Trends in Age at Marriage and the Onset of Fertility Transition in sub‐Saharan Africa

Abstract

Over the last 40 years, the question of the “African exception” has regularly come to the forefront in the discussion of fertility trends. In the 1980s, there was uncertainty about when fertility decline would commence throughout the region: while fertility was declining at a steady pace in Latin America and Asia, decline was evident in only a minority of sub-Saharan countries and, indeed, some countries showed fertility increase. As of the 1990s there was evidence of fertility decline in most countries of sub-Saharan Africa, and it appeared that sub-Saharan Africa was following the historical pattern of the other major regions. With a slow pace of fertility decline, or even stagnation at relatively high levels in various countries (Bongaarts 2008), the question of Africa's exceptionality has resurfaced. The fertility level in sub-Saharan Africa is the world's highest (5.1 children per woman versus 2.2 in Latin America and Asia in 2010–15; United Nations 2015a) (see Table 1). Compared with the experience of other world regions, sub-Saharan Africa stands apart not only in terms of fertility levels, but also with regard to a flatter age pattern due to longer birth intervals, the persistence of high ideal family size, and a low level of contraceptive use (Bongaarts and Casterline 2013). Sub-Saharan Africa also deviates from standard international patterns in terms of nuptiality. The traditional nuptiality regime has been defined by a particular combination of features, both for first marriage (early marriage for girls, a large age gap between spouses, almost universal marriage for both sexes) and for later conjugal life (polygamy, prompt and widespread remarriage for widowed and divorced women of childbearing age) (Lesthaeghe et al. 1989; United Nations 1988, 1990; van de Walle 1968). This dominant pattern existed with geographical differences, however, and also exceptions (especially Southern African countries). It has also been affected by significant changes over recent decades, especially through the increase in women's age at first union (Antoine 2006; Garenne 2004; Hertrich 2007; Lloyd 2005; Mensch, Grant, and Blanc 2006; Mensch, Singh, and Casterline 2005; Ortega 2014; Shapiro and Gebreselassie 2014; Westoff 2003), a narrowing gap between male and female age at marriage, and recent evidence of polygyny decline in Western Africa (Antoine and Marcoux 2014; Hertrich 2006). Despite these trends, sub-Saharan Africa still stands out in international comparisons both for the youngest age at first union for women and the largest age difference between spouses at first union (see Table 1). In 2010, the median age at first union was 21.2 years for women in sub-Saharan Africa, 1.4 years earlier than in Asia, and much earlier than in other parts of the world (25 to 28 years). Except in Southern Africa, the pattern is even earlier (20.3–20.5) at subregional levels, but it has an equivalent counterpart in the subregion of South Asia. The difference in age at union between males and females also remains significantly higher in Africa (5.5 years on average) than in the rest of the world, where the regional average is around 3 years or less. To what extent are these sub-Saharan fertility and nuptiality patterns bound up with each other? Are nuptiality changes part of the fertility transition? Is fertility decline possible in a context of early marriage? Is there empirical evidence of changes in age at marriage before or at the onset of the fertility transition? In this chapter I adopt a comparative approach to examining long-term trends in female age at marriage and fertility in sub-Saharan Africa, with a focus on continental countries having at least 1 million inhabitants. My database on nuptiality includes over 360 censuses and national surveys conducted in these 39 countries since the 1960s. I analyze the association between changes in age at first union and the onset of fertility transition, examining whether there is a typical pattern of association followed by most countries in the region. Questions about nuptiality changes in connection to the fertility transition are present in the classical literature on the demographic transition. The issue is of particular interest in sub-Saharan Africa, where the traditional marriage regime strongly supports high fertility. The idea that nuptiality change is part of the demographic transition was conceptualized in the 1960s. In Kingsley Davis's “theory of change and response” (1963), the restriction of nuptiality (through increases in age at marriage and/or in permanent celibacy) is, like emigration or limitation of marital fertility, one of the “multiphasic responses” to the sustained natural increase generated by mortality decline. Postponement of marriage is not a deliberate effort to reduce fertility; but, along with migration, it has often been a first collective response to demographic