4
Fertility
INTRODUCTION
The average growth rate of the population in sub-Saharan Africa has been around 3 percent per year since the 1960s, higher than in any other major region in the world. If these rates remain constant, the population will double in the next 23 years. The reason for Africa's rapid rate of population growth is well known: while mortality and fertility rates fell substantially in Latin America and Asia between 1965 and 1985, only mortality fell in sub-Saharan Africa; fertility remained relatively stable, well above the level required to replace the population.
The existing demographic regime in sub-Saharan Africa is unlikely to remain in effect for very long. According to data from the Demographic and Health Surveys (DHS), the level of fertility has already fallen by more than 10 percent in seven sub-Saharan African countries (Kenya, Botswana, Senegal, Zimbabwe, Zambia, Burundi, and Malawi). Not all of these fertility declines are completely plausible, but fertility certainly appears to have fallen in at least three DHS countries at the national level (Kenya, Botswana, and Zimbabwe), as well as in South Africa, for which DHS-type data suggest a substantial decline. Furthermore, fertility declines appear to have occurred in several other countries at the subnational level (e.g., southwestern Nigeria, southern Namibia, northern Tanzania). These changes were documented in a chapter on fertility trends in sub-Saharan Africa in a recent volume by the Committee on Population (see Cohen, 1993).
This chapter examines levels, trends, and determinants of fertility in Senegal. The next section combines recent census and survey results to build a general description of the pattern of fertility in Senegal and how it has changed over the last 20 years. This is followed by a section on explanations for these patterns, focusing on marriage, breastfeeding, and contraception, some of the most important determinants of fertility change.
LEVELS AND TRENDS OF FERTILITY
Overall Levels and Trends
The standard measure of fertility used here is the total fertility rate (TFR).1 Despite dramatic improvements in the quantity and quality of population data for Senegal since the early 1960s, our knowledge and understanding of fertility levels and trends in the country still relies mainly on findings from ad hoc single-round surveys. Since 1960, five surveys have furnished information that permits the estimation of fertility in Senegal at the national level: the Demographic Survey (DS) of 1960-1961 [1]; the National Demographic Survey (NDS) of 1970-1971 [2]; the World Fertility Survey (WFS) of 1978 [3]; and two rounds of the DHS (DHS-I, DHS-II) in 1986 [4] and 1992-1993 [7].2 The data from each of these sources are not all of comparable quality. For example, the 1960-1961 DS data are generally thought to have suffered from an underenumeration of births and deaths (Cantrelle et al., 1986) and to be of lower quality than the data from the other sources.
In addition to these surveys, fertility estimates are available from two censuses, conducted in 1976 [17] and 1988 [18]. The 1976 census questionnaire did not collect information on births, so fertility estimates for 1976 have been derived from the age distribution using stable population theory. The 1988 census contained a single question on births in the last 12 months. Unfortunately, because of mistakes in the coding and processing of the data, the TFR estimates produced at the time the official report of the 1988 census was being prepared were implausibly low. Consequently, fertility estimates were not included in official reports on the 1988 census, and to this day, fertility estimates from the 1988 census remain unpublished. In this report we have remedied the situation by adopting a simple correction procedure to estimate total and age-specific fertility rates (ASFRs) for 1988 that takes these data problems into account. (See Appendix B for a detailed explanation of the procedure used.)
Finally, there are a number of subnational studies [8-16] in which accurate birth registers have been maintained over several years. Although these small study areas are not representative of the entire country, they yield fertility estimates that may be more accurate than those derived from large-scale surveys and censuses. Hence, small-scale studies can provide a useful independent check on the accuracy of the other estimates.
In all, seven observations of the national TFR are available for the period 1960-1993. In principle, by putting these observations in sequence we can observe the trends in fertility over the last 30 years. However, since the quality of the data varies among the surveys, some estimates are considerably more reliable than others. Tests of internal and external consistency show that the quality of the data from several sources is often mediocre to poor. Furthermore, the methodology used to collect the information was not the same across all surveys. For example, the WFS and the DHS collected complete birth histories from all women, while the 1988 census only asked women about births in the last 12 months. This latter method is more prone to omission of births, particularly of children who died, so that it almost invariably produces lower estimates of the TFR.3 One needs to bear in mind such issues of varying data quality and methodology when comparing across survey instruments, and to remain reasonably cautious when interpreting small fluctuations in the overall rate.
Table 4-1 provides a summary of the available estimates of the TFR for the period 1960-1992 at the national level and for various subpopulations. Figure 4-1 shows national-level TFR estimates for the age range 15-34 (this age range has been used so that estimates of fertility from the three birth history surveys, WFS, DHS-I, and DHS-II can be used for periods up to 15 years prior to the survey dates) plotted against calendar time. Figure 4-2 shows the national age patterns of fertility, as the proportion of total fertility contributed by each age group.
The Brass P/F ratio technique—where by current fertility rates are cumulated and compared with reported lifetime fertility—provides a check of the internal consistency of each survey and a sensitive test for changing fertility. For cross-sectional surveys where fertility rates have not changed in the recent past, average parity equivalents (F) should be roughly equivalent to reported average parities (P). If fertility is rising, lifetime fertility will be lower than cumulated current fertility, particularly at higher age groups which have had more time for differences to emerge, and P/F ratios will fall with age. If fertility is falling, on the other hand, the opposite will be the case, and P/F ratios will rise with age. Figure 4-3 shows P/F ratios for the 1960-1961, 1970-1971, 1978, 1986, and 1992-1993 surveys.
The fertility estimates shown in Figure 4-1 come from different sources and data collection methodologies. The 1960-1961 DS [1] included questions on births and deaths over the preceding 12 months, as well as on live births and child survivorship during a woman's entire life. The TFR based on data for births during the preceding 12 months was estimated to be 5.4 children per woman. This figure may have underestimated the true TFR as a result of the omission of certain vital events.4 However, fertility rates and age-specific parties for the 1960-1961 DS are fairly consistent, at least above age 25 (see P/F ratios in Figure 4-3). If the P/F ratios are used to
TABLE 4-1 Levels and Trends in the Total Fertility Rate, 1960-1992
Year |
Source |
Subgroup |
TFR |
National Level |
|
|
|
1960-1961 |
DS |
n.a. |
5.4 |
1970-1971 |
NDS |
n.a. |
6.4 |
1976 |
Census |
n.a. |
7.0 |
1975-1978 |
WFS |
n.a. |
7.2 |
1983-1986 |
DHS-I |
n.a. |
6.6 |
1987-1988 |
Census |
n.a. |
5.9 |
1989-1992 |
DHS-II |
n.a. |
6.1 |
WFS and DHS Regions |
|
|
|
1975-1978 |
WFS |
West |
7.1 |
|
|
Center |
7.3 |
|
|
Northeast |
7.2 |
|
|
South |
7.4 |
1983-1986 |
DHS-I |
West |
5.9 |
|
|
Center |
7.1 |
|
|
Northeast |
6.6 |
|
|
South |
7.0 |
1987-1988 |
Census |
West |
5.5 |
|
|
Center |
6.0 |
|
|
Northeast |
6.0 |
|
|
South |
6.4 |
1989-1992 |
DHS-II |
West |
5.6 |
|
|
Center |
6.4 |
|
|
Northeast |
6.6 |
|
|
South |
6.5 |
Administrative Regions |
|
|
|
1967-1971 |
Ferry (1976) |
Cap-Vert |
6.6 |
1975-1978 |
WFS |
Dakar |
6.8 |
1983-1986 |
DHS-I |
Dakar |
5.5 |
1987-1988 |
Census |
Dakar |
5.0 |
|
|
Thiès |
6.2 |
|
|
Saint-Louis |
6.2 |
|
|
Tambacounda |
5.9 |
|
|
Louga |
5.7 |
|
|
Diourbel |
5.6 |
|
|
Fatick |
6.9 |
|
|
Kaolack |
6.1 |
|
|
Ziguinchor |
6.8 |
|
|
Kolda |
6.3 |
1989-1992 |
DHS-II |
Dakar |
4.9 |
Small-Scale Studies [8-16] |
|
|
|
1963-1965 |
Cantrelle et al. (1980) |
Niakhar |
6.8 |
1984-1990 |
Project Niakhar (1992) |
Niakhar |
7.8 |
Year |
Source |
Subgroup |
TFR |
Small-Scale Studies [8-16] |
|
|
|
1963-1965 |
Cantrelle et al. (1980) |
Paos Koto |
6.6 |
1981-1991 |
Pison and Desgrées du Loû (1993) |
Bandafassi |
6.2 |
1985-1992 |
Lagarde et al. (forthcoming) |
Mlomp |
5.0 |
Place of Residence |
|
|
|
1975-1978 |
WFS |
Dakar |
6.8 |
|
|
Other urban |
6.5 |
|
|
Rural |
7.5 |
1983-1986 |
DHS-I |
Dakar |
5.5 |
|
|
Other urban |
5.6 |
|
|
Rural |
7.3 |
1987-1988 |
Census |
Dakar |
4.8 |
|
|
Other urban |
5.6 |
|
|
Rural |
6.4 |
1989-1992 |
DHS-II |
Dakar |
4.9 |
|
|
Other urban |
5.4 |
|
|
Rural |
6.8 |
Ethnic Groups |
|
|
|
1975-1978 |
WFS |
Wolof |
7.2 |
|
|
Poular |
6.9 |
|
|
Serer |
7.9 |
|
|
Mandingo |
8.1 |
|
|
Diola |
6.3 |
|
|
Other |
7.0 |
1983-1986 |
DHS-I |
Wolof |
6.4 |
|
|
Poular |
6.4 |
|
|
Serer |
7.6 |
|
|
Mandingo |
6.8 |
|
|
Diola |
6.2 |
|
|
Other |
6.3 |
1987-1988 |
Census |
Wolof |
5.7 |
|
|
Poular |
6.1 |
|
|
Serer |
6.3 |
|
|
Mandingo |
6.0 |
|
|
Diola |
5.9 |
|
|
Other |
5.7 |
1989-1992 |
DHS-II |
Wolof |
5.6 |
|
|
Poular |
6.4 |
|
|
Serer |
7.2 |
|
|
Mandingo |
5.7 |
|
|
Diola |
5.7 |
|
|
Other |
6.5 |
Year |
Source |
Subgroup |
TFR |
Level of Education 1975-1978 |
WFS |
None |
7.4 |
|
|
Primary |
7.1 |
|
|
Secondary+ |
3.6 |
1983-1986 |
DHS-I |
None |
7.0 |
|
|
Primary |
5.7 |
|
|
Secondary+ |
3.6 |
1987-1988 |
Census |
None |
6.2 |
|
|
Primary |
5.7 |
|
|
Secondary+ |
4.1 |
1989-1992 |
DHS-II |
None |
6.6 |
|
|
Primary |
5.7 |
|
|
Secondary+ |
3.8 |
NOTES: See Appendix A for description of surveys: TFR = Total Fertility Rate, calculated for women aged 15-49. SOURCES: Cantrelle et al. (1980); Cohen (1993); Ferry (1976); Lagarde et al. (forthcoming); Project Niakhar (1992); Pison and Desgrées du Loû (1993); WFS, DHS-II data files and unpublished tabulations from the 1988 census. |
adjust reported TFR, the resulting estimates are 6.0 or 5.5, depending on whether the correction coefficient is derived from the group aged 20-24 or 25-29.
The 1970-1971 NDS [2] was a multiround prospective survey. The data from this survey have been analyzed elsewhere and are thought to be of reasonably good quality (Cantrelle et al., 1986). However, the P/F ratios calculated from this survey decline steadily with age (see Figure 4-3) and average 1.27 for the age groups 20-24 and 25-29; if this factor is applied as an adjustment, the TFR estimated for 1970-1971 becomes an unrealistic 8.2.
The 1978 WFS [3], 1986 DHS-I [4], and 1992-1993 DHS-II [7] were all ad hoc fertility surveys including birth histories. As a result of using similar methodology, the estimates should be more comparable than those from the other surveys. Birth histories are usually effective at coverage of births, but fertility estimates may be distorted by errors in reports of timing. In the DHS-I, some births were undoubtedly transferred from the 5 years before the survey to an earlier period so that interviewers could avoid asking women additional questions about their children under age 5 (Arnold, 1990). Consequently, DHS-I (and DHS-II) data underestimate fertility in the last 5 years. Table 4-1 shows estimates of TFR for the period 4 years prior to the survey dates, but these also appear to be underestimates. Questions have also been raised about the quality of the WFS data (Charbit et
![](/openbook/0309052807/xhtml/images/img00012.jpg)
FIGURE 4-1 Total fertility rates for women aged 15-34 from national-level surveys, 1960-1993. NOTE: See Appendix A for description of surveys.
al., 1985); the complex procedure for collecting birth histories in the WFS may have resulted in fertility being overestimated, particularly among women aged 25-39.5
In the 1988 census, women were asked about their births in the last 12 months. This time, the problem of identifying the correct length of the period was partly resolved by taking the census exactly 1 year after an important Muslim holiday: the end of Ramadan.6 After being adjusted (see Appendix B), the fertility estimates from the census for the period 1987-1988 are very close to those reported by the DHS-II for the period 1989-1992. On average, the ASFRs for the census are the same as those reported in the DHS-II for women under 30. On the other hand, the census estimates are slightly lower than the DHS-II estimates for women over age 30. This pattern might be explained by a relative undercount of births at all ages in the census in comparison with the DHS-II and a small fertility decline at ages under 30 during the short period between 1987-1988 and 1989-1992.
