6
Adult Mortality

Ian M.Timœus

INTRODUCTION

The aim of this chapter is to provide a largely descriptive account of levels, trends, and patterns of adult mortality in sub-Saharan Africa. A very incomplete picture emerges. Both the coverage and the accuracy of the data on which the chapter draws are far more limited than those for other components of African population dynamics such as child mortality or fertility. No data at all exist for most of the population of the region, and representative information on causes of death in adulthood is available from only a handful of studies of relatively small populations. Furthermore, very little information about the effect of the AIDS epidemic on national populations is yet available.

The shortage of data on adult mortality in sub-Saharan Africa largely reflects the inadequacy of vital registration systems, combined with the technical limitations of the methods that can be used to investigate the subject retrospectively (Timæus, 1991a). In addition, a substantial proportion of what we know about the demography of Africa derives from fertility surveys or from longitudinal studies mounted to investigate child health. Both types of inquiry usually cover too small a sample to be used to mea-

Ian Timæus is at the Centre for Population Studies, London School of Hygiene and Tropical Medicine



The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement



Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.

OCR for page 218
Demographic Change in Sub-Saharan Africa 6 Adult Mortality Ian M.Timœus INTRODUCTION The aim of this chapter is to provide a largely descriptive account of levels, trends, and patterns of adult mortality in sub-Saharan Africa. A very incomplete picture emerges. Both the coverage and the accuracy of the data on which the chapter draws are far more limited than those for other components of African population dynamics such as child mortality or fertility. No data at all exist for most of the population of the region, and representative information on causes of death in adulthood is available from only a handful of studies of relatively small populations. Furthermore, very little information about the effect of the AIDS epidemic on national populations is yet available. The shortage of data on adult mortality in sub-Saharan Africa largely reflects the inadequacy of vital registration systems, combined with the technical limitations of the methods that can be used to investigate the subject retrospectively (Timæus, 1991a). In addition, a substantial proportion of what we know about the demography of Africa derives from fertility surveys or from longitudinal studies mounted to investigate child health. Both types of inquiry usually cover too small a sample to be used to mea- Ian Timæus is at the Centre for Population Studies, London School of Hygiene and Tropical Medicine

OCR for page 218
Demographic Change in Sub-Saharan Africa sure adult mortality directly, and many of them have failed to collect data that can be used to estimate adult mortality indirectly. Adults tend to die of different diseases from children, and evidence has accumulated that neither the level of—nor the trend in—adult mortality is closely associated with child mortality (Murray et al., 1992). Thus, it is necessary to investigate adult mortality in Africa to provide a sound basis for the population estimates and projections that underlie planning in any sector. Evidence that adult mortality has stagnated at a high level in some countries (e.g., Timæus, 1984) further suggests that adult ill health is not an issue that can be left to look after itself. Equity and efficiency in health planning require consideration of the health problems of the poor throughout their lives. Against this background, this chapter attempts, to examine adult mortality in sub-Saharan Africa and the extent of the public health problem that such mortality poses. Given the limited knowledge of the subject, three questions seem of central interest. First, how does adult mortality in sub-Saharan Africa compare with other continents? Second, how does the mortality of adults compare with that of children? Third, are there distinctive variations in adult mortality patterns across Africa that raise issues for further investigation? SOURCES OF DATA The difficulties involved in measuring adult mortality in developing countries and the deficiencies of the data available on sub-Saharan Africa are well known to most demographers. The only mainland country south of the Sahara that has a civil registration system with sufficient deaths reported for it to be possible to use the national data to estimate adult mortality is South Africa. Even in this country the statistics exclude the four nominally independent “homelands” and are incomplete (though amenable to adjustment) for that part of the rest of the population that is not “White.”1 Although registration of adult deaths is more complete locally, particularly in some major cities, few such data have been published. Thus, nearly all of the estimates of adult mortality that can be made for sub-Saharan Africa derive from census or survey data. Useful information can be culled from a variety of sources, including national censuses, the Population Growth Estimation surveys of the early 1970s, the World Fertility Surveys (WFS), and the Demographic and Health Surveys (DHS), as well as other inquiries mounted by the statistical organi- 1   “White,” “Colored,” “Asian,” and “Black” are legal statuses defined in South African apartheid legislation and used in South African official statistics.

OCR for page 218
Demographic Change in Sub-Saharan Africa zations of particular countries. Unfortunately, there has been little growth in the availability of reasonably up-to-date information on adult mortality in sub-Saharan Africa during the last 15 years. Information from the 1980 round of censuses and from Phase I of the DHS is available for about the same number of countries as from the 1970 round of censuses and the WFS. Moreover, although the core questionnaire for Phase I of the DHS program included pertinent questions about survival of parents, these have been left out of the Phase II questionnaire. Similarly, a few countries that collected data about adult mortality in their 1980-round census failed to do so in their 1990-round census. Such decisions not to build on earlier efforts are regrettable. Even if the data on adult mortality obtained in an initial inquiry proved difficult to interpret, those on orphanhood, in particular, become much more useful once they have been collected repeatedly. A second limitation of the available information is that even in those countries that collect adult mortality data in their censuses, processing and publication of the results often take an inordinate length of time. Almost no information is available from the 1990 round of censuses at present, and some data collected in the 1980 round of censuses, including the orphanhood data from the 1985 census of Sierra Leone and 1986 census of Lesotho, could not be obtained for this study. Thus, although data have been collected to update—by a decade—many of the estimates presented here for the mid-1970s, it may be impossible to do the actual updating for several more years. Lack of data means that it is impossible to arrive at well-founded national estimates of adult mortality in many countries for any point during the 1970s or 1980s. All the different forms of information that can be used to measure adult mortality were considered during the preparation of this chapter, and some of the estimates presented are based on very inadequate data. Even with this catholic approach, the results cover only 24 countries and exclude several populous nations such as Nigeria, Ethiopia, and Zaire. In total, they refer to only about 40 percent of sub-Saharan Africa’s population. METHODS OF ANALYSIS The methods that can be used to estimate adult mortality in the absence of adequate vital registration have been reviewed in detail recently (Timæus, 1991a) and are discussed only briefly here. Examples illustrating these methods are supplied in the appendix to this chapter. The main source of direct estimates is retrospective questions about recent deaths in the household posed in censuses and single-round surveys. Such data are the primary source of information used to produce estimates for nine of the countries listed in Table 6–1: Mali, Togo, Cameroon, Congo,

