What do young people think makes their relationships good? Factors associated with assessments of dating relationships in South Africa

Abstract

Little is known about the factors and outcomes associated with young people’s subjective relationship assessments. Understanding what young people think makes their relationships ‘good’ or ‘bad’ would give us insight into what is important to them in their relationships as well as their decision-making and behaviour within them. Self-report data from 757 girls (mean age = 17.09 years) and 642 boys (mean age = 17.23 years) were analysed using logistic regression. Relationship primacy was significantly associated with positive relationships for girls and boys. Among girls, partner education and open communication about sexual and reproductive health were additionally related to relationship assessments. Among boys, very little quarrelling was the only additional factor associated with positive relationship assessment. Although relationship assessment was not associated with depression or problem drinking for either girls or boys, drug use was less likely among both girls and boys who reported having a positive relationship. Boys in positive relationships were also more likely to have used a condom the last time they had sex with their main partner. Intervention programmes should equip teenagers with skills to develop and maintain positive relationships.

Introduction

Young people’s intimate relationships are crucial for programmes and interventions to promote sexual and reproductive health, and prevent HIV and intimate partner violence (IPV). To improve relevance and acceptability, it is important that intervention developers have a nuanced understanding of teenagers’ conceptualisations, constructions, and experiences of these relationships. One important aspect of young people’s romantic relationship concepts is their ideas about what constitutes a ‘good’ relationship because it could provide insight into their motivations and choices to become involved in or stay in particular intimate relationships with particular partners. However, very little is known about what factors influence subjective assessments of what makes a relationship ‘good’ or ‘bad’.

Perceptions of relationship quality and satisfaction are based on each partner’s experience of a relationship and interpretation of interactions within it (Galliher et al. 2004). This assessment affects behaviour and emotional responses within the relationship (Galliher et al. 2004). Among adults, relationship satisfaction is linked with relationship functioning and the mental health and well-being of each member of a couple (Bradbury et al. 2000, Coleman 2011, Renshaw et al. 2011, Whitton and Kuryluk 2012). While there is an extensive literature of adult relationship satisfaction, less is known about factors and outcomes associated with young teenagers’ relationship satisfaction, specifically those in low resource contexts.

Studies, primarily conducted in North America, on adolescent romantic relationship satisfaction and quality have discovered associations between these assessments and teenagers’ individual characteristics and family and peer relationships (Thomas and Daubman 2001, Haugen et al. 2008, Roisman et al. 2009, Galliher and Bentley 2010). Very few studies have investigated how or which relationship characteristics and relationship behaviours influence relationship satisfaction among teenagers. Engaging in out-of-school activities and communication has been positively associated with relationship satisfaction (Carlson and Rose 2012). Low-level sexual activity such as kissing was associated with greater relationship satisfaction whereas higher-level activity such as intercourse was negatively correlated with relationship quality for younger adolescents (Welsh et al. 2005). There was no significant association between intercourse and relationship satisfaction among the older adolescents (Welsh et al. 2005) suggesting a developmental change in concepts of good relationships. There is also evidence that these concepts may be different for boys and girls. Among girls, a positive association between relationship quality and lower conflict and higher harmony ratings of their boyfriends was found whereas among boys, better relationship quality was associated with ratings of their own supportive behaviour and their ability to consider their girlfriend’s point of view during their interactions (Galliher et al. 2004).

We know very little about what factors young people take into account to assess their relationships and no published research to date has explored the influence of several relationship characteristics on young people’s assessments of their dating relationships. Understanding the factors influencing teenagers’ assessments will give us insight into what is important to them in their relationships and give clues to decisions and behaviours within them. All the published, peer-reviewed literature on this topic reports research conducted in the global North. It is essential to build an understanding of young people’s relationships in developing countries, such as South Africa, in order to inform public health initiatives to prevent IPV and HIV infection, and promote sexual and reproductive health and positive relationships. Therefore, the primary aim of this paper is to investigate what South African teenagers believe makes their relationships ‘good’ and a secondary aim is to explore whether this assessment has any impact on their health outcomes.