Data from the 1986 and 1992-1993 DHS surveys provide a useful consistency
![](/openbook/0309052807/xhtml/images/img00013.jpg)
FIGURE 4-2 Proportion of the total fertility rate attributable to each age group. NOTE: See Appendix A for description of surveys.
check on the accuracy of the 1988 census. Dividing all the women in the sample into 5-year cohorts, and assuming that each cohort in the 1986 DHS ages to become the next cohort in the 1992-1993 survey (i.e., those aged 15-19 in 1986 become those aged 20-24 in 1992—not exact, but good enough for present purposes), allows one to compare the cumulated ASFRs based on reported births in the year preceding the 1988 census with the change in the number of children ever born between the two DHS surveys. The denominator, commonly known as the period parity distribution, is obtained by cumulating the parity increments for different cohorts. For the age groups 20-24 and 25-29, the P/F ratios are close to 1, indicating a high degree of consistency between the census and the cohort parity changes. After age 30, the P/F ratios rise somewhat, but definitely not because fertility is falling, because the effects of falling fertility are purged by the cohort increment process. The most likely explanation for this pattern of P/F ratios is that it reflects underreporting of births by older women in the census. In summary, this analysis confirms our suspicions that there was
![](/openbook/0309052807/xhtml/images/img00014.jpg)
FIGURE 4-3 Brass P/F ratios. NOTE: See Appendix A for description of surveys.
differential coverage of births in the 1988 census. Specifically, the quality of reporting deteriorated for women over the age of 30. Before age 30, the fertility rates from the two DHS surveys and the 1988 census are consistent.
The fertility estimates in Table 4-1 permit different interpretations of fertility trends over the last 40 years in Senegal. The estimates for periods immediately prior to the surveys appear to show steeply rising fertility from 1960-1961 (TFR = 5.4) through 1970-1971 (TFR = 6.4) to 1975-1978 (TFR = 7.2), and then a decline from 1983-1986 (TFR = 6.6) continuing to 1989-1992 (TFR = 6.1), except for an anomalous drop to 5.9 in 1987-1988. However, when time series estimates covering the 15 years prior to the birth history surveys are considered, a different picture emerges. The estimates based on the birth history surveys show fertility (cumulated from age 15-34) declining gradually from 1960 to the late 1980s, with a few irregularities; the estimates from the 1960-1961 and 1970-1971 surveys are much lower (see Figure 4-1). The most pronounced irregularities in the birth history sequences are the points immediately preceding the DHS-I and DHS-II;
both these points are well below the general trend, and the DHS-I point is well below the corresponding time estimate from the DHS-II. These two points appear to be affected by misplacement of births in time beyond the local shifts around age 5 mentioned above.
Further analysis is required to clarify what actually happened to fertility since 1960. The P/F ratios from the 1960-1961 and 1970-1971 surveys fall smoothly with age, across all ages for 1970-1971, but to a minimum for the age group 35-39 for 1960-1961 (see Figure 4-3). The ratios from these two surveys taken together would be consistent with rising fertility affecting cohorts of women born between 1920 and 1950. However, P/F ratios from the 1978 survey show no pattern with age, showing no evidence of differences between lifetime fertility and fertility in the 5 years before the survey for cohorts born between 1930 and 1960. P/F ratios from the 1986 and 1992-1993 DHSs appear to rise with age, though not very sharply, suggesting slowly declining fertility. The 1986 and 1992-1993 series are essentially parallel, but with the 1992-1993 values slightly above the 1986 ones at all ages. Together, they suggest no major change in fertility trends between the surveys, but slightly higher overall omission of children ever born in 1992-1993.
The age pattern of P/F ratios from 1960-1961 and 1970-1971 could be explained by increasing omission of children ever born as age of mother increases, as well as by rising fertility. There are large differences in average numbers of children ever born for women at the end of their reproductive life across the surveys. Women aged 45-49 report an average of 5.4 children ever born in 1960-1961, 5.6 in 1970-1971, 7.2 in 1978, 7.3 in 1986, and 7.4 in 1992-1993. These averages would clearly be consistent with rising fertility from the cohort of women born prior to 1925 (aged 45-49 in 1970-1971) to the cohorts born after 1935. However, the age detail of the data suggests some omission from the earlier surveys. The women aged 45-49 in 1978 report an average of 7.2 births each, whereas the average number of children born reported by women aged 35-44 in 1970-1971 (of whom the 45-49 year olds in 1978 would be survivors) was only 5.7. It is unlikely that these women added on average 1.5 births in 8 years at the end of their reproductive lives; more likely is that women aged 35 and over in 1970-1971 underreported their lifetime fertility by half a child or more. A comparison of children ever born by cohort between 1960-1961 and 1970-1971 suggests the same type of error in 1960-1961 also.
The national-level data thus do not substantiate a large fertility increase in the 1960s and 1970s, though they do not rule out a moderate increase. Data from one of the population study areas, Niakhar, confirm a fertility increase (from a TFR of 6.8 in 1963-1965 to a TFR of 7.8 in 1984-1990) in one rural area of central Senegal. On balance, it seems likely that fertility
did rise somewhat in the 1960s and 1970s, though not by nearly as much as the change in TFR from 1960-1961 to 1978 indicates.
It is difficult to explain what caused fertility to increase in the 1960s. It is unlikely that the increase can be attributed to a general reduction in the level of primary infertility. The proportion of women aged 50 or more with no live births in the 1960-1961 DS was already quite low, only 4-7 percent. Fargues (1989) has suggested that fertility increases in North Africa may have resulted from reduced marital disruption, particularly divorce; a similar process could underlie a fertility increase in Senegal in the 1960s and 1970s.
Figure 4-2 shows relative age patterns of fertility from the five surveys. Though there are differences between the surveys—the proportion of fertility contributed by the age group 40-44 in 1960-1961 is low, as is the contribution of the age groups 30-34 and 35-39 in 1970-1971, partially balanced in the latter case by a high contribution by women aged 45-49—these differences are probably the result of sampling or other errors, and no major trends are evident in the age pattern of fertility from 1960-1990. There is one minor trend visible, however: the percentage of total fertility contributed by the age group 15-19 does appear to be falling over time, from nearly 15 percent in 1960-1961 to 13 percent in 1978, to 11 percent by 1992-1993.
The question of whether or not there has been a recent fertility decline also requires further elucidation. Parity-specific analysis of birth history data can be a useful way to detect small changes in fertility for a country at the beginning of a transition toward lower fertility (Brass and Juarez, 1983). Recent analysis of the DHS data using censored parity progression ratios (B60s)7 revealed that the pattern of fertility decline witnessed in Kenya, Botswana, and Zimbabwe is quite unusual in comparison with the Latin American and Asian experience (Working Group on Kenya, 1993). The typical pattern of developing countries outside Africa was for initial declines to take place primarily in the middle, and sometimes in the higher, parities. The decline among low birth orders occurred more slowly. In Africa, by contrast, where fertility reductions have occurred, they have occurred across all parities (Working Group on Kenya, 1993).
The B60s for women in both the DHS-I and the DHS-II are shown in Table 4-2. To simplify the table, neighboring parity progression ratios for the same cohort have been combined by multiplying consecutive indices together.8 This procedure has the advantage of dampening fluctuations in the estimates caused by small sample size and is particularly useful because no formula exists to calculate the standard errors associated with B60s: inferences about changes in fertility behavior can be drawn only by examining general patterns in the data (Brass and Juarez, 1983).
For the DHS-I data, the B60s show some weak evidence of a fertility
TABLE 4-2 Censored Cohort Parity Progression Ratios, DHS-I and DHS-II
|
Censored Parity Progression Ratios (B60s) |
|||
Survey and Women's Age |
1-3 |
3-5 |
5-7 |
7-9 |
1986 DHS-I |
|
|
|
|
20-24 |
.733 |
|
|
|
25-29 |
.785 |
.646 |
|
|
30-34 |
.822 |
.771 |
.631 |
.438 |
35-39 |
.804 |
.753 |
.665 |
.482 |
40-44 |
.781 |
.755 |
.675 |
.546 |
45-49 |
.787 |
.797 |
.731 |
.555 |
1992-1993 DHS-II |
|
|
|
|
20-24 |
.693 |
|
|
|
25-29 |
.730 |
.706 |
.554 |
|
30-34 |
.779 |
.745 |
.595 |
|
35-39 |
.800 |
.747 |
.641 |
.484 |
40-44 |
.841 |
.791 |
.690 |
.492 |
45-49 |
.783 |
.760 |
.736 |
.588 |
NOTE: See Appendix A for description of surveys. SOURCE: Calculations based on data from DHS-I and DHS-II. |
decline among women of middle parities, but the B60s are quite erratic in the lower parities. For example, DHS-I data indicate that a little under 80 percent of women born between 1937 and 1941 (i.e., aged 45-49 in 1986) progressed from a third to a fifth birth. The comparable figure for women born between 1957 and 1961 is 65 percent. The comparable figure for women born between 1952 and 1956 is 77 percent. The proportion of women moving from a fifth to a seventh birth shows a more consistent decline across cohorts, from 73 percent among women born between 1937 and 1941 to 63 percent among women born between 1952 and 1956.
Examined by themselves, the B60s from the DHS-II show clearer evidence of a fertility decline across all parities, in keeping with the pattern documented recently for other parts of sub-Saharan Africa (Working Group on Kenya, 1993). However, the results are much more ambiguous when the B60s from the DHS-I and DHS-II are compared side by side. For example, consider women aged 30-34 who have had five children. According to DHS-I, 63 percent of these women progress to parity 7. As one would anticipate when analyzing a fertility decline, this is a higher progression rate than reported in DHS-II (60 percent). At the same time, however, 71
percent of women aged 25-29 at the time of the DHS-II progressed from parity 3 to parity 5, compared with only 65 percent in the DHS-I. Perhaps these findings should not come as a great surprise given that the reported use of modern contraception in Senegal is very low (see below).
Regional Variations
In 1978, the WFS reported fertility rates for four grand regions of the country:9 west, comprising the two regions of Dakar and Thiès; center, comprising Diourbel, Fatick, Kaolack, and Louga; south, comprising Ziguinchor and Kolda; and northeast, comprising Saint-Louis and Tambacounda. Subsequently, these same regions were used in the DHS-I and DHS-II, making it straightforward to observe any regional variation in fertility over time. Fertility rates based on data from the 4 years preceding each survey are shown for each of these four regions in Table 4-1. Age-specific fertility schedules for each region for various points in time are presented in Figure 4-4.
In the late 1970s, the total level of fertility of the four grand regions was much the same (see Table 4-1); the TFR varied from 7.1 children per women in the west to 7.4 in the south. Since 1978, the level of heterogeneity has increased as fertility has started to decline. Fertility in the DHS-II ranges from 5.6 children per woman in the west to 6.6 in the northeast.
Age-specific fertility also varies by region (see Figure 4-4). Fertility among young women is lowest in the west; this region includes the cities of Dakar and Thiès, and is therefore by far the most urbanized region. On the other hand, for women over 35, fertility is lowest in the southern and central regions. These regional differences are also found in the census, in spite of its lower reliability (see Figure 4-5).
A fertility decline among women under age 25 seems to have taken place across all four regions, although the pattern of decline is quite different in each case. Comparing the WFS with the DHS-I reveals that fertility was higher in the former than in the latter in three of the four regions, although in two of the three cases the differences between surveys are slight. Only the central region exhibits no difference in the level of fertility between the two surveys. When the results of the DHS-II are added, however, the picture changes. First, the central region now exhibits lower fertility between the DHS-I and DHS-II. But the level of fertility is virtually unchanged between the two DHS surveys in the southern and northeastern regions. Only in the western region has fertility continued to decline, falling an additional 18 percent over its 1986 level and 32 percent over its 1978 level. One also observes an apparent fertility decline for women aged 25-29. The pattern is clearest in the west and northeast before 1986, less apparent in the center, and nonexistent in the south. As discussed below,
![](/openbook/0309052807/xhtml/images/img00015.jpg)
FIGURE 4-4 Changes in age-specific fertility rates, by grand region over time. NOTE: See Appendix A for description of surveys. SOURCES: Standard data files from WFS, DHS-I, and DHS-II.
![](/openbook/0309052807/xhtml/images/img00016.jpg)
FIGURE 4-5 Age-specific fertility rates, 1987-1988, by grand region. SOURCE: 1988 census, unpublished data.
the evidence of a decline in fertility among younger women is consistent with a rise in age at first marriage.
Findings from the 1988 Census
The 1988 census allows us to investigate variations in the level of fertility by the 30 departments within Senegal. Estimates of the ASFR in each department are given in Tables B-2 to B-11 in Appendix B. The TFRs are much higher in rural than in urban areas, but they vary from department to department apparently without much of a discernible pattern, although they appear to be slightly higher in the coastal departments in the southwest of the country. Unfortunately, because the quality of reporting probably deteriorates as one moves away from the capital, it is not clear that fertility is really higher in the coastal departments.
As discussed above, women under age 30 appear to have played a central role in a fertility decline in Senegal. Consequently, it is important to examine the regional pattern of fertility among these women for whom
the quality of data in the census is probably more reliable than that of the data for the older ages. Figure 4-6 shows the sum of the ASFRs for the age group 15-29. Fertility in urban areas is lowest in and around Dakar. In the south of the country, far from the influence of the capital, fertility levels in urban areas are generally higher than those elsewhere. These broad patterns of (1) lower fertility around Dakar, and (2) slightly higher fertility in the south than in the north of the country are also visible in the bottom half of the figure, which shows variations in fertility for the rural population in each department.
The geography of fertility rates before age 20 as reported by the census is comparable with that derived from the WFS, DHS-I, and DHS-II using the four grand regions defined above (see Table 4-3). But as we have seen above, this type of analysis masks considerable intraregional heterogeneity. In particular, the central region is a mix of departments, some of which are far closer to Dakar than others, and some of which are far more southerly than others—two important factors underlying fertility differentials in Senegal.
