Investigating old age exaggeration in South African population and survey data using near extinct generations methods

Ronald Richman, AIG South Africa
Rob Dorrington, University of Cape Town

In order to investigate possible age exaggeration in South African census and survey data, we adapt the near extinct generations method of Das Gupta (1991) to the South African environment. We propose a smoothing mechanism based on the Gompertz curve which is capable of dealing with age exaggeration and heaping in the death data. Since the method relies on completely reported death data, we correct the death data for under-reporting using Death Distribution methods before the application of the smoothed near extinct generations method. We find that significant age exaggeration is present in the Census data of 2011 from the age of 95 and provide a confidence band around our estimates. We also find that date of birth digit preference occurs in both the population and death data, which highlights the need for smoothing of the death data prior to using them to correct population estimates.