Published online by Cambridge University Press: 05 March 2018
Criminal violence differs from other conflicts because illegal cartels primarily use violence to eliminate rivals rather than overthrow the state. However, politicians’ ability to influence cartel behavior remains unclear. This article argues that politicians alter the use of violence by setting their jurisdiction’s police enforcement levels, but that cartels can bribe politicians to look the other way. Because cartels are uncertain about politicians’ corruptibility, not every bribe is successful. Following an election, cartels must invest resources into learning politicians’ level of corruption. Cartels only increase their level of violence after successfully bribing political leaders, which implies that local violence levels should increase the longer parties remain in office. The study formalizes this argument and tests its implications using data on homicides and political tenure from Mexico. The results link incumbency to violence and suggest Mexico experiences an additional 948 homicides for each year of increased political tenure after holding an election.
Department of Political Science, University of California, San Diego and Institute for Quantitative Theory and Methods, Emory University (email: pzachary@ucsd.edu); Department of Political Science, University of Pittsburgh (email: williamspaniel@gmail.com). Previous versions of this article were presented at the Midwest Political Science Association, International Studies Association and Western Political Science Association conferences as well as seminars at UC Riverside, UCLA and Rochester. We thank Sergio Ascencio, Regina Bateson, Peter Bils, Kevin Clarke, Jesse Driscoll, Kent Eaton, James Fowler, Omar García-Ponce, Matthew Gichohi, Thad Kousser, David Lake, Benjamin Laughlin, Alexander Lee, Ben Lessing, Brad LeVeck, John Patty, Aila Matanock, Will Moore, MaryClare Roche, Phil Roeder, Arturas Rozenas, Emily Sellers, Branislav Slantchev, Brad Smith, Barbara Walter, Michael Weintraub and anonymous referees for their feedback. We are indebted to Rzan Akel, Andrea de Barros and Jacob Boeri for excellent research assistance. Zachary acknowledges research support from the National Science Foundation. All errors remain the authors’ own. Data replication sets are available in Harvard Dataverse at https://dx.doi.org/10.7910/DVN/ZNDVPW and online appendices at: https://doi.org/10.1017/S0007123417000746.