OCR for page 218
Demographic Change in Sub-Saharan Africa Madagascar, Rwanda, Tanzania, Botswana, and Lesotho. Such questions have been included in many more national inquiries, but the results have frequently been disappointing. Often it is manifest that only a small minority of deaths have been reported. The data appear complete in only a few instances. Nevertheless, a range of techniques exists that can be used to assess such data and, in favorable circumstances, to adjust them for underreporting. The crucial assumption is that the degree of underreporting is the same at all adult ages. Of these techniques, those proposed by Brass (1975), Preston et al. (1980), and Bennett and Horiuchi (1981) were used to evaluate the data presented in this chapter. On this basis, reporting is taken as complete in six countries.2 In Mali, only about half the deaths, and in Togo slightly more than half, were reported. In Congo, only about one-third of men’s deaths and one-fifth of women’s deaths seem to have been reported. A second source of direct estimates of adult mortality is the multiround demographic surveys that have been conducted in several sub-Saharan countries. The results of many of these studies are now rather out-of-date, but they are the main sources used for three of the countries in Table 6–1—Côte d’Ivoire, Liberia, and Senegal. In all three, reporting has been accepted as complete.3 An additional source of adult mortality estimates for Africa is the information obtained from retrospective questions about the survival of respondents’ mothers, fathers (Brass and Hill, 1973), and spouses (Hill, 1977). Information on orphanhood has been collected more frequently than that on widowhood and has usually yielded better results. The estimates that result are somewhat out-of-date and, like direct estimates, can be biased by reporting errors. Such data are the primary basis for the remaining 11 sets of national estimates presented here (Benin, The Gambia, Ghana, Mauritania, Sierra Leone, Burundi, Kenya, Malawi, northern Sudan, Zimbabwe, and Swaziland) and provide important support for estimates based on recent deaths in five other countries, namely, Mali, Cameroon, Congo, Tanzania, and Lesotho. The basic orphanhood estimates were calculated by using the variant of the method proposed by Timæus (1992). Brass and Bamgboye’s (1981) method was used to determine the time reference of the results. 2   Some of these data exhibit signs of having been adjusted already, without it being documented in the sources from which they were obtained for this study. 3   Only summary measures of e15—the number of years that a person who survives to age 15 can expect to live—have been published for Senegal, so reevaluation of the data was impossible and 45p15—the probability of surviving from exact age 15 to exact age 60—had to be inferred from a model life table. It has been suggested that, if anything, the level of adult mortality is overstated (Cantrelle et al., 1986). In Côte d’Ivoire, the original survey study (Ahonzo et al., 1984) concluded that the reports of adult deaths were incomplete. My reevaluation of the results leads to a different conclusion.

OCR for page 218
Demographic Change in Sub-Saharan Africa TABLE 6–1 Survivorship from Age 15 to Age 60 by Sex, 1970s and 1980s   45P15   Region and Country Date Males Females Both Sexes Reliability Source Western   Benin 1978 .749 .779 .764 Very low Orphanhood Côte d’Ivoire 1978–1979 .646 .741 .694 Fair Multiround survey The Gambia 1978 .773 .812 .793 Fair Intercensal orphanhood Ghana 1982 .778 .880 .830 Low Orphanhood since marriage Liberia 1970–1971 .550 .584 .567 Fair Multiround survey Mali 1986 .579 .541 .560 Low Recent deaths and orphanhood Mauritania 1980 .782 .823 .803 Fair Orphanhood and recent deaths Senegal 1978 .652 .710 .682 Fair Multiround surveys Sierra Leone 1974 .466 .510 .488 Very low Orphanhood Togo 1981 .704 .760 .733 Very low Recent deaths Middle   Cameroon 1976 .644 .666 .654 Low Recent deaths and orphanhood Congo 1984 .656 .703 .680 Very low Recent deaths and orphanhood Eastern   Burundi 1981 .622 .699 .661 Low Orphanhood since marriage Kenya 1974 .714 .769 .742 Fair Intercensal orphanhood Madagascar 1974–1975 .487 .551 .518 Very low Recent deaths Malawi 1977 .741 .706 .723 Low Intersurvey orphanhood Rwanda 1978 .584 .629 .607 Low Recent deaths Tanzania 1988 .656 .675 .666 Very low Recent deaths and orphanhood Zimbabwe 1978 .801 .863 .833 Very low Orphanhood

