Demographic trends in China, 1953-1982

pHS Techrical Note 85-4 DEMOGRAMC TRENDS IN CHINA, 1953-1982 by Kenneth Hill July 1985 Population, Health and Nutrition Department World Bank The World Bank does not accept responsibility for the views expressed herein which are those of the author(s) and should not be attributed to the World Bank or to itc, affiliated organizations. The findings, interpre:ations, and conclusions åxe the results of research supported by the Bank; they do no necessarily represent official policy of the Bank. The designations employed, the presentation of material, and any maps used in this document are solely for the convenience of the reader and do not imply the expression of any opinion whatsoever on the part of the World Bank or its affiliatel concerning the legal status of any country, territory, city, area, or of its authorities, or concerning the delimitations of its boundaries, or national affiliation. PRN Technical Note 85-4 DEMOGRAPBIC TENDS IN CHIA, 1953-1982 AB S TRACT Extensive data, both recent and historical, on the population of China have recently become available. The paper describes the application of numerous consistency checks and adjustment procedures to these data, and draws conclusions about levels and treads of mortality, fertility and population size over the past three decades on the basis of the results obtained. Mortality fell quite rapidly from the early 1950s to the late 1970s, with the infant mortality rate falling from around 200 to below 40 per thousand live births, and expectation of life at birth increasing from around 40 years to nearly 68 years over the period. Fertility was apparently quite high and roughly constant, with a birth rate in the range of 40 to 45 per thousand population and a total fertility rate between 6.0 and 6.5, from the early 1950s to the mid 1960s, when a sharp decline set in such that by 1979 the birth rate was around 18 and the total fertility rate around 2.3. These fertility and mortality trends were interrupted by a major demographic crisis around 1960 during which fertility was halved and infant mortality was almost doubled; this crisis, during which some 30 million people died prematurely, was associated with a period, of severe food shortage and famine throughout most of the country. In the very recent past, both infant mortality and fertility appear to have increased somewhat, and the available demographic data suggest that China has by no means yet achieved its one child family policy, though fertility has declined greatly over the last two decades. Prepared by; Kenneth Hill, Consultant to the World Bank Population, Health and Nutrition Department July 1985 Demographic Trends in China, 1953-1982 Kenneth Hill Introduction For many years, very little hard information was available about the population of the People's Republic of China. Inferences and speculations about population trends were based on press reports, eyewitness accounts and occasional snippets of official information apparently based on a population registration system. This unsatisfactory situation has changed completely in the last four years, reflecting both a general switch to much greater openness on the part of the Chinese authorities and the successful execution of the 1982 population census. A flood of demographic data has poured out of China, including single-year age distributions from the 1953 and 1964 censuses, results from a 10 percent sample of households from the 1982 census, fertility rates for the period 1940-82 from a one-in-a-thousand fertility survey carried out in late 1982, and registration figures for end-year population, birth rate and death rate from 1950 to 1982. These data not only provide a basis for assessing current demographic trends in the PRC, but also for constructing demographic estimates for the entire period back to the establishment of the People's Republic. Not surprisingly, the sudden availability of extensive data concerning the world's most populous country has generated great interest in the demographic community, and several reviews of China's demographic history -2- have recently appeared.-' All these reviews stress the astonishing consistency of the information from the diverse sources. This consistency is remarkable for a country of China's level of social and economic development, but it is not perfect; the sources agree closely but not exactly. Information from the pooulation register is the most difficult to fit into the general picture; the figures on population size, birth rate and death rate for 1950-82 given in the 1983 Statistical Yearbook and reproduced in Table 1 not only fail to match closely comparable 'indicators from other sources but also appear to be internally inconsistent. The population sequence from 1956 to 1959 grows faster than the reported rate of natural increase (only possible with substantial net immigration), while from 1960 to 1964 the population growth is below the natural increase rate (oaly possible with net emigration); similar inconsistencies continue into the 1970's. Therefore, although general demographic trends over the last 30 years can now be established beyond any reasonable doubt, the exact course of short-term deviations from these trends remains ambiguous. It is the purpose of this review to describe both the general trends and short-term deviations more precisely. Summary This review uses information from the population censuses of 1953, 1964, and 1982, the 1982 one-in-a-thousand fertility survey, registered birth and death rates by calendar year, and the 1973-75 Cancer Epidemiology Survey. Judith Banister, "An Analysis of Recent Data on the Pop4ation of China," Paiulation and Develooment Review, vol. 10, June 1984, pp. 241-271; William P. Brass, Mortality in China over the past fifty years: indirect estimates from the 1982 Census. Paper presented at the international seminar on the 1982 Population Census, Beijing, March 1984; Ansley Coale, RaDid Pooulation Change in China, 1952-1982, Report no. 27, National Academy of Sciences Committee on Population and Demography, National Academy Press, Washington, D.C , 1984. TABIZ 1. China - Official Estimates of Total Populacion an:t Vital Rates, 1949-8 Year En: Apparen Mid-year Birth Death Natural ?cpulatin kIa Grcbth Cnsus Total Rate Rae It.ease Year , (millien) race (%) (illion) (per '000) (per OC0) (per 'OCO) 1949 541.7 - - 36.0 20.0 16.0 1950 552.0 1.9 - 37.0 18.0 19.0 1951 563.0 2.0 - 37.8 17.8 20.0 1932 574.8 2.1 - 37.0 17.0 20.0 1953 588.0 2.3 580.6 37.0 14.0 23.0 1954 60Z2.7 2.5 - 38.0 13.2 24.8 1955 614.7 2.0 - 32.6 12.3 20.3 1956 628.3 2.2 - 31.9 11.4 20.5 1957 646.5 2.9 - 34.0 10.8 23.2 1958 659.9 2.1 - 29.2 12.0 17.2 1959 672.1 1.8 - 24.8 14.6 10.2 1960 662.1 . -1.5 20.9 25.4 -4.6 1961 658.6 -0.5 - 18.0 14.2 3.8 1962 672.9 2.2 - 37.0 100 .27.0 1963 691.7 2.8 - 43.4 10.0 33.3 1964 705.0 1.9 694.6 39.1 11.5 27.7 1965 725.4 2.9 - 37.9 9.5 28.4 1966 745.4 2.8 - 35.0 8.8 26.2 1967 763.7 2.4 - 34.0 8.4 25.5 1968 785.3 2.8 - 35.6 8.2 27.4 1969 806.7 2.7 - 34.1 8.0 26.1 1970 829.9 2.9 - 33.4 7.6 25.3 1971 852.3 2.7 - 30.6 7.3 23.3 1972 871.8 2.3 - 29.8 7.6 22.2 1973 892.1 2.3 - 27.9 7.0 20.9 1974 I8.6 1.8 - 24.8 7.3 17.5 1975 924.2 1.7 - 23.0 7.3 15.7 1976 937.2 1.4 - 19.9 7.2 12.7 1977 949.7 .1.3 - 18.9 6.9 12.1 1978 962.6 1.4 - 18.2 6.2 12.0 1979 975.4 1.3 - 17.8 6.2 11.6 1980 987.1 1.2 - n.a. n.a. n.a. 1981 1000.7 L.4 - 20.9 6.4 14.5 1982 1015. å.5 1008.2 2141 6.6 14.5 Source: Statistical Yea-bock, 1983. -4- The analytical strategy followed is to convert the available indices of fertility and mortality into readily comparable forms, and then, in light of the errors likely to affect the different sources, to select best estimates from the available measures for each year. A consistent series of demographic indicators is then constructed from the best estimates for each year. ' The analysis is divided into two periods, 1953-64 and 1964-82, the population censuses marking the beginning and end of each period. The censuses are used to divide the periods because they represent crucial elements of the consistency evaluation. Table 2 shows final estimates of demographic indicators for the period 1953-64, and Table 3 completes the series for the period 1964-82. Figure 1 plots the birth rate, the death rate, dnd the infant mortality rate by year for the entire period. The principal features to be noted are as follows. Fertility remained approximately constant at a moderately high level until the late 1960's, apart from a very marked but short-lived reduction around 1960, and then fell sharply through the 1970's to a level less than half that of the 1960's, before turning upwards again slightly in the early 1980's. This upturn appears to have occurred primarily as a result of increased numbers of first births, with little increase in second or higher order births. Mortality fell from the early 1950's to the late L970's, though the decline was interrupted by a sharp mortality surge in the period 1958-61. Infant mortality fell from close to 200 per 1,000 live births to under 100 by the mid-1960's, to the upper 30's by 1978, though it rose to over 200 in the mortality surge around 1960, and increased slightly from late-1970's levels in the early 1980's. Expectation of life at birth increased from little more than 40 years in the early 1950's to around 55 years by the mid-1960's, and nearly 68 years by 1981. Both the pace of fertility decline from 1970 and the pace of mortality decline from the BLE 2. A Reconstruction of Detographic aialges in (hina: Selected Indicators, 1953-64 ?opulatim. (beginnir of Infant Tocal Growth Fiscal Eiscal year) Births Deaths Birzh Death Mortalicy -ertilicy Race Year (millions) (millions) (Millions) Race Race Race Race (rercEnc) 1953-54 582.6 26.08 13.99 44.3 23.8 186 6.39 2.05 1954-55 594.7 27.31 13.69 45.4 22.8 177 6.62 2.26 1955-56 608.3 24.82 12.86 40.4 20.9 167 5.96 1,95 1956-57 620.3 25.54 12.10 40.7 19.3 156 6.08 2.14 1957-58 633.7 25,80 12.24 40.3 19.1 144 6.07 2.12 1958-59 647.3 2,.11 13.87 .4.0 24.4 226 5.17 0o96 L959-60 653.5 18.88 21,37 28.9 32.8 205 4.38 -0.39 1960-61 651.0 14.72 19.17 22.7 29.5 132 3.39 -3.68 1961-62 646.6 20.88 13.79 32.1 21.2 236 4.71 1.09 1962-63 653.7 33.81 10.07 50.8 15.1 88 7.55 3.57 1963-64 677.4 30.71 11.73 44.7 17.1 96 6.82 2.76 1964-65 696.4 TJEt 3. A Recsrcrticn of Deoaphic Oanes in China: Selected Idcacars, 1964-8 Fiscal Year Mid-Year =irths Deachs Birth Desca TR Growh Year Populacicn Rate Race Rate (thov,sads) (housandi) (thausanrds) (Percent) 1964 696,387 28,808 11,211 40.9 15.9 96 6.19 2.30 1965 713,984 28,506 9,508 39.4 13.1 90 6.36 2.63 1966 737,982 26,781 8,798 36.1 11.9 84 5.48 2.42 1967 750,965 28,064 8,375 36.9 11.0 81 5.72 2.59 1968 770,634 30,059 8,102 38.5 10.4 77 6.38 2.81 1969 792,611 30,251 7,702 37.6 9.6 73 5.81 2.81 1970 815,160 29,486 7,279 35.7 8.8 69 5.78 2.69 1971 837,367 28,111 7,278 33.2 8.6 66 5.09 2.46 1972 858,200 27,438 7,109 31.6 8.2 61 4.87 2.34 1973 878,529 25,387 6,925 28.6 7.8 56 4.64 2.08 1974 896,991 23,60) 7,111 26.1 7.9 51 3.80 1. 