The production of mid-year population estimates is an important undertaking which
informs various stakeholders in policy formation and decision making. For instance,
national governments use mid-year estimates to allocate seats in parliament to various
constituents and public health sectors use them to monitor and improve service
delivery. Mid-year population estimates undoubtedly serve very important purposes that
affect lives of many people. As such, national statistical offices in various countries are
given the mandate to produce annual mid-year population estimates.
Statistics South Africa (Stats SA) assumes the function of producing and
publishing official mid-year estimates of the population in South Africa. Stats SA
produces its mid-year estimates using DemProj, population projection software which is
part of the SPECTRUM suite of policy models developed by the Futures Institute.
However, Stats SA does not publish full details of its adaptation of Demproj when
producing its mid-year estimates as it regards this as proprietary. Concerns have been
raised about the accuracy of the official mid-year estimates in terms of age distribution,
particularly for ages below 40 last birthday in 2011 (e.g. Dorrington 2013). As such, this
research critically analyses the method used by Stats SA to produce mid-year estimates
and assesses the feasibility of using administrative data to produce mid-year estimates
for South Africa.
The base population is adapted from the 2001 census population. Birth and death
registration data are used in a cohort component approach to produce alternative midyear
estimates for South Africa for the years 2002-2011. Prior to using these data, they
are adjusted for incompleteness of registration. Levels of completeness of birth and
death registration are estimated by extrapolating earlier estimates of completeness from
various researchers. The mid-year estimates obtained are compared with those
published by Stats SA in order to assess the relative quality of the two series of mid-year
estimates. The mid-year estimates for 2011 are also compared with the mid-year
population estimated from the 2011 census. These comparisons help identify the
mismatches to the census and their possible causes and as such, these may lead to
improved population estimates in the future, and a viable alternative method to that
currently being used by Stats SA.