Social Science and Medicine 52 (2001) 733–744 Relationship dynamics and teenage pregnancy in South Africa Rachel Jewkesa,*, Caesar Vunduleb, Fidelia Maforahc, Esmé Jordaand a Women’s Health Research Unit, Medical Research Council, Private Bag X385, Pretoria 0001, South Africa b Department of Health, Mmabatho, North West Province, South Africa c Department of Welfare, Pretoria, Gauteng, South Africa d Bio-statistics Unit, Medical Research Council, Tygerberg, W. Cape, South Africa Abstract Teenage pregnancy is extremely common in South Africa. Whilst its ‘problematic’ nature is a subject of debate, it reflects a pattern of sexual activity which puts teenagers at risk of HIV. Currently one in five pregnant teenagers is infected with the virus. This creates a new imperative to understand teenage pregnancy and the pattern of high risk sexual activity of which it is one consequence. This was an exploratory study undertaken to investigate factors associated with teenage pregnancy amongst sexually active adolescents in an urban and peri-urban context. The study used a matched case control design, with 191 cases and 353 school or neighbourhood, age-matched controls. Subjects were under 19 years and recruited from township areas of Cape Town. A structured questionnaire was used to obtain information on socio-economic factors, contraceptive knowledge and use, and sexual behaviour. Conditional logistic regression was used to analyse the relationship between teenage pregnancy and the factors investigated. The results presented focus on relationship dynamics and their association with the risk of pregnancy. Both groups of teenagers had been dating for a mean of two and a half years and about half were still with their first sexual partner. The partners of the pregnant teenagers were significantly older, less likely to be in school and less likely to have other girlfriends. The pregnant teenagers were significantly more likely to have experienced forced sexual initiation and were beaten more often. They were much less likely to have confronted their boyfriend when they discovered he had other girlfriends. Multiple modelling shows that both forced sexual initiation and unwillingness to confront an unfaithful partner are strongly associated with pregnancy and also related to each other. We argue that the associations are mediated through unequal power relations within the relationship which are reinforced by violence. We further discuss indicators of greater intimacy within relationships of the pregnant teenagers which may suggest that more of the pregnancies were wanted than was suggested. Both of these conclusions pose critical challenges for health promoters. # 2001 Elsevier Science Ltd. All rights reserved. Keywords: Sexual health; South Africa; Teenage pregnancy; Violence against women Introduction In South Africa the teenage years are characterised by generally socially sanctioned freedom and sexual experimentation for both genders, but particularly for young men (Wood & Jewkes, 1998). For most of the country, marriage is relatively late, pre-marital sexual activity is *Corresponding author. Tel.: +27-12-339-8525; fax: +2712-339-8582. E-mail address: rjewkes@mrc.ac.za (R. Jewkes). the norm and pre-marital childbearing and impregnation are socially accepted (Makiwane, 1998). Although only 3% of women under 20 years are married or live with a partner, 35% have been pregnant or have a child (Department of Health, 1999b). Teenage pregnancies commonly occur in women who are still at school. In the discourses of bio-medical professions, civil servants of the Departments of Health, Welfare and Education and public expressions of the broader community, teenage pregnancy is viewed as a major problem (e.g. Department of Health, 1998; Ncayiyana & 0277-9536/01/$ - see front matter # 2001 Elsevier Science Ltd. All rights reserved. PII: S 0 2 7 7 - 9 5 3 6 ( 0 0 ) 0 0 1 7 7 - 5 734 R. Jewkes et al. / Social Science and Medicine 52 (2001) 733–744 Ter Haar, 1989; Nash, 1990; Preston-Whyte, Zondi, Mavundla & Gumede, 1990). Concerns voiced centre around its impact on the mothers’ schooling, subsequent employment and earnings, financial position of her family’s household (maintenance from the father is rarely forthcoming), impact on the child of being born into relatively greater poverty, and obstetric complications (although currently there is evidence that adolescence per se confers no increased obstetric risk (Varga, 1998; Ncayiyana & Ter Haar, 1989)). These concerns are often underpinned (sometimes explicitly) by a notion that in this predominantly Christian country, teenage sexual activity is essentially immoral (Wood, Maepa & Jewkes, 1997). Alternative views of teenage pregnancy are usually only voiced in more private settings, sometimes by the same people who at other times articulate the more familiar discourse. In-depth research has revealed that many teenagers are encouraged to become pregnant by their partners to prove their love, womanhood and fertility (Preston-Whyte et al., 1990; Varga & Makubalo, 1996; Wood et al., 1997; Richter, 1996), encouraged by grandmothers to produce a baby for the home, and mothers often indicate that teenage pregnancy is infinitely preferable to the possibility of infertility1 caused by contraceptive use (Wood et al., 1997). Adolescent2 childbirth has become institutionalised and is a fairly typical stage in the domestic life cycle of families. The baby is usually accepted into the mothers’ family, given the protection of her ancestors, looked after by elder women and the mother is often able to return to school.3 The coexistence of these two competing strands of discourse, sometimes within the same family or even individual, can perhaps best be understood if wantedness of pregnancy is imagined as positioned on a continuum of degrees of wantedness, with movement possible at different times and stages of pregnancy (and after the birth), rather than placed in the static binary categories which are often assumed by researchers. Although the competing meanings of teenage pregnancy leave considerable scope for debate about its ‘problematic’ nature, the escalating HIV epidemic provides an additional dimension to bio-medical concerns. It creates a new imperative to understand teenage 1 This is widely perceived by women and family planning nurses to be a side-effect of progesterone based injectable contraceptives, particularly Depo Provera. 2 The terms ‘‘adolescent’’ and ‘‘teenage’’ are used interchangeably in this paper. 