Population Dynamics of Kenya (1993)

It is not possible to derive irrefutable conclusions about the determinants of demographic change from an analysis of data for any one country. Many factors are operating simultaneously with multiple interactions. Variations among social and regional population subaggregates contribute significant information, but it must not be forgotten that some of the major influences are likely to be common to all of them. These difficulties exist even when there is an abundance of reliable data, which is by no means true for Kenya. Probably the biggest handicap for the present purpose has been the limitation of quantitative measures for subaggregates of the population. For example, only crude and sketchy estimates of adult mortality and of migration can be derived for districts, so there is a considerable barrier to any study of the components of population growth.¹ Even basic indices that are known with fair precision at the national level, such as the total fertility and child mortality rates, can only be estimated roughly for units smaller than provinces. This limitation is a consequence of the comparatively small sample sizes in the household surveys that provide the most satisfactory detailed data. Similar conditions apply to the social and economic indicators whose links with demographic change are explored. The extent to which residents in rural areas participate in the activities and services of the main urban centers, such as Nairobi, Mombasa, and Kisumu, is hard to assess.

For all the reasons given, attempts to seek explanations for the significant changes in the demography of Kenya must be an assessment of consistencies. How far do the particular features of the country fit in with general theories and findings of demographic transition as deduced from world experience? The nature of the specific evidence gives more power to negative results than to positive ones. A contradiction has a validity on its own that may be argued away on the grounds of inadequate data or peculiar conditions but nevertheless must be recognized. A conformity is of much less weight because it is normally one that fits with many alternative interpretations.

The trends in child mortality and fertility have completely different characteristics, with the former being a continuous decline over several decades and the latter a very recent fall. This finding does not prove that they have no major causes in common but demonstrates that, if they do, the causes operate in distinctly separate ways. It is legitimate therefore to examine the two phenomena independently before raising questions about interactions. The limited and uncertain information on adult mortality makes it impracticable to study this component separately. However, it can be noted that there was a substantial decrease in adult mortality from the 1950s, although the bits and pieces of indirect evidence are inadequate to measure its size with any confidence. The moderate correlation between the levels of child and adult mortality by districts in the period 1969–1979 engenders confidence in the measures, because the two sets were derived by methods that were largely independent. There is then a plausible argument for the claim that trends in child and adult mortality from the 1950s to the 1970s were similar, although the materials are not sufficient to detect any variations in time and regional patterns that might contribute to the elucidation of determinants.

The continuance of the mortality change from the earlier periods and the small accretion of new data at the subaggregate level mean that there is little more to add to the thorough examinations in previous studies. The only novelty in this report is the greater concentration on trends. The steady declines in child mortality from the 1950s have been examined in Chapter 3. Despite the data problems for the most recent period, it is fair to deduce that the improvement in the 1980s was at an even faster pace than in the 1960s and 1970s. It might seem surprising that the unfavorable economic experience of the later period, relative to the earlier, cannot be detected in the child mortality record (see Working Group on Demographic Effects of Economic and Social Reversals, 1993, for an examination of the issue).

The cross-sectional analyses of district variations in child mortality found a number of associations that can broadly be described as environmental. These were with the incidence of malaria, the malnutrition measures of stunting and wasting, and ecological zones. In this classification, cultural factors are included with environmental because there is essentially no separation of the ecological effects from those due to the behavior of the communities living in the district units. In particular, the very high childhood mortality on the margins of Lake Victoria can be ascribed to diseases such as malaria or to the practices of tribal groups, particularly the Luo. Because there was no significant alteration in these ''environmental'' conditions over the period examined, they would not be expected to appear prominently as associated with trends. That turns out to be so and simplifies the consideration of what may be called "developmental" factors.

The social and economic advances of the country as a whole since independence in 1963 are outlined in Chapter 2. In light of the impressive progress, the strong improvement in child mortality is no surprise. But the overall gains provide no opportunities for a better understanding of the specific factors that were the most important. The large variations in child mortality trends among districts do present such opportunities. The variable that emerges as clearly significant for lower mortality is a higher standard of education. The level of education can be measured in a number of ways. Several alternative specifications were explored that showed very similar resulting associations with the trends in child mortality. The simplest is the proportion of adult women with no schooling, which has been adopted here to demonstrate the quantitative relations with trends in ₅q₀, the proportion of children dying by age 5 years. The correlation is strong and consistent, except for a few outlying observations for the more remote areas, where there are doubts about the accuracy of the measures of the mortality trend. The association between child mortality trends and adult literacy proportions is almost equally impressive, despite some reservations about the validity of the latter estimates. It is also relevant to note here that adult mortality in 1969–1979 is correlated with the education level of the women over districts.

