2. 356 Rogan, Hynie, Casale, Nixon, Flicker, Jobson and Dawad
to sexual behaviour on sexual behaviour outcomes among also inextricably linked to gendered HIV-risk behaviours
school-age young people in South Africa (see Figure 1). (Jewkes, Levin, Loveday & Penn-Kekana, 2003; Morrell,
Epstein, Unterhalter, Bhana & Moletsane, 2009). Jewkes,
Factors affecting HIV risk among young people in Dunkle, Koss, Levin, Nduna, Jama & Sikweyiya (2006)
South Africa and Jewkes & Morrell (2010) argue that there are deeply
In order to examine these interactions, the study employs entwined associations between particular constructions of
a social cognitive approach to measuring attitudes, beliefs, gender roles and risk of sexual coercion, intimate partner
intentions and perceived control over sexual behaviour. violence and HIV, which place young South African women
While social-cognitive models have been used extensively at heightened risk of HIV infection. Research in South Africa
in the planning and implementation of HIV-prevention suggests that intimate partner violence and sexual coercion
programmes (Campbell, 2003), few studies have investi- are associated with having multiple sexual partners, early
gated how individual-level factors (e.g. SES, gender and sexual debut, more recent sexual activity, unprotected sex,
age) interact with norms and attitudes towards sexual inconsistent condom use, unintended pregnancy and being
behaviour among young people. The following section HIV-positive (Andersson, Ho-Foster, Matthis, Marokoane,
reviews the confounding factors commonly associated with Mashiane, Mhatre et al., 2004; Dunkle et al., 2004a;
HIV risk among young people in South Africa. Hoffman, O’Sullivan, Harrison, Dolezal & Monroe-Wise,
2006; Jewkes, Dunkle, Koss et al., 2006; Jewkes, Dunkle,
Socioeconomic status Nduna, Levin, Jama, Khuzwayo et al., 2006; Maharaj &
Chief among the factors linking gender and HIV risk, socioe- Munthree, 2007). Moreover, Wood, Lambert & Jewkes
conomic disadvantage significantly influences a wide range (2008) argue that gender-based violence in South Africa
of sexual risk behaviours among both females and males often occurs in the context of the perceived ‘naturalness’
(Eaton, Flisher & Aaro, 2003; Kalichman, Simbayi, Kagee, of gender inequality in many communities, and in relation
Toefy, Jooste, Cain & Cherry, 2006). SES is likely to have a to the endemic use of violence against women as a part of
particularly significant effect on gendered sexual-behaviour everyday life.
outcomes and, among South African youths, low SES
has been found to have more consistent negative effects Psychosocial factors, including peer norms
on female risk behaviours than on male risk behaviours Behavioural and social cognitive studies have identified
(Hallman, 2004). For young women, lower SES has been a number of psychosocial processes and factors (such as
found to be associated with earlier sexual debut (Hallman, self-efficacy, self-mastery, risk perceptions and beliefs,
2005; Dinkelman, Lam & Leibbrandt, 2007), higher reporting peer and social norms, and self-esteem) that contribute
of transactional sex (Dunkle, Jewkes, Brown, Gray, McIntyre to sexual risk behaviour among young people in South
& Harlow, 2004b; Hallman, 2004 and 2005) and coerced sex Africa (Eaton et al., 2003; Visser, Schoeman & Perold,
(Hallman, 2004 and 2005), higher risk of early pregnancy 2004; Brook, Morojele, Zhang & Brook, 2006). While many
(Hallman, 2004), having multiple partners (Hallman, 2005), of these studies have ignored structural and individual-
lower chances of secondary sexual abstinence (Hallman, level factors (Eaton et al., 2003), several social-cognitive
2004), and lower instances of condom use at last sex studies have begun to explore how factors such as gender
(Hallman, 2005; Hargreaves, Bonnell, Morison, Kim, Phetla, and SES might be direct or indirect determinants of higher-
Porter et al., 2007). The relationship between gender, SES risk behaviour (cf. Brook et al., 2006). It is the interac-
and HIV is not universally accepted, however, and a number tions between individual beliefs, norms and attitudes, and
of studies have revealed either no clear causal relationship structural and individual-level factors, however, that are
(cf. Booysen, 2004) or the opposite relationship between likely to provide a more nuanced understanding of sexual
SES and HIV-risk behaviours and sexual experiences (cf. risk behaviour among young people (Eaton et al., 2003).
Wojcicki, 2005). Of the psychosocial processes, the interaction between
gender and peer norms, in particular, is perhaps the least
Sexual violence and coercion explored in the existing literature. However, earlier work in
The very high rates of sexual violence and coercion in other countries has suggested that gender may interact with
South Africa, including within the school environment, are peer norms in important ways. A study of sexual behaviour
among young people in Zambia (Magnani, Karim, Weiss,
Bond, Lemba & Morgan, 2002), for example, found that the
perception that peers were sexually active was positively
associated with ever having had sex among both boys and
Peer norms Gender girls. Among boys, peer influence was positively associated
and attitudes Sexual with both sexual activity and the number of lifetime partners.
to sexual
behaviour Socioeconomic
behaviours Protective behaviours involving more consistent condom
status use among girls, however, were significantly associated with
some measures of peer influence (Magnani et al., 2002).
In the South African context, peer pressure among
young people is strongly associated with negative views of
condom use and positive views of engaging in sexual activity
Figure 1: Schematic of the research hypothesis, applied to
adolescents in South Africa (MacPhail & Campbell, 2001; Brook et al., 2006). Eaton et al.
3. African Journal of AIDS Research 2010, 9(4): 355–366 357
(2003) note that the South African literature finds that the population census, 32% of the population was aged 0–19
peer pressure to be sexually active is generally high for both years; 44% of working-age adults were employed, 21%
boys and girls, but that boys seem to be more influenced by were unemployed and 35% were not economically active
it than girls. On the whole, the available literature seems (Statistics South Africa, 2001). The majority of people
to suggest that boys are more susceptible than girls to living in the study area are black South Africans (64%),
the influences of peer pressure, ‘deviant peers’ and social followed by Indians (16%), whites (15%) and coloureds
norms around sexual behaviour (MacPhail & Campbell, (5%) (Statistics South Africa, 2001).1 A large number of
2001; Brook et al., 2006). However, some evidence does residents live in informal settlements, with a few living in
suggest that girls are influenced to remain abstinent more formal dwellings and classified as low-to-middle-income
strongly by ‘positive’ peer norms than are boys (Dlamini, households. The area is flanked by established medium-to-
Taylor, Mkhize, Huver, Sathiparsad, De Vries et al., 2009). high-income residential areas (Odendaal, 2002).
