The study of female fertility according to whether the women are living in monogamous or polygynous unions is frequently limited by data availability. In this article, Solène Lardouxand Etienne van de Walleexamine this question for Senegal, where polygyny is widely practised, through an innovative use of the 1988 census that provides detailed information on the partners in unions. The completeness of their data allows the authors to qualify the results of previous studies based on smaller populations and, more importantly, to increase the range and power of the analysis. By examining the interaction between the age of the husband, that of the woman, and the number of co-wives, in particular, they reveal the complex influence of polygyny on fertility and produce some unexpected results. As well as a contribution to the study of fertility in Senegal, this article is a methodological exercise that increases our understanding of the relationship between monogamy, polygyny and fertility.
Polygyny, a form of union where a man is married to more than one wife, is widely practised in Senegal. Comparisons of fertility surveys in sub-Saharan Africa have shown that the prevalence of plural marriage in Senegal is among the highest on the continent, and has been relatively stable over time (Timaeus and Reynar, 1998; Locoh, 1995). The proportion of currently married women of childbearing age who were in polygynous unions was 48.5% according to the 1978 World Fertility Survey, 46.5% in 1986, 47.3% in 1992/93 and 46% in 1997 according to the Demographic and Health Surveys of Senegal (Ndiaye et al., 1997).
In this article, we will briefly review the main points on which the literature generally agrees concerning the relationship between polygyny and fertility, before discussing the merits of various sources and methodologies used to study the issue. The body of the article is devoted to an account of our findings on the fertility of unions by age of spouses and by number and rank of the wives, as obtained from an under-exploited source of data, the 1988 census of Senegal. The emphasis is methodological. We cannot do justice here to the vast anthropological literature on polygyny.
I - Polygyny
Although the numbers of men and women of childbearing ages are roughly equal in most human populations, the practice of polygyny is made possible by a large age difference between the spouses, and by the rapid remarriage of women after widowhood or divorce. In rural Senegal, the proportions single are very small after age 25 for women, and after age 35 for men. Men usually start their married life as monogamists, and may acquire additional wives later in life; conversely, the number of wives may be reduced by divorce or widowhood. The majority of women spend some of their married lives as co-wives. Polygynous families rarely include more than two or three wives. Islam, the main religion of the country, allows a maximum of four. However, a man may inherit the widows of one of his brothers, and occasionally the number of wives may rise to six or seven.
Polygyny and large families provide labour, physical security and prestige among peers and family members. Boserup (1970) associates characteristics of modes of production with the widespread practice of polygyny in rural areas. Goody (1976) hypothesizes the existence of a relation between hoe cultivation and polygyny, and plough cultivation and monogamy. In societies were the payment of bridewealth is the common practice, richer men tend to contract more polygynous unions (Timaeus and Reynar, 1998). Thus, increased wealth is a cause and a consequence of polygyny.
Urbanization and industrialization are processes that weaken extended family systems and lower the prevalence of polygyny. Matrimonial preferences may depend on the degree of access to education, to the media and to consumption of western goods. The lessening of the practice of postpartum abstinence and the increase of access of girls to education are social changes that may contribute to a decrease in the practice of polygyny (Timaeus and Reynar, 1998).
A number of studies have focused on the effects of polygyny on fertility. Most of these were concerned with African populations, although historical studies of the Mormons in America have also contributed to a better understanding of the topic (Anderton and Emigh, 1989; Bean and Mineau, 1986). Plural marriage tends to create a higher demand for women on the marriage market. In polygynous societies, female age at marriage tends to be early, marriage tends to be universal, and women tend to remarry early after widowhood or divorce (Antoine and Nanitelamio, 1995); where they exist, as in Senegal, widow inheritance or leviratic marriage ensure that even older women with children remarry systematically. The institution of polygyny in a society, therefore, may promote high fertility, because it tends to be associated with more time spent by women in the married state, and therefore more exposure to the risk of childbearing (Pison, 1986; Timaeus and Reynar, 1998). However, research has generally concluded that the fertility of individual women in polygynous unions was lower at all ages than the fertility of women in monogamous unions (Pison, 1986; Pebley and Mbugua, 1989).
Polygyny may reduce the fertility of individual married women for a number of reasons. The main effect operates through a lowered frequency of intercourse. Anderton and Emigh (1989) have distinguished between a “sexual competition model” where sexual relations are reduced more or less equally for each wife, and a “favouritism model” where the preferred wife, often the youngest one, maintains a high fertility level while the others are comparatively neglected. Madhavan (1998) noted that polygynously married women are concerned about their co-wives’ fertility since childbearing determines their status. A man may also keep his wives in separate locations, as for example when one looks after his farm in the countryside while another accompanies him in the city. When the co-wives live together with their husband on the same compound, having alternative sexual partners facilitates the observance of the long intercourse taboo after a birth and of extended periods of breastfeeding, thus potentially increasing birth intervals for each woman in turn. Blanc and Gage (2000), however, report that the difference in the duration of breastfeeding between polygynous and monogamous wives is small. Clignet (1970) argues that the authority and power of members of specific ages or rank may explain some fertility differences. Senior wives past the age of childbearing may influence the fertility of their younger co-wives by imposing the observance of traditional norms and taboos. The lower fertility of polygynous marriages may also be an artifact of selection, as a husband may be more likely to take an additional wife when he did not get all the children he wanted from his previous union. Timaeus and Reynar (1998) confirm that there is clear over-representation of childless women in polygynous unions. It is also sometimes suggested that multiple marital partners may facilitate the diffusion of venereal diseases that lead to secondary sterility, but this hypothesis has not been tested.
