7
Linkages Between Socioeconomic Factors and Demographic Change

In this chapter, we attempt to link socioeconomic factors to changes in child mortality and fertility. In the first two sections, we investigate these links at the district level. We also discuss the relation of child mortality to adult mortality and to fertility. Finally, we analyze contraceptive use at the individual level using a multivariate model.

CHILD MORTALITY AND SOCIOECONOMIC FACTORS

The most notable finding from the analysis of child mortality trends in Chapter 3 is the consistency of the changes over time for subregions (provinces and districts) from the 1950s onward. Despite the difficulties of determining precisely the level of child mortality in the later 1980s, it can be deduced with confidence that the rates continued to fall at a pace similar to the previous decade. This continuation was established for the country and the provinces (excluding the thinly populated Northeastern Province). The changes by districts in the most recent period cannot yet be estimated, but the regularity of the district child mortality declines prior to the 1970s, and the continuance of the province trends into the 1980s, make it certain that similar conclusions can be drawn about all, or almost all, of the smaller aggregates. The child mortality reductions of 1954 to 1974, given in Chapter 3, are then satisfactory indices of the subregional improvements. The estimation methods, which use the proportion of children dead by age groups of mothers from census reports, are capable of producing sufficiently accurate values for the broad levels of child mortality in given periods but not



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 139
Population Dynamics of Kenya 7 Linkages Between Socioeconomic Factors and Demographic Change In this chapter, we attempt to link socioeconomic factors to changes in child mortality and fertility. In the first two sections, we investigate these links at the district level. We also discuss the relation of child mortality to adult mortality and to fertility. Finally, we analyze contraceptive use at the individual level using a multivariate model. CHILD MORTALITY AND SOCIOECONOMIC FACTORS The most notable finding from the analysis of child mortality trends in Chapter 3 is the consistency of the changes over time for subregions (provinces and districts) from the 1950s onward. Despite the difficulties of determining precisely the level of child mortality in the later 1980s, it can be deduced with confidence that the rates continued to fall at a pace similar to the previous decade. This continuation was established for the country and the provinces (excluding the thinly populated Northeastern Province). The changes by districts in the most recent period cannot yet be estimated, but the regularity of the district child mortality declines prior to the 1970s, and the continuance of the province trends into the 1980s, make it certain that similar conclusions can be drawn about all, or almost all, of the smaller aggregates. The child mortality reductions of 1954 to 1974, given in Chapter 3, are then satisfactory indices of the subregional improvements. The estimation methods, which use the proportion of children dead by age groups of mothers from census reports, are capable of producing sufficiently accurate values for the broad levels of child mortality in given periods but not

OCR for page 139
Population Dynamics of Kenya precise measures by stages of childhood or calendar years. This conclusion follows because the dead children of women in any age group are spread over a range of dates of births and deaths. Cross-Sectional Relationships The information for the calculation of child mortality up to the 1970s has been available for some years. There are several examinations of the relations between the mortality estimates and the socioeconomic indices for provinces and districts. The most relevant for the present study are by Ewbank et al. (1986) and Blacker et al. (1987). In both of these, the child mortality estimates were derived by essentially the same techniques used in the present report. There are slight differences in detail, according to the measures employed (e.g., infant mortality in Ewbank et al.) and the location in time. There are also variations in the methods of adjusting the estimates in the small number of cases where the data are suspect. These divergences have a negligible impact on the investigation of socioeconomic linkages at the attainable level of precision. The main concern of these two papers was with the cross-sectional district variations at points of time, although there is some attempt to look at changes. The emphasis here is on determinants of trends. The two issues are, of course, closely linked but far from identical. For example, the basic environmental factors associated with particular causes of death, such as malaria, are likely to appear more prominently in cross-sectional investigations. Ewbank et al. (1986) reported the results of a regression analysis of district infant mortality rates from the 1969 and 1979 censuses and the intercensal changes. The explanatory variables included for the 1979 exercise were female literacy, the prevalence of malaria, the percentage urban, the number of kilometers of road, the population density, and the potential agricultural land per capita. A measure of availability of health services was constructed from the per capita availability of beds in hospitals and dispensaries, divided by the square root of the land area of the district. There were also a series of variables that assigned each district to one of five ecological zones. A similar analysis was applied to the infant mortality estimates from the 1969 census, but two variables—number of kilometers of road and prevalence of malaria—were omitted because of lack of data. In the regression analysis of change, the ecological zones were included along with the 1969 to 1979 differences for the other available variables. Several relations significant at the 10 percent level of probability were determined. In both 1969 and 1979, infant mortality was lower in districts with higher proportions of educated females age 25–29, with greater population densities, and with more high-potential agricultural land. Nyanza and Western provinces in the ecological zone defined by the Lake Victoria basin

