During the past decade, two demographic trends in developing countries have received considerable attention: the unprecedented size of the current cohort of adolescents and the steadily increasing percentage of women infected with HIV.1 Much of the acceleration in the spread of HIV among women has occurred among adolescents. In some parts of the world, notably Sub-Saharan Africa, the prevalence of HIV among women aged 15–24 is two to eight times that among men in the same age-group.2 In response, policymakers have increasingly sought to address the reproductive and sexual health needs of adolescents, particularly female adolescents. Most of the resulting policies and programs, however, pay surprisingly little attention to the large proportion of female adolescents who are married.

Child and adolescent marriage remains common in many parts of the developing world.3 Almost a third of the more than 330 million girls and young women aged 10–19 who currently live in developing countries (excluding China) were or will be married by their 18th birthday.4 In the majority of these countries, most of the sexual intercourse involving female adolescents occurs within marriage.5

Although the risk of HIV transmission between spouses is extremely low under certain conditions—specifically, when both partners are uninfected at the time of marriage and subsequently engage in sexual activity exclusively with each other—these conditions are often not met. When they are violated, as is frequently the case for women who marry at a very young age, sexual intercourse with a spouse is risky. Indeed, in some settings, married adolescents have higher rates of HIV infection than their sexually active unmarried peers.6 Thus, married female adolescents not only represent a sizable fraction of adolescents at risk of contracting HIV via heterosexual intercourse, but also a group with high rates of HIV infection.

Nonetheless, many policymakers and parents, and even young women themselves, continue to perceive marriage as a haven from the risk of HIV infection. Parents in Malawi, for example, encourage their daughters to marry early to protect them from HIV.7 Moreover, many international and national AIDS prevention messages encourage abstinence until marriage; these messages imply that sex within marriage is not only more socially sanctioned than premarital sex, but also somehow provides complete protection against HIV.

In this article, we present evidence from a series of international surveys about four important aspects of young women's lives. These data call into question the often deeply ingrained belief that marriage protects young women from HIV. First, for many adolescents—particularly the youngest brides—marriage greatly increases their potential exposure to the virus, because marriage results in a transition from virginity to frequent unprotected sex. Even among adolescents who are already sexually active, marriage generally leads to a dramatic rise in the frequency of unprotected intercourse, especially when pregnancy is desired. Second, the partners of married female adolescents are typically older—and, by virtue of their age, more likely to be HIV-positive—than the boyfriends of unmarried female adolescents. In addition, in polygamous societies, child and adolescent brides are more likely than older brides to be second or third wives. Third, women who marry as adolescents receive less formal education and have less exposure to the media than their unmarried peers, greatly reducing their opportunities to receive information about HIV/AIDS via these channels and potentially undermining their ability to negotiate safer sexual practices. Fourth, in addition to being mounted in venues that are relatively inaccessible to married adolescents, HIV outreach programs are often geared toward unmarried adolescents and other groups that are considered to be high-risk. These programs often promote protection strategies that are either inappropriate for married adolescents or difficult for them to implement, such as using condoms or abstaining from sexual activity.

DATA AND METHODS

The data in our analysis come from Demographic and Health Surveys (DHS) conducted between 1996 and 2004 in 22 African countries and seven countries in Latin America and the Caribbean. In each country, a nationally representative sample of women aged 15–49 was surveyed. The majority of surveys used the DHS+ version of the questionnaire, although an older version (DHS III) was used in five countries (South Africa, Niger, Togo, Côte d'Ivoire and Brazil). These instruments included identical questions about respondents' demographic and socioeconomic characteristics, marital status, recent sexual activity, pregnancy intentions and current contraceptive methods.* There were subtle differences between the DHS III and DHS+ in the questions regarding age at first sex, but these differences have little effect on reports of sexual activity.8

Our analyses focus on the sexual and social dynamics of 15–19-year-old adolescent females. The majority of our analyses rely on current status reports from married and sexually active unmarried young women in this age range; in a few analyses, we used data from an older cohort of women to avoid censoring younger adolescents who were not yet married or had not yet completed their schooling. All data are weighted using the probability sampling weights provided by DHS.