OCR for page 218
Demographic Change in Sub-Saharan Africa Southern   Botswana 1980 .555 .732 .646 Low Recent deaths Lesotho 1976 .503 .749 .627 Fair Recent deaths and orphanhood South Africa 1985 .638 .766 .702 Fair Vital registration Swaziland 1981 .561 .761 .663 Fair Intercensal orphanhood Northern   Sudan (northern) 1975 .695 .768 .732 Fair Orphanhood and widowhood   SOURCES: Data from Benin (Benin, n.d.); Côte d’Ivoire (Ahonzo et al., 1984); The Gambia (Blacker and Mukiza-Gapere, 1988); Ghana and Senegal (Timæus, 1991d); Liberia, Madagascar and Rwanda (Waltisperger and Rabetsitonta, 1988b); Mali (Mali, 1980 and provisional 1987 census tables); Mauritania (Timæus, 1987); Sierra Leone (Okoye, 1980); Togo (Togo, 1985); Cameroon (Cameroon, 1978, 1983); Congo (Congo, 1978, 1987); Burundi (Timæus, 1991c); Kenya (Mukiza-Gapere, 1989); Malawi (Timæus, 1991b); northern Sudan (Sudan, 1982); Tanzania (Tanzania, 1982 and provisional 1988 census tables); Zimbabwe (Zimbabwe, 1985); Botswana (Botswana, 1972, 1983); Lesotho (Timæus, 1984); South Africa (South Africa, 1988); Swaziland (Swaziland, 1980 and unpublished 1986 census tables).

OCR for page 218
Demographic Change in Sub-Saharan Africa In countries where more than one set of orphanhood data exists, it is possible to calculate more recent and reliable estimates by basing them on changes in orphanhood in adulthood between two inquiries (Timæus, 1991b). This approach was adopted in four countries: The Gambia, Kenya, Malawi, and Swaziland. In addition, several recent DHS surveys have asked whether deceased parents died before or after the respondent married. This information can also be used to produce more recent and reliable results than are obtained from the basic orphanhood method (Timæus, 1991c). The technique was used to produce the estimates for Ghana and Burundi, and to confirm the results of the multiround surveys in Senegal. One set of techniques for estimating adult mortality that could not be used to produce any estimates represents those based on intercensal survival and growth. In an attempt to increase the scope of the results, the integrated intercensal growth method (Preston, 1983) was applied to data from several countries. It yielded very erratic and implausible results. As one might expect, poor age reporting, high levels of international migration, and changes in census coverage combine to render such techniques useless in most African countries. Estimating adult mortality from African data involves a large element of judgment. Most of the results presented are based on data that have been subjected to some form of smoothing and adjustment. Decisions not to adjust certain data are to some extent arbitrary.4 Moreover, on the principle that estimates that are largely guesses are to be preferred to those that are complete guesses, results have been presented even for countries where the data are very difficult to interpret. Thus, none of the results is definitely accurate. The estimates obtained from the registration data for South Africa, multiround surveys, and several different sources of data that yield reasonably consistent results are assumed to be fairly reliable. The estimates are judged of low reliability in countries where different sources of data yield less consistent results; where only a single set of recent data on deaths exists, but it appears reasonably complete; or where the partitioning of the data on orphanhood into deaths before and after the respondent’s marriage provides a partial check on the results. Finally, the reliability of the estimates for countries in which only a single set of incomplete recent death data or orphanhood data is available must be considered very low. This category of estimates also includes those for Congo and Tanzania. Several sources of data exist for these two countries, but they yield inconsistent results and are difficult to interpret. 4   The mortality rates obtained from direct data on adult deaths were smoothed by fitting a two-parameter logit model life table based on the general standard (Brass, 1971) to observed survivorship from age 15. Time series of indirect estimates were smoothed by fitting regression lines, excluding any obviously discrepant points for extreme age groups.

OCR for page 218
Demographic Change in Sub-Saharan Africa ADULT MORTALITY LEVELS Estimates of adult mortality in the 24 countries for which information is available are shown in Table 6–1. Only data for 1970 or later are presented and the figures are the most recent available. Four of the estimates refer to the first half of the 1970s, ten to the late 1970s, and ten to the 1980s. The index presented is life table survivorship from exact age 15 years to exact age 60 years (45p15). It can be interpreted as the probability of surviving to old age, subject to surviving childhood, at the prevailing level of mortality. Unlike measures of life expectancy, survivorship from 15 to 60 years can be calculated without making assumptions about old-age mortality. It is the complement of the probability of dying between ages 15 and 60 (45q15), which has been adopted in a recent World Bank volume as the preferred index of adult mortality (Feachem et al., 1992). Perhaps the most striking feature of the results shown in Table 6–1 is that large differences exist in the level of adult mortality between different African countries. These estimates suggest that in Benin and The Gambia, more than 75 percent of those aged 15 would survive to their 60th birthday at the levels of mortality prevailing around 1980. In Ghana, Mauritania, and Zimbabwe, the equivalent figure is more than 80 percent. These figures represent moderate adult mortality by world standards. For example, survivorship from 15 to 60 years in Sri Lanka or in Trinidad and Tobago is similar to that in these lower-mortality countries of sub-Saharan Africa. Moreover, adult male survivorship is much lower in several East European countries. The apparently low level of adult mortality in these countries is somewhat surprising. Although it is possible that the orphanhood method has produced underestimates of adult mortality in all of these countries, the results for The Gambia, at least (see appendix), and Ghana (Timaeus, 1991d) exhibit a high degree of internal consistency. In Zimbabwe, adjusted registration data for Harare yield an estimate of adult survivorship for 1982–1986 that is almost identical to that in Table 6–1 (Moyo, 1991). Thus, the national estimate of survivorship may be somewhat too high but is unlikely to be grossly inaccurate. In contrast, in many African countries for which we have data, adult mortality is high. The estimated probability of surviving from age 15 to age 60 falls to less than 70 percent in half of these countries, and to less than 60 percent in Liberia, Sierra Leone, Mali, and Madagascar. Except in Mali, the highest estimates refer to the 1970s. They seem plausible: Some intensive, longitudinal studies of rural populations in Africa have documented even more severe levels of adult mortality. For example, in Bandafassi, Senegal, the probability of dying between 15 and 60 years of age still exceeded 50 percent during the 1970s (Pison and Langaney, 1985). Such elevated levels of adult mortality have been eliminated in most other parts