TABLE 4-3 Age-Specific Fertility Rates for Women Aged 15-19 by Grand Region, 1975-1992
Year |
Source |
Subgroup |
Age-Specific Fertility Rate (per 1,000) |
1975-1978 |
WFS |
West |
144 |
|
|
Center |
212 |
|
|
Northeast |
205 |
|
|
South |
281 |
1983-1986 |
DHS-I |
West |
103 |
|
|
Center |
207 |
|
|
Northeast |
175 |
|
|
South |
196 |
1987-1988 |
Census |
West |
82 |
|
|
Center |
173 |
|
|
Northeast |
165 |
|
|
South |
206 |
1989-1992 |
DHS-II |
West |
83 |
|
|
Center |
149 |
|
|
Northeast |
176 |
|
|
South |
187 |
NOTE: See Appendix A for description of surveys. SOURCES: WFS, DHS-I, and DHS-II data files and unpublished tabulations from the 1988 |
Supplemental Information from Three Small-Scale Demographic Surveillance Systems
Apart from the information provided by national surveys and censuses described above, fertility estimates are also available for some areas from small-scale surveys [8-16]. Of course, these surveys provide information on only a small fraction of the total population, and their small populations make their estimates prone to variability. However, these drawbacks must be weighed against the fact that smaller populations enable interviewers to conduct frequent visits to the same households. Thus the enumerators can check and correct information collected during previous rounds, as well as document very carefully any events that have occurred between visits. As a result, data collected from multiround surveillance systems are often of substantially higher quality than can be achieved from a single interview, especially with regard to date and age variables.
In Senegal, the oldest and largest of these multiround surveillance systems is at Niakhar/Ngayokhème [11], where demographic data have been collected intermittently since December 1962 under the auspices of the Institut Français de Recherche Scientifique pour le Développement en Coopération (ORSTOM). Figure 4-7 shows a series of estimates derived from the ORSTOM project for various points in time.
Estimates from the Niakhar project indicate that fertility rates were extremely high over the period 1984-1990 (Project Niakhar, 1992). The TFR was estimated to be 7.8 children per woman, approximately 2 children per woman more than the rate recorded in 1963-1965 by Cantrelle (1969). The increase in fertility occurred principally among women over the age of 25 (see Figure 4-7).10 Although a large increase, a comparison by Garenne and van Ginneken (1994) found a satisfactory consistency between the fertility of the Serer living in the villages covered by the ORSTOM laboratory of Niakhar and the fertility of the rural Serer population measured by the WFS in 1978 and the DHS-I in 1986.
Data are also available from two other longitudinal studies, although both are of smaller size than the ORSTOM study. The Bandafassi study [8] was initiated in 1970 in the department of Kédougou, Tambacounda region and has grown steadily to include records on approximately 8,500 individuals in 1993.11 The Mlomp study was initiated in 1984 in the department of Oussouye, Ziguinchor region, and includes information on approximately 6,500 individuals in 1990.12
Figure 4-8 presents fertility rates for 1981-1990 for the study area of Bandafassi. Although the reported ASFRs are rather high among teenagers, the overall level of fertility is relatively low (TFR = 6.2), invalidating the existence of a systematic association between early and high fertility in Senegal. Even within this small area, however, fertility rates vary greatly,
![](/openbook/0309052807/xhtml/images/img00018.jpg)
FIGURE 4-7 Changes in age-specific fertility rates, Niakhar study area, over time. NOTES: The baseline changed somewhat over the years, so the different fertility schedules are not completely comparable: the study area initially extended farther south, but not as far to the north, as the present area. See Appendix A for description of surveys. SOURCES: Cantrelle et al. (1980), Project Niakhar (1992).
from 6.7 children per woman among the Peul to 5.5 children per woman among the Bedik (Pison and Desgrées du Loû, 1993).
The study area of Mlomp [10] is unlike any other in Senegal. It is characterized by an extremely large amount of seasonal migration, and the majority of the adult population is absent for at least half of the year (Lagarde et al., forthcoming). Eighty percent of women aged 15-24 (almost all of them unmarried) migrate to seek employment as domestic servants in the main cities of Senegal or The Gambia. Marriage takes place relatively late in Mlomp: the mean age at first marriage in the area is estimated to be 24 for women and 30 for men (Enel and Pison, 1992). Marriage probably occurred earlier one or two generations ago (Enel and Pison, 1992). Once married, women usually stop their pattern of seasonal migration and stay all year in their village. The consequences of these patterns are apparent in Figure 4-9.13 Although Mlomp is a unique area, the rapid change in fertility shown in Figure 4-9 is an example of what might take place in Senegal in
![](/openbook/0309052807/xhtml/images/img00019.jpg)
FIGURE 4-8 Age-specific fertility rates, Bandafassi study area, 1981-1990. NOTE: See Appendix A for description of survey. SOURCE: Pison and Desgrées du Loû (1993).
the years to come: a rapid decline in fertility at younger ages, following a delay in marriage.
These multiround surveillance systems provide us with an independent check on the quality of the birth data in the 1988 census. Given that the census experienced serious problems with recording of deaths of children under age 1 (see Chapter 5), it is interesting to compare the census results with ASFRs obtained from the three multiround surveillance systems. As discussed earlier, the 1988 census restricted its investigation of fertility to a single question on the number of children women had had in the last 12 months,14 while all three multiround studies have invested considerable time and money in making repeated visits to the same households to ensure the accuracy of their data. Figure 4-10 reproduces the fertility schedules for the most recent periods for Niakhar, Bandafassi, and Mlomp. These data have been plotted with census data from the rural areas of the departments in which the surveillance systems are located.
In none of the cases is the match perfect, though encouragingly, there is
![](/openbook/0309052807/xhtml/images/img00020.jpg)
FIGURE 4-9 Changes in age-specific fertility rates, Mlomp study area, over time. NOTE: See Appendix A for description of survey. SOURCE: Lagarde et al. (forthcoming).
fairly close agreement between the Niakhar data and the census. The match is less satisfactory in the other two cases. Comparison of the ASFRs in the census with the Bandafassi data indicates that there may have been a sizable deficit of recorded births in the census in the department of Kédougou, one of the least developed parts of the country. Finally, the ASFRs for Mlomp and census data for rural Oussouye bear no resemblance to one another. However, the study area of Mlomp is considerably smaller than the department of Oussouye, has special socioeconomic characteristics, and probably should not be expected to be representative of that department.
Despite their limitations, local comparisons tend to support the hypothesis that births were underreported in the 1988 census. While fertility rates from the census for the youngest age groups, notably ages 15-19, correspond fairly well to those based on the Niakhar and Bandafassi data, the differences are greater among the older age groups, i.e., above age 20 in the census data for the department of Kédougou (compared with the Bandafassi data) and above age 30 in the census data for the department of Fatick
![](/openbook/0309052807/xhtml/images/img00021.jpg)
FIGURE 4-10 Comparison of age-specific fertility rates from 1988 census data with data from small-scale surveillance systems. NOTE: See Appendix A for description of surveys. SOURCES: 1988 census, unpublished data, Lagarde et al. (forthcoming), Project Niakhar (1992), Pison and Desgrées du Loû (1993).
(compared with the Niakhar data). The extent of underreporting probably varied from one region of the country to another. This fact must be borne in mind when studying demographic variations in fertility based on census-estimated rates.
Conclusions Drawn from Regional Differences
The above discussion highlights the considerable heterogeneity among administrative regions in both the level of fertility and the timing of births. The difference in total fertility exceeds two children per woman between the small-scale studies with the highest and lowest rates. Some regions have rates of childbearing among teenagers that are three to four times higher than those in other regions. Much of the diversity among regions occurs among women under 30, suggesting that the variations may be more closely related to the timing of entry into union than to the level of fertility within unions.
Evidence for a fertility decline is strongest in Dakar and the western region of the country. Elsewhere, as in the Niakhar region, the level of fertility has increased over the past 20 years. Thus, regional differentiation has become more pronounced since the 1970s.
Fertility Differences by Socioeconomic Characteristics
A fundamental problem facing population economists attempting to explain the relationship between income and fertility is that in every major region of the world over the last two centuries, childbearing has fallen as income has risen. At first blush, this runs counter to economists' expectations because if children are desirable, one should want to acquire more of them as one accumulates more wealth. In response, economists have tended to place increasing reliance on a vast literature commonly referred to as ''new home economics." This microlevel approach has been made famous over the last 30 years by Nobel Prize-winning economist Gary Becker.15
Within this framework, different sources of income exert differing effects on fertility. Higher male wages are assumed to increase fertility, while higher female wages are usually assumed to reduce fertility because the bearing and rearing of children competes with working in the labor force for women. Inspection of the recent empirical data appears to confirm many of the features of the neoclassical economic model (Schultz, 1994). In particular, countries with higher incomes, given their human capital resource base, were found to have higher fertility. At the same time, greater female education was found to be the dominant factor associated with a decline in fertility in both cross-sectional and longitudinal data (Schultz, 1994).
Microlevel Evidence
The 1978 WFS [3], the 1986 DHS-I [4], and the 1992-1993 DHS-II [7] allow the study of fertility differentials by some important socioeconomic characteristics, such as place of residence, ethnicity, and level of instruction. The existence of these three comparable national surveys allows us to analyze differences over time.
Figure 4-11 shows fertility differences by urban/rural residence. The vast majority of women still live in rural areas where fertility is very high. Although one small-scale study [11] indicated that fertility may have risen in Niakhar (see above), overall there appears to have been little change in fertility in rural Senegal over the past 15 years. The TFR in rural areas decreased slightly over this period, from 7.5 in the mid-1970s to 7.3 in the mid-1980s, and again to between 6.4 and 6.8 at the end of the 1980s and the beginning of the 1990s (see Table 4-1). The timing of fertility (i.e., in the total number of children ever born), like the age-specific pattern, also changed only slightly. In rural areas, childbearing still begins earlier and is ultimately higher than in the urban areas.
In urban areas, fertility has been declining since the mid-1970s. In the late 1970s, the WFS estimated fertility at 6.8 children per woman in Dakar and 6.5 children per woman in other urban areas. These levels have declined over the last 15 years. For the most recent period, 1989-1992, DHS-II data indicate that fertility is around 4.9 children per woman in Dakar and 5.4 in other cities. Some postponement of childbearing had already occurred in urban areas by the mid-1980s. In Dakar, a comparison of WFS and DHS-I data reveals that fertility decline had already occurred among women under age 35. In other cities, only women under age 30 have reduced their fertility. Overall, the changes in the timing of childbearing have led to a flatter fertility schedule that is less concentrated in the prime childbearing ages. There is virtually no indication that fertility has declined for women over the age of 30, except in Dakar. A decline in fertility among women over 30 would be significant because it would mark the beginning of a fertility decline within unions, which, when combined with the decline in fertility attributable to a delay in entry into union, can lead to a rapid change in the TFR.
Fertility was probably relatively stable in Dakar between the end of the 1960s and the mid-1970s.16Figure 4-12 displays the ASFR for Dakar in 1975-1978 calculated from the WFS [3] with the fertility schedule derived from the Cap-Vert (Dakar) fertility survey of 1972 [9] taken 6 years earlier. The TFR estimates are very close—6.8 and 6.6 children per woman, respectively—and we note a remarkable consistency between the two schedules.
Table 4-1 shows the variation in the TFR over time by major ethnic group. Cultural differences among ethnic groups have long been recognized
![](/openbook/0309052807/xhtml/images/img00022.jpg)
FIGURE 4-11 Age-specific fertility rates, by urban/rural residence. NOTE: See Appendix A for description of surveys. SOURCES: 1988 census, unpublished data and standard data files from WFS, DHS-I, and DHS-II.
![](/openbook/0309052807/xhtml/images/img00023.jpg)
FIGURE 4-12 Changes in age-specific fertility rates, Dakar Region, over time. SOURCES: Ferry (1976) and WFS standard data tape.
as important determinants of the level of fertility in Senegal,17 although the pattern of childbearing is broadly similar among the various groups. For example, fertility is considerably higher among the Serer, who live mainly in the Sine-Saloum region. The TFR for Serer women was 7.9 children per woman in 1975-1978, compared with a TFR of 7.2 for Wolof women or a TFR of 6.9 for Poular women. Although the TFR fell for all three groups between 1978 and 1986, it fell the least for the Serer. And in the DHS-II, the TFR for Serer women remains almost 1.0 child per woman higher than the TFR for the Poular and 1.6 children per woman higher than the TFR for Wolof women.18
The fertility schedules for the Diola, Mandingo, and other ethnic groups are based on smaller sample sizes, so they tend to be more erratic than the schedules for the Wolof, Poular, and Serer. However, Figure 4-13 shows fertility has declined significantly among young Mandingo women, who live predominantly in the south of the country. Thus, fertility variations by ethnic group partly confirm variations observed at the regional level.
Figure 4-14 shows variations in fertility over time separately for women
![](/openbook/0309052807/xhtml/images/img00024.jpg)
FIGURE 4-13 Age-specific fertility rates by ethnic group. NOTE: See Appendix A for description of surveys. SOURCES: 1988 census, unpublished data and standard data files from WFS, DHS-I, and DSH-II.
with different levels of education. The experiences of the three groups over time are very different. Women with no education had very high fertility in the mid-1970s. The TFR for the period 1975-1978 was 7.4 children per woman, a figure that has declined slowly over the last 15 years (see Table 4-1). On the other hand, fertility among women with only primary education fell sharply between 1975-1978 and 1983-1986; their TFR dropped from 7.1 to 5.7 children per woman over this period. Note, however, that the number of women with a primary education sampled in the WFS was quite small, so we must be careful not to read too much into small fluctuations in the shape of the fertility schedule, especially above age 35.
