Social Dynamics of Adolescent Fertility in Sub-Saharan Africa (1993)

Marriage is the demographic event most often used to estimate the time when regular sexual relations begin. In "natural fertility" populations, age at marriage is often a reliable determinant of when childbearing begins and of the number of children a woman will bear. But, although it may hold elsewhere, this relationship is quite tenuous in sub-Saharan Africa, in both rural and urban settings (see also Gyepi-Garbrah, 1985a).¹ Although we have presented evidence that suggests that the age at marriage is rising, understanding the economic and social dynamics of entry into marriage is essential to the study of adolescent fertility.

A key reason why age at marriage and entry into childbearing are so weakly linked stems from problems in defining marriage in Africa. Besides the fact that marital practices are extremely diverse, changes in ways of legitimizing unions and the rise of de facto forms of polygyny may be rendering even more fluid an institution that has long been recognized more as a process than as a discrete event. These definitional conundrums make it tempting to discard marriage as a meaningful link to entry into childbearing

and to rely on variables such as age at first sexual activity or age at first cohabitation, as various surveys have done.

Yet we cannot dismiss the fact that the principal transformation in adolescent fertility that the new African surveys are identifying is not a change in fertility per se, but a rise in rates of childbearing among never-married women. This trend is our most promising entry into several issues that vitally concern young women: the use of contraception, abortion, and the prospects for continuing school after a pregnancy. Because many of the problems of adolescent fertility appear to rise from condemnation of what is seen as premarital childbearing, the question of what marriage is and when it begins cannot be laid aside.

This chapter shows that the ambiguities in marriage leave considerable room for disputing marital status. When young women have few alternatives to immediate marriage and childbearing, becoming pregnant at the outset of a union, whether or not it is called a marriage, is likely to be highly welcome. But nowadays, when a young woman has other opportunities for training or when her partner appears ambivalent about the union, the resulting changes in the perception of marriage cast a more dubious light on a woman's pregnancy.

Despite extraordinary diversity across the continent, almost all African women marry, and they remarry quickly after divorce or widowhood (Smith et al., 1984). Women usually marry when they are comparatively young, from 16 to 18 years of age, and most women marry men substantially older than themselves. Besides these characteristics, dominant models of indigenous African marriage stress several key components. Three are noted below.

Because most African societies value women's productive and reproductive potentials, men and their families are expected to pay money or goods to the woman's family. These "bridewealth" payments also confirm the legitimacy of a union and its progeny (Radcliffe-Brown, 1950). The nature and quantity of bridewealth payments vary enormously. In general, richer families are expected to pay more. But a woman's family might forgo bridewealth altogether in exchange for a set of wealthy, powerful in-laws. Alternatively, they may prefer a man of humble origins who cannot provide tangible wealth but would be willing to work for them for several years, usually at periods of peak labor demand.

If we can visualize African marriage as a logical extension of elders' investment in the young, then it should be no surprise that the marriages of youth often reflect the political and economic agendas of their elders. Among the Bangangte of Cameroon the marriage rights that older men hold in their daughters (and sometimes their nieces) are key elements of political power; they give men the means to forge alliances with the families of prospective husbands and thereby to create ties to both patrons and clients (Feldman-Savelsberg, 1989). Because marriage represents a union between kin groups, the children born to it belong as much to their kin groups as to their parents. The legitimate affiliation of children to a family—their father's in patrilineal societies, or their mother's in matrilineal societies—is a critically important step toward the perpetual renewal of such alliances.

Because marriage is as much a link between kin groups as it is a union of a husband and wife, families mark the union ritually. Along the coast of western Africa, families offer drinks and kola nuts as a significant celebratory feature of customary marriage a symbol of recognition by the families of the union. Examples come from the Akan of Ghana (Oppong, 1974), the Abutia Ewe of Togo (Verdon, 1983), the Anlo Ewe of Togo (Nukunya, 1969), the Creoles of Sierra Leone (Harrell-Bond, 1975) and the Ibo of Nigeria (Meek, 1937, Uchendu, 1965).

Given the political potentials of marriage, family elders are anxious to control when youth marry and whom they marry. Especially in the past, adolescent women were expected to acquiesce to marriages to virtual strangers that had been arranged in their childhood. Some alliances were even arranged before a child's birth. If a female was born, she would become the wife of the man with whom an alliance was sought; if a male was born, he would become a lifelong client. Families are still heavily involved, particularly in first marriages (see, for example, Landberg, 1986 for the Kigombe of Tanzania). Some young women are under such strict control that they have few areas of choice. For them, adolescence can be a very long stage in life, often one with no clear end. Paradoxically, only the poorest adolescents in Africa can be considered full decision makers with respect to their marriages: Only those bereft of kin must make their own decisions in adolescence.

Because kin groups see the marriages of the young as a way to create links to other groups, they have vested interests in conjugal stability. Like many other groups, Akan families in Ghana may call meetings in an effort to resolve serious differences between spouses (Oppong, 1980). In eastern Africa, according to Klima (1970), an unhappy Barabaig wife who sought refuge with her family might be returned forcibly to her husband.

One of the most distinctive features of marriage in Africa is polygyny, a phenomenon that has considerable bearing on marriage for adolescents. In many countries a rapid rate of population growth and a wide age gap between spouses create a large pool of women to marry a smaller number of older men, whose ranks have been thinned by mortality. Polygyny is an extremely heterogeneous institution. A farmer may have two wives; a chief, many. Within polygynous households, wives are differentiated as well. Temporal precedence is crucial: The senior wife handles the household's administrative and fiscal tasks, and delegates burdensome chores to junior wives, each of whom ranks above any succeeding ones. The senior wife is often in charge of monitoring the actions of the younger wives around other men, watching for signs of deviation from marital loyalty. Three other categories of wives are important: The first is the ''official wife,'' required by an eminent urban man who must appear in official contexts with only one wife; she is generally the most educated, attractive, and cultured among his wives. Second, is the "beloved wife," who may receive better clothes, more money, and a separate house, and who may be seen more often in public with the husband. And third are the "outside" wives, who are not officially married to the man, may live elsewhere, and have more limited claims to his resources.

Political stature for a man often reflects the number of his wives and children, and a man who manages to marry more than one wife gains the possibility of numerous children. He also gains links offering political alliance and economic support with a wide range of families. Co-wives, if they are on good terms, share domestic chores and child care. Given a choice, however, most young women prefer monogamous marriage to gain greater leverage with their husbands and to liberate themselves from the work demands of domineering senior wives. Not surprisingly, as they grow older and gain junior wives of their own, their opinions of polygyny often improve.