3 The National Education Policy Act (1996) prohibits expulsion of a student because of pregnancy or refusal to take back students after the birth. Both of these have been common practice. pregnancy and the high risk sexual activity of which it is a consequence. The 1998 anonymous sero-prevalence survey of antenatal attenders showed that 21% of 15–19 year olds were infected with HIV (Department of Health, 1999a). Teenage sexuality and pregnancy have been the subject of considerable research in South Africa. For the most part this has been descriptive, taking the form of knowledge attitude and practices (KAP) studies of adolescents and pregnant teenagers (e.g. National Progressive Primary Health Care Network, 1996; Kau, 1988, 1991; Flisher, Ziervogel, Chalton, Leger & Robertson, 1993; Buga, Amoko & Ncayiyana, 1996) or qualitative studies of the circumstances of adolescent sexuality (e.g. Varga & Makubalo, 1996; Wood, Maforah & Jewkes, 1998; Wood & Jewkes, 1998). There is clinical research focusing on obstetric risks (Ncayiyana & Ter Haar, 1989), and research on health service use (e.g. Wood et al., 1997). Analysis of the relationships between possible risk factors and adolescent fertility has been notably missing. In an effort to fill this gap a case control study of factors associated with teenage pregnancy in Cape Town was undertaken. This paper presents data and discusses the dynamics of adolescent sexual relationships, and explores associations between these and adolescent pregnancy. Discussion of the meaning of other risk factors, particularly social and economic factors and sexual knowledge is presented elsewhere (Vundule, Maforah, Jewkes & Jordaan, 2000). Background Sexual activity starts in the mid-teens, often following closely on menarche, and teenagers, particularly boys, have multiple sexual partners (e.g. Naidoo, Williams, Knight & Bernstein, 1991; Sai, Rees & McGarry, 1993). Buga et al. (1996) interviewed 1072 school girls in the rural Eastern Cape (mean age 15.3 years) and found a mean age of sexual initiation of 14.9 years amongst the 76% of female respondents who were sexually active. Only 72% had reached menarche. Richter (1996), in her cross-sectional household study of 16–20 year olds living in Soweto, Umlazi and Khayelitsha, found 66% of respondents reported having ever had sex, with a mean age for sexual initiation of 16.4 years. Although teenage sexual activity and pregnancy are usually described as being on the increase (e.g. Chimere-Dan, 1995), very similar patterns of sexual and reproductive behaviour to those seen today are described in older urban ethnographies (e.g. Longmore, 1959; Pauw, 1963). The South Africa Demographic and Health Survey 1998 actually shows a reduction in the teenage birth rate (15–19 years) since 1980 (Department of Health, forthcoming). Contraception is rarely used at sexual initiation but most young women use it at some stage. The South R. Jewkes et al. / Social Science and Medicine 52 (2001) 733–744 Africa Demographic and Health Survey 1998 (Department of Health, 1999b) found that 64.4% of sexually active 15–19 year olds reported current use of a modern contraceptive method, this was rather higher than previous estimates which were around 25% (e.g. Buga et al., 1996) and is probably explained by the rather stringent definition of ‘sexually active’, which is taken as having intercourse in the four weeks prior to the interview. The pattern of contraceptive use was as expected, 76% were using injectable progesterones, 14% oral contraceptives, 6% used condoms for contraception. Oral contraceptives are not widely promoted to adolescents as they are regarded by nurses (and teenagers) as infinitely ‘forgettable’ (Wood et al., 1997; Mathai, 1997). Male opposition to contraceptive use is common and necessitates secrecy (Maforah et al., 1997). Richter (1996) found that fear of losing a partner was the most important barrier to contraceptive use. Injectable contraceptives are often preferred as they require no user involvement except attending for bi- or tri-monthly injections and can be secret. Unfortunately they often cause amenorrhoea, which is a particular problem in a cultural context where menstruation is perceived to be essential for bodily cleanliness (Jewkes & Wood, 1999). As a result sexually active teenagers commonly take ‘contraceptive breaks’ in order to see menstruation and may at this time fall pregnant. Health systems research has shown that nurses in family planning clinics often provide little information, method choice or discussion of side-effects, and women of all ages are frequently scolded by them (Wood et al., 1997; Mathai, 1997; Abdool-Karim, Preston-Whyte & Abdool-Karim, 1992). Since 1997 termination of pregnancy has been available on demand in the first trimester, but the South Africa Demographic and Health Survey 1998 shows that a year later only 53% of women knew this (Department of Health, 1999b). There is considerable misinformation about sexual health matters amongst young people (Boult & Cunningham, 1991; Richter, 1996; Varga & Makubalo, 1996). Sexual health education in the form of life skills have only recently been introduced as a compulsory part of the school curriculum but implementation is slow. Communication about sexual matters within families is very limited and teenage girls’ predominant sources of information are from the peers, boyfriends, and teachers (Richter, 1996) . Wood et al. (1998) found that male partners would often take advantage of communication gaps to emphasise notions of female availability and male sexual entitlement. Peers were not found to discuss intimate details of their sex lives and so the opportunities for alternative constructions of love and sex were limited. Within relationships communication in general between partners is often poor and discussion of sexrelated matters is usually only found in very close relationships (Varga, 1997). 735 South Africa is a very patriarchal society and as a result of its violent political history, the use of physical violence as a first line strategy to gain or keep a position of ascendance or resolve conflict is common (Simpson, 1991). In schools, there is a widespread problem of violence and lack of discipline, which involves both students and teachers. The 1998 South African Demographic and Health Survey found that school teachers were the most common perpetrators of rape of young girls (Department of Health, forthcoming). It is not surprising that violence is frequently found within dating relationships (Jewkes, Penn-Kekana, Levin, Ratsaka & Schrieber, 1999; Abrahams, Jewkes & Laubsher, 1999). Research has shown that young women are subject to assault (ranging from slapping to beatings with objects and stabbings) and sexual coercion (on a spectrum from begging to gang rape) by partners and others. For the most part these are deployed to ensure sexual availability, to discourage or punish infidelity, to assert control over the commencement and ending of relationships, and discourage attempts to undermine the boyfriends’ sexual success with other women (Wood & Jewkes, forthcoming; Wood & Jewkes, 1998; Varga & Makubalo, 1996; Varga, 1997). Taken on face value it can be difficult to understand why young people tolerate dysfunctional and abusive relationships. Obviously the pleasure derived from sex and islands of love in the relationships are part