Garenne and van de Walle (1989) have suggested the existence of yet another mechanism on the basis of longitudinal data from the demographic surveillance system of Ngayokheme in Senegal. They showed that the age of the husband has a consistent effect on fertility. Polygyny monopolizes younger women for the benefit of older men. As the number of wives increases, the age difference between the spouses increases too, while male fecundity decreases with age. The mechanism suggested by Garenne and van de Walle is not a lengthening of birth intervals (which would suggest a decline in the frequency of intercourse with age), but a growth in the proportion of infertile unions (a biological rather than a behavioural explanation.) In their study, the decline of female fertility with the increasing number of wives is mostly accounted for by the effect of the age of the husband, although another factor plays a minor role: the type and rank of union influence the likelihood that the spouses live together in the same residence.
In this article, we will test the following hypotheses. First, the fertility of wives decreases with the number of women married to the same man. Second, fertility increases with the rank of the wife in the polygynous union, in conformity with the “favouritism” model; the most recent wife will benefit from more attention from her husband, and be more exposed to sexual intercourse, suggesting that the rotation of the husband among his wives is unequal. Third, the age of the husband has a negative impact on the fertility of his wives. Testing each of these three hypotheses will require controlling for the other two factors (for instance, control for rank and age when we test the effect of number), since number, rank, and age of the husband tend to evolve together, and it is only exceptionally, for example, that a young husband would have three wives or more, or that a third wife would not be married to a man who is not much older than she is. In passing, we will test whether living conditions have a favorable impact on the fertility of wives. The information available is not adequate, however, to investigate this topic in depth; and so we use an index of household wealth as a control rather than as an element in a formally stated hypothesis.
We add two hypotheses which have not been tested in the literature. The fourth hypothesis is that the occurrence of a birth to one wife of a polygynist will affect the probability of childbearing by a co-wife. Pregnancies, births and postpartum will result in periods of abstinence for one wife, and may thus increase the probability of intercourse with another. Thus, we hypothesize that births by co-wives during the same year will be less likely. The fifth hypothesis is that the presence of a wife who is past her fecund years will have a negative influence on the fertility of the other wives, either because she will claim her share of bed time with her husband, thus lowering the frequency of intercourse for the others, or because she will enjoin her co-wives to respect intercourse taboos and to space their births.
Before turning to multivariate analysis to test our five hypotheses, we will show that polygyny in comparison to monogamy is associated with patterns of household structure where the spouses are less likely to live together, and hence, where women are less exposed to the risk of childbearing.
II - Data sources
Past studies of the relationship between polygyny and fertility have often been based on fertility surveys such as the Demographic and Health Surveys. These rely on small sample sizes, and on an even smaller number of polygynously married women, so that the analysis must be limited to the most frequently encountered triad (a man and his two wives) and to low ranks of wives (Bankole and Singh, 1998; Ezeh, 1997; Gage, 1995; Speizer, 1995). Typically, the surveys include few questions on marriage and polygyny. For example, the 1997 DHS for Senegal merely asks: “Does your husband/partner have other wives in addition to you?” (Earlier Senegalese fertility surveys also asked questions on the number of wives, and on rank.) There is no mention of the other wives’ ages, and no direct question on the age of the husband (even when men and women were interviewed as a couple, if the husband was polygynously married only one of his wives was linked to him as part of a couple). Timaeus and Reynar (1998, p. 160) note that “the natural unit for the study of polygyny is the union” and that there is numerical equivalence between the number of men and of unions. The sampling of women in the DHS, however, is done on an individual basis, so women are not linked to most of the characteristics of their husbands, including their age, or to the fertility of their co-wives. Women who are past childbearing age are not interviewed, even though younger co-wives owe their status of polygynous wives to their presence. It is not possible to ask whether there is a positive correlation between the fertility of co-wives (testing the possibility of an effect of competition), or a negative correlation (testing the possibility that there is a tendency to alternate births). The survey makes no distinction between co-resident and non co-resident wives, although there is a question on whether the respondent lives with her husband or partner.
Garenne and van de Walle (1989) used longitudinal data from the Ngayokheme demographic surveillance system in Senegal. The data are in many ways ideal because of their depth and specificity, but the study population is relatively small, totalling about 5,000 persons. They were collected between 1962 and 1981, thus providing many wives’ person-years of observation. The longitudinal nature of the data made it possible to follow the creation and dissolution of unions, and to attribute births to these unions with a great deal of certainty as to whether women were monogamously or polygynously married at the time of each birth. The denominators used to compute fertility rates were obtained by stringing together the person-years spent in the particular category of union. Marriages were considered to exist when both partners reported themselves as married, and the analysis was limited to pairs in which both spouses lived in the study area. Rank was inferred from the sequence of times at which women entered the union. In contrast to data from retrospective surveys, the characteristics of the husband and co-wives of a woman were known.
The present study is based on yet an entirely different approach. Data were extracted from the 1988 census of Senegal. The present research focuses on two regions of Senegal, Kolda and Tambacounda, bordering on the Gambia. Some characteristics of these regions are provided in Table 1. The dominant ethnic groups are the Peul and the Manding. Among all regions of Senegal, these two appear to be the least developed. In 1988, Kolda and Tambacounda scored the lowest on an index of the quality of home construction, they had the lowest proportion of households with electricity or running water, the lowest school enrollment of children aged 7-12, and the highest adult mortality (Pison et al., 1995, particularly Table 2-4). We restrict the study to rural areas in the two regions, in part because polygyny is less frequent in urban areas, but mostly because we are interested in the mechanisms prevailing among natural fertility populations, and it is less likely in the countryside that fertility might be affected by the diffusion of family limitation. Kolda and Tambacounda had the highest proportion of polygynously married women and the lowest age at female marriage among the regions. In a country where nuptiality changes are said to account mostly for a decline of fertility (Ndiaye et al., 1997), they stand out for their high proportion of married women from an early age on. In this article, we assume that fertility differences are not the result of birth control.