Moreover, peer influences are unlikely to act on their own
and several studies demonstrate how poverty and wellbeing, Data
in particular, moderate the effect of peer pressure on The data used in the present study were taken from the
sexual behaviour outcomes (Jama, Jewkes, Levin, Nduna, quantitative component of a quasi-experimental retrospec-
Khuzwayo, Duvvury & Koss, 2004; Brook et al., 2006). tive evaluation of the impact of a five-year multifaceted
Despite the important influences of SES, partner violence, HIV-prevention intervention aimed at secondary school
and psychosocial processes (particularly peer norms) learners. The survey consisted of a self-administered
on sexual risk behaviour among young people, empirical questionnaire distributed to 809 learners in Grade 11, at
evidence is needed to understand how gender interacts with two schools implementing the intervention and two control
these factors to shape sexual behaviour. This gap in the schools, all in Cato Manor. Fieldwork was conducted
literature seems particularly significant in the South African between October 2007 and March 2008. A self-adminis-
context, where gender has been so closely associated tered survey tool was created using largely validated
with a number of confounding factors — such as partner research instruments. It was refined for greater clarity and
violence and low SES. Furthermore, this is a context in for the local context after being pilot-tested with youths and
which peer influences on sexual behaviour have been found programme mentors. All questionnaires were printed in both
to be particularly strong. Underlining the importance of English and Zulu, with the two versions presented simulta-
understanding the gendered effects of peer norms, a recent neously on facing pages in each questionnaire booklet.
evaluation of Stepping Stones, the only HIV-prevention Bilingual facilitators assisted learners by introducing the
intervention aimed at young people in South Africa that project and survey instrument, presenting simulations with
has demonstrated a biological impact (i.e. a reduction in ‘test questions’ and then going through the survey with the
the prevalence of HSV-2: Jewkes et al., 2008), offers a learners, question by question. The questionnaires were
potentially important lesson. The intervention went beyond administered at the schools, with boys and girls separated
providing information on HIV and sexual behaviour and was into different classrooms.
based on a participatory approach that addressed gender The survey had four main components. The first part
inequities and encouraged group communication (Jewkes collected sociodemographic information, informed by
et al., 2008). The relative success of this programme past research with South African youths. The second
demonstrates the potential use of harnessing information on part measured the frequency of behaviours and experi-
how peer norms contribute to gendered sexual behaviour ences related to sexuality and sexual health, including past
outcomes. participation in an HIV-prevention programme and past
As such, this article explores how gender and SES sexual behaviour. The third part measured psychosocial
moderate the effects of norms and attitudes about higher- variables and employed internationally validated scales to
risk sexual behaviour among secondary school learners in measure hope (Snyder, Lopez, Shorey, Rand & Feldman,
a low-income community in KwaZulu-Natal province, South 2003), optimism (Scheier & Carver, 1985), self-esteem
Africa. We argue that improving the evidence base on (Rosenberg, 1965) and self-mastery (Pearlin & Schooler,
how gender and SES impact on peer norms will improve 1978). The fourth part of the survey was based on the
our understanding of some of the challenges facing both theory of planned behaviour (Ajzen, 1985) and captured the
HIV-prevention efforts and progress towards gender equality respondents’ attitudes, norms and intentions relating to key
within South African schools. HIV-risk-taking behaviour. The survey data were recorded
and analysed using SPSS 15.0 and Mplus. Ethical approval
Methods was granted by the three partner universities undertaking
the study.
Setting
The study was conducted in an urban semi-formal Outcome variables
community, Cato Manor, located within Durban, KwaZulu- The two main outcome variables considered in the analysis
Natal Province, on the east coast of South Africa. Cato are sexual activity (self-reported sexual debut) and reported
Manor has a long and politically charged history — number of lifetime sexual partners (see Table 1).2 Sexual
perhaps most famously as a site of forced removals during debut is recorded as a dummy variable. Analyses of the
the apartheid period, followed by land invasions in the number of sexual partners included only those learners
early years of the democratic era. According to the 2001 who reported ever having had sex. The number of sexual
4. 358 Rogan, Hynie, Casale, Nixon, Flicker, Jobson and Dawad
partners is a continuous variable, which, after removing individual-level factors (age, gender, experience of partner
outliers more than three standard deviations from the abuse, SES and exposure to HIV-prevention interventions).
median, ranged from ‘1’ to ‘10.’ Both of the outcome indica- Socioeconomic status was self-reported and taken from
tors used in this study are well-established risk indica- a question in which the learners were asked to select the
tors in the context of HIV (Eaton et al., 2003; Jewkes, statement that best described their respective households.
Dunkle, Nduna et al., 2006) as suggested by the fact that The five options ranged from not having enough money for
HIV-prevention programmes often aim to encourage sexual food or other basic items to having enough money for food
abstinence and reduce the number of sexual partners as well as luxury items. The responses were then recoded
(MacPhail & Campbell, 2001; Magnani, MacIntyre, Karim, into a three-point scale representing low, medium and high
Brown & Hutchinson, 2005). SES. All continuous independent variables were mean
centred.
Explanatory variables
Three clusters of independent variables (hypothesised to be Analysis
either potential risk factors or protective factors) are consid- Basic descriptive analyses were used to identify significant
ered in the multivariate analysis in the following section. The associations between the two dependent variables and
independent variables (see Table 1) hypothesised to predict a number of independent variables as well as the gender
a respondent having sexually debuted and the number of differences in sexual behaviour outcomes. In the multivar-
sexual partners reported; this included: 1) attitudes to sexual iate analyses, a binary logistic model was used to identify
behaviour (attitude to sex in the next three months, attitude the predictors of sexual debut (i.e. ever having had sex),
to sex before marriage, and beliefs about risks associated with not having had sex coded as ‘0’ and having had sex
with sex); 2) perceived social and peer norms (prescrip- as ‘1.’ A classic linear regression model was used to identify
tive parent norms and prescriptive partner norms about the predictors of the number of sexual partners reported
premarital sex, sexual risk-taking norms, descriptive norms by the learners who had sexually debuted. All independent
about adolescent sexual behaviour, and prescriptive peer variables demonstrating a significant bivariate associa-
norms about having sex in the next three months); and 3) tion with the respective outcome variable (i.e. ever had
Table 1: Overview of the outcome variables and explanatory variables
Variable Indicator Number of items Response type
Outcome variables
Ever had sex Ever engaged in vaginal or anal penetration (data 2 Yes/No
captured through a binary question and through the Continuous
reported number of lifetime sexual partners).
