The United Nations Population Division produces every two years revised versions of estimates and projections of population by age and sex with associated demographic indicators for 230 countries/areas.
While the increased availability of new censuses, surveys and vital statistics provides additional information, the information available for the same time period and geographic area is often inconsistent between sources and even within similar sources over time.
In this endeavor, a substantial part of country analyst's time is spent to collect and to derive basic demographic rates from as many sources as possible, to review and to evaluate them, and to reconcile differences between sources, estimation methods and data quality issues.
This presentation will use a number of case studies from Africa and Asia to highlight some of the challenges and issues in constructing robust time series of age-specific fertility and mortality rates. A range of robust statistical techniques able to deal with outliers, smoothing and non-linear trend analysis for interpolation/extrapolation with scarce and noisy data are examined and tested to derive annual demographic estimates for a selected set of countries covering a large diversity of situations in term of data quality, time series availability, geographic coverage, data collection and estimation methods.