Table 1 –
Selected characteristics of the regions of Kolda and Tambacounda in 1988
Characteristics Kolda Tambacounda Population (000) 592 325 Peul (%) 49.3 46.4 Manding (%) 23.6 17.4 Urban (%) 11.0 13.0 Households with access to running water (%) 4.0 9.0 Population aged 6+ with schooling (%) 18.0 18.0 School enrollment, ages 7-12 (%) Males 39.0 26.0 Females 19.0 17.0 Married women in polygynous union (%) 61.3 55.4 Never married women at 20-24 (%) 10.0 9.0 Sources: Pison et al., 1995; 1988 census of Senegal (République du Sénégal, 1990-91).
Table 2 –
Distribution of monogamously and polygynously married males by relationship to the head of household, and proportions living with all their wives (in %)
Relationship to head of household Monogamously married Polygynously married Numbers Distribution (%) Proportion linked (%)(a) Numbers Distribution (%) Proportion linked (%)(a) Head 41,189 50.7 81.4 34,865 73.1 75.7 Head of nucleus(b) 34,720 42.8 71.3 11,391 23.9 59.7 Son 13,763 17.0 76.4 4,521 9.5 65.7 Son-in-law 257 0.3 77.8 33 0.1 45.5 Father 227 0.3 79.3 106 0.2 50.0 Grandson 2,396 3.0 77.6 634 1.3 62.5 Brother 10,591 13.0 79.3 4,127 8.7 67.7 Other relative 3,296 4.1 73.8 852 1.8 48.7 Non-related 1,938 2.4 51.0 407 0.9 17.2 Other head(c) 2,252 2.8 8.5 711 1.5 12.7 Visitor of household head 2,415 3.0 0.0 578 1.2 0.0 Husband 69 0.1 0.0 329 0.7 0.0 Visitor of nucleus 2,562 3.2 0.0 469 1.0 0.0 Unknown 205 0.3 0.0 53 0.1 0.0 Total 81,160 100.0 71.8 47,685 100.0 69.6 (a) Linked men are co-resident with their wife (monogamists) or with all their wives (polygamists). (b) Households are generally composed of more than one nucleus. For instance, 42.8 % of monogamous men are heads of their own unit: these men are often sons or brothers of the head of the household (17% and 13% of the total monogamous men, respectively). (c) Other heads were coded as heads of household, although they are in a secondary nucleus. Source: Micro-data for rural Tambacounda and Kolda regions, Senegal census 1988.
Table 3 –
Distribution of monogamously and polygynously married females by relationship to the head of household, and proportions living with their husband and all their co-wives (in %)
Relationship to head of household Monogamously married Polygynously married Numbers Distribution (%) Proportion linked (%)(a) Numbers Distribution (%) Proportion linked (%)(a) Wife of head of household 37,790 46.9 88.7 77,923 68.7 78.7 Wife of head of nucleus(b) 30,034 37.3 82.5 22,934 20.2 63.1 of son 12,915 16.0 81.4 9,775 8.6 63.8 of son-in-law 243 0.3 82.3 121 0.1 28.9 of father 222 0.3 81.1 216 0.2 56.0 of grandson 2,199 2.7 84.5 1,308 1.2 63.1 of brother 9,986 12.4 84.1 9,008 7.9 67.0 of other relative 3,001 3.7 81.0 1,729 1.5 50.3 of non-related 1,238 1.5 79.9 421 0.4 34.4 of other head(c) 230 0.3 83.0 356 0.3 56.5 Head of household 624 0.8 0.0 1,029 0.9 0.0 Head of nucleus 5,174 6.4 0.0 5,080 4.5 0.0 Visitor of nucleus 6,867 8.5 0.0 6,386 5.6 0.0 Unknown 87 0.1 0.0 78 0.1 0.0 Total 80,576 100.0 72.4 113,430 100.0 66.8 (a) Linked women in monogamous union are co-resident with their husband; linked women in polygynous union co-reside with their husband and all their co-wives. (b) Households are generally composed of more than one nucleus. For instance, 37.3 % of women in monogamous union are wives of men heading their own nucleus: these men are often sons or brothers of the head of the household (16% and 12.4% of the total number of women in monogamous union are married to a son or a brother of the household head, respectively). (c) Other heads were coded as heads of household, although they are in a secondary nucleus. Source: Micro-data for rural Tambacounda and Kolda regions, Senegal census 1988.
Table 4 –
Age-specific total fertility rates of women in monogamous union and of women in polygynous union by rank of wife (linked and non-linked files)
Age Women in monogamous union Women in polygynous union Rank 1 Rank 2 Rank 3+ Linked files 15-19 0.301 0.335 0.274 0.241 20-24 0.324 0.309 0.308 0.283 25-29 0.305 0.275 0.279 0.274 30-34 0.249 0.209 0.210 0.179 35-39 0.182 0.153 0.144 0.143 40-44 0.083 0.066 0.076 0.052 45-49 0.046 0.034 0.030 0.023 Total fertility 7.4 6.9 6.6 6.0 % non-reports 3.3 2.9 3.6 3.8 Number of women 53,178 28,259 30,814 8,460 Non-linked files 15-19 0.240 0.287 0.237 0.227 20-24 0.298 0.303 0.286 0.261 25-29 0.276 0.288 0.260 0.257 30-34 0.224 0.211 0.203 0.178 35-39 0.147 0.176 0.142 0.141 40-44 0.067 0.078 0.061 0.051 45-49 0.026 0.042 0.034 0.038 Total fertility 6.4 6.9 6.1 5.8 % non-reports 5.1 4.2 5.1 5.8 Number of women 18,782 17,581 11,189 3,352 Source: Micro-data for rural Tambacounda and Kolda regions, Senegal census 1988.