Polygyny is more common in western Africa than in eastern or southern Africa, although about 15 to 30 percent of married men are polygynists in Kenya and Tanzania (Lesthaeghe et al., 1989). An estimated 50 percent of the Ashanti of Ghana practiced polygyny in 1945 (Fortes, 1954), and the proportion was estimated at 46 percent in the city of Ibadan in 1973 (Changing African Family Project, cited in Caldwell et al., 1989). Contrary to predictions that polygyny would inevitably disappear (see, for example, Goode, 1963), no overwhelming evidence points to its decline in the subcontinent

despite the welter of legal and religious codes that colonial as well as modern states have created to curb it (Pison, 1986; Lesthaeghe et al., 1989).²

Table 3-1 presents recent Demographic and Health Survey (DHS) data on the prevalence of polygyny among young married women ages 15 to 24. Western African countries show comparatively high levels. In Liberia, Mali, Nigeria, Senegal, and Togo over 30 percent of currently married women in this age range are in polygynous unions. Young women with no education are much more likely to be married to polygynists than are those with secondary or higher education. The largest educational differences are found in Kenya, where 2.4 times more uneducated women are in polygynous unions than those with secondary or higher education.

Urban areas almost always report lower rates of polygyny than rural areas (Gaisie, 1975; Brown, 1981; Pison, 1986; Lesthaeghe et al., 1989), a differential usually explained by declining economic returns from wife-children units in urban areas (Caldwell, 1969). Certainly for adolescents, polygyny is lower among urban women, as Table 3-2 shows. Mali shows

the largest difference: 36 percent of young rural women are in polygynous marriages, compared with 24 percent of urban ones.

Polygyny affects male adolescents most directly through the age at first marriage. In societies where polygyny is widespread, few young men aged 15 to 24 have more than one wife. Because polygyny has a strong accumulative dynamic, young men must compete with powerful senior men for their age mates as wives (Pison, 1986).

For women, the situation is quite different. In highly polygynous societies young women are the targets of intense competition and are likely to be courted almost as soon as they reach puberty. The practice of polygyny means that in theory any woman of reproductive age can marry, regardless of the paucity of marriageable men. Though most women are in monogamous unions at any one point in time, at some point in their lives many of their husbands will have other wives as well (Pison, 1986; see also Goldman and Pebley, 1989, on the formal demographic underpinnings of polygyny). Although polygyny is not necessarily associated with a low age at first marriage for women, it is associated with greater age gaps between husbands and wives (Pison, 1988; Goldman and Pebley, 1989; Lesthaeghe et al., 1989).

Lesthaeghe et al. (1989) argue that kinship organization plays a role in determining levels of polygyny. Matriliny, for example, seems to be asso-

ciated with low levels of polygyny because families retain rights in their daughters, and men prefer not to have multiple groups of in-laws with whom they must deal. Because matriliny makes polygyny less desirable, matrilineal societies may also have a narrower age gap between spouses or an older age at marriage for women, or both. However, DHS results (not shown) suggest that once the age at marriage begins to rise, whether a group is matrilineal or patrilineal seems to make little difference in early childbearing.

Marriage is generally considered the best indicator of exposure to the risk of childbearing. But in Africa, the tenuous link between these two events is partially attributable to measurement biases. Enumerators are more likely to code women who report that they are married as being older than those who report that they are unmarried (Lesthaeghe et al., 1989). Two other important factors make it difficult to analyze the relationship between age at first marriage and age at first birth: enormous variations in marital forms, and ambiguities in measuring the timing of marriage. That is not to imply that the rules defining marriage are themselves ambiguous. The legal aspects of the marriage process may be quite straightforward, and the expectations and the appropriate behavior of each party clearly defined. Rather, the measurement problem is that the multiple forms of marriage that exist in Africa, and the processual nature of marriage, make it difficult for researchers or even Africans to categorize people as being married or unmarried. Highlighting some social ambiguities surrounding the definition of marriage in Togo, Locoh (1988) argues that we should be cautious in using the term "marriage" as a demographic concept:

Many African societies recognize a range of union types: those endorsed by legal/statutory ceremonies, consensual undertakings, and various religious or customary procedures. So, do observed variations in marriage rates have empirical bases, or do they simply reflect subgroup differences in how marriage is defined? "Village wives," secondary marriages, and woman-to-woman marriages, though extreme examples, illustrate the problem. Among the Lele of Zaire, a "village wife" was considered married to all the men of a particular age in the village. This was not a low status position and the children of the village wife were considered legitimate even though their biological fathers might never be known (Douglas, 1963). And, among the Higi and Bulai of northern Nigeria, and some groups of the Jos Plateau region of Nigeria, a woman maintained several marriages simultaneously, each of which was socially regarded as a legal marriage (Meek, 1931; Chalifoux, 1990; Sangree, 1980). And in Sudan a childless Nuer woman could assume a male role. She could make bridewealth payments for a young woman who then bore children on her behalf (Gough, 1971); the father of the children had no legal rights over them or their mother.

It is important to recognize that while these types of marriages are good illustrations of the flexibility of African marriage customs, either they represent minor themes in the marriage system of most African societies, or they are practiced by very small groups. However, faced with this multiplicity of marriage customs, many surveys in Africa have relied on the marital status women report, recognizing that different definitions of marriage can produce enormous variations in reports of age at first marriage.

To avoid definitional entanglements, most Demographic and Health Surveys and World Fertility Surveys (WFS) focused simply on the date when cohabitation began. (The exceptions were the Sudan Fertility Survey and the Senegal Demographic and Health Survey, which used the date of consummation.) The most obvious disadvantage of this strategy, of course, is the elimination of married couples who do not cohabit. Coresidence is often only one event among many in a long process of conjugal negotiations that may last for several years. Using it as the defining criterion of marriage poses special problems in areas with high rates of migration and levels of urbanization. In contemporary Lomé (Togo), Locoh (1988) observes, about 25 percent of currently married women do not live with their husbands, a pattern emerging especially among younger women whose husbands are polygamous. Etienne (1986) reports of the matrilineal Baulé of Côte d'Ivoire that many women maintain visiting relationships with their husbands and may not establish permanent coresidence until several children are born (see also Fortes, 1950, for the Ashanti; for patrilineal groups, see Salamone, 1986, for the Dukawa of Nigeria, and Comaroff and Roberts, 1981, for the Tswana of South Africa).

Even the cautious designation of marriage as a stable union involving cohabitation would make some unions that were reported as marriages in the surveys fail local tests of legitimacy such as celebration, kin recognition, or church or civil sanction. On the other hand, unions that the surveys did not report as "marriages" may have been quite legitimate in local views.

Because one definition of marriage cannot cover all circumstances, the choice of the starting point of marriage for purposes of comparison depends on the goal of the analysis. For the DHS and WFS, which ultimately sought to estimate fertility, using date of cohabitation may have been the best choice for a wide age range of women. For studying adolescents, however, the group in most conjugal flux, choosing any one measure such as cohabitation to define the start of marriage inevitably brushes over considerable variation.

The wide array of union types makes it difficult to link fertility to the onset of marriage, a problem compounded by the processual nature of marriage in many African societies. Almost without exception, the WFS and DHS treated the beginning of marriage as a discrete event, and hence a person as either married or unmarried, a strategy that has obvious measurement advantages. Yet the boundaries between the two may be blurred in the many African societies that treat marriage as a process that evolves over months or even years through a sequence of events. These events may include the exchange of symbolic tokens, making installments on bridewealth payments, establishing a joint residence, or even the birth of a child (see, for example, Karanja, 1987, on the Yoruba of Nigeria; Harrell-Bond, 1975, on the Creoles of Sierra Leone). These events can best be seen as "conjugal testing," during which the partners and their families build mutual confidence by taking incremental steps toward a union. For young people, and young men in particular, access to authorized procreation, like access to other prerogatives of adulthood is achieved only gradually through demonstrating worthiness.

