Census data from 1991 from Botswana and Namibia were analyzed using the 'children ever born technique' to estimate child mortality. The method used is reviewed briefly, with emphasis on aspects not covered by the standard UN reference on mortality estimation, Manual X. The census data appear to be of very high quality for use with the children ever born technique. Between the mid-1970s and the early 1990s, both Botswana and Namibia experienced spectacular declines in under-five mortality (5q0) - estimates of 5 q 0 during this time span dropped from greater than 150% to below 75% for both countries. Rural areas have higher mortality rates than urban areas, but both rural and urban areas experienced similar declines. Plausible reasons for the dramatic decline in mortality are discussed, and prospects for the future are assessed.