OCR for page 218
Demographic Change in Sub-Saharan Africa of the world since the Second World War. World Bank estimates suggest that outside sub-Saharan Africa, 45q15 exceeds 30 percent in only nine countries (Feachem et al., 1992). In few, if any, countries outside sub-Saharan Africa is 45q15 as high as 40 percent. These estimates of adult mortality do not follow an obvious regional pattern. Hill (1991) and Blacker (1991) suggest that child mortality is higher in western than eastern Africa. An earlier study, based on fewer countries, suggested that this differential was also true of adult mortality (Timæus, 1991e). However, the estimates in Table 6–1, apart from those for the four southern African countries, which form a fairly homogeneous group, suggest that high adult mortality and low adult mortality are to be found on both the western and the eastern sides of the continent. This change in interpretation reflects, in part, differing definitions of the regions used and, in part, incorporation into the analysis of additional data that support a gloomier view of eastern African mortality. In addition, evidence presented later in this chapter suggests that the disappearance of this broad geographical contrast to some extent reflects differential mortality trends: The estimates presented here are centered on the late 1970s, about six years later than those in the earlier study. Gender Differentials There are extensive data indicating that gender differentials in child mortality in sub-Saharan Africa are small but usually favor girls slightly (e.g., Rutstein, 1984). Less is known about gender differentials in adult mortality. The subject is of particular interest because of the high maternal mortality that has been documented recently in many parts of the continent (Graham, 1991). Figure 6–1 compares the male and female estimates from Table 6–1. Given the range of different data sources and methods used to make these estimates, the results are remarkably consistent.5 They suggest that female survivorship in adulthood is slightly higher than male survivorship in most of sub-Saharan Africa. In the majority of this sample of countries, the gender differential in adult mortality is similar to that found in most other parts of the developing world and in the United Nations’ (1982a) Latin American and general families of model life tables and the four families of Princeton model life tables (Coale and Demeny, 1983). Only in two countries is there any evidence of excess female adult 5   When making the estimates, I attempted to consider each set of data for men and women on its own merits and not to look for patterns in the results until they were all available. Despite this precaution, my preconceptions have probably influenced somewhat the results shown in Figure 6–1.

OCR for page 218
Demographic Change in Sub-Saharan Africa FIGURE 6–1 Gender differentials in survivorship from age 15 to age 60. SOURCE: Table 1. mortality. The country with higher mortality is Mali. The estimates used here are based on census reports of recent deaths, and it is possible that the adjustments made for underreporting have led to the overestimation of female mortality. Orphanhood data collected at the same time suggest a more usual gender differential in adult mortality in Mali. The other country that apparently has excess female mortality in adulthood is Malawi. There is stronger evidence that this country is genuinely anomalous because both the results of the 1970–1971 multiround survey and the more recent orphanhood estimates shown in Table 6–1 support this conclusion. The four southern African countries included in Figure 6–1 stand out as having high mortality among adult men. The differential is somewhat attenuated in South Africa due to the more usual mortality patterns among the “White” population. Among the “Black” majority, however, the differential between male and female mortality is as large as in Swaziland or Botswana. These southern African countries are exceptional, not just within sub-Saharan Africa, but compared with the rest of the world. The absolute difference between adult male and female death rates in Lesotho is probably larger than in any other national population. The high mortality of adult men in southern Africa is almost certainly related to the importance of labor migration, particularly to the mines, in the economy of this part of the continent. Mining is in itself a hazardous occupation; the lifestyle associated with prolonged absences from home encourages heavy smoking and drinking; and the region suffers from an “epidemic” of tuberculosis that originated in the mining industry (Packard, 1990).

OCR for page 218
Demographic Change in Sub-Saharan Africa FIGURE 6–2 Survivorship in childhood and adulthood. SOURCES: Sources cited in footnote 6. Age Patterns of Mortality Figure 6–2 compares the estimates of adult survivorship from Table 6–1 with corresponding estimates of the probability of surviving to age 5, l(5).6 In many sub-Saharan African countries the relationship between the overall levels of child and adult mortality falls within the range of experience of the developed world during its mortality transition as encapsulated in the Coale and Demeny (1983) regional model life tables. In particular, in all the countries with high adult mortality, the relationship between survivorship to age 5 and survivorship from 15 to 60 years lies between those in the Coale and Demeny South and West families of models. In contrast, the five countries in which adult survivorship had risen to more than 75 percent by the end of the 1970s still had much higher child mortality than one would expect based on the Coale-Demeny models. The pattern is most extreme in The Gambia, where the estimate of adult survivorship is probably somewhat too high. However, as the section on mortality levels argues, it seems unlikely that adult survivorship has been overestimated in all five countries by as much as the 5 to 10 percentage points required to 6   For most countries the estimates of child survivorship made by the United Nations (1988) have been used, supplemented by those of Hill (1991). In a few countries, estimates based on information that has become available recently are used but no systematic search for such data was attempted.

