Ulric B. and Evelyn L. Bray Social Sciences Seminar

Abstract: We evaluate the effect on political selection of a program in Brazil that selects and trains new politicians. Three challenges hinder this evaluation. First, screening by the program induces selection bias. Second, outcomes such as electoral success are conditional on candidate self-selection. Third, when evaluating a training program holistically, one should not just control for selection, but also quantify its contribution. We overcome these challenges by combining rich data with a model that guides our empirical approach under the alternative assumptions of selection on observables vs unobservables. Further, we combine our theory with a decomposition exercise a la Oaxaca-Blinder, to quantify the relative contribution of screening vs treatment. Treatment increased candidacy by 17pp (from 16pp) and electoral success by 4.3pp (from 7.9pp). Treatment lowered barriers to political entry by increasing electability. But the overall effect of the program owed much to screening. Screening drove 30\% and 43\% of, respectively, increased candidacy and election rates, and shifted the pool of candidates to be more diverse, competent, and committed to democratic values. Renewing the political class faces tradeoffs: some of the traits favored by the program are not favored by voters and diminish the descriptive representation of low income people.

Joint work with Claudio Ferraz, Fred Finan, and Pedro Pessoa.