The fertility transition began just twenty years ago in sub-Saharan Africa, with large differences in timing between countries. In the pioneering countries of southern and eastern Africa (Botswana, Kenya and Zimbabwe), fertility is now below 4 children per woman, whereas it remains above 6.5 children per woman in the landlocked countries of the Sahel, where fertility has started to decline in urban areas only. Thus, sub-Saharan Africa remains the region with the highest fertility in the world: 5.4 children per woman versus 2.6 in Asia and Latin America, and 1.6 in industrialized countries (United Nations, 2003).

Sub-Saharan Africa also holds the record for age difference between spouses. The difference between mean ages at first marriage [1] averages 6.6 years in western Africa, 6 years in central Africa, 4.3 years in eastern Africa and 3.3 years in southern Africa. Despite a general trend downwards (Hertrich, 2001; Hertrich and Locoh, 1999), the age difference between spouses remains significantly higher in sub-Saharan Africa (5.4 years on average) than in the rest of the world, where it is around 3 years (3.2 years in Asia, 2.9 years in Latin America, 2.8 years in Europe, and 2.3 years in North America) (United Nations, 2000).

It seems justified to assume a link between high fertility and the large age differences between spouses that prevail in Africa. The literature interprets age difference between spouses as an indicator of inequality between the sexes and of the nature of the relationship between spouses. Conversely, improvements in women’s position and changing relations between spouses are among the most frequently cited factors in the fertility transition.

This article attempts to test the existence of an association between the age differences between spouses and contraceptive practice, on the basis of a body of recent Demographic and Health Surveys (DHS) conducted in 18 countries of continental sub-Saharan Africa.

The first section discusses potential links between age difference between spouses and reproductive behaviour, and sets forth some hypotheses. The next section presents the data and assesses their quality before describing the structure of age differences and producing regional models. The final section of the paper analyses the statistical association between the age difference between spouses and contraceptive practice after controlling for the effect of individual socioeconomic characteristics, and assesses the influence of the regional context (dominant mating model and diffusion of contraception) on this relationship.

This marriage system is in itself a key determinant of the high level of fertility in Africa. By ensuring that women spend most of their fecund lives in union, it maximizes the period during which they are exposed to the risk of pregnancy. Therefore, as in other regions of the world (Chesnais, 1986), an increase in age at first marriage has often been a key factor in the fertility decline of the African countries where it has begun.

In recent decades, nuptiality patterns in Africa have been marked by a general increase in the age of women at first marriage and a narrowing of the age difference between the sexes at first marriage [4] (Hertrich, 2001; Tabutin and Schoumaker, 2004). Polygyny remains a resilient institution, but recent data suggest that it has begun to decline in some countries and in advantaged social groups (Antoine and Pilon 1998; Antoine, 2002; Hertrich, 2003). Although these developments have been concomitant with a diversification of nuptiality regimes in sub-Saharan Africa, the region still stands apart from the rest of the world because change has been slower there than elsewhere (Tabutin and Schoumaker, 2004).

Although the distinctive characteristics of sub-Saharan nuptiality regimes are becoming less marked, they have persisted, and this may probably be attributed in part to their interdependence (a change in one feature has an impact on the whole system) and to marriage control mechanisms. Traditional marriage procedures required material resources (bridewealth) and symbolic features (codified rituals, the involvement of intermediaries, extension over time) that were not within the couple’s reach and required the intervention of elders (Radcliffe-Brown, 1953; Meekers, 1992; Hertrich, 1996). By controlling the allocation of girls and the access to wives, heads of families also secured the economic subordination of young men, who were dependent on their elders to marry (Meillassoux, 1982; Locoh, 2002). Thus, polygyny, early marriage of girls, prolonged celibacy of men, and large age differences all appear to be the manifestations of a structure of domination based on sex and age, and of a social organization that favours the extended family over the conjugal unit. The age difference between spouses may be linked to fertility not only as a feature of the nuptiality regime, but also as an indicator of the woman’s and the couple’s degree of autonomy.

Alongside other nuptiality variables (particularly age at marriage and polygyny), age difference between the spouses is frequently cited as a factor in the status of women, an indicator of the degree of inequality within the couple, but also of the role and position of women in society (Cain, 1984, 1993; Casterline et al., 1986; Hertrich and Locoh, 1999; Oppenheim Mason, 1984, 1993, 1997; Safilios-Rotschild, 1982; Thiriat, 2000; Véron, 1997; Ware, 1981).

