The existing sources of demographic data for South Africa have different strengths and limitations that make them inadequate for calibration of sample weights in post-apartheid South African household surveys. The official mid-year population estimates produced by Statistics South Africa do not cover the entire period, the previously used Actuarial Society of South Africa model has become outdated, and the updated Thembisa model does not produce estimates by population groups. We introduce the Centre for Actuarial Research and Centre for Actuarial Research-Statistics South Africa models, two sources of consistent demographic data disaggregated by age groups, sex, and population groups as well as by province from 1990 to 2022. When the two model estimates are used to calibrate the South African household survey series, there are no substantial differences in several of the estimates that are evaluated. However, there are notable differences in the White and Indian/Asian population groups between the Statistics South Africa estimates and the Centre for Actuarial Research model in the later years.