1975 913,4a) 21,797 7,073 23.7 7.7 46 3.39 1.6 1976 928,204 19,846 6,970 21.2 7.5 41 2.95 1.38 1977 941,080 19,340 6,430 20.4 6.8 39 2.74 1.36 1978 953,990 19,137 6,124 19.9 6.4 38 2.60 1.36 1979 967,023 17,547 6,236 18.0 6.4 43 2.31 1.16 1980 978,334 19,704 6,403 20.0 6.5 41 2.54 1.35 1981 991,635 23,121 6,583 23.1 6.6 41 2.96 1,65 19& 1,008,173 Figre 1 lIrth Rate, Death Ruate, and Infant Mortality Rate by Fiscal Year, 1953-54 to 1981-82:* China * 1 1 t -t 1* i aAO 2 40unix)au -X. lisa LAGOfon -7- early 1950's, despite the setback around 1960, were remarkably rapid, and represent remarkable achievements for a population the size of that of China. There can be little doubt that the fall in fertility and rise in mortality in the period 1958-61, which has left a clear imprint on the Chinese age distribution, resulted from the social chaos associated with the Great Leap Forward and a major famine that followed closely after it (Ashton et al., 1984). This demographic crisis cost some 30 million premature.deaths and some 33 million lost or ostponed births, and represents a catastrophe of unprecedented magnitude. China has achieved a remarkable reduction in both fertility and mortality for its general level of development. The upturn in fertility in the early 1980'. seems to have occurred as a result of a change in policy regarding age at first marriage, and the resultant increased rate of family formation (marriages followed by first births) rather than from a substantial increase in second or higher order births. It may be seen as a tempo effect rather than as a quantum effect, therefore, and can be expected to be short-lived. However, the one-child family has clearly not yet become the norm, and fertility declines below the levels of the late-1970's, with a total fertility rate around 2.5, seem unlikely in the near future in the absence of further policy tightening. The rise in infant mortality in the early 1980's; though minor, is probably also a result of existing policy--the rise is accounted for by higher female infant mortality rates, which may well result from female infanticide, Even without this factor, however, the pace of mortality decline appears to have slowed in the late-1970's and early 1980's, a not altogether unsurprising result, given the combination of already low achieved mortality levels and the primary health care focus of the health services. Further mortality reductions will be increasingly expensive and achieved more slowly than over the past 30 years. Data Adiustments The data used in this study have in general not been adjusted, though interpolation has been used to obtain measures for compatible time periods and estimates of demographic parameters have been smoothed to eliminate excessive fluctuations. Two dxceptions have been made to the non-adjustment rule, however: the age distributions from the three censuses have all been adjusted to correct for clear shortages of males of military service ages, and fertility rates from the 1982 fertility survey have been adjusted to take account of apparent reporting of births by Chinese rather than Western calendar dates. The justification for, and form of, the adjustments are given below. Sex ratios (males per female) calculated from the single-year age distributions from the 1953, 1964, and 1982 censuses in general change quite smoothly with age, but in all three cases fall sharply for young adult males. In the case of the 1953 and 1964 censuses, these patterns are not repeated for the same cohorts at the next census, implying that.young adult males were under-represented in the age distributions, but were adequately represented at older ages at the next census. For 1982, there is no later age distribution against which to compare the census age distribution, but the reappearance in 1982 of the same pattern strongly supports a recurrence of the same error. Table 4 shows the sex ratios of the population aged 15 to 35 for 1953, 1964, and 1982, together with the sex ratios of corresponding cohorts at the later census (1982 for 1964) or censuses (1964 and 1982 for 1953). The sex ratios for cohorts by census are remarkably consistent, except for the cohorts aged TABIE 4. Sec Ratios (males per fe-ale) fot Ages 15-35 for 1953, 1964, and 19&, tcgether with Sex Ratios for Corresping Cchorts a Subsequent Censuses; Qina. Sex Racio Sex Ratio Sex Ratio Age 1953 Age Cchorts 1964 Age Ccorts 191 Age Ccharts n 1953 In 1964 n 1.98 In 1964 Ln 198Z In 19& 15 1.189 .LS3 1.156 1.115 1.097 1.060 16 1.109 1.141 1.138 1.118 L.100 L.0S5 17 1.056 1.133 1.123 1.088 1.102 1.050 18 1.061 1.122 1.105 1.045 1.094 .G25 19 1.086 1.133 1.120 1.064 1.112 0.995 20 1.064 1.124 1.126 1.088 1.125 0.973 2t 1.057 1.129. 1.121 1.084 1L36 0,991 22 1.044 L.126 . 140 1.080 L.129 L.030 23 i (o2 1.109 1.123 1090 1.145 1.073 24 1.037 1.105 1.100 i.099 1.139 1.087 25 1.040 1.101 1.104 1L130 1.141 1.076 26 1.068 1.117 L.097 1.153 f,.156 1.063 27 1.059 1.099 1.081 1.141 1.138 1.071 28 1.051 1.093 1.067 1.133 1.123' 1.063 29 1.052 1.077 L.047 1.122 l.105 L-18 30 1.050 1.077 1-.037 1.133 1.120 1.055 31 1.071 1.077 1.027 1.124 L.L26 1.064 32 1.063 1.069 1.019 1.129 1.121 1.101 33 1.058 1.058 1.012 1.126 1.140 1.097 34 1.051 1.047 0.978 1.109 .123 1.100 35 1.060 1.047 0.972 L.105 L100 1.102 16 to 30 in 1953, aged 17 to 25 in 1964, and aged 18 to 22 in 1982, all of which show low sex ratios-too few males--relative to neighboring ages or subsequent enumerations. It is assumed that too few males were included in the age distributions at these ages, and the age distributions have been adjusted by applying interpolated sex ratios to the female populations to estimate the correct number of males. This adjustment did not affect the total populations for 1953 or 1982, because in both cases the available age distributions do not add up to the total reported populations; the 1953 total includes persons not classified by age, assumed to include the missing males, and persons not surveyed, while the 1982 age distribution excludes the institutional population, s.-ummed to include the missing males. The adjustment increases the-1964 population somewhat, since although the age distribution includes persons of not stated age, the sex ratio of the not stated population is 1.006, suggesting that for both males and females the' not-stateds represent true non-response, and do not include the missing young adult males. The second adjustment is made to the total fertility rates from the 1982 fertility survey. In a-comparison of registered births with births calculated using fertility rates from the 1982 fertility survey, Coale (1982, p. 27) noted that the estimates of registration completeness thus obtained were uniformly low for Chinese leap years, which consist of 13 lunar months as opposed to the 12 lunar months of the non-leap years. For the period 1954 to 1980, the completeness estimates for leap years all fall below a three-year moving average of the annual estimates, the average deviation being -4,6 percent. Coale's convincing explanation of this phenomenon is that births were reported, at least in part, to the fertility survey in terms of the Chinese rather than the western calendar, thus increasing fertility rates for - 11 - leap years and reducing those for all other years. If all births were reported in this way, births in leap years would be exaggerated by five percent, while b-irths in other years would be reduced by three percent. For the period 1950 to 1969, the average deviation is close to this maximum five percent, so we have assumed that all births for this period in the fertility survey were reported by the Chinese calendar. Age-specific and total fertility rates for leap years were thus divided by 1.05, and rates for other years by 0.97. For the period from 1969 to 1980, however,'the average deviation is only -3.1 percent, so we have assumed that from 1969 onwards, two-thirds of births were reported by the Chinese calendar and one-third by the Western calendar. For this period, age-specific and total fertility rates for leap years were divided by 1.034, those for non-leap years by 0.980. Demographic Rates 1953-64 The estimation methodology applied to this period is based on the comparison of measures from different sources: the vital registration system (annual births and deaths), the censuses of 1953, 1964, and 1982 (survival of births and previous census population), and the 1982 fertility survey (age-specific and total fertility rates by calendar years 1953 to 1964). These measures are not directly comparable as they stand. The censuses were all held at mid-year, so the age distributions of children indicate survivors of mid-year to mid-year ("fiscal year") births, whereas the other indicators refer to calendar years; further, the age distributions of children reflect the joint effects of both fertility (births in a one year period) and mortality (survivorship to the census enumeration). Total fertility rates are not comparable with birth rates, since the latter are affected by the population age distribution, though the measures must co-vary - 12 - closely over short time periods. A comparison of the different indicators for the period of interest is obviously crucial, however, to an evaluation of the underlying demographic rates. Accordingly, we have converted the available indicators into comparable figures for "fiscal years." There is more scope for examining the consistency of measures of fertility and of implied child mortality than of adult mortality since there are no independent indicators of the timing of adult deaths beyond the registered crude death rates; intercensal population loss indicates the total number of adult deaths, but not their timing. We thus concentrate on fertility and child