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 51
Population Dynamics of Kenya 4 Fertility Trends Although reports of births in recent periods preceding the census and surveys were collected, they are subject to biases of coverage and time location errors. Adjustments can be made but not with any certainty. It is simpler to examine the trends in fertility through the ratios of children ever born to women in age groups. These measures are shown in Table 4-1 for the three censuses (1962, 1969, 1979), the Kenya Fertility Survey (KFS), the Kenya Contraceptive Prevalence Survey (KCPS), and the Kenya Demographic and Health Survey (KDHS). Values are given separately for the first and third rounds of the National Demographic Survey (NDS) (1977 and 1983); the data from the second round have not been closely analyzed. Panel A of the table gives the mean parities, that is, the total children born divided by the total women in the age group; panel B shows the percentage of women in each age group reporting no children; panel C presents mean births to mothers (instead of all women). On a first inspection of panel A, the pattern of change over time is clear enough. The mean parities of the younger women show little change up to 1984. The levels for the older women increase steadily until the late 1970s. However, the large-scale operations—the NDS and the 1979 census—give lower mean parities for women over 40 years of age in comparison with the KFS and KCPS, which were close to them in time. These differences suggest that there was underreporting of births at the census and the NDS of 1983. It seems likely that there was at least as much underreporting at the earlier censuses, and this conclusion is supported by the configuration. The mean parity of 5.07 for women aged 35–39 in 1962 rises by about 1.5 children for the same
OCR for page 52
Population Dynamics of Kenya Table 4-1 Birth Measures from Censuses and Surveys Age Group of Women 1962 Census 1969 Census 1977 NDS 1978 KFS 1979 Census 1983 NDS 1994 KCPS 1989 KDHS A. Mean Parities per Woman 15–19 0.36 0.35 0.33 0.35 0.32 0.29 0.35 0.28 20–24 1.65 1.88 1.83 1.84 1.85 1.75 1.96 1.58 25–29 3.01 3.65 3.72 3.76 3.65 3.56 3.96 3.47 30–34 4.20 5.11 5.55 5.55 5.38 5.36 5.70 5.01 35–39 5.07 6.00 6.67 6.82 6.47 6.66 7.04 6.48 40–44 5.61 6.44 7.25 7.59 7.02 7.43 7.84 7.36 45–49 5.90 6.69 7.46 7.88 7.17 7.65 8.15 7.63 B. Percentage of Childless Women 15–19 79.1 75.5 77.3 73.9 78.4 77.8 73.2 78.6 20–24 36.8 24.7 24.2 19.1 26.6 22.9 19.7 21.5 25–29 22.3 11.1 7.6 5.4 10.9 7.3 4.9 5.3 30–34 17.8 8.2 4.6 3.3 7.9 3.9 4.2 2.9 35–39 15.3 7.6 5.1 1.6 7.0 3.6 2.6 2.2 40–44 14.4 7.8 4.1 3.4 7.5 2.9 3.1 2.3 45–49 13.7 8.0 4.6 2.8 7.6 2.8 3.3 2.8 C. Mean Births per Mother 15–19 1.71 1.45 1.45 1.34 1.48 1.32 1.31 1.31 20–24 2.61 2.50 2.41 2.27 2.52 2.27 2.44 2.01 25–29 3.87 4.11 4.02 3.98 4.10 3.84 4.16 3.66 30–34 5.11 5.56 5.82 5.74 5.85 5.58 5.95 5.16 35–39 5.99 6.50 7.03 6.93 6.96 6.90 7.23 6.63 40–44 6.55 6.99 7.56 7.86 7.59 7.65 8.09 7.53 45–49 6.84 7.26 7.82 8.11 7.76 7.87 8.43 7.85
OCR for page 53
Population Dynamics of Kenya cohort by 1969, too large an increase to be accounted for by additional births at these late ages. Part of the discrepancy can be attributed to the inclusion of women who did not report numbers of children with childless women at the censuses and NDS. The lower percentages of childless women in the 1977 and 1978 surveys at ages beyond which first births rarely occur (panel B), compared with the measures in 1969 for the same cohorts, must be due to error. This source of bias is eliminated in panel C in the measures of mean births per mother because women who did not report children are not included in the measure. However, the trends remain similar to those of panel A, although they are reduced in magnitude. There remains uncertainty about whether the apparent increase in fertility in the 1960s and 1970s was real or due entirely to improved birth reporting. On balance, the evidence suggests that there was an increase in fertility. The relatively lower estimates from the 1948 census analysis were supported by some small surveys, and the mean births per mother by age group in 1962, 1969, and 1979 are consistent with a gradual rise in fertility for cohorts born in the 1930s. A similar conclusion was reached from the detailed analysis of the KFS birth histories (Henin et al., 1982). More recent attempts to examine fertility trends in Kenya use