An evaluation of the determinants and implications of panel attrition in the National Income Dynamics Survey (2008-2010)

Abstract

In the absence of established longitudinal panel surveys in South Africa, national cross-sectional household survey data are frequently used to analyse change. When these data are stacked side by side, however, inconsistencies both in time trends and between household- and person-level data are found. This study uses a new set of weights calibrated to the Actuarial Society of South Africa 2003 model projected totals using a cross-entropy estimation approach. These weights are favoured because they produce consistent demographic and geographic trends. The calculated weights are similar to the initial sample weights (and hence retain the survey design benefits) but match to a series of age-sex-race and province marginal totals that are consistent over time. The weights are publicly available for the 14-year period between 1994 and 2007.