In many countries, but particularly in east and southern Africa, HIV has spread beyond small groups with high-risk behaviour into the general heterosexual population. Both theoretical  and empirical studies  have identified the rate of sexual partner acquisition to be an important determinant of the sexual transmission of HIV. This rate tends to be higher among sexually active young people and falls after marriage or the formation of other forms of stable partnership. Therefore, teenagers and young people have become an important focus for preventative campaigns and interventions and for research. Of course, risk behaviour does not stop with the advent of marriage or stable partnerships, particularly among men. In Europe and the USA, the proportions of men in a marriage or steady relationship who reported two or more sexual partners in the past year ranged from 5 to 10%[3,4]. In low-income countries, there is a wider variability in this measure, from less than 5% up to 50%.
This paper integrates the themes of premarital and extramarital sexual behaviour. It assesses the extent to which individuals display a consistency in their sexual conduct over the life course. In particular, the research question addressed here is whether characteristics of early sexual life, such as age at sexual debut and number of premarital partners, predict the propensity later in life to seek extramarital partners. Several considerations lead to an expectation of consistency in sexual life-styles. Either heterogeneity in the family environment and relationships before the initiation of sexual activity or the strength of the biological sex drive may leave a lasting imprint on sexuality that in turn influences both pre- and post-marital behaviour. Familial characteristics found to be associated with more active adolescent sexual activities include parental marital disruption [6–8], weaker measures of religious belief [9,10] and more favourable parental attitude to premarital sexual activity . Male sex hormones have been associated with sexual motivation in both sexes and in subsequent sexual debut in boys [12,13]. Genetic markers [14,15] and physical signs of pubertal development [16–18] have been associated with age at sexual debut in both sexes. To the extent that either of these causal pathways is valid, early sexual conduct is a marker for later conduct but not a determinant: both are an expression of a common pre-existing cause and are less amenable to change by intervention. A third possible set of influences carries very different causal implications. Rather than positing that propensities towards certain forms of sexual conduct are fixed at an early age and are relatively immutable, the assumption here is that attitudes towards sex and patterns of sexual interaction are learned during adolescence and that this process of sexual socialization has a profound effect throughout life. Therefore, interventions designed to modify behaviour in adolescents may not only reduce the exposure to risk during adolescence itself, but may also be protective in later life . Under this more sociological model it is likely that peer norms exert a major influence on the timing and nature of sexual expression [6,7,18], but which programmes are effective at promoting sexual behaviour change in adolescents is not yet evident .
Although the underlying mechanisms are still unclear, research studies in Europe and the USA show that there is a persistent link between age at sexual debut and sexual behaviour later in life. In particular, the 1992 French survey of sexual behaviour of adults (aged 18–69 years) found that a young age at first intercourse in men was associated with a higher mean lifetime number of partners, a higher mean number of partners in the past 12 months, and a higher mean rate of union dissolution and reformation . In the United States National Health and Social Life Survey of sexual behaviour, a multivariate analysis reported that individuals (aged 18–59 years) who were virgins at age 18 had a lower number of partners in the past 12 months compared with those who had had sexual intercourse by this age . In the UK Sexual Attitudes and Lifestyles Survey a multivariate analysis also found that an age at sexual debut below 16 years in men and women significantly predicted increased odds of two or more sex partners in the past 12 months and 10 or more lifetime partners . In the developing world, evidence from two studies in Tanzania suggested that younger age at debut is associated with a higher numbers of partners later in life [23,24].
The objective of this paper is to assess the nature and strength of the link between characteristics of early sexual conduct and the probability of extramarital sex later in life, drawing upon cross-sectional survey data on married men from four low-income sites.