It is even more difficult to draw conclusions for the group of women with secondary or higher education: the fertility schedule estimated from the WFS is quite erratic, reflecting the small sample size. There is, however, little doubt that women in this group have lower fertility than women with less education. The TFR among the former group has changed little over the period—the TFR has remained around 3.6 children per woman—but the peak of the fertility schedule may have moved from ages 25-29 to ages 30-34.19
Macrolevel Evidence
The cross-sectional relationships among various socioeconomic, fertility, and marriage indicators can be analyzed using data from the 1988 census.20 Three fertility indicators are used in this analysis: the period TFR for women aged 10-54 (TFR10-54), the period TFR for women aged 15-29 (TFR15-29), and the period ASFR for women aged 15-19 (ASFR 15-19). The marriage indicators used are the singulate mean age at first marriage for men and women (SMAMm, SMAMf), the proportion of men aged 20-24 who have ever been married, and the proportion of women aged 15-19 who have ever been married.
The correlation coefficient is a useful statistic that provides a measure of the association between two random variables. Correlation coefficients between selected socioeconomic and demographic indices at the department level (N = 30) are provided in Table 4-4.
Until recently, there was little evidence of any parity-specific fertility control among married women in Senegal, so it is not surprising to find a low correlation between the TFRs and the various measures of socioeconomic development displayed in Table 4-4. When these indicators are correlated with the fertility rate of females aged 15-29, the associations become more pronounced. Fertility among women aged 15-19 is even more strongly associated with socioeconomic status, so that a department's level of socioeconomic development becomes a good predictor of the level of adolescent childbearing. Figure 4-15 is illustrative of the departmental
![](/openbook/0309052807/xhtml/images/img00025.jpg)
FIGURE 4-14 Age-specific fertility rates by level of education. NOTE: See Appendix A for description of surveys. SOURCE: 1988 census, unpublished data and standard data files from WFS, DHS-I, and DHS-II.
TABLE 4-4 Correlation Coefficients Between Socioeconomic and Demographic Indicators
|
Fertility Indicators |
Nuptiality Indicators |
|||||
Socioeconomic Indicator |
TFR |
TFR15-29 |
ASFR15-19 |
% Males Married, 20-24 |
% Females Married, 15-19 |
SMAMm |
SMAMf |
Housing |
-.365 |
-.754 |
-.879 |
-.703 |
-.818 |
.744 |
.812 |
Electricity |
-.551 |
-.833 |
-.799 |
-.548 |
-.656 |
.584 |
.673 |
Latrine |
-.502 |
-.709 |
-.682 |
-.416 |
-.488 |
.434 |
.511 |
Water |
-.544 |
-.812 |
-.846 |
-.462 |
-.598 |
.476 |
.579 |
Composite |
-.516 |
-.822 |
-.843 |
-.557 |
-.673 |
.585 |
.674 |
Urban |
-.440 |
-.747 |
-.780 |
-.671 |
-.766 |
.695 |
.781 |
Ever school |
-.225 |
-.606 |
-.727 |
-.852 |
-.898 |
.890 |
.920 |
School enrollment rate (male) |
-.028 |
-.435 |
-.639 |
-.882 |
-.916 |
.925 |
.933 |
School enrollment rate (female) |
.008 |
-.416 |
-.628 |
-.871 |
-.909 |
.936 |
.933 |
NOTE: See Chapter 2 for details on how the socioeconomic indicators were constructed. SOURCE: Unpublished tables, 1988 census. |
![](/openbook/0309052807/xhtml/images/img00026.jpg)
FIGURE 4-15 Departmental variations in fertility by the composite facilities indicator, women aged 15-19. NOTE: The composite facilities indicator is the arithmetic mean of (1) the percentage of households obtaining their water supply from a faucet or standpipe, whether placed inside or outside the home; (2) the percentage of households with electricity; (3) the percentage of households with toilets or latrines; and (4) the percentage of households whose houses have been built using good construction techniques. See Chapter 2 for more details. SOURCE: 1988 census, unpublished tables.
relationship between age-specific fertility of women aged 15-19 and the level of socioeconomic development.
As seen later, age at first marriage is an important proximate determinant of fertility among women aged 15-19. The correlations between various indicators of marriage and socioeconomic status reported in Table 4-4 are rather large. Figure 4-16 is illustrative of this relationship at the department level. One region is a notable outlier on the figure: Ziguinchor. Given the department's socioeconomic status, fewer young women in Ziguinchor are married than would be expected, probably because a high proportion of Diola women (the major ethnic group in the area) tend to spend their early adult years working in urban areas, and consequently tend to marry relatively late.
In summary, fertility rates and socioeconomic indicators tend to be
![](/openbook/0309052807/xhtml/images/img00027.jpg)
FIGURE 4-16 Departmental variations in nuptiality by the composite facilities indicator, women aged 15-19. NOTE: See note for Figure 4-15. SOURCE: 1988 census, unpublished tables.
inversely related. Fertility for younger women tends to be more strongly related to socioeconomic status than does fertility for older women. The percentage of males and females married is also inversely related to socioeconomic status, the relationships being stronger for females than for males. With one notable exception, Ziguinchor, mean age at marriage tends to increase with socioeconomic status.
Summary of Fertility Levels and Trends
Senegalese women have high fertility. On average, each woman has about 6 children over the course of her reproductive life. Although a substantial number, this figure actually represents a decline of approximately 1 child per woman from the fertility level observed at the end of the 1970s. It is possible that this decline was preceded by an increase in fertility during an earlier period. Until recently, there was a general reluctance to accept any signs of recent fertility decline in Senegal, but it now appears that such a decline has in fact occurred. This decline started first in the capital and extended progressively, although slowly, to other regions of the country. Until now, this decline has affected only the urban population.
Examination of differentials by region, place of residence, education, and ethnicity reveals that the western region, which is the most educated and urbanized, is the one with the lowest fertility. Variations in reproductive behavior by ethnicity may also explain some regional differences.
What other variables might help explain these changing patterns of fertility? The decline in ASFRs among women aged 15-19 and 20-24 indicates that the fertility decline may be linked to a delay in the entry into union (Sadio, 1985). However, some of the other observed fertility differentials, such as those between women with secondary education and women with no education, are very large and cannot be explained solely by differences in the timing of marriage. Other proximate determinants, such as contraceptive use, abortion, and sterility, also play a role in limiting total fertility. The relative importance of these factors is quantified below using the proximate determinants framework of Bongaarts and Potter (1983).
PROXIMATE DETERMINANTS OF FERTILITY
Overview
Regardless of the survey examined, ASFRs in Senegal remain high over a woman's entire reproductive period. However, as noted above, fertility rates have fallen over the last 15 years, particularly among the young. Why have these changes taken place? It is analytically convenient to differentiate factors influencing fertility into two distinct classes: (1) a set of biological ''proximate" determinants that exert a direct influence on fertility, and (2) all social, cultural, and economic factors that can influence fertility only through these proximate determinants. Originally specified by Davis and Blake (1956) as "intermediate variables," the major proximate determinants were identified by Bongaarts (1978) and adopted into a straightforward framework so that the relative importance of each could be easily assessed. The following is a complete list of the quantitatively important proximate determinants:
-
the proportion of women married (or more strictly the proportion of women in sexual unions);
-
the contraceptive prevalence and effectiveness rate;
-
the induced abortion rate;
-
the length of the postpartum infecund interval (primarily the result of breastfeeding patterns);
-
the frequency of coitus;
-
the level of spontaneous intrauterine mortality; and
-
the prevalence of primary or secondary sterility.
Each proximate determinant has a different degree of importance in explaining variations in fertility. It has been shown, however, that differences in fertility rates among populations are due predominantly to variations in only four of the determinants: the proportion married, the level and effectiveness of contraceptive use, the incidence of induced abortion, and the length of the postpartum infecund interval (Bongaarts and Potter, 1983). In sub-Saharan Africa, high levels of pathological sterility may also lead to important variations in fertility among certain ethnic groups (Frank, 1983; Bongaarts et al., 1984). The respective effects of the main proximate determinants on fertility can be quantified using the following model:
TFR = Cm * Cc * Ci * Ca * Ip * TF,
where:
TFR = total fertility rate,
Cm = index of marriage,
Cc = index of contraception,
Ci = index of postpartum infecundability,
Ca = index of induced abortion,
Ip = index of sterility, and
TF = potential fertility.21
The effect of each of the proximate determinants is quantified by an index that varies between 0 and 1. The model has been constructed in such a way that when an index is close to 1, the proximate determinant will have a negligible inhibiting effect on total fertility, whereas when the index is close to 0, the proximate determinant will have a large inhibiting effect.
Table 4-5 shows the Bongaarts indices, estimated using data from the WFS [3], DHS-I [4], and DHS-II [7]. The level of fertility is scarcely reduced by contraceptive use (Cc), or pathological sterility (Ip). Both these indices are close to 1, and together reduced potential fertility by less than 2 percent in 1978 and 5 percent in 1986 and 1992. The 4-percent increase in fertility reduction can be attributed to a recent increase in the use of modern contraception. On the other hand, the inhibiting effect of postpartum infecundability (Ci) is very large. This index alone reduced potential fertility by 37 percent in 1978, 42 percent in 1986, and 41 percent in 1992.22 The importance of postpartum infecundability to the reduction of fertility is typical in Africa and is mainly due to extended periods of breastfeeding (see Jolly and Gribble, 1993).
The second smallest index—the second largest fertility-inhibiting index—is Cm, the index of marriage. It is calculated as the ratio of the TFR
TABLE 4-5 Indexes of the Effect of Proximate Determinants on Fertility in Senegal, WFS, DHS-I, and DHS-II
Year and Survey |
TFR |
Cm |
Cc |
Ci |
Ca |
Ip |
1978 WFS |
7.2 |
.935 |
.992 |
.625a |
1.000 |
.992 |
1986 DHS-I |
6.6 |
.898 |
.973 |
.576 |
1.000 |
.976 |
1992 DHS-II |
6.1 |
.843 |
.951 |
.595 |
1.000 |
1.000 |
NOTES: See Appendix A for description of surveys; TFR, Cm, Cc, Ci, Ca, and Ip are defined in the section Proximate Determinants of Fertility and in endnote 21. a The WFS did not ask women whether they were currently amenorrheic, so i was imputed using the formula i = 1.753 exp(.1396B - .001872B2), where B is the duration of breastfeeding (Bongaarts and Potter, 1983). SOURCES: République du Sénégal (1981), Ndiaye et al. (1988) and unpublished tabulations of DHS-II data. |
to the TMFR, the total marital fertility rate. As Table 4-5 demonstrates, this index fell from .935 in 1978 to .898 in 1986 to .843 in 1992. Thus, the decline in the TFR, which fell from 7.2 in 1978 to 6.1 in 1993, was the result mainly of a change in marriage patterns. As we shall see below, the major change was a trend towards later age at first marriage for women.
For all three surveys, the index Ca has been set to 1.000. Owing to the sensitivity of the issue, no data were collected on abortions in any of the three surveys. However, Coeytaux (1988) and others have suggested that the incidence of induced abortion in sub-Saharan Africa is nontrivial and higher than generally realized. Given that age at first marriage in Senegal is rising, the likelihood of conceptions occurring before marriage and being aborted has probably also risen, so research into the prevalence of abortion and its impact on fertility in Senegal should be a priority.
In the following subsections, the effects of each index are discussed in greater detail.
Marriage and Fertility
Bongaarts' index of marriage (Cm) takes into account the global effect of marriage on fertility, including age at first marriage, marital stability, and the proportion of women ever married. These elements are discussed separately below, along with the relationship between the age at first marriage and age at first birth, and the role of polygyny.
![](/openbook/0309052807/xhtml/images/img00028.jpg)
FIGURE 4-17 Percent never married, by age. NOTE: See Appendix A for description of surveys. SOURCES: République du Sénégal (1974, no date, and 1988 census, unpublished tables) and standard data files from WFS, DHS-I, and DHS-II.
Age at First Marriage
To examine changes in marriage patterns over time, we observe the proportion of men and women never married by age at various points in time. Figure 4-17 plots the percent of women (top half of the figure) and men (bottom half of the figure) never married, by age, for different census and survey dates. Table 4-6 provides more detailed information on the percent of women who have never married by ages 15-19 and 20-24 for various subgroups of the population.