the parity progression ratio (PPR; the proportion of women going from an nth to an (n + 1)st birth). This index is a sensitive measure of fertility change but is very robust to data errors, for example, in time location of births and confusion between childlessness and failure to report. These ratios can be calculated directly for women past the ages of childbearing. In practice, the end of childbearing can be taken as 40 years because births after that age have a negligible effect on the PPRs except at very high birth orders. From the Kenya censuses, PPRs for age groups of women were calculated by Feeney (1988), up to 70–74 years of age for 1962 and 1979, and up to 60–64 years for 1969. He plotted the PPRs of each order on a time scale of year of birth of the woman so that the measures for the same cohorts at the three censuses were at coincident points on the horizontal axis. Blacker, in an unpublished note, extended Feeney's work by adding PPRs from 1969 and 1979 reports of births to women still in the childbearing period. The extrapolation to the end of reproduction was based on the fertility rates by birth order calculated from the births reported for the 12 months before the censuses. The method was developed by Brass (1985). Blacker also added measures from older women (more than 46 years of age) at the 1948 census and from those aged 45–49 years at the KFS and the KCPS. His graphs are reproduced in Figure 4-1. The picture that emerges from the investigation of these measures is a coherent one. The comparable PPRs are higher for each successive census but only slightly for 1979 versus 1969. The 1962 values vary rather erratically over time, but the 1969 and 1979 trends are more regular. At every
OCR for page 54
Population Dynamics of Kenya Figure 4-1 Kenya parity progression ratios.
OCR for page 55
Population Dynamics of Kenya
OCR for page 56
Population Dynamics of Kenya
OCR for page 57
Population Dynamics of Kenya birth order there is a steady decline in the PPRs that extends from the 1970s back as far as the records go, that is, to women born around the beginning of the century. The pattern of the PPRs confirms that there was improved reporting in the censuses. It is possible that better reporting by younger women than by older contributed to the apparent rise in the PPRs over time. On the other hand, the closeness of the 1969 measures to the 1979 values for the same cohorts reported when 10 years older makes this possibility implausible. The extra PPRs derived by Blacker from the reports by younger women at the 1969 and 1979 censuses refer to more recent time periods. They suggest a cessation of the rise in the 1970s, and even possibly a decrease, but the synthetic nature of the calculations, projected by the use of current fertility reports, must signal caution. Blacker pointed out that the steepness of the rise in the PPRs between 1920 and 1970 increases with parity. This pattern casts doubts on the conclusion that the fertility increases were due mainly to decreases in pathological sterility. Blacker suggested that the main causes may have been biosocial factors, such as shorter birth intervals due to declines in breastfeeding and postpartum abstinence, along with lengthening reproductive lives. However, there are virtually no data on the proximate determinants of fertility prior to the late 1970s. The surveys between 1977 and 1984 provide mean parities by age group of women that fluctuate but reveal no clear trend (Table 4-1, panel A). The sample errors, particularly for the older women, are appreciable, and the variations can be accounted for by chance and by small systematic biases in coverage and age reporting. The deduction is that the tendency for fertility to increase had ended. Of course, women at the end of their reproductive periods around 1980 had their peak childbearing in the 1960s. The substantial rise in fertility was probably a product of factors in the 1940s and 1950s, a finding that agrees with the analysis of the KFS birth histories (Henin et al., 1982) and an unpublished analysis in 1981 by the Panel on Tropical Africa of the Committee on Population and Demography. In contrast to the surveys bracketing 1980, the 1989 KDHS gives mean parities that are distinctively lower. The striking reductions are for the younger age groups of women. Thus, the 2.01 and 3.66 mean births per mother at the KDHS for age groups 20–24 and 25–29 years, respectively (Table 4-1, panel C), are below all the other levels in the series from 1962 onward. The implication is that a sharp decline in fertility occurred in the 1980s. The 1983 round of the NDS is consistent with this conclusion, with slightly lower mean parities than the 1977 first round and the 1978 KFS for women age 20–34 years. The measures from the 1984 KCPS are contradictory, however, since they are the highest of the whole series. Nevertheless, the authors of the KCPS report discerned a slight decrease in current fertility from the KFS level on the basis of recent births. The evidence of
