Essays on Public Choice and Political Competition

The efficacy of campaign funds towards electoral success has been the subject of much study among both political scientists and economists. This thesis contributes to this literature, both from a theoretical and empirical perspective, by studying the effects of advertising expenditures on electoral outcomes using data from federal congressional district races in the U.S. from 2010 to 2016. Based on empirical evidence of strategic complementarity of advertising expenditure, Chapter 1 of this dissertation estimates a model of political competition in which candidates simultaneously choose advertising expenditures from a fixed endowment of contributions in order to maximize their expected vote. I find a positive effect of candidate advertising on vote share and a smaller negative impact of third party negative advertising on Republicans and incumbent Democrats, but no corresponding impact on non incumbent Democrats. The effect of independent advertising from third party interest groups in support of candidates has no significant impact. There is unsurprisingly a strong incumbency advantage for both types Democrats and Republicans. Counter-factual simulations based on the estimated model show that efficiency gains can be made by banning third party advertising. Chapter 2 reconsiders the determinants of election outcomes while assuming the under lying data generating process follows a Dirichlet process, leading to the so called Dirichlet Regression. The assumption of a Dirichlet process is appropriate in the context of elections since vote shares, combined with abstention rates, are compositional. Results from this exer cise corroborate the results from Chapter 1 - a significant effect of advertising by candidates, but a negligible effect of third party advertising. Moreover, Dirichlet regression explains a higher proportion of the variation of election outcomes compared to estimating a normal distribution while providing comparable scores on other measures of goodness of fit. Estimates from Chapters 1 and 2 suggest that advertising expenditure by Democrats has a stronger effect than for Republicans. One way to quantify this effect is to consider the average cost of a vote. My results suggest that the cost of a vote for Democrats is approximately two thirds that for Republicans - a finding that is robust across model parameterizations and specifications. Chapter 3 extends the game theoretic model of Chapter 1, and characterizes candidate beliefs about voter preferences in terms of beta distributions. Doing so allows me to charac terize equilibria in which candidates choose different policy platforms in equilibrium. I show that when candidates have a common but not necessarily uniform prior belief about voter preferences, they will choose the same policy in equilibrium, yielding the well known mean and median voter theorems found in early models of political competition. When candidates have different priors or when there is a positive abstention rate even with a common uniform prior, my model predicts that candidates will choose different policies in equilibrium. As such, my model can rationalize the observation that candidates campaign on different policy platforms. Moreover, this result of policy divergence in equilibrium presents a contribution that has alluded the bulk of earlier theoretical work