of the answer, but important too are the broader meanings which these have in their lives. Most young people grow up in townships or rural areas where there are few recreational facilities and opportunities for advancement are limited. Prospects of employment after leaving school are limited as adult unemployment is currently about 34% (Statistics South Africa, 1998). In this context substantial efforts are invested in the few available and affordable opportunities for entertainment and arenas where success may be achieved and selfesteem gained. Sexual relationships are one of these and acquiring and keeping the ‘right’ partners becomes a central preoccupation for both young men and women and an important arena in which respect among peers may be won or lost (Wood & Jewkes, forthcoming). For young women, in addition, there are opportunities for material gain, and sex is often used as a currency for exchange, particularly with older men (Wood & Jewkes, 1998). Methods A matched case-control study design was used to ensure an adequate number of pregnant teenagers and to control for age, which was anticipated to be the most important confounder. Cases were African adolescents 736 R. Jewkes et al. / Social Science and Medicine 52 (2001) 733–744 under 19 years who booked at two antenatal clinics in African townships and informal settlements in Cape Town namely Gugulethu and Khayelitsha between July and November 1995. Controls were adolescents who were not pregnant, said they had never been pregnant and were matched for age and school, if the case was still in school at the time of conception, or matched for neighbourhood, if the case was not in school. Ethical approval was granted by the Medical Research Council’s Ethics Committee. All the study subjects were recruited by two trained field workers. Each day during field work they visited the two antenatal clinics and invited all the pregnant women attending under 19 years to participate in the study. Their ages were double checked against the age given in their clinical records. The controls were agematched teenagers who said they were not pregnant and did not have a history of pregnancy. Each case was asked to name eight women of the same age who were at their school (not necessarily from the same class), if they were attending school, or who lived in the same neighbourhood, if they were not in school. The possible controls named seven and eight were approached first and invited to participate in the study. This was done to reduce bias as it was anticipated that they would be less ‘case-like’ than the first few mentioned. None of the subjects refused to participate in the study. A structured questionnaire was used to interview all the study subjects. The development of the questionnaire was informed by in-depth interviews that were undertaken with 24 pregnant teenagers recruited from one of the two antenatal clinics that participated in the survey (Wood et al., 1998), and some previously tested questions from other surveys. The questionnaire contained questions on: demographic and socio-economic characteristics, substance abuse, circumstances around sexual activity, attitudes towards adolescent sexual activity, sources of information regarding sex and sexuality, contraceptive use and attitudes towards contraception and, for cases, circumstances around pregnancy. The questionnaires were translated into, and administered in Xhosa, the first language of the subjects. A total of 597 subjects (199 cases and 398 controls) were recruited into the study. On first inspection of the data, it was apparent that 42 (10.5%) of the selected controls had never had intercourse and 23 of these had never had boyfriends. They were excluded from further analysis as they were not at risk of pregnancy and they were too few in number for reliable conclusions to be drawn about them as a sub-group. A further three controls were dropped because they had not provided information about contraceptive use. As a result of their exclusion, 29 cases had only one control (162 had two controls) and eight cases were dropped from the study as both their controls were thus excluded. As a conse- quence the analysis presented here is based on a total of 544 subjects (191 cases and 353 controls). All the teenagers had had at least one boyfriend and all had had sexual intercourse. Conditional logistic regression was used to investigate the relationship between the outcome pregnant/not pregnant and a set of prognostic factors (for example demographic information, household and socio-economic influences, attitudes, etc.). Each age matched set was considered as a stratum. Model-building was achieved by requesting backward as well as forward selection of predictor variables. The best subset selection was requested for the selection of nine variables out of a possible 14 for the final model. The set of variables for the final model was also selected on feasibility grounds. As in the case of ordinary logistic regression, the conditional risk ratios (computed by exponentiation of the parameter estimates) were used to interpret the relationship of pregnancy to the prognostic factors considered. If the risk was larger than 1, an increment in the factor would increase the hazard rate. If the risk ratio was less than 1, an increment in the factor would decrease the hazard rate. The risk ratio should be interpreted as an estimate of the incidence-density ratio (Nurminen, 1995), which is considered to be a measure of the instantaneous rate of development of disease outcome in a population. For the continuous covariates, the risk ratio gives the change in the log odds for an increase of ‘‘1’’ unit in the covariate. The Log-likelihood for the model are presented as summary measures of goodness-of-fit. Results The mean age of both cases and controls was 16.4 years (SD of 1.10). Their mean age at first menstruation had been 3 years previously, at 13.7 years (SD 1.10) for cases and 13.5 years (SD 1.10) for controls. All the teenagers had had a boyfriend and had sexual intercourse for the first time at a mean age of 14.0 years (SD of 1.60 for cases and 1.40 for controls). Most (97.4%) of the 191 pregnancies in this study were not planned and 34.1% (65) teenagers said they had considered terminating the pregnancy. Table 1 shows some indicators of social circumstances and the relationships of the teenagers with their boyfriends, focusing on the first ever and current relationships. The pregnant teenagers came from larger households, were less likely to live with their father, less likely to have a TV at home and were more likely to live in a home made of materials other than bricks, i.e. wood, zinc or plastic. Both groups of women started dating at roughly the same age and had been dating for about 2.5 years at the time of enrolment in the study. The boyfriends were 737 R. Jewkes et al. / Social Science and Medicine 52 (2001) 733–744 Table 1 Social circumstances and boyfriends