OCR for page 58
Population Dynamics of Kenya fertility reduction from the KFS and KDHS comparison is examined in detail below. EVIDENCE OF DECLINES IN FERTILITY FROM RECENT SURVEYS There have been many analyses of the KFS data. The main aim here is to establish a base for the study of subsequent fertility change in the 1980s. This brief exposition follows closely the approach in the report by Henin et al. (1982) published by the World Fertility Survey program. However, for the convenience of later comparisons, the birth measures are arranged by cohorts of women in time periods rather than in time periods by age groups of women. In the latter form, the age-specific rates are for the conventional five-year groups; in the former, for the intervals in which a cohort of women moves from one age group to the next (e.g., from 20–24 to 25–29 years). Table 4-2A shows the age-specific birth rates for cohorts of women, denoted by their ages at the time of the KFS, in five-year time intervals before the survey. The cumulated values of these rates from the beginning of childbearing up to five yearly time points are given in Table 4-2B. These represent what the mean parities would have been for the same women reporting in the same way at surveys 5 years, 10 years, and so on previously. Summation of the rates within time periods gives the cumulated fertilities up to the highest age groups of women reporting; these are increasingly truncated as the past recedes. The measures for the preceding four time periods are shown in Table 4-2C. Comparison of the cumulated fertilities at equivalent ages along the top diagonal suggests that there had been a quite remarkable trend, with a sharp increase at 10–14 years before the survey followed by a substantial reduction thereafter. Simple, approximate estimates of the fertility up to the 45–49 age group (very close to the total fertility) are obtained by assuming that the age-specific rates removed by the truncation are the same as for the women aged 45–49 years at the time of the survey. If, in fact, a decline had occurred, these extrapolations would lead to an under-rather than an over-estimation. The outcome is a total fertility of 9 births per woman in the 10–14 years before the survey, falling to 8.3 in the previous 5 years. The level indicated for 1963–1967 (10 to 14 years before KFS) is well above that deduced from census data near that time; the apparent trend is highly implausible and hard to explain in terms of changes in nuptiality, fertility control, or biological factors. Examination of the age-specific birth rates in Tables 4-2A and 4-2B reveals that the distributions change over cohorts. The cumulated totals to equivalent ages rise over time for cohorts up to 30–34 years but decrease for
OCR for page 59
Population Dynamics of Kenya Table 4-2A Births per 1,000 Women by Age Groups and Time Period, KFS 1978 Time Preceding Survey (years) Age at KFS 0–4 5–9 10–14 15–19 20–24 25–29 30–34 35+ Total Births 15–19 334 11 345 20–24 1,439 383 21 1,843 25–29 1,804 1,463 449 25 3,741 30–34 1,641 1,852 1,570 479 34 5,576 35–39 1,402 1,711 1,850 1,379 437 49 6,828 40–44 1,029 1,530 1,787 1,614 1,209 372 10 7,551 45–49 615 1,138 1,582 1,577 1,557 1,112 292 25 7,898 Total 8,264 8,088 7,259 5,074 3,237 1,533 302 25
OCR for page 60