Little is known about the marital lives of men. Many of the national demographic surveys in the country have been restricted to women, reflecting a long tradition of research on nuptiality, primarily as a determinant of a woman's fertility. The bottom half of Figure 4-17 provides information on
TABLE 4-6 Percent of Women Never Married, Aged 15-19 and 20-24, 1978-1992
|
|
|
% Never Married |
|
Year |
Source |
Subgroup |
15-19 |
20-24 |
National Level |
|
|
|
|
1970-1971 |
NDS |
n.a. |
57 |
15 |
1976 |
Census |
n.a. |
55 |
21 |
1978 |
WFS |
n.a. |
41 |
14 |
1986 |
DHS-I |
n.a. |
57 |
23 |
1988 |
Census |
n.a. |
57 |
30 |
1992-1993 |
DHS-II |
n.a. |
69 |
36 |
WFS and DHS Regions |
|
|
|
|
1978 |
WFS |
West |
64 |
26 |
|
|
Center |
30 |
10 |
|
|
Northeast |
28 |
4 |
|
|
South |
29 |
6 |
1986 |
DHS-I |
West |
77 |
40 |
|
|
Center |
44 |
8 |
|
|
Northeast |
38 |
12 |
|
|
South |
50 |
20 |
1992-1993 |
DHS-II |
West |
85 |
52 |
|
|
Center |
62 |
25 |
|
|
Northeast |
54 |
21 |
|
|
South |
58 |
28 |
Administrative Regions |
|
|
|
|
1988 |
Census |
Dakar |
83 |
52 |
|
|
Thiès |
66 |
33 |
|
|
Saint-Louis |
52 |
25 |
|
|
Tambacounda |
27 |
9 |
|
|
Louga |
43 |
18 |
|
|
Diourbel |
48 |
18 |
|
|
Fatick |
52 |
22 |
|
|
Kaolack |
41 |
19 |
|
|
Ziguinchor |
78 |
45 |
|
|
Kolda |
30 |
10 |
Small-Scale Studies |
|
|
|
|
1963 |
Cantrelle (1969) |
Sine |
40 |
9 |
1963 |
Cantrelle (1969) |
Saloum |
27 |
5 |
1984 |
Enel et al. (1994) |
Mlomp |
99 |
89 |
1984 |
Delaunay (1994) |
Niakhar |
50 |
11 |
1991 |
Delaunay (1994) |
Niakhar |
62 |
13 |
|
|
|
% Never Married |
|
Year |
Source |
Subgroup |
15-19 |
20-24 |
Small-Scale Studies |
|
|
|
|
1992 |
Pison and Desgrées |
Bandafassi |
|
|
|
du Loû (forthcoming) |
Peul |
35 |
4 |
|
|
Malinké |
74 |
7 |
|
|
Bedik |
88 |
21 |
Place of Residence |
|
|
|
|
1978 |
WFS |
Dakar |
73 |
31 |
|
|
Other urban |
62 |
24 |
|
|
Rural |
29 |
5 |
1986 |
DHS-I |
Dakar |
81 |
44 |
|
|
Other urban |
79 |
38 |
|
|
Rural |
37 |
8 |
1988 |
Census |
Dakar |
85 |
57 |
|
|
Other urban |
79 |
44 |
|
|
Rural |
40 |
16 |
1992-1993 |
DHS-II |
Dakar |
91 |
58 |
|
|
Other urban |
82 |
56 |
|
|
Rural |
53 |
19 |
Ethnic Groups |
|
|
|
|
1978 |
WFS |
Wolof |
48 |
18 |
|
|
Poular |
29 |
8 |
|
|
Serer |
45 |
11 |
|
|
Mandingo |
24 |
14 |
|
|
Diola |
77 |
21 |
|
|
Other |
40 |
11 |
1986 |
DHS-I |
Wolof |
64 |
26 |
|
|
Poular |
43 |
11 |
|
|
Serer |
60 |
22 |
|
|
Mandingo |
53 |
21 |
|
|
Diola |
89 |
49 |
|
|
Other |
54 |
27 |
1992-1993 |
DHS-II |
Wolof |
73 |
40 |
|
|
Poular |
55 |
25 |
|
|
Serer |
74 |
31 |
|
|
Mandingo |
66 |
41 |
|
|
Diola |
91 |
62 |
|
|
Other |
72 |
27 |
Level of Education |
|
|
|
|
1978 |
WFS |
None |
29 |
8 |
|
|
Primary |
63 |
24 |
|
|
|
% Never Married |
|
Year |
Source |
Subgroup |
15-19 |
20-24 |
Level of Education |
|
|
|
|
|
|
Secondary+ |
90 |
47 |
1986 |
DHS-I |
None |
41 |
11 |
|
|
Primary |
85 |
40 |
|
|
Secondary+ |
95 |
68 |
1988 |
Census |
None |
44 |
19 |
|
|
Primary |
87 |
55 |
|
|
Secondary+ |
95 |
73 |
1992-1993 |
DHS-II |
None |
56 |
22 |
|
|
Primary |
86 |
52 |
|
|
Secondary+ |
96 |
82 |
NOTE: See Appendix A for description of surveys. SOURCES: Cantrelle (1969); Delaunay (1994); Enel et al. (1994); Pison and Desgrées du Loû (forthcoming); WFS, DHS-I, and DHS-II data files and unpublished tabulations from the 1988 census. |
men's marital status by age from three independent sources. Entry into marriage occurs at a much slower rate for men than for women, so that the mean age at first marriage for men is quite high—27.6 years. By age 24, only about 30 percent of men have married. Figure 4-17 demonstrates that there has been virtually no change in the age at first marriage for men.
The top half of Figure 4-17 presents similar information for women. Despite the potential ambiguities surrounding the definition of marriage, 23 the data show a clear trend towards later age at first marriage. In 1978, the WFS found that 14 percent of women aged 20-24 and 41 percent of women aged 15-19 had not been married. In 1986, the DHS-I found that the percentage of women who had never married had risen to 23 percent for women aged 20-24 and 57 percent for women aged 15-19. In the DHS-II, these percentages had increased again, to 36 and 69 percent, respectively. Figure 4-18 shows that there is considerable heterogeneity in marriage patterns among the four grand regions: women marry much later in the west than in the three other regions. These differences reflect important differences in education attainment and urban/rural residence, as explained below, as well as differences in the accepted age at first marriage among the various ethnic groups. The Poular, for example, marry relatively early, while the Serer and the Diola marry somewhat later, in part because women from these two latter ethnic groups often migrate to urban areas for a short period to accumulate goods before they marry (Ndiaye, 1992:18).
![](/openbook/0309052807/xhtml/images/img00029.jpg)
FIGURE 4-18 Percent of women never married, by age and region. NOTE: See Appendix A for description of surveys. SOURCES: Standard data files from WFS, DHS-I, and DHS-II.
A comparison of the findings of the WFS, DHS-I, and DHS-II demonstrates that signs of changing marriage patterns are appearing in all four grand regions of the country. The changes are partly attributable to increasing levels of education and urbanization. Figures 4-19 and 4-20 show the percent of women who have never married for various socioeconomic groups. Marriage occurs earliest among women with little or no education and among women living in rural areas. However, there is a perceptible increase in the age at marriage within each educational level and within both urban and rural areas. Antoine and Nanitelamio (1991) recently documented a rise in the number of single women in Dakar, but Figure 4-20 demonstrates that the phenomenon is not restricted to the capital. Further, one of the largest rises in age at first marriage has been for women with primary school education. Even among women with no education, there are still signs of a pattern towards later marriage. The immediate demographic consequence of these changes is an increase in the period of adolescence for females. In the past, early marriage meant the abrupt end of a woman's childhood and a transition into being a wife and a mother. Now, with later marriage, women can experience a period of adolescence before they marry (Working Group on the Social Dynamics of Adolescent Fertility, 1993).
Marital Stability
Information on divorce and remarriage comes principally from the WFS because it was the only national-level demographic study that collected complete information on marital histories. The DHS-I, DHS-II, and census questions on marriage were rather limited and concerned current marital status and age at first marriage. Because of the high quality and uniqueness of the WFS data, they have been analyzed in considerable detail (République du Sénégal, 1981; United Nations, 1983b; Smith et al., 1984; Ndiaye, 1985). These analyses revealed the following findings:
-
In 1978, 71 percent of married women were in their first marriage. For those who were not, the principal reason for the dissolution of their marriage was divorce or separation: 20 percent of the married women had been divorced or separated. The remaining 9 percent of married women had been widowed.
-
The survivorship rate of marriages in Senegal is low compared with that in Kenya and Lesotho (two other sub-Saharan African countries where a WFS was conducted). On average, women are married 1.32 times over the course of their life, and 5 percent of women are married 3 or more times.
-
Senegal's relatively high (by African standards) marital instability is
![](/openbook/0309052807/xhtml/images/img00030.jpg)
FIGURE 4-19 Percent of women never married by age and education. NOTE: See Appendix A for description of surveys. SOURCES: Standard data files from WFS, DHS-I, and DHS-II.
![](/openbook/0309052807/xhtml/images/img00031.jpg)
FIGURE 4-20 Percent of women never married, by age and urban/rural residence. NOTE: See Appendix A for description of surveys. SOURCES: Standard data files from WFS, DHS-I, and DHS-II
-
largely compensated for by rapid and high rates of remarriage. Among women who are widowed or divorced, 90 percent remarry within 5 years.
-
Marriages are generally more stable in rural areas, among women who have never attended school, and among women who have given birth to at least one child within the first 5 years of the marriage (Ndiaye, 1985). Marriages are less stable among urban women, literate women, and the Poular, who have the youngest average age at first marriage of any ethnic group. The high divorce rate among the Poular is probably the result of the high percentage of arranged marriages. Why should urbanization and literacy affect the probability of divorce? Both increase the probability that the woman has paid employment in the wage sector, thus increasing a woman's independence. Women in urban areas may also be less stigmatized by a separation or a divorce than women in rural areas.
In summary, the percentage of women who are single between marriages at any time is low, and the average length of time women spend married is high. Consequently, divorce and remarriage have little role to play in explaining fertility differentials in Senegal, either over time or among various socioeconomic groups.
Proportion of Women Ever Married
In the case of Senegal, the ultimate proportion of women ever married has not varied significantly over time: the proportion of never-married women aged 45-49 was less than 1 percent in all three surveys and was estimated to be 1.8 percent based on the 1988 census. Thus, the index Cm has not been affected by a change in the proportion of women ever married.
Age at First Marriage and Age at First Birth
Marriage is the legal and cultural institution that sanctions childbearing, and the Bongaarts formula assumes that all fertility occurs within marriage. How good is this assumption in Senegal? Illegitimate fertility in the late 1970s was relatively low in Senegal: 6 percent of first births took place outside of marriage according to the WFS (Guèye and Ferry, 1985). Has the delay in marriage brought about a delay in the age at first birth, and if so, is it responsible for the decline in fertility among young women noted above?
One simple measure of the extent of early childbearing is the proportion of women of a given age having had their first birth before the age of 20. To minimize biases due to the omissions and errors of dating that are frequent for older women (Arnold and Blanc, 1990), we consider only women aged 20-29 at the time of the survey. In 1978, 65 percent of all women
aged 20-29 had had their first birth before the age of 20. These percentages fell only slightly in the 1986 DHS-I: 61 percent of women aged 25-29 and 60 percent of women aged 20-24 had had their first birth by age 20. In the 1992-1993 DHS-II, these proportions fell by about 10 percent, so that 57 percent of women aged 25-29 and only 52 percent of women aged 20-24 had had their first birth by age 20. This represents a more substantial decline than the one that took place between 1978 and 1986, but is it more or less than what one might expect from an increase in the age at marriage?
The trends in teenage marriage and teenage childbearing are displayed in Figure 4-21, which shows the percentage of women marrying by age 20, the percentage giving birth by age 20, and the percentage having a premarital birth before age 20. There are two points for each survey: one represents women aged 20-24 at the time of the survey, while the other represents women aged 25-29. These women were aged 20-24 5 years before the survey, so their information has been plotted 5 years to the left of the survey date. Hence the three surveys have been combined to show the experiences of women aged 20-24 at various points in time. The observations from different age groups and surveys fit together closely over time, arguing for the general quality of the data. The solid downward-sloping line at the top of the figure represents the proportion of women aged 20-24 marrying by age 20. The downward slope of this line means a decline in early marriage. The downward-sloping line below the early marriage line represents the percentage of women giving birth by age 20. Again, a downward-sloping line represents a decline in early childbearing. However, the gap between the two lines narrows over time, indicating that early childbearing has declined more slowly than early marriage. These changes could be the result of an increase in the frequency of premarital conception, or a shortening of the interval between marriage and first birth. 24 We return to this topic below.
As noted above, there are few reported cases of premarital births in Senegal. There are two reasons for this. First, the young age at first marriage implies that a large proportion of unmarried women are young and still infecund. Second, older women often misreport age at first birth if they were unmarried at that time. However, Figure 4-21 suggests that there has been a slight increase in premarital childbearing. The triangles scattered across the bottom of the figure represent the percentage of women who reported having a premarital birth before age 20. (Note that these triangles have been plotted on the second scale displayed on the right of the figure.) The upward-sloping dotted line fitted to these points shows, though it does not fit tightly, the upward trend in premarital births. Thus, the figure indicates that there has been a slight weakening of the link between marriage and childbearing.25
The observation of a certain weakening of the relation between age at
![](/openbook/0309052807/xhtml/images/img00032.jpg)
FIGURE 4-21 Early marriage, early childbearing, and premarital births over time. NOTE: See Appendix A for description of surveys. SOURCE: Standard data files from WFS, DHS-I, and DHS-II.
marriage and age at first birth is confirmed using local data sets. For example, marriage occurs very late in Mlomp, especially by rural standards, and many young women have their first—and sometimes even their second—child while still unmarried (Enel et al., 1994; Lagarde et al., forthcoming).
The Bandafassi survey [8] reveals that the relationship between age at first marriage and age at first birth can be quite complex. This survey covers three ethnic groups (Peul, Malinké, and Bedik), each of which lives in a different village.26 Age at first marriage and age at first birth differ among the groups, and the correlation between the two is weaker than one would have expected. According to the survey, women in the Peul ethnic group marry, on average, at age 16.6 and have their first child at age 17.9, or 13 months later. On average, Malinké women marry at age 19.4, but often they have already had their first child; their average age at first birth is 4 months earlier than their age at first marriage. Finally, among the
Bedik, premarital births are even more common: for these women, the average age at first marriage is 20.6 years, compared with an average age at first birth of 19.2 years, or 14 months earlier (Pison and Desgrées du Loû, 1993).
The Niakhar data [11] allow us to observe the change in the timing of first marriage since the beginning of the 1960s. Comparing the estimates reported by Cantrelle (1969) with those of Delaunay (personal communication, 1993), it is clear that the proportion of women aged 15-19 and 20-24 who are single has increased over the last 30 years. But over the same period, the fertility of women aged less than 25 has remained unchanged (see above and Figure 4-7).
Although the relationship between age at first marriage and age at first birth may be complex locally, it should not overshadow the main point: age at first marriage has increased, and, whether as a result or not, age at first birth has increased and fertility fallen. The general relationship between early marriage and early childbearing at the regional level in Senegal is shown in Figure 4-22. The figure plots the percentage of women aged 15-19 who are single and the ASFR for women aged 15-19. The data are from the 1988 census. Each of the ten administrative regions of the country is represented by two data points, one for rural women and one for urban women. The figure demonstrates the strong negative relationship between the proportion of women single and the extent of teenage childbearing.