OCR for page 139
Population Dynamics of Kenya experienced significantly high infant mortality rates as did, to a lesser extent, the dry eastern and coastal areas (Coast Province plus Kitui and Machakos districts). A substantial part of the district infant mortality variation was accounted for by the socioeconomic and environmental variables (the R2 coefficients were .70 and .80 for the 1969 and 1979 analyses, respectively). However, the analysis of changes in infant mortality between 1969 and 1979 revealed only a weak relation with the independent variables (R2 equaled .26), the only significant factors being the percentage of educated females among those aged 25–29 years and a slightly better performance in the Nyanza and Western provinces ecological zone. Ewbank and colleagues also utilized the reports on malnutrition in the third Kenya nutrition survey in 1982. The prevalence of stunting (short heights at given ages of children) is an indicator of chronic malnutrition. Children were classed as stunted if their heights were less than 90 percent of the World Health Organization (WHO) standard. The proportions stunted in each district were added as a variable in the regression analysis for 1979 already described. The extra contribution to the explanation of the mortality variation was negligible. A reduced multiple regression retaining only the female education and stunting variables resulted in both having significant effects at the 5 percent level, with 50 percent of the variation in infant mortality accounted for. However, when the proportion of outpatient cases due to malaria was included in the regression, the stunting association became nonsignificant. It appears that a complex of socioeconomic and environmental factors contributed to the district variations in child mortality, but the parts played by each cannot be disentangled. National-level differentials in child mortality figures and trends by education of the mother are presented and discussed in Chapter 3 . The 1979 census provided data on proportions of children dead by age group of mothers, subdivided by both district and education. Measures of child mortality can be derived from these proportions by the same methods applied previously for the district totals. Kibet (1981) provided such measures using the index 2q0 (the probability of death in the first 2 years of life), which is obtained by a slight modification of the proportions of children dead for mothers aged 20–24 years. For Kenya as a whole, the 2q0 for women with a secondary education was only 37 percent of the corresponding value for women with no education. In the great majority of districts the ratio did not vary much from the national average. In some of the more remote areas where it did, the numbers of women with a secondary education were low and the sample errors large. The broad regularity of the relation can be illustrated from the districts with the lowest overall 2q0, Nyeri, and the highest, South Nyanza. In the former, the secondary education mortality index is 39 percent of the no-schooling value; in the latter, the proportion is 43 percent. As pointed out by Ewbank et al., this result suggests that the

OCR for page 139
Population Dynamics of Kenya relation between maternal education and child mortality is not an artifact of environmental factors or cultural characteristics. As they noted, however, education is not the only important determinant because women with secondary education in South Nyanza had a higher 2q0 (.104) than the uneducated in Nyeri (.079). Similar reversals apply in many comparisons between pairs of districts from the upper and lower ranges of child mortality. Although Blacker et al. (1987) used much the same measures as Ewbank and colleagues, their approach was different. They examined, using scatter diagrams, how district variations in child mortality (as described by 5q0) related to single socioeconomic and environmental measures. Their aim was to illuminate the nature of the associations by highlighting deviant districts. The causes of the deviation could have been biases in the measures due to data errors or real differences in the patterns among subregions of the country. Blacker and colleagues found that many of the relations explored produced no signs of significant interactions. The only variables to which child mortality was related were female education, percentage of outpatients with malaria, stunting (less than 90 percent of the WHO height-age standard), wasting (less than 80 percent of the WHO weight-age standard), percentage of households with no piped water, persons per health facility, and population density. Although there were suggestions of correlations between child mortality shortly before the 1979 census and health indicators, none is very impressive except that with the prevalence of malaria. The stunting and wasting indices derived from the Third Rural Nutrition Survey of 1982 tended to be large in the high-mortality districts of the Coast Province and small for the low-mortality Central Province. But over the bulk of the districts there was little association, and the diagrams showed a wide scatter with only a slight tendency for the 5q0 values to increase with the stunting and wasting proportions. The configuration of the plot for 5q0 and the percentage of households without piped water (also derived from the 1982 Nutrition Survey) were very similar to the stunting and wasting results, although with notable deviations in some districts. Thus, some of the highest child mortality areas of the Coast were classified as good in terms of water supply. Because only a small proportion of households had piped water, however, the analysis was not a powerful one. The persons per health facility (hospitals, health centers, clinics, and dispensaries) in each district were weighted by the square root of the land area to obtain an "access" variable for 1979. The overall plot of child mortality against the health facility access variable indicated no association but if the small number of outliers (five) with extremely high (i.e., poor) access measures but average mortality were ignored, some tendency for improvement with better access emerged. Again the relation came largely from the favorable position of the Central Province districts and the unfavorable placing of some coastal regions.