of the teenagers Mean number of people in household (SD) Materials for house is not bricks Did not live with biological father Does not have a television Mean age with first boyfriend (SD) Mean age difference of girl and first boyfriend (SD) Mean age difference of girl and current boyfriend (SD) Mean number of current boyfriends (SD) Mean number of boyfriends ever (SD) Mean number of sexual partners ever (SD) Current boyfriend still at school (n) % Pregnant % Not pregnant Risk ratio 95% CI 5.0 (1.9) 65.3 (124) 74.4 (142) 49.2 (94) 14.0 (1.4) 3.5 (1.8) 5.1 (3.3) 1.2 (0.4) 2.0 (1.0) 1.8 (1.0) 47.6 (91) 3.8 (1.5) 42.5 (147) 52.1 (184) 26.4 (93) 13.8 (1.3) 2.7 (1.3) 4.0 (2.5) 1.1 (0.5) 1.9 (1.0) 1.8 (1.0) 68.0 (240) 1.48 3.09 2.62 2.98 1.06 1.41 1.16 1.91 1.06 1.08 0.42 1.31–1.66 2.02–4.74 1.76–3.90 1.97–4.51 0.92–1.24 1.23–1.61 1.08–1.24 1.25–2.93 0.89–1.26 0.89–1.31 0.29–0.62 some years older than the teenage girls and the mean age difference was greater for the current than the first partner. Having greater mean age differences between both the first and current boyfriend and girlfriend were risk factors for pregnancy. Twenty-six women reported having more than one current boyfriend and this was significantly more common amongst the cases. There was no difference in the mean number of sexual partners ever; the range for the cases was 1–6 and for controls 1– 4. The pregnant teenagers were significantly more likely to have boyfriends who had left school. Table 2 describes some aspects of the circumstances of the teenagers’ first experience of sexual intercourse. Most had intercourse with their first boyfriend, while 9.2% (17) of cases and 5.1% (17) of controls said that their first sexual experience was either rape, incest or sex with someone other than a boyfriend. Most commonly the first sexual partners were described as some one the women had just met. Initiating sexual activity with a steady boyfriend or someone known for a while was protective against subsequent pregnancy. Less than a third of teenagers said they were willing to have sex the first time. Most commonly they reported being persuaded to do so, but a substantial number said they were forced or raped. This was significantly more common amongst the pregnant teenagers. The great majority of teenagers agreed with statements that intercourse happened on the spur of the moment, that people of their age seem to do it and that it was a natural follow on in the relationship. The majority of teenagers said they were in love, but this was significantly less likely for the pregnant ones. Very few used contraception the first time. Almost half of the respondents said their first sexual relationship was still continuing at the time of interview. There was no difference between the two groups in the median duration of the first sexual relationship (when this was with a boyfriend), which was 9 months. The interquartile range was much wider for controls (0.5–24 months) than cases (0.5–18 months). Very few young women (4.7% cases vs 2.9% controls) reported that the relationship did not continue after the first episode of intercourse. The first sexual relationships of the women who became pregnant were significantly shorter than the controls. At the time of interview, all the teenagers reported that they were either still in their first relationship or had moved onto a second or subsequent partner. Where this had occurred the majority of both groups had had their next sexual partner within a month, but it was significantly more likely to happen within a week with the pregnant teenagers. Table 3 presents some other aspects of the dynamics of the teenagers’ sexual relationships. The great majority of both groups of teenagers reported having sex at least once a week, but the pregnant ones were significantly more likely to report this. The majority of teenagers feared that they would be beaten or their partners would leave them if they refused to have sex. These fears were significantly associated with pregnancy. Pregnant teenagers were significantly less likely to cite love as the main motivation for sex and more likely to identify fear. Fear of losing friends or pressure from friends were not significant risk factors for pregnancy. The proportion of each group who said they had had sex for a present or money was very similar in each group (21.1% cases vs 18.8% controls). Fear of presents stopping was not a risk factor for pregnancy. The majority of teenagers reported having had sex against their wishes and one in ten had been raped. Experience of coercive sex was a risk factor for pregnancy. Two thirds of teenagers had been beaten by a boyfriend, with pregnant teenagers reporting more episodes. A greater frequency of beatings was a risk factor for pregnancy. Very few women had left 738 R. Jewkes et al. / Social Science and Medicine 52 (2001) 733–744 Table 2 Sexual dynamics: first relationship Sexual partner at first intercourse: boyfriend father/step-father/uncle rape other Relationship with first sexual partner: just met recently known for a while steady relationship Sexual initiation: willing persuaded forced or raped How sex came about: it happened on the spur of moment I was curious to see what it was like people my age seem to do it natural ‘‘follow-on’’ in relationship I was drunk I wanted to lose virginity I was in love No contraception used Duration of first sexual relationship: still continuing >3 months 1–3 months 51 month Interval until next sexual partner: 1 week or less 1 week to 1 month 1 month to 3 months >3 months still with first or no second partner % pregnant (n) % not pregnant (n) Risk ratio 95% CI 90.8 (165) 5.4 (10) 2.7 (5) 1.1 (2) 94.9 (318) 1.5 (5) 1.2 (4) 2.4 (8) 0.49 – – – 0.23–1.04 – – – 82.2 (157) 11.0 (21) 6.8 (13) 53.3 (188) 36.5 (129) 10.2 (36) 1.00 0.20 0.47 – 0.11–0.34 0.23–0.97 25.7 (49) 42.4 (81) 31.9 (61) 30.3 (107) 51.6 (182) 18.1 (64) 1.00 0.98 2.35 – – 1.53–3.61 86.4 (165) 46.6 (89) 83.8 (160) 83.8 (160) 2.1 (4) 3.1 (6) 68.1 (130) 88.4 (169) 84.9 (299) 53.7 (189) 92.9 (327) 91.2 (321) 0.0 1.4 (5) 90.9 (320) 82.3 (291) 1.13 0.75 0.37 0.49 – – 0.19 1.83 0.68–1.88 0.52–1.08 0.19–0.69 0.27–0.89 – – 0.11–0.32 1.04–3.23 44.7 (85) 15.3 (29) 5.8 (11) 34.2 (65) 49.3 14.0 11.7 25.1 (173) (88) (41) (49) 1.00 1.08 0.54 1.62 – 0.62–1.86 0.26–1.14 1.04–2.51 15.3 (29) 27.4 (52) 7.9 (15) 6.4 (12) 43.2 (82) 3.7 (13) 27.6 (97) 13.9 (49) 5.6 (20) 49.2 (173) 1.00 0.28 0.15 0.24 0.22 – 0.13–0.57 0.06–0.35 0.08–0.70 0.11–0.44 boyfriends as a result of being beaten. Most of the women said that their boyfriend had other girlfriends, and this was a protective factor against pregnancy. Most women said they did nothing when they found out, but cases were more likely to have fought with the other girl. Confronting the boyfriend about it was strongly associated with protection against pregnancy. Table 4 presents the multiple regression model. In the model, teenage pregnancy was most strongly associated