The processual nature of marriage poses obvious difficulties for distinguishing the end of an informal union from the beginning of a formal marriage (see Ngondo a Pitshandenge, 1988, on the difficulties this phenomenon entails for legal authorities). Identifying a general sequence of events that, once completed, confirms a legitimate marriage and establishes a definitive marriage date is one strategy for coping with this problem. The Côte d'Ivoire survey of 1980–1981 under the auspices of the WFS lends itself to this purpose. It identified three critical stages in the marital process: cohabitation, consummation, and celebration. Analyzing data from this survey, van de Walle and Meekers (1988) and Meekers (1992) found

the most common type of union to be a compressed sequence in which cohabitation, ceremony, and sexual relations with the first husband began within a month of each other, though ceremonies often came first. This type of marriage accounts for about 46 percent of all first marriages. In such marriages women's ages were quite low—from 13 to 14 years. Many of these marriages occurred among the Mandé Nord and African immigrant groups, both largely Islamic and patrilineal. By contrast, matrilineal groups, such as the Baulé and other Akan, and urbanized and literate women manifested another kind of sequence: "free unions," in Meekers' phrase, in which ceremonies followed sexual relations and cohabitation typically after a very long interval. Such marriages accounted for 22.2 percent of first unions. (Because surveys typically look at only one slice in time, it is difficult to appreciate that consensual unions often become stages toward more formalized ones. Using these same data, Brandon, 1990, found that 65 percent of all consensual unions are terminated or transformed by the fifth year; 42 percent become formal unions, and only 23 percent are dissolved.)

As these analyses suggest, it is difficult to find one reliable reference point from which to date all marriages. Different groups may attach quite different meanings to any one element of the process. Among the Ga of Accra, celebratory beverages may confirm the start of negotiations for a high-status marriage, the beginning of cohabitation, or the paternity for future children (Robertson, 1984.) As early as 1795, Afzelius (1967) reported that the Susoos of Sierra Leone used such drinks to cement the betrothal of unborn children. Further, the definition of marriage may change over time: Older cohorts may emphasize socially defined unions, whereas younger ones emphasize legal codes or religious ceremonies as criteria of legitimacy.

Regardless of the criteria used, a great deal is at stake in marriage—property, children, labor, and political alliances—and disputes frequently erupt over whether the conjugal sequence has actually been completed. If relations deteriorate between two families that are linked by a marriage, then heated arguments may ensue over whether the couple is really married. The money that the man's family previously handed over as bridewealth may suddenly be construed as payments on a past loan that need not be returned; and the food or drinks that were consumed together are now recalled not as marriage preliminaries but as casual tokens of hospitality. Finally, because the process of building a marriage can be so drawn out, the relationship may actually end before the completion of bridewealth payments or bride service. Disputes may arise afresh far into the future, especially when inheritance time arrives, about whether the marriage really occurred. (See Comaroff, 1980, and Comaroff and Roberts, 1977, for discussions of such complexities.)

The measurement difficulties that such ex post facto redefinitions pose

cannot be minimized. Awusabo-Asare (1988) found that in Ghana, unions that are dissolved after marriage payments are finalized are more likely to be reported retrospectively as marriages than those for which payments were not completed. And people are more likely to report unions that end as marriages than those that do not last (van de Walle, 1993). For a number of reasons, then, using fixed criteria to measure change in marriage, whether from retrospective or time-series data, is inherently fraught with problems.

These difficulties in identifying and measuring marriage have obvious bearing on how births are perceived and counted. For analyses that assume that only one event in the process definitively determines a marriage, the proportion of births that might be defined as "premarital" varies according to which indicator of marriage is used. Analyses of the Côte d'Ivoire WFS reveal that 21 percent of first births would be defined as premarital if the date of celebration, that is, the exchange of drinks and kola nuts, is used to indicate the first union (van de Walle and Meekers, 1988; Meekers, 1992). These results are reported in Table 3-3. If the date of consummation, that is, sex with the man who eventually becomes the first husband, is used, then about 10 percent of all first births would be premarital; by using cohabitation, 17 percent would be premarital. The data from this survey show considerable variation even within Côte d'Ivoire. All three of the events among the Mandé Nord seem to give similar results: 11 percent of first

births would be premarital by using either cohabitation or celebration as indicators, and 8 percent by using consummation. Among the Baulé people, on the other hand, 17 percent of first births occurred before consummation of first marriage, 47 percent occurred before cohabitation, and 50 percent before celebration.

In the end it may be a mistake to force into a specific order all the events that collectively form a marriage and then single out one that marks the beginning. For a man, "legitimate" access to a partner and to reproduction depends less on rigidly following the prescribed cultural steps than by proving himself worthy—for example, by providing labor, goods, and political support to his family, which must be persuaded to provide bridewealth, and to potential in-laws, who seek a dutiful son-in-law. Because mutual testing between partners and their families is the basic agenda of the marriage process, the exchange of drinks, say, or the birth of a child may be the symbolic event that nudges the union further toward what the families eventually are willing to call a marriage; but it does not necessarily finalize the marriage. Instead, the meaning that a particular event takes on within the sequence depends largely on how the spouses and their families are getting along at the time.

If we accept this more fluid view of marriage, two critical points emerge. First, among groups that stretch out the marriage process, it is difficult to argue that marriage affects entry into childbearing rather than the reverse: It may be equally true that unions that are fertile are more likely to endure and thus end up being called marriages.

Second, that marriage is fluid in nature suggests that a "premarital" birth—one that precedes the conclusion of the marriage process—is by no means the same thing as an illegitimate birth. No one can deny the implications of legitimacy; but legitimacy refers less to a birth after marriage than to a man's willingness to assume the social and economic responsibilities of fatherhood. Hence if, in a patrilineal society, children are born to an unmarried woman, the man can usually claim them by compensating her family. In most cases, men are anxious to claim their children, and to maintain ties with them even if the union itself does not last. In Botswana, in the early 1970s, as many as one-third of households were headed by women (Peters, 1983:106); subsequent micro studies suggested that "a Tswana 'unmarried' woman is not necessarily out of touch with her children's father's family ... not only with respect to material support but also to social recognition" (Peters, 1983; see also Kerven, 1984, and Peters, 1984). Even children whom a father temporarily rejects can be legitimized later, or they can be formally sponsored by someone other than the father.

Yet just as a premarital child is not illegitimate, marriage alone does not ensure a child's legitimacy. A married woman's child by an outside lover may have difficulty gaining paternal recognition. While the lover

may deny paternity to avoid adultery charges, the husband, if he is angry about the mother's indiscretion, may refuse to accept the child. Affiliation with the mother's family is possible, though rights as a full member under such conditions are uncertain.

