Réalisé sous la co-direction de Pierre Tremblay / En utilisant une approche économique du crime, cette thèse s’intéresse à la capacité des délinquants à minimiser leurs risques d’arrestation ou de condamnation et à l’impact de cette adaptation différentielle sur la manière dont on devrait interpréter le profil du délinquant qu’offrent les statistiques policières ou correctionnelles. Les études qui ont utilisé cette approche ont surtout insisté sur les coûts pénaux, mais omettent de prendre en considération les bénéfices que procurent les crimes aux délinquants et leur vulnérabilité différentielle aux risques, pourtant essentiels à la validation de cette approche. En considérant les expériences pénales comme un signal d’information entre les délinquants et les acteurs du système de justice criminelle, ceux-ci devront interpréter ce signal avec du bruit : l’évitement pénal. Cette distorsion du signal induit chez les policiers, les juges ou les agents correctionnels l’image trompeuse du délinquant inefficace. Notre première analyse, basée sur une méthode d’estimation des populations, évalue l’intensité de cette distorsion à travers les différentes étapes du système pénal (arrestation, condamnation, incarcération). Ce bruit ne se distribue pas de façon aléatoire, mais est influencé par les caractéristiques des délinquants. Les délinquants les moins vulnérables à la détection se révèlent plus expérimentés dans la réalisation de leurs délits : plus âgés, plus mobiles, plus spécialisés. Notre deuxième analyse s’est intéressée à l’effet de l’évitement pénal sur notre interprétation de la récidive. Les données de délinquance autorévélée d’un échantillon de détenus, mises en parallèle avec les données pénales, ont permis d’estimer la capacité individuelle d’évitement pénal. Lorsqu’on prend en considération cette disposition, les antécédents pénaux deviennent un indicateur des échecs auxquels les délinquants font face durant leur trajectoire délinquante plutôt que de la poursuite même de cette trajectoire, créant l’illusion de l’efficacité de sa prédiction par les antécédents pénaux. L’intensité des coûts pénaux antérieurs réduit la probabilité de la poursuite de la carrière. Les bénéfices soutirés des activités criminelles augmentent cette probabilité, alors que les opportunités d’emplois conventionnels la réduisent. Notre troisième analyse reprend une stratégie d’estimation des populations qui combine une variante adaptée aux déplacements urbains des délinquants et procure une perspective plus macrosociologique à nos observations des chapitres précédents parce qu’elle étudie l’évolution des populations délinquantes de différents groupes ethniques en fonction de la surveillance policière à laquelle ils font face. Nos résultats indiquent que, pour les délits dont le signalement est principalement proactif, un surplus de surveillance d’un groupe induit une baisse de surveillance auprès des autres groupes. Il en résulte que l’effet préventif obtenu dans le premier cas est annulé par un effet incitatif dans le second cas, celui-ci étant suffisant pour observer une augmentation générale de la population délinquante. Cet effet Harcourt vient corriger l’illusion de l’efficacité du profilage criminel ou des interventions policières ciblées. En conclusion, on constate qu’en se basant seulement sur des données pénales, on entretient l’illusion de l’efficacité pénale. Le bruit causé par l’évitement pénal ou l’impunité différentielle des délinquants a des répercussions importantes sur notre compréhension du comportement adaptatif des délinquants. / Using an economic approach to crime, this thesis focuses on offenders’ ability to minimize their risks of arrest or conviction and on the effect of this differential adaptation on the way we should interpret offenders’ profile based on police and correctional statistics. Studies using the approach mostly insisted on penal costs, but overlooked the benefits that offenders gain from crime and their differential vulnerability to risks, essential to its interpretation. Considering penal experiences as an information signal between offenders and criminal justice system officials, both parties would need to interpret this signal with a serious amount of noise: penal avoidance. This signal distortion induced to police officers, judge or correctional officers a misleading depiction of inefficient offenders. Our first analysis, based on a population estimates method, assess the intensity of this distortion through the different steps of the penal system (arrest, conviction, incarceration). This noise is not randomly distributed, but influenced by offenders’ characteristics. Offenders who are less prone to detection are be more experienced in the achievement of their offences: older, more geographically mobile, more specialized. Our second analysis looked at the effect of penal avoidance on our interpretation of recidivism. Self-reported offending data from an inmate sample, in conjunction with penal data, allowed us to assess individual penal avoidance ability. When taking into consideration penal avoidance, past penal experiences appear to be an indicator of failure faced by offenders during their criminal career rather than its true continuation, creating the illusion of the efficient prediction of recidivism using past penal experiences. The intensity of past penal costs reduces the probability of the continuation of the criminal career. The benefits of crime increase this probability while legitimate opportunities reduce it. Our third analysis, also using a population estimates methods combined to a variation adapted for offenders’ urban mobility, gives a macro perspective to our past observations and allowed us to examine the evolution of offenders populations from different ethnic groups in function of the intensity of police surveillance they each are subject to. Our results suggest that, for offenses with more proactive reporting, an increased surveillance of one group results in the decreased surveillance of other groups. The preventive effect on the first group is then cancelled by an incentive effect on the second. This effect alone suffices to increase the total criminal population. This Harcourt effect corrects the illusion of criminal profiling or targeted strategies efficiency. In conclusion, we observe that relying solely on penal data to assess offenders behaviours maintain an illusion of penal efficiency. The noise caused by offenders’ penal avoidance or differential impunity influences our understanding of offenders’ adaptive behaviors.

