1 Do Police Reduce Crime? A Reexamination of a Natural Experiment * John J. Donohue † Daniel E. Ho ‡ Patrick Leahy § Stanford Law School Stanford Law School Stanford Law School Abstract We reexamine a natural experiment first studied by Di Tella and Schargrodsky (2004, “DS”). In response to a 1994 terrorist attack against a Jewish Community Center in Buenos Aires, the government implemented 24-hour police surveillance on city blocks with Jewish institutions. Using a control group of blocks without Jewish institutions, DS applied difference-in-differences, finding that increased policing substantially reduced car theft. We explain how the reallocation of police resources from unprotected to protected blocks, shifts in criminal activity to avoid 24-hour police patrols, and a parking prohibition on protected blocks undermine the original design. The intervention may have displaced, rather than deterred, crime, invalidating the original control group. To investigate this possibility, we reanalyze the data with two modifications. First, we disaggregate the original control group into near and far blocks, with displacement much more likely to affect near blocks. Second, to reduce model sensitivity, we match exactly on all covariates, including neighborhood and the full pretreatment car theft time series. Consistent with displacement, we find that crime increases on near blocks relative to protected blocks, but that crime rates on protected and far blocks are indistinguishable. I. Introduction Do police reduce crime? Assessing the policing effect has vexed scholars for generations. Governments may deploy police to areas with expected high crime, seriously confounding observational inference. To break the simultaneity of crime and police, one approach has been to * We thank Andrea Chin, Alex Holtzman, Tim Shapiro, and Olga Zverovich for research assistance, Rafael Di Tella and Ernesto Schargrodsky for data, and Jennifer Arlen, Yun-chien Chang, Daniel Chen, Ming-Jen Lin, Dan Rubinfeld, two anonymous referees, and participants at the International Conference on Empirical Studies of Judicial Systems at Academia Sinica for helpful comments. † J.D., Harvard Law School; Ph.D. in Economics, Yale University. C. Wendell and Edith M. Carlsmith Professor of Law; Stanford Law School, Address: 559 Nathan Abbott Way, Stanford, CA 94305; Email: donohue@law.stanford.edu. ‡ J.D., Yale Law School; Ph.D. in Government, Harvard University. Professor of Law, Stanford Law School; Address: 559 Nathan Abbott Way, Stanford, CA 94305; Email: dho@law.stanford.edu, URL: http://dho.stanford.edu. § M.S. in Statistics, Stanford University. Research Fellow, Stanford Law School; Address: 559 Nathan Abbott Way, Stanford, CA 94305; Email: pleahy@law.stanford.edu.