pressure, as it is easier to adopt than a restriction on marital fertility (United Nations 1990). Ansley Coale (1967, 1974) subsequently distinguished two steps in the fertility transition: first a “Malthusian transition” in which general fertility is lowered by the restriction of marriages; and, second, a “neo-Malthusian transition” in which a decrease in marital fertility resulted from the deliberate choice of couples to limit the number of their children. This two-step approach was later adopted by Jean-Claude Chesnais (1986) in his extensive work on demographic transition. Based on his assessment of the countries where fertility declined before the 1980s (i.e. excluding sub-Saharan Africa) and despite exceptions in Latin America, he concluded that nuptiality transition could be considered as a first step in the fertility transition across a large part of the world: “In all countries where there is appropriate statistical information, control of marriages preceded birth control by couples” (ibid., p. 381). More broadly, the robustness of this theory relies on a two-stage process where nuptiality change is a prelude to deliberate birth control; it acts as a regulator of general fertility but can occur before the sustained fertility decline that fixes the onset of fertility transition.1 The pronatalist nature of the traditional African nuptiality system has been widely documented2 and can be summarized by two aspects. First, it maximizes the span of a woman's reproductive life that is assigned to reproduction. Unlike pre-transitional Europe, where late marriage and permanent celibacy restricted the potential of fertility, in sub-Saharan Africa the traditional fertility-inhibiting factors operate mainly within marriage by means of the postpartum infecundability that results from long breastfeeding and postpartum abstinence (Page and Lesthaeghe 1981). A woman's life course is structured by marriage and reproduction: she is married at a young age; and if the marriage ends (through divorce or widowhood), she quickly remarries—at least while she is still of childbearing age. In the 1980s, the proportion of reproductive time spent out of union was usually below 20 percent (United Nations 1986), with an average of around 15 percent (Bongaarts, Frank, and Lesthaeghe 1984) and values below 10 percent in various Western African populations. Polygyny3 is one of the keys to the smooth running of this system, as it makes the marriage market more flexible. Indeed, in case of marital disruption, a woman can remarry rapidly without waiting for a single partner to become available (Locoh 2006; Hertrich 2006). The second aspect of the association between the nuptiality and high-fertility regimes is related to the organization of the conjugal unit and of gender relations. Institutional arrangements converge to limit the conjugal unit to its reproductive tasks and to impede conjugal intimacy and autonomous decision-making. The traditional marriage system largely contributes to building weak relationships between spouses and, therefore, to hindering the elaboration of common and independent fertility decisions (Caldwell 1982; Lesthaeghe 1980; Lesthaeghe et al. 1989; Mason 1993; National Research Council 1993; Ryder 1983). In addition large age gaps between spouses creates a distance between them as a result of the generational and cultural gap between the partners and reinforces the subordinate position of the wife. Polygyny and the high risk of marital disruption are other causes of a frail conjugal bond, because they create uncertainty and a climate of distrust between spouses (Antoine 2006; Hertrich and Locoh 1999). The weakness of the conjugal bond is seen as enhancing fertility through different paths. First, couples have little incentive to question normative behaviors when there is little privacy and opportunity for discussion between spouses. Husbands and wives usually have separate budgets and therefore little opportunity to discuss the full costs of childrearing (all the more so given that the costs of children are often spread across a larger family network). Second, the frailty of conjugal bonds provides women a powerful rationale for high fertility. In rural patrilineal societies, where women have limited access to land and economic assets, having children is a critical means to securing access to household resources and to consolidating their status in relation to husbands, in-laws, and possible co-wives. According to a recent study (Lambert and Rossi 2016), high fertility remains a strategy for women in the face of uncertainties and family rivalries in Senegal. According to these considerations, it makes sense to anticipate that fertility transition requires—or at least would be facilitated by—a loosening of traditional marriage patterns. Depending on the analytical approach, one can expect changes in nuptiality and fertility trends to be either simultaneous or sequential. The first case (simultaneity) refers to a direct, mechanical