The fertility information in the census of Senegal consists of births during the year previous to the census. There is no information on children ever born, but this is not a liability since it cannot be assumed that all of a woman’s children were conceived when she had the same marital status or the same rank as at the time of enumeration. In this study, we assume that her present marital status prevailed during the previous year. A monograph on the demography of Senegal, sponsored by the National Research Council of the United States, has used the 1988 census data on births to estimate fertility (Pison et al., 1995). The data were deemed satisfactory for the purpose, although there were a number of non-responses that had to be allocated. Since women may have omitted births immediately followed by the child’s death, the level of fertility may be slightly underestimated (ibid., p. 47). In the present study, women who did not report the number of their births during the previous year were not included in the fertility computations. This would amount to an assumption that non-responding women had the same fertility as responding ones. Since interviewers may have entered no answer where women reported no birth (a common error), this assumption may result in a slight overestimation of fertility. Fewer than 4% of women did not respond, and the differences by women’s age or by polygyny status were not important (see Table 4 for more details). We assume that the biases, if they exist, are equally distributed among various categories of women, and negligible in the analysis. We compute age-specific fertility rates from cross-sectional indicators that refer only to the year previous to the census date; women can have 1 or 0 delivery. For each age group of women, we compute the ratio of births over the number of women by rank and by number of wives in the union.
The census is generally, but unjustly, considered as an inferior kind of data source for the study of subjects of such complexity. It has clear limitations, of course — for example it includes no fertility or marriage history — and yet it has obvious advantages over the Demographic and Health Surveys. First of all, there are large numbers of cases, even for wives of rank three and higher; and not only on women aged 15 to 49 years. Secondly, it provides information on all the members of a residential unit. The census of Senegal adopts the de jure principle, i.e. it enumerates all the persons for whom the household is the place of usual residence. The unit of enumeration is the household, which is made up for the most part of persons who are related by blood or marriage. Absent residents and visitors are included, but the structure of the household is defined in relation to a “head of household” who may be present or temporarily absent at the time of the census. The order of enumeration is in principle determined by the relationship to the head. For example, a first wife is listed immediately after the head of household, with her unmarried children. Other wives, with their children, are listed in order of their rank. Married children have their own nucleus and are thus enumerated with their own family unit.
The information on relationship is especially detailed in the Senegal census of 1988. A nucleus (noyau) refers to a unit consisting either of a married couple and their children, or of one person and his or her children. The relation of the head of the nucleus to the head of household is given, and other members of the nucleus are defined in relation to its head. This system provides details on the connections between various conjugal units, monogamous and polygynous, included in the household (and not only to the head of household). As part of marital status information, the census provides the number of wives of a man, and the rank of each wife up to the third wife (the closing category being three and over). Some of the wives of polygynists do not live together in the same household, or with their husband; even in this case, the census provides the number of wives of a man, and their rank. It is not only possible to identify the wives of the head of household, but also those of various heads of nuclei, for example those headed by a brother or a son of the head of household. Moreover, the census allows us to distinguish between married members of the household who do not have their spouse or spouses living with them, and those who do.
In order to analyse this dataset, the first task was to assemble files including a husband, his wife or all his wives, with their characteristics and those of their household; these files are designated hereafter as “the linked files”. They include only complete conjugal units, linking sets of co-resident monogamous or polygynous partners
In some instances, the relationship is provided with.... This was possible because the number of wives of each man and the rank order of each wife were given. If a nucleus included a man married to n wives, and the number of his co-resident wives was also equal to n, the conjugal unit was included in the linked file. Since the highest value of n listed in the census was 3+ — meaning “three wives or more” for each man, and “third or higher rank” for women — we have probably included in the file some conjugal units from which higher-rank wives were missing. These would be units where a polygynist husband was coded as having three or more wives, where wives of ranks 1 and 2 and at least one wife of rank 3 or more were found. If one or more wives of rank 3+ were living elsewhere, we would still treat the household as “linked”, for lack of more information
The imprecision involved in the procedure is probably.... When a household contained several wives whose rank was recorded as 3+ (i.e. “third or higher”), it was possible to give a rank (up to eighth) corresponding to the order in which the wives of an individual were listed in the census.
Strictly speaking, the persons in the linked file are not representative of all married persons in the population, but they constitute a sub-population where the relations between all the members of conjugal units can be investigated. The ratio of married women to married men is higher among the non-linked (1.6) than among the linked (1.5), and the ratio of polygynously married women to polygynist husbands is also higher (2.6 versus 2.3). We suspect that this may be due to the fact that some polygynist husbands are living outside of the rural area or of the region where some of their wives reside. The levels of education of the linked and unlinked men and women are not significantly different. Non-linked polygynists belong to households that have the same standard of living and the same access to factors of production as the linked polygynists; non-linked monogamists appear to be marginally better off than monogamists in the linked file. In general, the two sub-populations appear very similar.
III - Characteristics of the population under study
The rural population of Kolda and Tambacounda lives mostly in large extended family households. The average size of the household is slightly above ten persons, and 40% of household members do not belong to the nuclear family (spouses and children) of the head. The preferred residential form of polygyny is co-residence in the same household or compound; very few wives of polygynists live separately as head of their own household. It is quite common that sons or brothers of the head of household head their own nucleus that includes their wife or wives, their children and other members, although this occurs less and less frequently as the age and the number of wives of the head of nucleus increase. In the rural population of Kolda and Tambacounda, 51% of men and 73% of women aged above 12 are married. Among the married, 37% of the men and 58% of the women are in polygynous union. As expected, a majority of married men are heads of household, and a majority of married women are wives of heads. It is striking, however, that many of the married persons are unaccompanied members (i.e. without their spouse or children) of the head’s own nucleus, or of another nucleus.