By reinforcing gender domination with age domination (Bozon, 1991, 2003), the age difference establishes an inequality of status within the couple, as the wife is doubly subordinated to her husband, both as a woman and as his junior. The age difference also has an indirect effect on women’s autonomy through the associated components of the nuptiality regime. Because women spend almost their entire adult lives in a union, because of early first marriage and expedited remarriage procedures, they are restricted to their role as wives and mothers. This also signifies a lack of recognition of a personal space for women outside male tutelage. The premarital period, which encourages the development of individual projects and the acquisition of job skills and status outside the domestic sphere, is truncated for women. Life for widowed and divorced women outside the framework of marriage is restricted (social disapproval, obstacles to economic independence). In this context, by having many children, a woman not only gains social recognition, but also consolidates her status within the family space in relation to her spouse, her co-wives and her husband’s family. Large age differences can thus be linked to the former “uncontrolled” fertility model, because they contribute to limiting women’s opportunities for personal initiative and depriving them of an identity model other than that of wife and mother.

At the individual level, a large age difference generates distance between the spouses, first because age is fundamental to establishing a relationship of subordination, and second because it creates a generational and cultural gap between spouses who are at different stages in their life cycle and who had different experiences of pre-marital life. In populations where large age differences are the rule, these objective constraints of the marital bond are reinforced by other marriage practices. When marriage is the result of family strategies, couples are not formed or defined by joint decisions of the partners, especially with respect to fertility. Moreover polygyny creates a climate of distrust between the spouses not only in polygynous unions, where “the dependence and submission of the co-wives [toward their husband] are reinforced by the competition and inequality between them” (Antoine, 2002), but also in monogamous marriages, because some men can use the potential arrival of a co-wife as “a threat to ensure their wives’ subordination” (Hertrich and Locoh, 1999). The economic and family organization (separate budgets, division of labour, child fostering within the kinship network…) does not encourage the development of common interests or the establishment of a unit of conjugal decision-making where independent fertility decisions can be reached and norms can be discussed. Large age differences are an element in an institutional system that impedes conjugal intimacy and contributes to reducing the couple to a unit of biological reproduction. The denial of the couple as a decision-making unit appears to be a societal choice (Caldwell, 1982; Lesthaeghe, 1980; Ryder, 1983).

The view that increased couple autonomy facilitates the fertility transition is a matter of controversy. It was hinted at during the debates on modernization in the 1960s (Goode, 1963; Parsons, 1960, 1968), and developed by John Caldwell (1982) [6] in his theory of the reversal of wealth flows between generations. The theory considers the emergence of an independent conjugal unit, bound by emotional ties, as a preliminary to a decline in fertility. This nuclearization theory was widely criticized in the 1980s and 1990s as ethnocentric and evolutionist, and because it lacked scientific verification (Bartiaux and Tabutin, 1984; Locoh, 1988; McDonald, 1992; Segalen, 1988; Vimard, 1993). However, the conjugality hypothesis has gained new currency in the past fifteen years in research on decision-making processes with respect to contraception. Karen Oppenheim Mason (1993) maintains that stronger emotional bonds between spouses are likely to encourage family planning in two ways: through improved communication between the spouses and through a greater convergence of interests [7]. Several studies have shown that discussion of family planning issues among spouses is associated with more frequent contraceptive use (Oheneba-Sakyi and Takyi, 1997; for a summary, see Beckman, 1983) [8]. More recent research on the reproductive expectations of both sexes has found that women accept that their husbands play an important, if not predominant, role in fertility decisions (Watkins et al., 1997; Dodoo et al., 1997; Isiugo-Abanihe, 1994), and that decisions about family size taken jointly by the spouses influence the probability of contraceptive use (Andro and Hertrich, 2001; Bankole et al., 1995; Becker, 1996; Dodoo, 1995; Dodoo et al, 1997; Ezeh, 1991, 1993; Lloyd, 1996; Lasee and Becker, 1997; Phillips et al., 1997).

In addition to this general hypothesis, the statistical regularity of the relationship and the mediating effect of context need to be assessed.