Population Dynamics of Kenya Table 4-2B Cumulated Births per 1,000 Women by Age Group, KFS 1978 Time Preceding Survey (years) Age at KFS 0 5 10 15 20 25 30 15–19 345 20–24 1,843 404 25–29 3,741 1,937 474 30–34 5,576 3,935 2,083 513 35–39 6,828 5,426 3,715 1,865 486 40–44 7,551 6,522 4,992 3,205 1,591 382 45–49 7,898 7,283 6,145 4,563 2,986 1,429 317 Table 4-2C Cumulated Births per 1,000 Women by Time Period, KFS 1978 Time Preceding Survey (years) Age at KFS 0–4a 5–9 10–14 15–19 30–34 5,218 35–39 6,620 5,420 40–44 7,649 6,950 5,677 45–49 8,264 8,088 7,259 5,074 Extrapolated to 45–49 8,264 8,703 9,012 8,409 Adjusted 7,910 8,053 8,120 7,836 a Small undercount because no births recorded for women currently 10–14. Table 4-2D Births per 1,000 Women by Age Group and Time Period: Observed and Model Time Preceding Survey (years) Age at KFS Source 0–4 5–9 10–14 15–19 Total Births 30–34 Observed 1,641 1,852 1,570 479 5,576 Model 1,693 1,823 1,551 499 35–39 Observed 1,402 1,711 1,850 1,379 6,828 Model 1,387 1,653 1,778 1,513 40–44 Observed 1,029 1,530 1,787 1,614 7,551 Model 946 1,342 1,598 1,721 45–49 Observed 615 1,138 1,582 1,577 7,898 Model 344 947 1,342 1,599
OCR for page 61
Population Dynamics of Kenya older cohorts (when comparing along the diagonals in Table 4-2B). The variation is relatively slight up to the 35–39 age cohort, but sharp for the two oldest groups of women. Thus, for fertility up to age 30–34, 37 percent (2,083 divided by 5,576) was reported as before age 20–24 for the 30–34 age cohort, but only 31 percent (1,429 divided by 4,563) for the 45–49 age cohort. This difference might occur because of a move to earlier childbearing, but it seems unlikely and there is no evidence for it. A more convincing explanation (particularly since the same feature has occurred in several birth history surveys of populations with a low literacy level) is recording error in which the births are located too near the present time. Corrections are made by fitting the observations for cohorts aged 30–34 and 35–39 years by the relational Gompertz model (Brass, 1981). The two parameters of the model distribution are taken as rounded averages over the two cohorts. The observed age-specific rates for the two oldest cohorts are then replaced by the values from the model distribution, with their sums constrained to equal the observed cohort totals. As can be seen from Table 4-2D, the model values for the 30–34 and 35–39 cohorts are fairly close to the observed birth rates. Therefore, only the distributions for the two oldest cohorts have been adjusted. The total fertilities (to age 45–49 years), extrapolated from the adjusted measures in the same way as for the observed measures, are also given in Table 4-2C. The fluctuation in total fertility over the 20 years 1958–1977 has now largely disappeared, leaving a roughly constant level of about 8 children per woman for that period. Similar exercises allowing for systematic changes in the distribution of fertility by age of woman have been carried out. The estimates differ in detail, but the general conclusions are the same. An alternative explanation of the time displacement of births in the reports of older women is age error. Perhaps, these women were on average younger than the recorded age span, and hence the cumulated fertilities were to earlier points of the reproductive period than specified. However, the bias would have to be very large, with women in the 45–49 years age group really about 3 years younger than reported relative to the 30–34 age group. Moving the average ages to bring the locations of the cohort time-period birth distributions into agreement would not bring them into near coincidence because those for the cohorts under age 40 have a more peaked shape than those for the older women. Nor, of course, would the time-period total fertilities be much changed, leaving the trend anomalies unexplained. Thus, the time-period shift seems to be much the most plausible explanation although there may easily be a contribution from systematic age biases. Tables 4-3A, 4-3B, and 4-3C present the fertility measures from the KDHS maternity histories in the same form as Tables 4-2A to 4-2D for the KFS. Thus, the age-specific birth rates for cohorts and time periods are in
OCR for page 81
Population Dynamics of Kenya increased use of contraception.2 It is also important to bear in mind that these calculations are based on birth records for cohorts collected at one survey, the KDHS, and are thus insensitive to data errors due to possible lack of comparability in the procedures and efficiency of the two surveys, the KFS and KDHS. The consistency of the findings from the two approaches is a cogent argument for their essential validity. FERTILITY DECLINES BY DISTRICT The declines in fertility show distinct regional differences, but the provinces are not sufficiently homogeneous to be accepted as the most effective aggregates for the study of patterns. Unfortunately, the information for analysis at the district level is