RESULTS

Rates of Early Marriage and HIV Status

Marriage before age 20 is common in Sub-Saharan Africa (Table 1). In all of the African countries we studied, with the notable exception of South Africa, at least four of every 10 women marry before reaching their 20th birthday. In Latin America and the Caribbean, between one-quarter and two-thirds of young women marry during their teenage years. Some of the national data mask considerable local variation; the proportions of young women married by age 18 in the Amhara region of Ethiopia (80%; data not shown) and in Kayes, Mali (83%), are substantially higher than the national rates for those countries (49% and 65%, respectively).

The HIV status of most adolescents in Africa and Latin America is unknown, although a handful of studies suggest that the prevalence of HIV among married adolescent females is relatively high. In some urban areas of Kenya and Zambia, for example, the prevalence of HIV infection among married adolescent females is 33% and 27%, respectively, whereas the prevalence among sexually active unmarried females in these areas is 22% and 17%.9 Nationally representative samples reveal similar patterns, albeit at lower infection rates. For example, recent DHS data show that the prevalence of HIV infection is greater among married female adolescents aged 15–19 than among their unmarried sexually active counterparts in countries such as Kenya (6.6% vs. 2.5%), Tanzania (3.1% vs. 2.5%) and Cameroon (3.9% vs. 2.0%), although these differences are statistically significant only for Kenya.

Sexual Activity and Pregnancy Intentions

Table 2 examines the relationship between marital status and recent unprotected sex among young women aged 15–19. In all countries except South Africa and Namibia, more than half of adolescent females who had had unprotected sex (i.e., sex without a condom) during the previous week were married; in 18 of the 29 countries, more than 80% were married. These findings reflect both a lower frequency of sexual activity and greater condom use among unmarried female adolescents than among married female adolescents. If unmarried adolescents are less likely than their married peers to report being sexually active, having sex frequently and having sex without a condom, then our data may overestimate the proportion of unprotected sexual activity that occurs within marriage; nonetheless, these numbers suggest that across tremendously diverse cultural settings, marriage remains the principal relationship context in which adolescent females have sex.

Table 2 also presents the proportions of married, unmarried and sexually active unmarried young women who had had unprotected sex in the previous week. That proportion was strikingly higher among married adolescents than among unmarried ones. Even when the cohort of unmarried respondents is restricted to those who were sexually active, the proportion of adolescents who had had recent unprotected sex was significantly higher among those who were married than among those who were not.

For many adolescent wives, the first year of marriage—when couples are often trying to conceive their first child—may be the most sexually active period in their entire life. Figure 1 (page 82) shows the frequency of unprotected sex by marital duration among women who married before age 20 in Burkina Faso, Malawi, Mali, Mozambique and Niger—the five Sub-Saharan countries with the highest proportion of females 18 or younger who are married (data for all 29 countries are available upon request). In many African nations, including most of those in Figure 1, the proportion of adolescent wives who had had unprotected sex during the previous week declined by 10–20 percentage points between the first year of marriage and the second. Much of this decline may be attributable to these women becoming pregnant with or giving birth to their first child.

Differences in pregnancy desires may partially explain the differences in the frequency of unprotected sex between married and unmarried adolescents. To investigate this possibility, we examined the proportion of adolescents who had had recent unprotected sex by pregnancy desires and marital status (Table 3, page 83). Although the desire to become pregnant was more common among adolescents who were married than among those who were unmarried but sexually active, it does not fully account for the differences in frequency of unprotected sex: In nearly every country, married adolescents were significantly more likely than sexually active unmarried adolescents to have had unprotected sex in the previous week, regardless of pregnancy intentions. These differences were generally greater among women who did not wish to become pregnant, particularly in South and East Africa and Latin America, but even among those actively seeking pregnancy in the next two years, married adolescents were, on average, about three times§ as likely as unmarried adolescents to have had unprotected sex in the past week (data not shown).

