Quality and Consistency of DHS Fertility Estimates, 1990 to 2012

This report evaluates the quality of birth history data from 182 DHS surveys conducted in 69 countries since the 1990s (DHS Phase II onward). It focuses on levels and trends in estimated fertility. Fertility trends are reconstructed using birth history data from DHS surveys; the regularity of trends in fertility and their consistency across surveys draw a broad picture of the quality of data from the 69 countries. The results of the reconstruction indicate that fertility data are very good in some countries (e.g., Armenia, Colombia, Indonesia, Morocco), acceptable in many countries (e.g., Jordan, Kenya, Zimbabwe), and poor in other countries (e.g., Benin, Ethiopia, Niger, Nigeria, Pakistan). The study identified discrepancies between the fertility estimates, published in the survey reports (for the three years preceding the survey) and the reconstructed fertility estimates, and explored various data quality issues that may explain these discrepancies. Displacement of recent births in the birth history table of the DHS questionnaire because of the child health section is widespread, but it is only marginally related to discrepancies between published and reconstructed fertility trends. Three other factors may account for the differences: 1) omission of recent births, 2) the Potter effect, and 3) differences in sample composition. Overall, the analyses presented in this study indicate that DHS fertility estimates are of good or acceptable quality in the majority of surveys, but that taking published fertility figures at face value could be risky in some contexts. Inferring fertility trends by comparing recently published fertility data from successive surveys may lead to erroneous trend results.