Blank space: demographic estimation data-sparse contexts

This thesis explores novel approaches for the estimation of demographic outcomes in contexts where data are limited. In the first part, we investigate the potential of an emerging non-traditional data source for demographic research: online genealogies. Harnessing FamiLinx, a big genealogical database with over 86 million observations, we show that the availability of accurate and non-missing demographic information in online genealogical data is selective. Our findings reveal that individuals with a non-missing value in a demographic variable are more likely to present non-missing data in the other demographic variables, and to be embedded in family networks, whose members exhibit demographic information of superior quality and completeness. In the second part, we develop a Bayesian method for estimating the total fertility rate (TFR) indirectly from defective data. By combining online genealogical data from FamiLinx populations with information from more reliable sources, the proposed method allows to obtain TFR estimates for seven European countries and the United States during the historical period 1751-1910, a time when many of these countries lacked well-functioning civil registration systems. In the third part, we build a Bayesian model for the estimation of subnational male and female TFRs. Using real data from the United States and simulated data from Australia, we demonstrate that the proposed method can produce reasonably accurate TFR estimates in contexts, such as small areas, where data tend to be sparse and highly variable. Throughout the second and third parts of this thesis, we leverage indirect estimation techniques within a flexible Bayesian modeling framework. This approach allows to incorporate multiple data sources, to capture regularities in demographic trends over time and space, and to account for uncertainty.