OCR for page 139
Population Dynamics of Kenya The examination of how child mortality varied with population density did not reveal very much. The 5q0 values were high in the densely populated districts bordering Lake Victoria but low in the equally congested areas of Central Province. In some of the districts of low density, for example, on the coast, a large proportion of the population was settled in subregions that, if the data could be disaggregated, would show much increased measures of persons per unit area. The increase in 5q0 with the percentage of new outpatients with malaria was consistent and convincing. The scatter diagram is reproduced in Figure 7-1. Although the index of malaria prevalence was less than ideal, Blacker et al. note that it agreed well with spleen enlargement rates where these were available. Only three districts seem to fit distinctly poorly with the apparent relation of 5q0 to the malaria prevalence index. In two of these, Baringo and Kisii, there was reason to doubt the validity of the malaria measures because they were not consistent with spleen rates. The latter implied levels that would improve agreement with the regression line. The largest proportion of new outpatients with malaria was recorded for Bungoma where child mortality was high but rather lower than expected on the basis of the correlation. Figure 7-1 Child mortality (1974) and malaria (1975–1984).

OCR for page 139
Population Dynamics of Kenya The large differentials in childhood mortality by education of the mother are noted in Chapter 3 of this report and are documented in many studies of Kenya. The analogous relation for district aggregates appeared in the Ewbank et al. (1986) multivariate analyses of the 1969 and 1979 census estimates. The Blacker et al. (1987) scatter diagram by districts for 1979 of 5q0 plotted against the percentage of women aged 15 years or more with no schooling was consistent with the other findings. However, in districts with lower levels of female educational attainment, the association was far from regular. There were two clusters that were inconsistent with the relationship shown by the other units. In four districts of the Lake Victoria basin (Kisumu, Siaya, South Nyanza, and Busia), child mortality was exceptionally high but education levels were near average. In a group of districts in the thinly populated areas of northern Kenya plus Narok and Kajiado on the border with Tanzania, educational levels were low but mortality was estimated to be moderate. The latter cluster included several of the districts where there was most doubt about the accuracy of child mortality estimates, but the view that there may be favorable environmental factors to offset the adverse effects of poorly educated mothers in these districts is not implausible. Blacker et al. also calculated the ratio of 5q0 from the 1979 census to the corresponding value from the 1969 census and plotted results by district against the improvements in female education (measured as the proportion of women over 15 years with no education in 1979, divided by the corresponding proportion in 1969). Although the points were widely scattered, particularly where the improvements in education were small or even retrograde, there was a distinct relationship. Again, it tended to be dominated by the districts of Central Province, where there were substantial gains in female education and child mortality, and the coast plus the more remote northern areas where changes in both were small. Relationships with Trends Most of the association of child mortality with socioeconomic indicators investigated by Ewbank et al. and by Blacker et al. were at certain points in time (mainly around 1979). But the emphasis in the present study is on trends. Therefore, we have focused on comparing the mortality declines and socioeconomic measures by district. The data utilized are essentially the same as in the other analyses, although there are some differences in specification. The conceptual framework is, however, a little different. The mortality trends are taken to represent a process of change that operated consistently from the 1950s to the present. The socioeconomic indicators measured late in that period are assumed to reflect the development over the range of time in the relevant sectors: education of women, provision of health services, etc. In Figure 7-2, the district child mortality de-

OCR for page 139
Population Dynamics of Kenya Figure 7-2 Child mortality declines (1954–1974) and female education (1979). clines from the mid-1950s to the mid-1970s are plotted against the percentages of women aged 15–44 years with no schooling as reported in the 1979 census. Nairobi and Mombasa were excluded because of obvious selection problems due to their large movements in population. The association between the two variables is apparent. The districts that deviate from the general regression (Kajiado, Isiolo, Wajir, and Elgeyo-Marakwet) are remoter areas for which there is some doubt about the reliability of the mortality estimates. However, it is also possible that environmental factors are affecting the death rates differently there than in the rest of the country. The group of districts in the lake basin that were outliers in the relation between child mortality level in 1979 and female education (Kisumu, Siaya, South Nyanza, and Busia) now fit in well, because the mortality decreases are moderate as is educational attainment, although the death rates were high at all points of time. Several of the more remote areas with a small proportion of educated women, but moderate child mortality, recorded only a slight fall in the latter or even a rise (Garissa, Turkana, Marsabit, Samburu, and Mandera). Thus, the association of the trends in child mortality with the level of female education reached in 1979 is clearer and tidier than the cross-sectional relations found in the earlier studies. A similar exercise,

