This repository contains our replication project of The Economic Costs of Organised Crime: Evidence from Southern Italy (Pinotti, 2015) for the OSE Scientific Computing class at Bonn University held during the Winter Semester 2021-2022.

The issue of organized crime represents a source of potentially adverse socio-economic repercussions across a plethora of communities worldwide, in part due to its prevalence in some form or the other in almost every country. To quantify its impact in the case of the infamous Italian mafia, we replicate the Pinotti (2015) investigation into the economic cost of organized crime in Italy. In doing so, we recreate its main results showing the effect of organized crime on GDP and, in the process, we gain insight into the complexities underlying the Synthetic Control Method (SCM).

This project begins with an overview of the topic of organized crime followed by an analysis of the paper featuring descriptive statistics, graphs, and a battery of robustness checks. We use this empirical context to introduce and motivate an implementation of the SCM, with emphasis on gradually building up the several optimization steps. This allows us to showcase the flaws of our earlier SCM optimization functions and gain some economic intuition in the process. Our implementations draw on the variety of algorithms available in Python's rich scientific libraries which can be used for our intermediate computations. Specifically, we draw on some of the state-of-the-art quadratic programming solvers including CVXOPT, OSQP, CPLEX, ECOS, and SLSQP using convenient wrappers in the form of CVXPY, qpsolvers, and scipy.optimize.

A final point of departure is the examination of a conflicting result by Becker and Klößner's (2017) response paper to Pinotti (2015). We attempt to reconcile their different findings by investigating some unexpected, recent developments in the literature on the computational challenges of SCM.

We recommend nbviewer to view the notebook as it provides a clear and seamless experience including any interactive graphs. Alternatively, the entire GitHub repository dedicated to the project can be downloaded and this notebook can be replicated locally via Jupyter Notebooks. A number of packages and libraries are used throughout the notebook, a full list with links to the respective online documentation is made available in the opening section of the notebook as well as in environment.yml.

The auxiliary folder contains .py files with code used in the main notebook, these files are organised according to the section of the notebook in which they are used. The dataset folder provides the original dataset provided by Pinotti as well as the German reunification dataset and the Basque Country dataset used in the main notebook.

The replication project rests on the following two papers:

• Pinotti, Paolo. The Economic Costs of Organised Crime: Evidence from Southern Italy. The Economic Journal 125.586 (2015): F203-F232.
• Becker, Martin, and Stefan Klößner. Estimating the economic costs of organized crime by synthetic control methods. Journal of Applied Econometrics 32.7 (2017): 1367-1369.