mortality, and derive trends in adult deaths from registered deaths by year. Comparisons of the alternative sources of demographic information are made as follows. First, the annual numbers of registered births and deaths are estimated from the reported crude birth and death rates and the geometric mean of the beginning and end-year populations, on the assumptio that this is the inver se of the procedure used to calculate the reported annual rates. Numbers of births and deaths for fiscal (mid-year to mid-year) years are then estimated by interpolation, fitting a fourth order polynomial to cumulated births and deaths for four consecutive calendar years centered on the fiscal year of interest. Total fertility rates for fiscal years are obtained by a'pplying the same interpolative procedure to the total fertility rates for calendar years reported by the 1982 fertility survey. The 1953 census age distribution of females aged 4 and over in 1953 is then projected EQrward by single ralendar years to 1964, using smoothed intercensal survival ,probabilities varying by fiscal year to approximate actual mortality trends suggested by the registered crude death rate; migration was assumed to be negligible. Fiscal year births were then calculated from the age-specific and total fertility rates for fiscal years from the 1982 survey applied to the - 13 - projected mid-year female populations. Registered births can then be compared with the projected births for each fiscal year, and apparent completeness of registration can be calculated. The births for each fiscal year from both registration and from the projection, largely determined by the fertility rates from the 1982 survey, can then be compared with the numbers of survivors of those births recorded by the 1964 and 1982 census enumerations; the most convenient form of comparison is apparent survivorship ratios, which can be computed by sex by applying an assumed sex ratio at birth of 106 males per 100 females to the births. Table 5 shows registered births, projected births using the fertility survey rates, implied completeness of birth registration, and cohort survivorship probabilities by sex to 1964 and 1982 by fiscal years from 1953-54 to 1963-64. The estimates of birth registration completeness fall for the first two years, settle around 80 percent for 1955-56 to 1960-61, with the exception of a high value for 1959-60, and then rise abruptly for the last three years, a rise that continues, as shown in Table 13, to nearly 95 percent in the mid-1960's. The survivorship probabilities to both 1964 and 1982 are remarkably consistent by sex though the sex differential changes from a male advantage for births prior to 1960 through approximate equality for 1960 to 1963, to a female advantage for 1963-1964; the sex differentials in survivorship are more marked to 1964 than to 1982. The survivrship ratios are also reasonably consistent by cohort and by end point (1964 or 1982); the general pattern is of a gradual increase in survival with year of birth, consistent both with gradually falling infant and child mortality and with shorter life span, but there are several notable deviations from the general pattern. Births in 1954-55 have higher than expected survival to both 1964 and 1982; births in 1958-59 have much lower than expected survival to both - 14 - A&. 5. Esci=ced amber of birhs based on registracion ar: cbmaied brom 1953 cnsus projecic ard 19 fåertility surver; i=:liied cclec ss of biTh regist-ation-.ad i=plied-I surirsip of bir.h c&crn to 1964 ard 19& c-nsuses: China, fiscal years 1953-54 to 193-6 Suvivcrship Raics o Fisca ear 1/ irhs (thccsmaIds) BL-:h harCchort.-- Projecticn Lmlied kcordin co che kceordir to che Ficlardi Registr,acica 196( C--SLUs 194 ce=S year Regtratic Survi Crpleecss Mles Femies -les -males 1953-54 22,269 26,081 .854 . .750 .733 .717 .713 1954-55 21,303 25,645 .831 .817 .792 .773 .763 1955-56 19,545 24,&4 .787 .770 .744 .725 .721 1956-57 20,908 25,537 .819 .792 .762 .748 .734 1957-58 20,846 25,sm .CG .811 .78 .765 .744 1958-59 17,793 Z,13 .835 .686 .669 .652 .62 0959-M 15,198 17,5>Z $66 .831 .811 * .819 1960-61 11,957 L4,716 .813 .793 .786 * .751 1961-62 17,819 20,875 .354 .752 .750 * .781 19E2-63 28,060 20,398 .908 .983 .989 * .915 1963-64 28,780 30,705 .937 .925 .944 K. Tocal 224,278 i4,736 .47 n.a. n.a. n.a. d.a. WM, Male surivor"hip to 198Z for so:e y-ears rot given, sire mle peuacin -ts adjusced ou basi3 of feale populition for ages 18-22. For expLa katic t of calculaciena, see texc. - 15 - 1964 and 1982; births in 1959-60 have somewhat higher than expected survival, especially to 1982; births in 1961-62 have lower than expected survival to 1964, though their survival to 1982 is not clearly out of line; and births in 1962-63 have higher than expected survival which is particularly marked to 1964. Table 5 compares measures from four supposedly independent sources, and the variations observed provide powerful support for this independence. However, the nature of the sources provides some basis foi drawing conclusions about the causes of the variations for each birth cohort. We discuss these variations below, by fiscal year of birth. i) 1953-54: completeness of birth registration is high, but all the survivorship probabilities are reasonable. Since the two censuses and the fertility survey are consistent, birth registration is probably abnormal, a conclusion strengthened by the observation that the official birth and death rates through 1953 are rounded and may not be based on actual registrations. We therefore accept the fertility survey births and their survival probabilities. ii) 1954-55: completeness of birth registration is rather high, but all the survival probabilities are rather high. Birth registration completeness might come out high as a result of a continuati,on of the birth rate rounding proposed for 1953-54, but there is no obvious reason why the mortality risks for this cohort should have been lower than those of adjacent cohorts, or why the population aged 9 in 1964 and 27 in 1982 should both have been over-reported. We therefore conclude that the - 16 - fertility survey rates for 1954-55 are some seven percent too low, a conclusion that would reduce birth registration completeness to around 77 percent and cohort survivorship ratios by seven percent. iii) 1955-56 to 1937-58: all the measures are approximately consistent, so the fertility survey rates and survivorship probabili- ties are accepted. iv) 1958-59: completeness of birth registration i's about normal, but. the survivorship probabilities are all low. Since it is unlikely that the population aged 5 in 1964 and 23 in 1982 was consistently under-enumerated, the measures are accepted as consistent but indicating sharply heavier child mortality for this than for earlier birth cohorts. v) 1959-60: comple.teness of birth registration is high, and the survivorship probabilities are rather high, to 1964, and high to 1982. The female population aged 22 in 1982 may be over-reported, since it is an officially advocated lower age limit for female marriage, but there is no reason to suppose that the population aged 4 in 1964 was over-reported. We therefore conclude that registration completeness was about the normal 80 percent, implying survivorship to 1964 somewhat lower than normal, and that the fertility survey rates for this year are some six percent too low. vi) 1960-61: all the measures are approximately consistent, so the fertility survey rates and survivorship probabilities are accepted. - 17 - vii) 1961-62: birth registration completeness is rather high, suggesting too few births from the fertility survey, but the survivorship probabilities to 1964 are low, suggesting too many births; survival to 1982 is not so low, but the number of females aged 20 in 1982 may be inflated by age at marriage laws. On balance, we accept the fertility survey results and survivorship pr(babilities, excepting that for females to 1982, and conclude that birth registration completeness started to rise during this fiscal year. viii) 1962-63: birth registration completeness is high, survivorship to 1964 impossibly high, and survivorship to 1982 for females also implausibly high. A tempting conjecture would be that one-year-olds were over-reported in 1964, by an upward age shift from age 0, a common error in developing country age distributions. However, female survivorship from 1964 to 1982, age 1 to age 19, is a not-implausible 0.925, so any error in 1964 was followed by an error of similar magnitude at A quite different age in 1982. A more likely explanation seems to be that the fertility survey rates for 1962-63 are too low, possibly as a result of the fiscal year interpolation procedure used here. However, even a registration completeness of 80 percent, the lowest that can reasonably be assumed, would reduce survivorship probabilities only to around 0.87 to 1964, consistent with an infant mortality rate of only about 110 per - 18 - thousand live births, which in turn would be consistent with somewhat higher survivorship than that observed from 1964 to 1982. There are thus probably errors in several of the sources: the interpolated fertility survey rates are too low, the 1964 population aged 1 too large, the 1982 female population aged 19 possibly too small (some 19-year-olds having been reported as age 20); all we can conclude is that birth registration was more complete, that fertility was higher than the interpolated rates, and that infant mortality was much lower than its average level over the preceding decade. ix) 1963-64: though-birth registration completeness is high, so is survivorship to 1964; survivorship from the 1964 census to 1982 is somewhat low, though it incorporates an element of high infant and early child mortality. It is hard to evaluate these measures effectively, because of the importance of the age pattern of early child mortality, but it is not evident that any one is inconsistent with the others. We therefore accept the general level of fertility indicated by the 1982 survey and the level of child mortality shown in Table 5. From this year-to-year analysis, we draw the following conclusions about fertility; 1. Birth registration was approximately 80 percent complete from 1954-55 to 1960-61, then improved to over 90 percent complete by 1963-64; the - 19 - official birth rate for calendar year 1953 was probab1- not based directly on registered births. 