OCR for page 139
Population Dynamics of Kenya Figure 7-3 Child mortality declines (1954–1974) and adult literacy (1981–1988). with the percentage of females among pupils at secondary schools as the education variable, also indicates a fair correlation—but less impressive than demonstrated by Figure 7-2. It seems likely that the mortality declines by district would also be associated with adult literacy, and this relation is confirmed by Figure 7-3. The literacy data are taken from the Rural Literacy Surveys of 1981 and 1988. The indicator used is an average of the proportions of adults classed as literate (both sexes) in 1981 and 1988. Literacy was defined in the surveys as the ability to read and write in any language. The average was taken to reduce sample errors and individuals biases. Although in most districts the change from 1981 to 1988 was small or plausible, there were notable exceptions, for example, Narok (20 percent in 1981, 39 percent in 1988) and Kwale (57 percent in 1981, 35 percent in 1988). Only 30 districts were available for the plotting of points because of the exclusion of urban areas such as Nairobi and Mombasa and some amalgamation of remote districts. The regularity of the association between mortality improvement and higher adult literacy is impressive. The outlying points are for Elgeyo-Marakwet, where the estimated trend in child mortality is hard to believe, Kwale, and Kajiado. If the more plausible 1988 measure of adult

OCR for page 139
Population Dynamics of Kenya literacy is taken for Kwale, it is no longer an outlier. The fact that adult literacy (covering both sexes) is as closely related to the child mortality trend as female education is, suggests that the associations are the product of general rather than specific influences. Exploration of the links of socioeconomic variables with fertility decreases, presented later in this chapter, suggests that female employment might be related to child mortality as well. The numbers of women employed in the modern sector in each district were taken from the employment survey and divided by the corresponding total women aged 15–44 years in the district at the 1979 census. The coverage of the employment survey is by no means complete, and the age range of women workers is not restricted to 15–44 years. Nor will employment always be in the district of residence, particularly near urban areas such as Nairobi and Mombasa. Nevertheless, the indices give a rough guide to the intensity of female employment in the modern sector. The declines in child mortality by district are plotted against the female employment measures in Figure 7-4. Our hypothesis is that the greater exposure to external ideas through employment in the modern sector is favorable to the spread of healthier child care practices. The scatter diagram indicates an association, but it is far from linear < div class="captionblock"> Figure 7-4 Child mortality declines (1954–1974) and female employment (1980).

OCR for page 139
Population Dynamics of Kenya and is determined mainly by the Central Province districts with large child mortality improvements and fairly high employment indices, as well as those districts with poor mortality declines and low female employment. The latter are geographically spread in Coast, Rift Valley, Northeastern, and Eastern provinces. When the female employment index exceeds 5 percent (27 of the 40 points, excluding Nairobi), there is little, if any, rise in the improvement of child mortality with the employment measure. The lack of much relation between child mortality in the late 1970s and the availability of health facilities was noted by Blacker et al. In Figure 7-5, the same health facility indicator is plotted against child mortality trends. Again, there is little relation between the variables, except that provided by the Central Province districts with the five largest mortality declines and good provision of health services. If these are omitted, the correlation becomes negligible. Whatever association exists is due to a coincidence of developments in a limited region of the country. Much the same conclusions apply to the relations between child mortality trends and other health indicators, namely, the prevalence of malaria and of wasting and stunting, as well as the availability of piped water. For all of these indicators, the Central districts have favorable measures. The nutritional indicators are Figure 7-5 Child mortality declines (1954–1974) and health facilities.