Um dos problemas que afetam as cidades brasileiras é a falta de uma metodologia adequada para a implantação de equipamentos coletivos públicos (escolas, postos de saúde, entre outros), e da utilização de modo racional desses equipamentos. Esse foi o ponto de partida para esse trabalho, cujo objetivo principal foi desenvolver uma ferramenta de análise espacial para auxiliar o poder público no planejamento e na gestão dos serviços públicos de educação e saúde, no que concerne basicamente à melhor localização das unidades básicas de atendimento e à melhor distribuição dos usuários por essas unidades, buscando reduzir os custos de transporte. Após uma avaliação da atual política de distribuição e utilização dos equipamentos coletivos de educação e saúde no Brasil e no mundo, foi realizada uma análise dos conceitos relativos às ferramentas de apoio a problemas de decisão de caráter espacial, particularmente os Sistemas de Apoio à Decisão Espacial (SADE). De posse dessa base conceitual, buscou-se formular as bases de um sistema (ou uma metodologia) que apoiasse a implantação de novos equipamentos coletivos e uma utilização eficiente dos equipamentos já existentes, tendo como parâmetro principal de comparação o custo de deslocamento dos usuários. Esse sistema deveria ser adequado à realidade das cidades médias brasileiras, de maneira geral hoje inseridas num cenário de falta de planejamento e de ausência de bases de dados estruturadas (e atualizadas). Ainda, como proposta adicional, procurou-se incorporar como ferramentas de apoio ao sistema algumas técnicas emergentes que, embora relativamente pouco utilizadas no planejamento urbano até o presente momento, apresentam grande potencial para tal. São elas: os Autômatos Celulares (ou CA, Cellular Automata) e as Redes Neurais Artificiais (RNA). Os fundamentos do sistema concebido foram transpostos para uma aplicação prática desenvolvida em um Sistema de Informações Geográficas (SIG) através de um estudo de caso conduzido na cidade de São Carlos (SP), cujos resultados demonstraram que quando se pensa em otimizar os custos de deslocamento, a principal ação a ser empreendida é a redistribuição da demanda às unidades de oferta, antes de se pensar na abertura de novas unidades. Mais importante do que os resultados numéricos obtidos, foi a confirmação de que é possível se utilizar as diversas ferramentas de planejamento e gestão de modo integrado. A partir dessa metodologia, um investimento em desenvolvimento de software pode levar à construção de um efetivo Sistema de Apoio à Decisão Espacial. De maneira mais geral, pode-se afirmar que a obtenção de dados é o grande obstáculo para pesquisas dessa natureza. A montagem de uma base de dados sólida e confiável é, sem dúvida, o ponto crucial para a execução de projetos potencialmente bem sucedidos / One of the main problems faced by brazilian cities is the lack of adequate methodologies for the implementation and rational use of public service facilities (such as schools and health care centers). That was the starting point in the definition of the main objective of the present work, which is the development of a spatial analysis tool for seeking an optimal arrangement of primary health and education facilities, in order to reduce transportation costs. The use of such a tool in the public administration is important not only for planning but also for management purposes. The study starts with an investigation of approaches that have been used in real cases, in cities of different countries, to define health and education facility types and their distribution. A careful analysis of location-allocation concepts and analysis tools for the solution of spatial problems, with particular emphasis on the Spatial Decision Support Systems (SDSS), was also conducted. The system (or methodology) proposed, which was meant to support the process of location of new facilities and also an efficient use of the existing ones while reducing transportation costs, was based on that conceptual framework. The system was directed to the context of brazilian medium-sized cities, where planning concepts are not widely applied and even the databases available are sometimes very limited and frequently outdated. The system design looked for both traditional and innovative approaches and computer-based techniques, such as Cellular Automata (CA) and Artificial Neural Networks (ANN). These techniques, although not yet extensively applied in urban planning and management, seem to be promising alternatives for those activities. The integration of the different tools and techniques was tested in a practical application carried out in the city of São Carlos (SP) and developed in a Geographic Information System (GIS) environment. The results drawn from the application are clear: the main action to reduce the transportation costs in the case studied was a re-distribution of the demand. It was considerable, despite the fact that not a single facility has been added. Another important outcome of the research was the confirmation of the hypothesis that the distinct planning and management tools tested could be used in an integrated approach. Therefore, the methodology discussed here could effectively become a Spatial Decision Support System only by means of software development for building the system interfaces. As a general conclusion, the data gathering problems still seem to be the main barrier for research development in this area. Consequently, the construction of solid and reliable databases is undoubtedly a key point to start any potentially successful project