However, differences in SMAM only reflect part of the age differences between spouses, which depend not only on the difference in the ages at which men and women first marry, but also on the frequency of marriage termination and remarriage (which can vary with the age difference between spouses) and especially on mating patterns (marriages between single people account for only a portion of first marriages between individuals). In addition, SMAM is an indicator derived from aggregate data, and does not provide any information about the distribution of age differences and therefore does not lend itself to analyses at the individual level.

The only individual analyses of age differences between the spouses are by John Casterline and his team (Casterline et al., 1986), based on World Fertility Surveys conducted during the late 1970s in 28 developing countries (9 in Africa, of which 5 in sub-Saharan Africa, 11 in Asia and 8 in Latin America). Mead Cain (1984, 1993) used those results on age difference as a marker of the status of women and the strength of patriarchy, and linked them to fertility levels. His research indicates the existence of an important correlation, for these 28 countries, between the median age difference and the fertility level. That correlation is mainly the result of a contrast between two groups of countries: the countries of Latin America and Southeast Asia (with small age differences and low fertility) on the one hand, and the countries of Africa and southern and western Asia (with large age differences and high fertility) on the other. The analyses thus distinguished between countries where transition was already ongoing and those where it had not yet begun. Within each of the two groups, there was no clear link between age difference between spouses and the fertility level. The situation may have changed since the start of fertility decline in Africa. It is now possible to re-examine the question using more recent DHS data.

The question on age of the spouse was addressed only to women in union and refers only to their partner at the time of the survey. The DHS use a broad definition of marriage or union, based on the respondents’ declarations rather than on criteria of legal or customary recognition. This approach is justifiable in sub-Saharan Africa, where marriage procedures vary and are often lengthy and complex (Radcliffe-Brown, 1953; Meekers, 1992; Hertrich, 1996) and where many unions go unrecorded. It should be noted, in particular, that women who do not live with their spouses are included among women in union and that they were asked the question about their spouse along with other married women. This paper deals with women in union in general and uses the terms “marriage”, “couple”, “husband”, “spouse” and “partner” regardless of whether the union is recognized formally or not, and whether or not the partners live together.

The fact that the age of the spouse is known for the current union only is an important limitation. The changing pattern of marital unions over different cohorts cannot be examined, since women from different cohorts were at different stages of their married lives at the time of the survey. This limitation is particularly important because marital mobility is high. One-third of women aged 45-49 had been married more than once, compared with 7% of women aged 20-24.

The relationship between changes in age difference between spouses and the fertility transition can therefore only be investigated on the basis of a comparative analysis of cross-sectional indicators. However, this approach is facilitated by the diversity of the countries surveyed, in terms of fertility as well as of age difference between spouses. A wide range of stages in the fertility transition are represented, from countries where modern contraception is practically non-existent (Chad) to countries where it is widespread, such as Zimbabwe, Gabon and Kenya, where more than half the women have already used a modern method of contraception. There is also a wide variety of situations in terms of age difference between spouses, with averages ranging from 6 years in Malawi to 14 years in Guinea.

The reliability of the data on age differences nevertheless warrants attention, especially as this indicator is based on two statements. It is calculated as the difference between the age (in years) of the interviewed woman and of her spouse. The interviewer records the woman’s date of birth (and then calculates her age as the difference with the date of the interview) if she knows it, or if not, her age. The age of the spouse is based on a direct question to his wife. To assess the quality of the data on age difference, we will therefore examine the quality of the data on the ages of the two spouses, in terms of completeness and regularity of their distribution.

The data on age are available for all women and for almost all their spouses (Table 1). Only two countries (Guinea and Mozambique) report more than 3% of missing data on the husband’s age. However, this high degree of completeness does not mean that the data are accurate and reliable. In 12 of the 18 countries covered, more than 40% of the women cannot specify their month of birth. In six countries, most of them in West Africa, this is true of more than 80% of the women.