OCR for page 139
Population Dynamics of Kenya poor for the Coast districts, which also show little fall in child mortality. For the other districts, the plots show a wide scatter of points with little correlation. The configuration of the graphs provides no support for the claim that these socioeconomic factors are significantly related to falls in child mortality. Relationship with Adult Mortality In Chapter 3 the relation between adult and childhood mortality levels is examined briefly, following Blacker et al. (1987). As pointed out, the only usable data are for the 1969 to 1979 census interval, and therefore the emphasis on longer-term trends in childhood mortality cannot be pursued for adults. Blacker et al. also examined relations of adult mortality levels estimated for 1969–1979 to population density and to female education by district. The former exercise revealed little. As would be expected from the association between child mortality and female education as demonstrated by Ewbank et al. (1986) and Blacker et al. (1987), and the association between child and adult mortality, higher levels of life expectancy at age 15, e(15), tended to be associated with more schooling. The correlation was modest, however, and somewhat weaker than the link between child mortality trends and female education. It is true that the northern districts with their low adult life expectancies and poor schooling fit in well with the regression, but there were outliers scattered over the provinces (e.g., Kisumu, Taita-Taveta, Narok, and Kilifi). A plot of the e(15) measures against the percentage of women with no schooling is shown in Figure 7-6. In view of the measurement uncertainties, it would be unjustified to conclude that female education was more closely related to child than to adult mortality. Rather, the education index reflects factors that are significant for both. FERTILITY AND SOCIOECONOMIC FACTORS The extensive previous analyses of fertility in Kenya have focused on cross-sectional variations in levels and patterns of fertility in relation to socioeconomic and environmental influences. There have also been some attempts to examine the causes of the increases in fertility in the 1950s and 1960s. But evidence of declining fertility in the 1980s became available only with the dissemination of the data from the 1988–1989 Kenya Demographic and Health Survey (KDHS). Thus, there has been no analysis of the relation of the declines to socioeconomic factors. The extraction of reliable measures of the fertility decreases for aggregates smaller than provinces is not straightforward. The estimates made in Chapter 4 for 17 districts out of the 41 total (comprising about two-thirds of the population) are clearly not precise but can be used with caution to explore relations to socioeconomic

OCR for page 139
Population Dynamics of Kenya A logistic regression is used to analyze the factors associated with contraceptive use. The dependent variable indicates whether or not a woman is currently using contraception. The sample includes nonpregnant women currently in a union. Individual-level characteristics from the KDHS and district-level characteristics taken from various government surveys and documents are used as explanatory variables.2 Results The results of the analysis are presented in Table 7-2. Most of the factors operate on contraceptive use as expected. Availability of family planning demonstrates the strongest positive association with contraceptive use. A woman living in a district with at least 11 service delivery points (SDPs) per 100,000 people is 2.75 times as likely to use contraception as a woman living in a district with fewer than 11 SDPs per 100,000 people, net of other individual-and district-level effects. Density of family planning services clearly has a positive effect on increased contraceptive use, either by providing the means to attain limited family size or by introducing the concept of fertility control not previously considered by the woman. Hammerslough (1991a) gave evidence for the former explanation in an analysis of survey data from the Kenya Community Survey, which was conducted in 1989 in 260 of the rural sample clusters used in the KDHS. This survey gathered community information from adult residents, including information on family planning service availability. Using a multivariate framework, he concluded that although accessibility to services increased at the same time as demand for contraception rose, it did not initiate the increase in contraception. However, the increased availability of services probably did induce more women to use efficient methods instead of traditional ones. Interestingly, at the individual level, family planning accessibility, as measured by a woman's travel time to a family planning source, is not significantly associated with contraceptive use. This result is not totally 2   The sources of community-level data for the multivariate analysis are as follows: number of service delivery points by district, 1991: Health Information System Unit, Ministry of Health computer file, November, 1991; kilometers of paved road per 10,000 people, 1991: provided by the Roads Department in the Ministry of Transportation and Communication; percentage of the rural population literate, 1988: Kenya (1990); percentage of the female population employed in the modern sector, 1980: employment data from 1980 Kenya employment survey divided by women aged 15–44 years at the 1979 census; and percentage of the district that is urban, 1989: Kenya (1991b).

OCR for page 139
Population Dynamics of Kenya TABLE 7-2 Logit Estimates of Probability of Contraceptive Use Among Nonpregnant Women Currently in Union, Kenya, 1989 Explanatory Variables β SE(β) Odds Ratio Individual Level       Travel time to family planning source       (More than one hour)       Less than one hour .001 .094 1.00 Household electricity       (No electricity)       Electricity .972a .184 2.64 Education       (None)       1–6 years .317a .114 1.37 7+ years .658a .117 1.93 Type of household flooring       (Mud floor)       Other than mud floor .497a .103 1.64 Woman's group       (Does not belong)       Belongs .365a .087 1.44 Listens to radio       (Less than once a week)       At least once a week .306a .099 1.36 Type of residence       (Rural)       Urban -.477a .155 0.62 Age       (15–24)       25–39 .621a .116 1.86 40+ .567a .147 1.76 Religion       (Muslim or other)       Christian .290 .195 1.34 District Level       Availability of family planning       (Fewer than 11 service delivery points per 100,000 people)       At least 11 service delivery points per 100,000 people 1.012a .145 2.75 Rural literacy       (Less than 50% of the rural population literate)       At least 50% of the rural population literate .973a .124 2.65