with having frequent sex (once a week or more) without injectable contraceptive protection, forced sexual initiation, lack of TV ownership, having a larger household size, not living in a brick house, not living with the biological father, talking openly about sex with a boyfriend, and perceptions that most friends are pregnant. The model chi-squared for this model was highly significant (p=0.0001). The five variables which were considered for the final model but were not entered are: opinions about abortion, in love on sexual initiation, getting information about sex and sexuality from a medical source and confronting an unfaithful boyfriend. They were not entered as they were found to be strongly correlated with other variables and their inclusion caused the model to be unstable. For example, when the ‘‘girls reaction to unfaithful boyfriend’’ was included in the multiple model it was found to be strongly correlated with the variables ‘‘TV ownership’’, ‘‘frequency of sex’’ and ‘‘forced sexual initiation’’ already in the model. If these variables were then excluded, the resulting model chi-squared was highly significant (p=0.0001) but it was not statistically correct to pursue this as the model was unstable. Table 5 shows the interactions between feelings of readiness of the teenagers when they first had sex and 739 R. Jewkes et al. / Social Science and Medicine 52 (2001) 733–744 Table 3 Sexual dynamics Intercourse>1 time a week Did not use injectable contraceptives in past year Fears if sex refused: beatings he would leave laughed at lose friends presents stop Main reasons for sex: love fear peers do it Experience of coercive sex: sex against wish ever been raped ever beaten by boyfriend Number of episodes of beatings: median (range) quartiles mean (SD) Beaten and left boyfriend: not beaten beaten and left beaten but did not leave talks openly about sex with boyfriend perception that most friends are pregnant boyfriend has other girl Reaction when discovered about other girlfriend (n=406): fought with girl did nothing left him confronted him confronted boyfriend when found out he had another girl % pregnant (n) % not pregnant (n) Risk ratio 95% CI 94.7 (180) 84.1 (159) 72.5 (253) 42.9 (148) 9.22 7.57 4.19–20.29 4.56–12.57 77.9 (148) 65.3 (124) 39.2 (74) 6.4 (12) 45.0 (85) 72.1 62.9 33.5 13.0 41.2 (250) (219) (116) (45) (141) 1.85 2.18 1.61 0.46 1.50 1.10–3.11 1.05–4.51 1.02–2.55 0.23–0.90 0.92–2.45 48.7 (93) 48.2 (92) 3.1 (6) 72.9 (256) 19.7 (69) 7.4 (26) 1.00 4.54 0.74 – 2.87–7.18 0.29–1.91 72.0 (137) 11.1 (21) 59.8 (110) 59.7 (113) 9.4 (33) 60.1 (203) 1.76 1.18 1.04 1.20–2.59 0.64–2.16 0.72–1.50 10 (1,21) 4–10 4.6 (5.2) 5 (1,20) 3–10 3.3 (3.9) – – 1.08 – – 1.03–1.13 40.8 13.6 45.7 37.6 51.6 68.6 39.8 15.8 44.4 20.4 21.1 83.2 (139) (55) (155) (71) (73) (288) 1.00 0.88 1.10 2.69 6.01 0.43 – 0.51–1.52 0.74–1.65 1.72–4.19 3.65–9.88 0.27–0.68 2.5 (7) 66.4 (184) 14.8 (41) 16.3 (45) 13.6 (48) – – – – 0.14 – – – – 0.05–0.39 (75) (25) (84) (71) (97) (129) 19.4 (25) 58.9 (76) 17.8 (23) 3.9 (5) 2.1 (4) Table 4 Multiple regressiona Variable Risk ratio 95% CI Higher mean number of people living in home Did not live with biological father Did not live in a brick house Did not own a TV Sex once a week or more Did not use injectable contraceptives Most friends have been pregnant Forced sex Talked openly about sex with boyfriend 2.44 3.26 5.09 10.33 30.81 24.35 4.38 14.42 4.72 1.71–3.46 1.28–8.30 2.01–12.88 3.35–31.82 6.72–141.37 7.34–80.79 1.45–13.23 3.76–55.33 1.62–13.82 a Model Chi-square is 206.61 with 9 degrees of freedom. statements describing how intercourse first came about. The data presented is only for those first having sex with a boyfriend. The table shows some substantial differences in the proportion of young women who identified with particular statements describing how intercourse came about depending on their feelings of readiness. It also shows differences in their patterns between those who were pregnant and the control group. 740 R. Jewkes et al. / Social Science and Medicine 52 (2001) 733–744 Table 5 Relationships between feelings when they first had sex with a boyfriend and statements describing how intercourse came about (n=497) Percentages Relationship Description of feelings at sexual initiation Pregnant n=167 Not pregnant n=330 Feelings on sexual initiation and perception that it ‘‘happened on the spur of moment’’ willing 80.0 89.7 persuaded forced willing 90.0 82.7 97.8 87.9 70.7 95.9 persuaded forced willing 28.6 34.6 86.7 45.4 5.2 94.9 persuaded forced willing 77.1 98.1 82.2 93.1 94.8 93.8 persuaded forced willing 80.0 98.1 4.4 89.7 93.1 5.1 persuaded forced willing persuaded forced 4.3 0.0 91.1 80.0 30.8 0.0 0.0 98.0 97.1 67.2 Feelings on sexual initiation and ‘‘curious to see what it was like’’ Feelings on sexual initiation and perception that ‘‘people my age do it’’ Feelings on sexual initiation and perception that it was a ‘‘natural ‘follow on’ in the relationship’’ Feelings on sexual initiation and ‘‘wanting to lose virginity’’ Feelings on sexual initiation and being ‘‘in love’’ Discussion Aspects of the circumstances of sexual initiation were associated with risk of pregnancy, conceived both at that sexual episode and at subsequent ones. Most of the teenagers had intercourse with their first boyfriend, who most described as some one the girls had just met. This finding was also reported by Wood et al. (1997) and Flisher et al. (1993) in their research in the same geographical area. Boyfriends of the pregnant teenagers were significantly older than those dated by controls, a finding which has been reported in studies in other countries (e.g. Millar & Wadhera, 1997) and would result in greater power inequalities in these relationships. Less than a third of teenagers said they were willing to have sex the first time. Most commonly they reported being persuaded to do so, but a substantial number described being forced or raped. This was significantly more common amongst the pregnant teenagers. The term ‘‘forcing’’ is commonly used in South Africa and is distinguished from rape. A similarly high prevalence (28%) of forced sexual initiation was reported by both Buga et al. (1996) and has been described in the qualitative research of Wood et al. (1998) in Khayelitsha and of Wood & Jewkes (1998) in Umtata. Richter (1996) found that 28% of girls reported having ever been ‘‘forced against their will to have sex’’. One in ten of the cases first had intercourse with someone other than a boyfriend, most commonly this was an older male relative; the proportion was nearly twice that of the controls. Most of the women had quite lengthy first sexual relationships. The cases’ first sexual relationships with boyfriends were significantly shorter than those of the controls, which was probably due to the circumstances of sexual initiation. The interval to the next partner was also significantly shorter for the cases. Thus after sexual initiation, both