Method

Recruitment and data collection

This paper presents an analysis of the baseline data from participants in the Stepping Stones randomised control trial prior to their participation in the intervention; a detailed description of the study methodology and outcomes are available elsewhere (Jewkes et al. 2006b). Trial participants were recruited from 70 small, peri-urban towns or rural village sites near Mthatha in the Eastern Cape province of South Africa. Poor literacy levels and unemployment in this region are high; families rely on a combination of subsistence farming, informal trading, social grants or government pensions, and support from family members who are employed in other areas of the province or country. In general, participants in the trial were quite poor (Jewkes et al. 2006b). The trial condition was randomised at the level of town/village. Recruitment of 20 girls and 20 boys took place primarily at high schools at each study site. Trial participants were isiXhosa-speaking volunteers who were aware of the study content (i.e., sexual and reproductive health and intimate relationships) and they ranged in age from 15 to 26 years old; this method of recruitment resulted in a non-probability sample of study participants.

Structured questionnaires were administered in isiXhosa in individual, face-to-face sessions with trained, sex-matched interviewers; the measures relevant to this study are described in detail below. We present an analysis of data from 757 girls and 642 boys aged 15–18 years who had an intimate partner – a boyfriend or girlfriend – in the year before baseline interviews.

The ethics committees of the University of Pretoria and University of Witwatersrand granted approval for the trial. Participation was voluntary and informed consent was obtained from participants.

Measures

Background information

Information about participants’ age, socioeconomic status (SES), and adverse childhood experiences were collected. Age in years was calculated using the date of birth provided by participants. SES was measured using several items including the ownership of a television and a car in the home, food availability in the home, and how easy it would be to access ZAR100 (approximately USD14) for a health emergency. A principal component factor analysis was run on these items to obtain a factor score. Higher scores indicate relatively higher SES within the study context described earlier. Cronbach’s alpha was 0.56 for girls and 0.60 for boys.

Childhood trauma was included because these early experiences can play a role in individuals’ later psychosocial functioning, including their intimate relationships (Banyard et al. 2001, Banyard and Cross 2008). Seventeen items, adapted from the Childhood Trauma Questionnaire (Bernstein et al. 1994), asked participants whether they had experienced emotional or physical neglect, or emotional, physical, or sexual abuse before the age of 18 years. Participants responded to each item on a 4-point scale from Never to Very Often. A principal component factor analysis was run on these items to obtain a factor score. Higher scores indicate more childhood adversity. Cronbach’s alpha was 0.75 for girls and 0.72 for boys.

Relationship information

Participants were asked about their current or most recent intimate relationship with a boyfriend or girlfriend. The measures included relationship assessment, age difference between partners, partner status as main or casual partner, relationship equity, communication, and relationship conflict and violence. These items were chosen for this study because they are typically addressed in intervention programmes aiming to prevent IPV and poor sexual and reproductive health (SRH) outcomes (Jewkes et al. 2006b, Jewkes et al. 2008, Jewkes 2010, Jewkes et al. 2011, Mathews et al. 2012).

The main outcome variable in the study was measure by a single item with four response categories asked participants how they would assess their current relationship with their main partner. Responses were dichotomised into positive assessments (‘excellent’ or ‘good’ ratings, coded as “1”) and negative assessments (‘ok’ or ‘not good’ ratings, coded as “0”). The use of a single item measure allowed for a fully subjective interpretation by participants. This condition made possible a unique exploratory analysis to understand which relationship factors may have played a role in young peoepl’s assessments.

Age difference between partners was calculated using the main partner’s age reported by the participant and the participant’s age. A positive number indicates that the boyfriend was older and a negative number indicates that the girlfriend was older.

Participants provided information about the highest level of education obtained by their partner. For girls’ partners, this variable was dichotomised into partners who had achieved matric (Grade 12, high school diploma) or further education and those who had not achieved this level. For boys’ partners, this variable was dichotomised into partners who had achieved at least Grade 8 (first year of high school) or further and those who had not.

Participants were asked whether their current partner was their main partner and also whether they were that partner’s main partner; these items were combined to identify mutual main partnership versus those that were not. This measure does not indicate monogamy; either partner may have had another, secondary or casual partner (Dunkle et al. 2004, Jewkes and Morrell 2010).