If marriage is neither a necessary nor a sufficient condition of paternal legitimacy, then a premarital birth—if that is how we interpret a birth that occurs before the completion of the marriage process—is not necessarily an embarrassing case of illegitimacy. If the entire conjugal process is proceeding smoothly, one event leads to, and legitimizes, the next. Hence a birth to a young woman that occurs before her union is fully recognized may be a welcome event in the continuum toward a legitimate marriage. Only if the union is dissolved—or if the man dies before it is fully concluded, and inheritance is suddenly at stake—will questions of "legitimacy" likely arise.

In sum, although we might assume that one of the main functions of marriage in Africa is to confer recognized paternity on a child, the mother's marital status at the time of her delivery is at best a rough proxy for the child's standing in the family; still, it is the best we have. The rest of the chapter expands on these observations as we turn to consider recent changes in marriage.

Vast changes in legal codes, educational opportunities, and avenues for employment, as well as the advent of new theologies, have not simplified the marriage process in Africa. Indeed, they have added elements that make it even more complex. In many countries, new forms of marriage are appearing, and we may be witnessing an even greater attenuation of the conjugal process. Although kin participation in youths' marriages seems to be ebbing, conjugal payments and polygyny seem to be taking subtle new forms. All these trends render even more difficult the analyst's task of determining the age at first marriage and its effect on reproduction. To document the outcomes, we examine several commonly discussed domains of change: economics, legislation, religion, education, and urbanization

In all societies, how to acquire the material resources to marry and support a family is a question of wide concern. The specific issue that concerns us here is how economic well-being affects the age at marriage in Africa. A key area of debate has been property dynamics: the relationships between land supply and population density. By far the most influential work on this topic is that of Boserup (1965, 1970, 1990). Until recently, Boserup contends, low population densities, male out-migration, and sub-

sistence agricultural methods were conducive to large family size in most of Africa (Boserup, 1990). Changes in these elements—constriction of the land base, combined with intensified requirements for permanent farming—eventually encourage people to have smaller families, through reducing polygynous marriage or the number of births per wife, so that they may pass on a viable farm to the next generation. The implication, of course, is that age at first marriage increases as well.

No one now disputes Boserup's main contention that population density and agricultural change measures interact in this general direction. But the mechanisms, strength of influence, time frame of change, and sequence of steps in the unfolding dynamic are all understudied for the African case. Lesthaeghe (1989a) argues that unlike the case of preindustrial Europe, the age of marriage in Africa was too low overall to function as a culturally recognized regulator of fertility. The vast majority of African agricultural societies had quite low population densities, so that there was no shortage of arable land and therefore people did not need to establish the appropriate age at first birth according to basic economic variables. Using WFS data from the late 1970s, Lesthaeghe et al. (1989) found that when literacy and polygyny were controlled, population pressure failed to produce significant effects on age at marriage.

DHS findings from Ondo State, Nigeria, and from western Kenya, two of the areas of highest population density in the subcontinent, suggest that the implications of population pressure for age at first marriage are far from clear. Even where we find an association between marriage age and population density, the causation may be unclear. Southern Nigerian societies have high ages at marriage; indeed, the women in the Yoruba population of Ondo State probably have the highest age of marriage for any African population in which childbearing typically begins after marriage: 20.2 years for women now 25 to 29 years old (Medical/Preventive Health Division, Ondo State and Institute for Resource Development, 1989). Yet we cannot rule out the effects of other factors on this high age at marriage. Southern Nigeria has very high rates of schooling for women and many opportunities for female occupational training and specialization in trade and the informal sector. If population density is a factor for Nigeria, therefore, it probably works through these occupational and education structures, rather than through land pressure per se. In western Kenya, DHS results for the densely populated Nyanza and Western provinces actually indicate lower median ages at first marriage and first birth for women ages 20–49 than are true for the rest of the country (Kenyan National Council for Population and Development and Institute for Resource Development, 1989). The social and cultural concomitants of the western Kenyan pattern are well known (for example, Hakansson, 1988): high rates of male out-migration for work on the railroads, high rates of polygyny in urban areas, maintenance of both urban and

rural residences, and rotation of wives and children through the various family homes and enterprises.

In sum, there is little support for the view that high population density in agricultural societies affects the age at marriage or childbearing for women in the way that Boserup's theory predicted, at least within the time frames for which we have evidence in contemporary Africa. Raising the age of marriage is only one method people can use to control the number of heirs to property. Even where increasing population density in agricultural societies does seem to be associated with a rising age of women at marriage or with lower fertility, the effect may be indirect: through an increase in occupational specialization for women that requires longer training or through a change in needs for youthful farm labor. Similarly, rather than curtailing fertility or resorting to celibacy or infanticide, which would drastically restrict their options if not breach cultural presumptions, people may try to thin out competing claims on property by drawing on methods of family control that are already in the cultural repertoire. Such methods may include marginalizing some women by declaring them undesirable to marry or defining specific children as ineligible for inheritance.

Apart from population density and the land base, the economic issue that looms large for age at marriage in Africa is the material resources that allow a young man to marry in a socially acceptable way. Young wage earners nowadays who obtain access to resources independently of family elders gain a significant advantage in choosing their own partners (Robertson, 1984). They also, of course, gain more resources to form temporary relationships with women who are not their wives.

Although many segments of society have abandoned bridewealth, strictly defined, other demands have sustained the high costs of marrying for a young man. Those by the woman's family usually include future help with food costs, court fees, medical treatment, and younger children's school fees. New expectations of financial support arise as fast as the economy changes. In one contemporary variant of child betrothal, a wealthy older man may be asked to educate a young girl who has been designated as his future wife (for southern Ghana, see Vellenga, 1986).

Although more economic freedom from elders gives young men more independent access to conjugal or sexual partners, many are expected to shoulder most, if not all, of their marriage expenses alone. As a result, mustering the resources to marry may take more time than in the past (Isiugo-Abanihe, 1988). Downturns in personal fortune, commodity prices, or employment may prolong the process (see, for example, Hollos and Leis, 1989.)

An extreme case of rising marriage costs is that of the Igbo in Nigeria. Isiugo-Abanihe (1988) reports that demands for exorbitant amounts of bridewealth mushroomed with the oil boom of the 1970s. Grooms were expected to buy televisions, radios, refrigerators, motorcycles, clothes, and trinkets for their

future in-laws. Some even had to buy a car or land or to build a house for their fathers-in-law. These incidental expenses represented steep financial outlays, often amounting to much more, even after accounting for inflation, than previous bridewealth amounts. Consequently, the age at marriage for both men and women has increased rapidly. For women who married in 1967 or before, 17 years was the mean age at marriage; among those who married after 1970, 24 years was the mean. Among men, the means were 26 years before and 32 years after 1970. Even among the uneducated, ages at marriage have increased substantially.

A factor related to the new economic barriers to marriage may be the monetization of bridewealth (see, for example, Ngubane, 1989; Comaroff and Comaroff, 1990). According to Nagashima (1989:190), among the Iteso in Kenya the transition from cattle to cash as bridewealth was instrumental in prolonging the marriage process, transforming it ''from the traditional immediate payment to lengthy, piecemeal payments.'' To be sure, whether the marriage process is lengthened or shortened by changes in the financing may be related to social class. As cash began to replace cattle as bridewealth among the Giriama in southeastern Kenya, wealthy parents began to shorten the time of payment because they wanted cash as investment capital to buy land and plant tree crops (Parkin, 1972). Parkin's findings suggest that when bridewealth is high, poorer people may take longer to pay it, both because the payer is too poor to acquire the extra amount and because the receiver is more anxious to secure income than capital.