groups of teenagers reported remaining almost continuously in sexual relationships and at the time of interview, all the teenagers were either still in their first relationship or had moved onto a second or subsequent partner. Most had started this second relationship within a month of the first terminating. These patterns may be explained by ethnographic data from the former Transkei region which highlights the overwhelming preoccupation of young people with acquiring and keeping boyfriends and girlfriends (Wood & Jewkes, 1998). Both women and men were often R. Jewkes et al. / Social Science and Medicine 52 (2001) 733–744 reluctant to terminate, even unsatisfactory relationships, particularly as for women this entailed a risk of physical violence. The significant difference between cases and controls with respect to beating was not in its occurrence but its frequency. Coercive sex appeared to be in some respects different from beatings. Most of the teenagers said that at some time they had had sex against their wishes, but cases were significantly more likely to report this. The frequent beatings and forcing of sex are manifestations of male attempts to forcefully control the relationship. The infrequency with which women left after being beaten suggests that to a large extent they were successful. Although an informal transactional aspect of sexual relationships is often said to heighten women’s vulnerability (McGrath et al., 1993), particularly in reducing the ability to insist on condom use, in this study it was not found to be a risk factor for pregnancy. Many relationships had some form of exchange basis which teenagers came to rely on or enjoy, as indicated by the number of teenagers who feared that the presents would stop if they refused to have sexual intercourse. There was no difference between the two groups in the prevalence of reports of having engaged in sex explicitly for presents or money. Rather than being seen as ‘‘sex work’’ this reflects a view of sex in terms of reciprocity which is widespread in Africa (Caldwell, Caldwell & Quiggin, 1989). There may have been differences, however, in the types of presents given, their relative value to the young women and the frequency of frankly transactional sex. These could be explored in further research. Assertiveness and intimacy More equal power relations seemed to be protective against pregnancy. This is seen in the reaction of young women to their partners’ infidelity. When asked about their reactions when they had discovered their boyfriends’ unfaithfulness, the most common response from both groups was that they did nothing, which may reflect the fact that some of the young women would have been the second girlfriend of a man who was married or had another main girlfriend. Cases were much more likely to report that they fought with the other girlfriend, a practice has been discussed in qualitative research (Wood & Jewkes, 1998). Controls were more likely than the cases to confront their boyfriend about the other relationship. A willingness to do this was a strong protective factor against pregnancy, and it was also strongly correlated with forced sexual initiation. We hypothesise that this is an indicator of greater equality of power distribution within the relationship and that this power could also 741 be deployed with respect to protection against pregnancy. There are some indications to suggest that the relationships of the young women who get pregnant were more strongly valued and perhaps more intimate, in the sense of the two members of the relationship being more reliant on each other. One sign is the frequency of intercourse, which is one measure of intimacy. The most substantial risk ratio in the model was associated with having sex once a week. This may be explained partly through the biological mechanism of greater exposure to risk of pregnancy in the absence of effective contraception, but its negative correlation with the variable on confronting an unfaithful boyfriend, i.e. disempowerment, indicates that other factors may be in play. One explanation is that these relationships were more important to the young women, which is why they are more prepared to fight other women who threaten them and are less willing to risk losing them through confrontation. In general these women’s partners were less likely to have other girlfriends, which may reflect the fact that these women were more important to them. Another indicator of intimacy may be the partners talking ‘‘openly’’ about sex, as Varga (1997) suggests that this is unusual in more casual relationships. The position in the multiple model may also be explained if the man took advantage of his younger girlfriend’s sexual naivity to impose a version of the meaning of sex and its role in a relationship (cf. Wood et al., 1998). If these relationships were more intimate and more important to the players concerned it is possible that more of the babies in this study were, if not planned, or at least perceived to be wanted before or around the time of conception, than is suggested by the ‘‘unplanned’’ nature of most of them. This would be in keeping with a dynamic model of ‘‘wantedness’’. The coexistence of greater violence and greater intimacy or investment in a relationship is possible given the particular circumstances of these young women’s relationships, i.e. where violence was more or less ‘normal’, abuse episodes often followed by short periods of remorse and gifts (Wood & Jewkes, 1998), enormous importance placed on relationships by girls (particularly) from deprived homes and few alternative sources of entertainment and self-esteem building than acquiring and keeping boyfriends (see above). Understanding forced sex Table 5 presents descriptive data which is of interest to a discussion of power within sexual relationships and in understanding the meaning of forced sexual initiation. Almost all the controls who said they had been willing, described events in the same way. They were in love, curious about sex and it happened on the spur of the 742 R. Jewkes et al. / Social Science and Medicine 52 (2001) 733–744 moment. As might be expected, all the controls who said they wanted to lose their virginity fell into this group. Those who were persuaded to have sex were in many respects similar, but significantly less likely to have been curious. The group who described sexual initiation as forced differed. Whilst the majority of these indicated that they had been in love at the time, the proportion was significantly lower than those of the other two groups. Almost none said they were curious to see what it was like. Being in love and being curious about sex were thus the most important predictors of willingness, which is what would have been expected from the findings of the qualitative research (Wood et al., 1998). Similar patterns are found for the cases who described themselves as willing. Most said they were in love and curious