Number of sexual partners Total number of reported lifetime sexual partners. 1 Continuous
Explanatory variables
Negative attitude to not Displays a negative attitude to abstaining from 3 5-point Likert-type scale*
having sex until marriage premarital sex or displays an intention to engage in sex
before marriage.
Opposed to premarital sex Is opposed to the idea of sex before marriage. 3 5-point Likert-type scale*
Belief in HIV and sexually Acknowledges the specific health risks (i.e. HIV/STIs) 7 5-point Likert-type scale*
transmitted infection (STI) associated with sexual activity.
risks
Avoiding premarital sex: Perceives that parents are opposed to sex before 2 5-point Likert-type scale*
parent norms marriage.
Avoiding premarital sex: Perceives that most recent partner is opposed to sex 2 5-point Likert-type scale*
partner norms before marriage.
Prescriptive peer norms Feels that it is acceptable for both male and female 4 5-point Likert-type scale*
supporting sexual activity peers to be sexually active.
Descriptive peer norms Perception that male and female peers are abstaining 2 5-point Likert-type scale*
supporting abstinence until from sex before marriage.
marriage
Experienced partner abuse Ever been ‘punched or kicked’ by a partner. 1 Yes/No
Exposed to an HIV Ever participated in or been exposed to any type of 1 Yes/No
intervention HIV-prevention programme (no definition provided).
Socioeconomic status 3-point scale (low, medium and high), collapsed from 1 5-point,single response
the 5-point self-assessment. self-assessment of household
wellbeing.
*5-point Likert-type scale: ‘Strongly disagree’ to ‘Strongly agree’
Note: ‘Number of items’ refers to the number of questions used to measure each indicator; several individual-level explanatory variables are
not included in this table due to the simplicity of their construction (e.g. gender and age).
5. African Journal of AIDS Research 2010, 9(4): 355–366 359
sex or the number of lifetime partners) were included in all learners in the sample, girls, for example, reported signif-
the multivariate models. Thus, the final regression models icantly lower intentions to have sex in the next three months
included all variables that: 1) were significantly associated (F(1, 792) = 60.749; p = 0.0001) and lower intentions
with the outcome variable at the 5% level of confidence; 2) to engage in sex before marriage (F(1, 785) = 6.856; p =
mediated the effects of the other independent variables on 0.009). Girls reported significantly less support from their
the outcome variable; 3) displayed a significant interaction parents to have sex in the next three months (F(1, 787) =
with gender, or gender and SES; or 4) had a strong theoret- 52.397; p = 0.0001) and less support to have sex before
ical basis for inclusion. marriage (F(1, 789) = 62.536; p = 0.0001) as compared
As in other studies exploring gender differences in sexual with the male learners. Each of these associations holds
behaviour (cf. Magnani et al., 2002), gender interactions when controlling for both age and sexual debut; however,
were included in the models in order to capture the effect of the association between gender and the intention to have
gender and the interactions between gender and the other sex before marriage was not significant (p = 0.105) when
independent variables over and above the main effects. controlling for age and sexual debut.
Therefore, in both models presented in the following section,
second- and third-order interaction terms were added to Sexual activity
explore the moderating effects of gender and of gender Ever had sex
and SES, respectively. Main effects were entered in the first Several key bivariate associations (not shown) with having
step; two-way interactions (with gender) were entered in the sexually debuted suggest that hypothesised factors — such
second step; and three-way interactions (with gender and as SES and intimate partner violence, in particular — may
SES) in the third step. The interaction steps were accepted be important predictors of earlier sexual debut. Among the
if the R2 change was significant at the 5% level. sexually active learners, a significantly larger number of
female learners reported partner violence compared with the
Results females who were not sexually active. Significantly fewer
of the sexually active females described their households
Sample description as having a high SES in comparison to the sexually active
The age of the participants (n = 809) ranged from 16 to males and the non-sexually active females. Interestingly, a
23 years. Respondents who were under age 16 or who significantly larger proportion of the sexually active female
spoke neither English nor Zulu were removed from the learners reported having been exposed to an HIV-prevention
sample (n = 26). An overwhelming majority (98.9%) classi- programme or intervention as compared with the male
fied themselves as black and 90.6% indicated Zulu as learners who had sexually debuted.
their home language. Of the participating learners, 44.8% In the multivariate analyses, both main effects and
were males and 55.2% were females. Just over half of two-way interactions were significant. Table 3 shows the net
the respondents (50.7%) indicated that they had repeated effects of the predictor variables on ever having had sex,
a school year due to failing. Roughly 50% reported either using a logistic regression model. Measures of attitudes and
not having enough money for food or for other basic items, norms towards premarital sex were included in the model,
while just 2.9% reported coming from households with but attitudes to having sex in the next three months were
enough money for luxury items. The mean household size not significantly associated with sexual debut and were
was 5.7 members, and most learners (63.6%) reported that therefore excluded. In the first step, those learners opposed
their mother was the primary caregiver. to premarital sex were shown to have lower odds of having
As in other studies of adolescents in South Africa, the sexually debuted. Prescriptive norms supporting sexual
boys reported having more sexual partners than did the activity among peers were associated with increased odds
girls, and a significantly larger proportion of boys (70.6%) of having sexually debuted. Parent norms were not signifi-
than girls (46.5%) reported having sexually debuted by cantly associated with sexual debut, but the perception that
the time of the survey (Table 2). Sexually active boys also potential sexual partners would abstain from sex before
reported a significantly younger median age at sexual marriage (partner norms) significantly reduced the odds of
debut (age 14) compared with sexually active girls (age 16). having sexually debuted. As expected, older learners and
Among sexually active girls, 19.2% reported having ever those who had been exposed to partner violence had signif-
been diagnosed with a sexually transmitted infection (STI) icantly higher odds of having sexually debuted. Participation
(cf. 18.9% of the sexually active boys) and 9.3% reported in some form of HIV-prevention programme and being
having ever been forced to have sex by a partner. Just over female significantly reduced the odds of having sexually
one-fifth of sexually active girls participating in the study debuted. Controlling for the other factors, the effect of SES
reported having been pregnant at some point in the past. on sexual debut falls away.