In considering the effects of economic forces on marriage, the case of labor migration in southern Africa looms large, whether we view it as a response to shrinking land supply or to scarce bridewealth resources. Gulbrandsen (1986) and Murray (1980, 1981), among many others, argue that the combination of abandonment of polygyny and circular labor migration to South Africa has profoundly transformed the entire regional system of marriage and family relations. Because of the long absences of men, many relationships never culminate in a marriage, or do so very late, and thus lead to a large number of births that we would count as nonmarital. In their prime working years, young men typically migrate for work most of the year. During their brief visits home, they strike up temporary relationships with women who are left behind to maintain the farms. Because a young man must still pay bridewealth to marry legally, his parents, who are taking care of their daughters—and their daughters' children—may actually discourage him from marrying in order to retain greater control over his earnings (see also Izzard, 1985). The result is not surprising, as Gulbrandsen (1986:17) explains:

Upon their retirement in middle age, male migrants are anxious to marry and set up a home independently of their parents. They tend to marry women already in their thirties with several children. Some men are reluctant to support a predecessor's children, but others are happy to acquire children, who constitute potential resources.

As for women, those who marry and move into their husbands' compounds report tensions with in-laws, with whom they compete for their husbands' support. Unmarried women who bear children by usually absent partners may remain as dependent members of poor households that are maintained by parents and brothers (later by sons) or head their own small impoverished households. Needing to support their children, they may also begin to migrate periodically to the city to look for work, leaving their children with their mothers (Bryant, 1977).

Through this complicated picture of marriage and fertility in southern Africa, the fact that stands out is that for much of the region, the demands and opportunities created by massive labor migration are making childbearing—and even childbearing—largely independent of marriage. The resulting gap between physiological maturity and economic viability, especially for men, likely extends the period in which any children they father are likely to be born in "premarital" unions.

New legal codes have aimed to supplant the family with the state as the institution granting legal status to marriages (see, for example, Meekers, 1990, for Côte d'Ivoire). They have also sought to shift the definition of marriage from an alliance between two families to the basis of a new conjugal family. Such measures, according to predictions by authors such as Goode (1963:202), would "become forces in their own right, supporting the emergence of new family systems." They have also aimed to create greater scope for individuals to make their own partner choices. During the colonial period, countries such as Nigeria took legislative measures against unwanted arranged marriages primarily by establishing a minimum legal age at marriage and abolishing child marriage for girls (Omideyi, 1983; Segamba et al., 1988). Since independence many African countries themselves have sought to liberalize laws to accommodate a wider range of marriages such as legal/statutory, religious, and consensual arrangements. Nowadays, the minimum age of marriage and the conditions of consent are set out in the legal codes of a number of countries. As Table 3-4 reveals, the legal

minimum age of marriage for women varies enormously according to major administrative divisions, or religious or ethnic groups. Some countries have established a range of ages: from no minimum in some areas to 21 in other areas of Ghana, 9 to 18 years in Kenya, and 9 to 16 years in Nigeria and Côte d'Ivoire.

Practices such as polygyny and bridewealth have also been the targets of reform in some African governments (see Ngondo a Pitshandenge, 1988, for a general review). Objecting to what they perceived as the sale of women, colonial authorities often attempted to discourage if not to abolish bridewealth practices (Phillips and Morris, 1971). However, efforts to broaden the scope of allowable marriages, rather than constrict choices, have been among the most successful of legal initiatives. People nowadays marry under many different religious and civil arrangements. On the whole, how

ever, customary marriage processes still predominate. In Côte d'Ivoire in 1980–1981, for example, fewer than 2 percent of women in formalized first unions were in civil unions; the rest entered marriage through customary marital procedures (Meekers, 1992). Even among the educated, Western marital procedures have not completely supplanted customary practices. An analysis of the postenumeration survey of the Ghana 1960 census (Tetteh, 1967) revealed that most couples who had a church or civil wedding did so after carrying out the customary rites (see also Crabtree, 1950; and Oppong, 1974). These findings suggest that a civil ceremony may be the final stage in a sequence of marital events among the educated.

Despite the ideals that these laws aimed to achieve, reality has often been quite different. Though legal measures have suppressed most outright efforts to betroth infants or small girls, ages at marriage are still quite low in some countries. Part of the problem lies in the difficulty of enforcing laws on marital age in countries where few people actually know their ages, and where a marriage does not need to be reported to the authorities unless it is accompanied by a civil or religious ceremony. Another obstacle is the very strong influence that religion exerts in some countries. The tenets of Islam in particular pressure families to ensure their daughters' purity by marrying them when they are young. Moreover, many countries recognize the legitimacy of several legal systems governing marriage: Ghana recognizes customary, Islamic, Christian, and ordinance unions (Pool, 1972); and Sierra Leone recognizes customary, statutory (including Christian), and Islamic marriages (Harrell-Bond, 1975). The ages at which marriage is permissible under different guidelines may vary markedly, even for the same country. The fact that many countries set out a range of minimum ages rather than a single age implies that compromises had to be struck to accommodate diverse, potentially contradictory marriage practices.

Another reason why reality does not always correspond with legally stipulated marital ages is that countries frequently provide for overruling laws regarding marriage and for abrogating the consent of the partners. To take the example of the Zaire code of 1985, the minimum age of marriage for men is 18 and for women 15; "however, the conciliation court may for serious reasons be legally empowered to grant a waiver of the age limit. The court shall decide on such a request by any person who justifies the need for a waiver" [clause 352, emphasis supplied] (Annual Review of Population Law, 1985). The new Rwanda code of 1988 is similar. The minimum age of marriage is set very high, at 21 for both parties, but the code provides that this requirement can be set aside "for serious reasons" (Annual Review of Population Law, 1988). In countries with no national legislation on minimum age, the question is left to the customary courts. Yet the legal framework in one such country, Niger, expresses a situation that applies implicitly even in countries with formal laws: "The consent of

the spouses is not considered as one of the essential conditions for the validity of the marriage, especially when it is the first marriage of a boy or girl living under paternal authority. The marriage is decided upon by the heads of the two families" (Knapp, 1979:N-48). In Nigeria a valid customary marriage is based explicitly on the premise that "[t]he actual contracting parties were the parents, not the intended spouses, which is traditional under Nigerian customary law" (Annual Review of Population Law, 1981:82).

For one reason or another, legal attempts to bar early marriage often fail. According to the Cameroon Fertility Survey of 1978, part of the World Fertility Survey, 8 percent of urban women and 25 percent of rural women were married before the minimum legal age of 15 years (United Nations, 1989). And although the minimum age in Côte d'Ivoire is 18 years, data from the WFS of 1980–1981 showed that 40.9 percent of urban women and 42.6 percent of rural women were married before 18. In Senegal, where the minimum legal age is 16 years, the WFS of 1978 revealed that 15.5 percent of urban women and 35.6 percent of rural women first married before age 15 (United Nations, 1989).