effect of nuptiality on fertility. It is conceptualized through the framework of the proximate determinants of fertility. At the population level, other things being equal, a decline in the time spent in union (i.e., having a regular sexual life) will lower fertility. Modeling fertility by using a large body of international data has confirmed nuptiality as one of the four key proximate determinants of fertility (the three others being contraception, postpartum infertility, and abortion) (Bongaarts 1978, 1992). According to this outline, the inhibiting effect of delayed nuptiality increases, on average, in the first stage of fertility transition, but the impact of contraception becomes dominant and much stronger as the fertility transition progresses (Bongaarts 1992). In some regions, like North Africa in the 1970s and the 1980s, the postponement of marriage was a leading cause of fertility decline (Westoff 1992; Ouadah-Bedidi and Vallin 2000). In sub-Saharan Africa, the picture is more mixed. Data and studies are fragmentary concerning the impact of nuptiality on fertility at the outset of the transition. Country-level studies usually provide evidence of changes in the age at first union at the onset of fertility decline. For Eastern and Southern Africa, Harwood-Lejeune (2001) estimates that one-sixth to one-third of the fertility declines in the 1980s and early 1990s is explained by rising age at marriage. Two recent large-scale comparative studies (Garenne 2014; Shapiro and Gebreselassie 2014) conclude that delayed marriage in most countries in the region had a small impact on fertility decline when compared to the overwhelming contribution of contraception. However, these studies examine long periods of time (comparing the results from the most recent DHS to those from the first available one or to the estimates at the onset of fertility decline); therefore, the possible effect of nuptiality at the onset of fertility decline is difficult to capture and is probably underestimated because it is diluted over time and superseded by the impact of contraception. In the second approach, nuptiality changes first—that is, before and possibly as a precursor to fertility decline. This is a possible scenario if the mechanical inhibiting effect of delayed nuptiality on fertility is counterbalanced by other changes, for instance if there is an increase in marital fertility. Here, the possible link between nuptiality and fertility decline does not necessarily have to be understood in a deterministic way: a single factor (for instance, increases in level of education) may both raise the age at marriage and increase contraceptive uptake. The general assumption is that the delay between later marriage and fertility decline corresponds to a period of change in the context of reproduction, especially in terms of increased individual autonomy and, possibly, conjugal autonomy. This type of scenario (delayed age at marriage without simultaneous fertility decline) has been considered for Africa by Chojnacka (1993, 1995). The comparative analysis of long-term trends in age at marriage and fertility throughout sub-Saharan Africa presented below will provide the opportunity to examine the occurrence of both scenarios. The objective here is to describe historical trends in age at first union and fertility and to examine the temporal relationship between the two trends, especially during the period around the beginning of fertility decline. Tracing long-term demographic trends across sub-Saharan Africa is difficult. Although the availability of data has increased significantly since the 1980s, the situation was previously fragmentary. The quality of data and the comparability between sources are additional obstacles to obtaining consistent series. One usual solution is to limit the analysis to a single source (for instance, using retrospective data from one survey or several surveys from the same program, such as the DHS). By contrast, the approach used here seeks to take into account all available national censuses and surveys since 1950. The objective is to extend as far as possible the time span considered and to increase the robustness of the data by taking advantage of cross-validation between sources. The cost, however, is that the statistical series are disconnected from other kinds of indicators. For instance, while indicators on nuptiality and fertility from the DHS could be linked with indicators on contraception, education, etc. (since they are computed from the same databases), this is not possible with the present series because they are derived from different sources and further harmonized. For fertility I use TFR series from the UN World Population Prospects (WPP), which are provided in 5-year periods since 1950 (UN 2015a). These series, which have been constructed by taking into account multiple sources and varying methods of estimation (Alkema et al. 2011, 2012), are certainly the most reliable data on African fertility. Unlike