Partner violence among the sample was high, with 28.2% of In the second step (two-way interaction model), gender
the girls (and 7.2% of the boys) having reported being struck (being female) moderated the effects of several variables.
by a partner at some point. A significantly higher percentage Gender and age form a significant interaction, with older
of the sexually active girls (63.4%) had been tested for HIV girls more likely to have had sex. In the main-effects model,
compared with only 16% of the sexually the active boys. a stronger belief in the risk of HIV or STIs was positively
The girls participating in the study also reported signifi- associated with having sexually debuted, but being female
cantly different sexual behaviour intentions and social and reverses the relationship between sexual debut and
peer norms compared with their male counterparts. Among understanding the risks associated with having sex. In
6. 360 Rogan, Hynie, Casale, Nixon, Flicker, Jobson and Dawad
Table 2: Proportion and means of selected HIV-risk indicators for the Grade 11 learners, by gender
Males Females p-value
(n = 362) (n = 446)
Ever had sex 70.6 46.5 0.0001
Had sex in the past year 27.3 20.6 0.025
Age at sexual debut* 14 16 0.0001**
Median number of lifetime sexual partners* 5 1 0.0001**
Ever been pregnant* – 20.9 –
Ever impregnated* 9.2 – –
Ever been tested for HIV* 16.0 63.4 0.0001
Ever been diagnosed with an STI* 18.9 19.2 0.319
Ever been physically abused by a partner 7.2 28.2 0.0001
Ever been physically forced to have sex* 7.8 9.3 0.353
Ever engaged in transactional sex* 11.0 7.1 0.338
Participated in any HIV-prevention programme 56.4 67.7 0.001
Low socioeconomic status 53.3 47.3 0.295
*Includes only those reporting ever having had sex (245 males; 197 females)
**p-value for F-statistic for difference in means; all other p-values are reported for χ2 analyses
addition, being female and perceiving that both male and Discussion
female peers were abstaining from sex before marriage
(descriptive norms) significantly reduced the odds of having The results of this study contribute to the existing body
sexually debuted. Three-way-interaction variables modelling of literature on gender and HIV risk in two ways. First,
gender, SES, and selected norms and attitudes were not they demonstrate that the effect of peer norms on sexual-
significant and are not described here. behaviour outcomes differs between males and females.
Moreover, the results provide empirical evidence concerning
Multiple sexual partners the direction of the relationships (i.e. risk-enhancing versus
Table 4 demonstrates the effects of the independent protective effects) between peer norms and different sexual
variables on the number of sexual partners reported by behaviours. Perhaps most notably, the finding that the
sexually active learners. As the model suggests, age, perception that their peers are abstaining from premarital
having experienced partner violence, attitudes to having sex sex has a protective effect against both sexual debut and
in the next three months, and prescriptive peer norms about the number of lifetime partners among girls (but not among
sex were all significantly associated with having a greater boys) is important. Second, the study demonstrates that,
number of sexual partners. Gender is highly significant (p ≤ over and above the effects of all the other explanatory
0.0001) in the first step, with being female resulting in fewer variables, having experienced partner violence is strongly
partners, but is not significant after the interaction variables and positively associated with both sexual debut and a
are included. Being exposed to an HIV intervention of some greater number of lifetime partners. Overall, the study’s
type and being opposed to premarital sex among peers findings point to the role that gender plays in moderating
had significant negative associations with the number of the effects of attitudes and norms on sexual behaviour
sexual partners in the main-effects model. After controlling outcomes. In particular, the results suggest that peer norms
for other variables, the number of sexual partners was not and experience with partner violence are particularly strong
significantly associated with SES or with perceived parent factors associated with gendered sexual behaviour, while
or partner norms. SES (when controlling for other factors) was not found to be
Two second-order interaction variables add to the explan- a significant predictor.
atory power of the model. Female learners seem to be
particularly affected by perceived peer norms about avoiding Lessons about gender and peer norms
sex before marriage (i.e. perceiving that their peers are Most studies in South Africa have found boys to be more
abstaining from sex before marriage), as gender interacted sensitive to peer influences than are girls (e.g. MacPhail
significantly with this set of norms to predict having had & Campbell, 2001; Eaton et al., 2003; Brook et al., 2006).
fewer sexual partners (p < 0.05). The effect of attitudes on However, the findings presented here suggest that girls’
premarital sex was less clear, as being female moderates perceptions of what their peers are doing are likely to be
the effect of being opposed to premarital sex, such that very important to them, and, as in other studies (e.g. Dlamini
being female and being opposed to having premarital sex et al., 2009), these perceptions are found to impact on their
is associated with having had a higher number of sexual decision to remain abstinent. As such, more work should
partners. Perceived parent and partner norms about be directed to understanding the different types of peer
premarital sex did not interact with gender. Once again, the influences and the ways in which young males and females
three-way-interaction variables were not significant and so respond. Efforts to standardise the way that peer influences
this step was not included in the final equation. are measured would contribute to our greater understanding
7. African Journal of AIDS Research 2010, 9(4): 355–366 361
Table 3: Adjusted odds ratios (with confidence intervals) for ever having had sex (n = 809 adolescents)
Block 1 Block 2
Attitudes
Negative attitude to not having sex until marriage 1.510 0.861
(0.805–2.831) (0.334–2.220)
Opposed to premarital sex 0.701** 0.639**
(0.572–0.860) (0.452–0.905)
Belief in HIV and STI risks 1.292 2.200**
(0.869–1.921) (1.119–4.323)
Norms
Avoiding premarital sex: parent norms 0.971 0.937
(0.819–1.152) (0.724–1.213)
Avoiding premarital sex: partner norms 0.837** 0.880
(0.716–0.980) (0.699–1.109)
Prescriptive peer norms supporting sexual activity 1.284** 1.308*
(1.068–1.543) (0.952–1.798)
Descriptive peer norms supporting abstinence until marriage 0.932 1.128
(0.803–1.081) (0.886–1.435)
Individual-level factors
Age 1.403*** 1.175
(1.232–1.598) (0.959–1.440).
Female (vs. male) 0.570** 0.565**
(0.392–0.827) (0.383–0.832)
Experienced partner abuse (vs. ‘not’) 2.166** 2.129**
(1.370–3.413) (1.327–3.45)
Exposed to an HIV intervention 0.669** 0.658**
(0.471–0.951) (0.459–0.944)
Socioeconomic status 0.960 0.971
(0.758–1.216) (0.683–1.382)
Interactions
Gender * age 1.365**
(1.043–1.786)
Gender * descriptive peer norms supporting abstinence until marriage 0.723**
(0.528–0.990)
Gender * belief in HIV and STI risks 0.463*
(0.197–1.090)
Nagelkerke R2 0.335 0.362
Model χ2 222.79*** 244.177***
Block χ2 222.79*** 21.378**
*p ≤ 0.10
**p ≤ 0.05
***p ≤ 0.0001
of the effects of peer norms on sexual behaviour. less likely to have sexually debuted.