Attempts to suppress bridewealth have been almost as unsuccessful as efforts to regulate the age at marriage. For example, among the Iteso of Kenya, Nagashima (1989) reports, the girl's family now demands cash as compensation for the costs of educating her and for the loss of her income. Even where the amount of bridewealth has declined, it would be difficult to separate the effects of legal sanction from bridegrooms' economic hardship. And if families fear legal retribution for demanding bridewealth, increasing expectations that suitors will pay a girl's school fees or initiation expenses may stem from families' efforts to disguise outright bridewealth.

As for polygyny, legal measures seem to have had little direct effect. For example, although Côte d'Ivoire abolished polygyny in 1954, data from the WFS of 1980–1981 disclosed that about 43 percent of currently married women aged 15–49 were in polygynous unions (Direction de la Statistique, 1984). Western Africa shows the least decline in polygyny. The overall incidence of polygyny in Ghana remained virtually unchanged between the 1960 and 1970 censuses; indeed, it rose in each educational category over the decade (Aryee and Gaisie, 1981). Yet, just as legal measures may have converted bridewealth payments into more legally acceptable forms, a similar generalization seems applicable to the outcome of legislation on polygyny. Even though the practice of having more than one legal wife at a time may have declined, particularly in urban areas, some researchers claim that it has been replaced by new types of union formed between women and already married men (van de Walle and Kekovole, 1984; Clignet, 1987; Lacombe, 1987). Lacking the social recognition that is usually accorded to civil or even to customary unions, these new unions have been labeled "outside" marriages (Baker and Bird, 1959; Harrell-Bond, 1975; Karanja,

1987), "informal arrangements" (Pellow, 1977), and in francophone countries, le deuxième bureau (Lacombe, 1983).

For Botswana, although there may be some problems with the data, there is some evidence that new legal measures concerning access to property and credit underlie some part of the country's very low rates of marriage among adolescent women as documented in the recent DHS. (The importance of these rights, in fact, may suggest that one reason for these low rates was that young women were quite anxious to report to the DHS interviewers that they were unmarried.)

To be sure, education may be playing some role in the rising age at marriage. Indeed, Botswana is one of the few African countries where females outnumber males in secondary school (World Bank, 1988). Although, according to Gulbrandsen (1986:21), educated women may be choosing to remain single "rather than being beaten by a foolish, illiterate husband," women may simply be educating themselves out of the marriage market. But female education is rising in several southern African countries, and marriage rates have declined more sharply in Botswana than in many of its neighbors. Little research directly addresses this contrast; but a key difference in Botswana with respect to property rights, documented by Larsson (1992), suggests that legal changes in other domains may affect marriage profoundly. Whereas in most other southern African countries women must marry in order to obtain access to property, in Botswana unmarried women have the same legal rights to property, credit, and business rights as do married men. This legal advantage, conferred after independence in 1966, benefits rural women with respect to rights in farm land; it also helps urban women who seek to acquire plots to build or to buy houses for rental income or self-employment activities. Larsson (1992:10) argues that

The key point is that only unmarried women in Botswana have these extraordinary rights. Married women, by contrast, are subject to the husband's authority in property matters concerning land and housing. Moreover, they cannot obtain loans without the husband's permission, nor can they lodge court complaints. Unlike women in countries such as Zimbabwe and Zambia, unmarried women in Botswana, though their rights to land and property are by no means consistently enforced, have more legal leverage if they try to press claims (Hansen and Ashbaugh, 1991). By contrast to women in other southern African countries, then, Botswana women have more options to fall back on if they do not wish to marry particular men. Many appear to see delaying or even forgoing marriage as their best route to security.

What are the effects of new religions on the age of women at marriage and on premarital fertility?³ On the whole, Islamic societies have lower levels of female education and ages at marriage for women than those with Christian majorities (Lesthaeghe et al., 1989). (For men, the age at marriage varies widely among ethnic groups that practice Islam, from very high among the Hausa to fairly low among the Swahili.) Because of their intense concern with female purity, Islamic groups try to ensure that most births occur within the bounds of marriage by compressing the gap between physiological maturation and marriage; indeed, they sometimes reverse the order. Lesthaeghe et al. (1989) point out that very young ages of marriage for women produce low overall rates of premarital fertility. Whether the cause stems from efforts to place tight controls on women's reproduction (for example, J. Goody, 1973) or from Islamic doctrine per se, Muslims are less likely than Christians in the same country to report having had a premarital birth (see Table 3-5).

Yet we cannot assume that religious doctrine itself systematically produces these patterns or even that they hold consistently. One exception to the usual Islamic patterning of marriage, sexuality, and fertility among young women appears among the Twareg camel herders and traders in the Sahara (Worley, 1991). Though devout Muslims, they differ markedly from other African Muslim populations in the freedom of sexual action they accord women, especially those from high-status families. The marriage process lasts a considerable time. The groom takes a year or two to amass wedding gifts, less for the parents, the typical pattern, than for the bride. The bride may even decide that she needs another year or more before separating from her kin. Even then, marriage does not immediately result in cohabitation because "[t]he parents want to be sure that her husband appreciates her and shows her respect" (Worley, 1991:331). During the first two or three years of marriage, "[b]oth sexes may have many affairs ..." (Worley, 1991:333). Wives eventually bear highly valued children in whom they retain rights in case of divorce.

More systematic research might use such variations to ask how Islam is related to adolescent reproductive behavior. Influences could stem from (1) religious convictions deriving from various schools of Islamic law, (2) patterns that predate the advent of Islam, or (3) a tendency of Muslims to assimilate regional norms into their marital patterns. In Nigeria, DHS data on age at first marriage by religion support the thesis that differences be

tween Muslim and non-Muslim populations in the same province may be small, perhaps because the conservative nature of predominantly Muslim areas affects ages at marriage of both populations. In Ondo State, for example, ages at first marriage for Muslim and Protestant women aged 25–29 are almost identical at 19.7 and 19.8 years, respectively (Medical/Preventive Health Division, Ondo State and Institute for Resource Development, 1989).

Ages at marriage in predominantly Christian societies are generally higher, perhaps because of their greater emphasis on formal schooling and on monogamy. Yet a widespread perception in Africa is that many of these religious injunctions have had perverse effects, producing a large number of premarital births. One explanation blames Christian tenets themselves, in particular, the insistence that followers be monogamous and abandon practices such as female circumcision, widow inheritance, bridewealth, and partial sexual relations (ngweko) for the unmarried (see, for example, Kenyatta, 1971; Ahlberg, 1991). Yet in a wider social context that permits and even values polygyny, some men may have responded to injunctions to be monogamous by marrying one woman in the church while simultaneously taking on "outside" wives (see, for example, Mann, 1985; Karanja, 1987) or becoming "sugar daddies" to young mistresses (Dinan, 1983). Some observers argue that Christianity's demand to delay sexual relations until marriage means that young men in eastern Africa are no longer taught traditional methods of sexual play without risking pregnancy; as a result most proceed with full sexual relations rather than abstaining completely. As

early as 1936 Krige pointed to evidence of inconsistencies between ideal norms and practical reality in southern Africa:

It is quite difficult to muster convincing evidence about either the effectiveness of the older, indigenous checks on adolescent fertility or Christianity's effects in suppressing them. In some countries Christian mission activity may well have been responsible for the erosion of indigenous controls over unsanctioned adolescent sexuality and fertility. But if it has, then areas with numerous Christian adherents should have the highest rates of nonsanctioned adolescent sexuality and reproduction. The new DHS results point to a counterexample: Burundi.