2. Fertility was quite high in the early and mid-1950's, with total fertility between 6.0 and 6.5 and the crude birth rate in the low forties. Fertility then fell in 1958-59 by nearly 20 percent, fell again slightly in 1959-60, fell again by about 25 percent in 1960-61, recovered somewhat in 1961-62, then rose very sharply to a peak in 1962-63, with total fertility over 7.0 and the birth rate close to 50 per' thousand, before declining slightly to 1963-64. Turning to child mortality, we have in Table 5 survivorship ratios for adjusted births by sex and fiscal years to the 1964 census. These survivorship ratios reflect the cumulative effects of different age and period mortality rates as each cohort ages from birth to census enumeration, and they cannot be interpreted as measures of mortality level in the fiscal year of birth, even though the mortality rates suffered will normally be highest in the first year of-life. However, by a stepwise procedure we can attempt to estimate period mortality levels from these survivorship probabilities. The survivorship probablity for male (or female) births in 1963-64 represents only the mortality level of that fiscal year, and approximates the life table function L./l(.. For each sex, we can find the implied mortality level in a model life table family, and assume that the same mortality level applied to all birth cohorts as they passed through the year. Using ratios 1LJIL- for the implied level, we can reverse-survive each 1964 single year cohort to mid-1963. We can then repeat the process for male (or female) births in 1962-63, calculating the survivorship ratio to mid-1963, finding the model mortality level of the corresponding xL0/lcy, assuming that this level applies to all older cohorts, and projecting - 20 - backwards to mid-1962. We continue this procedure backwards year by year to 1953, finding model mortality levels for each year that are consistent with the observed cohort survivorship probabilities, the child mortality pattern of the model, and the assumption of a single period mortality level applying to all cohorts-passing through the year. Using the Coale-Demeny (1966),L model life tables, it turns out to be impossible to apply this method across all cohorts. The fluctuations in survivorship probabilities are too sharp fr6m cohort to cohort to be able to asstme period levels applying to all cohorts without obtaining negative mortality for at least one period, even using the "East" mortality pattern, which has the highest infant mortality relative to child mortality, and can therefore represent the widest.fluctuations of any of the four families. However, by adjustiag the survivorship probabilities to 1964 in Table 5 for the ages 1, 4 nd 9 in accordance with the above discussion of birth registration and enumeration completeness, loy using the East family, and by smoothing somewhat the extreme fluctuations in survivorship probabilities for cohorts aged 2, 5, 6 and 7 (for males) or 8 (for females) in 1964, it is possible to obtain a reasonable sequence of child mortality levels and infant mortality rates by sex for fiscal years from mid-1953 to mid-1964. The levels and infant mortality rates obtained by period, and the implied cohort survivorship probabilities, are shown in Table 6. The fluctuations not accounted for may reflect cohort effects, whereby-particular birth cohorts that experienced high initial mortality continued to experience mortality rates above period averages, presumably because they were weakened by their. 1/ Coale, A.J. and P. Demeny. Regional 'Model Life Tables and Stable Populations. Princeton University Press. Princeton, N.J., 1966. - 21 - BE 6. COhort Survivorship Patios to 1964 and Correspding Hrrality Lavels and iant brtalicy 2aces, by Sex: Chinra, Fiscal Yezrs 1953-54 to l963-% Yales Femaes Cohort Cohort Survivorshio Ratiøs Suzvivorsnip Ratios Cbserved East Infant Gserved East Znfant Fiscal (birth to Mo"rtality Mortality (birth to Mrtality otality Year 1964) v£del Level mate 1964) Model level Lvel 1953-54 .750 .750 L2.5 185 .733 .733 11.1 178 l954-55 (.760)* .760 13.0 176 (.738)* .738 10.6 187 1953-56 .770 .770 13.2 172 .744 .755 (11.5)** 71 1956-57 .792 .773 (14.2)** 155 .762 .762 12.3 L58 1957-58 .811 .756 (15.0)** L4t .780 .742 (14.0)** 132 1958-59 .686 .724 (10,0)** 235 .669 .109 (8.0)** 237 1959-50 (.77Z)* .772 11.9 195 (.753)* .753 9.1 215 1960-61 .793 .793 15.9 126 .786 .786 13.6 138 1961-62 .752 .795 (10.0)** 235 .750 .803 (8.0)** 237 1962-63 (.392)* .892 18.0 93 (.910)* .910 17.5 83 l963--64 .925 .923 17.0 108 .944 .944 17.5 83 * Mjusted observaidL"s. *~ Levels soothed to avoid as.sumption of impossible or üipausible levels i prior years. - 23 - TÆZ7. Sm:othed Sasc r:alee13i under age 10 by sex: China, fiscal years 1953-54 to 1963-64 Fiscal year ~ Males . Fales l953-54 L.2.5 l0.6 L954-55 13.0 11.1 1955-56 13.6 11.6 1956-57 i4.2 l2.3 1957-58 L3.0 13,0 1958-59 10.0 8.0 1959-60 11.9 91 1960-Lý 13.9 13.6 1961-62 W0.0 8.0 1962-53 18.0 17.5 1963-M4 17.0 L7.5 ZT SL 8. Tocal Regised Deachs, Esdceied teaCs Ln:pr Age 10, ani Cenpari of Regiscered De.achs co Registered 3irths: China, Fiscal Years 1953-4 to 1.96 ~aRio: Åp:iër Age O Registered ceaths Lter 10 Tocal Registered ard Srwy Ceaths tc veer (chusan:Is) (Lheoaalå) DE-azhs(c:smi) it 195354 7,926 6,610 083 26,081 0.30 1954-55 7,65 6,w:9 0.56 27,312* 0.28 t9551-56 7,275 6,088 0.4 Z4,&4 Ø.29 956-37 6,399 5,633 0,s 25,537 0.27 1957-58 7,220 5,193 0.7n- z5,"co 0.28 195-59 8,389 8,691 L.C4 22,1:13 0.38 1959- 13,92 6,491 0.47 ls,S* 074 1960-1 13,7& 3,263 0.24 14,716 0.94 1961.2 7,5324 7,062 0.94 20,875 0.36 19&-63 5,494 3,197 0.49 33,813* 0.19 1963-54 7,514 3,85 0.51 30, 705 0.24 ctca 94,47 6.,662 0,66 270,658 0.35 dn the basis of incerploaced suiweship ratios. -24- ages does not vary closely with the sequence of deaths under age 10. The ratio of deaths under age 10 (which we hope are close to the true numbers) to total registered deaths, assumed to suffer from under-registration, varies around 80 percent to 1957-58, rises to 104 percent for 1958-59, falls to below 50 percent for 1959-60 and 1960-61, rises to 94 percent for 1961-62, and then falls around 50 percent for 1963-64. The period with peak numbers of registered deaths, 1959-61, is a period with relatively few estimated deaths under 10, giving rise to the lo ratios of child to total registered deaths, whereas the years of high estimated deaths under 10, 1958-59 and 1961-62, have only moderately higher total registered deaths, giving rise to the high ratios of child to total registered deaths. There are three possible explanations'for the lack of consistency between the series of registered deaths and estimated child deaths. The first is that the series of deaths under 10 is wrong. Accepting this explanation would mean dismissing the consistency between the 1964 age distribution, the 1982 fertility survey results, and registered births by year. Also few births occurred in the period 1959-61, so even a doubling of infant deaths would have had little effect on total deaths; thus, the huge numbers of registered deaths are unlikely to reflect child deaths in the main, and thus provide little information about them. This point is reinforced by the sequence of ratios of registered deaths to adjusted births; these ratios are fairly constant through the mid-1950's, rise to a peak for 1959-61, and then fall to low levels for 1962-64. Thus, it is possible that the sequence of child deaths is wrong, but the sequence of registered deaths does not prove it. The second explanation is that the completeness of death registration varied sharply year by year, with registered deaths thus not reflecting the trend in actual deaths at all closely. This explanation does not seem very - 25 - likely, since the completeness of birth registration apparently remained approximately constant, at least until 1962, and a coverage change in deaths would be expected to be associated with a similar change in births. The third explanation is that the surge in registered deaths in the period 1959-61 resulted from very high adult mortality in the two years, not reflected in child deaths. This explanation would require that the underlying cause of the high adult mortality did not have a marked impact on child mortality. Though this explanation is not on the face of it very plausible, it is possible that severe famine accompanied by strict food rationing, or other method of food allocation, giving preference to young adults and children could produce such a differential, an account that would be consistent with the very low survivorship ratios for the elderly observed between 1953 and 1964. Famine might also exacerbate various morbid conditions differentially, resulting in marked differentials by age in the effect on mortality; in this respect it is interesting to note that the brunt of the mortality increase around 1960 appears to have been borne by the population already alive in 1920 and thus that lived through the 1919 influenza epidemic. The demographic evidence tends to favor this third explanation. Both age-specific growth rates and, intercensal survivorship probabilities (shown in Table 10) suggest that, on average over the period, mortality above age 40 was substantially higher between 1953 and 1964 than it had been prior to 1953, the excess being particularly severe for males, The sharp drop in survivorship ratios at age 35 (in 1953) would be consistent with a famine combined with a rationing scheme giving priority to those under age 40 during 1959-61. Thus, the annual numbers of registered deaths are not necessarily inconsistent with the estimated numbers of deaths under 10, since the huge numbers of deaths registered for the period 1959-61 probably reflect very - 26 - largely deaths over age 40. Child deaths may have increased in 1958-59 before rationing measures were put into effect, may have fallen to more normal levels in 1960-61 as a result of targeted rationing, and may have increased again in 1961-62 as rationing was relaxed. The obvious alternative hypothesis, that child mortality was actually very high in 1959-61, but that births were severely under-registered because of the high infant mortality, and that births for the period recorded by the 1982 fertility survey were under-reported badly for a similar reason, fails to explain why survivorshi'p ratios to 1964 are actually