CAPS Seminar: Vincent Pons

Red, Blue and Grey Abstract pattern

K450, CGIS Knafel Building, 1737 Cambridge Street, Room K450, Cambridge, MA 02138

Please join us for the CAPS Seminar with Vincent Pons, Michael B. Kim Associate Professor of Business Administration, Harvard Business School, for his talk:

“Keep your Enemies Closer: Strategic Candidate Adjustments in U.S. and French Elections”

The CAPS Seminar is a monthly series highlighting research by CAPS affiliates and fostering discussion among members of our community. 

Light refreshments will be provided following the discsussion. 

Talk abstract:
A key tenet of representative democracy is that politicians should adjust their discourse and policies to the voters who elect them. The Median Voter Theorem (MVT) predicts that this will happen due to candidate incentives: if candidates are fully strategic, they will adjust to their opponent’s platform until converging to the median voter. In practice, countervailing forces may prevent this convergence from taking place. First, candidates may not be able or willing to deviate from their own intrinsic preferences, as in citizen-candidate models. Second, incentives to adjust to opponents’ platform may be low when polarization is strong and few voters are moderate. In such cases, politicians may choose to mobilize their base instead of persuading swing voters.

Despite its importance in political economy, we lack direct rigorous tests of the convergence mechanism underlying the MVT. In the U.S., there is a large consensus that Democrat and Republican politicians do not all adopt the same positions, including those running in the same constituency (Erikson and Wright, 1997). Even though politicians do not fully converge to the median voter in their electorate, this does not preclude some convergence mechanism from being at play. Furthermore, there is some evidence in the literature supporting the MVT prediction that politicians gain from moderation (Canes-Wrone et al., 2002; Hall, 2015; Handan-Nader et al., 2022), but these studies do not reveal whether politicians act on this moderation incentive or not. Researchers have found mixed evidence that changes in voter preferences lead to changes in candidates’ positions and congressional votes (Ansolabehere et al., 2001; Lee et al., 2004), while others have showed that changes in the composition of the electorate lead to changes in policies (Miller, 2008; Cascio and Washington, 2012; Fowler, 2013; Fujiwara, 2015). These changes may nevertheless be driven by the entry of new candidates rather than policy adjustments and convergence effects among the same candidates. Finally, recent studies have provided descriptive evidence that candidates moderate their discourse in the second round of two-round elections (Le Pennec, 2020; Acree et al., 2022). However, these papers do not provide any causal evidence that such moderation is driven by an attempt to move closer to the platform of competing candidates.

In this paper, we provide direct evidence that candidates adjust their discourse to their competitors. We show that they converge to each other both in ideological tone and rhetorical complexity. We derive these results by focusing on two-round elections in the U.S. (a two-party system) and France (a multiparty system). Specifically, we collect the content of websites of candidates at the elections for the U.S. House of representatives between 2002 and 2016. Our dataset contains 3,194 primary election websites and 5,392 general election websites. We also use campaign manifestos issued by French candidates to the parliamentary and local elections held between 1958 and 2022. Our dataset contains 46,779 first-round manifestos and 10,298 second-round manifestos. We use text-as-data methods to construct several measures of interest: the ideological tone of the website (in the U.S.) or of the manifesto (in France), the level of language complexity, the prevalence of different general topics (e.g., welfare and quality of life vs. economy), and an agnostic measure of text similarity between any two documents. We derive three sets of results.

First, we use these data to test whether candidates adjust their discourse to the voters they target. In two-round elections, the same candidate needs to address a broader and more diverse set of voters in the second round, as compared to the first round. If candidates adjust their discourse to their target voters, we should expect them to converge to the center between the first and second rounds. Indeed, restricting our attention to candidates who compete both in the primary (first round) and the general election (second round), we find that the distributions of ideology and complexity scores shift toward the center of their respective scale, both in the U.S. and in France. We do not see such pattern for the distribution of topics.

Second, we test whether the convergence between rounds is driven at least in part by an effort to adjust to other candidates qualified for the second round. We exploit cases in which the identity of candidates qualified for the second round is quasi-random, by focusing on elections in which candidates narrowly won their primary (in the U.S.) or narrowly qualified for the runoff (in France). Using a regression discontinuity design, we find that second-round candidates converge to the platform of their actual opponent, as compared to the platform of the runner-up who did not qualify for the last round. In the U.S., the overall convergence is mostly driven by convergence in complexity while in France it is mostly driven by convergence in ideology. There is no significant convergence in topics in either country.  

Third, we compare candidates’ level of adjustment depending on their own extremeness as well as their opponent’s extremeness. The convergence mechanism underlying the MVT should be more powerful when the opposing candidate is a real threat and when one is not already in the center. Consistent with this hypothesis, we find that candidates converge less, both in ideology and complexity, when facing an opponent who is ideologically extreme as opposed to a more moderate opponent. This is particularly true in the U.S. Conversely, candidates who are extreme themselves (either in ideology or complexity) converge more to their opponent than moderate candidates do.  

Our findings have broad implications for our understanding of electoral competition. They suggest that, even though politicians do not fully converge to the median voter, the convergence mechanism underlying this prediction is real. Candidates strategically converge to their competitors, and they do so in several complementary ways: politicians adjust both the ideological tone and the complexity of their discourse to match that of their opponent.