Characteristics of Sexual Partners

One reason that married adolescents are usually thought to have a lower risk of HIV infection than their unmarried peers is that married adolescents are believed to have fewer sexual partners. Contrary to expectations, however, in most countries the mean number of partners reported by married adolescents exceeded the number reported by sexually experienced unmarried adolescents (Table 4, page 83). There was, however, considerable variation within countries in the number of sexual partners reported by unmarried, sexually experienced adolescents; the majority reported having had one sexual partner in the past year, but up to a quarter were engaging in "secondary abstinence" (i.e., they had refrained from sex for more than a year), and a small but potentially quite important proportion reported having had two or more partners.

This last subgroup of unmarried young women may be at considerable risk, as recent research suggests that having multiple partnerships, particularly concurrent ones, is strongly associated with HIV infection.10 In all countries with available data except Kenya and Namibia, unmarried female adolescents were more likely than their married counterparts to report having had two or more sexual partners within the past year; the proportion of unmarried respondents with multiple partners ranged from fewer than 1% (Kenya) to almost 19% (Ethiopia).

Despite the lower proportions of married adolescents having multiple partners, several factors call into question the relative "safety" of husbands as opposed to boyfriends.11 First, husbands are, on average, older than boyfriends. Second, DHS data indicate that in most countries, the age gap between spouses was larger if the woman married before age 18 (Table 5, page 84). For these women, the average spousal age difference ranged from 4.7 years in Guatemala to 14.2 years in Guinea. The age difference was typically a year or two smaller if the wife was 18 or older when the marriage began.

Polygamy is widely practiced in most of Sub-Saharan Africa: In recent cohorts, 25–43% of married women aged 20–24 in western Africa and roughly 10–15% of those in eastern and southern Africa were in polygamous marriages.12 The DHS data reveal that in all countries in Sub-Saharan Africa except Namibia and South Africa (where adolescent marriage is uncommon), women who married before age 18 were more likely than older brides to have polygamous husbands (Table 5); in many of these countries, the difference was substantial.

Restricted Access to Education and Information

In addition to differing from their unmarried peers with respect to sexual behaviors and partner characteristics, married young women also spend less time in school and have less exposure to the media. As Table 6 (page 85) shows, a woman's age at first marriage is positively related to her total years of schooling; in all 29 countries, women who married when they were 18 or older had more education than those who married at a younger age. The starkest difference was in Nigeria, where women who were at least 18 when they married achieved, on average, 9.3 years of schooling, while those who married before they were 18 remained in school for only 2.5 years. Differences in educational attainment by age at first marriage were evident both in countries with low levels of overall education, such as Burkina Faso, Ethiopia and Mali, and in countries with higher levels of education, such as South Africa, Peru and Zimbabwe.

Although schools are one of the primary venues for HIV education, another route by which messages concerning reproductive health can reach targeted populations is the media; once again, however, women who married at an early age were at a striking disadvantage. In all countries with available data, married adolescents were significantly less likely than unmarried adolescents to watch television; in all but one, they were less likely to read the newspaper weekly (Table 6). Differences with respect to radio exposure were less dramatic but still significant in more than half of the countries.

Lack of Appropriate HIV Prevention Strategies

Perhaps because of this limited access to information about HIV from educators and the media, married female adolescents were less likely than sexually experienced unmarried female adolescents to know one or more ways of avoiding HIV (Table 7, page 86). Even when compared with all unmarried adolescents, many of whom had not yet had sex, married adolescents were often poorly informed.