higher for the births reported for 1959-61 than for the births reported for 1958-59 or for 1961-62; if infant deaths led to omissions of births between 1959 and 1961, then they would be expected also to lead to omissions in 1958-59 and 1961-62. If we accept the sequence of deaths under age 10 in Table 8, it is clear that the registration of deaths under age 10,was substantially less complete than the registration of deaths at age 10 and over. The increase of child deaths of 3.5 million from 1957-58 to 1958-59 is associated with an increase in registered deaths of only 1.2 million, while the decline in child deaths of 3.9 million from 1961-62 to 1962-63 is assad-iated with a decline in registered deaths of only 1.0 million. Taken together, these two observations suggest that the registration of child deaths was only about 30 percent complete. We can estimate total intercensal deaths as the 1953 population plus intercensal births minus the 1964 population, giving 156,875 thousand. Deaths at age 10 or over ca'a be estimated by subtracting deaths under 10, giving 94,213 thousand. if 30 percent of deaths under 10 w. ere registered, then 78,848 thousand deat4s at 10 or over were registered, implying a completeness of registration of 80.5 percent. Marked differentials in registration completeness between child and adult deaths are commonly observed - 27 - in developing countries, so such a difference for China is not implausible. That the completeness of registration for adult deaths is roughly the same as that for births is not necessarily implausible either, since births rapidly followed by death may be much less likely to be registered than births that survive a year or more. Table 9 shows the numbers of deaths by year under age 10, 10 and over, and total, assuming that the registration completeness of deaths under 10 was 30 percent, while that-for deaths over 10 was 80.5 percent. The sequence of deaths 10 or over by year looks quite reasonable,' except perhaps for the increase in 1963-64, a year in which death registration completeness may have increased, as it apparently did for birth registration. The sequence of registered deaths is the only available source of information about the timing of adult mortality between 1953 and 1964, but survivorship ratios from 1953 to 1964 by age group provide an indication of the age pattern of intercensal mortality. Survivorship ratios by sex for initial age groups aged 10 and over in 1953 are shown in Table 10 with corresponding "East" mortality levels. In terms of mortality levels, the survivorship ratios plunge for initial age groups over 30 in 1953; for males, the levels continue to drop as age increases, whereas for females they stabilize around level 7. It seems unlikely that the mortality patterns implied by the 1953-64 survivorship ratios were standard features of Chinese mortality, since the age-specific growth rates (Table 10) fall sharply for males from about age 45 to negative values above age 65, and fall sharply for females above age 65'to close to zero if mortality had been falling prior to and during the period, the age-specific growth rates would be expected to increase with age above age 40. A further indication that the 1953-64 experience was not typical of China prior to 1953 is the changes in the age distribution above age 40 from 1930 (taken from the Chinese Farmers Survey, - 28- TABLE 9. Estimated deaths (husnds) under age 10 and age 10 or over China, ýiscal years 1953-54 to 1963-64 Fiscal ya%ieLr age 10 Age 10 Ir older Tocral 1953-54 6,610 7,382 13,992 1954-55 6,609 7,077 13,686 1955-56 6,088 6,768 12,856 1956-57 5,633 6,470 11,103 1957-58 5,193 7,045 12,238 1958-59 8,691 7,18 15,873 1959-60 6,491 14,877 21,368 1960-61 3,263 15,903 19,166 1961-x2 7,062 6,727 13,789 1962-63 3,197 6,875 10,072 1963-64 3,825 7,908 11,733 Total 62,662 94,213 156,875 tZ l0 Intercewsal Age- ard Se:-specific Sar--ivL Ratios, 1953-54, lapLied E,azt thralicy Levels, ard Growtýh Races by Ag Gep and Sex: hin-a, 1953-54 to 1963-64 yales emes IntercensaG=?Wth incercen=alGot Age Group Strvival .4rtality Ratý $rvival 'brtalicy Race (1953) "ates I.lvel (pexcnm), Rates Level (percent) 10-14 .924 5.7 3.83 .959 12.7 4.58 15-19 ,940 1l.9 1.30 .949 12.8 L 61 20-24 .929 11.3 0.66 .933 U.$ 0.57 35-29 .921 11L4 L.42 .925 LI.S LL3 30-34 .893 10.0 ,.07 . .883 .0 1.47 35-39 .818 5.6 1.13 .852 6.1 0.91 40-44 .754 3.9 0.90 .830 6,2 L.02 45-49 .687 3.1 0.48 482 7.8 0,55 50-54 .584 1.7 0.47 .728 8.0 0.83 55-59 .448 < t.0 0.16 .601 7.3 Z27 60-64 .313 < 1.0 0.00 .458 7.2 0,96 65-69 - - -0.64 - 011 70-74 - - -102 - --0.37 75-59 - - -'0.08 - 0.31 80- - -0.54 - -0,22 - 29 - and admittedly not necessarily representative of all China) through 1953 to 1964. The age distributions, shown in Table 11, show increasing proportions above age 60 from 1930 to 1953, consistent with declining mortality, but falling proportions for the same age groups from 1953 to 1964. The information thus suggests that, on average over the period, mortality above age 40 was substantially higher between 1953.and 1964 than it had been prior to 1953, the excess being particularly severe for males. The sharp drop in survivorship ratios at age 35 (in 1953) would be consistent with a large number of famine-associated deaths at ages 40 and over during the period 1958-62. We lack year-by-year information on the age pattern of mortality, so we have no basis for the direct calculation of annual life tables or of summa'ry measures such as expectation of life at various ages. However, we have annual estimates of deaths and population over age 10, so we can calculate death rates over a-e 10. By assuming the population over age 10 to be approximately stable, we can obtain rough estimates of expectation of life at age 10, using the death rates over age 10 to interpolate within the Coale-Demeny model stable populations. Rough estimates of expectation of life at birth for each fiscal year can then be obtained by splicing on the child life table levels arrived at in Table 7. Results are shown in Table 12. Expectacion oE life at birth rose by almost one year per year between 1953 and 1964, despite falling to a level almost 10 years below the 1953-54 level in 1959-60. Life expectancy at age 10 rose by only some two and a half years over the period. major gains were clearly made in child mortality reductions, with less impressive gains for adults. -”」’&.&‘「十루．7&7’··，「｀．깍「”’써”》맥”k어？컫 31 Demograohic Rates 1964-82 for the period 1964-32, we have one information set less to wark with, than for the period 1953-64, since we do not have an age distribution for well after the end of the period, but we also have several additional information sets: a distribution of deaths by age for the years 1973-75 from the Cancer Epidemiology Survey (CES),'a life table for 1981 based on deaths by age in 1981 from the 1982 census, and children ever barn and children surviving by age group of mother from the 1982 census, providin- the information needed for the indirect estimation of child mortality. As a first step, however, we can repeat the analysis applied to the period 1953-64: calculating fiscal year births and deaths from the registration system, calculating fiscal year births from the interpolated female population and fertility survey age-specific fertility rates, calculating child survivorship probabilities for the births to 1982, and estimating period child mortallity levels recursively from the survivorship probabilities. Table 13 shows the births calculated from the two sources, the birth registration cdmpleteness implir-d for each year by the fertility survey rates, and the survivership rates to 1982 by sex, assuming a sex ratio at birth of 106 males per 100 female births.' No birth or death rates for 1980 have bae'n published, so the 1980 re--istered birth rate w.as*estimated by interpolating between 1979 and 1981 an the basis of fertility survey total. fertil--ty rates for the three years. The birth re-istration completeness implied by the fertility survey starts at around 95 percent for the first four fiscal years, 1964-68, drops to around 1131 percent for the four fiscal -years 1968-72, possibly aS:SOC4ated with t'he Cultural Revolution, rises to around 92 percent for the three fiscal years 1972-75) then falls steadily for the five fiscal years 1975-80, before rising 州 T-1 Iq Mr, M T M wm_ 33 sharply for the last two fiscal years of the period, 1980-81 and 1981-82, the estimate of the latter year reaching more than 100 percent. The survivarship rates to 1982 are remarkably consistent by sex, showing a slight female advantage for the first two ears, 1964-66, then a slight male ad-vantage for 13 years, 1966-79, and a marked male advantage for the last three years, 1979-82. The survivorship ratios show considerable variations by year, however, with high values for the birth cohorts of 1965-66, 1968-69, 1970-71, 1973-74,'1976-77, 1978-79, 1979-80, and 1981-82. Each of these y'ears includes a Chinese leap year, so we may have over-corrected the fertility survey births for leap years somewhat, or the 198, population may itself be affected by leap year effects, if aae were reoorted by the Chinese calendar. The general trend of the survivorship rates does not seem to reflect the variations in implied birth registration completeness, however, supporting the conjecture that re-istration completeness really did vary during the period. The perio from 1976 to 1982 merits special attention. Implied registration completeness falls for four years to 0.861 in 1979-80, the lowest level -for the whole 18-yeaz,period, and then increases sharply for the last two years to the highest observed level with complete registration in 1981-82. The survivorship ratios to 1982 implied by the fertility survey births fluctuate sharply, but are tmplattsibly high for four 'ears (1976-77, 1978-79, 1979-80, and 1981-82) out of the si.,c, particularly for males. Such shar-o fluctuations -'in birth registration completeness, which had been in the range of 90 to 95 percent for the previous 12 years, seem unlikely to be corr _,ct-, though the implementation of the one-child family policy could have led to lower registration completeness in the first four years followed by an offi,cial crackdown resulting in higher completeness in the last two years. Rowever, if registration completeness was in fact essentially'constant --end` - 34 - throughout, the