Table 2 indicates the relationship to the head of household of married men in monogamous and polygynous unions, as well as the proportions of these married men who reside in complete conjugal units, together with all their wives. Table 3 presents similar information on married females.
In the linked files, we estimate that 72.4% of the wives of monogamous men and 66.8% of the wives of polygynous men live with their husband. The missing wives may be found in other households, either as heads of nucleus if they have children living with them, or as members of another nucleus. The large number of married “visitors” (as we call them in Tables 2 and 3) unaccompanied by their family illustrates the fluidity of the household structure, and the instability of residence, even when recorded in a de jure census. These visitors are particularly frequent among the nuclei headed by “other relatives”, by unrelated people, or by the so-called “other heads” who are undefined, but may also be unrelated transitory persons boarding on the premises. Visitors are also found among close relatives of the head of household. Many marriages start while the husband lives in the household of his father, his brother or some other relative; and many women spend extended periods at the beginning of their marriage in their own family of origin, before moving to their husband’s home; or they may make long visits there later in their married life. The status of head of household is acquired at an older age than marriage; the median age of household heads is 46 years. Few married women are head of household, however, because an absent husband will often be reported as the de jure head of household in their stead. Female headship is reserved mostly for widows who have not remarried at the time of the census.
The raw data from the census do not provide any insight into the circumstances that explain why wives and husbands are not found together. But these circumstances may account for some of the fertility differentials between co-resident wives and other married women. In Table 4, we present a simple computation of age-specific and total fertility rates by type of union, distinguishing between those who are in the linked files because they are co-residing with their husband and their possible co-wives, and those who are in the non-linked files because they are visitors or are not living with all their partners. The table also includes the percentage of women who did not report the number of their children born during the year before the census. They were left out of the denominator used to compute the rates.
This simple comparison does not do justice to all the possible influences that could account for the fertility differentials. In particular, it does not distinguish fertility by number of wives or by age of the husband and wives. This we attempt to do with the linked files in the next section. On the other hand, the analysis in Table 4 is more inclusive, and considers women who do not live either with their husband or with all their co-wives. Co-residence is almost always associated with higher fertility, and the difference is most marked for monogamous couples; the advantage for first wives appears to be limited to the first age group. However, more polygynous units have some of their members absent (i.e. the proportions of persons linked in Tables 2 and 3 is lower than in monogamous units). Our data confirm that the fertility of monogamous wives is higher than the fertility of each wife of polygynous husbands, except for the age group 15-19. Monogamous women are much more likely than polygynous women to live in the same household as their husband; and wives of monogamists who are not living with their husband have lower fertility than co-resident spouses. On the whole, the latter situation represents a more anomalous situation than when a polygynous husband does not live with all his wives, and is more likely to have an impact on fertility.
In the linked files, the unit of observation is the marital union, and it includes all partners in the union, i.e. one husband with his wife or all his wives (with the possible exception of unidentified wives of rank 3+ living elsewhere). Table 5 shows the distribution of wives in monogamous and polygynous unions by their characteristics such as age, number and rank of wife. As we use cross-sectional indicators, the effect of age should be carefully considered (Pison, 1986). The median age of monogamous men is 38, of polygynously married men with two wives is 46, and with three or more wives is 53 (see Table 6)
The ages used in this study were recomputed from the.... The younger age of monogamously married women compared with the higher median ages of lower rank polygynously married women suggests that marital status is not stable over time. Most monogamously married women may belong to a polygynous union in the future or did so in the past. Wives of rank 1 tend to be older than wives of ranks 2 and 3 or more.
Table 5 –
Percentage distribution of wives’ characteristics in monogamous and polygynous unions by rank (linked files)
Variables Monogamous unions Polygynous unions 2 wives 3 or more wives Rank 1 Rank 2 Rank 1 Rank 2 Rank 3+ Age (%) 15-24 39.3 14.4 36.7 3.6 12.2 25.0 25-29 20.3 20.4 22.5 9.5 18.0 19.7 30-34 13.1 19.1 14.8 16.0 18.9 16.1 35-39 8.7 14.8 10.2 17.3 17.7 13.2 40-49 10.8 19.0 10.1 30.1 21.1 15.4 50+ 7.0 11.6 5.3 22.6 11.6 10.2 Missing 0.8 0.7 0.4 0.8 0.5 0.4 Total 100.0 100.0 100.0 100.0 100.0 100.0 Ability to read or write (%) Yes 2.4 1.5 1.8 1.3 1.4 1.9 No 97.3 98.3 97.9 98.4 98.2 97.6 Missing 0.3 0.2 0.3 0.3 0.4 0.5 Total 100.0 100.0 100.0 100.0 100.0 100.0 Median age of wife 25.0 33.0 27.0 40.0 35.0 30.0 Mean age difference between spouses(a) 12.5 12.6 18.3 13.2 18.4 22.6 Fertility rate the year prior to the census(b) 0.26 0.18 0.22 0.09 0.15 0.18 Total females N=132,305 58,307 25,572 25,572 7,618 7,618 7,618 % (row) 44.1 19.3 19.3 5.8 5.8 5.8 (a) Median and mean age differences do not differ significantly. (b) Women who did not report their number of births were not included in the computations. Source: Micro-data for rural Tambacounda and Kolda regions, Senegal census 1988.
