Just as persons, firms are more or less prone to be victimized depending on their behavior, location or attributes. For this project, we merge victimization surveys with firms census, in the context of Mexico, to predict extortion through applying cutting-edge statistical learning techniques. Our best model predicts extortion with XX percent of accuracy (XX percent points better than the baseline estimations). An extrapolation of the results was done to obtain incidence at firm level for the whole universe. The spatial correlation between our predictions and other sources of data is positive and strong. Further research describes next steps to explore the role of extortion as a structural cause of firm attrition.

1. crime_econ_completo.Rmd: has the complete code of the research

2. Draft_June_2018.Rmd: has the source code of the first draft