This section estimates the extent of underreporting of births in annual surveys, the 1995 sample census and the 2000 census long form. All these ‘surveys’ suffered ‘complete’ underreporting of births, either detected by the post-enumeration survey, or revealed in later statistical investigations. Other factors contributing to underreporting of births should also be taken into account, such as underreporting of births due to infant deaths, or the non-reporting of births due to the deaths of young mothers. Moreover, as shown in Table 4.5 in Chapter 4, in the large-scale sample census and census long form, there was data-specific underreporting, the ‘partial’ underreporting of births, primarily due to migration of young mothers or infants being misreported as in-migrants or adopted children.
It is more difficult to determine the extent of ‘complete’ underreporting of births than that of ‘partial’ underreporting. The analyses in Chapter 4 suggest that the 1982, 1990 and 2000 censuses may have underreported births by 4 percent, 8 percent and
11 percent, respectively. But caution is required because the extent of underreporting for the 1990 and 2000 censuses were estimated by using the 2000 census and education statistics, respectively. As already discussed, the latter two sources may have been to some extent inflated. Meanwhile, the analysis in Chapter 5 reveals that annual surveys may have underreported births by 6 to 9 percent, not more than 10 percent in the 1990s. While the SSB post-enumeration surveys suggested 6-7 percent underreporting in annual surveys in 1993-94, the upper limit of less than 10 percent underreporting from 1991-1996 is estimated by employing the 1998-2003 education statistics. Again, caution is required in estimating the exact extent in each year. Chapter 5 also shows that annual surveys included more, on average 1.5 percent of out-of-plan births, through misreporting second or higher-order births as first-order births from 1996 to 1999, rather than the normal practice of outright underreporting of higher-order out-of-plan births.
Although it can be safely assumed that annual surveys, including the 1995 sample census but not the 2000 census long form, have less than 10 percent ‘complete’ underreporting based on previous analyses, it is hard to determine the exact extent in each year. To reduce the arbitrariness in fertility adjustment, this study assumes two scenarios in estimating the extent of underreporting of births in selected sources from 1991 to 2000: (1) lower estimate: the results from the post-enumeration surveys in 1993-94 probably reflected the ‘true’ extent of underreporting of births in annual surveys, and (2) higher estimate: the extent of underreporting derived from the comparison with education statistics representing the average level over the period, but the relationship between each survey may have been the same. In the first scenario, it is assumed that annual surveys in 1991-92 suffered from the same extent of underreporting as the 1993 annual survey, while annual surveys from 1996 to 1999 had on average 1.5 percent lower levels than the 1993 survey. It is also assumed that the 1995 sample census and the 2000 census underreported births by 7 percent and 10 percent, respectively (Column 1 in Table 6.3). In the second scenario, it is assumed that annual surveys underreported slightly more births, but the relationship between surveys is similar as in the first scenario. Therefore, the annual surveys underreported births by 9.3 percent from 1991 to 1993, 8.9 percent in 1994, 7.8 percent from 1996 to 1999, and the 1995 sample census and 2000 census by 8 percent and 11 percent, respectively (Column 2 in Table 6.3).
Both the 1990 and 2000 censuses and the 1995 sample census were discovered to have ‘partial’ underreporting of births, as examined in Chapter 4 (see Table 4.5). However, this pattern seemed very weak in the annual survey, probably because survey interviewers devoted more effort to detecting underreporting. The extent of ‘partial’ underreporting of births calculated by the same method was around 0.2-0.3 percent during the whole period, which was roughly equal to the yearly death rate of women of reproductive age. On the other hand, all people are supposed to be enumerated in the census or sample census, and many births and younger children were probably reported but were placed in the category of non-own-births under such circumstances.
It was found that in some Asian countries, including China, newborns who died immediately after birth easily went unreported both as births and infant deaths (Coale and Banister 1994). In this case, the underreporting of infant deaths also meant the underreporting of births. In the absence of other strong evidence, this research chooses the surveillance data of infant deaths as a reference from the Ministry of Health, developed with help from the United Nation’s Children’s Fund (UNICEF) in the early 1990s (MOH 2000). Table 6.2 compares the infant death rates reported from the health systems and those from the SSB surveys. It appears that the observed infant death rates in annual surveys are slightly lower, with a difference ranging from one per thousand in 1994 to 5 per thousand in 2000, which is the estimated extent of underreporting of births due to infant deaths in this thesis.
Table 6.2 Comparison of infant death rates between the MOH and SSB sources in China, 1991- 2000 (per thousand)
Year 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 MOH 50.2 46.7 43.6 39.9 36.4 n.a. 33.1 33.2 33.3 32.2
SSB n.a. 42.3 41.6 38.8 32.3 34.0 37.4 33.1 26.7 26.3 Sources: The MOH series are from the MOH (2000); the SSB series are from the SSB (1991-2002).
In general, for annual surveys in the 1990s, and the 1995 sample census and the 2000 census long form, the extent of birth underreporting can be estimated by including the following underreporting: (1) the ‘complete’ underreporting due to concealment
of births from respondents, (2) the ‘partial’ underreporting due to migration or misreporting as in-migrants or adopted children, (3) underreporting due to infant deaths, and (4) underreporting due to the deaths of young mothers. Consequently, the estimated extent of underreporting of births in these surveys in each year is given in Table 6.3. It is very likely that the actual underreporting rates are slightly higher or somewhat lower, but this estimation is certainly a reasonable choice based on the available evidence and data.
Table 6.3 Estimated extent of underreporting of births in annual surveys, 1995 sample census and 2000 census long from in China, 1991-2000
‘Complete’
underreporting rate Other underreporting rate
Estimated underreporting rate Year Scenario 1 (percent) Scenario 2 (percent) ‘Partial’ underreporting rate (percent) Due to mothers’ deaths (percent) Due to infant deaths (percent) Scenario 1 (percent) Scenario 2 (percent) (1) (2) (3) (4) (5) (6) (7) 1991 6.9 9.3 0.3 0.4 7.6 10.0 1992 6.9 9.3 0.3 0.5 7.7 10.1 1993 6.9 9.3 0.3 0.2 7.4 9.8 1994 6.4 8.9 0.3 0.2 6.9 9.4 1995 7.0 8.0 6.27 0.1 13.4 14.4 1996 5.5 7.8 0.3 0.2 6.0 8.3 1997 5.5 7.8 0.3 0 5.8 8.1 1998 5.5 7.8 0.3 0 5.8 8.1 1999 5.5 7.8 0.3 0.7 6.5 8.8
Notes: The underreporting rate due to infant deaths in 1991 and 1996 is interpolated with reference figures in neighbouring years; the underreporting rate in 1997 was assumed as 0 due to the SSB reported infant mortality rate higher than the MOH source; the ‘partial’ underreporting rates are from Table 4.5, using reverse survival methods based on the 1995 sample census and 2000 census long form tabulations.