Moreover, there appear to be few protective strategies available and accessible to adolescent wives. In all 14 countries for which data are available, married adolescents reported either doing nothing or limiting their sexual activity to one partner (their husband) as their primary strategy for avoiding HIV infection (data not shown). Neither of these strategies offers an effective means of reducing the degree of risk. Although sexually experienced unmarried young women, like their married counterparts, tended to have only one sexual partner, in several countries they reported using condoms or abstaining from intercourse as their most common protective strategy. Among all unmarried young women, the most popular protective strategies in the overwhelming majority of countries were either to abstain from sex or to not initiate sexual activity. These options are, for the most part, not available to young women who are married.

The possibility of refusing sexual relations within marriage is virtually eliminated in many instances by the threat of violence. Forty-three percent of married adolescent women in Ethiopia and 64% of those in Mali stated they felt that husbands are justified in beating their wives if the women withhold sex (results not shown). Coerced sex, particularly coerced first sex, has been linked to poor reproductive health outcomes for young women, including increased risk of HIV infection.13

DISCUSSION

We identified four associations that point to the need for HIV prevention programs that focus on married adolescents. First, marriage coincides with intensified exposure to unprotected sex. For a large fraction of young brides in developing countries, marriage marks the transition from no sexual exposure to regular sexual relations. Even for those who have engaged in premarital sex, marriage may result in a dramatic increase in the frequency of unprotected sexual activity. These changes may be driven in part by the desire for pregnancy.

Second, the husbands of adolescent wives tend to be much older than their spouses. Because of their older age, these men are generally more sexually experienced than the boyfriends of unmarried adolescents; as a result, husbands are, at least in some settings, more likely than boyfriends to be HIV-positive. In one study, for example, the male partners of adolescent wives were more likely than the partners of unmarried adolescents to be infected with HIV in Kisumu, Kenya (30% vs. 12%), and Ndola, Zambia (32% vs. 17%).14

Nonetheless, it is possible that husbands are less likely than boyfriends to transmit the virus.15 This argument rests on findings that viral loads and infectivity rates are quite high in the first few months following infection and then typically decline until symptoms of AIDS appear.16 If HIV-positive boyfriends, due to their young age, have been infected more recently than HIV-positive husbands, then one might expect the infectivity of boyfriends to be higher, on average, than that of husbands. No reliable data are available to support or refute this speculation. We note, however, that the prevalence of HIV infection continues to rise for men in many countries until age 35,17 suggesting that a large number of new infections occur among men in their late 20s and early 30s—an age-group that includes the husbands of many adolescents. Clearly, the relationship between infectivity rates and the age of the male partner warrants additional research.

Another potentially important demographic finding is that the absolute age difference between spouses tends to be larger when women marry before age 18. In general, the evidence linking large spousal age differences and greater power imbalances within marriage is weak and sometimes conflicting.18 Nonetheless, a recent National Academy of Science panel on adolescents in developing countries concluded that "there is reason to believe that marriages of young women and older men are less equitable" than other marriages.19 If young women married to much older husbands have less power in the relationship, then they may have less ability to negotiate strategies to protect against HIV or to influence their husbands' behavior, whether his use of condoms within the marriage, his engagement in extramarital sex or his use of condoms with nonmarital sexual partners. In addition, in Sub-Saharan Africa, young brides are more likely than older brides to enter into polygamous unions and hence share their sexual risks with their husbands' other wives.

Third, adolescent wives have less access than their unmarried counterparts to social and public sources of information and support.20 Our findings suggest that married young women spend less time in school—the primary setting for HIV programs—than do unmarried young women. Similarly, married adolescents are less likely to be exposed to the mass media, another important source of messages designed to help curb the spread of HIV. Eliminating these gaps in access to HIV prevention messages is of vital importance. Moreover, in patrilocal societies, where females leave their natal villages to live with their husbands, adolescent wives' contact with former friends and family members may be severely curtailed. In Amhara, Ethiopia, for example, 15% of married female adolescents reported having visited with same-sex friends outside the home in the previous week, compared with 24% of unmarried female adolescents.21 This lack of social contact may be important, as emerging research suggests that the advice and experiences of friends and family can have strong, positive effects on the adoption of reliable HIV prevention strategies.22