Table 1
Completeness of data on age and age heaping (women in union, weighted totals)

The distribution of the ages reported by the respondents for themselves, and even more for their spouses, confirms the inaccuracy of the data in many countries. An over-representation of “rounded” ages, ending in 5 or 0, is observed in most countries. If all the statements were accurate, those ages would account for around 20% of statements; yet the proportion recorded is below 25% for the woman’s age in only six countries and for the man’s age in only three countries (Table 1). Age heaping is observed in all ten west African countries, with a concentration of rounded ages usually higher than 30% for both sexes, and even exceeding 40% for the man’s age in four countries. In the other regions, where women are more aware of their age, attraction for rounded ages is non-existent or low (below 30%) in all countries except Ethiopia. The quality of data on men’s age is doubtful in five countries, particularly Ethiopia, Kenya and Tanzania, where more than 30% of the ages recorded are multiples of 5. However, even though these distortions seem high, they are comparatively lower than in the World Fertility Survey and the first DHS, in which values ending in 0 or 5 in some cases represented more than two-thirds of the data entries; the situation was even worse with census data (Goldman et al., 1985; IRD, 1990).

Though examining the age distribution only addresses one aspect of the errors in the responses on age, it is sufficient to show that reported ages cannot be considered as accurate quantitative values. Women are not aware of the exact age of their spouses, and the errors in their responses are probably not random. Age is recognized as an attribute of male power and analyses of census data show that over-reporting of age among men is higher than among women above age 50 [11]. It is therefore likely that the husbands’ ages – and consequently the age differences between the spouses – are overestimated. Countries of the Sahel, which show some of the most marked patterns of age distortion, also show the most heaping with respect to age differences (Table 1). It is therefore possible that the respondents (or the interviewers) tended to estimate the age of the men on the basis of the expected age difference between spouses. However, it is reassuring to note that there is much less distortion in the distribution of age differences than in the distribution of ages.

The inaccuracy of age reporting suggests that individuals see age from a qualitative rather than a quantitative viewpoint. For this reason, the age difference lacks the necessary accuracy to serve as a continuous variable in this paper. The information will be used mainly to distinguish between two contrasting marital situations:

These two categories have deliberately been given “extreme” limits to ensure homogeneity despite the errors in estimated age differences. To take other situations into account as well, four categories will be considered in order to analyse the relationship between age difference and reproductive behaviour:

The analysis examines the relation between these four types of union and reproductive behaviour. The probability that women will adopt new reproductive behaviour is measured in terms of whether they “have ever used” or “currently use” modern contraception. The aim is not so much to measure contraceptive use per se, as to identify the women who, by practising birth control, can be considered as demographic innovators. The analysis therefore had to include women who had used modern contraception in the past but were not using it at the time of the survey, either because they were not at risk of conceiving (pregnant women, breastfeeding women, sterile women, or women whose husband is absent), or because they had not yet reached the desired number of children (particularly the younger women).

Before turning to the results of the analysis, we present an overview of the structure of age differences in the 18 countries.

The main differences between the regions lie in the modal values and dispersion, but they nevertheless have similar profiles. The curves are asymmetrical, with low representation of negative values and a concentration of values around the mode, and are skewed to the right. Everything suggests that mating is impossible unless a minimum age difference is respected. As a result, unions where the age difference between the spouses is much smaller than the modal value are marginal, and even more so when the woman is older than her husband. The concentration around modal values reinforces the hypothesis of the social control of mating. It is also likely that these observed distributions exaggerate the real situation, as the norm may have influenced statements.

In all countries, the age differences are larger and more dispersed for remarried women than for women in a first union (Figure 3). Of particular note, marital situations in which the woman is older than her husband become less exceptional. In countries outside the Sahel, they concern almost one remarried woman in ten on average, compared with fewer than 3% of women in first marriages. In the Sahel, however, this type of age difference remains rare (3% of remarried women and less than 1% of women married for the first time). The variation in age difference between first and subsequent marriages can be attributed partly to the fact that women enter union early in Africa, so a first marriage to a younger man is unlikely. This type of match becomes accessible as a woman’s age increases, and is therefore more common in the case of remarriage. Furthermore, since women’s remarriages are generally less controlled than their first marriages (Locoh and Thiriat, 1995), mating patterns are more flexible and therefore more diverse. This is confirmed by the higher representation of large age differences in cases of remarriage. Since first marriages account for most unions (8 in 10 on average in each group of countries), this category reproduces the regional differentials observed for all unions (largest age differences in the Sahel, lowest in southern, central and eastern Africa, and intermediate in the Gulf of Guinea countries). By contrast, with respect to age differences at remarriage, the Sahel group stands out, and the countries of the Gulf of Guinea have the same profile as southern, central and eastern Africa.