An especially interesting finding is that being female and
perceiving that males and females of a similar age are Lessons about violence and coercion
abstaining from sex decreased the odds of having sexually The high prevalence of rape, intimate partner violence,
debuted. Perceived abstinence among peers was even and sexual coercion in South Africa (Jewkes et al., 2003;
associated with having fewer sexual partners among the Andersson et al., 2004) and the level of reported partner
girls who had already sexually debuted. This finding, in violence in this study lead to the unfortunate conclusion that
particular, is an important addition to the South African litera- violence against women remains a common occurrence
ture in light of the prevailing view that peer norms, especially and implies that violence is a significant risk factor for HIV
among males, encourage sexual risk behaviour (rather than among female youths (over and above the effects of all
having a protective effect). Earlier work (Dlamini et al., 2009) other factors). While the findings related to gender, attitudes,
has offered preliminary evidence of the protective effects and perceived peer norms emphasise the gendered nature
of peer norms among females in South Africa, but did not of sexual decision-making, the role of partner abuse cannot
control for other factors (i.e. in a multivariate context). In be ignored. In both models, having experienced partner
addition to the protective effects of peer norms, there is also violence in the form of striking or punching (admittedly, a
evidence that females are acting on an understanding of the fairly narrow definition of partner violence) was highly signif-
risks involved in engaging in sex, as the female learners icant as a main effect and it remained significant after the
with a higher recognition of the risk of HIV and STIs were interaction variables were included.
8. 362 Rogan, Hynie, Casale, Nixon, Flicker, Jobson and Dawad
Table 4: Multiple regression coefficients (with standard errors) for variables predicting the number of sexual partners reported by the sexually
active Grade 11 learners (n = 442)
Block 1 Block 2
Intercept 5.546*** 5.401***
(0.195) (0.200)
Attitudes
Negative attitude to not having sex in the next three months 0.987** 1.948**
(0.415) (0.575)
Opposed to premarital sex –0.349** –0.537**
(0.126) (0.172)
Norms
Avoiding premarital sex: parent norms 0.014 0.036
(0.104) (0.130)
Avoiding premarital sex: partner norms 0.004 0.040
(0.101) (0.129)
Prescriptive peer norms supporting sexual activity 0.217* 0.271*
(0.118) (0.164)
Descriptive peer norms supporting abstinence until marriage 0.127 0.362**
(0.105) (0.146)
Individual level factors
Age 0.246** 0.301**
(0.084) (0.114)
Female (vs. male) –3.514*** –1.784
(0.282) (3.015)
Experienced partner abuse (vs. not) 0.568** 0.597**
(0.296) (0.298)
Exposed to an HIV intervention –0.387* –0.323
(0.228) (0.226)
Socioeconomic status 0.220 0.101
(0.159) (0.202)
Interactions
Gender * opposed to premarital sex 0.510**
(0.254)
Gender * descriptive peer norms supporting abstinence until marriage –0.504**
(0.211)
R2 0.481 0.513
R2 change 0.481*** 0.032**
F 25.395*** 15.713***
(df1 = 14; df2 = 384) (df1 = 25; df2 = 373)
F change 25.395*** 2.242**
*p ≤ 0.10
**p ≤ 0.05
***p ≤ 0.0001
As noted in many studies, we find that being the victim of Lessons about socioeconomic status
partner abuse is almost exclusively the domain of females SES was not found to be a significant predictor in either
and that it is positively associated with sexual risk behaviour of the models and did not interact with gender and
(in this study, having sexually debuted and having multiple selected norms and attitudes. This may be the case, in
sexual partners) after controlling for other factors. Moreover, part, due to the subjectivity of the SES measure employed
the finding that, among girls, being opposed to premarital (self-reported) and to the fact that the majority of the
sex was associated with having had a higher number of learners reported medium or low SES. The validity of
sexual partners suggests that some of the girls had likely self-reporting SES among youths in a developing-country
experienced coercion. This particular interpretation, while context, where the respondents live in the same impover-
certainly plausible, illustrates the difficulties associated with ished community and have relatively similar levels of SES,
making causal inferences with respect to norms, attitudes to the best of our knowledge, has not been reviewed in the
and sexual-behaviour outcomes. However, the link between literature. Moreover, while several studies (e.g. Jama et
partner violence and various sexual risk behaviours — such al., 2004; Brook et al., 2006) have found that poverty or a
as unprotected sex, early sexual debut, multiple partner- low SES can moderate the effects of peer influences, the
ships, sex with older men, coerced sex and transactional relationship between SES and sexual-behaviour outcomes
sex — has been well noted in the South African literature is far from clear, with much of the variance in the findings
(Andersson et al., 2004). based on the different measures of SES employed (Wojcicki,
9. African Journal of AIDS Research 2010, 9(4): 355–366 363
2005). While the findings of this study offer no indication that where males and females can collectively ‘renegotiate their
a low (or high) SES has any bearing on sexual-behaviour peer identities’ and where individuals may gain confidence
outcomes or on the effects of peer norms, we argue that it is while learning how to negotiate sexual interactions. Striking
still important to control for SES (in spite of the challenges) a balance between recognising the different role of peer
when modelling the determinants of sexual behaviour — influences among males and females while allowing
particularly in light of the emphasis in the recent literature both genders to collectively form peer identities appears
on SES and sexual risk behaviour. to be a key challenge. The findings of this study suggest
that partner violence and sexual coercion should not be
Implications for policy and programming overlooked when peer identities are negotiated. Indeed,
The findings presented here have important implications for Campbell (2004) emphasises that peer education interven-
policy and programming. Not only are sexual-risk profiles tions are less likely to succeed in affecting peer norms if the
different for girls, but the determinants of sexual-behaviour participants are not able to adequately address the social
outcomes are significantly different between girls and boys. factors (e.g. partner violence) impeding behaviour change.
HIV prevention, ‘life orientation’ and reproductive health The findings here, together with those from the evalua-
interventions need to take these differences into account. tions of Stepping Stones (Jewkes et al., 2008) and IMAGE
The finding that perceived peer norms are more important (Pronyk et al., 2008), clearly point towards the need to
for girls than for boys might suggest that group interventions move beyond health-awareness approaches and to explic-
and other forms of support may influence girls more than itly address the underlying gender roles and inequities that
their male peers. As such, the school environment is likely facilitate gender-based violence and sexual risk behaviour.