Data and methods
In response to the HIV pandemic, the then Global Programme on AIDS launched a programme of large-scale surveys in 1988. Two main questionnaires were designed for administration to general populations. The more commonly used instrument focused on knowledge, attitudes and belief. The second instrument, the Partner Relations (PR) questionnaire, collected more detailed information about sexual behaviour. In particular, most of the PR surveys collected the following information about early sexual experience: age at first intercourse, whether the first partner became a marital partner, the duration of time that the first partner was known before intercourse occurred, the number of premarital partners, and age at first marriage. The definition of marriage was a broad one. It included regular partnerships that had lasted a year or longer, regardless of their legal, religious or traditional status. Respondents were also asked how many current spouses and regular partners they had. Later in the questionnaire, married individuals were asked whether they had experienced intercourse with anyone in the past 12 months apart from their spouse or regular partner and, if so, with how many different partners. These were defined as non-regular partnerships. This sequence of questions represents a valuable opportunity to assess the links between characteristics of early sexual experience and the probability of having extramarital partners in the recent past.
Four PR surveys out of the total of about nine such surveys were selected for the analysis on the following criteria: collection of key variables; availability of a clean, well documented file; and a minimum of 50 men who reported extramarital sexual contacts in the previous 12 months. Three nationally representative surveys – Côte d'Ivoire, Tanzania and Thailand – and one survey of a capital city, Lusaka, fulfilled these criteria. All four surveys used clustered probability samples, interview schedules in the appropriate local language, and face-to-face interviews conducted by specially trained staff. Fieldwork was conducted in 1989 or 1990. Full details of the methodology of the PR surveys have been published .
The analysis was restricted to men, because of evidence suggesting that women are more likely to under-report extramarital sex [3,4,26] and the low prevalence of extramarital sex evident in the datasets. The analysis further excluded men who had been married or in a regular union for less than one year to ensure that all respondents were fully exposed to a complete period of risk. After these exclusions, the effective sample sizes were 1028 (Côte d'Ivoire), 1085 (Tanzania), 649 (Lusaka) and 683 (Thailand).
All data were analysed using stata 5.0 . Bivariate analysis was used to assess the premarital predictors of having extramarital intercourse (EMI) in the past year. Odds ratios (OR) were calculated for the risks of having EMI compared with a base category. χ2 tests were used to test (at the 95% level) for heterogeneity in OR and linear trends in OR where appropriate. The distribution of the number of premarital partners was positively skewed in all sites. Therefore, the natural logarithm of the number of premarital partners was used to assess linear trends by the age of debut. Predictors of EMI were also assessed using multivariate logistic regression. Adjusted OR were calculated controlling for the premarital predictors and age. Other behavioural and sociodemographic variables were included if they significantly (at the 95% level) improved the logistic model (assessed using likelihood ratios tests). These other potential confounders were education, years lived in the locale, urban/rural residence, absence from home for more than a week in the past year, co-residence with wife, a polygamous partnership and whether the respondent defined the current relationship as a marriage or as a regular partnership.
Table 1 summarizes the background characteristics of the four study populations. Some schooling was universal in Thai men and least common in Côte d'Ivoire, where 35% of the sample had no education. The four samples had similar rates of immigration. Between 6 and 10% of respondents had immigrated in the past year. The entire Zambian sample was urban and was carried out in the capital city, Lusaka. Approximately 50% of Ivoirean and Tanzanian men in the sample live in an urban area, whereas only 26% of the Thai sample do. Absence from home for at least 4 weeks in the preceding year was only asked about in the survey in Côte d'Ivoire and Tanzania, where 59 and 29% of men, respectively, reported it. Cohabitation with their spouse was most common in Thailand (96%) and least common in Côte d'Ivoire (65%). The prevalence of formal or informal polygyny varied between 39 and 20% in the three African sites but was much lower in Thailand (3%). In Thailand and Tanzania, over 90% of men defined their main partnership as a marriage rather than as a less formalized union, whereas in Côte d'Ivoire and Lusaka between 60 and 70% defined themselves as married.
The period prevalence of EMI in the past year was highest in Côte d'Ivoire (34%). Approximately 25% of men in Tanzania and Lusaka reported EMI in the past year, whereas the prevalence of EMI was lowest in Thai men (16%).
Descriptive information about the circumstances of sexual debut and premarital sex is given in Table 2. The age category that includes the modal age of sexual debut was the same in all four sites at 15–19 years and the median age was also similar (17–18 years). There were pronounced variations in sexual debut before the age of 15 years. In Côte d'Ivoire and Lusaka, nearly 20% of married men reported early debut, compared with 9 and 4% in Tanzania and Thailand, respectively.