Table 6 –
Percentage distribution of husbands’ characteristics in monogamous and polygynous unions by number of wives (linked files)
Characteristics Monogamous unions Polygynous unions 2 wives 3 wives or more Age 20-39 52.0 30.5 11.0 40-49 20.9 29.1 26.3 50-59 13.7 21.3 29.2 60+ 13.4 19.1 33.5 Total 100.0 100.0 100.0 Median age 38.0 46.0 53.0 Ability to read or write Yes 9.6 10.1 12.5 No 90.4 89.9 87.5 % 100.0 100.0 100.0 Standard of living(a) Low (0-1 point) 53.6 45.0 32.9 Lower middle (2-3 points) 37.9 42.9 48.8 Upper middle (4-5 points) 8.2 11.6 17.7 High (6-7 points) 0.3 0.4 0.6 Total 100.0 100.0 100.0 Means of production owned None 40.9 31.4 23.0 Ox 16.7 15.3 12.6 Horse 21.3 23.9 24.1 Plough 15.2 20.9 27.4 Cart 5.9 8.6 12.9 Total 100.0 100.0 100.0 N = 91,498 58,307 25,573 7,618 (a) Scale of 0 to 7, obtained by adding points for the following household characteristics: Flush toilet: 2 points; or Pit latrine: 1 point. Interior well: 1 point; or Interior tap: 2 points. Roof, Concrete or Zinc-Slate-Tile: 1 point; Walls, Brick or Cement: 1 point; Radio: 1 point. Source: Micro-data for rural Tambacounda and Kolda regions, Senegal census 1988.
The mean age difference between spouses is of great interest. It increases with the rank and number of wives: it is comprised between 12.5 years for monogamous unions and 22.6 years when the rank of the wife is 3 or higher (Table 5). Figure 1 compares age differences between spouses by type of marriage. Men marry wives from a wide range of age groups. Moreover, as men become older, they marry additional wives of higher ranks who are usually much younger than themselves. (In cases where the age difference between a husband and his wife is negative the wife is older than her husband. These are probably cases of inherited wives.)
Distribution of age differences between husband and wives by type of union, number and rank of wives (in %) Age difference = age of husband– age of wife (in years)
Source: Micro-data for rural Tambacounda and Kolda regions, 1988 Senegal census.
The age difference between a husband and his wife in monogamous unions is almost equal to the age difference between a husband and the first wife in unions with two wives and in unions with three or more wives (mean age difference of 12 to 13 years). In contrast, as shown in Table 5, the difference between the mean age of the husband and of his second wife is 18.3 and 18.4 years in unions with two wives and three or more wives, respectively. Figure 1 shows that the range of age differences between the husband and the second wife is larger than between the husband and the first wife.
One of the main reasons for a man to have multiple marriages is to have many children (Blanc and Gage, 2000). Thus, we present husbands’ characteristics (Table 6) before describing wives’ fertility. As in the case of women, polygyny varies with age. Men in monogamous unions are in large proportion under age 50 (72.9% of them), whereas those in poly-gynous unions, especially in unions with three or more wives, are mostly found to be older than 50.
The level of wealth is higher in polygynous households than in monogamous households
Education is a variable that is often controlled in.... We use a scale of standard of living based on characteristics of the dwelling unit and its equipment, and a variable for the means of production available. Horses are the item that is the most widely owned in monogamous unions. But almost 41% of these unions do not possess any of those means of production. In the next section we see how these indicators are associated with levels and trends of fertility of monogamous and polygynous wives.
IV - Results
The linked files make it possible to analyse the fertility of married women by age, number and rank of wives in the marital unit. Before discussing results of logistic regression on the occurrence of a birth during the year prior to the census, we present graphs that depict the relationship between fertility rates over the one-year period and age of both husband and wives of different ranks in unions with one, two, or three wives or more.
Figures 2 to 4 show fertility rates for wives in monogamous union, and for wives of ranks 1 to 3 or higher in polygynous unions, as a function of their age and of the husband’s age.
Age-specific fertility rates by age of wife and husband in monogamous unions (number of children per woman the year prior to the census)
Source: Micro-data for rural Tambacounda and Kolda regions, 1988 Senegal census.
Age-specific fertility rates by age of wives and husband in polygynous unions with 2 wives (number of children per woman the year prior to the census)
Source: Micro-data for rural Tambacounda and Kolda regions, 1988 Senegal census.
Age-specific fertility rates by age of wives and husband in polygynous unions with 3 wives or more (number of children per woman the year prior to the census)
Source: Micro-data for rural Tambacounda and Kolda regions, 1988 Senegal census.
Regardless of their age, the fertility of monogamous wives declines as the age of their husband increases (Figure 2). Fertility is lower in polygynous than in monogamous unions, but the pattern of the curves is very similar for wives in monogamous and for first wives in polygynous unions with two wives, with the exception of first wives under 30 years whose age-specific rate is higher when the husband is older than 60 years than when he is 50-59 (Figure 3A). In contrast, the fertility of second wives does not vary much by age of the husband when he is younger than 60 when we control for age of wives. This result suggests that below 60 years, the influence of age of the husband on the fertility of second wives is small. Thus, the effect of the husband’s age is clearest when he has only one wife. In polygynous unions, the effect of rank confuses the picture.
Figures 4A to 4C describe age-specific fertility rates in unions with three or more wives
The small number of cases may account for the irregular.... Figure 5A to 5C show about the same information as previously but this time from the husband’s side. Age-specific fertility rates are always smaller for first wives than for wives of ranks 2 and 3 +. The impact of the age of the husband seems to differ by rank of wife, and the wives of higher order seem to be the “favourites” (Figure 5). Figure 4B shows that there is almost no difference between the fertility rates of second wives when the husband is younger than 60
For rank 3 +, age-specific fertility rates do not decrease....
Age-specific fertility rates by age of husband and wives and by rank of wife in polygynous unions with 3 or more wives (number of children per woman the year prior to the census)
Source: Micro-data for rural Tambacounda and Kolda regions, 1988 Senegal census.
