Page 21 - IJPS-2-2

Danan Gu, Runlong Huang, Kirill Andreev, et al.

Our results revealed that eastern-coastal provinces had relatively lower rates of underestimation
on average and less distortion in age trajectories of mortality compared with provinces in western
China. Several reasons may explain these regional differences. First, eastern-coastal provinces are
the most developed and urbanized areas in China; and where death registration is relatively com-
plete compared to other areas in western China (Huang and Poston, 2000; Wang, Wang, Cai et al.,
2011). In the Chinese Disease Surveillance Point system, eastern coastal provinces generally had
less underestimation of death than in provinces in other parts of China (Wang, Wang, Cai et al.,
2011). Second, the population is primarily Han ethnicity in eastern China; whereas in some western
provinces ethnic minorities accounted for a relatively large share of total population. Because Han
Chinese use a lunar calendar (i.e., animal years) to remember his/her birth date and other important
life events — such as marriage — age misreporting is relatively lower than among ethnic minorities
in China (Zeng and Gu, 2008; Zeng and Vaupel, 2003). It is also possible that less steep slope of
mortality at very old ages in less developed western provinces may be due to mortality selection that
dropped those frail persons out, leaving more robust ones in the cohort, a common phenomenon at
very old ages in general populations (Zheng, 2014).
The underestimation of mortality at older and oldest-old ages was more pronounced in the
2010 census than in the 2000 census. However, the age trajectory of mortality was more accurate in
2010 census than in the 2000 census. These seemingly counterintuitive findings are not unreasonable.
First, improved age-trajectories of mortality in the 2010 census likely reflects less age-exaggeration
in the 2010 census than in the 2000 census. Indeed, this is consistent with recent improvements in
the household registration system (or hukou system) in China (Zhai, Chen and Li, 2015). Second, the
higher rates of underestimated mortality after age 70 in the 2010 census (vs. the 2000 census)
may be attributable to higher rates of underestimated mortality at ages 60 to 70 in the 2000 census.
The higher underestimation of mortality in the 2000 census for ages 60 to 70 (i.e., lower observed
mortality rates) likely produced artificially lower estimates for mortality after ages 70. In other
words, it is possible that actual rates of underestimated mortality were smaller in the 2010 census
than in the 2000 census. Third, underreports of death were more severe in the 2010 census than in
the 2000 census. It has been reported in China many families either do not report the deaths of their
older parents or relatives within the required time-period following their deaths (or “postpone” the
actual dates of death) to continue receiving pensions or other social benefits (Liu, 2011; Wang and
Liu, 2016; also see Appendix A: Note 4 for more information). Such underreports may be more se-
rious in the 2010 census because of possible increased motivations for higher pensions and more
social benefits at older ages due to improvements in social security and other social welfare systems
in both urban and rural areas in China after 2000.
Gender differences in the underestimation and age-trajectories of mortality are also noteworthy. In
almost all cases, females matched the age trajectory of the Kannisto model better than males, yet
females had higher rates of underestimation than males. Again, this seemingly counterintuitive find-
ing is not entirely unexpected. Males may have a greater underestimation of mortality at ages 60
to 70 than females, which in turn, may produce an artificially lower underestimation of mortality at
ages 70 or older in males. According to the 2009 DSP assessment survey by the China Center for
Disease Control and Prevention for in the 2006, 2007, and 2008 DSP regular surveys, males had a
higher rate of underestimation than females in all age groups except at ages 0 to 5 (Wang, Wang, Cai
et al., 2011). Therefore, it is possible that females had a lower rate of underestimated mortality at
ages 70 and older relative to males. It also may be possible that underreports of death are more prev-
alent at all ages in females than in males. More research is clearly warranted to better understand
gender differences in the underestimation of mortality at oldest-old ages in the Chinese census(es).
Results also demonstrated crossovers between the observed death rates at some ages in several
provinces in the 2000 and 2010 censuses and the observed death rates in Japanese females in the
2000s. Such crossovers are not likely due to mortality selection — in which frail people exit the co-
hort before reaching oldest-old ages, leaving more robust individuals in the cohort. Instead, we be-

International Journal of Population Studies | 2016, Volume 2, Issue 2 15