The length of acquaintance with the first sexual partner before intercourse occurred is mainly of interest because it reveals the proportion who first had sex with someone whom they had met that day. Such partners were no doubt composed largely of ‘casual pick-ups’ and sex workers. The relevant question was not asked in Lusaka. In the remaining three sites, this type of encounter was most common in Thailand (35%), a reflection of widespread prostitution, moderately common in Tanzania (19%) and least common in Côte d'Ivoire (10%).
The interval in years between sexual debut and marriage was estimated indirectly from answers to questions on age at debut and age at marriage. In all four sites, ages at debut and at marriage were within one year for 30–40% of men. Many of these men were probably virgins until marriage but, from existing data, it was impossible to confirm this. The median interval was similar in all sites, at 2–3 years. However, within each site there was considerable heterogeneity between men, with approximately a quarter of married men reporting an interval of 5–9 years and a further 11–19% an interval of 10 or more years.
Another indicator of the nature of the first sexual partnership was the proportion of men who married their debut partner. This was around a third of men in Tanzania, Lusaka and Thailand and approached half of the men in Côte d'Ivoire (44%).
The last premarital sexual experience variable to be examined was the number of premarital partners. This question was not asked of men who reported marrying their debut partner, on the assumption that the number was zero. This analysis followed the assumption, which was unlikely to be true for all men, but it is probable that this group of men had fewer premarital partners than those who did not report marriage to their debut partner. Heterogeneity within populations in premarital sexual experience was also evident in the number of premarital partners. Four to 10% reported one other premarital partner, whereas 24–46% reported five or more. Thai and Lusakan men were most likely to report many premarital partners (median of three). In the other sites, the median number of premarital partners was lower, two in Tanzania and none in Côte d'Ivoire.
Table 3 presents a bivariate analysis of the relationship between the characteristics of premarital sexual experience and the odds of EMI in the 12 months preceding the survey. In all four sites, a significant trend of reducing odds of EMI with increasing age at sexual debut was evident.
The length of previous acquaintance with the first sexual partner had a large and statistically significant association with EMI in Côte d'Ivoire and Thailand. As expected, men whose first sexual intercourse was with a stranger were much more likely to report EMI than other men. In Tanzania, the odds were reversed; however, this difference was not significant.
The trend in odds of EMI and the time interval between sexual debut and marriage was significant in Tanzania and Thailand. In both populations, the men who experienced sexual debut and marriage within the same year were less likely to report EMI than other men. In the other two sites, there was no evidence of any association.
There was a consistent, strong and statistically significant difference in extramarital sexual activity across all four sites between men who married their first sexual partner and those who did not. The odds of EMI in the latter groups compared with the former ranged from 1.87 in Côte d'Ivoire to 7.24 in Thailand.
The last factor to be considered in this preliminary analysis is the number of premarital sexual partners. In all sites a significant trend of increasing odds of EMI with an increasing number of premarital partners was found. The significantly higher odds of EMI associated with five or more premarital partners ranged from 2.5 in Côte d'Ivoire to 9.3 in Thailand.
These relationships were reassessed in a multivariate analysis. The number of predictive factors was first reduced from five to three. The variable ‘married debut partner’ was not included in the model because of its overlap with the number of premarital partners. The variable ‘time between debut and marriage’ was also omitted because of its strong correlation with the number of premarital partners. A clear trend in reducing the number of premarital partners with increasing age at debut was seen (Table 4), although it was decided to retain both factors, along with the length of acquaintance with the debut partner in the multivariate model. Age was also included in the final model along with the other potential confounders listed previously if they were found to improve the model significantly. The adjusted effects of premarital sexual characteristics on EMI are shown in Table 5.
In the bivariate analysis (Table 3), age at sexual debut emerged as a strong and significant predictor of EMI. After adjustment for the length of acquaintance with the debut partner, number of premarital partners, and the demographic and socioeconomic factors shown, the effects of age at debut attenuated, but did not disappear. In all the study populations, men who experienced later sexual debut at age 20 or more were less likely to have had recent extramarital partners than those whose debut occurred before 15 years of age. A significant trend in reducing the odds of EMI was evident in Côte d'Ivoire and Tanzania with increasing age at debut. A similar trend was apparent in Lusaka but was not statistically significant. In Thai men however, no trend was apparent.