It is difficult to conclude on the role of the husband’s age on this basis, beyond the obvious fact that younger men in polygynous unions with more than two wives appear generally more fertile than older men, but that a clear effect emerges only when the age of the husband exceeds 60. We turn to multivariate analysis to disentangle the complex effects of age of the partners, rank, and number of wives.
Table 7 presents tests of our five hypotheses in a compact way. The dependent variable is the occurrence of a birth from a wife; this variable is dichotomous. We applied three logistic regressions: one column of the table is devoted to all types of marriages together; comparable results are shown for monogamous unions and for polygynous unions. In the three models, the age of the woman and the wealth of the household (which has a significant positive impact on fertility) are control variables. The variable “squared age of wife” is meant to account for the curvature of the fertility distribution, and it improves the predictive power of the model. An odds ratio smaller than one indicates that the probability of the occurrence of a birth among the wives who have the characteristics described by the modality of a given independent variable is smaller than the category of reference, all other things being equal. For instance, when the number of wives of a polygynist increases by one unit, the probability of the occurence of a birth to one of the wives decreases. Inversely, when the rank of wife increases by one unit, the probability of the occurrence of a birth increases. We tested the effect of an interaction between number of wives and rank but it turned out not to be significant and did not improve the prediction of the model.
Table 7 –
Main results of three logistic regressions on occurrence of a birth. wives in all unions, in monogamous and in polygynous unions
Variables All unions Monogamous unions Polygynous unions Odds ratio (std. err.) Odds ratio (std. err.) Odds ratio (std. err.) Age of wife 1.23 (0.02)*** 1.22 (0.02)*** 1.24 (0.03)*** Squared age of wife 0.91 (0.00)*** 0.92 (0.00)*** 0.91 (0.00)*** Age of husband 20-39 (Ref.) 1.00 1.00 1.00 40-49 0.95 (0.02)** 0.93 (0.03)* 0.97 (0.02) 50-59 0.82 (0.02)*** 0.77 (0.03)*** 0.89 (0.03)*** 60+ 0.57 (0.02)*** 0.52 (0.03)*** 0.65 (0.03)*** Rank of wife(a) 1.10 (0.02)*** – 1.11 (0.02)*** Number of wives(b) 0.82 (0.01)*** – 0.75 (0.02)*** Birth to a co-wife Yes – – 1.78 (0.04)*** No (Ref.) – – 1.00 Presence of 1st wife over 50 Yes – – 1.15 (0.07)*** No (Ref.) – – 1.00 Standard of living(c) 1.06 (0.01)*** 1.04 (0.01)*** 1.08 (0.01)*** Number of cases 126,231 53,444 72,787 Pseudo R2 0.0813 0.0504 0.1053 -Log likelihood 60,560.36 29,030.71 31,185.29 (a) 1 for women in monogamous union and wives of rank 1 in polygynous union, 2 for wives of rank 2, and 3 for wives of rank 3 or higher. (b) The variable includes wives of monogamists and of polygynists with 2 wives, and 3 wives or more. (c) See Table 6. *** p<0.001; ** p<0.01; * p<0.05. Source: Micro-data for rural Tambacounda and Kolda regions, Senegal census 1988.
The first hypothesis, namely that the fertility of each wife decreases with the number of wives, is confirmed. So is the second hypothesis: the wife of highest rank, a relative newcomer in the marital unit and perhaps the “favourite” of the moment, is most likely to have given birth in the previous year. As for the age of the husband, it appears to have a stronger effect for monogamists than for polygynists; for the latter the difference is substantial only after age 50.
Finally, our last two hypotheses are not confirmed. Childbearing by one wife during the previous year increases the odds of a birth to a co-wife. And the presence of a senior wife has a slight effect that is significant only at the 5% level.
We look more closely at these relationships in Table 8, where we compute six logistic regressions by number and rank of the wives. In most cases, the odds of a birth from wives in monogamous or polygynous unions decrease significantly as the husband’s age increases, all other things being equal (including the wife’s age). This result confirms our third hypothesis. There are exceptions, however. The negative effect of husband’s age on the fertility of wives is unambiguous for monogamous wives and for the first wives of polygynists. For second or third wives, however, the effect is either not significant or positive before the husband has reached age 60.