Burundi is heavily influenced by Christianity: 68 percent of all women enumerated in the 1979 census were Catholic and 8 percent were Protestant (Segarnba et al., 1988). Whether Christianity or some other factor is the cause, Burundi has minimal reported polygyny (10 percent of women; Segamba et al., 1988). Although very few women have any secondary education, the country has a relatively high age at marriage for women (see Table 2-7). Even so, few women report having had a premarital birth (see Table 2-10), and fewer reported premarital sexuality than in any other country except Mali (see discussion below concerning Table 3-7). The case of Burundi therefore does not support the hypothesis that Christianity inevitably erodes controls on adolescent sexuality. (It is worth pointing out that Burundi is in a sense a subpopulation because of its small size. Similar anomalies would undoubtedly surface elsewhere by examining more within-country variations.)

Although Christianity's influence over marital and premarital moral conduct is variable, Christianity may have had a stronger, though more subtle and possibly unintended, influence on children's legitimacy than on marriage itself. A study by Mann (1985, 1988) of nineteenth-century elite families in Lagos, Nigeria, provides important evidence that supports this hypothesis. Mann points out that Europeans in Lagos

These domestic values translated into clear economic and legal stipulations. Christian marriages were to form conjugal property estates, whereby a man shared his income during life, and his inheritance at death, with his

wife and their children, not with his siblings and their children, as local Yoruba culture dictated (Mann, 1985). According to local culture, a man acquired children by securing reproductive rights in a wife; but he could also legally acquire his illegitimate children by compensating his lover's family or husband. Through its unyielding insistence on monogamy and on marriage as a bond between husband and wife, Christianity undermined an "outside" child's legitimate link to a paternal kin group. A man's children who were born to an "outside" wife were necessarily illegitimate, regardless of the man's willingness to recognize them.

The spread of education is a singularly important change affecting marriage in Africa. In many regions of the Third World, including sub-Saharan Africa, higher levels of education and literacy of women are associated with higher ages at first marriage (Casterline and Trussell, 1980; McCarthy, 1982; McDonald, 1985; Trussell and Reinis, 1989). In the WFS, the median age at first marriage for women under age 25 increases as one moves from illiteracy to full primary education, from one year in Benin to over three years in Cameroon, Senegal, and Nigeria (Lesthaeghe et al., 1989). Because formal education, more than almost any other innovation, appears to postpone or attenuate the marriage process, it creates a potential for interpreting pregnancies as "premarital."

Table 3-6 demonstrates that the more education a woman has, the later she will marry; women aged 20–24 with no education were about twice as likely to be married in their teenage years as those with secondary or higher education. The table also suggests that there are different thresholds at which education affects the propensity to marry as a teenager. In eight out of our eleven countries, the likelihood of teenage marriage differs little between uneducated women and those with primary schooling, so that it appears to be secondary or higher education that significantly defers marriage (see also Westoff, 1992). These results may have changed considerably from 40 or 50 years ago, when primary education was such an innovation for women and when children often did not enter school until age 10 or 12. Women with primary schooling thus may have married much later than those with no education.

Greater education is also generally associated with more reported premarital sexual activity, as Table 3-7 shows. Exceptions include Ghana and Mali, with little differences in premarital activity between uneducated women and those with primary schooling, and Botswana and Kenya, with little difference between women with primary and with secondary education. Ondo State, Nigeria, may comprise another important exception to the education/premarital sexual activity link, although the connection is by inference from

low levels of premarital births. (Whereas most regional anomalies are obscured in nationally representative surveys, Ondo State merits special attention because its population is larger than some of the national populations considered in this report.) Ondo State, despite its relatively high levels of female secondary education, has very low levels of premarital fertility. Whether those levels are achieved through abstention, or by the use of contraceptives, or both is unclear (Caldwell et al., 1992, argue for the second for the Ekiti District).

High levels of education appear to be associated with greater prevalence of ''outside marriages'' as well as with more premarital sexual activity. From a woman's point of view, education and formal polygyny mix poorly. Karanja (1987) points out that young educated women are less eager to be polygynous wives; they resent becoming junior wives to women with less education or even senior wives in polygynous marriages (see also Brandon and Bledsoe, 1988). Because highly educated women face a marriage market restricted to potential partners who are as educated as themselves, they often prefer unsanctioned relationships with older men of wealth, hoping that their roles as "outside wives" will help them advance into higher social echelons.

Urban-rural differences in the probability of teenage marriage present a potentially crucial dimension of change. DHS data show that in Kenya and Mali, whether they lived in urban or rural areas made relatively little difference in the likelihood that women had married before they were 20. (This finding is based on the responses of women aged 20–24 at the time of the surveys.) In Botswana and Burundi, rural women were actually less likely to marry as teens than their urban counterparts, but again the differences are small (Table 3-8). For most DHS countries, however, the demands and opportunities of urban life appear to have substantially delayed marriage among adolescents. Once again, inferences about causal direction should be drawn carefully. When rural youth with prospects for employment or education in the city leave the countryside their departure effectively raises the ages of urban marriage and lowers those in the rural areas (see also Findley and Williams, 1991).

Although place of residence is strongly associated with age at marriage, its association with premarital sexual activity is much less clear (see Table 3-9). In Burundi, Liberia, Nigeria, and Uganda young urban women were much more likely to have engaged in premarital sex than rural women. In Ghana, Kenya, Mali, Togo, and Zimbabwe the differences are small. Botswana actually has substantially higher rates of premarital sexual activity in rural areas. This anomaly may simply reflect the extraordinarily long periods

during which many of Botswana's rural women now remain unmarried, as we pointed out earlier.

The variability in the urban-rural differences from one country to another, as important as they are, suggests that African cities differ so much in their occupational structures and residence patterns that they do not constitute a unitary social phenomenon, and therefore do not have a uniform, predictable impact on behavior patterns such as fertility rates.

How these forces of economics, law, religion, formal education, and urbanization affect adolescent sexuality and childbearing in Africa is not easy to sort out. Nevertheless, we can offer a few general observations.

The constant incorporation of new symbols that legitimize marriage and the invocation of new legal and religious codes deepen the inherent ambiguities in marital status. They also create more potential for drawing out the marriage process and for testing out multiple partners before marriage. (Unfortunately, we cannot reliably ascertain whether marriage has indeed become a more attenuated process than in past generations; see also van de Walle, 1993).

