Diego Firmino Costa Da Silva, Raul da Mota Silveira Neto

Educational policies are always suggested as having potential impact on violence reduction in poor neighborhoods of cities, but there is very few researches supporting this direct link. In order to contribute to this discussion, the objective of this work is to measure the influence of the implantation of full-time high school on its neighborhood violence, measured by crime and police occurrences of different kinds (homicide, thefts, crime against property, bodily injury, and traffic-accidents), considering the case of Recife city, Brazil. With the rise of spatial econometrics, it seems prudent to consider the possibility of spatial interaction when dealing with observational units representing geographical areas. Delgado & Florax (2015) points out that, in the case where at least part of the consequence is result of spatial interactions, the stable unit treatment value (SUTVA) hypothesis is no longer valid. Considering the case in which some neighborhoods received full-time schools (treatment), this would not imply that residents of different neighborhoods could not study in such schools. In fact, there is no law in Brazil to prevent a student to enroll in a school of different neighborhood of his or her home, but it is expected that they will study in their neighborhood or in neighborhoods near the residence. In this sense, we expand the differences-in-differences (DID) model to take the possibilities of spatial interactions. Because their flexibility, we estimate the DID versions of the Spatial Durbin Model (SDM), calculating direct and indirect marginal effects, of Spatial Durbin Error Model (SDEM), and of Spatial Lag of X (SLX). The strategies allow us to verify not only the effect of full-time schools on crime of a particular neighborhood, but also on its neighbors’ criminality. Considering that incidence of indirect effect of treatment may be different for treated and untreated neighborhoods, we estimate two different indirect effects for treated neighborhood having two types of neighbors, treated and untreated. Using a data panel containing 94 contiguous neighborhoods of the Recife city between 2000 and 2010, the preliminary results indicate that full-time schools contributed to the reduction of homicides, bodily injury and crimes against property in treated neighborhoods and also, in some cases, in their neighbors. These results provide evidence on the extent to which education policies would play a role in reducing urban crime in the neighborhoods, something that is widely discussed in several spheres, but still with little empirical evidence for metropolises with high violence rates.

Diego Firmino Costa Da Silva (p), Raul da Mota Silveira Neto, José Luis Ratton

Homicides rates in Brazil are very high, even considering Latin American experience (UNICO, 2012). Although generally associated with the social economic variables, such as very high income inequality (Menezes et al., 2014), there is some consensus that drug traffic has an important role in explaining these above Latin America’s average homicide rates. Very few works, however, have deal with this relationship appropriately, and to best of our knowledge, none research have explored the relationship from a neighborhood perspective. In this research, we use a unique panel data set about neighborhoods of the City of Recife, Brazil, to analyze the influence of local drugs occurrence on homicide rates in the period 2000-2010. Besides including information about different kinds of crimes in the neighborhoods, the data set present socioeconomic information of the neighborhoods, and, thus, allows to controls not only for fixed neighborhoods, effects but also for potential variations on neighborhoods social environment. Because crime variables generally present strong spatial dependence, we consider and test for different spatial econometric models, all of them exploring the panel nature of the data (XLS, SDEM, SDM). The results indicate that, although there is a partial strong association between drugs occurrence and homicide rate in neighborhoods of the City of Recife, after controlling for space and time fixed effects and for traditions socioeconomic variables (such as income level and inequality), the association between drugs occurrence and homicide is not significant. On the hand, the evidence from our spatial panel models indicates that socioeconomic characteristics of the neighborhoods of a location, such as income level and inequality, substantial and significantly affect its homicide occurrence. Thus, our set of evidence obtained for the city of Recife, Brazil, confirms the importance of the neighborhoods’ social environment for the occurrence of urban crimes (spatial dependence), but it does not support the notion that drugs occurrence by itself is an important determinant of urban homicide rate.

Marina Toger (p), Itzhak Benenson, Sarit Weisburd

With growing urbanisation as well as upsurge in human population, increasing attention is being paid to so-called smart policing, where limited resources are deployed in an optimised fashion to achieve maximum visibility, shortest response times and, finally, minimise the crime. At the heart of smart policing, there is a need to analyse correlation between spatiotemporal patterns of the presence of police, and the occurrences of crime. While some literature showed that relocating police forces did not significantly impact crime occurrence, other researchers showed that concentrated police presence decreases probability of crime in these areas.
Recently available high-resolution spatial data on police presence and emergency calls to police enable revisiting this question using the new big-data analysis techniques. In this project, we analyse detailed GPS records of police vehicles together with crime locations time-stamped using 911 calls records in Dallas, Texas USA. We develop a statistical model of spatiotemporal dependence between the police vehicles and emergency calls and detect both spatial and temporal thresholds over which presence of law enforcement affects the probability of misconduct. The critical issue of significance of the revealed dependency is studied with the help of statistical simulation, based on the randomised datasets: Void of any interaction effect between police and crime and several kinds of possible deterrence. In this way we investigate the dependency of the frequency of crime, by the types of the incidents such as violent crimes, burglaries, public disturbances, etc.
Our approach captures the data in fine grain with exact time and location in order to detect effects that are lost in aggregation. Instead of grouping incidents by area, which cloaks differential effects in specific places within that unit, the aggregation of police-crime interactions based on recurrent sequential patterns and effects paints a nuanced picture of the complex impacts and specific possible strategies for optimisations.