OCR for page 218
Demographic Change in Sub-Saharan Africa APPENDIX As the body of this chapter emphasizes, only limited and defective data on adult mortality exist for sub-Saharan Africa. Analysis of this information is far from straightforward. To illustrate the procedures involved and the difficulties that arise in interpretation of the results, this appendix presents applications of various methods of estimation for some of the data analyzed in this chapter and explains how the summary indices presented in Table 6–1 were derived. The first important form of data is on deaths by age during a specified period. Such data are collected by registration systems or by asking direct questions in censuses and single- or multiround household surveys. Reporting may be incomplete or, with retrospectively collected data, may refer to the wrong reference period. In addition, inaccurate reporting of either ages at death or ages of the living population used as a denominator to calculate rates may bias the estimates. Several methods exist for evaluating such data. Most of them are based on the relationship between the number of people of any age and the number of deaths above that age. If one adjusts for population growth, these numbers should be the same because nobody can live forever. To exploit this identity, one has to assume that reporting is equally incomplete at all adult ages. On this basis, various approaches can be used to compare the two quantities and assess the shortfall in reported deaths (Brass, 1975; Preston et al., 1980; Bennett and Horiuchi, 1981). Sampling errors, deviations from the assumption of constant underreporting of deaths by age, the distorting effect of past mortality change and migration on the population’s age structure, and age misreporting can all make such comparisons difficult to interpret. Such methods of evaluation have been applied to all the direct data used in this chapter for which sufficiently detailed tabulations are available. Figure 6–A.1 illustrates the application of Preston et al.’s (1980) method to registration data on male deaths among “Blacks” in South Africa in 1985. It compares the estimated population between each age and age 75, calculated from deaths adjusted for growth at 2.6 percent per annum, with the enumerated population in the same age range according to the 1985 population census. If all the assumptions held and reporting was perfectly accurate, apart from being incomplete, the plot would be a horizontal straight line. Although it is not, the assumptions involved appear to have held up reasonably well until about age 55. Thus, the analysis provides rather convincing evidence that relative to the census, only about 72 percent of adult male deaths were reported. In fact, in absolute terms, the registration of deaths of “Black” men is certainly worse than this because the 1985 census enumeration was also incomplete. It is, however, only relative in-

OCR for page 218
Demographic Change in Sub-Saharan Africa FIGURE 6–A.1 Ratio of estimated to reported population, “Black” men, South Africa. SOURCE: South Africa (1988). consistencies between the two sources that bias the estimated death rates. A similar analysis for “Black” South African women suggests that relative to the census, about 59 percent of adult female deaths were registered in 1985. Once the degree to which reports of adult deaths are incomplete is known, it is a straightforward matter to adjust them upward before calculating age-specific death rates and life table measures of mortality and survivorship. Besides the problem of incomplete death reporting, life tables calculated from developing country data tend to be distorted by age reporting, sampling, and other errors. To reduce the effect of such problems, the life tables were smoothed prior to extracting the final estimates of 45p15. The approach adopted is to fit a two-parameter logit relational model life table based on the general standard (Brass, 1971). These models exploit the fact that after the logistic transformation is applied, the differences between any two life table survivorship functions, l(x), are approximately linear. Thus, plotting the logits, Y(x), of the observed survivorship function against those of the standard, Ys(x), one can fit a line to the points for ages at which the data seem reliable and extrapolate from them to age ranges in which the observed data are clearly biased. Figure 6–A.2 illustrates the procedure

OCR for page 218
Demographic Change in Sub-Saharan Africa with data on “Black” female deaths in South Africa in 1985. The radix of the life table is at age 15 years, and reported deaths were multiplied by a factor of 1.695 before calculating death rates and measures of survivorship. Inspection of Figure 6–A.2 reveals that the line representing the observed data bends downward for the oldest ages on the right-hand side of the plot. This deviation indicates that the reported mortality of elderly women is lower than one would expect based on the data for younger age groups and the age pattern of mortality in the general standard. It is likely that reporting of deaths of elderly women is particularly incomplete or that their ages at death tend to be exaggerated. Therefore, the last three points were discarded and a model life table was fit to the points for ages 20 to 60 years. This procedure yields a line with an intercept of –0.34, implying an overall level of adult mortality among “Black” women in South Africa that is a little lower than in the standard, and a slope of 1.26, implying relatively heavy mortality in middle age compared with early adulthood. The prob- FIGURE 6–A.2 Logit survivorship, compared with standard, “Black” women, South Africa. SOURCE: Figure 6–A.1.