Polygyny is another factor underlying the greater dispersion of age differences, but always in favour of larger values. By definition, polygynous marriages involve men who are already married, and consequently older on average than unmarried men (most of whom were never married) [13]. It is therefore logical for them to have a larger average age difference with respect to their wives. This relationship may also reflect a selection effect of individuals least involved in building a conjugal unit, a position that would account for both the practice of polygyny and the large age difference. The practice of polygyny varies strongly between regions. It involves on average 41% of the women surveyed in the Sahel, 33% in the Gulf of Guinea, and 20% in other regions. However, the regional differentials in age differences observed for all unions cannot be attributed to the uneven representation of polygyny, since they are found in both the monogamous and polygynous sub-groups.

The relationship observed at the regional level also appears at the individual level. In every country analysed, women married to much older men are less engaged in new reproductive behaviour. For women from the 18 countries taken together, when the age difference between spouses is small, the likelihood of ever having used modern contraception is 2.4 times higher than when the age difference is large, and around one-third higher than when the age difference is moderate. The effect of age difference is both strong and highly significant (below the threshold of 1%). It is observed in all the countries in the study, but to varying degrees. For example, the ratio of the use of modern contraception between the extreme groups (small age differences and age differences of 15 years or more) ranges from 1.1 to 1.4 in 11 countries, from 1.5 to 1.8 in 5 countries, and exceeds 2.5 in two countries (Appendix III, Figure 4). Contraceptive use by couples with an intermediate age difference (5 to 14 years) is generally closer to that of couples with a small age difference than to that of couples with a large age difference. There is no regional pattern associated with differentials in contraceptive practice between categories of age differences.

Table 2
Logistic regression odds ratios for the determinants of the use of a modern method of contraception for 18 countries of sub-Saharan Africa

The reference category for the models in the multivariate analysis was that of women in a moderately non-egalitarian union (the largest group, comprising 53% of women), aged 25-29, with fewer than four surviving children, uneducated, living in a rural area, without a defined religion or animist, in a polygynous union and married for the first time.

Table 2 gives odds ratios rather than coefficients to facilitate interpretation of the results. A ratio significantly above one indicates a positive relationship between the independent variable and the indicator tested; a ratio significantly lower than one indicates a negative relationship between the two variables.

The results confirm the effect of age difference between spouses on women’s contraceptive practice. As expected, the women’s socioeconomic characteristics play an important role in the relationship, though the effect of age difference remains highly significant and in line with the expected pattern, even after controlling for confounding variables. The probability of having ever used modern contraception is higher when the age difference between spouses is small (excluding the atypical group of women at least 5 years older than their husbands). For women in egalitarian couples, the probability of using contraception is higher than for women in the strongly non-egalitarian group (OR = 1.095, versus 0.794), and slightly higher than for women in the reference intermediate group.

Except for marriage rank, all the variables entered into the analysis have a significant effect on the probability of using contraception. In line with the convergent results of studies on reproductive behaviour in developing countries, the woman’s level of education has by far the largest impact. The probability of having ever used modern contraception increases steadily and rapidly with the level of education, reaching a ratio of 1 to almost 12 between women with no education and women who have reached secondary education. Schooling is an indicator of both social background and woman’s status. It fosters the adoption of new fertility ideals through access to more diversified sources of information and to opportunities for personal fulfilment outside the roles of wife and mother. Within the couple and the extended family, education gives a woman more power to negotiate or impose innovative behaviour. After the level of education, the place of residence, another indicator of social background, has the next biggest impact on contraceptive practice. Urban dwellers are much more likely to use modern contraception than women living in rural areas (OR = 1.19, versus 1), and this can also be attributed to differentials in terms of diffusion of information, social control, and access to and use of health and family planning services. The “religion” variable mainly sets Christian women apart, with a higher probability of using contraception than women who practise traditional religion or say they have no religion. Muslim women do not differ significantly from the latter.

The polygynous or monogamous nature of the union, which may be considered as another indicator of the nature of the marital relationship, has a similar (slightly smaller) effect to the age difference between spouses: being in a monogamous couple significantly increases the probability of using contraception.

