Two-ballot versus plurality rule: an empirical investigation on the number of candidates

Very Preliminary. Please do not cite without permission. Duverger claimed more than 50 years ago that the number of candidates in elections should be a function of electoral rules. Both his lawand hypothesissuggest the number of candidates vying for seats in elections to be tightly linked to characteristics of the elec-toral process such as its degree of proportionality and the presence of runo¤s. Here we test the validity of Duvergers claim using data from municipal elections in Brazil. Our study di¤ers from others in the \u85eld in two important dimensions. First, by using municipal data we avoid the usual problems that plague statistical analysis using cross-country data. Sec-ondly, we have a truly exogenous source of variation due to a change in electoral legislation introduced by the constitutional reform of 1988: simple plurality remained the rule only in municipalities with less than 200,000 voters, and a second-ballot became mandatory for the others above that threshold. This allows for a neat identi\u85cation strategy using panel data. Our main \u85nding is that elections with runo¤s lure greater numbers of candidates in municipalities with su ¢ ciently high levels of heterogeneity