OCR for page 139
Population Dynamics of Kenya Explanatory Variables β SE(β) Odds Ratio Paved roads       (Fewer than 25 kilometers of paved road per 10,000 people)       At least 25 kilometers of paved road per 10,000 people .278a .100 1.32 Urbanization       (Less than 6% urban)       At least 6% urban -.390a .106 0.68 Female employment       (Less than 5.5% of female population employed in the modern sector)       At least 5.5% of female population employed in the modern sector .097 .126 1.10 NOTES: Reference category in parentheses; χ2(16) 470.56; pseudo R2 = .1174; pseudo R2 = (Lo-Lm)/Lo, where Lo is the value of the likelihood-ratio χ2 statistic for the model with intercept only, and Lm is the value for the model in which other variables have been included. a p ≤ .01 (two-tailed test). unexpected and may reflect several methodological problems in measuring accessibility in this way. The nearest SDP may provide poor-quality services and, thus, may not be the clinic the woman would use. Moreover, women may choose to travel to a source that is not in their immediate surroundings to ensure that their use of contraception is kept private (Committee on Population, 1991). District rural literacy and whether the woman's household had electricity are second only to number of family planning SDPs in having strong positive associations with contraceptive use. Women who lived in districts where at least 50 percent of the rural population was literate were more than 2.5 times as likely to use contraception than women who lived in rural areas with lower levels of literacy, regardless of their own educational attainment. There is also an individual-level education effect. Women with one to 6 years of schooling were more likely to use contraception than women with no schooling. Women with more than 7 years of education were even more likely to do so. Njogu (1991), in a study of the individual-level factors

OCR for page 139
Population Dynamics of Kenya related to contraceptive use, also showed that educational level was strongly associated with contraceptive use in the KDHS and KFS. With regard to household electricity, a woman living in an electrified home was more than 2.5 times as likely to use contraception than a woman living in a house without electricity. This variable is probably a reflection of level of household income. Another income-related variable, household flooring, also has a significant association with contraception. Several variables that are generally associated with exposure to modern ideas and values show a positive effect on contraceptive use. A woman who belongs to a women's group, listens to the radio at least once a week, or lives in a district with at least 25 kilometers of paved road per 10,000 people is significantly more likely to use contraception than a woman without one or more of these characteristics. Similar results were found in a study by Hammerslough (1991b) in which he used community-level data from the Kenya Community Survey and KDHS individual-level survey data to test the association between membership in a women's group and contraceptive use. Focus group discussions, reported in this same study, indicated that women's groups served as an interface between their members and modern sector organizations. It is suggested that this relationship fosters contact with family planning associations, as well as providing support for using contraception. It is often postulated that increased urbanization also exposes women to modern values and institutions, and thus should have a positive association with contraceptive use. In this analysis, both measures of urbanization—whether the woman lives in an urban or rural area, and whether she lives in a district that is more than 6 percent urban—have significant negative effects on contraceptive use. It may be that once factors associated with urban residence, such as income level, accessibility to family planning services, and modernization indicators, are controlled, there is nothing about urbanization itself that is significant in increasing contraceptive use and in fact it can be a negative influence. Another explanation may be that in Kenya the distinction between urban and rural residents is often ambiguous. There is significant movement of people back and forth between rural and urban areas, with people often maintaining close ties to their rural communities rather than developing permanent ties in the urban area where they have gone to work. In such a case, the urban-rural dichotomy, with people living in urban areas having different and "more modern" values than people in rural areas, may not hold. The percentage of the female population employed in the modern sector of each district was also tested for its relationship to individual contraceptive use. As with urbanization, it was expected that women living in a

