WORKING PAPERS
Criminogenic Effects of Mass Forced Displacement
This study seeks to identify the causal effect of Syrian refugee influx on crime in Turkey for the 2009-2016 period. Using provincial data on new criminal cases handled by the chief prosecutor’s office and new criminal charges in assize courts and the penal court of first instance, an instrumental variables Poisson regression is employed to take the endogenous nature and the intensity of refugee settlement and the count nature of crime into account. Results indicate that the expected number of new criminal cases handled by the chief prosecutor’s office increases by 31 to 58 percent as a consequence of mass displacement. However, several falsification exercises assert that our confidence in these causal effects is severely undermined, suggesting that the true effect is null. Consistently, Syrian refugee influx has no effect on new criminal charges in assize courts and evidence is very brittle for those in penal courts. The study also highlights the drawback of relying on incorrect standard errors in the light of recent advances
on valid inference for IV and shows that some prior evidence is weakened or washed away.
​
JEL Classification: F22, J68, K42, C23, C26
Keywords: instrumental variables Poisson, crime, refugees, Syrians, Turkey
Causal Identification in Social Sciences Using Observational Data
This paper addresses identification challenges in social sciences using observational data, through illustrative and tailored examples from the economics of crime and criminology. It highlights the benefits of addressing the causal problem by incorporating structural causal models and causal graphs, rather than relying solely on the statistical problem, to avoid pitfalls such as mediation fallacy, collider bias, and the use of inadmissible instruments or covariate sets. Through case studies, the paper emphasizes the necessity to obtain the full causal model to draw correct inferences. The ultimate goal is to introduce researchers into the fundamental principles and methodologies that underpin robust causal analysis, emphasizing the distinction between identification and estimation and equip them with several tools for rigorous causal inference in their empirical work.
​
JEL Classification: A12; C18; C51; K14
Keywords: structural causal model, causal graphs, crime research, sentencing, Pearl's causal framework
Political Conservatism and Gun Ownership in the United States: Evidence from a Regression Discontinuity Design
Employing a regression discontinuity (RD) design, this study reveals that the direction of the association between political conservatism and gun ownership does not reflect the direction of the causal effect. Gun ownership levels are 2.9 percentage points lower on average, in counties where Republican presidential candidates barely won, compared to where they barely lost. The proposed mechanism is that an anticipation of an outcome that will favor an administration less friendly to gun purchasing leads to those in contested areas to buy extra guns. The falsification tests indicate that the assumptions of RD design hold and that the estimate of RD treatment effect has a causal interpretation. The results may have severe consequences in the identification of the direction of causal effect of gun ownership levels on violent crime rates.
​
Keywords: Gun ownership; discontinuity; US presidential elections; local average treatment effect; Simpson's paradox