Lastly, the demographic variables relative to the woman’s position in the life cycle have a significant effect. The probability of using contraception increases logically with the number of surviving children. Women who have at least five surviving children are more than two-thirds more likely to use contraception than women with fewer offspring. Age has a complex effect because it reflects both period and cohort effects. As a result of the period effect, contraceptive use increases with age. The period of exposure to the probability of using contraception, and the probability of having reached the desired family size increases with age. Meanwhile, as a result of the cohort effect, contraceptive use is lower at older ages: the youngest cohorts have grown up in an environment more open to innovation and have been more exposed than their seniors to campaigns promoting contraception. The combination of period and cohort effects explains why the odds ratios are below one for all age groups compared with the reference category of women aged 25-29. The probability of using contraception is significantly lower than the reference category because of the period effect for women under 25 and because of the cohort effect for women aged 35 and over.

Mating patterns, in terms of the age difference between spouses, appear to be a factor in the adoption of new reproductive behaviour. For women with the same socioeconomic and cultural characteristics, the closer they are in age to their spouses, the greater the likelihood that they will use modern contraception. This effect is only partly absorbed by related indicators, namely the woman’s level of education and the monogamous or polygynous nature of the union. Thus the age difference between spouses has a clear independent effect on contraceptive use.

The analysis of contraceptive practice based on individual independent variables has its limitations because it overlooks the context in which individual behaviour occurs. In other words, women are assumed to act in a neutral space, influenced solely by their individual socioeconomic attributes. However, the psychological and social costs of using modern contraception are probably not the same in an environment where contraception has become common and thus enjoys a degree of social legitimacy, as in an environment where contraception is unusual and represents a transgression of social and moral rules. Similarly, being married to a man close in age probably does not have the same significance or the same implications in a society where large age differences are the norm and where unions of this type constitute in themselves new behaviour, as in a society where small age differences between spouses are common.

The variable describing the dominant type of mating pattern in the region was defined on the basis of the proportion of women married to a man at least 15 years her senior among all the women from the region interviewed in the individual survey. Three regional groups were defined: a first group where less than 10% of women have an age difference of 15 years or more with their spouses; a second group where the proportion is between 10 and 25%, and a third group where it is more than 25%. The variable describing the diffusion of contraception in the region was defined on the basis of the proportion of women surveyed who have ever used a modern method of contraception. Three groups were also defined for that variable: less than 15%, 15 to 34%, and 35% or more of women who have ever used modern contraception.

These contextual variables were calculated at the level of the regions defined in the Demographic and Health Surveys, i.e. 4 to 12 subdivisions depending on the country. These regions are too large to ensure cultural homogeneity, which would have been the ideal situation for our purpose. In the absence of more precise information about the geographical location of the women, it was not possible to make a more detailed breakdown.

The reference category for Model 2 is defined in the same way as for Model 1, with the added components of residence in a region where modern contraception was not widespread (less than 15% of current or ever use) and where couples with large age differences are rare (less than 10% of the couples).

Interpreting the effect of age difference between spouses on contraceptive practice is more complex in this model because it integrates both the age difference coefficient and the coefficient measuring the interaction between the age difference and the contextual variables. The effect of age difference between spouses on the probability of using modern contraception is estimated by multiplying those two coefficients. For example, Model 2 can be used to compare two women with the same individual characteristics, but who differ in terms of the age difference with their husbands (5-14 years for the first woman; and more than 15 years for the second woman) and the regional conjugal model (fewer than 10% of strongly non-egalitarian couples in the region where the first woman lives; and more than 25% in the region where the second woman lives). The first woman belongs to the reference category, with an odds ratio equal to 1. For the second woman, this odds ratio is 0.87 (0.731 multiplied by 1.19), i.e. lower than that of the first woman.

Adding contextual variables significantly increases the share of explained variance, as confirmed by the likelihood ratio, already highly significant in the previous model. The results of the second model are consistent with those of the first (Table 2) in that the variables relating to individual socioeconomic characteristics generally operate in the same direction in the two regressions, but their effect is either weakened (level of education, place of residence, type of union) or reinforced (woman’s age, number of surviving children) when regional indicators are taken into account. The effect of religion is no longer significant after controlling for the contextual variables.

