| dc.description.abstract | This thesis consists of three essays which tackle two different dimensions of public security studied in the context of the Economics of Crime: recidivism and the criminology of place. The first essay investigates the profile of the inmates who return to prison and the main determinants of prison recidivism in Santa Catarina state. The dataset consists of 22.723 inmates released in 2013-2014 from 47 institutions, with a follow-up period until November 2018. Survival/Duration analysis methods were used to estimate a non-parametric statistical model which accounts for data right censoring. The recidivism rate observed for Santa Catarina, during the follow-up period, reached 31,7%, lower than the most common national figure of 70%. Model results confirmed the significance on recidivism for gender, religion, race, civil status, education, type of offence, length of detention, number of criminal condemnations, previous escape or evasion and prison conditions. The second essay investigates the impact of inmate’s labor and education programs on recidivism over a period of nearly five years for 21.274 prisoners released from Santa Catarina prisons in 2013 and 2014. Two methods were used to control for selection bias: first, the propensity score matching (PSM) was used to find an adequate comparison group and, second, it was allowed for unobserved heterogeneity with a frailty term. Both were later introduced in separated hazard models through a Cox regression to assess the effect of prison-based education and labor programs. When restricting the participation criterium in the programs to 90-day or more, the frailty model showed a hazard ratio of 0,68 for prison study programs, and 0,86 for prison job programs, thus, indicating that program participation reduced the odds of recidivism in 32% and 14%, respectively. Therefore, this study confirmed that there are programs being executed in Santa Catarina that are effectively working for prisoner rehabilitation. Finally, the third essay assesses space variations of homicides occurrences in Belo Horizonte/MG during the period 2014-2018. Georeferenced crime data for murder and attempted murder were counted together for each census block representing interpersonal violence with fatal and non-fatal victims. Negative binomial regression models are employed to account for data overdispersion and to accommodate excessive zero counts. Negative relationships are found for income concentration index, population density and banks/betting houses per area, while positive relationships are found for the percentage of young population, percentage of illiterate population ages 18 and above, proportion of rented households, infrastructure index, downtown neighborhood, bars/restaurants per area and transport nodes per area. In general, it was found that a great majority of the crimes were concentrated at a small number of blocks. It is suggested that factors based on the theories of social disorganization, routines activity, situational prevention and crime pattern can work together in planning local strategies against crime and designing multidimensional public security policies. | |