in the case of fertility, there are no ready-to-use harmonized data series on nuptiality, and a specific database was constructed. The indicator used is the median age at first union for women.4 Both series are available at the national level only. Most of the analysis focuses on continental sub-Saharan Africa and countries with at least 1 million inhabitants in 2010, a total of 39 countries. To examine trends in age at marriage, I use statistical tables on marital status by sex and age from INED's pan-African database on nuptiality (Hertrich 2007; Hertrich and Lardoux 2014) (see Appendix5). For the 39 countries considered here, the database includes 362 national censuses and surveys carried out since 1950—9.3 per country on average. These data are extensive enough to trace long-term trends in age at marriage since at least the 1970s for 31 countries, and since the 1960s for 24 countries. For 31 countries, the trends can be followed up to at least 2010; for 3 countries the data end between 2001 and 2005. Period estimates of age at first union were calculated from these cross-sectional data on marital status by sex and age using the approach proposed by Hajnal (Hajnal 1953; United Nations 1984). The series of proportions of never-married individuals by age can be equated with that of a theoretical cohort and summarized by a standard indicator such as mean age or median age at first marriage. I use the median age at first union rather than the singulate mean age at marriage (SMAM), which is difficult to interpret when nuptiality is changing. In sub-Saharan Africa (with the exception of Southern Africa), where marriage is nearly universal, occurs at young ages for women, and is concentrated within a narrow age range, the median age at first union captures the current pattern of nuptiality, which is that of the young cohorts (aged 15–24 years) reaching the age at marriage at the time of the survey. Precise information on age at first marriage is difficult to obtain because people in many African countries do not have good knowledge of ages and dates, and also because African marriage is often a process (rather than an event) involving various ceremonies and stages, and this leads to varying interpretations of the timing of entry into union (van de Walle 1968; Meekers 1992; Hertrich and Locoh 1999; Antoine et al. 2009; Hertrich 2013). The issue is especially important when using retrospective data and this leads to a preference for cross-sectional indicators (Lesthaeghe 1989; van de Walle 1968, 1993). Yet, errors may also arise with period data—for instance, concerning the marital status of women who have uncertain or transitional marital status. There may also be errors of age reporting, depending on women's marital status (with age transfer toward younger ages for never-married women, and toward later ages for married women) (Pullum 2006). Such distortions may be further exacerbated by the design of the survey or census (criteria of eligibility, status of the respondent, more inclusive approach to conjugal union by surveys as compared to censuses, lower coverage of unmarried women by surveys, etc.). Systematic evaluation shows that estimates of median age at first union in sub-Saharan Africa tend to be underestimated by individual surveys when compared to census data (Hertrich and Lardoux 2014). To take into account these discrepancies between census and survey estimates, trends in median age at marriage were adjusted for each country to obtain harmonized series (see Appendix). Annual series were computed by linear interpolation between the existing points (census or survey dates). For the maps related to past periods with missing data, approximate estimates were chosen, when possible, in order to draw an overall picture. The WPP data on fertility are given by five-year period and the annual series of fertility were computed by linear interpolation between the existing points. TFRs were rounded to one decimal.6 A few countries have surprising or irregular trends in nuptiality. This is especially the case for the Central African Republic and Somalia7 and, to a lesser extent. Swaziland and Zimbabwe. For completeness of coverage, I show the trends in these countries but I consider them to be outliers. Because I aim to document how changes in age at first union relate to the onset of fertility transition, the definition of the onset of fertility transition is critical. Two approaches coexist. One is to consider that fertility transition has begun when consistent, continuous, and irreversible fertility decline is observed; a common rule (Bongaarts and Casterline 2013; Casterline 2001) is to date the onset of fertility at the year the TFR reached a level 10 percent below its maximum value. The other approach (Alkema et al. 2011) uses the point of maximum fertility prior to sustained fertility decline as the onset of fertility transition. Here I consider that these two dates delimit the period when fertility decline begins. The date