Finally, married young women who are cognizant of HIV risks often rely on remaining faithful to their husbands—and hoping that their husbands remain faithful to them—as their only viable protection strategy. Alternative or back-up strategies that are used by unmarried adolescents, such as refraining from sexual activity or using condoms, are rarely considered by or even thought feasible for married adolescents. Currently, recommended strategies for protection and risk reduction are to abstain from sexual activity, to reduce the frequency of sexual activity, to change to a safer partner, to use a condom, to know one's own and one's partner's HIV status and to maintain a mutually monogamous relationship with an uninfected partner. Younger brides may be willing and indeed anxious to know their own and their partner's HIV status, but they may face substantial difficulties obtaining their partner's compliance with testing and mutual disclosure of results. In addition, although they may be highly motivated to remain faithful, their ability to influence their husband's sexual behavior might be minimal.

Limitations

Although many of our analyses show consistent and significant differences between married and sexually active unmarried adolescents, we emphasize that these are associations rather than causal relationships. Being married at a young age is clearly associated with engaging in frequent unprotected sex, having older sexual partners and having less exposure to sources of information, but these results do not prove that getting married causes these differences. One possible interpretation is that adolescents with these characteristics—for example, those with limited educational opportunities—are more likely to choose or be pressured into early marriage. Alternatively, perhaps these associations are driven by both causal and selection effects.

Measurement error, particularly for self-reported data, may also bias our results. For example, whether an adolescent reports herself as being married depends on her interpretation of marriage, which in many cultures is a complex, sometimes fluid arrangement open to considerable ambiguity and variation. Despite our reliance on self-reported data, we found clear differences in sexual behaviors, age of partners, pregnancy intentions, and other social and economic attributes associated with marital status.23

Accurate reporting of sexual behavior, especially among adolescents, is notoriously unreliable.24 Reporting errors may vary systematically by whether or not the respondent is married. For example, if unmarried young women are more likely than married ones to underreport their sexual behaviors, then the differences between the two groups will be overestimated. In Cameroon, Kenya and Tanzania, 1–2% of the unmarried adolescents who claimed never to have had sex were HIV-positive. Although some of these young women may have been infected from sources other than sex, the magnitude of these numbers suggests that some unmarried respondents were reluctant to disclose their sexual activity. Similarly, differences in the number of sexual partners may be underestimated if married adolescents were less likely than unmarried adolescents to disclose multiple partnerships. Comparisons between married and unmarried young women should be interpreted in light of these potential biases.

Finally, both current status and recall data are subject to bias.25 Current status measures of adolescents do not take into account that married adolescents tend to be older than unmarried ones, even within the 15–19-year-old age-group. Recall bias regarding age at first intercourse and age of first marriage in this age-group is probably minimal, however, as both events are likely to have occurred recently.

Policy Implications

Unmarried young women with multiple partners have received the lion's share of attention in HIV prevention programs and continue to require very specific and targeted strategies. On the other hand, married adolescents—despite facing significant social and behavioral risks—have been marginalized by adolescent HIV/AIDS policies and have not been central in reproductive health programs aimed at adult married women.26 Indeed, in many studies, married female adolescents have been portrayed as a low-risk group. Providing effective policy and programmatic interventions for these young women may prove particularly challenging. Yet, overcoming these challenges is important for at least two reasons. First, helping these young wives to avoid becoming infected could serve as a critical firewall, preventing the shift of the HIV epidemic from a concentrated to a generalized one. In developing countries where HIV is currently prevalent only among sex workers and their clients, the wives of these clients are frequently the next group to be infected. Second, although epidemiologic models of HIV transmission rarely identify young married women as important vectors, most of these women are at the beginning of their most intensive childbearing period; thus, protecting them could greatly reduce mother-to-child transmission of HIV as well as dramatically lower the probability of dual-parent orphanhood for their children.