about sex. A small proportion, similar to that of the control group indicated that they had actively wanted to lose their virginity. The proportion in love was lower than for the controls, possibly reflecting the fact that more of the cases had only just recently met their first sexual partner. The group who said they were persuaded were similar to the controls in that most said they were in love and many fewer said they were curious. Four percent of cases who were persuaded said they actually wanted to lose their virginity, suggesting perhaps an element of scripted refusal. The cases who were forced to have sex differed markedly in their pattern of responses to the controls who were forced. They were half as likely to report being in love, although a third did report this, but much more likely to report some curiosity. The proportion who were ‘‘curious’’ was even slightly higher than amongst the persuaded cases. Thus quite a few of those who said they were forced described feelings as part of the circumstances of sexual initiation which most closely reflected willingness to have sex. This raises the possibility that amongst the forced group there may have been women who were willing but represented the event as forced to the interviewer to save face in the interview. Another explanation is that some may have been forced by their partners in situations where otherwise they would have consented, this practice is discussed in Wood & Jewkes (forthcoming). Buga et al. (1996) assert that forced sex cannot be regarded as the same as rape. The women interviewed apparently agreed with this as only one control and two of the cases explicitly identified their experience of forcing as ‘rape’ when this was asked in a separate question. Adolescents normally reserve the term ‘rape’ for the actions of strangers or groups of men (sometimes including a boyfriend) (Wood et al., 1998). The patterns of responses to questions describing the circumstances of sexual initiation, however, suggest that an outsider would view it as partner rape and so perhaps it is best equated with the Western notion of date rape. Overall the cases were significantly less likely than controls to perceive that sexual activity was normal amongst their peers and a natural follow on in the relationship. The cases who were forced, however, were much more likely than other cases to perceive sexual activity to be normal amongst their peers. We hypothesise that this reflects an idea that sexual initiation, sex in relationships and possibly also subsequent pregnancy, are more age-associated inevitabilities than something many teenage women seek, particularly desire or feel they have much control over. The act of forcing would serve as an assertion of male control over the sexual domain of the relationship, a reminder for the young women of prevailing power relations. This obviously requires further research, but may explain the strong association between forced sexual initiation and pregnancy occurring often years later, which is shown in the multiple regression model. These findings suggest that aspects of the dynamics of sexual relationships are of considerable importance in teenage pregnancy. There are other factors in multiple regression model and the meanings of these and mechanisms of risk are discussed in detail elsewhere (Vundule, Maforah, Jewkes & Jordaan, 2000). Conclusions This study should be regarded as an initial exploration of risk factors for pregnancy amongst sexually active adolescents, as essentially hypothesis-generating rather than testing. None the less it is important as it represents a first attempt to move beyond descriptive studies of adolescent sexual practices in South Africa. There are limitations on the interpretation of the data imposed by the study design and analysis. Ideally in this study sexual activity would have been a matching criterion for the controls, but logistically this was not possible as the relevant information could only be reliably gained after an extended period of interviewing. A decision was therefore made to accommodate it in the analysis stage, although we recognise that this is non-ideal and resulted in the loss of some controls as well as cases. The ‘‘snap shot’’ nature of the study design, unfortunately prevents us from gaining an understanding of the extent to which attitudes and relationships change after the pregnancy. Further insights would require a cohort study. Despite the study’s limitations, the findings suggest that it is not the early initiation of sexual activity per se which puts teenagers at risk of pregnancy, but other aspects of the circumstances in which sexual relationships are enacted. We hypothesise that the high risk associated with forced sexual initiation and unwillingness to confront an unfaithful partner is mediated through inequalities in power relations within the relationships. For the teenagers who become pregnant their relative lack of power in the relationship may be heightened by the substantial differences in ages of the R. Jewkes et al. / Social Science and Medicine 52 (2001) 733–744 partners, the greater likelihood that the man is working and a home environment of greater poverty. Within the relationship the boyfriends assert control through coercive sex, both initially and subsequently, and frequent beatings. Further research into power relations in adolescent relationships and their relationships with adolescent fertility is clearly needed. The associations suggested by the study point towards a need for far greater emphasis in sexual health promotion interventions on gender equality and physical and sexual violence. We further hypothesise that part of the mechanism of risk associated with frequency of intercourse is in relation to its position as an indicator of greater intimacy in the relationship. It may in fact partly reflect wanted pregnancies. These pose a critical challenge for health promoters as education and contraceptive availability will not prevent them. If this hypothesis were proven, it would indicate that meanings of teenage pregnancy which are found in lay people’s private discourses need to be given more focus in a broad intervention strategy. This would need to address the impact of poverty on teenagers’ aspirations and options. Acknowledgements This study was funded by the Urbanisation and Health Programme of the Medical Research Council. We would like to thank the following people for their assistance in undertaking this project: our field workers Nosisi Dingani and Lebohang Sekholomi; Zodumo Mvo for help with data cleaning; and the respondents, who gave their time and shared their experiences with us. We are also grateful to Groote Schuur Hospital for give us permission for the study, to staff of the Midwife Obstetric Units for allowing us to identify and interview our cases on their premises and to staff of the schools for assisting us in locating the controls. References