Working Papers

Poster - Best CIVICA Poster 2023 🎉

Political Science version (on demand)

Abstract: I investigate how radical-right candidates affect voters' choice of whether to cast a ballot and whom to vote for, jointly with the effect on mainstream parties' decision to form electoral alliances. I use a Regression Discontinuity Design exploiting the fact that, in France, the first-round vote share discontinuously determines who advances to the second round of the election. I find that the presence of a radical-right candidate in the second round raises turnout rates. Results at the voting station level suggest that the main driver of the boost in turnout rates comes from loyal radical-right voters who only vote when their favorite candidate runs. The effects are stronger when the radical right's expected probability of winning is high and when political polarization is high. I also find that left-wing parties respond by coordinating more often on a single candidate. Regarding electoral outcomes, simply having radical-right candidates in a race switches the winning probabilities of the opponents. In three-candidate races, radical-right parties decrease vote shares of conventional right-wing parties more than vote shares of conventional left-wing parties, leading to a lower probability of winning for the conventional right. In two-candidate races, top runners whose main challengers come from the radical right, are almost certain to win, while this is not the case when top runners compete against candidates from a mainstream party.




Abstract: We study the effectiveness of allocating audits via well-defined, objective rules, rather than ad hoc and opaque bureaucratic discretion. We do so in a high-stakes context of documented political bias and where independent evaluators have called on the agency to shift from discretionary allocations to rule-based ones: audits of labor union finances in the United States. Informed by evaluations of the agency, we use machine learning tools to predict audits under each district director in order to infer which ones follow well-defined rules. We then use the turnover of directors to evaluate the performance of rules rather than discretionary allocations. We find that rule-based allocations perform worse than discretionary ones, resulting in fewer cases, fewer audit findings, and fewer convictions, and particularly driven by performance declines in the area around the district office where bureaucrats’ private information is likely best. We study the heterogeneity of rules used and investigate whether the most promising ones can be replicated by other districts. Our results suggest some caution in the movement to replace bureaucrats’ discretion with algorithmic approaches to governance.




Abstract: Using rich data on Swedish politicians, this paper documents the importance of the first experience of an election on future political participation. A Difference-in-Discontinuity design, based on the interaction between the date of birth and the timing of elections, allows us to compare the behavior of individuals who can vote and run for office in an election for the first time to individuals of similar age who participate for the second time. We find that, while turnout rates of both groups remain close at around 82\%, more demanding measures of political engagement, such as running for office and being elected, rise by 10-14% and 60-70%, respectively, with the second participation. We discuss the role of parties' screening in generating these results. We then explore the consequences of lowering the minimum age limit to 16 years old on the age profiles of politicians.



Work in Progress


Winner of the third CIVICA call