The overall objective of this research is to investigate whether using year of death data
to produce estimates of time location is a better approach than the method developed
by Brass and Bamgboye (1981) and whether estimates of mortality produced using year
of death data are any better than those derived using the conventional orphanhood
method. In this research, year of death data from the 2008 National Income Dynamics
Study (NIDS) are used to estimate the time location of mortality rates as suggested by
Chackiel and Orellana (1985). Estimates of mortality derived from using year of death
data are then compared to estimates derived using the conventional orphanhood
methods from the 1996 and 2001 censuses, the 2007 Community Survey and the 2008
NIDS survey. Estimates of mortality rates over time derived from the ASSA 2003
model are used as the benchmark to determine which estimates best represent the level
and trend of mortality in South Africa. Year of death data from the NIDS survey
contain a high non-response rate, so to correct for that problem, deaths with unknown
year of death were distributed proportionately. Results and analysis show that
distributing deaths with unknown year of death data proportionately according to those
with known year of death results in time reference points being located more in the
recent past than those produced by using the Brass and Bamgboye (1981) method. The
analysis also shows that estimates of mortality produced from using the year of death
data are not internally consistent and for some years underestimate mortality. Since the
NIDS data contain data on survival of parent at ages 5 and 15 of respondent, the
research uses the data to impute year of death to deaths with unknown year of death.
Results show us that the estimates produced from using these data are internally
consistent and stretch for a much wider span of years.