Smoothing Period Parity Progression Ratios for Survey Data from Developing Countries

Stuart H. Sweeney, University of California, Santa Barbara

Fertility stalls are typically identified from a time series of TFR. Since stalls universally occur in developing countries, the surveys used to calculate TFR appear at irregular intervals and have complex survey designs. An alternative measure is the period parity progression ratio that can be used to identify long run patterns covering years without surveys. Standard estimation of period PPRs use each survey separately, and the overlapping estimates from different surveys will likely be misaligned and each will be biased downwards near the interview year. This paper describes and evaluates a pooled-survey estimator of period PPR based on local likelihood estimation. The estimator yields a single composite period PPR that spans the full set of survey years, and uncertainty is captured and expressed as part of the estimation process. Following the description of the statistical theory, the estimator is assessed using simulated data and applied to Guatemala.