Dialogue in Political Advertising: Evidence from US Political Campaigns 2012-2018

Federico Ciliberto (University of Virginia)

October 24, 2022, 14:15–15:30

Room Auditorium 4

Industrial Organization seminar


We investigate whether political candidates engage in ``dialogue'' over campaign themes, and what determines the extent of that dialogue in US Political Campaigns, 2012-2018. We define ``dialogue'' as the occasion where both candidates air political ads that discuss the same campaign theme. We characterize the level of dialogue, and how it varies across races, and across time within races. We propose a political game where candidates' choose which campaign themes to advertise. The model micro-founds the similarity index, which is what we use to quantify dialogue, as the expected probability that two candidates discuss the same theme. We show that similarity and race competitiveness are functions of the underlying strengths of the political themes. Then, we identify and explore three necessary conditions for dialogue: (i) candidates address the same theme; (ii) candidates address the same broadcast channel; (iii) candidates advertise at the same time of the day. We ask whether candidates's ads satisfy one or more of these conditions in both a race-level and race-week-level context. We also run simple Kolmogorov-Smirnov tests to compare whether the notion of ``race-level'' dialogue is empirically different from a ``race-week-level'' notion of dialogue. Finally, we investigate the determinants of dialogue, such as the time to the election, the type of election, and how close the race is. Our preliminary findings are as follows. At the aggregate level, there is much evidence of messaging overlap across the two parties, whether we look at time block, themes, or TV channel. However, we show that selection is crucial: There are many races where there is no concomitant advertising. When looking at individual races, and, specifically, over race- weeks, there is much less overlap, even after controlling for selection: substantial heterogeneity across race-weeks in terms of where and when candidates air ads, and what they discuss in those ads. The Kolmogorov-Smirnov tests clearly reject the hypothesis that dialogue over a race is distributed as over race-weeks, implying much less dialogue. Next, we show that there is statistically significant evidence that in closer races the candidates's themes overlap more, but the (economic) magnitude is really small - around 1 pct effect. Much of the difference is between Solid vs any other degree of race competitiveness, while the difference is smaller when comparing Tossup vs Leaning vs Likely. Overall, we do not find strong evidence of more dialogue in closer races. However, we do find strong evidence that closer races lead to more advertising (the selection effect). So, what we find is that closer races see more advertising but not more dialogue. These conclusions can also be drawn for Senate races (more ad, but not more dialogue); and as we get close to the election date (more ads as the date of the election nears, but not more dialogue).