration is correlated across issue dimensions, which provides a measure of the voter’s ideological consistency and allows each dimension’s outcome to inform voters about other dimensions. As a result, when choosing policy the incumbent must consider how potential outcomes on each dimension will influence subsequent voter beliefs and preferences--and thus their own electoral chances. Electorally disadvantaged incumbents have incentives to experiment with informative policies, which in our setting are extreme. Conversely, electorally secure politicians have incentives for moderation. Our multidimensional setup allows us to study the conditions under which the voter’s ideological consistency encourages or discourages these incentives--and the degree to which they are moderated by the relative policy informativeness across dimensions. On one hand, greater ideological consistency increases the electoral returns of using policy to shape information on every dimension, and may thus exacerbate policy distortion across all policy domains (complementarity effect). But greater consistency also implies that learning in one dimension has larger spillovers across all issues (substitution effect), which may instead induce policymakers to reduce policy distortion on all but a few components of their policy program. We identify conditions under which each effect dominates. In an extension, we investigate how strategic policymaking changes if the electorate contains single-issue voters. Finally, we relate our work to previous models of partisan identification by analyzing how our results change under the assumption that voters only observe their realized payoff, which represents a one-dimensional summary of separate policy outcomes.

Authors

• Jesse M. Crosson, Trinity University

• Federica Izzo, UCSD

• Gleason Judd, Princeton University