fertility survey recorded too few births in 1980-81 and 1981-82, and too many for the years 1976-80; such an error might also arise from the one-child family policy, if respondents to the survey pushed back the dates of recent births to give the impression that they occurred further in the past. The survivorship ratios indicate errors either in the fertility survey births (too few) or in the census population (too many) for the birth cohorts of 1976-77, 1978-79, 1979-80, and 1981-82. The census age distribution may itself be distorted by reporting errors induced by the one-child family policy, again by exaggerating the age of young children, and thus pushing back the reported date of birth. It will not be possible to sort out this tangle of errors until another age distribution or fertility history sequence becomes available. It is clear, however, that there are errors, and it seems likely that both the fertility rates and the age distribution are affected by them, given the implausible sequence of birth registration completeness in Table 13 (assuming the fertility survey to be correct) and the even more implausible sequence that would result from assuming the age distribution to be correct. We there7fore conclude that both the fertility survey and the age distribution from the 1982 census are distorted for this period by date or age reporting errors, and assume that birth registration completeness remained constant over the period at its average level over the period of 90.9 percent. The total fertility rates in Table 3 for the years 1976-82 are the fertility survey rates adjusted for level so as to maintain this level of registration completeness for each year. Althou&h the exact trend in fertility in the years immediately preceding the 1982 census remains unclear, all the sources show a rise in fertility for 1981 and 1982. The registered birth rate increases from around 18 per thousand in 1979 to around 21 per thousand in 1981 and 1982, the total - 35 - -fertility rate from the fertility survey increases from 2.24 in 1980 to 2.63 in 1981, and the 1982 census population aged 0, born in 1981-82, is nearly 20 percent larger than the population aged 1. Most of this increase in fertility appears to have resulted from changes in the timing of fertility rather than in changes in general fertility -level. The sum of age-specific first birth rates recorded by the 1982 fertility survey increased from 0.869 in 1980 to 1.162 in 1981. Cumulated first birth rates indicate the proportion of women who will ultimately become mothers given the current period rates throughout their lives; thus, the 1980 sum indicates that 86.9 percent of women would become mothers given 1980 first birth rates, probably somewhat below the proportions of each cohort of Chinese women who will actually have at least one child, and thus indicating some prostponement of first births. The 1981 sum indicates that 116 percent of women would ultimately become mothers given the 1981 first birth rates, clearly an impossibility and indicating a bringing forward of first births, no doubt related to the relaxation in 1980 of some local restrictions on age at female marriage. Changes from 1930 to 1981 in the sums of second birth rates (0.564 to 0.624), and third-plus birth rates (0.802 to 0.839) were mtore modest, increasing by 10.6 and 4.6 percent respectively. Thus, the fertility survey data suggest that most of the fertility increase in the early 1980's resulted from a change in age at family formation rather than from changes in individual fertility level within families. The rise in fertility should not therefore be seen as a major public rejection of official.fertility policy; there was some increase in the rates of bearing second and higher order children, but most of the fertility increase came about from a bunching of first births in 1981 (and probably also in 1982), Once this timing effect has run its course, fertility can be - 36 - expected to fall again, With more normal first-birth rates, the 1981 total fertility rate would have been around 2.4. Turning to child mortality, the survivorship ratios by cohort fluctuate too sharply from year to year to be able to use the step-by-step procedure applied to estimate annual child mortality levels for the period 1953-64. We have therefore applied the procedure to a three-year moving average of the calculated ratios. Wp also have some independent information on the pattern of child mortality. The 1981 life table indicates a "North" pattern of child mortality, whereas the 1973-75 CEM indicates a "West" pattern. We have therefore used a "West" pattern from 1964 to 1976, and a "North" pattern for 1976 to 1982. Survivorship ratios for male cohorts aged under five in 1982 could not be fitted with plausible mortality levels, so. fiscal year levels were assumed in such a way as to show some slight downward trend in child mortality and to fit observed survivorship by 1975-76, For females, the 1981 life table indicates that female child mortality was about one level lower (that is, heavier mortality) than male mortality, However, the survival ratios for female births in the years 1979-82 indicate a much larger female disadvantage than one mortality level* We have therefore assumed that the excess female disadvantage represents excess female early infant deaths associated either with infanticide or intentional neglect, and we have estimated one level of mortality for female infants and another, taken as one mortality level below the corresponding male estimates, for female mortality after age 1. Table 14 shows the observed survivorship ratios (three-year moving averages of the values in Table 13), the cohort survivorship ratios implied by the step-by-step process, and the period mortality levels obtained by the process. Some local smoothing of period levels was applied to avoid extreme year-by-year fluctuations (indicated by - 37 - TABrz 14. Cchort Suri.otshio Racics co L982 md correso:rdig Pericd' :aliy lavels by S.ex; Cirna, Fiscal Years 19&'-53 co V981-82 ale .rale Cohorc Sutrivorshio ?acios ,t;w' 0eiod sk.-:t s vrsnio Racios '%sc' Period FiLcai Cbsered* :del 42rcality Lavel** Cbserved* Ytdel Morality Lavei* Year infu=y chiichcd 1965-é .3665 .8665 16.5 .682 .3692 (15.) 196-67 .8765 ,8776 (16.5) .8753 .8753 16.0 1967-68 .8901 ,8901 17.3 .8880 .843 (11.0) 19e-9 .8993 48993 18.0 .8968 .3968 L6.7 1%9-70 .9174 .9052 (18.5) .9140 .9018 (18.0) k970-71 ;904c .9106 (18.5) .8991 .9036 (17.0) 1971-72 .9169 .9169 18.9 .9119 .9119 17.7 1972-73 .9231 .9231 19.1 .9189 19189 17.9 1973-74 .9316 .9316 t9.7 .9275 .9275 18.5 1974-75 49394 .9394 20.4 .9346 .9346 19.4 1975-76 .9461 .9461 20.6 .9401 .9401 19.3 1976-77 .9499 .9552 (21.3) .9439 .9493 (20.0) 1977-7s .635 .9610 (21.8) .9563 9563 20.6 1978-79 .9721 .9658 (22.2) .9613 .%18 .. 1979-0 .9660 .9679 (22.2) .9514 .9514 18.8 21.3 980-31 .9740 .9715 (22.5) .9570 .9570 19.2 21.5 1981-32 - .9760 22.5 - - 19.2 21.5 *1,hree-ar :mug averags of iscal year cohorz vatues. ** e trel used for 1976-77 lo 1981-32; cal secochin co avoid 51u::aciEcns applied Cor ''ers -hen :he chseved e d ::del ratics do rc are (mnna:icy lwveLs in are.ches. ........F - 38 - mortality levels enclosed in parentheses). The resulting sequences of mortality level show fairly steady declines in child mortality over the period, and roug.hly constant male-female differentials of about one mortality level at least until 1979. Final estimates of child mortality level from this stepwise process, shown in Table 15, were obtained by some further smoothing. The estimated infant mortality rates, also shown in Table 15, fit in reasonably well with the sequence estimated for the period 1953-64 in Table 6. The excess fem ale infant deaths in the years 1979-82 total around 375,000, and increase the female infant mortality rate from around 35 to over 45 per thousand live births. The period infant and child mortality levels in Table 15 can be compared with three independent sources. The -1982 census included-a-question on household deaths by age and sex in 1981, and a life table has been constructed from the reported deaths (Jiang et al., 1984).'' This life table indicates somewhat higher male infant and child mortality and higher female child mortality than the stepwise sequence, but indicates lower female infant mortality. Part (a) of Table 16 compares the "North" mortality levels under age 10 from the life table with the stepwise results. We do not usually expect a retrospective question on deaths to overestimate mortality, particularly in childhood, but accepting the 1981 life table childhood mortality, at least for males (the 1981 life table shows no female infanticide effect), would mean rejecting the cohort survivorship rates in Table 14 as far back as the mid-1970's, even assuming that child mortality had not fallen at Jiang, Zheng-hua, Zhang Wei-min, and Zhu Li-wei. The Preliminary Study to the Life Expectancy at Birth for China's Population. Paper preparet for International Seminar on China's 1982 Population Census. Beijing, China; March 1984. - 39 - .aBI. 15. -imaces of Thc.mc and child mr-aLicy Lavels, and iznea, Mrta1icy ace, by Sex, China, Fiscal Years 1964-6§ Co 1981-32 aIes F=rales yea'r L.avel FýamiyraiyRt Ifny Ciedily ly MraiyRace 1964-63 16.0 %t'99 '15.0 se' 93 1963-ý6 16.5 %sc' 92 15.5 %st' 38 1966-67 17.0 %s 86 .0 82 196768 17.3 ;éet' 83 IG.3 sc' 78 196-% 17.6 %st 79 16.6 75 1%9-70 . 