Among the contextual variables, as might have been expected, the extent of diffusion of modern contraception in the region has a decisive effect on the individual practice of contraception. For the same individual characteristics, the probability of ever having used a modern method varies in a ratio of 1 to 5 depending on whether the woman resides in a region where use of modern contraception is rare (less than 15% of women have ever used or currently use) or in a region where it is widespread (at least 35% of women). By contrast, the dominant age difference category does not affect the probability of women’s use of contraception, after controlling for individual characteristics and diffusion of contraception in the region.

By contrast, the dominant marital model in the region has a significant influence on the relationship between age difference between spouses and contraceptive practice. When women live in an environment where the non-egalitarian model is common (at least one-quarter of couples), the probability that they use contraception is closer to that of other couples (OR = 0.830 versus 0.731), probably because the category of non-egalitarian couples is a more heterogeneous group in that case.

The comparative advantage of women who are less than 5 years younger than their husbands over that of women who are 5 to 15 years younger is only significant in regions where contraceptive use is well established (i.e. at least 15% of women have ever used it). Only in that context do women in age-egalitarian couples have a higher probability of using contraception than women who are between 5 and 15 years younger than their husbands. Contrary to the initial assumption, therefore, couples with a small age difference do not have a pioneering position in the onset of the fertility transition. When contraceptive use is only in its earliest stages, the couples most egalitarian in age do not have a significantly higher probability of adopting contraception than couples with a moderate age difference (i.e. with an age difference of between 5 and 15 years). However, when contraceptive use spreads, couples close in age do show a higher probability of using contraception, and even more so when contraceptive practice intensifies in the area where they live. When between 15 and 35% of women in the region use contraception, couples with a small age difference are more likely to use contraception than other couples, and this probability is even higher in regions where contraception is very widespread (i.e. used by at least 35% of women).

By contrast, the other contextual variable entered into our analysis, namely the dominant mating model in the region, has no influence on the probability of age-egalitarian couples using contraception compared with couples having a moderate age difference. The coefficients measuring the interaction between age difference and the proportion of highly non-egalitarian couples do not have a significant impact on the relationship between the two variables of interest for the comparison of the two types of couple.

The results strongly confirm this hypothesis: the smaller the age difference between spouses, the higher the probability of using modern contraception. This relationship is partly influenced by individual socioeconomic characteristics, but it remains significant even after controlling for these characteristics, which include the level of education, a key indicator of woman’s status, and the monogamous or polygynous nature of the marriage, another indicator of intimacy.

Large age differences between spouses appear to hinder contraceptive practice, probably because they are not conducive either to individual decision-making by the woman who is in a subordinate position, or to the elaboration of shared conjugal projects by persons who are at different stages in their life cycles and who do not share the same generational culture. The difference in contraceptive use between unevenly matched couples and other couples does not decrease with the diffusion of contraception, and the differential increases when only a small proportion of couples have a large age difference, probably because this category includes the couples most attached to traditional values in the area of reproductive and conjugal behaviour.

This selection effect might have been apparent for couples with a small age difference through higher contraceptive use when they represent a minority category of the population and in a context of low diffusion of contraception, with the conjugal unit showing the way to acceptance of the psychological and social costs of a practice that has not yet acquired social legitimacy. However, the analysis disproves this. On the one hand, the probability that age-egalitarian couples use contraception is independent of their share in the total population. On the other hand, they stand out most strongly from the other categories of couple when contraception becomes widespread in the region. They are not pioneers, but adopt contraception faster when it begins to spread in a significant way. Being in such a couple is therefore not a driving force, but a factor accelerating the new reproductive behaviour.

Our data do not shed light on the mechanisms behind higher contraceptive use by women who are close in age to their husbands. Is it because these women enjoy more independence in family-planning decisions, or because these marriages favour joint decisions by the couple about family size? Rather than conflicting, the two processes can be considered as representing pathways towards access to contraception for women in age-egalitarian couples, pathways that do not exist or are unusual for women married to men much older than themselves.

Distribution (%) of age differences between spouses⁽¹⁾

Distribution (%) of age differences between spouses by monogamous or polygynous status of the marriage and marriage rank (first or subsequent marriage) for the woman, by region⁽¹⁾

Proportion (%) of women who have ever used modern contraception according to the age difference between spouses

Multivariate logistic regressions of the age difference between spouses on socio-economic variables (18 countries). Odds ratio

Distribution (%) of women according to the variables entered into the multivariate analysis (18 countries)