Um dos problemas que afetam as cidades brasileiras é a falta de uma metodologia adequada para a implantação de equipamentos coletivos públicos (escolas, postos de saúde, entre outros), e da utilização de modo racional desses equipamentos. Esse foi o ponto de partida para esse trabalho, cujo objetivo principal foi desenvolver uma ferramenta de análise espacial para auxiliar o poder público no planejamento e na gestão dos serviços públicos de educação e saúde, no que concerne basicamente à melhor localização das unidades básicas de atendimento e à melhor distribuição dos usuários por essas unidades, buscando reduzir os custos de transporte. Após uma avaliação da atual política de distribuição e utilização dos equipamentos coletivos de educação e saúde no Brasil e no mundo, foi realizada uma análise dos conceitos relativos às ferramentas de apoio a problemas de decisão de caráter espacial, particularmente os Sistemas de Apoio à Decisão Espacial (SADE). De posse dessa base conceitual, buscou-se formular as bases de um sistema (ou uma metodologia) que apoiasse a implantação de novos equipamentos coletivos e uma utilização eficiente dos equipamentos já existentes, tendo como parâmetro principal de comparação o custo de deslocamento dos usuários. Esse sistema deveria ser adequado à realidade das cidades médias brasileiras, de maneira geral hoje inseridas num cenário de falta de planejamento e de ausência de bases de dados estruturadas (e atualizadas). Ainda, como proposta adicional, procurou-se incorporar como ferramentas de apoio ao sistema algumas técnicas emergentes que, embora relativamente pouco utilizadas no planejamento urbano até o presente momento, apresentam grande potencial para tal. São elas: os Autômatos Celulares (ou CA, Cellular Automata) e as Redes Neurais Artificiais (RNA). Os fundamentos do sistema concebido foram transpostos para uma aplicação prática desenvolvida em um Sistema de Informações Geográficas (SIG) através de um estudo de caso conduzido na cidade de São Carlos (SP), cujos resultados demonstraram que quando se pensa em otimizar os custos de deslocamento, a principal ação a ser empreendida é a redistribuição da demanda às unidades de oferta, antes de se pensar na abertura de novas unidades. Mais importante do que os resultados numéricos obtidos, foi a confirmação de que é possível se utilizar as diversas ferramentas de planejamento e gestão de modo integrado. A partir dessa metodologia, um investimento em desenvolvimento de software pode levar à construção de um efetivo Sistema de Apoio à Decisão Espacial. De maneira mais geral, pode-se afirmar que a obtenção de dados é o grande obstáculo para pesquisas dessa natureza. A montagem de uma base de dados sólida e confiável é, sem dúvida, o ponto crucial para a execução de projetos potencialmente bem sucedidos / One of the main problems faced by brazilian cities is the lack of adequate methodologies for the implementation and rational use of public service facilities (such as schools and health care centers). That was the starting point in the definition of the main objective of the present work, which is the development of a spatial analysis tool for seeking an optimal arrangement of primary health and education facilities, in order to reduce transportation costs. The use of such a tool in the public administration is important not only for planning but also for management purposes. The study starts with an investigation of approaches that have been used in real cases, in cities of different countries, to define health and education facility types and their distribution. A careful analysis of location-allocation concepts and analysis tools for the solution of spatial problems, with particular emphasis on the Spatial Decision Support Systems (SDSS), was also conducted. The system (or methodology) proposed, which was meant to support the process of location of new facilities and also an efficient use of the existing ones while reducing transportation costs, was based on that conceptual framework. The system was directed to the context of brazilian medium-sized cities, where planning concepts are not widely applied and even the databases available are sometimes very limited and frequently outdated. The system design looked for both traditional and innovative approaches and computer-based techniques, such as Cellular Automata (CA) and Artificial Neural Networks (ANN). These techniques, although not yet extensively applied in urban planning and management, seem to be promising alternatives for those activities. The integration of the different tools and techniques was tested in a practical application carried out in the city of São Carlos (SP) and developed in a Geographic Information System (GIS) environment. The results drawn from the application are clear: the main action to reduce the transportation costs in the case studied was a re-distribution of the demand. It was considerable, despite the fact that not a single facility has been added. Another important outcome of the research was the confirmation of the hypothesis that the distinct planning and management tools tested could be used in an integrated approach. Therefore, the methodology discussed here could effectively become a Spatial Decision Support System only by means of software development for building the system interfaces. As a general conclusion, the data gathering problems still seem to be the main barrier for research development in this area. Consequently, the construction of solid and reliable databases is undoubtedly a key point to start any potentially successful project

This paper presents methods and results of county-level population estimation using Landsat Thematic Mapper (TM) images of Denton County and Collin County in Texas. Landsat TM images acquired in March 2000 were classified into residential and non-residential classes using maximum likelihood classification and knowledge-based classification methods. Accuracy assessment results from the classified image produced using knowledge-based classification and traditional supervised classification (maximum likelihood classification) methods suggest that knowledge-based classification is more effective than traditional supervised classification methods. Furthermore, using randomly selected samples of census block groups, ordinary least squares (OLS) and geographically weighted regression (GWR) models were created for total population estimation. The overall accuracy of the models is over 96% at the county level. The results also suggest that underestimation normally occurs in block groups with high population density, whereas overestimation occurs in block groups with low population density.