limited by the sample sizes of the KFS and the KDHS. When tabulations from the 1989 census of women by age and the number of children born to them are available by district, it will be possible to compare the measures with the corresponding data from the 1979 census. Although problems from the underreporting of births in the censuses by the older women will still remain, it is probable that satisfactory allowances can be made and the trends estimated. The direct measures of current fertility based on the census reports of births in the preceding year are subject to substantial error and cannot be used without adjustments. The fertility declines in Kenya between the two censuses should be large enough for the effects to be evident in the mean parities, although the magnitudes of these cohort changes will not be the same as for the time-period movements. In the KDHS there was some oversampling to provide larger numbers of households in selected districts. Tabulations of births by age group of women and 5-year calendar periods have been made for the 16 districts with the largest sample sizes. The smallest group included is 167 women for Nakuru; the largest excluded, 111 for Kiambu. Nairobi has already been included among the provinces in earlier analysis. Our evaluation of birth reporting at the KDHS concluded that the fertility calculated for the period 10 to 14 years before the survey was in close agreement with the corresponding measure from the 5 years preceding the KFS. Accordingly, the change at the district level is estimated from the comparison of the cumulated rates for women up to age 40 in the 5 years preceding the KDHS with those from 10–14 years before. The upper limit of 40 years is necessary because of the truncation of maternity histories with increasing time in the past. 2 The role of contraceptive use as a proximate determinant of Kenyan fertility is examined more thoroughly in the next chapter.
OCR for page 82
Population Dynamics of Kenya Table 4-10 presents the percentage declines in fertility as measured for selected districts and also for provinces. For the latter, the reductions found previously from the comparison of KDHS and KFS estimates (Tables 4-6A to 4-6C) are given in the last column and can be compared to KDHS estimates in the first column. The bases of the two sets of province indices differ in the age ranges of women (less than 40 and up to 50 years), as well as the source and calendar period of the earlier fertility measure (1974–1978 from the KDHS and 1973–1977 from the KFS). Nevertheless the agreement is rather satisfactory except for Nairobi. The discrepancy for the capital is probably due to the less stable nature of the population, leading to both erratic change and reporting errors. In Central, Coast, Eastern, and Western provinces, the fertility declines calculated solely from KDHS data are a few percentage points lower than the corresponding values from the KDHS-KFS comparison, and for Nyanza and Rift Valley slightly above. The agreement at the province level confirms that the measures of district change based on KDHS alone are valid, but the problem of sample variability remains. To provide some check, the changes in fertility from 5–9 years prior to the KDHS to 0–4 years before for women under 45 were also calculated and are shown in the middle column of the table. The estimates from the two series are not, of course, independent but provide a useful guide to the possible existence of major anomalies. The results are, generally, satisfyingly consistent. On average, the fertility declines from 1979–1983 to 1984–1988 were about two-thirds the corresponding percentages of 1974–1978 to 1984–1988. The largest discrepancies are for Kirinyaga, Kisii, and Uasin Gishu, where the second series suggests that the already large fertility decreases may have been underestimated, and for Bungoma and Kisumu where increases are indicated rather than small reductions. Thus the relative order of the changes is little affected. The one exception is Mombasa where a moderately large fertility decline in the first series becomes a small one in the second. The correlation coefficient between the rankings of the two series (Kendall's) is .68, which indicates a strong association. The selected districts in the Coast and Rift Valley record fertility declines that are close to the overall measures for their provinces. The district variations in Central Province are greater, although all the reductions are substantial; the 36.5 percent reduction for Kirinyaga is the largest of all the declines. In Nyanza, the fertility decrease of 13.5 percent conceals two small reductions in Kisumu and South Nyanza, as well as two large ones in Kisii and Siaya. The differential for Kisii is not particularly surprising because the district has several characteristics, including ethnic composition, that distinguish it from the rest of Nyanza; the result for Siaya is, however, unexpected. The two selected districts of Western Province have
