This thesis examines demographic manifestations of son preference in three parts. Part I develops a simulation model that formalises the decision to practice prenatal sex selection through a "ready, willing and able" framework. The model is calibrated to South Korean and Indian sex ratio at birth (SRB) trajectories. Simulations reveal how SRB distortions in both countries have emerged despite declining son preference due to the rapid diusion of ultrasound combined with growing propensities to abort as a result of weakening norms for large families. Part II examines the potential role of big data to indirectly estimate the SRB at the subnational level in India. States with distorted SRBs tend to display a relatively high Google search activity for ultrasound. SRB "now-casts" generated using search volumes perform better than lagged variable models in high birth registration states. Part III examines the relationship between prenatal sex selection and postnatal excess female child mortality in two studies. The first applies lifetable techniques to decompose population changes in child sex ratios into a fertility component attributable to prenatal sex selection and a mortality component attributable to sex-differentials in postnatal survival. This study finds that although reductions in numbers of excess female deaths have accompanied increases in "missing" female births in all countries experiencing SRB distortions, excess female mortality has persisted in some but not in others. The second study uses birth histories of the Demographic and Health Surveys for six countries that have witnessed SRB distortions - India, Nepal, Pakistan, Azerbaijan, Armenia and Albania - to examine if differential mortality change by sex can be explained by the uptake of prenatal sex selection. This study finds that changes in prenatal sex selection only explain mortality change in India. Across all countries, although patterns of mortality disadvantage are concentrated amongst less educated mothers, prenatal sex selection is strongest among the better educated. Differential sorting into the two behaviours offers an explanation for why the effect for prenatal sex selection is generally weak.

La prévention situationnelle représente aujourd’hui dans de nombreux pays un secteur de recherche stratégique de la lutte contre le crime. Apparue au milieu des années 1970 au sein du laboratoire de recherche du ministère de l’intérieur britannique, cette nouvelle spécialité a pris la forme d’une ingénierie dont l’objectif est de développer des solutions techniques empêchant le passage à l'acte des délinquants, par une intervention sur les situations particulières lors desquelles des délits semblables sont commis ou pourraient l'être (cambriolage, vol de véhicule, vandalisme, etc.). Ce que l’on appelle désormais la « science du crime » se fonde sur l’assemblage d’une pluralité de savoirs pratiques, évolue entre des laboratoires de recherche et des secteurs professionnels variés (police, urbanistes, etc.), s’appuie sur des modalités d’administration de la preuve qui passent par la déduction mathématique (modélisation statistique) et intègre ses inventions théoriques dans des innovations sociotechniques (des dispositifs de prévention et de réduction des risques). Cette thèse retrace le développement de la prévention situationnelle en se déplaçant dans l’espace et le temps afin d’atteindre les lieux de sa fabrication et de rentrer dans l’intimité des controverses à travers lesquelles elle prend forme. En décrivant cette science du crime en train de se faire - des laboratoires gouvernementaux jusqu’à sa standardisation technique dans les instances de normalisation européenne, en passant par les politiques de recherche et le travail d’instrumentation - nous rendons visibles toutes les entités (théories, chercheurs, gouvernement, instruments, catégories statistiques, modèles de risque, délinquants, victimes, normes techniques, etc.) auxquelles la prévention situationnelle s’attache et se détache. Nous montrons ainsi que les liens concrets tissés entre les chercheurs et leurs différents alliés vont bien au-delà des relations entre les personnes. Ils vont jusqu’à toucher le contenu même de la prévention situationnelle. Au final, il s’agit de représenter la prévention situationnelle sous la forme d’un collectif assumant sa responsabilité politique. / In many countries today, situational crime prevention is a strategic research sector in the battle against crime. Originating within the Home Office Research Unit in the UK during the mid 1970s, this ‘new technology’ has the purpose of developing crime prevention solutions by intervening in situations where crime commonly occurs. What has now come to be called “crime science” is based on an array of practical knowledge, evolves between research laboratories and various professional sectors (police, town planning, etc.), uses evidence-based research, and implements its theoretical discoveries in socio-technical innovations (prevention and risk reduction systems). This thesis retraces the development of situational crime prevention technology to have a closer look at the controversies from which it takes its shape. By describing this crime science-in-the-making, from state laboratories and international policy transfers, from research studies and instrumentation, we reveal all the entities (researchers, government, theories, instruments, statistical classes, risk models, offenders, victims, technical standards, etc.) to which situational crime prevention has become tied, and untied. Thus, we demonstrate that concrete links weaved between researchers and their different allies go far beyond personal relationships, touching the very core of the technology. As such, situational crime prevention is constituted as a collective, political entity.