After adjustment, the strong link among men in Côte d'Ivoire and Thailand between EMI and the length of acquaintance with the debut partner before intercourse occurred became less pronounced but retained significance. In these two sites, men whose sexual debut involved a stranger were more likely to report EMI than other men. However, in Tanzania, the unexpected relationship in the opposite direction became stronger and attained significance.
The adjusted association between the number of premarital partners and EMI was not essentially different from the unadjusted association, although the significant trend was lost in Côte d'Ivoire and Lusaka. Men reporting five or more such partners were significantly more likely than men with fewer partners to have engaged in recent extramarital sex in all populations.
Table 5 also summarizes the effects of the potential confounding variables. There is no evidence of significant trends in current age and EMI. Men in Côte d'Ivoire who are between 25 and 34 years old were found to be more likely to engage in EMI compared with younger men. Whereas in the other sites, as age increased the odds of EMI generally reduced, a result that reached significance in Lusaka.
Increasing education was found to be associated with higher odds of EMI in Côte d'Ivoire. In addition, those men in Côte d'Ivoire who defined their current main partnership as ‘regular’ rather than marital were three times more likely to report EMI. Co-residence with the main partner was associated with lower EMI in Tanzania and Thailand. Finally, polygamous Tanzanian men and men living in urban areas reported more EMI in the past 12 months.
This analysis, using survey data from four developing country populations, showed that characteristics of premarital conduct, in particular age at sexual debut, and number of premarital partners, were associated with the probability of extramarital sex later in life. Although statistically significant effects on EMI of these two characteristics of early sexual conduct were not found in all four populations, there was nevertheless an important consistency in the patterns of the results. Those men who postponed sexual debut until age 20 or more were less likely to report EMI than men with younger reported ages of debut in all sites. In addition, men with five or more premarital partners were significantly more likely to report EMI in all sites. Because of the relatively strong relationship between age at sexual debut and the number of premarital partners, it is not justified to attempt distinctive interpretations of the two relationships. More plausibly, both characteristics probably reflect the same underlying behavioural pattern of sexual precocity.
These findings, together with similar results in Europe, the USA and a different population in Tanzania [4,22–24,28], strongly suggest that these associations over the life course may be a feature of many, or even all, societies. They may also hold for women as well as men. The European and US analyses showed a strong link between precocious age at first intercourse and the number of partners later in life among women. In much of the developing world in general, the more strict social control on expressions of female sexuality make similar analyses, based on reported behaviour, more difficult than in Europe and North America, where attitudes are more liberal: indeed there is strong evidence that women under-report extramarital sex [29,30]. Other study designs, using individual in-depth interviews with open questions asked of a more limited number of women, may be more appropriate than standardized surveys to elicit reliable information on extramarital partners of women in low income settings.
Also among men, data on reported sexual behaviour are likely to suffer from desirability and memory biases. Although cross-sectional interview surveys on sexual behaviour will always be vulnerable to accusations of yielding poor quality data, it was reassuring that response rates on such sensitive topic in the four study sites were high (more than 92%) . The data were also internally consistent. Results of tests of internal consistency on the age at first sexual relations and the age at first regular partnership showed high consistency of 92–98% across the study sites. Similar checks, performed on data from the Demographic and Health Surveys, also suggest that data on the timing of first sex are quite robust [31,32]. The possibility that individual differences in the willingness to report both premarital and extramarital partnerships account for the statistical links cannot be excluded. However, as shown in this study, the relative strength and persistence of the associations throughout age groups in different settings makes this hypothesis unlikely.
Although an unusual amount of detail was obtained in the PR surveys about the circumstances of sexual debut, this analysis was nevertheless constrained by the lack of information on a number of topics that would have helped to understand more fully the nature of the associations. In particular, a greater understanding of the possible role of first sexual encounters in shaping later sexual careers would have been added by data on the age of the first partner, the place and other circumstances of first sex, whether the relationship continued and whether contraception, including condoms, was used.