OCR for page 83
Population Dynamics of Kenya TABLE 4-10 Declines in Cumulated Fertility by Province and District Decline (%) Province and District 1974–1978 to 1984–1988 Women Under 40 Yearsa 1979–1983 to 1984–1988 Women Under 45 Yearsa 1973–1977 to 1984–1988 Women Under 50 Yearsb Central 26.7 21.4 31 Kirinyaga 36.5 33.5 Muranga 20.6 12.4 Nyeri 24.2 18.6 Coast 21.0 22.6 27 Kilifi 19.3 15.7 Mombasa 19.8 5.2 Eastern 13.3 12.1 16 Machakos 11.8 6.9 Meru 21.7 15.1 Nyanza 13.5 8.4 13 Kisii 19.3 18.9 Kisumu 6.6 -1.2 Siaya 27.4 18.5 South Nyanza 6.6 6.3 Rift Valley 22.8 16.2 20 Kericho 22.3 11.4 Nakuru 23.3 18.7 Uasin Gishu 25.8 22.6 Western 1.3 -1.8 5 Bungoma 8.1 -4.8 Kakamega 10.2 4.8 Nairobi 14.09.1 26 a Estimated from births in time periods reported in the KDHS. b Estimated from births reported in the five years prior to the KDHS and KFS.
OCR for page 84
Population Dynamics of Kenya small fertility declines, and the province measure is even lower, implying that such is the case for the one district excluded, Busia. The selected districts comprise 17 of the 41 total but include more than two-thirds of the Kenyan population. Many of the excluded areas are geographically remote from the central area and are thinly populated, but those characteristics do not make their fertility changes less interesting—rather, the contrary is true. In particular, the results for Coast Province demonstrate that the declines for the districts not selected must on average have been a little greater than the overall provincial measure of 21 percent to balance the 19.3 percent for Kilifi and 19.8 percent for Mombasa. Of the remaining districts, three (Kwale, Lamu, and Tana River) have typical Coast Province features of low educational level and high child mortality with little improvement; the other, Taita, does not conform to these characteristics but includes only one-quarter of the population of the selected districts and slightly more than one-tenth of the province as a whole. A further check was applied by comparing the mean parities by age groups of women for the selected districts at the 1988–1989 KDHS with corresponding measures from the 1979 census. There is clear evidence of the underreporting of births by older women in 1979, but the mean parities for younger women are close on average to measures from the 1977–1978 KFS. Another reason for omitting the older women in the comparisons is the fact that a considerable proportion of their births took place before the fertility decline was clearly established. After investigation it was decided that only the first four 5-year age groups (i.e., 15–34 years) should be retained to provide a measure of fertility as a sum of the mean parities. Various alternative systems of weighing the mean parities were examined, but the resulting changes in the estimates of fertility declines from the 1979 census to the 1988–1989 KDHS were too small to justify the added complication. Table 4-11 gives the ratios of the fertility index from the 1988–1989 KDHS to the 1979 census for provinces and the selected districts. These are generally slightly lower than the measures presented earlier, probably because of relative underreporting of births at the census, but in Nyanza the discrepancy is in the opposite direction. The possible effects of age misstatements and sample errors should not be forgotten. It is arguable whether adjustments for the presumed birth reporting errors at the 1979 census should be made for districts, but on balance, their use seems to improve the measures of change. The adjustment made assumes that the proportional discrepancy between the 1979 census and the 1977–1978 KFS for a province also applies to districts in the province. The adjusted ratios are then translated into percentage declines in the fertility index. The last column repeats for comparison the district fertility reduction between 1974–1978 and 1984–1988 calculated from the KDHS birth histories. The province divergences
