John Duggan's Research Site

In this section, I provide links to papers and other content added in the last year or so. Dates indicate the date of the post and sometimes differ from the date of the linked paper. 

September 8, 2026  "Frontiers of Formal Theory:  Spatial Models of Elections"  These are slides from a talk at the American Political Science Association.  The goal was to identify a number of big problems that have been left open after 50 years of research on spatial models of elections.   My thesis is that the field should provide better "workhorse" models that incorporate more realistic candidate/party objectives, multiple dimensions, multiple candidates, and multiple winners (i.e., post-election policy making), and that doing so will convey benefits for research in applied modeling.  I restrict attention to models in which, prior to voting, candidates/ parties commit to platforms in a continuum of policies.

August 6, 2026  "Multidimensional Elections" (with Avi Acharya)  We establish existence of pure strategy equilibria in a model of two-candidate competition with electoral uncertainty that is general with respect to the dimensionality of the policy space, voter preferences, and the amount of uncertainty.  Elections are typically both competitive (in the sense that each candidate wins with positive probability) and meaningful (in the sense that the candidates adopt distinct platforms).  Candidates may place positive weight on policy and office; these weights may differ across candidates; and one may have an electoral advantage over the other.  We provide conditions for the existence of an aggregate voter and a characterization of the limits of equilibria as uncertainty is removed from the model.  We demonstrate the tractability of the model in applications to distributive politics. Here are slides from a talk.

July 4, 2026 "Partisan Alignment and Extremism in Multidimensional Elections" (with Avi Acharya)  We apply our multidimensional elections framework to study the effect of candidate extremism (how far the candidates are from the aggregate voter) and alignment (how small the angle between them from the perspective of the aggregagte voter).  In the quadratic model, we explicitly solve the symmetric model in any number of dimensions and show that extremism hurts voter welfare, as does alignment of candidate preferences -- an insight that cannot be generated in one-dimensional models.  We fully characterize platforms in the non-symmetric model with low noise, and we show that the negative effects are preserved.

January 8, 2026  "Candidate Competition with Electoral Uncertainty: The Stochastic Valence Model"  This paper develops a tractable formal model as a tool for understanding the effects of valence and electoral motivations in a unidimensional model of elections between two candidates with possibly asymmetric office and policy motivations.  Electoral uncertainty arises from an unobserved valence shock.  Existence of equilibria in pure strategies is verified for general utility functions and any log concave density of the valence shock, and the effects of office motivation and valence are fully worked out in the especially tractable case of linear loss functions, i.e., Euclidean utility.  The equilibrium is solved for analytically for the special case of elections with logistically distributed shocks.  Here are slides from a talk.

August 11, 2025  "The Direction of Interest Group Influence:  Campaign Advertising, Policy Equivalence, and Voter Welfare" (with Zuheir Desai)  We examine the influence of money on elections when an interest group can contribute to informative campaign advertising, and challenger policy positions are endogenous.  We show that the direction of influence does not matter for policy outcomes:  whether the group actively offers money for policy, or passively responds to policy choices, the equilibrium policy outcomes are the same.   Voter welfare is either single-peaked or monotonically increasing in the cost of advertising.  This paper continues the project of our 2021 paper, which assumed the challenger's position is exogenously given.

August 2, 2025  "Candidate Divergence in Models of Electoral Uncertainty:  Mixed Motivations"  This paper provides conditions for divergence of equilibrium platforms in locational models of elections under the assumption of mixed motivations, i.e., candidate may place positive weight on both policy and office.  The main result is that for generic pairs of policy weights for the candidates, all equilibria feature distinct platforms for the candidates.  While this is well known for the case of pure policy motivation, from a 1985 article by Randy Calvert, the case of general, mixed motivations has been overlooked.  

June 12, 2025  "Accountability in Markovian Elections" (with Jean Guillaume Forand)  We study the ability of elections to solve the dynamic programming problem of a representative voter in a dynamic citizen-candidate framework.  We assume a finite number of states and general utilities and state transition probability, thereby allowing for non-trivial dynamics.  We show that politicians may manipulate the state to affect their electoral prospects, resulting in suboptimal payoffs for the voter; and we show that for equilibria satisfying a reciprocity condition, equilibria become approximately optimal as the voter becomes patient. The latest version exploits results on undiscounted dynamic programming to examine in detail the structure of equilibria as the players become patient (see our note for details).  This paper is published in Games and Economic Behavior: 151: 183--217.

October 15, 2024  "Relaxing Continuous Differentiability:  Some Interesting Examples"  Continuously differentiable functions have nice properties.  They are continuous; the contour set at point with non-zero gradient has a nice manifold structure; Lagrange's theorem establishes a first order condition for constrained maximization problems at solutions satisfying the constraint qualification; and around any solution at which a system of equations is non-singular, the implicit function theorem yields a unique selection of solutions, and this selection is differentiable.  This note shows that these properties do not hold when continuous differentiability is relaxed to directional differentiability.  

October 1, 2024  "A Note on Farsighted Decision Making in Stationary Dynamic Programming Problems" (with Jean Guillaume Forand)   We write up some notes on undiscounted dynamic programming.   Each optimal policy rule determines a unique ergodic distribution, and we show that as the discount factor goes to one, the limit of these ergodic distributions is optimal:  there is no other policy rule that generates an ergodic distribution with a higher expected payoff for the decision maker.