Table 8 –
Odds ratios of a birth for wives in monogamous and polygynous unions. logistic regressions on the occurrence of a birth from wives of ranks 1, 2, 3 or higher
Variables Monogamous unions Polygynous unions 2 wives 3 wives or more Rank 1 Rank 2 Rank 1 Rank 2 Rank 3+ Odds ratio Odds ratio Odds ratio Odds ratio Odds ratio Odds ratio (std. err.) (std. err.) (std. err.) (std. err.) (std. err.) (std. err.) Wife’s age 1st wife 15-24 (Ref.) 1.00 1.00 – 1.00 – – 25-29 0.99 (0.03) 0.90 (0.04)* – 0.69 (0.11)* – – 30-34 0.82 (0.03)*** 0.73 (0.04)*** – 0.59 (0.10)** – – 35-39 0.58 (0.03)*** 0.56 (0.04)*** – 0.45 (0.08)*** – – 40-49 0.23 (0.01)*** 0.25 (0.02)*** – 0.18 (0.03)*** – – 50+(a) 0.17 (0.02)*** 0.05 (0.01)*** – 0.05 (0.02)*** – – 2nd wife 15-24 (Ref.) – – 1.00 – 1.00 – 25-29 – – 0.92 (0.04)* – 1.15 (0.12)* – 30-34 – – 0.72 (0.04)*** – 0.75 (0.08)** – 35-39 – – 0.49 (0.03)*** – 0.49 (0.06)*** – 40-49 – – 0.18 (0.02)*** – 0.25 (0.04)*** – 50+(a) – – 0.05 (0.01)*** – 0.05 (0.02)*** – 3rd wife or higher 15-24 (Ref.) – – – – – 1.00 25-29 – – – – – 0.99 (0.08) 30-34 – – – – – 0.59 (0.06)*** 35-39 – – – – – 0.50 (0.06)*** 40-49 – – – – – 0.13 (0.02)*** 50+(a) – – – – – 0.03 (0.01)*** Husband’s age 20-39 (Ref.) 1.00 1.00 1.00 1.00 1.00 1.00 40-49 0.93 (0.03)** 0.87 (0.04)** 1.09 (0.04)* 0.89 (0.10) 0.93 (0.10) 1.07 (0.11) 50-59 0.77 (0.03)*** 0.58 (0.04)*** 1.11 (0.05)* 0.62 (0.09)** 0.95 (0.10) 1.22 (0.13) 60+ 0.48 (0.03)*** 0.38 (0.04)*** 0.79 (0.05)** 0.31 (0.06)*** 0.70 (0.09)** 0.91 (0.11) Birth from the 1st wife – 1.00 1.69 (0.06)*** 1.00 2.53 (0.28)*** 1.52 (0.18)*** 2nd wife – 1.73 (0.06)*** 1.00 2.55 (0.28)*** 1.00 2.10 (0.16)*** 3rd or higher wife – – – 1.84 (0.18)*** 2.24 (0.17)*** 1.00 Presence of 1st wife over 50 Yes – – 0.97 (0.08) – 0.95 (0.13) 1.19 (0.13) No (Ref.) – – 1.00 – 1.00 1.00 Standard of living(b) 1.04 (0.01)*** 1.06 (0.02)*** 1.08 (0.01)*** 1.03 (0.04) 1.11 (0.04)** 1.12 (0.03)*** Pseudo R2 0.0486 0.1124 0.0648 0.1586 0.1160 0.0983 N 53,622 25,290 25,290 7,511 7,511 7,511 -Log likelihood 29,184.21 10,623.25 12,640.19 1,956.38 2,802.53 3,212.97 (a) Very few women older than 50 years have a birth. (b) See Table 6. *** p<0.001; ** p<0.01; * p<0.05. Source: Micro-data for rural Tambacounda and Kolda regions, Senegal census 1988.
While previous research (Garenne and van de Walle, 1989) suggested that the fertility of wives decreases systematically as the husband ages, our results show that the age of the husband has a positive small impact when he is younger than 60 for wives of the highest rank (second wife for a bigamist). Only after 60 does the physiological effect of age on male fertility manifest itself; the husband may have less interest in sexual relations.
We found that decreases in the odds of having a birth are larger for monogamous wives and first wives in polygynous unions when the husband ages (50% to 70% decrease of the odds of a birth when the husband is older than 60 years compared to when he is 20-39 years), than for second and third or higher order wives (reduction by 10% to 30% when husbands age from 20-39 years to over 60).
In polygynous marriages, the variable impact of the age of the husband and the significant role of rank on the fertility of wives suggests that the rotation for sexual intercourse among wives of all ranks is not regular. There may be some favouritism towards younger wives of higher ranks (who joined the union more recently) with whom the husband has sexual intercourse more often. Such a higher frequency could compensate the loss of fecundity as husbands get older. The significant decrease after 60 years may be due to his sub-fecundity and probably to less frequent coitus.
Table 8 also examines the impact of a birth to a co-wife on the fertility of a wife of a specific rank in the same union. The effects are highly significant in the aggregate. For instance, in unions with two wives, when the second wife has a child, the probability of a birth to the first wife is substantially increased. Similarly, Figure 6 shows that without any control for age of the husband, the odds of the first wife giving birth is higher when the second wife has a birth too.
First wife’s fertility rates by age and by 2nd wife’s occurrence of a birth in polygynous unions with 2 wives (number of children per woman the year prior to the census)
Source: Micro-data for rural Tambacounda and Kolda regions, 1988 Senegal census.
This unexpected finding seems to run counter the idea that there is a tendency for women in a polygynous marriage to alternate their births. Why then, a strong positive association between the probabilities of childbearing? The evidence is insufficient to suggest a convincing behavioural explanation. A simple reason might be that what these co-wives who have a birth during the same year have in common is a fecund husband, or simply one who was present during the year before the census, while none of the wives of a sterile or absent husband would have a child.
Lastly, the estimation of different models by number of wives and their rank within the union shows that the presence of a first wife older than 50 does not have any significant effect on the fertility of younger and higher-order co-wives (Table 8), while when polygynous unions are considered all together a slightly significant effect appears (Table 7). We know of no other study that has tested this effect, and the result is new.
The present study represented to some extent a methodological exercise. Censuses are the great under-exploited data source for Africa. Their content varies a great deal. The census of Senegal in 1988 contains extraordinarily detailed information on household structure and on nuptiality. It constitutes a first-rate source of information on polygyny, and could be used in this study to investigate the relationship between polygyny and fertility. In the rural areas selected for this study, the population is not strongly differentiated by income or education. A regime of natural fertility still prevails, and we assumed that the main intermediate factor influencing fertility was the frequency of intercourse. Because the census contains many more marital units than could ever be interviewed in a survey, it was possible to disentangle the various factors that characterize marriages: the ages of husband and wives, the number of wives living with their husband, and their rank. All these factors have a significant effect on fertility. The age of husband has a significant effect and the effect is clearest in monogamous couples. Wives of high rank tend to have a higher fertility than wives of low rank, after controlling for age and number of wives. This we interpret as evidence of “favouritism” by the husband, resulting in a higher frequency of intercourse with the latest arrival among his wives. A surprising result is that the probabilities of multiple wives having a birth during the same year are positively associated. Finally, the presence of first wives past the age of childbearing did not affect the fertility of their younger co-wives.