OCR for page 85
Population Dynamics of Kenya TABLE 4-11 Declines in Fertility Indices by Province and District Mean Parity Index Ratios 1998–1999 KDHS to 1979 Census (per thousand) Province and District Reported Adjusted Reduction Based on Adjusted % Cumulated Fertility Reduction 1974–1978 to 1984–1988 from KDHS (%) Central 889 879 12.1 26.7 Kirinyaga 815 806 19.4 36.5 Muranga 848 839 16.1 20.6 Nyeri 901 891 10.9 24.2 Coast 889 847 15.3 21.0 Kilifi 890 848 15.2 19.3 Mombasa 882 840 16.0 19.8 Eastern 904 898 10.2 13.3 Machakos 881 875 12.5 11.8 Meru 920 914 8.6 21.7 Nyanza 911 959 4.1 13.5 Kisii 830 874 12.6 19.3 Kisumu 854 899 10.1 6.6 Siaya 1,035 1,090 -9.0 27.4 S. Nyanza 916 964 3.6 6.6 Rift Valley 925 893 10.7 22.8 Kericho 980 946 5.4 22.3 Nakuru 666 643 35.7 23.3 Uasin Gishu 908 877 12.3 25.8 Western 981 951 4.9 1.3 Bungoma 969 939 6.1 8.1 Kakamega 985 955 4.5 10.2 Nairobi 915 759 24.1 14.0 (reported to adjusted) are only around 1 percent in Central and Eastern provinces, and 3 percent in Rift Valley and Western, but rise to 5 percent in Coast and Nyanza. The disagreement for Nairobi is much larger, but there are particular problems of obtaining accurate reports from this fluid population. As already noted, the declines in the mean parity index are not measuring the same effects as the direct time-period reductions. There should, however, be a reasonably consistent relationship. If Nairobi is excluded, the decreases in the mean parity index are overall roughly three-fifths of the change in the direct time-period fertility measures. For individual provinces there is a moderate fluctuation about this relation but no striking anomaly. The adjusted mean parity decline for Nairobi is well above the expected value, but the unadjusted is considerably below. Discrepancies at the district level are greater than at the province level in accord with expec-
OCR for page 86
Population Dynamics of Kenya tation, but the general order of the fertility declines is largely preserved. Thus, Nakuru and Kisumu record substantially greater fertility reductions in the mean parity ratios than in the time-period measures, but in the former case the shift is from high to the highest and in the latter from the lowest to moderately low. The divergences for Kericho and Machakos are of more concern because the former ranks well up for fertility reduction by the direct time-period measurement but near the lowest for the mean parity ratio, and the shift for the latter district is the reverse. The most striking anomaly of all is for Siaya, which shows the second highest decline in the cumulated fertility measure but a rise in mean parities (whether adjusted or not). It is clear that care has to be exercised in drawing conclusions about fertility change for these districts in which the evidence is contradictory. SUMMARY The evidence from censuses and surveys indicates that fertility fell in the late 1970s and the 1980s, probably with a more rapid rate of decline in the late 1980s. Total fertility decreased from approximately 8.2 births per woman in 1973–1977 to 6.7 births per woman in 1984–1988. The decline was notable in that it occurred in almost every subgroup. All age groups contributed to the decline, with the middle and later reproductive ages contributing the most. Fertility declined 23 percent in urban areas, followed closely by a decrease of 17 percent in rural areas. The greatest fertility reduction among the provinces occurred in Central Province (31 percent). The fertility decline in Coast Province, noted for its substantial Muslim population, was also high (27 percent). Western Province was an anomaly, with a small decline of only 5 percent, which may be more a result of the data than a reflection of what actually occurred. Fertility reductions in percentage terms were very similar regardless of women's educational level. Fertility declines occurred at all birth orders, a pattern that is very different from that observed in Latin America and Asia, where fertility declines started in the middle parities and moved successively to the higher and then lower birth orders (see Caldwell et al., 1992, for related discussion on the differences of fertility declines in Africa). An analysis of the patterns of decline among other sub-Saharan African countries revealed a pattern similar to Kenya in those countries experiencing fertility change.