The Sexual Relationship Power Scale was adapted to assess relationship equity (Pulerwitz et al. 2000, Dunkle et al. 2004). Thirteen statement items each had four response options ranging from Strongly Agree to Strongly Disagree. For girls a typical item would read: “When [name of partner] and I disagree, he gets his way most of the time.” For boys, the parallel item would read: “When [name of partner] and I disagree, I get my way most of the time.” Higher scores indicated greater equity or power-sharing in the relationship whereas lower scores indicated a more gender-conservative relationship in which boyfriends had more control. A principal component factor analysis was run on these items to obtain a factor score. Cronbach’s alpha was 0.73 for girls and 0.58 for boys.

Communication openness with partner was measured by six items with a four-point response option ranging from Strongly Agree to Strongly Disagree. Items ask about whether the participant talks with his or her partner about disagreements and how free the participant feels to discuss specific topics (e.g., hopes and fears, HIV, pregnancy). Lower scores indicate higher levels of communication openness between partners. Cronbach’s alpha was 0.65 for girls and 0.59 for boys. Four items assessed the level of communication about SRH issues between the participant and her or his partner. Participants answered yes or no to each item including discussions about contraception, HIV prevention, sex, and having children together. Higher scores indicate more communication. A principal component factor analysis was run on both the openness and SRH-specific items to obtain a factor score.

Quarrelling with the partner was assessed with a single item asking how often the participant and his or her partner argue; the four response options ranged from Never to Often. Relationship conflict was measured by five items asking how often a participant and her or his partner argued about specific issues (e.g., money, fidelity, spending time together). A principal component factor analysis was run on these items to obtain a factor score. Cronbach’s alpha was 0.74 for girls and 0.74 for boys. Emotional, physical, and sexual IPV measures were adapted from the World Health Organisation’s instrument (World Health Organization 2000). Girls were asked about victimisation by a male partner and boys were asked about perpetration against a female partner. Follow up questions after each section of emotional, physical, and sexual IPV were asked to ascertain whether the current partner was the perpetrator (for girls) or victim (for boys) of these behaviours. Three dichotomous items were created from these items to indicate victimisation (for girls) or perpetration (for boys) of each subtype of IPV.

Health outcomes

Health outcomes measures included safe sex practice, depression, and substance use. Participants were asked whether they had used a condom correctly the last time they had sex with their partner. This item was used as the safe sex practice health outcome measure.

The Centre for Epidemiological Studies on Depression Scale (CES-D) was used to assess depressive symptomology (Radloff 1977). This screening tool consists of 20 items about feelings and behaviours typical of depression with a four-point response scale ranging from Rarely or Never to Most or All of the Time. A score range from 0 to 60 was possible; participants who scored 16 and above were considered to have significant depressive symptomology and a dichotomous variable was created using this cut point. Cronbach’s alpha was 0.92 for girls and 0.89 for boys.

Problem drinking was measured using the AUDIT screening instrument (Saunders et al. 1993). A score of 8 or higher was used to differentiate those who had a drinking problem and those who did not (in the past year). Cronbach’s alpha was 0.88 for girls and 0.71 for boys. Participants were asked whether they had ever used drugs including marijuana, benzene, methaqualone, injectable drugs, or any other drug; a dichotomous variable was created to differentiate participants who had ever used any drug from those who had never used any drug.

Data management and analyses

All statistical analyses were conducted in STATA 12. All procedures used in data analysis took into account the study design, viewing the baseline study as a stratified, two stage survey with participants clustered within towns/villages. Descriptive analyses were conducted stratified by participant sex and relationship quality assessment. Means were used for continuous variables and percentage and frequencies were used for categorical variables (e.g., how many participants had experienced IPV). Means were compared using T-tests and proportions using chi-square tests. In all analyses, estimates were carried out using standard methods for estimating standard errors from complex multistage sample surveys (Taylor linearization).

Logistic regression analyses were used to show the associations between relationship factors and relationship assessment (the outcome). All logistic regression models adjusted for age, SES, Childhood Trauma, and study site. Models were built separately for girls and for boys. In each case, models began with the inclusion of the control variables and all relationship measures. Systematic backward elimination of relationship variables was applied (if p>0.05) to arrive at the most parsimonious model for girls’ and for boys’ relationship quality assessment.