Abdool-Karim, Q. A., Preston-Whyte, E., & Abdool-Karim, S. S. (1992). Teenagers seeking condoms at family planning services } Part 1. A user’s perspective. South African Medical Journal, 80, 301–302. Abrahams, N., Jewkes, R., & Laubsher, R. (1999). ‘‘I do not believe in democracy in the home’’. Men on relationships with and abuse of women. In Medical Research Council Technical Report. Tygerberg: Medical Research Council. Boult, B. E., & Cunningham, P. W. (1991). Black teenage pregnancy in Port Elizabeth. Early Childhood Development & Care, 75, 1–70. Buga, G. A. B., Amoko, D. H. A., & Ncayiyana, D. (1996). Sexual behaviour, contraceptive practices and reproductive 743 health among school adolescents in rural Transkei. South African Medical Journal, 86, 523–527. Caldwell, J., Caldwell, P., & Quiggin, P. (1989). The social context of AIDS in Subsaharan Africa. Population and Development Review, 15, 185–234. Chimere-Dan, O. (1995). Fertility and family size preferences in the Eastern Cape Province, South Africa. A study of the Transkei sub-region. In Population Research Programme. Johannesburg: University of Witwatersrand. Department of Health. (1998). National framework and guidelines for contraceptive services. First draft document for discussion. Republic of South Africa, 17 September. (1999a). 1998 National HIV sero-prevalence survey of women attending public antenatal clinics in South Africa. Pretoria: Health Systems Research and Epidemiology, Department of Health. (1999b). South Africa Demographic and Health Survey. Preliminary report. Pretoria: Department of Health. Department of Health. (forthcoming). South Africa Demographic and Health Survey. Final report. Pretoria: Department of Health. Flisher, A. J., Ziervogel, C. F., Chalton, D. O., Leger, P. H., & Robertson, B. A. (1993). Risk-taking behaviour of Cape Penninsular high-school students. Part VIII. Sexual behaviour. South African Medical Journal, 83, 495–497. Jewkes, R., & Wood, K. (1999). Problematizing pollution: dirty wombs, ritual pollution and pathological processes. Medical Anthropology, 18, 163–186. Jewkes, R., Penn-Kekana, L., Levin, L., Ratsaka, M., & Schrieber, M. (1999). ‘‘He must give me money, he mustn’t beat me’’ } Violence against women in three South African Provinces. Pretoria: Medical Research Council Technical Report. Kau, M. (1988). Sexual behaviour and contraceptive use by adolescent pupils in the Republic of Bophuthatswana. Curationis, 11, 9–11. Kau, M. (1991). Sexual behaviour and knowledge of adolescent males in the Molopo region of Bophutatswana. Curationis, 14, 37–40. Longmore, L. (1959). The dispossessed } a study of the sex-life of Bantu women in and around Johannesburg. London: Jonathon Cape. Makiwane, M. B. (1998). Adolescent pregnancy and reproductive health in Transkei (rural South Africa). African Journal of Reproductive Health, 2, 41–48. Mathai, M. (1997). A study of the knowledge and problem solving ability of family planning nurses in Mdantsane. Unpublished thesis, MPhil (Maternal and Child Health), University of Cape Town, Cape Town. Maforah, F., Wood, K., & Jewkes, R. (1997). Backstreet abortion: women’s experiences. Curations, 20, 79–82. McGrath, J., Rwabukwali, C., Schumann, D., Pearson-Marks, J., Nakayina, S., Namande, B., Nakyobe, L., & Mukasa, R. (1993). Anthropology and AIDS: the cultural context of sexual risk behaviour among urban Baganda women in Kampala, Uganda. Socail Science and Medicine, 36, 429–439. Millar, W. J., & Wadhera, S. (1997). A perspective on Canadian teenage births, 1992–4: older men and younger women?. Canadian Journal of Public Health, 88(5), 333–336 (Sept– Oct). 744 R. Jewkes et al. / Social Science and Medicine 52 (2001) 733–744 Naidoo, L. R., Williams, P., Knight, S. & Bernstein, A. (1991). A survey of knowledge about AIDS and sexual behaviour among students at the University of Natal, Durban Campus. University of Natal AIDS Committee, Durban Working Group. Nash, E. (1990). Teenage pregnancy } need a child bear a child?. South African Medical Journal, 77, 147–151. Ncayiyana, D. J., & Ter Haar, G. (1989). Pregnant adolescents in rural Transkei. South African Medical Journal, 75(4), 231–232. National Education Policy Act. (1996). No. 27. South African Government, Cape Town. National Progressive Primary Health Care Network (NPPHCN) (1996). Youth speak out for a healthy future. Johannesburg: NPPHCN. Nurminen, M. (1995). To use or not to use the odds ratio in epidemiologic analyses?. European Journal of Epidemiology, 11, 365–371. Pauw, B. (1963). The second generation. Cape Town: Oxford University Press. Preston-Whyte, E., Zondi, M., Mavundla, G., & Gumede, H. (1990). Teenage pregnancy, whose problem? Realities and prospects for action in KwaZulu/Natal. South African Journal of Demography, 3, 11–20. Richter, L. (1996). A survey of reproductive health issues among urban black youth in South Africa. Final Grant Report. Pretoria: Medical Research Council. Sai, F., Rees, H., & McGarry, S. (1993). Reproductive health and family consultancy } national review and recommendations. Johannesburg: Commission of the European Communities Special programme for South Africa. Simpson, G. (1991). Explaining sexual violence: some background factors in the current socio-political context. Project for the Study of Violence. Johannesburg: University of Witwatersrand. Statistics South Africa (1998). The people of South Africa population census, 1996. Census in brief. Pretoria: Statistics South Africa. Varga, C. A. (1997). Sexual decision-making and negotiation in the midst of AIDS: youth in KwaZulu/Natal, South Africa, Health transition Review 45–67 (Supplement 3 to Vol. 7). Varga, C. & Makubalo, L. (1996). Sexual (non) negotiation among black African teenagers in Durban, 28 31–38. Varga, C. (1998). Placing adolescent child-bearing in perspective: antenatal health-care utilisation and pregnancy outcome. Paper presented to 4th Reproductive Priorities Conference, Van Der Bijl Park, South Africa, 18–21 August. Vundule, C., Maforah, N. F., Jewkes, R. & Jordaan, E. (2000). Risk factors for teenage pregnancy amongst African adolescents in metropolitan Cape Town: a case control study, South African Medical Journal (in press). Wood, K., Maforah, F., & Jewkes, R. (1998). ‘‘He forced me to love him’’: putting violence on the adolescent sexual health agenda. Social Science and Medicine, 47, 233–242. Wood, K., & Jewkes, R. (1998). ‘‘Love is a dangerous thing’’: micro-dynamics of violence in sexual relationships of young people in Umtata. Pretoria: Medical Research Council Technical Report. Wood, K. & Jewkes, R. (forthcoming). ‘‘Dangerous’’ love: reflections on violence among Xhosa township youth. In R Morrell (Ed.), Changing men in South Africa. Pietermaaritzburg, South Africa: University of Natal Press. Wood, K., Maepa, J., & Jewkes, R. (1997). Adolescent sex and contraceptive experiences: perspectives of teenagers and clinic nurses in the Northern Province. Pretoria: MRC Technical Report.