OCR for page 218
Demographic Change in Sub-Saharan Africa ability of surviving from 15 to 60 years (45p15) is 70.4 percent in the fitted life table, as opposed to 70.8 percent according to the unsmoothed data. The final estimates for South Africa presented in Table 6–1 are based on similar analyses for the “Colored” and “Asian” populations. These were combined with the unadjusted life table for the “White” population, weighting by each group’s share in the total population. Because the official statistics do not cover the four so-called independent homelands, the mortality of the adult population of these areas was assumed to be the same as that of the “Black” population in the rest of South Africa. This assumption may be unjustified but is unlikely to bias the national estimates appreciably. The second important form of information used to estimate adult mortality is that on the survival of respondents’ mothers and fathers. The questions required are simply, Is your mother alive? and Is your father alive? They can be administered in censuses and single-round household surveys. No attempt is made to collect the ages at death of parents. Instead, these ages are implicitly assumed from respondents’ ages by using demographic models. The information used is the proportion of respondents in each 5-year age group with a living mother (or father). This proportion is closely related to the life table probability of surviving from about the average age at which women (or men) have children to that age plus the current age of the respondents. By using this fact, regression-based procedures have been developed for estimating life table survivorship from data on orphanhood (e.g., Timæus, 1992). Each 5-year age group yields a separate estimate of adult mortality. Although the estimates refer to different age ranges, they can all be converted to 45p15 by using any one-parameter system of model life tables, without reducing their precision greatly. The younger the age group of respondents, the more recently did their parents die, on average. If it is assumed that the level of mortality has changed steadily, it becomes possible to estimate the date at which the mortality of the cohort of parents reported on by an age group of respondents equaled the level of mortality prevailing in the population (e.g., Brass and Bamgboye, 1981). In this way, the series of estimates obtained from respondents aged 5 to 55 years in a single inquiry can be used to infer the past trend in the level of adult mortality over a decade or more. The kind of results obtained from the orphanhood method of estimating adult mortality are illustrated in Figure 6–A.3 by using data from Swaziland. Questions about the survival of mothers and of fathers were asked in both the 1976 and the 1986 censuses, yielding the four lines shown in the figure. Swaziland illustrates a common problem with estimates of adult mortality obtained from orphanhood data. Each of the four sets of results suggests that adult mortality is declining. Because the questions have been asked twice, however, it is possible to compare estimates from the two sources for the early 1970s. They are clearly inconsistent. The reports of younger

OCR for page 218
Demographic Change in Sub-Saharan Africa FIGURE 6–A.3 Adult survivorship estimated from orphanhood, Swaziland. SOURCE: Data from Swaziland (1980) and unpublished 1986 census tables. respondents in the earlier census indicate lower mortality than the reports of older respondents in the later census. Similar inconsistencies characterize the data on several other eastern African countries that have asked about orphanhood more than once. They include Kenya, Malawi, northern Sudan, and Tanzania. Elsewhere in Africa, however, successive surveys have yielded much more consistent results, permitting one to have reasonable confidence in the accuracy of the data. Results for one such country, The Gambia, are shown in Figure 6–A.4 (only maternal orphanhood data are available from both censuses). Others include Cameroon, Lesotho, and Mauritania. Accumulated experience makes it clear that when such inconsistencies between two sets of data arise, they stem from underreporting of orphanhood at early ages (e.g., Timæus, 1991a,b). Data supplied about young orphans often refer to a foster parent or stepparent, rather than to a dead natural parent. Each set of estimates exaggerates the decline in mortality, and the most recent estimates of adult mortality from both inquiries are probably too low. In Swaziland at least, it remains likely that adult mortality has declined somewhat. Every age group of respondents reported more living parents in

OCR for page 218
Demographic Change in Sub-Saharan Africa FIGURE 6–A.4 Adult female survivorship estimated from orphanhood, The Gambia. SOURCE: Blacker and Mukiza-Gapere (1988). 1986 than in 1976. If questions about orphanhood have been asked more than once, the effect of underreporting of orphanhood in childhood can be reduced by analyzing data on the incidence of orphanhood between the two inquiries among young adults. This analysis is done by using the two sets of proportions to construct a synthetic cohort based at age 20. Life table survivorship is estimated from the proportion of each age group in this synthetic cohort with a living mother (or father) by using regression coefficients developed for the purpose (Timæus, 1991b). All the estimates refer to the same intersurvey period, but they tend to differ somewhat because of sampling and age reporting errors. Averaging the results for different age groups, produces point estimates such as those shown in Figures 6–A.3 and 6–A.4. In The Gambia, the synthetic cohort estimate for the 1973–1983 intercensal period emphasizes the consistency of the two sets of orphanhood results. In contrast, in Swaziland the intercensal estimates for both men and women indicate much higher mortality than estimates for 1981 made from 1986 data on children. They do, however, fall in line with the earlier estimates from each census, which are obtained from respondents aged 25 to 45 years. These results suggest that a fairly slow decline in adult mortality has oc-

OCR for page 218
Demographic Change in Sub-Saharan Africa curred in Swaziland. Moreover, the trends for men and women are fairly consistent. Thus the intercensal results seem plausible. The results from applying this approach to Kenya and Malawi are similar to those for Swaziland (Timæus, 1991b). Therefore, intercensal and intersurvey orphanhood estimates of 45p15 were also adopted for these countries. Unfortunately, in northern Sudan and Tanzania the method worked less well, probably because the accuracy of the reports changed markedly between inquiries. In several countries the only information available on adult mortality is a single set of orphanhood data. In Burundi, Ghana, and Senegal, such data were obtained in a DHS survey that also asked women whether they were orphaned before or after they were first married. Data from a single survey on orphanhood since first marriage share the advantages of synthetic cohort data computed for countries that have asked about orphanhood repeatedly. They reflect the recent incidence of orphanhood among young adults and are unaffected by underreporting of orphanhood in childhood. The information analyzed for each five-year age group is the proportion of women with a living mother (or father) among those women whose mother (or father) was alive when the woman first married. Regression methods exist for estimating life table survivorship from these data (Timæus, 1991c). The results for different age groups refer to similar dates. Therefore, like synthetic cohort data, they are usually averaged to produce a single recent estimate. In the three countries with this form of data, underreporting of orphanhood in childhood does not seem to be a problem. Estimates from orphanhood since marriage are consistent with the trend estimated from lifetime orphanhood but are more up-to-date (see Timæus, 1991c,d). They are judged moderately reliable and presented in Table 6–1. Finally, in several countries, only one set of orphanhood data was available for this study and the supplementary question on whether respondents were orphaned before or after marriage was not asked. In some of them, useful data on recent deaths were available, but in Benin, Sierra Leone, and Zimbabwe they were not. This circumstance makes it impossible to determine whether orphanhood among young children is underreported or not. Thus, the most recent estimates cannot be trusted. In such countries, most significance was accorded to the results obtained from respondents aged 15 to 40 years. Data about children were discarded if they suggested an accelerating decline in the level of adult mortality. Similarly, discrepant data obtained from respondents aged 40 to 55 years were ignored. The estimation methods are less reliable for these age groups, and the respondents are likely to exaggerate their own ages, thereby biasing the results. To smooth out the effects of imprecision in the data and estimation methods, the remaining estimates were regressed on the dates to which they apply. The predicted values of 45p15 for the most recent date at which the data seem reliable are presented in Table 6–1.

