Page 10 - IJPS-2-2

Assessments of mortality at oldest-old ages by province in China's 2000 and 2010 censuses

ants belong to Uyghur and other ethnic groups for which age-reporting is not reliable. Thus, Coale
and Li concluded that death rates in China as a whole up to age 100 in the 1982 census were rela-
tively accurate if data from Xinjiang were excluded (Coale and Li, 1991: 298–300). These distor-
tions in mortality associated with Xinjiang were consistent with the reported age-exaggerations in
population counts in Xijiang noted by other researchers (e.g., Liu, 1991; Wang, 2012; Yang, 1988).
Accordingly, Huang and Poston (2000) found that Xinjiang and Tibet had much larger devianc-
es between fitted and observed death rates than other provinces; whereas provinces in eastern China
had smaller deviances.
In sum, the number of oldest-old adults has grown rapidly in China (United Nations, 2015) as
mortality rates at older ages have declined in recent decades (Gu, Gerland, Li et al., 2013; Zhou,
Wang, Zhu et al., 2016). With the exception of a few studies (Coale and Li, 1991; Duan and Shi,
2015; Huang and Poston, 2000; Wang, 2013; Zeng and Vaupel, 2003), our knowledge about un-
derreporting and age-trajectories of mortality at oldest-old ages in China remains limited. Even
less clear is the accuracy of oldest-old mortality in the two latest censuses (2000 and 2010) and vari-
ations at the provincial-level. Because most approaches in the existing literature in China suffer from
some methodological limitations, it is critical to re-examine the underestimation of mortality at old-
est-old ages using more appropriate methods. The purpose of this study is to examine the quality of
age-specific death rates at oldest-old ages by province in the 2000 and 2010 Chinese censuses
with comparisons to high-quality data from countries in the HMD. Specifically, we focused
on comparisons of conditional probabilities of dying at specific ages between Chinese and HMD
data and comparisons of age trajectories of mortality between observed and those from the Kannisto
model.

2. Data and Methods
2.1 Data Sources and Age-specific Death Rates

Data used in this study came from the 100 percent province-age-sex-specific tabulations of de facto
population at the time of the 2000 and 2010 censuses and number of deaths in the past twelve
months prior to the census (National Bureau of Statistics of China, 2002; 2012). The age-
sex-specific tabulations for the entire country and the average population in the past twelve months
were also used to estimate age-specific death rates for the entire country. The age-sex-specific death
rates by province in the two censuses were estimated by the number of deaths in the past twelve
months and the average population in the same period. The age-sex-specific average population by
province during the twelve months prior in a census was estimated from age-sex-specific popula-
tion counts at census and the number of deaths in the past twelve months based on cohort compo-
nents by assuming the even distribution of deaths and populations in neighboring age groups (with-
out consideration of migration). Because the migration flow at old ages — especially at oldest-old
ages — in the censuses was very small, the omission of migration would not introduce noticea-
ble biases. Furthermore, as the National Bureau of Statistics of China only published province-spe-
cific population and deaths in the form of five-year age groups, the death rates used in the present
study were measured by five-year age groups from ages 60–64 to ages 95—99. The reason why we
included age-specific death rates at ages 60-80 is that we aimed to establish some relationship be-
tween mortality at these ages and mortality at oldest-old ages. Mortality at childhood was not relia-
ble in the Chinese census due to severe underreporting (see Appendix A: Note 1) and adult mortality
at ages 15–59 also was not reliable due to their high domestic migration.
The sex-specific annual life tables of the 13 countries used by Thatcher and colleagues (1998)
from 1950 to the latest available year in the HMD (mostly until 2014) were also used and considered
as the standard criterion to evaluate the accuracy of the Chinese data. The HMD data are available at
its official website (www.humanmortality.org).

4 International Journal of Population Studies | 2016, Volume 2, Issue 2