18.0 %st' 74 17.0 %st 71 1970-71 18.3 71 17.3 67 1971-72 13.6 67 17.6 %e2' 64 l972-73 190 63 18.0 60 1973-74 19.5 57 18.5 55 1974-75 20.0 52 19.0 3'50 19735-76 20.5 Ws' 46 19.5 %5t' 45 1976-77 21.3 37 10.0 N=ch 40 1977-78 21.3 trth 33 20.6 35 1978-79 22.2 %rth 30 2LIh 31 1979-0 zZ. h 30 18.8 21.2 %rth' 50 1980-61 22.5 27 19.2 21.5 treh £6 1981-N2 22.3 27 19.2 21.5 Mrth 46 - 40 - TABLE 16 TANE 16. Ccmpairisons of Stepu-ise Estiratec of Child Mnrtality Levels with Independent Estinates; C1iria, Variouo Years, 1970-82 Part (a): Cap>aison with 1981 Life Table Age Males Female x Life Table Step-vise Life Table Step-ise Ox qrt' Level 'mC' Level nqx 'orth' Level 'orth' Level 0 .03556 21.5 22.3 .03372 20.9 19.2 1 .01600 20.9 22.5 .01778 20.3 21.5 5 .00650 21.6 22.5 .00537 21.2 21.5 Part (b): Campari-on with Banieter-Preston 1973-75 Life Table Age Boh Sex. x Life Table S te-.ise 'weit' Level nx Wecl ievel 1973-74 1974-75 0 ,05629 19.0 19.0 19.5 1 001998 19.0 19.0 19.5 5 .02393 12.0 19.0 19.5 Part (c): C~parisco vith Estimatec fran 1982 Prportione Dead of Children Evr Bor 1982 Cen-us Proportion of Average Cort Average Parity C~hort Age West Model Weferee Step-vie Age of Ciildren Dead Parity frc Fertiliry Survey Parity x X0o Wvel Date tev1e ther 1977 1982 Satio 20-24 .0621 .420 .020 .414 .0483 2 .0708 18.6 1980.8 21.9* 25-29 .0637 1.593 .526 1.588 .3312 3 .0669 19.2 19794 21.7* 30-34 .0754 2.763 2.055 2.766 .74P 5 .0741 19.1 1975.4 19.7 35-39 .0975 3.802 3.490 3.781 .9230 10 .0947 TB.3 1971.7 18.0 *Irchl rather than 'S ' odel. - 41 - all over the period. The 1981 life table for childhood is thus inconsistent with the fertility survey births and 1982 age distribution of children under age 10. Since the life table itself must be based in part on the 1982 age distribution for rate denominators, omission of young children from the census age distribution (or an upward shift in their ages) would result in overestimates of child mortality in the life table, but would also result in overestimates of early child mortality from the stepwise process. The difference between the two sets of estimates is therefore due largely to errors either in deaths reported for 1981 or in births obtained from the fertility survey. The most likely way in which deaths might be over-reported is by the inclusion of deaths that occurred either before or after 1981. Such an error is a possibility, especially for China with two calendar hystems being used. However, it remains possible that the estimates of child Mortality for the early 1980's in Table 15 may be underestimating child- mortality in the early 1980's by a Coale-Demeny level or so, and overestimating slightly the pace of child mortality decline for the period 1975-82. The second independent source for child mortality is the 1973-75 CES, which covered some 90 percent of China's population, and provides deaths by five-year age groups for both sexes combined for the study population. The CES also provides an age distribution for the study population, but the reference date for this population is not clearly defined. However, Banister and Preston (1981)1' have derived an adjusted life table from the basic data. Panel (b) of Table 16 shows age-specific mortaity rates under age 10 from this life table, together with the implied "West" model life table Banister, J. and S. H. Preston. Mortality in China. Population and Development Review, Vol, 7, No. 1, pp. 98-110. -42- mortality levels. Under age 5, the life table and stepwise mortality levels agree closely, but,the life table shows very much heavier mortality for the age group 5-9, a differential with the mortality levels under age 5 not matched by any model life table system or reliably recorded real life table. Under age 5, the life table supports the stepwise sequence, while for age group 5-9, the life table is probably wrong. The third independent basis for child mortality estimates is the 1982 census data on children ever born and children surviving by age group of mother. A number of procedures have been developed for obtaining child mortality estiates from proportions dead of children ever born, founded on pioneering work by Brass (1964).-" The early methods assumed constant fertility and mortality.in the past, clearly-inapplicable to China in 1982, but more flexible.methods have recently been introduced to obtain reference dates for the estimates and to allow for changing fertility (U.N., 1983).1 We use here a method that estimates period child mortality risks under conditions of changing fertility using ratios of average parities (numbers of children ever born) for cohorts at two dates five years apart to allow for changing fertility, The 1982 census provides average parities by age group for mid-1982, but not for mid-1977. We have therefore used the single year of age/single calendar year-specific fertility rates from the 1982 fertility survey to estimate average parities by cohort at the end of 1981 and at the end of 1976, and used these parities to calculate cohort parity ratios. 17 Brass, W. Uses of Census or Survey Data for the Estimation of Vital Rates. Paper prepared for the African Seminar on Vital Statistics, Addis Ababa, 14-19 December 1964. United Nations. Manual X. Indirect Techniques for Demographic Estimation". Population Studies, No. 81. Department of International Economic and Social Affairs. United Nations. New York, N.Y., 1983. - ~-' 43- Results are shown in part (c) of Table 16. It may be noted in passing that the average parities thus obtained from the fertility survey are remarkably similar to the average parities calculated from the 1982 census, indicating that the age-specific rates from the fertility survey for the cohorts considered are highly,consistent with the lifetime fertility reported to the 1982 census. The estimates of child mortality level shown in Table 16 (c) do not agree closely with the stepwise estimates.* The agreement is adequate for the early and mid-1970's, but the indirect estimates for iround 1980 indicate much higher child mortality than either the stepwise process or the 1981 life table; indeed, they suggest an increase in child mortality from 1979 to 1981. In applying the indirect estimation procedure, it is common to find estimates of child mortality based on reports of women aged 15-19 and (to a lesser extent) 20-24 that are heavier, than those based.on reports of women aged 25-34; we believe that selection effects often distort the estimates for younger women. However,'these problems do not generally extend to women aged 25-29, who are also in this application providing high estimates of child mortality relative to the stepwise estimates. It is possible, however, given the high age at first birth in China in the late-1970's and early 1980's--the modal age at first birth in 1980 and 1981 was 24 years-that these selection biases have continued to a higher than normal age, affecting even the proportions dead of women aged 25-29. Thus, these indirect estimates canfirn, the general level of child mortality in the early and mid-1970's, but provide an additional indication that there was little change in child mortality from the mid-1970's onwards, and that the stepwise process is underestimating child mortality for this period. However, the indirect procedure is not immune from error, and could in this instance overestimate recent child mortality if abortions and still-births were included in reports of children ever born and -44- children not surviving. Accepting the indirect estimates would'imply that there was a substantial underestimation of children under age 10 by the 1982 census relative to coverage at other ages, necessary to explain the underestimates of the stepwise process. The group census-taking format of the 1982 census might omit children differentially, though checks with registers were supposedly made. In conclusion, these comparisions with independent sources confirm 'the general trend of child mortality decline from 1964 to the mid-1970's, but indicate heavier child mortality, and a slower rate of mortality decline from 1975 onwards than does the stepwise process. For this recent period, we are unable to draw firm conclusions about child mortality levels and trends. All the data sources used may contain errors, and we have no firm basis for choosing between the various estimates. In what follows, we accept the stepwise estimates because such acceptance maintains consistency between the key sources-the 1982 age distribution, birth registration, and the fertility survey, In the summary section, however, we accept final estimates of infant and child mortality that are more or less consistent with the 1981 life table, and thus a compromise between the stepwise and indirect estimates, though our conclusions for this period must be heavily qualified by admissions of uncertainty. Turning to adult mortality, we can proceed as for the period 1953-64-estimate total intercensal deaths and the deaths before age 10 (using the period mortality levels in Table 15) by fiscal year, then estimate total deaths at ages 10 and over, and then distribute these deaths by fiscal year on the basis of assumed trends in death registration completeneness, both under and over age 10. For the period 1964-82, however, we have two additional pieces of information: deaths by age group from the CES, and the 1981 life -45- table. Rather than use the CES calculated or adjusted life tables, which are based on a suspect population age distribution, we have derived a life table above age 10 directly from the age distribution of deaths using age-specific population growth rates between 1964 and 1982. Deaths by age in any population can be converted into deaths in the stationary or life table population by expanding the deaths by age by the cumulated populati'on growth rates below the age in question. The resulting life table will be correct, so long as the age distribution of deaths and the growth rates are correct; that is, 'regardless of absolute registration coverage, so long as coverage is constant by age. The calculations and derived life table measures--survivorship from age 10 and age-specific mortality rates--are shown- in Table 17. The age-specific mortality rates are then applied to a 1974 age distribution above age 10 for all China obtained by interpolating between the 1964 and 1982 census age distributions to estimate true deaths over age 10 in 1974. Total deaths over age 10 are calculated as 4981 thousand for 1974, for which year child deaths are estifqated as 1999 thousand, and registered deaths totaled 6572 thousand. It is then possible to calculate combinations of death registration completeness below and above age 10 consistent with these values: if deaths over age 10 were completely registered, deaths under age 10 were 79.6 percent covered, whereas if deatis under age 10 were completely registered, deaths over age 10 -ere 91,8 percent covered. We then used the ratio of "true" deaths under age 10 to "true" deaths over age 10, 0.4013, and the corresponding ratio from CES deaths, assumed to be based on registration, 0.3743, to estimate the true combination.of registration completeness under and over age 10. The estimates obtained indicate that deaths over age 10 were 96.0 percent registered, whereas deaths under age 10 were 89.5 percent registered. - 46 - TABLE 17 TL8E 17. Estimation of 1974 Life Table fran Cancer Epideniology Survey