Tesi doctoral - London School of Economics and Political Science. Departament of Social Policy, 2018

Inclou bibliografia general (p. [166]-180) i índex: gràfics, imatges i figures, de mapes i de taules.

The rapid fertility declines and increases in contraceptive prevalence rates in the last decades in Latin America occurred concurrently with increases in teenage and unintended childbearing. The factors behind, as well as possible future demographic consequences of this unique pattern of fertility change are still poorly understood. This thesis advances knowledge of three aspects of reproductive behaviour in Latin America: dynamics of contraceptive use in relation to an unintended birth experience, educational disparities in motherhood-timing and possible future of cohort fertility. I explore an untapped potential of the Demographic and Health Surveys (DHS) and Population Censuses for Peru, Ecuador, Colombia and Brazil, and employ demographic and statistical modelling techniques that have not been previously fully exploited in the Latin American context. First, I show how the longitudinal DHS reproductive “calendars” can be analysed using event-history models to advance the understanding of contraceptive choices of women who experience unintended pregnancies in Colombia and Peru. The study uncovers the importance of considering patterns of both pre- and after-birth contraceptive behaviour to inform the organization of postpartum family planning programmes in both countries. Second, using census data, I provide the first estimates of cohort first-birth age-specific schedules disaggregated by education level for Ecuador, Colombia and Peru. I document vastly increasing educational disparities in motherhood-timing during the fertility transition and discuss the potential factors behind this process. The analyses uncover a drastic increase in teenage fertility among women who drop-out of secondary school, indicating a need for an intervention. Lastly, using indirectly reconstructed fertility rates from censuses, cohort fertility is forecasted for total population and by education in Brazil. The study shows how a Bayesian model for fertility forecasting can be applied in the Latin American context where the childbearing pattern has been distinct from other world regions and where there is a scarcity of time-series of fertility rates. The study reveals the evolution of educational differences in completed fertility and shows that emerging low period fertility levels in Brazil might not necessarily correspond to women’s equally low lifetime fertility in the future. Overall, the substantive findings improve the understanding of the reproductive behaviour disparities in Latin America and serve as inputs for the design of policies to alleviate them. The novel use of data and application of methods are important for the development of future research agendas on fertility change and for the collection of fertility data in the region.