The relationship between delayed marriage and fertility is particularly clear in Dakar. Table 4-6 shows that the percentage of women aged 15-19 who are unmarried has increased over time from 73 percent in 1978 to 91 percent in 1992-1993. Over the same period, the percentage of women having their first child before age 20 has fallen, and the fertility rate for this age group has decreased significantly (see above and Figure 4-11).
The Role of Polygyny
Polygyny is perhaps the most distinctive feature of African marriage, and any discussion of the effect of marriage on fertility would be incomplete without a brief examination of the relationship between polygyny and fertility. This relationship has been described often, so we only summarize the more important elements here. When discussing the influence of polygyny on fertility, it is important to distinguish between micro- and macrolevel effects (Pison, 1986). At the microor individual level, women in polygynous unions are usually thought to have lower fertility than those in monogamous unions for a number of reasons. First, coital frequency is likely to be lower because the husband's attention is divided among several competing wives. Second, coital frequency and fecundity tend to decline with age, and for the practice of polygyny to be sustainable, there must be a
![](/openbook/0309052807/xhtml/images/img00033.jpg)
FIGURE 4-22 Urban/rural variation in percentage of women aged 15-19 never married and age-specific fertility rates. SOURCE: 1988 census, unpublished tables.
large difference in the average age of spouses. Polygynous wives thus tend to have older-than-average husbands whose fecundity is lower. Third, polygyny makes it easier to maintain traditional birth-spacing practices, such as a postpartum taboo on sexual relations until the child is weaned. Fourth, African society places high premiums on childbearing. Consequently, if, as the result of male or female infertility, a monogamous marriage does not produce any children, that may be sufficient incentive for a man to look for a second wife. This can introduce a selection bias, so that a disproportionate number of subfecund men and women are in polygynous relationships.
At the macro- or societal level, polygyny is often associated with a number of lifestyle factors (early marriage and almost universal remarriage for widows and divorcees) that tend to raise the total length of time a woman spends in marriage. These factors tend to raise fertility.
A number of studies based on Senegalese data have contributed to the debate on the relationship between polygyny and fertility (United Nations, 1983b; Ndiaye, 1985; Pison, 1986; Garenne and van de Walle, 1989). Garenne
and van de Walle (1989:282) studied the relationship between polygyny and fertility among the Serer using time-series data collected from the Ngayokhème study [11] during 1962-1981. The authors found that fertility of monogamous wives exceeds that of polygynous wives, and the fertility of the latter decreases as their rank in the union increases. They attribute this finding to two factors: first, polygynous wives spend more time separated from their husbands; second, older men probably have lower fecundity and coital frequency.
Using WFS data, Ndiaye (1985) found few differences in fertility over the first 5 years of marriage between women in monogamous and polygynous marriages. Furthermore, the author concludes that failure to have children in the first 5 years of a marriage is not associated with the probability that a husband takes a second wife.
Table 4-7 shows the percentage of currently married women in polygynous unions, by age, in the three most recent national-level demographic surveys, WFS, DHS-I, and DHS-II. Senegal has one of the highest rates of polygyny in western Africa: in 1992-1993, 49.2 percent of women aged 30-34 and 68.4 percent of women aged 40-44 were in polygynous unions. One-quarter of all currently married women aged 15-19 are also in polygynous unions, and the vast majority of these young women must have married directly into a polygynous relationship. The increasing occurrence of polygynous unions with age reflects the fact that husbands may marry a second or third wife after a few years, so that even if women enter a monogamous marriage they are likely to lead some part of their married life in a polygynous relationship.
The incidence of polygyny differs across socioeconomic and ethnic boundaries (Table 4-8). Polygyny is usually more common among women that have no formal education and those who live in rural areas. For example, 51 percent of currently married women who have not had any formal education are in polygynous unions, compared with only 29 percent of those with secondary education or higher. Similarly, 51 percent of currently married women in rural areas are in polygynous unions, compared with only 40 percent in urban areas. Women with more education are less likely to accept being in a polygynous union. 27
The incidence of polygyny is dwindling, at least in part because of increasing urbanization and female education. Table 4-7 lends some support to this hypothesis among women aged less than 35. The percentage of these women reporting that they are in polygynous unions tends to be generally smaller in the DHS-II than it was in the DHS-I, and smaller in the DHS-I than in the WFS. If correct, this represents a very recent trend in marriage, because both the WFS and Garenne and van de Walle (1989) suggest that polygyny was rising during the 1960s and 1970s. However, while polygyny may be declining, other forms of union that closely re-
TABLE 4-7 Percentage of Currently Married Women in Polygynous Unions by Age, 1978-1973
National Survey |
15-19 |
20-24 |
25-29 |
30-34 |
35-39 |
40-44 |
45-49 |
Total |
1978 WFS |
30.6 |
35.5 |
46.5 |
57.7 |
56.5 |
65.5 |
64.8 |
48.5 |
1986 DHS-I |
27.1 |
33.2 |
40.9 |
56.0 |
63.7 |
59.7 |
62.0 |
6.5 |
1992-1993 DHS-II |
25.0 |
32.9 |
42.5 |
49.2 |
58.1 |
68.4 |
59.6 |
47.3 |
NOTE: See Appendix A for description of surveys. SOURCE: Ndiaye et al. (1988:16) and unpublished DHS-II tables. |
TABLE 4-8 Percentage of Currently Married Women in Polygynous Unions, by Age and Various Socioeconomic Variables, 1992-1993
Variable |
15-19 |
20-24 |
25-29 |
30-34 |
35-39 |
40-44 |
45-49 |
Total |
Level of Education |
|
|
|
|
|
|
|
|
None |
24.7 |
35.0 |
46.3 |
55.3 |
61.2 |
69.7 |
61.9 |
50.5 |
Primary |
27.8 |
25.6 |
24.5 |
28.3 |
45.9 |
55.8 |
27.3 |
32.1 |
Secondary+ |
16.7 |
17.2 |
26.8 |
20.0 |
35.1 |
65.4 |
12.5 |
28.8 |
Place of Residence |
|
|
|
|
|
|
|
|
Urban |
24.7 |
26.3 |
31.6 |
33.7 |
51.8 |
64.4 |
55.3 |
40.2 |
Rural |
25.1 |
35.6 |
48.2 |
58.7 |
61.5 |
70.3 |
61.8 |
50.7 |
Ethnic Group |
|
|
|
|
|
|
|
|
Wolof |
27.7 |
33.4 |
46.8 |
54.0 |
59.9 |
74.7 |
67.8 |
51.4 |
Poular |
22.3 |
38.3 |
43.6 |
48.2 |
61.3 |
69.8 |
52.8 |
46.6 |
Serer |
17.6 |
27.6 |
35.5 |
38.6 |
54.5 |
56.7 |
57.1 |
41.1 |
Mandingo |
22.2 |
28.6 |
56.5 |
58.5 |
63.5 |
68.5 |
44.4 |
50.3 |
Diola |
66.7 |
32.0 |
17.1 |
28.2 |
36.4 |
51.6 |
43.8 |
34.6 |
Other |
30.8 |
26.9 |
31.0 |
56.9 |
52.4 |
69.4 |
58.3 |
44.4 |
NOTE: See Appendix A for description of survey. SOURCE: Derived from unpublished DHS-II tables. |
semble polygyny (for example, outside wives) may be increasing. It is improbable, therefore, that a change in polygynous behavior would have influenced fertility significantly.
Summary of the Effect of Marriage Patterns on Fertility
Analysis of the influence of marriage on fertility shows that the current trend for women to delay first marriage is associated with lower fertility in the younger ages. These changes are occurring predominantly in Dakar and, to a lesser extent, in other urban centers. At the same time, the slight rise in premarital births indicates that these changes in marriage patterns do not necessarily reflect an increase in the age at first sexual encounter.
Breastfeeding and the Length of Birth Intervals
In a population where contraception is scarcely used, the average length of time between births is determined primarily by the duration of postpartum
infecundity and the duration of postpartum abstinence. The duration of infecundity (i.e., the length of the anovulatory interval) is determined primarily by the duration and intensity of breastfeeding, which tends to delay the normal resumption of menses.28 It is analytically convenient to distinguish between factors that are designed to restrict the total number of children, and those that are not parity-specific which are adopted after each birth regardless of parity to protect the health of the mother and child, rather than to limit family size. One example of a nonparity-specific practice is postpartum abstinence. Consequently, it is factored into the index of postpartum infecundability (Ci), rather than the index of contraception (Cc).
In Senegal, as in many other sub-Saharan African countries, people have a clear idea of what length of time constitutes an appropriate birth interval. A quick pregnancy following a birth is viewed as overly burdensome and fatiguing for the woman. It also leads to premature weaning for the most recent child, which is viewed as particularly harmful to the child's health (Ferry, 1981). Consequently, traditional practices have been established to ensure that birth intervals remain relatively long.
Table 4-9 provides various estimates of the median length of birth intervals and the median duration of breastfeeding, postamenorrhea, and abstinence both for the whole country and for various subgroups and specific areas. Women in Senegal maintain long birth intervals—around 30 months, on average—with birth intervals being slightly shorter in urban than in rural areas.
In Senegal, breastfeeding is essentially universal and lasts a long time. The median duration of breastfeeding is around 19-20 months, with little change over time (see Table 4-9). This is the primary reason why birth intervals in Senegal are rather long (Cantrelle and Léridon, 1971; Ferry, 1977, 1981; Cantrelle and Ferry, 1979; Guèye and Ferry, 1985; Sow, 1994a,b). Postpartum abstinence overlaps the period of breastfeeding and so has little additional effect on fertility. As shown in Table 4-9, the length of postpartum abstinence is relatively short. Averaging over the various surveys indicates that the median duration of postpartum abstinence is around 6 months. This is longer than the prescribed 40 days under Islamic code, but is still too short to have a great influence on the length of the subsequent birth interval.
Fertility behavior in the region of Sine-Saloum has been studied in depth over the past 30 years. Table 4-9 provides various estimates of the length of birth intervals made by different authors at different times on the various subpopulations of the region. There is a general consistency among the sources, despite the methodological differences among the various surveys. Breastfeeding and amenorrhea are particularly long in this region, but postpartum abstinence is usually short. Thus these sources support the
hypothesis that the length of time a woman breastfeeds is the most important determinant of long birth intervals.
In a population where lengthy breastfeeding is the norm, the main determinant of the length of breastfeeding is usually the survival of the child. If the child dies, breastfeeding ceases, and the mother quickly regains her fecundity. The difference in the length of the subsequent birth interval between women whose children die at 3 months and women whose children die at 18 months can exceed 12 months (Cantrelle and Léridon, 1971; Ferry, 1976; Lagarde et al., forthcoming). Because child mortality decreased notably in the 1970s and 1980s, one may expect a slight increase in the median duration of breastfeeding and consequently in birth intervals. At the national level, the length of breastfeeding has risen, on average, since 1978, but this has not translated into a significant change in the length of birth intervals. The implication is that changes in breastfeeding practices have occurred, such as reduced breastfeeding intensity, that have changed the relationship between breastfeeding duration and amenorrhea duration.
Closely related to the long duration of breastfeeding, the duration of postpartum amenorrhea in Senegal is also relatively long, though there are large differences between urban and rural areas. The median duration of postpartum amenorrhea varies from 16-18 months in rural areas to around 12 months in urban areas. Note that factors other than breastfeeding may also influence the postpartum period, particularly in rural areas. Chronic malnutrition associated with physical activity can contribute to long periods of amenorrhea, especially among rural women (Rosetta, 1989). However, the duration of postpartum amenorrhea decreased by approximately 2 months between the two DHS surveys, even though the length of breastfeeding appears to have risen. Again, this is consistent with a decline in the intensity of breastfeeding—which is partly a determinant of the duration of infecundity—accompanying the earlier introduction of solid food.29
In summary, birth intervals have remained relatively long, primarily as a result of extended breastfeeding. The length of postpartum amenorrhea is shorter in urban than in rural areas, although one might expect the reverse to be true on the basis of child mortality differentials alone (child mortality is lower in urban areas). The explanation lies in differences in the length and intensity of breastfeeding between rural and urban areas. On average, breastfeeding is shorter in urban than in rural areas by about 2 months. However, the difference in amenorrhea between urban and rural areas is even greater—3-5 months—suggesting either better nutrition or less intense breastfeeding in the urban areas. Finally, postpartum abstinence may be declining, although the reported decline between the DHS-I and DHS-II is too rapid to be convincing.