Death2 by Age atd Age-specific Grm-th Ratec, 1964-82, and Estimated Deachs over Age 10 in 1974: Qina 1974 Life Table Funcion9 1974 Estimated Age Group Both Sexeg Population r Deachs 5dx l(xIO)/ 0 5m; Mid-Year Deathe x,x+4 1964 1982 5 x 73-75 Population 10-14 86959 132048 .02321 95 100.7 1.0000 .00104 98019 102 15-19 63179 126793 .03870 81 100.2 .9948 .00104 92367 96 20-24 52319 75724 .02054 112 160.7 .9896 X00169 $5051 144 25-29 50804 92761 .03M45 104 170.8 .9813 .00181 61455 111 30-34 47037 73095 .02449 102 193.6 .9725 .00207 50716 105 35-39 41460 54306 .01499 133 278.6 .9625 .00302 49028 148 40-44 35897 48474 .01669 168 380.9 .9480 .00420 45391 191 45-49 31071 47457 .02353 221 554.1 .9283 .00627' 38874 244 50-54 26693 40925 .02374 291 821.1 .8997 .00967 32906 318 55-59 22728 33971 .02233 375 1187.3 .8572 01486 26736 397 60-64 17797 27431 .02404 512 1820.3 .7958 .02515 22182 558 65-69 11718 21306 .03321 559 2293.2 .7017 .03693 16811 621 70-74 7430 14374 .03666 653 3190.1 .5831 .06593 11037 728 75-79 3784 8621 .04575 485 2911.4 .4181 .08786 5977 525 80+* 1826 5058 ,05660 579 5171.6 .2675 .2000 3470 694 total 500702 802344 40262 4470 19334.5 59.67 4981 *xeccatica of life at age 80, e(80) ässumed to be 5.0 years. Unber-- are in chcusande. - 47 - We then apply the 1981 life table to an estimated 1981 population obtained by reverse projecting the 1982 age distribution by one year. Deaths over age 10 total 5282 thousand. Deaths under age 10, taken from the stepwise procedure child mortality estimates. applied to annual births (assuming constant registration completeness of 0.909 from 1976 onwards) total 986 thousand. Registered deaths totaled 6361 thousand in 1981, implying an ..overall completeness of death registration of 101.5 percent. This implied over-registration of deaths in 1981 is a'further indication that the stepwise estimates of child mortality may be too low for the years up to 1982. We assume that death registration was in fact virtually complete in 1981, both below and above age 10. 'Total deaths between 1964 and 1982 are found as the difference between the 1964 poptslation (696,387 thousand) plus intercensal births (447,002 thousand) and the 1982 population (1,008,175 thousand), giving 135,214 thousand deaths, of which 115,670 thousand were registered. Given deaths by fiscal year under 10 from the stepwise process, and assuming that in 1964-65 deaths under 10 were 45 percent registered while deaths over age 10 were 80 percent registered, total deaths by year were found, assuming that registration completeness improved both for child and adult deaths in such a way as to pass through the estimates derived above for 1974 and 1981 and also to give the correct total of intercensal deaths. The results are shown in Table 18. The fiscal year total deaths are shown again in Table 3, where,the population evolution from 1964 to 1982 is obtained from annual births and deaths. The sequence of deaths in Table 18, and death rates in Table 3, look plausible enough, but the year-to-year changes should be interpreted with cautilon, since the annual sequence of completeness estimates.consists of a series of guesses passing through two empirically estimated points. Thus, the - 48 - TABME 18. Regiscered Descb9, Escimared child and Adult Deaths, Registered Desh? and Death Registracicn Ca=piecene2s, ad Total .t:imated Deatbe: OCirma, Fiscal year 1964-82 Deachs under Age 10 Death over Age 10 Total Fisca1 I egisared ,.cimaced Registncic Regi2teed Reire Registr-acin Stiaced Etimated Year Deakths Destby Cmpleteres Death eaths lete5e Derhe machs 1964-65 7508 4197 0.45 1889 3619 0.80 7014 11211 1965-66 6564 3872 0.50 1936 4628 0.82 5636 9508 196&-67 6384 3492 0.55 1921 4463 0.84 5306 8798 1967-68 6335 3359 0.:0 2015 4320 0.86 5016 8375 1965-69 - 6369 3334 0.65 2167 4202 0o.8 476 8102 1969-70 6299 3194 0.70 2236 4063 0.90 4508 7702 1970-71 6143 3041 0.75 2281 3852 0.91 4238 7279 1971-72 6364 2820 0.80 2256 4108 0.92 4458 7278 1972-73 6363 2545 0.83 2112 4251 0.93 4564 7109 1973-74 6342 2175 0.86 1871 4472 0.94 4750 6925 1974-75 6655 1823 0.89 1632 5023 0.95 5288 7111 1975-76 6714 1513 0.91 1377 5337 0.96 5560 7073 1976-77 6643 1373 0.92 1270 5373 0.96 5597 6970 1977-78 6202 1154 0.94 1085 5117 0.97 5276 6430 1978-79 5921 957 0.95 909 5012 0.97 5167 6124 1979-W0 6091 1006 0.96 966 5125 0.98 5230 6236 1980-t 1 6266 920 0.97 892 5374 0.98 5483 6403 1981-62 6507 968 0.98 949 5558 0.99 5615 653 Totil 115670 41743 0.75 29762 85908 0.92 93471 135214 Anesarn iri d)cuzands. - 49 - general trends are probably approximately correct, but may well miss short-term fluctuations. For instance, if mortality over age 10 increased in a particular year while death registration completeness fell, the methodology applied here would not detect the mortality increase. Average post-child mortality for the period 1964-82 can be estimated from intercensal survivorship ratios from 1964 to 1982. Table 19 shows the survivorship ratios and the implied "West" mortality levels. Intercensal survivorship ratios for long intervals such as that between 1964 and 1982 under conditions of changing mortality tend to overweight period mortality levels for which the cohort death rates are highest. Thus, the survivorship ratio for the cohort aged 0-4 in 1964 will overweight mortality levels early in the intercensal period, when most of the cohort deaths occurred, whereas the ratios for cohorts aged 50 and over in 1964 will overweight mortality levels toward the end of the period when most of the cohort deaths were occurring. The most robust estimat es of average intercensal mortality levels should thus be obtained from the survivorship ratios for the central range of ages, where mortality rates are not changing rapidly with age. For cohorts aged 10-49 in 1964, the average "West" mortality level for males is 20.1, implying an expectation of life at age 10 of 58.6 years; comparable figures for females are a "West" level of 19.3 and an expectation of life at age 10 of 60.5 years. The 1974 life table derived in lable 17 indicates an expectation of life at age 10 for both sexes of 59.7 years, very close to the both sexes average of 59.5 years obtained from intercensal survival. Two other features of Table 19 may be noted. First, the mortality levels for males over age 45 in 1964 fall quite sharply with age, indicating higher old-age male mortality than would be exected on the basis of young and middle-age mortality; females show a similar, though less pronounced, pattern. Second, the survivorship -50 - ZMU 19. Intcrcen!al bbrtalicy Estimaec by Sex ftc Ccorr Survival; Chirw, 1964-62 Maleg F~ales Chort Cebam Pomulatiom SvivoriWip t Cchorr. Pculation Survivorhip Age, 1964 1964 1982 Racio Vortatity, tvel 1964 1982 Ratio Mralicy Level 0-4 51814 49373 .9529 18.1 49033 46383 .9460 16.9 5-9 49621 46878 .97 16.1 45219 43511 .9622 17.8 10-14 45314 43178 .9529 18.6 41645 4036 .9688 19.6 15-19 33205 31523 .9493 19.3 29974 28533 .9526 18.7 20-24 27805 26309 .9462 19.8 24514 23125 .9433 18.7 25-29 27036 25315 .9363 20.1 23768 22293 .9379 19.3 30-34 24891 23373 .9390 22.0 22146 2070. .9348 20.4 35-39 21739 19226 .8844 20.7 19721 17580 .3914 19.0 40-44 18565 15497 .3347 21.0 17332 15017 8664 20.1 45-49 15819 11342 .7170 19.0 15z52 11926 .7819 1.7 50-54 13384 7943 .5935 18.3 -13309 9136 .6863 18.8 55-59 10823 4627 -4275 16.6 11905 6297 .5289 17.6 Averge Level 10-49 20.1 19.3 - 51 - ratios for males and females aged 5-9 in 1964 are lower than expected, suggesting either high adolescent mortality relative to mortality at other ages, an over-enumeration of 5-9 year-olds in 1964, or an under-enumeration of 23-27 year-olds in 1982, relative to overall enumeration completeness. We found no clear indication of relative over-enumeration of 5-9 year-olds in 1964, so the survivorship ratios suggest that there may have been some under-enumeration of young adults of both sexes in 1982. Such a relative under-enumeration might hel'p explain the problems with the stepwise child mortality estimation procedure, since too few women would result in too few estimated births on the basis of the 1982 survey fertility rates, and thus survivorship ratios of births to the 1982 census that were too high. The procedure used to obtain rough estimates of expectation of life at birth and at age 10 by fiscal year for the period 1953-64 cannot be aonlied to the period 1964-82. As the intercensal growth rates in Table 17 show, the population over age 10 was by no means stable over the period--age-specific growth rates vary sharply by age-and the death rates over age 10 by fiscal year are affected by the changing age structure, with the effects of falling mortality being attenuated by a gradual aging of the population. We do, however, have estimates of expectation of life at birth and age 10 from the 1981 life table, and of expectation of life for both sexes at age 10 from the 1974 life table; this latter value can be used as a basis for estimating both expectation of life at age 10 by sex, assuming the male-female differential implied by the intercensal survival ratios in Table 19, and expectatioa of life at birth by sex, adding in the average of the stepwise child mortality estimates for 1973-74 and 1975-75. Our estimates of life expectancy for 1974 and 1981 are given in Table 20. 52 - TABLU 20. Zxpecacion of Lie at BLth ard at Age 10 by Sex; China, 1974 and L981 Year Males Femles e(o) e(10) e(o) e(10) L974 63.6 58.8 65.1 60.6 1981 66.4 60.4 68.3* 63.4 * Adjuced for estimated effects of ecess infant fezale &aths. Pouulation Age Distribution, Mid-1980 For population projection purposes, it is useful to have an age distribution for mid-year 1980 as a starting point for projections by standard five-year periods. Table 21 provides such an age distribution, obtained by reverse-projecting the 1982 age distribution (after adjustment for military-age males and redistribution of the population under age six to conform to the assumption of constant birth registration completeness from L976 to 1982) using the 1981 life table. TABLE ZL Ptulation by Age wid Sex, Mid-1980, Ci.-a Males Famles 0-4 48W47 45384 5-9 64030 60073 10-14 66060 62201 S-1%9 5629 51489 X-Z4 44351 41150 25-29 45554 4762 30-34 34150 30972 35-39 27061 23895 25654 22A26 45-49 24101 21365 50-54 20470 18427 55-59 16831 13984 60-64 12876 13080 65-69 9356 IM78 70-74 5934 7572 75-79 3052 4581 1368 2565 Total 504125 474209 1877F