Caroline Le Pennec (Berkeley)

“Strategic Campaign Communication : Evidence from 30,000 Candidate Manifestos”

Abstract: This paper uses a novel dataset of more than 30,000 candidate manifestos issued between 1958 and 2017 to study how politicians strategically use their campaign communication when they are constrained by party politics. Existing models predict that, under electoral pressure, candidates should compete on policy issues and converge to the policy position that is preferred by their electorate. However, in many contexts, politicians run for election in local races under a national party ticket and cannot deviate from their party platform. How do politicians run strategic campaigns when their ability to adjust their policy position is limited? To address this question, I exploit a unique institutional setting—the French legislative elections—in which individual candidates compete in two-round elections and issue their own campaign manifesto before each election round. This empirical design provides variation in the competitive environment within race, as the set of competitors changes from the first to the second round and runoff candidates have an incentive to adapt their campaign strategy accordingly. It also provides a systematic record of campaign messages issued by the same candidate at each stage of the competition. Using various methods of computational text analysis, I show that candidates who benefit from the support of fewer base voters in their district are more likely to strategically moderate their discourse in the runoff and appeal to other voters. They do so by emphasizing consensus-based non-policy issues (e.g. individual traits or local issues) as opposed to their party platform. In addition, I find that discourse moderation predicts higher politician quality and better constituency service once elected. While we tend to pay more attention to party-level competition, my results indicate that candidate-specific campaigns can give voters information that matters for representation and is hidden from national party platforms—such as information on individual politician quality.