Competition between candidates is an important phenomenon in electoral systems where voters choose not just among different parties but among individual politicians, such as SNTV and OLPR, sometimes referred to as candidate-centered. In such systems, both parties and politicians compete for the voters’ preference and the competition between individual politicians doesn’t necessarily respect party lines: non-partisans as well as co-partisans may fight each other for votes. Competition between candidates, particularly from the same party, is frequently held responsible for the pitfalls of democratic politics on countries which adopt candidate-centered systems, such as lack of electoral accountability and proneness to particularistic policy-making.
However, current measures of competition between candidates are wanting. Usual measures include different candidates-to-seats ratios, such as the number of candidates a party fields versus the number of seats it expects to win, and algebraic transformations of the vote margins between elected and unelected candidates from the same party or list. In reality, these types of measures measure the level of competition within a unit, such as party or a list, and not the competition that actually takes place between candidates. Moreover, by ignoring the competition the actually takes place, they may provide an inaccurate measure of the level of competition.
Such measures are maximally accurate if we assume that everybody within the unit of interest actually competes with everybody else. But that may not be case. Recent studies have been pointing out that, rather than going after the whole district, candidates in OLPR may target narrower geographic areas within it. If copartisans running for the same district target different geographic areas within it they’re not actually competing with each other: they’re going after different voters. This calls forth the need for measures of competition between candidates that are sensitive to such dynamics.
In this paper we propose such a measure and apply it to the Brazilian lower house elections. Brazilian lower house elections take place under OLPR and are typically regarded as extremely candidate-centered. Our approach consists in comparing how candidates’ votes are distributed across the district territory. The intuition is that candidates whose votes are similarly distributed are target the same geographic areas and thus are actually competing with each other, while candidates whose votes are differently distributed are not. In practice, we use the horizontal cluster, an index of concentration borrowed from the regional economics literature, for the candidates’ votes in the least aggregated but still geographically meaningful unit to estimate candidates’ ideal points and check how close or distant those points are to each other.
Our paper has potentially wide-ranging implication for the literature on candidate-centered electoral systems. If competition between candidates, particularly from the same party, is actually responsible for the pitfalls of democracy in countries which adopt these systems we should strive for better, more accurate measures of the phenomenon.