OCR for page 218
Demographic Change in Sub-Saharan Africa REFERENCES Ahonzo, E., B.Barrere, and P.Kopylov 1984 Population de la Côte d’Ivoire. Abijan: Ministère de l’Economie et des Finances. Benin No date Enquête sur la Fécondité au Benin. Cotonou: Bureau Central de Recenscement. Bennett, N.G., and S.Horiuchi 1981 Estimating the completeness of death registration in a closed population. Population Index 47:207–221. Blacker, J.G.C. 1991 Infant and child mortality: Development, environment and custom. Pp. 75–86 in R.G.Feachem and D.T.Jamison, eds., Disease and Mortality in Sub-Saharan Africa. New York: Oxford University Press for The World Bank. Blacker, J.G.C., and J.Mukiza-Gapere 1988 The indirect measurement of adult mortality in Africa: Results and prospects. In African Population Conference, Dakar, 1988, Vol. 2. Liège: International Union for the Scientific Study of Population. Botswana 1972 Report on the Population Census, 1971. Gaborone: Central Statistics Office. 1983 1981 Population and Housing Census: Administrative/Technical Report and National Statistical Tables. Gaborone: The Government Printer. Brass, W. 1971 On the scale of mortality. Pp. 69–110 in W.Brass, ed., Biological Aspects of Demography. London: Taylor and Francis. 1975 Methods for Estimating Fertility and Mortality from Limited and Defective Data. Occasional Publication. Chapel Hill, N.C.: International Program of Laboratories for Population Statistics. Brass, W., and E.Bamgboye 1981 The Time Location of Reports of Survivorship: Estimates for Maternal and Paternal Orphanhood and the Ever-Widowed. Centre for Population Studies Working Paper 81–1. London School of Hygiene and Tropical Medicine, London. Brass, W., and K.Hill 1973 Estimating adult mortality from orphanhood. Pp. 111–123 in International Population Conference, Liège, 1973, Vol. 3. Liège: International Union for the Scientific Study of Population. Cameroon 1978 Recensement General de la Population et de l’Habitat d’Avril 1976, Tome 2. Yaoundé: Direction de la Statistique et de la Comptabilité Nationale. 1983 Enquête Nationale sur la Fécondité du Cameroun, 1978, Rapport Principal. Yaoundé: Direction de la Statistique et de la Comptabilité Nationale. Cantrelle, P., I.L.Diop, M.Garenne, M.Gueye, and A.Sadio 1986 The profile of mortality in Senegal and its determinants. Pp. 86–116 in Determinants of Mortality Change and Differentials in Developing Countries. E.85.XIII.4. New York: United Nations. Coale, A.J., and P.Demeny 1983 Regional Model Life Tables and Stable Populations, 2nd ed. New York: Academic Press. Condé, J., M.Fleury-Brousse, and D.Waltisperger 1980 Mortality in Developing Countries. Paris: Organisation for Economic Co-operation and Development.

OCR for page 218
Demographic Change in Sub-Saharan Africa Congo 1978 Recensement General de la Population du Congo, 1974, Tome 4. Brazzaville: Centre National de la Statistique et des Etudes Economiques. 1987 Recensement General de la Population et de l’Habitat de 1984, Tome 3. Brazzaville: Bureau Central de Recensement. Fargues, P., and O.Nassour 1988 Douze Ans de Mortalité Urbaine au Sahel. INED Travaux et Documents 123. Paris: Presses Universitaires de France. Feachem, R.G., D.T.Jamison, and E.R.Bos 1991 Changing patterns of disease and mortality in sub-Saharan Africa. Pp. 3–27 in R.G.Feachem and D.T.Jamison, eds., Disease and Mortality in Sub-Saharan Africa. New York: Oxford University Press for The World Bank. Feachem R.G., T.Kjellstrom, C.J.L.Murray, M.Over, and M.A.Phillips, eds. 1992 The Health of Adults in the Developing World. New York: Oxford University Press for The World Bank. Graham, W.J. 1991 Maternal mortality: Levels, trends and data deficiencies. Pp. 101–116 in R.G. Feachem and D.T.Jamison, eds., Disease and Mortality in Sub-Saharan Africa. New York: Oxford University Press for The World Bank. Hill, A.L.L. 1991 Infant and child mortality: Levels, trends and data deficiencies. Pp. 37–74 in R.G. Feachem and D.T.Jamison, eds., Disease and Mortality in Sub-Saharan Africa. New York: Oxford University Press for The World Bank. Hill, K. 1977 Estimating adult mortality levels from data on widowhood. Population Studies 31:75–84. Hutt, M.S.R. 1991 Cancer and cardiovascular diseases. Pp. 221–240 in R.G.Feachem and D.T.Jamison, eds., Disease and Mortality in Sub-Saharan Africa. New York: Oxford University Press for The World Bank. Mali 1980 Recensement General de la Population, Décembre 1976. Résultats Définitifs, Tome 1. Bamako: Bureau Central de Recensement. Moyo, I.M. 1991 Mortality levels, patterns and trends in the city of Harare, Zimbabwe. Unpublished M.Sc. thesis, London School of Hygiene and Tropical Medicine. Mukiza-Gapere, J.H.G. 1989 Biases and errors in the orphanhood method of estimating adult mortality: An empirical examination. Unpublished Ph.D. thesis, University of London. Murray, C.J.L., G.Yang, and X.Qiao 1992 Adult mortality: Levels, patterns and causes. Pp. 23–111 in R.G.Feachem, T. Kjellstrom, C.J.L.Murray, M.Over, and M.A.Phillips, eds., The Health of Adults in the Developing World. New York: Oxford University Press for The World Bank. Okoye, C.S. 1980 Mortality Levels and Differentials in Sierra Leone, Vol. 2. Freetown: Central Statistics Office. Omondi-Odhiambo, J.K.Van Ginneken, and A.M.Voorhoeve 1990 Mortality by cause of death in a rural area of Machakos District, Kenya in 1975– 1978. Journal of Biosocial Science 22:63–75.