The process of socialization and the nature of family life during adolescence may strongly condition other aspects of individual futures; for example, early marriage increased the probability of divorce . Here, early sexual initiation and multiple premarital partners may establish a pattern of sexual conduct that persists into later sexual lifestyle patterns. However, the validation of any such explanation would require very detailed prospective data and the alternative more biological and psychological explanations cannot be ruled out.
If adolescent socialization provides at least part of the explanation, what are the implications for HIV/sexually transmitted disease (STD) prevention? Delaying age at first sex is a strategy that many national programmes are already promoting for their young population, and this study provides additional arguments to support such efforts. This study suggests that effective behavioural change interventions in adolescents may be protective against HIV/STD not only in adolescence, but also throughout the life course. Attempting to reduce the proportion of young people who have first sexual relations below a certain age using school-based strategies is also an effective response adopted by some national programmes [34,35]. This strategy will probably not modify the associations that our study has shown, but by delaying the median age at first sex it will considerably reduce its importance at the population level as a predictor of risk behaviour for HIV/STD.
The authors gratefully acknowledge the following individuals: Mohamed Ali, Martine Collumbien, Roger Ingham, Louisiana Lush, Cicely Marston, Sara Thomas, Jerry Wheeler, Brent Wolff and Basia Zaba for valuable advice and comments on an earlier draft of this paper.
1. Garnett GP, Anderson RM. Strategies for limiting the spread of HIV in developing countries: conclusions based on studies of the transmission dynamics of the virus.
J Acquir Immune Defic Syndr Human Retrovirol 1995, 9: 500 –513.
2. World Bank. Strategic lessons from the epidemiology of HV, in confronting AIDS: public priorities in a global epidemic
. Oxford: Oxford University Press; 1997. pp. 53 –102.
3. Leridon H, van Zessen G, Hubert M. The Europeans and their sexual partners.
In:Sexual behaviour and HIV/AIDS in Europe
. Hubert M, Bajos N, Sandfort T (editors). London, UCL Press; 1998. pp. 165 –196.
4. Michael T, Butzer R, Feinleib J, Joyner K. The number of partners.
In:The social organization of sexuality – sexual practices in the United States
. Laumann E, et al
. (editors). Chicago: The University of Chicago Press; 1994.
5. Carael M. Sexual behaviour.
In:Sexual behaviour and AIDS in the developing world
. Cleland J, Ferry B (editors). London: World Health Organization; 1995.
6. Kiragu K, Zabin LS. The correlates of premarital sexual activity among school-age adolescents in Kenya.
Int Family Planning Perspect 1993, 19: 92 –97.
7. Murray NJ, Zabin LS, Toledo Dreves V, Luengo Charath X. Gender differences in factors influencing first intercourse among urban students in Chile.
Int Family Planning Perspect 1998, 24: 139 –144.
8. Newcomer S, Udry JR. Parental marital status effects on adolescent sexual behavior.
J Marriage Family 1987, 49: 235 –240.
9. Thornton A, Camburn D. Religious participation and adolescent sexual behaviour and attitudes.
J Marriage Family 1989, 51: 641 –653.
10. Paul C, Fitzjohn J, Eberhart-Philips J, Herbison P, Dickson N. Sexual abstinence at age 21 in New Zealand: the importance of religion.
Soc Sci Med 2000, 51: 1 –10.
11. Thornton A, Camburn D. The influence of the family on premarital sexual attitudes and behavior.
Demography 1987, 24: 323 –340.
12. Udry J, Billy JO, Morris NM, Groff TR, Raj MH. Serum androgenic hormones motivate sexual behavior in adolescent boys.
Fertil Steril 1985, 43: 90 –94.
13. Udry JR, Talbert LM, Morris NM. Biosocial foundations for adolescent female sexuality.
Demography 1986, 23: 217 –230.
14. Rodgers JL, Rowe DC, Buster M. Nature, nurture and first sexual intercourse in the USA: fitting behavioural genetic models to NLSY kinship data.
J Biosoc Sci 1999, 31: 29 –41.
15. Miller WB, Pasta DJ, MacMurray J, Chiu C, Wu H, Comings DE. Dopamine receptor genes are associated with age at first sexual intercourse.