Differentials in child mortality by education of the mother have been a focus of interest since at least the studies of Caldwell (1979) and Behm (1979). The large size and near universality of the effects are well established, but the mechanisms by which they arise are still controversial. It is generally thought that well-educated women tend to be wealthier, to reside in better houses and environments, and to be less constrained in decisions about their children's welfare. In Chapter 3, the ₅q₀ values by the education of the mother as calculated from the Kenya Fertility Survey (KFS) and the Kenya Demographic and Health Survey (KDHS) data are shown. In both surveys, child mortality for women with no schooling was about twice that

of women in the highest education group with more than 9 years of school. But the reduction in child mortality between the surveys was much the same for all schooling categories. In particular, the decline for women with no schooling was nearly identical with the improvement for Kenya as a whole, despite the residual nature of the group due to the rapid advance of female education. Only a small fraction of the child mortality reduction was a consequence of the changing composition of women by grade of schooling. The indications are that the same features were present in earlier child mortality improvements.

The configuration of these results for individuals and district aggregates does not support the view that the driving force in the child mortality trends is the specific benefit at the family level of an educated mother. Individual education may be and probably is significant for the existence and preservation of differentials at points in time, although there may be confounding with other social and economic determinants. The strong reduction in child mortality for the offspring of women with no schooling demonstrates that more general influences are operating. At the district aggregations, the similarity of the associations of child mortality trends with female education and adult literacy (both sexes), as well as the adult mortality correlation with female education, again suggest general rather than specific forces.

In sum, what is notable is the extension of child mortality declines to mothers with little schooling if the educational achievement of the district is good. Taken together, the relationships suggest that the underlying determinant of the child mortality reductions is the increase in knowledge related to the capacity of the family to control its own environment. Obviously formal education is an important component, but improvements in child care tend to spread over a community even if they begin in only limited sections of it. The apparent lack of much association at the district level between child mortality trends and health indicators, particularly health facilities, is disappointing. Although the measures of the latter are crude, they should be enough to give a rough guide to the opportunities for child care. But any relation may be lost among other influences.

The fertility decline of 20 percent in a decade occurred sharply, following 20 years of high and probably slightly rising fertility rates (Chapter 4). The decline was primarily a consequence of the increased use of contraception to control births (Chapter 5). The effect extended to every province of the country, except possibly the remote Northeastern Province for which there is no information. Although the size of the decline varied by regions, it was substantial in widely spaced districts. The simultaneous initiation of

the downward trend in nearly every area strongly suggests that the determinants contained a powerful central component. There are no signs of a gradual spread from the more favored or developed regions to the less advanced. Indeed the linkages by district of the fertility reductions with the developmental indicators of education, urbanization, mortality, and population density are extremely weak. It is true that the quantitative relations of these to the fertility declines can be analyzed only for 17 of the 41 districts because of data limitations. The excluded areas contain all the remote districts with poor development and, possibly, little fertility decline, although this is not known with certainty. However, the restriction of the units to the more populous districts does not detract from the evidence they provide. Of particular note are the districts of the Coast Province for which the fertility decline was 27 percent, second only to Central Province. A greater part of Coast's decline compared to other regions may have been due to later ages at first birth, but the contraceptive component was also considerable. Yet, apart from Northeastern Province, the Coast had the lowest values of development indices for education, literacy, and child mortality improvements.

Further insights into the nature of the fertility transition come from the analysis of the trends in parity progression ratios by cohorts, calculated from the KDHS birth histories. The striking feature of the pattern is the similarity of the trends in size and timing at all birth orders. It appears that the pressures toward birth limitation had their impact on parents of different ages and family sizes at the same time. The gradual spread of birth limitation from the middle parities to the higher and then to the lower orders, typical in the early stages of fertility transition in Latin America and Asia, does not hold in Kenya.

The only socioeconomic indicator having a possible association with fertility declines by district is employment in the modern sector. It is clear that some employees in the towns, particularly Nairobi, Mombasa, and Kisumu, reside in surrounding rural areas. Their transference to the districts of residence would decrease fertility and raise modern sector employment. How large the changes would be is hard to guess, but they might shift a rather uncertain correlation to an impressive one. The associations are much the same for female and male employment. It cannot then be claimed that any effects operate through the mothers rather than the fathers, although such a mechanism cannot be rejected.

Although the observations on fertility declines by district are too restricted and unreliable to support a valid multivariate analysis of the linked factors, much can be gained by investigation of the contraceptive data from the KDHS. It has been established in Chapter 5 that increased use of contraception is closely tracked by fertility reductions. Because contraceptive use was so low at the KFS, with the partial exception of a few subcat-

egories, the multivariate analysis of the KDHS reports of contraception by individuals can serve as an effective surrogate for a direct examination of usage trends. This analysis is presented in Chapter 7.

The proportion of the variation in contraceptive use contributed by the independent variables is small—a result that is consistent with the failure to find much linkage in the crude district comparisons. Nevertheless, the exercise yielded several interesting results. Some of these are negative, notably the lack of differences by religion, in spite of the claims that Muslims, mainly in Coast Province, would be more resistant to the spread of family planning. The small differences in use of contraceptives by age of women from 25 years on fits with evidence from the parity progression ratios.