to be an important space for shaping peer norms and this
presents a unique opportunity in light of emerging evidence Study limitations
that school-based peer-education programmes can impact Finally, several caveats must be offered. First, the study
on peer-group norms (see Visser, 2007). Community-wide cannot claim to be representative of young people in South
interventions may also form an important part of addressing Africa or even of young people in Cato Manor. The sampling
peer influences. While discussing the significance of peer design was intended to match control and treatment
influences in the Zambian context, Magnani et al. (2002) groups and not to be representative of young people from
noted the growing conviction that community interventions the community. Second, as with most studies that rely on
may be the most effective way of combating negative peer self-reports of respondents’ sexual behaviour (cf. Magnani et
influences among young people. al., 2002; Karnell, Cupp, Zimmerman, Feist-Price & Bennie,
In terms of the content and structure of interventions, 2006; Jewkes et al., 2008), the possibility of reporting bias
lessons from the existing body of literature on gender and and error must be declared when the questionnaires ask
HIV risk in South Africa seem to suggest that interven- young people to self-report sensitive information about
tions that focus on changing sexual behaviour or health their personal or sexual lives. In this study, this bias may
knowledge, rather than on the gender norms that shape have been exacerbated by a lack of privacy as the youths
behaviours, are less likely to succeed (Harrison et al., completed the questionnaire. While all efforts were made to
2010; Jewkes & Morrell, 2010). Morrell et al. (2009) further ensure that the learners had sufficient desk space to ensure
argue that many approaches to achieving gender equality privacy, the reality is that the classrooms were crowded and
in South African schools have been challenged by the need some learners had to share a desk. Third, this study only
to effectively reach peer groups through which gender looked at two sexual risk-taking behaviours and did not
norms and identities are created. Conversely, programmes consider condom use, for example, among sexually active
such as Stepping Stones (Jewkes et al., 2008) and IMAGE youths. Fourth, the question format and the scales that were
(Pronyk, Kim, Abramsky, Phetla, Hargreaves, Morison et included in the questionnaire were validated in a Western
al., 2008) have met with some success due, in large part, setting. A concerted effort, however, was made to adapt the
to their emphasis on ‘collective critical thinking,’ group language to fit the local context; to that end, the question-
communication, and normative gender identities (Harrison naire was carefully piloted prior to being distributed to the
et al., 2010). Notably, these relative successes have been respondents.
observed despite the well-known difficulties in implementing Fifth, and perhaps most critically, causality cannot be
peer education programmes (for a fuller discussion, see inferred from the analyses presented here. As in other
Campbell & MacPhail, 2002). The findings of this study social cognitive models, both independent and dependent
can contribute to the design of similar inventions and may variables may be ‘reciprocally determining’ (Bandura, 1977),
help to identify which types of peer norms are important in and one cannot necessarily make assumptions about the
shaping gender roles and norms. direction of causality between, for instance, attitudes to
We argue that understanding the effects of peer norms, having sex in the immediate future and having had sex
in particular, is a critical first step towards engaging with the in the past. For example, among the girls, it is not clear
formation of gender identities and the design of interventions why being opposed to having premarital sex was associ-
that can impact on established gender norms in the context ated with having had a greater number of sexual partners.
of HIV and AIDS. It is crucial, however, that interventions As outlined above, one possible interpretation may be that
take into account the role of partner violence in the relation- negative sexual experiences (e.g. coercion or forced sex)
ships of young people. As Campbell (2004) and Campbell have created this view. Another interpretation is that there
& MacPhail (2002) argue, peer education requires a space are contradictions in youths’ perceptions of ideal behaviours
10. 364 Rogan, Hynie, Casale, Nixon, Flicker, Jobson and Dawad
and actual outcomes, among females, in particular. These research interests include: gender, survey and sampling methodol-
caveats notwithstanding, the findings presented here offer ogies, health, poverty and inequality, evaluation methodologies and
important insights into the gendered differences in sexual reproductive health.
behaviour and decision-making among young people in Michaela Hynie is an associate professor in the Department
South Africa. of Psychology and an associate director of the York Institute of
Health Research at York University. She is generally interested in
how to use research as a means for social change, both directly,
Conclusions through the process of research itself, and indirectly, by generating
research findings that can be used for activism. The content of her
This study addresses a critical gap in the South African research falls into three broad categories: culture, immigration and
literature in terms of the relationship between gender, social health inequities; how basic interpersonal or social-psychological
norms and sexual behaviour among adolescents in South processes are affected by culture; and, sexual behaviour and safer
Africa. In particular, the findings suggest important ways of sex, with a focus on culturally appropriate interventions and the
understanding how gender relates to behavioural outcomes. evaluation of international initiatives.
Interventions seeking to effect behaviour change among Marisa Casale is a researcher with HEARD at the University of
young people should acknowledge gender differences and KwaZulu-Natal, and a doctoral candidate with the Department of
Psychology at the University of Cape Town. Her research focuses
attempt to develop context-appropriate ways of addressing
on southern Africa, with key interest in HIV prevention among
peer influences through, for example, the use of peer youths, and caregiver and child health.
educators, group and community interventions, and school- Stephanie Nixon is an assistant professor in the Department of
based life-orientation programmes. However, such interven- Physical Therapy and the Dalla Lana School of Public Health, as
tions should not shy away from the social or individual-level well as academic director of the International Centre for Disability
factors (chief among these being partner violence) that and Rehabilitation, at the University of Toronto. She is also a
could prevent adolescents’ behaviour change. research associate with HEARD at the University of KwaZulu-Natal.
The need for a ‘collective negotiation of peer identities’ Sarah Flicker is an assistant professor in the Faculty of
suggests that peer-based interventions, in particular, should Environmental Studies at York University, and an Ontario HIV
target communication between boys and girls. Similarly, the Treatment Network Scholar. Her expertise is in the areas of adoles-
cent HIV prevention and support, participatory methodologies, and
results of this study also suggest that such interventions
community-based research.
would do well to avoid emphasising traditional constructions Geoff Jobson (previously a researcher at HEARD) is a senior
of girls as passive actors in regard to sexual decision-making. researcher at the Peri-Natal HIV Research Unit at the University of the
On the whole, the findings suggest that the relationship Witwatersrand. His research interests include: masculinity and sexuality,
between gender and HIV risk is more complex than often and HIV-prevention and care models for people living with HIV.
conceptualised and is significantly associated with a combina- Suraya Dawad is a researcher at HEARD at the University of
tion of individual-level and psychosocial factors. Future work KwaZulu-Natal; her interests include health systems and the costing
should seek to identify, in greater detail, the specific types of comprehensive-care models.
of peer norms or pressures that are linked with sexual-
behaviour outcomes in the South African context. In terms References
of programming, this study highlights the imperative that
HIV-prevention programmes should move beyond focusing Ajzen, I. (1985) From intentions to actions: a theory of planned
on health knowledge to directly address the underlying behavior. In: Kuhi, J. & Beckmann, J. (eds.) Action-Control: From
factors that impact on gender norms and behaviours. Cognition to Behavior. Heidelberg, Germany, Springer.
