Disentangling the Effects of Ad Tone on Voter Turnout and Candidate Choice in Presidential Elections.

This article investigates the distinct effects of positive and negative advertising in U.S. presidential elections on voter turnout and candidate choice. Using data from the 2000 and 2004 elections, the authors develop a structural model that allows advertising tone to differentially influence both relative candidate vote shares and overall turnout, addressing empirical challenges such as endogeneity, measurement error, and highly correlated advertising variables through a novel instrumental variables approach combined with machine-learning (LASSO IV) techniques. The findings reveal that negative ads primarily increase a candidate's relative vote share by decreasing the opponent's attractiveness but slightly reduce turnout, whereas positive ads significantly stimulate turnout but have a smaller effect on relative share. Counterfactual simulations demonstrate that ad-tone strategies can influence turnout by millions and potentially swing close elections, though when both candidates optimize their ad-tone mix, electoral advantages tend to cancel out. The study underscores the importance of disentangling ad-tone effects and employing robust empirical methods to accurately assess political advertising's impact.