OCR for page 139
Population Dynamics of Kenya district with more women employed in the modern sector would be exposed to different values and institutions, as well as there being a trade-off between raising children and working outside the home. Living in a district with at least 5.5 percent of the female population employed in the modern sector did not have a significant relationship to contraceptive use. However, this factor was the most significant variable in determining whether a contraceptor used a modern or traditional method. (The results from this logistic regression are not shown.) Female employment was also the most significant factor in determining whether a woman would use a modern contraceptive versus no contraceptive (results also not shown). Finally, the effects of two other individual-level variables were tested: age and religion. Women age 25 years and older are more likely to use contraception than younger women. But the estimated probabilities of use for age groups 25–39 and 40 or over did not differ significantly from each other: Women in both groups were more likely to use contraception than women age 15–24, but women age 25–39 were not more likely to use contraception than women age 40 or more, and vice versa. This finding is consistent with the results in Chapter 4, which indicate that fertility declines occurred in both middle and later age groups. Religion is not significantly associated with contraceptive use. This finding is not unexpected, given the steep fertility declines during the 1980s in the Coast Province, an area that is largely Muslim (see Chapter 4). Other variables not measured in this model that could affect contraceptive use are cultural norms valuing high fertility and child mortality. A contraceptive prevalence differentials study (Population Studies and Research Institute, 1991) of six districts conducted in 1990 noted that these two factors were associated with areas of low to medium contraceptive prevalence. Survey respondents, when asked about the value of children, noted that children provided additional household income as well as psychological satisfaction. Women also mentioned the costs of children, indicating that feeding, clothing, and educating children were expensive. In addition, the availability of future economic opportunities, particularly for men, was cited as a concern. Although many of the variables examined in this multivariate analysis are significant, it is important to note that the model as a whole explains little of the variation in the probability of using contraception. The pseudo R2 is .12. This result supports our earlier hypothesis in Chapter 4 that the declines in fertility seem to stem from a central force and not necessarily from individuals or regions with particular development or other characteristics. This result is consistent with the weak linkage between fertility and socioeconomic variables at the district level.

OCR for page 139
Population Dynamics of Kenya SUMMARY This chapter explores relationships between socioeconomic factors and changes in fertility, mortality, and contraceptive use. The declines in child mortality were strongly associated with female education and adult literacy at the district level. Little relationship was found to the availability of district-level health services. Examination of associations between the declines in fertility and several socioeconomic factors here and in Chapter 4 revealed very weak relations to education, urbanization, and child mortality. Employment in the modern sector showed a stronger relationship at the district level and was the only developmental indicator that was related to the fertility declines in a significant way. Multivariate analysis of contraceptive use highlighted the strong relationship between individual-level contraceptive use and the number of family planning service delivery points in a district. Other significant associations were found with district rural literacy, individual education, household electricity and type of flooring, district road density, membership in a women's organization, and weekly radio listening. Female employment in the modern sector was the strongest indicator in determining whether a contraceptor used a modern versus a traditional method. Caution must be taken in interpreting the results from these models, however, because only a small part of the variation in individual contraceptive use has been explained.

OCR for page 139
Population Dynamics of Kenya Appendix Tables 7A-1A and 7A-1B follow on pages 164–167.

OCR for page 139
Population Dynamics of Kenya TABLE 7A-1A Socioeconomic Indicators for Districts—Education and Employment Indices   Education Employment Province and District Women Age 15–44 with No Schooling, 1979 (%) Adults Literate, 1981–1988a (%) Index of Female Employment in Modern Sector, 1980b (%) Index of Male Employment in Modern Sector, 1980c (%) Nairobi 17.4   52.0 109.1 Central         Kiambu 24.3 71 23.3 40.2 Kirinyaga 36.9 61 12.8 14.9 Muranga 26.8 66 11.7 22.9 Nyandarua 33.4 65 11.6 20.4 Nyeri 19.7 68 14.7 21.5 Coast         Kilifi 84.9 32 6.1 17.0 Kwale 82.1 46 4.3 17.8 Lamu 75.1   5.0 24.2 Mombasa 40.5   31.9 81.1 Taita Taveta 38.1 60 11.6 43.5 Tana River 84.3   4.1 12.1 Eastern         Embu 38.4 51 7.0 17.5 Isiolo 84.7   12.5 18.5 Kitui 64.2 51 3.3 9.3 Machakos 31.1 56 5.4 14.6 Marsabit 94.3   3.2 10.6 Meru 46.5 50 5.9 12.1 Northeastern         Garissa 95.4   1.4 8.6 Mandera 97.7   0.4 4.5 Wajir 97.4   0.4 3.8