Andersson, N., Ho-Foster, A., Matthis, J., Marokoane, N., Mashiane,
Notes V., Mhatre, S., Mitchell, S., Mokoena, T., Monasta, L., Ngxowa,
N., Pascual Salcedo, M. & Sonnekus, H. (2004) National
1 We use the same racial classifications employed by Statistics cross-sectional study of views on sexual violence and risk of HIV
South Africa in its household surveys and censuses. infection and AIDS among South African school pupils. British
2
This article focuses exclusively on these two sexual behaviour Medical Journal 329(7472), pp. 329–352.
outcomes and not on condom use (or other protective Bandura, A. (1977) Social Learning Theory. Englewood Cliffs, New
behaviours) at last sex. Forthcoming work will examine condom Jersey, Prentice-Hall.
use and intentions to use condoms in the context of both the Booysen, F. (2004) HIV/AIDS, poverty and risky sexual behaviour in
intervention and gendered sexual behaviour. South Africa. African Journal of AIDS Research (AJAR) 3(1), pp.
57–67.
Acknowledgments — This work was supported by the Health Brook, D., Morojele, N., Zhang, C. & Brook, J. (2006) South African
Economics and HIV/AIDS Research Division (HEARD) at the adolescents: pathways to risky sexual behaviour. AIDS Education
University of KwaZulu-Natal (South Africa), York University and Prevention 18(3), pp. 259–272.
(Canada) and the University of Toronto (Canada). We acknowledge Campbell, C. (2003) ‘Letting Them Die’: Why HIV/AIDS Prevention
the important contribution of our entire multicounty research team, Programmes Fail. Oxford, UK, James Currey/The International
in particular our field researchers. We also thank all of the learners, African Institute.
teachers and school administrators that gave so generously of Campbell, C. (2004) Creating environments that support peer
their time. education: experiences from HIV/AIDS. Health Education 104(4),
pp. 197–200.
The authors — Michael Rogan (previously a researcher at HEARD) Campbell, C. & MacPhail, C. (2002) Peer education, gender and
is currently a part-time lecturer and doctoral fellow at the School the development of critical consciousness: participatory HIV
of Development Studies at the University of KwaZulu-Natal. His prevention by South African youth. Social Science and Medicine
11. African Journal of AIDS Research 2010, 9(4): 355–366 365
55, pp. 331–345. Jewkes, R., Dunkle, K., Nduna, M., Levin, J., Jama, N., Khuzwayo,
Coates, T., Richter, L. & Caceres, C. (2008) Behavioural strategies N., Koss, M., Puren, A. & Duvvury, N. (2006) Factors associated
to reduce HIV transmission: how to make them work better. The with HIV serostatus in young rural South African women:
Lancet 372, pp. 669–684. connections between intimate partner violence and HIV.
Dillard, K. (2002) Adolescent Sexual Behaviour: Socio-Psychological International Journal of Epidemiology 35(6), pp. 1461–1468.
Factors. Washington, D.C., Advocates for Youth. Jewkes, R., Levin, J., Loveday, A. & Penn-Kekana, L. (2003)
Dinkelman, T., Lam, D. & Leibbrandt, M. (2007) Household and Gender inequalities, intimate partner violence and HIV-preventive
community income, economic shocks and risky sexual behaviour practices: findings of a South African cross-sectional study. Social
of young adults: evidence from the Cape Area Panel Study 2002 Science and Medicine 56(1), pp. 125–134.
and 2005. AIDS 21(supplement 7), pp. 49–56. Jewkes, R. & Morrell, R. (2010) Gender and sexuality: emerging
Dlamini, S., Taylor, M., Mkhize, N., Huver, R., Sathiparsad, R., perspectives from the heterosexual epidemic in South Africa
De Vries, H., Naidoo, K. & Jinabhai, C. (2009) Gender factors and implications for HIV risk and prevention. Journal of the
associated with sexual abstinent behaviour of rural South African International AIDS Society 13(6), pp. 1–11.
high school-going youth in KwaZulu-Natal, South Africa. Health Jewkes, R., Nduna, M., Levin, J., Jama, N., Dunkle, K., Puren, A.
Education Research 24(3), pp. 450–460. & Duvvury, N. (2008) Impact of Stepping Stones on incidence
Dunkle, K., Jewkes, R., Brown, H., Gray, G., McIntyre, J. & Harlow, of HIV and HSV-2 and sexual behaviour in rural South Africa:
S. (2004a) Gender-based violence, relationship power and risk cluster randomised controlled trial. British Medical Journal
of HIV infection in women attending antenatal clinics in South 337(a506), pp. 1–11.
Africa. The Lancet 363, pp. 1415–1421. Kalichman, S., Simbayi, L., Kagee, A., Toefy, Y., Jooste, S., Cain, D.
Dunkle, K., Jewkes, R., Brown, H., Gray, G., McIntyre, J. & Harlow, & Cherry, C. (2006) Associations of poverty, substance use, and
S. (2004b) Transactional sex among women in Soweto, South HIV-risk behaviors in three South African communities. Social
Africa: prevalence, risk factors and association with HIV infection. Science and Medicine 62(7), pp. 1641–1649.
Social Science and Medicine 59(8), pp. 1581–1592. Kalipeni, E., Craddock, S. & Ghosh, J. (2004) Mapping the AIDS
Eaton, L., Flisher, A. & Aaro, L. (2003) Unsafe sexual behaviour in pandemic in Eastern and Southern Africa: a critical overview. In:
South African youth. Social Science and Medicine 56, pp. 149–165. Kalipeni, E., Craddock, S., Oppong, J. & Ghosh, J. (eds.) HIV and
Hallman, K. (2004) Socioeconomic Disadvantage and Unsafe AIDS in Africa: Beyond Epidemiology. Malden, Massachusetts,
Sexual Behaviors among Young Women and Men in South Blackwell.
Africa. New York, The Population Council. Karnell, A., Cupp, P., Zimmerman, R., Feist-Price, S. & Bennie, T.
Hallman, K. (2005) Gendered socioeconomic conditions and (2006) Efficacy of an American alcohol and HIV-prevention
HIV-risk behaviours among young people in South Africa. African curriculum adapted for use in South Africa: results of a pilot study
Journal of AIDS Research (AJAR) 4(1), pp. 37–50. in five township schools. AIDS Education and Prevention 18(4),
Hargreaves, J., Bonnell, C., Morison, L., Kim, J., Phetla, G., Porter, pp. 295–310.