TABLE 4-9 Median Length of Birth Intervals and Duration of Breastfeeding, Amenorrhea and Postpartum Abstinence (in months)
Year |
Source |
Subgroup |
Birth Interval |
Breastfeeding |
Amenorrhea |
Abstinence |
National Level |
|
|
|
|
|
|
1978 |
WFS |
National |
33.4 |
18.5 |
— |
— |
1986 |
DHS-I |
National |
31.5 |
18.8 |
16.2 |
7.9 |
1992-1993 |
DHS-II |
National |
32.4 |
20.1 |
14.3 |
3.5 |
1978 |
WFS |
Urban |
32.3 |
17.6 |
— |
— |
1986 |
DHS-I |
Urban |
31.2 |
16.2 |
12.4 |
6.7 |
1992-1993 |
DHS-II |
Urban |
32.1 |
18.4 |
11.8 |
3.6 |
1978 |
WFS |
Rural |
33.8 |
19.1 |
— |
— |
1986 |
DHS-I |
Rural |
31.7 |
20.2 |
18.1 |
8.4 |
1992-1993 |
DHS-II |
Rural |
32.6 |
20.9 |
15.8 |
3.4 |
Local Studies in the Sine-Saloum Region |
|
|
|
|
|
|
1973 |
Lacombe (1973) |
Fakao |
33.8 |
— |
— |
— |
1973 |
Lacombe (1973) |
Diahanor |
34.5 |
— |
— |
— |
1963-1968 |
Cantrelle and Léridon (1971) |
Niakhar |
30.9 |
24.3a |
— |
2.3a |
1963-1968 |
Cantrelle and Léridon (1971) |
Paos-Koto |
30.9 |
24.0 |
— |
2.3a |
1968-1969 |
Cantrelle and Ferry (1979) |
Thiénaba |
32.6a |
23.7 |
17.9 |
2.3a |
1981-1982 |
Mbacké (p.c.) |
Thiénaba |
— |
21.0 |
17.0 |
2.0 |
1981-1982 |
Mbacké (p.c.) |
Fissel |
— |
22.0 |
17.0 |
3.0 |
1981-1982 |
Rosetta (1989) |
Ndiaganio |
— |
22.8 |
19.0 |
— |
1991 |
Project |
Niakhar |
31.0 |
26.0 |
— |
— |
|
Niakhar |
|
|
|
|
|
|
(1992) |
|
|
|
|
|
Other Rural Studies |
|
|
|
|
|
|
1980-1992 |
Pison and Desgrées du Loû (forthcoming) |
Bandafassi |
32.9 |
— |
— |
— |
1985-1992 |
Lagarde et al. (forthcoming) |
Mlomp |
29.6 |
— |
— |
5.0 |
Semi-rural Studies |
|
|
|
|
|
|
1968-1969 |
Ferry (1981) |
Khombole |
32.3a |
19.7a |
15.4a |
2.4a |
NOTES: —, not available; p.c., personal communication; see Appendix A for description of surveys. a Mean duration used instead of median. SOURCES: Cantrelle and Ferry (1979), Cantrelle and Léridon (1971), Ferry (1981), Lacombe (1973), Lagarde et al. (forthcoming), Project Niakhar (1992), Pison and Desgrées du Loû (forthcoming), Rosetta (1989) and WFS, DHS-I, and DHS-II data files. |
Contraception
Though prolonged breastfeeding and postpartum abstinence result in avoidance of pregnancy and so may be considered a form of contraception, the traditional intention was not to restrict family size but rather to ensure adequate spacing between births as a way of maximizing the number of surviving children. Despite (or because of) effective traditional birth-spacing practices, use of modern contraception had a relatively late start in Senegal. Although efforts to promote use began before 1980 (Osmanski et al., 1991; République du Sénégal, 1990), it was not until that year that a 1920 French law outlawing the use of modern contraceptives was repealed (République du Sénégal, 1991). Since 1980, political support for family planning has increased within the government, and international donor agencies have played a substantial role in building the country's family planning effort (République du Sénégal, 1990). Both the number of family planning clinics and the number of family planning acceptors increased substantially between 1985 and 1990 (Osmanski et al., 1991). (See the appendix to Chapter 5 for a general description of the expansion of health services, of which family planning has been a part.)
Not surprisingly, knowledge of contraceptives has increased over the last 15 years. As can be seen in Table 4-10, knowledge of modern methods of contraception increased more than threefold in the 15 years between 1978 and 1992-1993. By 1992-1993, over 70 percent of currently married women knew of a modern method. This increase in knowledge has taken place in both rural and urban areas and at all educational levels.
Table 4-10 also shows the increases that have taken place in contraceptive use. Modern contraceptive use is still very low (less than 5 percent of currently married women reported that they were current users in 1992-1993), but the proportion of users has increased over the recent past. Predictably, contraceptive use is much more frequent in cities (where 12 percent of women use modern contraception, according to DHS-II data) than in rural areas (1 percent), and among women with secondary or higher (29 percent) or primary (13 percent) schooling than among those with no formal education (2 percent). The most popular modern method of contraception is the pill; almost half of the currently married women who were current users of modern methods in 1992-1993 reported using the pill. Other methods of contraception reported include intrauterine devices (IUDs) (1 percent), female sterilization (0.4 percent), and condoms (0.4 percent).
The relatively frequent use of contraception in urban areas, especially among women who have attended school, partly explains why marital fertility has decreased among that population. Elsewhere, for the great majority of the population, contraception is not yet sufficiently widespread to have had a noticeable influence on the total fertility rate.
TABLE 4-10 Percentage of Currently Married Women Aged 15-49 Who Know of Modern Contraception and Who Use Contraception
|
|
|
|
Contraceptive Users |
|
Year |
Source |
Subgroup |
Knowledge of Modern Methods |
Traditional Method |
Modern Method |
National Level |
|
|
|
|
|
1978 |
WFS |
n.a. |
20.4 |
3.3 |
0.6 |
1986 |
DHS-I |
n.a. |
67.6 |
9.0 |
2.4 |
1992-1993 |
DHS-II |
n.a. |
70.3 |
2.7 |
4.8 |
Place of Residence |
|
|
|
|
|
1986 |
DHS-I |
Urban |
87.5 |
7.5 |
6.7 |
|
|
Rural |
58.1 |
9.6 |
0.3 |
1992-1993 |
DHS-II |
Urban |
88.4 |
4.3 |
11.8 |
|
|
Rural |
61.6 |
1.9 |
1.4 |
Level of Instruction |
|
|
|
|
|
1986 |
DHS-I |
None |
64.0 |
8.8 |
1.0 |
|
|
Primary |
86.9 |
9.0 |
5.9 |
|
|
Secondary+ |
97.3 |
10.7 |
22.2 |
1992-1993 |
DHS-II |
None |
65.5 |
1.9 |
2.2 |
|
|
Primary |
93.7 |
6.3 |
12.8 |
|
|
Secondary+ |
98.6 |
7.2 |
29.3 |
NOTE: Injection and sterilization were included as modern methods in 1986 and 1992-1993, but not in 1978; see Appendix A for description of surveys. SOURCES: République du Sénégal (1981:Table 4.2.1B, p. 265); DHS-I: Ndiaye et al. (1988:Tables 4.1, 4.3, 4.7, and 4.8, pp. 43, 45-46, 52, 54-55); DHS-II: unpublished tables supplied by Macro International, Inc. |
Additional information about the potential demand for family planning can be obtained by examining women's responses to questions relating to (1) their desire for (and timing of) more children, and (2) their reported ideal family size. Differences in women's responses to such questions have been observed between the DHS-I in 1986 and the DHS-II in 1992-1993, suggesting that the potential demand for family planning may be increasing. For example, the percentage of currently married women who want to have another child decreased from 79 percent in the DHS-I to 70 percent in the DHS-II (Table 4-11). At the same time, the proportion stating they wanted no more children changed little, from 19 to 20 percent, but those undecided increased from 2 to 7 percent. Although this change in preferences may be
TABLE 4-11 Percent Distribution of Currently Married Women by Desire for Children, According to Number of Living Children
|
Number of Living Children |
|||||||
Preference |
0 |
1 |
2 |
3 |
4 |
5 |
6+ |
Total |
Wants to have another child |
|
|
|
|
|
|
|
|
1986 DHS-I |
97.0 |
97.2 |
94.7 |
88.6 |
80.1 |
66.3 |
34.4 |
79.3 |
1992-1993 DHS-II |
89.3 |
91.5 |
89.1 |
80.4 |
74.8 |
58.8 |
34.8 |
69.9 |
Wants no more children |
|
|
|
|
|
|
|
|
1986 DHS-I |
1.9 |
2.1 |
4.2 |
10.9 |
18.8 |
31.0 |
61.6 |
19.0 |
1992-1993a DHS-II |
1.0 |
1.4 |
4.2 |
9.1 |
16.6 |
28.3 |
52.6 |
20.1 |
Undecided |
|
|
|
|
|
|
|
|
1986 DHS-I |
0.9 |
0.7 |
1.1 |
0.5 |
0.9 |
2.7 |
3.8 |
1.6 |
1992-1993 DHS-II |
2.8 |
3.0 |
4.3 |
8.2 |
6.1 |
9.6 |
9.5 |
6.6 |
Declared infecund |
|
|
|
|
|
|
|
|
1986 DHS-I |
— |
— |
— |
— |
— |
— |
— |
— |
1992-1993 DHS-II |
6.6 |
4.0 |
2.3 |
2.3 |
2.5 |
3.4 |
3.0 |
3.2 |
NOTE: See Appendix A for description of surveys. a Includes women who have been sterilized. SOURCES: DHS-I: Ndiaye et al. (1988:Table 5.1, p. 68); DHS-II: unpublished tables supplied by Macro International, Inc. |
partly the function of a change in the wording of the questionnaire, it appears to have occurred across almost all parities and all age groups (see Tables 4-11 and 4-12).
Another indicator suggesting that the demand for family planning may increase in the future is the decline, since 1978, in the number of children women reported as "ideal." The ideal number of children for all women fell from 8.5 in 1978 to 6.8 in 1986, and then to 5.9 in 1992-1993. As can be seen in Figure 4-23, this trend is apparent for all age groups. However, 5.9 children per woman is still very close to the potential supply of children, so that with the trend towards later age at first marriage, the small decline in ideal family size can be realized without resorting to contraception.
Abortion
As in most developing countries, there are few firm statistics on the extent of abortion in Senegal. Although reliable statistics are unavailable,
TABLE 4-12 Proportion of Currently Married Women Aged 15-49 by Desire for Another Child, According to Age of Woman
|
Age of Woman |
|||||||
Preferences |
15-19 |
20-24 |
25-29 |
30-34 |
35-39 |
40-44 |
45-49 |
Total |
Wants to have another child |
|
|
|
|
|
|
|
|
1986 DHS-I |
98.5 |
98.2 |
91.8 |
78.1 |
61.7 |
45.8 |
33.9 |
79.3 |
1992-1993 DHS-II |
94.2 |
93.7 |
86.1 |
72.7 |
57.6 |
34.8 |
20.2 |
69.9 |
Wants no more children |
|
|
|
|
|
|
|
|
1986 DHS-I |
1.2 |
1.7 |
7.6 |
19.4 |
36.5 |
50.5 |
59.0 |
19.0 |
1992-1993a DHS-II |
1.8 |
1.7 |
7.3 |
17.6 |
33.0 |
47.8 |
52.1 |
20.1 |
Undecided |
|
|
|
|
|
|
|
|
1986 DHS-I |
— |
0.1 |
0.6 |
2.3 |
1.6 |
3.7 |
7.1 |
1.6 |
1992-1993 DHS-II |
3.8 |
4.0 |
5.4 |
8.5 |
7.2 |
8.5 |
10.8 |
6.6 |
Declared infecund |
|
|
|
|
|
|
|
|
1986 DHS-I |
— |
— |
— |
— |
— |
— |
— |
— |
1992-1993 DHS-II |
0.2 |
0.4 |
1.2 |
1.3 |
2.0 |
8.8 |
17.0 |
3.2 |
NOTES: Totals by age for each year may not equal 100; in both years, 0.1 percent of cases were missing. See Appendix A for description of surveys. a Includes women who have been sterilized. SOURCES: DHS-I: Ndiaye et al. (1988:Table 5.2, p. 68); DHS-II: unpublished tables supplied by Macro International, Inc. |
![](/openbook/0309052807/xhtml/images/img00034.jpg)
FIGURE 4-23 Variations in ideal family size over time. NOTE: See Appendix A for description of surveys. SOURCES: Standard data files from WFS, DHS-I, and DHS-II.
illegal abortion is generally considered to be an increasingly serious problem, especially in urban areas (Ferry, 1981; Pillsbury, 1990; Diouf, 1994). Because abortion can carry a fine that ranges from 20,000-50,000 CFA (US$40 to US$100) and 6-10 months in prison, both for the woman undergoing the abortion and for the performer, abortions are often carried out clandestinely. Botched abortions are a major cause of all hospital admissions for women of reproductive age. In 1988, abortion complications were one of the two major reasons for the emergency transfer of women from Pikine to Dakar (Guèye et al., 1989, cited in Pillsbury, 1990).
Limited data on abortions are available from several sources. Information on abortion from the WFS is limited because the relevant questions did not distinguish between spontaneous and induced abortions, but it appears that between 10-12 percent of women reported having had an abortion (Diouf, 1994). Sow (1985) estimates that at the time of the WFS, women had, on
average, 0.3 abortions over their lives, although the basis for this estimate is unclear.
Because of the moral and religious sensitivity surrounding the issue, no questions on abortion were asked in the 1986 DHS-I, but there was a rather roundabout attempt to measure the abortion experience of women in Senegal in the 1992-1993 DHS-II. In DHS-II, women were asked whether they had ever had an unwanted pregnancy, and, if so, what they had done about this pregnancy. Out of 6,299 women, 1,131 admitted that they had had an unwanted pregnancy. Of these women, 1,058 carried the pregnancy to term and 37 were currently pregnant. Only 19 women admitted that the pregnancy was terminated, either spontaneously or by induced abortion, although a further 15 women admitted that they had attempted to induce an abortion but had failed.*
The prevalence of abortion in the WFS survey appears to be in reasonably close agreement with a survey undertaken in 1986 in Pikine. Out of 9,196 pregnancies, 804 or 8.7 percent of them resulted in spontaneous or induced abortions (Antoine and Diouf, 1989:507). The frequency of abortion varied by age of women, marital status, ethnic group, level of instruction, and number of children ever born. Abortion in Pikine appeared to be particularly frequent among 15-19 year olds and among unmarried women (Diouf, 1994). However, abortions were also fairly common among women over age 30 and among married women.
In the Bongaarts model presented above, the index measuring the effect of abortion on fertility is set to 1 for the WFS and the DHS-I. The assumption is that, given the general high demand for children, deliberate abortion is not important in determining final fertility outcomes. This assumption almost certainly leads to a serious underestimation of the extent of abortion among adolescent schoolgirls. Taking the DHS-II figures at face value implies an index of Ca of .999, an unsatisfactory result and no improvement over the basic (poor) assumption of zero abortion. Clearly, much more research on abortion is needed, especially among young women, although the effect of abortion on total fertility is probably of little importance.30
Sterility
For pathological or other reasons, some couples will never bear children (primary infertility). Others lose the ability to bear children following an earlier birth (secondary infertility). Sterility can occur either naturally, primarily by aging, or pathologically, primarily by complications resulting
from infection with a sexually transmitted disease, which in Africa is most frequently gonorrhea.
