Comparative research on corruption has always faced challenges on how to reliably measure this phenomenon. Indicators based on perceptions of or experience with corruption are the most common approaches, but these methods have also faced criticism regarding limitations to their conceptual and measurement validity. A number of scholars have thus sought to develop alternative, more objective, measures of corruption. Following this line of research, this paper relies on audit reports from Brazilian municipalities to construct a concrete indicator of political corruption. Data collection exploits the setup of randomized multiple audit rounds to construct a unique panel of 140 municipalities covering five administrative terms between 1997 and 2013. A first empirical application of data is presented, testing the potential deterrent effect of electoral accountability on future corruption levels.