J., Watts, C. & Pronyk, P. (2007) Explaining continued high HIV MacPhail, C. & Campbell, C. (2001) ‘I think condoms are good but
prevalence in South Africa: socioeconomic factors, HIV incidence aai, I hate those things’: Condom use among adolescents and
and sexual behaviour change among a rural cohort, 2001–2004. young people in a South African township. Social Science and
AIDS 21(supplement 7), pp. S39–S48. Medicine 52, pp. 1613–1627.
Harrison, A. (2008) Young people and HIV/AIDS in South Africa: Magnani, R., Karim, A.M., Weiss, L., Bond, K., Lemba, M. &
prevalence of infection, risk factors and social context. In: Abdool- Morgan, G. (2002) Reproductive health risk and protective
Karim, S. & Abdool-Karim, Q. (eds.) HIV/AIDS in South Africa. factors among youth in Lusaka, Zambia. Journal of Adolescent
Cambridge, UK, Cambridge University Press. Health 30, pp. 76–86.
Harrison, A., Newell, M.-L., Imrie, J. & Hoddinott, G. (2010) HIV Magnani, R., MacIntyre, K., Karim, A.M., Brown, L. & Hutchinson,
prevention for South African youth: Which interventions work? P. (2005) The impact of life skills education on adolescent sexual
A systematic review of current evidence. BMC Public Health risk behaviours in KwaZulu-Natal, South Africa. Journal of
10(102), pp. 1–12. Adolescent Health 36, pp. 289–304.
Hendriksen, E., Pettifor, A., Lee, S.-J., Coates, T. & Rees, H. (2007) Maharaj, P. & Munthree, C. (2007) Coerced first sexual intercourse
Predictors of condom use among young adults in South Africa: and selected reproductive health outcomes among young women
The Reproductive Health and HIV Research Unit National Youth in KwaZulu-Natal, South Africa. Journal of Biosocial Science
Survey. American Journal of Public Health 97(7), pp. 1241–1248. 39(2), pp. 231–244.
Hoffman, S., O’Sullivan, L., Harrison, A., Dolezal, C. & Monroe- Morrell, R., Epstein, D., Unterhalter, E., Bhana, D. & Moletsane, R.
Wise, C. (2006) HIV risk behaviors and the context of sexual (2009) Towards Gender Equality: South African Schools during
coercion in young adults’ sexual interactions: results from a diary the HIV and AIDS Epidemic. Pietermaritzburg, South Africa,
study in rural South Africa. Sexually Transmitted Diseases 33(1), University of KwaZulu-Natal Press.
pp. 52–58. Njau, B., Mtweve, S., Barongo, L., Manongi, R., Chugulu, J.,
Human Sciences Research Council (HSRC) (2008) South African Msuya, M., Mwampeta, S., Kiwale, B., Lekule, J. & Jalipa, H.
National Prevalence, Incidence, Behaviour and Communication (2006) The influence of peers and other significant persons
Survey, 2008. A Turning Tide among Teenagers? South Africa, on sexuality and condom-use among young adults in northern
Cape Town, HSRC Press. Tanzania. African Journal of AIDS Research (AJAR) 6(1), pp.
Jama, P., Jewkes, R., Levin, J., Nduna, M., Khuzwayo, N., Duvvury, 33–40.
N. & Koss, M. (2004) Peer pressure to have sex: risk factors Odendaal, N. (2002) ICTs in development — Who benefits? Use of
and associated sexual practices. Paper presented at the 15th geographic information systems on the Cato Manor Development
International Conference on AIDS, Bangkok, Thailand, 11–16 July Project, South Africa. Journal of International Development 14,
2004. pp. 89–100.
Jewkes, R., Dunkle, K., Koss, M., Levin, J., Nduna, M., Jama, O’Sullivan, L., Harrison, A., Morrell, R., Monroe-Wise, A. & Kubeka,
N. & Sikweyiya, Y. (2006) Rape perpetration by young, rural M. (2006) Gender dynamics in the primary sexual relationships of
South African men: prevalence, patterns and risk factors. Social young rural South African women and men. Culture, Health and
Science and Medicine 63(11), pp. 2949–2961. Sexuality 8(2), pp. 99–113.
12. 366 Rogan, Hynie, Casale, Nixon, Flicker, Jobson and Dawad
Pearlin, L.I. & Schooler, C. (1978) The structure of coping. Journal
of Health and Social Behavior 19(1), pp. 2–21.
Pronyk, P., Kim, J., Abramsky, T., Phetla, G., Hargreaves, J.,
Morison, L., Watts, C., Busza, J. & Porter, J. (2008) Combined
microfinance and training intervention can reduce HIV-risk
behaviour in young female participants. AIDS 22, pp. 1659–1666.
Rosenberg, M. (1965) Society and the Adolescent Self-Image.
Princeton, New Jersey, Princeton University Press.
Scheier, M.F. & Carver, C.S. (1985) Optimism, coping and
health: assessment and implications of generalized outcome
expectancies. Health Psychology 4(3), pp. 219–247.
Snyder, C., Lopez, S., Shorey, H., Rand, K. & Feldman, D.
(2003) Hope theory, measurements, and applications to school
psychology. School Psychology Quarterly 18(2), pp. 122–139.
Statistics South Africa (2001) Census 2001. Pretoria, South Africa,
Statistics South Africa.
Varga, C. (2003) How gender roles influence sexual and
reproductive health among South Africa adolescents. Studies in
Family Planning 34(3), pp. 160–172.
Visser, M. (2007) HIV/AIDS prevention through peer education and
support in secondary schools in South Africa. Journal of Social
Aspects of HIV/AIDS 4(3), pp. 678–694.
Visser, M., Schoeman, J. & Perold, J. (2004) Evaluation of HIV/
AIDS prevention in South African schools. Journal of Health
Psychology 9(2), pp. 263–280.
Walker, L. & Gilbert, L. (2002) HIV/AIDS: South African women at
risk. African Journal of AIDS Research (AJAR) 1(1), pp. 75–85.
Wojcicki, J. (2005) Socioeconomic status as a risk factor for HIV
infection in women in East, Central and Southern Africa: a
systematic review. Journal of Biosocial Science 37, pp. 1–36.
Wood, K. & Jewkes, R. (1997) Violence, rape and sexual coercion:
everyday love in a South African township. Gender and
Development 5(2), pp. 41–46.
Wood, K., Lambert, H. & Jewkes, R. (2008) ‘Injuries are beyond
love’: Physical violence in young South Africans’ sexual
relationships. Medical Anthropology 27(1), pp. 43–69.
Wood, K., Maforah, F. & Jewkes, R. (1998) ‘He forced me to love
him’: Putting violence on adolescent sexual heath agendas.
Social Science and Medicine 47(2), pp. 233–242.