of maximum TFR could be regarded as the “eve” (the point of initiation or early beginning) of fertility decline. I reserve the term “onset” for the date when the TFR is 10 percent below the eve point. As fertility decline is slow in many African countries, the time lag between the two points is usually substantial, 14 years on average (see Table 2 and endnote 6). If one considers that fertility transition is confirmed when TFR is 10 percent lower than the historical maximum, then Niger is the only country still in a pretransitional situation. The pioneering comparative work on trends in African nuptiality, based on a systematic consideration of censuses and surveys, was carried out in the early 1980s by Lesthaeghe and colleagues (1989), then updated by van de Walle (1993). It showed an increase in women's age at first union but remained cautious on the future of such trends. Over the last 20 years, a growing number of studies (though less extensive and often limited to retrospective data) have provided additional evidence on the increase in women's age at marriage (Garenne 2004; Hertrich 2007; Lloyd 2005; Mensch, Grant, and Blanc 2006; Mensch, Singh, and Casterline 2005; Ortega 2014; Shapiro and Gebreselassie 2014; Tabutin and Schoumaker 2004; Westoff 2003). My data confirm the substantial change in first marriage patterns across the continent. The pattern of early female marriage,8 which was a pillar of the sub-Saharan nuptiality system, has been substantially altered. Figures 1 and 2 show how the trend toward later age at marriage spread to the entire continent during the last five decades. In the 1960s, an early marriage pattern was clearly dominant. The median age was below 18 in most countries, with the main exception of Southern Africa, where late marriage was already the rule, and to a lesser extent some neighboring countries from Middle and Eastern Africa. In the greater part of the continent during the 1970s, the age at first marriage for women began to increase significantly. By 1980, only a minority of countries (mainly in Western Africa) still had a median age at first marriage under 18. Over the following decades, the increase spread to much of Western Africa, and the pattern was confirmed in other regions. In the mid-1990s, the early marriage model that had been dominant 50 years before had practically disappeared. The one exception by 2010 was Niger.9 The main standard now is a median age over 19.5 years at the beginning of conjugal life, and the figure exceeds 21 years in a large number of countries. Trends in women's age at first union and total fertility, 1965–2010 SOURCES: UN 2015a for fertility; author's database on African nuptiality for median age at first union. Long-term trends in women's median age at first union, by country SOURCES: Author's database on African nuptiality. The delay in women's first union has been a systematic trend (Figure 2). While the assumption of a past decline in age at first marriage in selected countries had arisen from retrospective data (Garenne 2014), this is not supported by the cross-sectional estimates. Indeed, the only cases in which data suggest a decrease (CAR, Swaziland) or stability (Somalia, Zimbabwe) are those with problematic data. In terms of geographical differences Southern Africa stands out, both because late marriage was already common 50 years ago and because age at marriage continued to increase rapidly in most countries. Except for Lesotho, median age at first marriage in most countries now exceeds 25 years and in some it approaches 30 years. The nuptiality pattern in Southern Africa represents a special combination of late marriage for both sexes, significant numbers of people who never married, small gender difference in ages at marriage, frequent marital dissolution, low remarriage, and low levels of polygyny (Timæus and Graham 1989). The development of this marriage system has been attributed in large part to widespread labor migration (especially in mines), which institutionalized circular migration and spousal separation and more broadly affected domestic arrangements and relations between sexes and generations. Apartheid policies, women's education and labor force participation, and high bride prices are cited as additional factors (Timæus and Graham 1989; Hosegood, McGrath, and Moultrie 2009; Moore and Govender 2013). Marriage is no longer considered to be the normative context for sexuality and fertility. Informal unions and childbearing outside marriage are common. Only in North Africa are levels and trends in age at first union similar to those in Southern Africa (not included in this study; see Ouadah and Vallin 2000, 2013). While a median age at marriage for women above age 25 has never been observed in other continental sub-Saharan countries, there are cases of sharp increases in age at first marriage in each region: Gabon and Cameroon in Middle Africa, Rwanda in Eastern Africa, and most countries in the Gulf of Guinea in Western Africa (Figure 2). For Eastern and Middle Africa, the general picture is that of a slow but regular increase in age at marriage from the 1970s to the 1990s; the increase has apparently stalled since then, with a median age of around 20–22 years in most countries. Considered to be the region with the most traditional features of African nuptiality, Western Africa, however, does not deviate from the general pattern of change. Indeed the postponement of women's first marriage is occurring in all these countries, but with differences in timing and pace of change. Three sub-regions can be distinguished in Figure 2: the Gulf of Guinea (Nigeria to Liberia) with pronounced and long-term changes since the 1960s; the Western Coast (Sierra Leone to Mauritania), where the increase in age at marriage began later, usually in the 1980s; and the Sahelian countries (Niger to Guinea), with slower but continuous trends. Differences among countries in Western Africa clearly increased during the last 50 years. In the 1960s and 1970s, all countries in the region shared a similar pattern of early marriage (16–18 years), while the range of variations today is larger (17–23 years). The comparison of cross-sectional indicators provides a first insight into the relationship between fertility and patterns of age at marriage. As expected for the recent period, there is a negative correlation between them (Figure 3): the higher the median age at first marriage in the country, the lower the TFR. As shown on the map (Figure 1), the contrast is particularly marked between Southern Africa (where marriage is especially late and fertility low) and the remote Sahelian countries (where fertility remains over 6 children per woman and women marry at earlier ages). A less expected result is that the correlation at the country level (although statistically significant) was weak in the past. Until the late 1980s, less than one third of the variance was captured by the correlation (Figure 3). In other words, the differences in fertility levels during the pretransitional period did not overlap with the differences in ages at marriage. However, looking at when the fertility transition began in the 1990s with growing diversity between countries, the correlation becomes stronger, as if nuptiality matters when things begin to change. In this scenario, nuptiality seems to be the forerunner, as its pattern correlates more closely with the fertility level 10 years later than with the level of the same year. To further investigate the relationship and temporal shift between changes in nuptiality and fertility, I will proceed in two steps: first, by examining nuptiality in the period of early fertility decline; second, by expanding the time horizon to include the years preceding the transition. Fertility decline in most sub-Saharan countries began slowly, taking an average of 14 years for a 10 percent decrease in TFR, with regional means ranging from 12.5 years (Southern Africa) to more than 15 years (Western Africa) (see Table 2). The period when fertility decline begins is characterized by changes in nuptiality, specifically in the collapse of early marriage patterns in the countries where it was still common. On the eve of fertility transition (the historical point of maximum fertility prior to sustained decline), a median age at first marriage below 19 was still dominant in Eastern Africa (67 percent of the population) and particularly widespread in Western Africa (90 percent of the population). At the onset of fertility decline (when TFR is 10 percent below the maximum), this pattern became a minority in both regions (13 percent in Eastern Africa, 21 percent in Western Africa) and became even less common in Central and Southern Africa, where it was already rare. In most of sub-Saharan Africa (86 percent of the population), the median age at marriage exceeded 19 years at the onset of fertility transition. According to these data, an early marriage pattern seems incompatible with a sustained fertility decrease. Focusing on countries with consistent data, there is no empirical evidence of fertility transition in a context where the median age at marriage was below age 18 at the time of the onset. Between the eve and the onset of fertility decline, most countries (80 percent of the sub-Saharan population) experienced a postponement in women's age at first marriage. Exceptions exist only in Middle Africa, where the fertility transition began later in a context in which age at marriage was already close to 20 years. In contrast, the rise in age at marriage was systematic in Western Africa, where early marriage was the rule. But what is the picture before the fertility decline? For 25 countries, the time series on age at marriage begin at least 5 years before the initiation of fertility decline (i.e., maximum TFR), making it possible to examine the connections between nuptiality and fertility over a larger time horizon. Figure 4 depicts, for each of these countries, the trend in the