The significance of educational attainment at the individual level and the rural literacy rate at the district level may appear to contradict the comments above on the weakness of the links between developmental indicators and fertility declines. This is not necessarily the case, however, because contraceptive use by the better-educated women at the KFS was far from negligible. Here, there is a distinction between fertility trends and cross-sectional differences at points of time.

There is a cluster of variables that have significant relations with contraceptive use and that might be indicators of economic standards or of openness to external influences. These are the presence of electricity, a house floor other than mud, and a habit of radio listening at the household level. Participation in a women's organization is also a factor in contraceptive use.

The sharp fertility transition affecting nearly all regions, educational groups, and birth orders gives weight to the view that the changing social and economic factors for Kenya as a whole, from the late 1970s onward, were a major stimulus to fertility reduction. The three outstanding features of the process were the deterioration in real wages, the rise of living standards, and the growing strength of the family planning program.² The economic conditions provided the push, and the programs the opportunity. The key proximate determinant in the fertility declines was the increase in contraception (see Chapter 5). Of course that alone does not prove that the supply of family planning services was a significant element in the expanding usage, compared to the demand for family limitation. Rather, the evidence lies in the vigor with which the provision of services was pursued by the Kenyan government and international agencies as outlined in Chapter 6. No doubt the impressive social and economic developments from the 1950s

to the 1970s, along with the associated declines in child mortality, established the necessary conditions that made fertility reduction possible. But the configurations of change suggest that the more immediate impulse came from the the economic problems of the 1980s and the vigor of the family planning program.

Interpretation of the association at the district level between fertility declines and employment in the modern sector (Chapter 7) must be very speculative. The data weaknesses preclude a close examination of the characteristics of the relation. It is difficult to see it as operating through direct economic pathways of income. Certainly the developed districts of Central Province tend to rank high for both variables, but there are others with such measures that are not as advanced economically, for example, Kericho and Uasin Gishu. Employment in the modern sector may lead to greater awareness of a broader range of ideas in which family planning is seen as a rational solution to economic pressures. If this is so, there are other possible concomitants that might have a bearing. Thus, the massive expansion of the Kenyan tourist industry has increased the contacts with overseas visitors and the potential for communication of attitudes. It is at least interesting that tourism has its largest impact in the Coast Province where the decline in fertility appears so inconsistent with socioeconomic development. Unfortunately, the data to pursue this hypothesis are lacking. The multivariate analysis of contraceptive usage at the time of the KDHS reveals significant factors that can be interpreted as indicators of greater exposure to family control ideas, but a more direct influence of economic conditions cannot be excluded.

What indications for future population dynamics come from the analysis of past and current trends? Any such forecasts must assume a reasonable degree of political and economic stability. The histories of Uganda, Mozambique, and Ethiopia show how precarious such an assumption may be in sub-Saharan Africa. The mortality reductions in Kenya have been well established for some 50 years. They will continue in the absence of major catastrophe, such as a dramatic increase in AIDS.³ The pace of mortality improvement may well slacken due to slower social and economic progress, especially because very low death rates require multiplied effort. The experience of other countries suggests, however, that the momentum built up has yet to expend itself.

In some respects the assessment of the prospects for fertility is more uncertain; in others, more secure. If deteriorating economic conditions were a factor in the initiation of fertility declines, it might be that a reversal in the deterioration would push birth rates up.⁴ The acceptance of family planning, once securely established, does not seem to be easily reversible. It has been demonstrated that the rise in contraception is remarkably widely spread by region, social condition, age of women, and family size. The process has not been one of gradual diffusion. As in the case of the mortality trend, the internal momentum should carry it forward toward the substantially lower fertility consistent with desired family sizes. The ideal family size reported at the time of the KFS was 6.2 children, and at the Kenya Contraceptive Prevalence Survey, 5.8. Between those surveys from 1978 to 1984, the level of contraceptive use was much too low to achieve these stated ideals, but a large increase in usage followed. At the time of the KDHS in 1988–1989, the ideal family size had fallen to 4.4 children. It seems reasonable to expect a considerable move toward attaining this ideal in the next decade. Whether there is some intermediate plateau short of replacement levels where the trend would stop (as has happened in other countries) cannot be guessed.

This report is a case study of Kenya. Detailed investigations of the patterns of demographic change in other countries of sub-Saharan Africa are outside its scope. Indeed the demographic data sources are much more restricted elsewhere. However, it seemed relevant for comparative purposes to produce estimates of the trends in parity progression ratios over cohorts for countries of Latin America, Asia, and Africa. The striking finding was that the Kenyan pattern presented similar reductions in size and timing across all birth orders, as contrasted with the pattern of Latin America and Asia, where declines spread from the middle parities. However, the pattern in Zimbabwe, Botswana, and Nigeria showed features very much like those of Kenya. This finding gives good support to the views that the Kenya transition is not unique but is shared by other countries of sub-Saharan Africa, and that the pattern emerging in sub-Saharan Africa is different from that of other regions.