OCR for page 87
Population Dynamics of Kenya APPENDIX TABLE 4A-1 The Proportion of Women Progressing to the Next Birth Within 5 Years, KDHS, All Women Parity Progression Ist–2nd 2nd–3rd 3rd–4th 4th–5th B60 unadjusted Cohort age 15–19 .9246 1.0000 .5625 .0000 20–24 .8744 .8832 .8122 .5952 25–29 .8926 .8975 .8638 .8697 30–34 .8747 .9081 .8907 .8893 35–39 .9056 .8934 .8877 .8818 40–44 .8966 .8961 .9049 .8598 45–49 .9121 .8917 .9008 .9001 Cohort age (5 years later) 20–24a .8808 .5362 .2779 .0000 25–29a .9080 .9198 .9199 .8203 30–34a .8762 .9264 .9188 .8889 35–39a .9082 .9005 .9048 .9052 40–44a .8974 .9047 .9167 .8688 45–49a .9137 .8975 .9038 .9039 Indices of relative change 20–24/25–29a 0.9630 0.9602 25–29/30–34a 1.0187 0.9687 0.9402 .9783 30–34/35–39a 0.9631 1.0085 0.9844 .9824 35–39/40–44a 1.0091 0.9875 0.9684 1.0149 40–44/45–49a 0.9813 0.9985 1.0012 .9512 B60 adjusted 20–24 .8533 .8247 25–29 .8861 .8589 .8083 .8352 30–34 .8699 .8866 .8597 .8537 35–39 .9032 .8792 .8733 .8690 40–44 .8951 .8903 .9018 .8562 45–49 .9121 .8917 .9008 .9001
OCR for page 88
Population Dynamics of Kenya Parity Progression 5th–6th 6th–7th 7th–8th 8th–9th 9th–10th B60 unadjusted Cohort age 15–19 .0000 .0000 .0000 .0000 .0000 20–24 1.0000 .0000 .0000 .0000 .0000 25–29 .9180 .9093 .8549 .3388 .0000 30–34 .8620 .8076 .8980 .7195 .6169 35–39 .8321 .8557 .8266 .8086 .8342 40–44 .8777 .7966 .7955 .7874 .7094 45–49 .8174 .7950 .8018 .7427 .6763 Cohort age (5 years later) 20–24a .0000 .0000 .0000 .0000 .0000 25–29a .4302 .1417 .0000 .0000 .0000 30–34a .8713 .8370 1.0000 .8769 .4020 35–39a .8484 .9241 .8362 .9540 .7319 40–44a .9005 .8369 .8771 .8361 .8164 45–49a .8303 .8391 .8202 .8231 .7891 Indices of relative change 20–24/25–29a 25–29/30–34a 1.0536 1.0864 30–34/35–39a 1.0160 0.8739 1.0740 0.7542 0.8428 35–39/40–44a 0.9240 1.0224 0.9424 0.9670 1.0218 40–44/45–49a 1.0570 0.9494 0.9699 0.9567 0.8989 B60 adjusted 15–19 20–24 25–29 .8546 .7326 30–34 .8112 .6744 .7871 .5183 .5236 35–39 .7984 .7717 .7329 .6871 .6213 40–44 .8640 .7548 .7777 .7106 .6080 45–49 .8174 .7950 .8018 .7427 .6763 a Truncated to give the same age range of births as for the next lower age group.
OCR for page 89
Population Dynamics of Kenya Figure 4A-1 Unadjusted and adjusted B60s for the parity progressions.
OCR for page 90
Population Dynamics of Kenya Fourth to Fifth Birth Fifth to Sixth Birth Sixth to Seventh Birth
OCR for page 91
Population Dynamics of Kenya Seventh to Eighth Birth Eighth to Ninth Birth Ninth to Tenth Birth
Representative terms from entire chapter: