Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
NOTES 1. 2. The analyses based on the sample assumes that the survivors, i.e., those in the sample, do not differ in their fertility experience from nonsurvivors. There is much evidence that this assumption is true and that if there are any differences, they are trivial and do not affect the analyses. The life tables are constructed by an extension of a method devised by Preston and Coale (1982). In brief, the construction involves calculating the number of persons who cross each exact age x in the intercensal period, an estimate derived by interpolation between the number enumerated in each cohort in the earlier and later censuses. When the data are accurate, the method yields a life table that is an exact expression of intercensal mortality by age. A full description of the method is contained in Coale (1984). 3. A more precise estimate can be obtained for each single year of age by allowing for the fact that the proportion surviving from one age to another is not an exactly linear sequence. More appropriate, slightly nonlinear interpolation is achieved by deriving interpolation factors from a model life table at about the right level of mortality. 4. The births derived from the age-specific fertility rates are for calendar years (January 1 to December 31); the births compatible with the census age distributions are for "fiscal" years (July 1 to June 30) because the censuses were taken as of midyear. Fiscal-year births were estimated by a division of the calendar year births into first-half-year and second-half-year births. The difference between first-half-year and second-half-year births was taken 73
74 as one-fourth the difference between the births in the preceding year and in the following year. These half-year births were then recombined on a fiscal year basis. 5. It should be noted that a "stopping ruler--no more births after a boy is born, but possibly another after the birth of a girl--is well known not to increase the male/female ratio. 6. The age-specific rates for 1981-82 were obtained by assuming the continuation to the first half of 1982 of the 1981 rates. 7. It is also possible that older women understate the number of births that occurred to them a long time ago and that such understatement may contribute to the low estimates of fertility in the 1940s from this survey. However, the rates are in agreement with data about Chinese farmers, and the agreement of the numbers projected from the births in the 1950s to the 1982 census with the single-year distribution in that census is a convincing indication of no substantial understatement of the births reported as occurring nearly 30 years before the survey. 8. For cohorts that began entering marriage before 1950, the proportion ever married at each age was determined by subtracting first-marriage rates from 0.999, assigned to the age above which no more first marriages in the cohort were reported. 9. This method of life table construction is fully described in an article by Ansley Coale (1984). Constructed examples with artificially perfect data show that the life table is exact if the data used are exact.