In the absence of fertility regulation, a reasonable estimate of the level of infertility is given by the proportion of women aged 40-49 who have never had a live birth. In a population with a ''normal" level of sterility, one would expect around 5 percent of women to be nulliparous (Frank, 1983), although the level depends strongly on the age pattern of first births. In Senegal, the figure ranges from 2.7 to 4.6 percent across the last three national-level surveys, WFS, DHS-I, and DHS-II. The small number of nulliparous women in each sample does not permit us to establish a trend between surveys. Other local studies confirm the general level of sterility (Ferry, 1981) or indicate slightly higher levels, possibly reaching 8 percent, as in Bandafassi (Pison and Desgrées du Loû, forthcoming). In summary, the prevalence of primary sterility in Senegal appears rather normal, and sterility scarcely diminishes the reproductive potential of Senegalese women.
SUMMARY AND CONCLUSIONS
The 1978 WFS established that fertility in Senegal was very high in the mid-1970s—7.2 children per woman. There are some indications that fertility may have been lower two or three decades earlier and increased in the preceding years. Fertility declined by around one child per woman between the mid-1970s and the beginning of the 1980s, and the differences in the fertility levels among various regions and socio-cultural subgroups have increased.
The decline has occurred almost exclusively among women under age 30. A comparison of ASFRs between 1975-1978 and 1989-1992 reveals that the decline in fertility among women aged 15-19 is approximately twice as large as that among women aged 20-29, which, in turn, is almost twice as large as that among women over age 30. After age 30, fertility has remained much the same.
The driving force behind these changes has been a trend towards later marriage. In some areas of Senegal, for example in Casamance, women have always married relatively late. However, the larger trend towards later marriage probably began in Dakar in the early 1980s and has been spreading to other urban areas ever since. The pattern of later marriage is closely linked to formal education, although signs of change are emerging even among women who have never attended school. Little of the fertility decline appears to be attributable to either a decrease in ideal family size or an increase in the use of modern contraception. Nationally, use of modern contraception has enjoyed only modest success. Contraceptive use among currently married women increased from 1 percent in 1978 to 5 percent in
1992-1993. However only certain subgroups of the population use modern contraception, so that the absolute number of users is still very low.
In summary, these features indicate that Senegal's small fertility decline is unlike those that have occurred recently in other sub-Saharan African countries. In Botswana, Kenya, and Zimbabwe—usually considered the three countries in the vanguard of an African fertility transition—fertility declines are associated with increased use of modern contraception. In Senegal, the decline is associated with a trend toward later marriage. In addition, although Botswana, Kenya, and, to a lesser extent, Senegal and Zimbabwe have all experienced a decline in teenage marriages, Senegal has not experienced such a dramatic separation of teenage marriage and teenage fertility. Consequently, Senegal has not seen the same increase in premarital fertility as these other countries.
The Senegalese pattern matches more closely, but not exactly, the pattern found in certain Northern African countries during the first phase of their fertility declines. For example, most of the first decreases in-fertility observed in countries such as Algeria, Egypt, and Tunisia can be attributed to later age at first marriage (Fargues, 1989; National Research Council, 1982). It is interesting to note that in these countries, the initial phase of fertility decline was followed immediately by a second phase linked to a substantial decline in the demand for children and a corresponding increase in modern contraceptive use among married women. Whether Senegal follows this pattern and experiences a second phase of fertility decline immediately following the first remains to be seen.31 Much will depend on what happens to the demand for children. There has been a trend towards wanting fewer children that extends across all parities and all age groups of women. However, current preferences still lie very close to the physiological maximum, assuming a continued regime of delayed marriage and long birth intervals.
If the education sector can overcome its fiscal constraint and continue to make advances in primary and secondary school enrollment for women, further declines in fertility can be expected in the near future. In rural areas, a further reduction in fertility can be achieved solely by later marriage. Urban areas, particularly Dakar, have already experienced most of the decline in actual fertility that is achievable solely by an increase in the age at marriage. Future fertility reductions will have to await greater coverage of modern contraception.
If the government of Senegal wishes to influence the timing and shape of the fertility decline, it must implement strong policies that target both girls and women of reproductive age. For girls, policy should be aimed at increasing formal education. For women, action should be taken to promote the availability of contraception while increasing women's functional literacy and reducing their domestic burden. When combined with interventions
aimed at improving maternal and child health, such policies should have a good chance of making women more receptive to the use of modern contraception.
NOTES
1. |
The TFR expresses the total number of children a woman would have if she experienced the same level and pattern of fertility throughout her reproductive life as those in effect at the time of the survey. An advantage of using the TFR over other measures of fertility, such as the crude birth rate, is that it is independent of the age structure of the population. |
2. |
See Appendix A for a complete list and brief description of all the major surveys used in this report. |
3. |
Apart from not reporting children that died, respondents frequently are unable to identify the beginning of the reference period, and this can also lead to systematic bias. Fortunately, this was probably not a serious problem in the 1988 census (see endnote 6). |
4. |
For example, the infant mortality rate based on data for deaths occurring during the preceding 12 months was estimated to be 93 deaths per 1,000 live births, or approximately half the presumed actual rate. |
5. |
There was also a tendency in the WFS to shift the dates of births towards the survey date, and there were a number of omissions of births for more remote time periods (Guèye, 1984; van de Walle and Foster, 1990). |
6. |
The 1988 census was conducted between May 20 and June 3, but May 27, 1988, was adopted as the official reference date. The holiday of "Korité," which marks the end of Ramadan and is widely celebrated in Senegal, fell on May 29 the year before, nearly 12 months to the day before the census reference date. Consequently, the holiday was used as a reference point to identify any births and deaths that occurred in the 12 months preceding the census. |
7. |
B60s measure the proportion of women in an age cohort (a group of women defined by 5-year age groups at the time of the survey) who, having attained an nth birth, go on to an (n + 1)th birth within 60 months. Thus B60s are a form of censored parity progression ratios (CPPRs). B60s have been designed to correct for an inherent bias in the usual calculation of CPPR suggested by Rodriguez and Hobcraft (1980) that results from systematically excluding women with long birth intervals (see Brass and Juarez, 1983)- |
8. |
That is, we report the probabilities that a woman in a particular cohort who has had her nth birth goes on to have an (n + 1)th birth within 60 months of her nth birth, and subsequently goes on to have an (n + 2)th birth within 60 months of her (n + 1)th birth. |
9. |
The term "grand region" is used throughout this report in reference to the four WFS and DHS regions, as opposed to the ten administrative regions. |
10. |
Other data from local studies in the Sine-Saloum region are not presented here because of the small population size from which they are taken, their poor quality that is acknowledged by those responsible for their collection, and the lack of published data. |
11. |
See Pison et al. (1991), Pison and Desgrées du Loû (forthcoming), and Appendix A for additional information. |
12. |
See Pison et al. (1991), Lagarde et al. (forthcoming), and Appendix A for additional information. |
13. |
Note that given the longitudinal nature of the survey, the omission of births in recent periods is probably not a large problem. For the most recent period, 1985-1992, the estimates of the fertility rate are based on the number of recorded events between multiple rounds of the survey. Thus the most recent estimates take advantage of the longitudinal nature of the survey. Fertility estimates for the earlier periods are from retrospective biographical data. |
14. |
The census did not include an additional question on the number of children a women had ever borne, which would have enabled us to use indirect estimation techniques to correct for underreporting of births, nor did it ask one further question, on the number of children that had died, which would have provided information for the estimation of child mortality. |
15. |
Briefly, this neoclassical model holds that households decide on the number of children to have by attempting to maximize household utility subject to time and income constraints. Children, both in quantity and quality, enter the problem in exactly the same way as other goods. Together with other competing sources of satisfaction, they create household demand, which can be satisfied by a combination of (1) income from either a household's wealth or the labor-force participation of its members, and (2) time. Because the bearing and rearing of children is an extremely time-intensive activity, especially for women, having children competes with other activities, such as working in the labor force. Hence, women are forced to trade off their time among (1) working in the labor force so that they can purchase additional goods and services, (2) bearing and raising children, and (3) doing other things. The problem is further complicated by the fact that the wages and opportunities for women are likely to be functions of previous labor-market experience, which in turn depends on the timing and frequency of earlier children in a woman's early working life (Willis, 1973; Becker, 1981; Schultz, 1981). Note, however, that this demand-side approach is often viewed as being overly simplistic in developing countries because it largely ignores the economic contribution of children through farm work and child-minding activities—which can be considerable—and the role of children as care providers for their parents in old age (Anker and Knowles, 1982). These factors provide an important incentive for high fertility. Finally, the neoclassical model assumes that individual households act alone and ignores the role extended families play as decision makers and, in some cases, as child rearers. |
16. |
Ferry (1981) observes that fertility rates in Dakar were at least as high as, if not higher than, those in rural areas in the late 1960s. |
17. |
See, for example, Lacombe (1972) who provides estimates of fertility rates by ethnic group from as early as 1957. |
18. |
Comparisons among ethnic groups, and in particular between the Serer and the other groups, are made difficult because of a tendency for Serer women to declare themselves to be Wolof in surveys. This practice, which has been termed "Wolofization" (Ferry, 1977), is more prevalent in urban than in rural areas. Thus the Serer ethnic group disproportionately comprises women who remain in rural areas, which introduces an important selection bias into the data. |
19. |
Secondary education of women is also likely to be correlated to the probability that the husband works in the formal sector, which itself can have an independent effect on fertility outcomes (see Mboup, 1993). |
20. |
See Chapter 2 for a description of each of the socioeconomic indicators. |
21. |
TFR = total fertility rate; Cm = index of marital fertility calculated using the formula Cm = TFR/TMFR, where TMFR is the total marital fertility rate. (Both rates are estimated for the 4 years prior to the survey.) Cc= index of contraception, calculated as Cc = 1-1.08ue, where u is the current contraceptive use prevalence rate among women in sexual unions, and e is the average effectiveness of the chosen contraceptive methods, calculated as the weighted average of the proportion of women using each method times the effectiveness of the method. Following Jolly and Gribble (1993), we have divided other methods into other traditional methods and other modern methods. The following effectiveness levels have been assigned to specific methods: sterilization = 1.0; IUD, Depo-Provera, Norplant = .95; pill = .90; other modern methods = .70; traditional methods = .30. Breastfeeding and postpartum abstinence are not included in contraception, since they are not parity-specific practices and are adopted to protect maternal and child health, rather than as a means to limit family size. Furthermore, there is substantial variation among surveys in the proportion of women practicing abstinence as a method of contraception, which probably has more to do with differences in definitions and interviewers' instructions than with changes in women’s behavior. Finally, traditional methods, such as the use of ritual charms or potions (gris-gris), have been excluded; in essence, these methods are equated with using no contraception. Ci= index of postpartum infecund interval, calculated using the formula Ci = 20/(18.5 + i), where i is the mean number of months of postpartum infecundability (estimated as the mean number of months of postpartum amenorrhea or abstinence, whichever is longer) for women in union. Ca = index of induced abortion, taken to be 1.00 in the absence of additional information; and Ip = index of sterility calculated as Ip = (7.63 - .11s)/7.3, where s is the proportion of women aged 40-49 who have never had any children. |
22. |
It would be a mistake to read too much into the differences in the Ci index among surveys. The WFS did not ask women whether they were currently amenorrheic, so the Ci index for the WFS was imputed using a formula provided by Bongaarts and Potter (1983), based on the duration of breastfeeding. |
23. |
When examining age at first marriage in any African setting, it is important to remember that marriage in Africa may be better described as a process than an event (van de Walle, 1968, 1993; Lesthaeghe et al., 1989; Pison, 1989; Meekers, 1992; Working Group on the Social Dynamics of Adolescent Fertility, 1993). Unlike births or deaths, entry into marriage may take place over an extended period of time. Consequently, the date on which a marriage occurs is subject to several interpretations. First, marriages are usually marked by a ceremony and the transfer of a bridewealth that can range from symbolic tokens to large transfers of cash or goods spanning several years. However, the payment of bridewealth, the ceremony, the cohabitation of spouses, and the consummation of the marriage often occur several months apart and not necessarily in the same order (Meekers, 1992). Among the Poular, for example, consummation of marriage may occur several years after a marriage cer- |
26. |
The Malinké belong to the Mandingo group, and the Bedik a small group also called the Tenda, are close to the Bassari. |
27. |
"Polygyny is often perceived by more educated women as being a type of union that is incompatible with their social aspirations" (République du Sénégal, 1981:81). |
28. |
The calculation of Ci can be done directly on the basis of the duration of postpartum amenorrhea or indirectly on the basis of a model linking the duration of amenorrhea to that of breastfeeding. On the basis of data from the DHS-I, comparison of direct and indirect results shows that the period of infecundability is longer if one considers declarations of postpartum amenorrhea than if one infers the length of amenorrhea indirectly from the duration of breastfeeding (Jolly and Gribble, 1993). |
29. |
We have no data on the intensity of breastfeeding over time. |
30. |
An alternative approach is to calculate Ca indirectly from Table 4-5. Assuming a level of total fecundity of 15.3 children per woman (Bongaarts and Potter, 1983) and applying the formula TFR = Cm x Cc x Ci x Ca x Ip × 15.3, implies levels of Ca between .82 and .88 across the three surveys. |
31. |
An important element in these other countries' fertility decline is that as women’s educational status has improved, the gain due to more widespread contraception and the loss due to shorter breastfeeding—the most important determinant of amenorrhea—have practically canceled each other out (Fargues, 1989). In contrast, recent data from the DHS-II do not indicate that the average length of breastfeeding is decreasing in Senegal. |