Another major trend has been the diminution of gerontocratic control over marriage (see, for example, Robertson, 1984), leading to growing autonomy in the choice of partner, especially among educated urban families (Omideyi, 1983; Lesthaeghe et al., 1989; see also La Fontaine, 1972, and E. Goody, 1973, among many others, for changes in rural areas). The results of the Togo DHS substantiate this point. It asked, "For your last union, was your husband chosen by your family, you yourself with the advice of your family, or you yourself without the advice of your family?" Of women in their first unions, 27 percent of those aged 20–24 reported that their families alone chose their husbands; while among the 45-to 49-year-olds, 46 percent reported that their families chose their husbands. Analogous patterns emerge when women are grouped by residence and education. Among rural women, 38 percent said their families chose their husbands, compared with only 14 percent of urban residents. And among women with no education, 40 percent said their families chose their husbands, compared with only 14 percent of women with some education.

Predictably, many people point to the demise of elders' authority as the root of modern evils, especially the rise of what they see as moral degeneration among the young. Yet from the perspective of young women who do not want to grow up to find their conjugal futures were decided long ago in their childhood, this aspect of the demise of elders' authority is a decided improvement.

One of the frequently noted concomitants of elders' loss of control over marriages is an apparent decline in full-scale marriages and an associated rise of consensual unions (see E. Goody, 1973; Bleek, 1978; and Robertson, 1984 for Ghana; Southwold, 1973, for Uganda; La Fontaine, 1972, for the Gisu of the Kenya-Uganda border; Etienne, 1986, for the Baulé of Côte d'Ivoire; and Meekers, 1992, for Togo). Once again, it is difficult to verify this trend; people may be more likely to label a union as unofficial if it has not been mediated by elders.

Closely related to the rise of consensual unions is the dramatic and controversial transformation in polygyny. Karanja (1987) argues that low-status women can achieve social mobility by becoming "outside wives" of elite men (see also Mann, 1985). Other studies, however, point to the inequality inherent between the legally married wife and the outside wife, and to the consequent inequalities among their children.

Whatever their effects on women individually, factors such as education, Christian ideals of monogamy, and aspirations to social mobility have powerful effects on the market for formal marriage in Africa. They limit the pool of appropriate partners for high-status women to those men who are educated, affluent, and unmarried, and they tempt women to delay entry into formal union in hopes of better matches.

To better understand the social dynamics that foster these demographic outcomes, we need to depart temporarily from the search for precise categories and measurements, and squarely confront the ambiguous nature of conjugal unions under conditions of sweeping change. Looking at transformations in this light suggests that women may be construing their marital statuses according to the expectations of prestigious groups. In Freetown, Sierra Leone, uneducated women are likely to report themselves fully married after customary marriage rites have been performed. But better-educated women are more likely to report being single or "engaged" if they have had no civil or religious ceremony. This observation, if accurate, implies that to these women defining a union as "nonmarital" is preferable to admitting being in a customary union (Brandon and Bledsoe, 1988), a strategy that may help women with aspirations for elite status to disassociate themselves from polygyny.

A man facing economic and social pressures can initiate several potential conjugal links, but minimize his costs by eventually selecting a principal wife—usually the one with the most education or the most prestigious family connections—and marginalizing the rest as outside wives who lack full legal status under modern statutory or ordinance codes (see also Mann, 1985). The outside or country wife tends to live in a more rural area or a poorer neighborhood, and her children go to less prestigious schools.

In the past, even formal co-wives were never perfectly equal. Yet factors such as education appear to be exacerbating inequities among out-

side wives and their respective children. The status of such women and the children they bear is often murky (Phillips and Morris, 1971; Vellenga, 1986). Elite men can use this ambiguity in court cases either to deny the responsibilities associated with paternity or to sue other men for "adultery" damages.

Western Kenya presents an archetypal case of the combined dynamic of ranking and marginalization, whereby some low-status women are demoted in multiple-partner unions and others are kept in a remote holding pattern of the marriage process. Hakansson's (1988) work shows that educated and better-connected men are investing heavily in high-status marriages both because they can control their wives' cash income and because of the powerful affinal networks they can tap. High-status women and their children likely fare quite well, but lower-status women confront increasingly dire prospects because of land shortages. The still-extant colonial "customary law" ensures women's rights in land only in the context of bridewealth marriage. Lacking acreage that is considered sufficient to support a large family, most men contract legal marriage to only one wife at a time. They delay legal marriage but strike up temporary relationships with women for domestic services and personal comfort. A poor woman is caught in a bind: Lacking alternatives, she must establish a much-needed relationship with a man, but runs the risk of being sloughed off, destitute, in the competition over diminishing resources. For the same reason of land shortage, her own family members want to move her out quickly, especially if she has children who will compete for family land. A rejected trial wife may thus end up with no home and fast-disappearing chances that a subsequent partner will want to support her children. Women, therefore, are marrying later not necessarily to prolong a school career; rather, they may be going from one man to another, increasingly less likely to marry as they accumulate children.

The tendency for individuals to draw out the period of conjugal testing and make relationships more contingent has immediate relevance for adolescent fertility. The strong preference is for births to occur within marriage. But for many young women, pregnancy (like marriage) is a process of testing relationships with potential partners and weighing the possible benefits of continuing them against the costs of forgoing career options or having the father abscond.

The fluidity and ambiguity of marriage can work either to the advantage or to the disadvantage of individuals. Some women can take advantage of the ambiguities of marriage to test men's reactions and then respond

accordingly. Once pregnant, for instance, a woman may decide to proceed with the pregnancy and bear the child if the man appears to be showing paternal interest and is supportive; if he proves uninterested, she may try to abort in order to continue her school career or find another man.

For men, the trend in Africa appears to be one of marginalizing certain unions, rather than delaying unions or forgoing them altogether. New pressures brought by education and economic competition appear to be intensifying differences among a man's partners who previously would have lived together, shared household resources, and acknowledged each other as co-wives. To be sure, for many of these women, becoming an outside wife is a significant move up the social mobility scale. But the fact remains that women who are acknowledged as "official" wives enjoy relative security within the conjugal unit, whereas lower-status women find themselves stalled as outside wives in a remote orbit of the marital process by men who must maintain a monogamous appearance.

At a broader level, although the problems associated with premarital adolescent reproduction seem to point to marriage as the main determinant of legitimate reproduction, a premarital pregnancy is not necessarily unwanted or illegitimate. Viewing marriage as a process means that the birth of a child—the first child, in particular—may be an important step toward a union, rather than a logical outcome of it. In view of these ambiguities, making a distinction between conceptions before and within marriage may be largely inappropriate. It is not clear, then, that marital status is the best variable for distinguishing between sanctioned and unsanctioned childbearing.

The next chapter looks beyond marriage to some further conditions of entry into childbearing. It shows that in many African societies, people may place less emphasis on marriage for defining legitimate childbearing than on children having an acknowledging, supportive father or on the parents completing a successful transition to adulthood. Because the DHS does not immediately suggest other variables that may be as important to the legitimation of a pregnancy, marital status and age at marriage must remain the index questions. We caution, however, that this is a default strategy: Marriage is at best a proxy for what are likely to be considered locally much more important criteria of reproductive entitlement.