Methodological considerations for students of Mexican legislative politics: selection bias in roll-call votes

Abstract

This paper examines the nature of the data available for studying legislative behavior in Mexico. In particular, we evaluate a potentially serious problem: only a subset of roll-call votes have been released for the critical transition period of 1998-2006. We test whether this subset is a representative sample of all votes, and thus suitable for study, or whether it is biased in a way that misleads scholarship. Our research strategy takes advantage of a partial overlap between two roll call vote reporting sources by the Chamber of Deputies: the site with partial vote disclosure, created in 1998 and still in place today; and the site with universal vote disclosure since 2006 only. An examination of the data generation and publication mechanisms, comparing different estimations of legislative behavior, reveals that omitted votes reduce the precision of estimates but do not introduce bias. Scholarship of the lower chamber can therefore proceed with data that we make public with the publication of the paper.