J Biosoc Sci 1999, 31: 43 –54.
16. Zabin LS, Smith EA, Hirsch MB, Hardy JB. Ages of physical maturation and first intercourse in black teenage males and females.
Demography 1986, 23: 595 –605.
17. Udry JR. Age at menarche, at first intercourse, and at first pregnancy.
J Biosoc Sci 1979, 11: 433 –441.
18. Smith EA, Udry JR, Morris NM. Pubertal development and friends: a biosocial explanation of adolescent sexual behaviour.
J Health Soc Behav 1985, 26: 183 –192.
19. Mann J, Tarantola D, Netter T, Cohen M. Prevention.
In:AIDS in the world
. Mann J, Tarantola D, Netter T (editors). Cambridge, MA: Harvard University Press; 1992. pp. 325 –448.
20. Aggleton P, Rivers K. Interventions for adolescents. In: Preventing HIV infection in developing countries. Biomedical and behavioral approaches. Gibney L, DiClemente D, Vermund S (editors). New York: Klumer Academic/Plenum Publications; 1999. pp. 231 -255.
21. Bozon M. Reaching adult sexuality: first intercourse and its implications. From calendar to attitudes.
In:Sexuality and the social sciences. A French survey on sexual behaviour
, English ed. Dartmouth; 1996. pp. 143–175.
22. Johnson A, Wadsworth J. Heterosexual partnerships.
In:Sexual attitudes and lifestyles
. Johnson AM, et al
. (editors). Oxford: Blackwell Scientific Publications; 1994.
23. Mnyika KS, Klepp KI, Kvale G, Ole King'ori N. Determinants of high-risk sexual behaviour and condom use among adults in the Arusha region, Tanzania.
Int J STD AIDS 1997, 8: 176 –183.
24. Konings E, Blattner WA, Levin A. et al. Sexual behaviour survey in a rural area of northwest Tanzania.
AIDS 1994, 8: 987 –993.
25. Ferry B, Deheneffe J, Mamdani M, Ingham R. Characteristics of surveys and data quality.
In:Sexual behaviour and AIDS in the developing world
. Cleland J, Ferry B (editors). London: World Health Organization; 1995. pp. 11 –42.
26. Cleland J. Risk perception and behavioural change.
In:Sexual behaviour and AIDS in the developing world
. Cleland J, Ferry B (editors). London: World Health Organization; 1995. pp. 157 –192.
27. StataCorp. Stata Statistical Software: Release 5.0 User Manual
. Stata Press; 1997.
28. Bozon M, Leridon H. Sexuality and the social sciences. A French survey on sexual behaviour.
29. Adeokun L. Research on human sexuality in Pattern II countries.
In:Human sexuality: research perspectives in a world facing AIDS.
Chouinard A, Albert J (editors). Ottawa: IDRC; 1989.
30. Cleland J, Ferry B. Sexual behaviour and AIDS in the developing world.
In:Social aspects of AIDS
. Aggleton P (editor). World Health Organization; 1995.
31. Blanc A, Rutenberg A. Assessment of the quality of data on age at first sex, age at first marriage and age at first birth in Demographic and Health Surveys.
In:Assessment of DHS-1 data quality, DHS methodological reports.
Columbia, MD: Institute for Resource development; 1990. pp. 41 –82.
32. Meekers D. Sexual initiation and premarital childbearing in sub-Saharan Africa.
Population Studies 1994, 48: 47 –64.
33. Ying H. Patterns of divorce risk in the 1970s and 1980s for Swedish women with a gymnasium education.
Yearbook of Population Research in Finland 1996, 33: 44 –59.
34. Kirby D, Short L, Collins J. et al. School-based programs to reduce sexual risk behaviours: a review of effectiveness.
Publ Health Rep 1994, 109: 339 –360.
35. DiClemente RJ. Preventing HIV/AIDS among adolescents.
:Schools as agents of behavior change [Editorial].
JAMA 1993, 270: 760 –762.
Keywords:© 2000 Lippincott Williams & Wilkins, Inc.
adolescent; Africa; age at first sex; age at sexual debut; HIV/AIDS; non-regular sex; risk factors; sex behaviour; SE Asia