Prescribing specific policy or programmatic interventions is beyond the scope of this paper. Each country (or region within a country) will need to assess the degree and specific sources of HIV risk for young married women.** Our analyses suggest that in nearly every country we examined, including several with mature AIDS epidemics, married adolescents may face higher risks than unmarried adolescents via greater exposure to unprotected intercourse, large age differences with their partners and limited access to information. There are two notable exceptions to this general pattern: South Africa and Namibia. In these two countries, fewer than 50% of sexually experienced adolescent females are married and levels of polygamy are relatively low. Thus, in these countries, the needs of adolescent wives may not be as great or as pressing as the needs of unmarried, sexually active young women. In contrast, in some countries not included in our study, the risk of HIV infection via early marriage may be particularly high. In India, for example, the low level of premarital sexual activity and the high proportion of adolescent females who are married suggest that adolescent sexual activity occurs primarily within marriage.27 Married monogamous women are considered to be at low risk. However, one study in urban India found that 14% of married monogamous women whose only source of risk was their husband were HIV-positive.28

In short, specific policy interventions must be tailored to sexual and behavioral profiles of young women in individual countries. The potential effect of increasing young women's age at marriage on HIV risk would depend on how the sexual behaviors of these young women (and their partners) would change. Some policymakers fear that delaying marriage would be accompanied by an increase in sexual activity among unmarried adolescents that would more than offset the decline in sexual activity among married adolescents. However, data from Latin America and Africa show that although the context in which adolescent sexual relationships occur has changed over time (from within marriage to before marriage), the overall percentage of sexually active young women has remained constant, or even declined slightly, as the age of marriage has risen.29

In countries where child and adolescent marriage is common and premarital sexual activity among young women is strongly curtailed, delaying marriage by itself might not result in a longer interval between sexual debut and marriage and might actually delay the often involuntary loss of virginity for many young women. Furthermore, in these countries, delaying the age at marriage until at least 18 might afford these women the opportunity to extend their education, develop livelihood skills and strengthen their self-esteem, all of which may help them create healthier and more stable marital unions when they wed.

In other countries, making sexual intercourse within marriage safer may be a more important strategy. The process of getting married can itself be used as an opportunity to reach both brides and grooms with information about HIV/AIDS and to provide HIV testing. Both religious and state institutions could incorporate these elements into the marriage process. Reducing spousal age differences in countries where men marry much younger brides may also help minimize risks associated with marriage, as intergenerational sex is believed to play an important role in perpetuating the HIV epidemic in many countries.30 Although the large age differences between unmarried young women and their older, wealthier sexual partners (sometimes referred to as their "sugar daddies"31) have received much media and research attention in parts of Sub-Saharan Africa, in reality the age difference between adolescent wives and their husbands is, on average, much larger than that between unmarried young women and their sexual partners.32

Finally, our analyses suggest that the first year of marriage is a period of unusually intense sexual activity for many women. Targeted outreach programs for the recently married have been proposed.33 Although the use of condoms within marriage is still relatively unpopular,34 it could be encouraged as a contraceptive and protective strategy for newly wed couples seeking to delay childbearing. Helping recently married couples solidify their unions and adjust to living together could also help reduce the number of women in one of the highest risk categories, those who have previously been married.35

The combination of policies designed to delay marriage until at least age 18 and of policies that recognize and lower HIV risks within marriage could be instrumental in making marriage safer. As with unmarried young women, some subgroups of married adolescents may be at much higher risk than others. Adolescents married to uninfected partners who remain sexually exclusive face minimal risks, whereas those married at young ages to much older men who have or have had multiple partners are acutely vulnerable—especially if these young women are trying to become pregnant. Policymakers have long sought to reduce adolescent HIV risks by discouraging risky premarital and extramarital sexual activity; it is time to also promote concurrent and complementary strategies to reduce adolescent marital sex and foster safer marital sex practices.