OCR for page 218
Demographic Change in Sub-Saharan Africa Packard, R.M. 1990 White Plague, Black Labor. London: James Currey. Pison, G., and A.Langaney 1985 The level and age pattern of mortality in Bandafassi (Eastern Senegal): Results from a small-scale and intensive multi-round survey. Population Studies 39:387– 405. Preston, S.H. 1983 An integrated system for demographic estimation from two age distributions. Demography 20:213–226. Preston, S.H., A.J.Coale, J.Trussell, and M.Weinstein 1980 Estimating the completeness of reporting of adult deaths in populations that are approximately stable. Population Index 46:179–202. Rutstein, S.O. 1984 Infant and Child Mortality Levels, Trends and Demographic Differentials. WFS Comparative Studies, 43. Voorburg, Netherlands: International Statistical Institute, Central Bureau of Statistics. South Africa 1988 South African Statistics, 1988. Pretoria: Central Statistical Service. Sudan 1982 The Sudan Fertility Survey, 1979: Principal Report, Vol. 1. Khartoum: Department of Statistics. Swaziland 1980 Report on the 1976 Swaziland Population Census, Vol. 2. Mbabane: Central Statistical Office. Tanzania 1982 1978 Population Census, Vol. 5. Dar es Salaam: Bureau of Statistics. Timæus, I.M. 1984 Mortality in Lesotho: A Study of Levels, Trends and Differentials Based on Retrospective Survey Data. WFS Scientific Reports, 59. Voorburg, Netherlands: International Statistical Institute, Central Bureau of Statistics. 1987 Adult mortality in Mauritania. In Evaluation de l’Enquête Nationale Mauritanieanne sur la Fécondité. WFS Scientific Reports, 83. Voorburg, Netherlands: International Statistical Institute. 1991a Measurement of adult mortality in less developed countries: A comparative review. Population Index 57:552–568. 1991b Estimation of mortality from orphanhood in adulthood. Demography 27:213–227. 1991c Estimation of adult mortality from orphanhood before and since marriage. Population Studies 45:455–472. 1991d New estimates of adult mortality from DHS data on the timing of orphanhood relative to marriage. Pp. 857–874 in Proceedings of the Demographic and Health Surveys World Conference, Washington D.C., 1991, Vol. II. Columbia, Md.: Institute for Resource Development/Macro International, Inc. 1991e Adult mortality: Levels, trends and data sources. Pp. 87–100 in R.G.Feachem and D.T.Jamison, eds., Disease and Mortality in Sub-Saharan Africa. New York: Oxford University Press for The World Bank. 1992 Estimation of adult mortality from paternal orphanhood: A reassessment and a new approach. Population Bulletin of the United Nations 33:47–63. Togo 1985 Recensement General de la Population et de l’Habitat, 9–22 Novembre, 1981, Tome 4. Lomé: Bureau Central de Recensement.

OCR for page 218
Demographic Change in Sub-Saharan Africa United Nations, Department of International Economic and Social Affairs 1982a Model Life Tables for Developing Countries. E.81.XIII.7. New York: United Nations. 1982b Demographic Yearbook, 1980. E.81.XIII.1. New York: United Nations. 1987 Demographic Yearbook, 1985. E.86.XIII.1. New York: United Nations. 1988 Mortality of Children Under Age 5: World Estimates and Projections, 1950–2025. E.88.XIII.4. New York: United Nations. Waltisperger, D., and T.Rabetsitonta 1988a Un bilan de trente ans de mesures directes de la mortalite adulte en Afrique. In African Population Conference, Dakar, 1988, Tome 2. Liège: International Union for the Scientific Study of Population. 1988b Un bilan de trente ans de mesures directes de la mortalite adulte en Afrique: Annexes. Unpublished tables, Centre Français sur la Population et le Developpment, Paris. World Bank 1988 Social Indicators of Development 1988. Baltimore, Md.: Johns Hopkins University Press for The World Bank. Wurrie, F. 1979 Analysis of mortality by cause of death for the western area in Sierra Leone, 1972– 1975. Unpublished M.Sc. thesis, London School of Hygiene and Tropical Medicine. Zimbabwe 1985 Main Demographic Features of the Population of Zimbabwe. Harare: Central Statistics Office.