OCR for page 139
Population Dynamics of Kenya   Education Employment Province and District Women Age 15–44 with No Schooling, 1979 (%) Adults Literate, 1981–1988a (%) Index of Female Employment in Modern Sector, 1980b (%) Index of Male Employment in Modern Sector, 1980c (%) Nyanza         Kisii 38.5 47 3.8 10.8 Kisumu 35.7 49 11.8 30.2 Siaya 49.9 44 2.2 6.8 South Nyanza 49.6 39 3.4 8.1 Rift Valley         Baringo 59.9 38 8.6 14.0 Elgeyo Marakwet 49.7 36 5.0 8.8 Kajiado 70.8 25 7.8 12.5 Kericho 49.0 46 11.7 40.1 Laikipia 45.4 46 20.4 40.4 Nakuru 38.8 52 22.3 48.5 Nandi 42.2 48 10.7 34.4 Narok 83.7 30 3.5 8.1 Samburu 91.9   2.6 20.3 Trans Nzoia 47.8 52 16.4 30.6 Turkana 96.3   0.6 4.6 Uasin Gishu 41.4 54 23.4 36.8 West Pokot 88.0 42 2.8 6.6 Western         Bungoma 35.3 52 4.8 11.3 Busia 56.0 42 6.0 7.9 Kakamega 40.3 51 5.3 16.4 a Percentage of adults (both sexes) literate: average of 1981 and 1988 surveys. b Percentage of female employment in the modern sector, 1980: numbers of women employed in modern sector from 1980 employment survey divided by women aged 15–44 years at 1979 census. c Percentage of male employment in the modern sector, 1980: numbers of men employed in modern sector from 1980 employment survey divided by men aged 15–44 years at 1979 census.

OCR for page 139
Population Dynamics of Kenya TABLE 7A-1B Socioeconomic Indicators for Districts—Population Density, Health, and Economic Indices   Health Economic Province and District Population Density, 1979 (per km2 Outpatients with Malaria, 1976–1984 (%) Children Wasted, 1982a (%) Children Stunted, 1982b Index of Health Facilitiesc Households with No Piped Water, 1982 (%) Nairobi 1,210           Central             Kiambu 280 6.5 1.2 17.5 577 75.0 Kirinyaga 202 7.9 1.9 24.5 329 77.4 Muranga 261 12.9 4.2 24.8 623 76.2 Nyandarua 66 3.6 2.0 12.4 490 82.4 Nyeri 148 5.4 3.0 18.5 525 78.6 Coast             Kilifi 34 25.4 5.1 42.1 1,203 74.7 Kwale 34 24.6 4.9 38.5 1,035 86.0 Lamu 6 20.2 5.1 42.1 240 74.7 Mombasa 1,622 23.4     178   Taita Taveta 8 28.6 4.7 14.7 648 80.5 Tana River 2 28.6 5.1 42.1 905 74.7 Eastern             Embu 96 21.8 2.0 22.3 504 89.1 Isiolo 1 24.4     436   Kitui 15 24.5 1.8 30.0 1,974 100.0 Machakos 72 21.3 2.9 23.1 1,526 95.3 Marsabit 1 17.0     2,023   Meru 83 18.1 3.3 16.8 1,265 68.4 Northeastern             Garissa 2 21.1     2,069   Mandera 3 24.3     4,151   Wajir 2 20.6     2,860  

OCR for page 139
Population Dynamics of Kenya   Health Economic Province and District Population Density, 1979 (per km2) Outpatients with Malaria, 1976–1984 (%) Children Wasted, 1982a (%) Children Stunted, 1982b (%) Index of Health Facilitiesc Households with No Piped Water, 1982 (%) Nyanza             Kisii 395 32.3 5.0 33.1 690 99.0 Kisumu 230 30.9 3.4 19.8 482 87.5 Siaya 188 29.7 6.3 36.6 711 98.1 South Nyanza 143 32.2 1.5 25.3 973 99.5 Rift Valley             Baringo 20 10.9 6.4 19.4 566 93.0 Elgeyo Marakwet 65 12.7 2.1 18.6 179 98.1 Kajiado 7 16.5 2.5 19.8 699 93.2 Kericho 161 17.5 3.0 18.1 435 85.7 Laikipia 13 9.4 6.4 19.4 617 93.0 Nakuru 90 9.6 2.3 34.5 832 85.8 Nandi 109 22.8 3.3 12.1 364 79.6 Narok 13 15.0 2.5 19.8 842 93.2 Samburu 4 16.9     784   Trans Nzoia 124 20.1 2.8 19.1 645 86.8 Turkana 2 19.7     904   Uasin Gishu 89 19.3 2.7 17.8 412 95.2 West Pokot 17 24.8 2.1 18.6 671 98.1 Western             Bungoma 163 38.5 2.0 24.7 1,042 89.3 Busia 183 31.6 2.1 21.1 658 99.3 Kakamega 294 33.0 2.0 26.7 1,489 96.5 a Percentage of children wasted (i.e., low height for age): less than 80% of WHO standard (1982 Nutrition Survey). b Percentage of children stunted (i.e., low weight for height): less than 90% of WHO standard (1982 Nutrition Survey). c Index of health facilities: persons per health facility (hospitals, health centers, clinics, and dispensaries) multiplied by the square root of the land area.