Separate logistic regression analyses for girls and boys were built to investigate whether relationship quality assessment is an independent variable associated with health outcomes or behaviours: condom use at last sex with main partner, significant depressive symptoms, and substance use (problem drinking in the past year and ever drug use). These models included terms for age, SES, Childhood Trauma, and site.

Results

Participants

Participants were Black, isi-Xhosa speaking young people between the ages of 15 and 18 years (mean = 17.09 among girls, mean = 17.23 among boys). From the full baseline sample, 493 girls and 654 boys age 19 years and older were excluded. A further 159 girls and 38 boys were excluded because they did not have an intimate or dating relationship in the past year and another 6 girls and 33 boys were excluded because they did not provide a response for the relationship quality variable. These exclusions left 757 girls and 642 boys.

Descriptive data stratified by sex and relationship assessment are presented in Table 1 and summarised here. Most girls (n=749, 98.9%) were currently attending high school; 49.1% had completed grade 9 and 34.4% had completed grade 10. Similarly, most boys (n=635, 98.9%) were still in school; 47.2% indicated they had completed grade 9 and 25.4% grade 10. On average, girls and boys experienced some but not extreme levels of childhood adversity; most common were experiences of corporal punishment at home. Approximately one fifth of girls reported currently experiencing depression, but very few reported problem drinking or ever using drugs. Among the boys, 13.7% were experiencing depression, almost one quarter met criteria for problem drinking, and over one third had ever used drugs.

Table 1.

Descriptive data for sociodemographic and relationship descriptor variables, depression, and substance use stratified by relationship assessment

	Girls (N=757)				Boys (N=642)
	Total (n=757)	Positive Assessment (n=575)	Negative Assessment (n=182)		Total (n=642)	Positive Assessment (n=576)	Negative Assessment (n=66)
	Mean (SE)	Mean (SE)	Mean (SE)	p	Mean (SE)	Mean (SE)	Mean (SE)	p
Age (years)	17.09 (0.04)	17.12 (0.04)	17.02 (0.06)	0.260	17.23 (0.03)	17.26 (0.04)	17.18 (0.11)	0.439
SES	−0.02 (0.07)	−0.06 (0.08)	0.09 (0.11)	0.289	0.09 (0.09)	0.13 (0.09)	−0.07 (0.20)	0.209
Childhood Trauma	0.02 (0.05)	−0.02 (0.05)	0.13 (0.08)	0.071	−0.04 (0.04)	−0.06 (0.06)	0.001 (0.10)	0.633
Age difference between partners (years)	3.18 (0.10)	3.27 (0.10)	2.92 (0.17)	0.058	1.10 (0.05)	1.11 (0.06)	1.08 (0.17)	0.865
Relationship Equity	0.04 (0.05)	0.10 (0.06)	0.13 (0.08)	0.192	0.11 (0.05)	0.09 (0.06)	0.30 (0.17)	0.114
Communication Openness with Partner	−0.03 (0.05)	−0.07 (0.05)	0.11 (0.07)	0.017	0.11 (0.05)	0.08 (0.06)	0.35 (0.19)	0.079
Communication about Sexual and Reproductive Health	0.04 (0.04)	0.09 (0.04)	−0.12 (0.07)	0.003	−0.19 (0.47)	−0.14 (0.05)	−0.43 (0.15)	0.069
Relationship Conflict	0.002 (0.07)	−0.02 (0.08)	0.06 (0.15)	0.660	0.007 (0.04)	−0.03 (0.05)	0.30 (0.18)	0.059
	% (n)	% (n)	% (n)	p	% (n)	% (n)	% (n)	p
Partner Educationa	30.5% (233)	33.7% (194)	21.4% (39)	0.002	69.9% (449)	71.4% (411)	57.6% (38)	0.036
Mutual main partnership	82.2% (627)	88.0% (506)	66.5% (121)	0.000	86.0% (552)	89.8% (517)	53.0% (35)	0.000
Quarrelling with partner
Never	19.8% (150)	19.5% (112)	20.9% (38)	0.125	29.9% (192)	30.7% (177)	22.7% (15)	0.000
Rarely	46.9% (355)	45.9% (264)	50.0% (91)		45.8% (294)	46.9% (270)	36.4% (24)
Sometimes	27.1% (205)	29.0% (167)	20.9% (38)		17.9% (115)	17.4% (100)	22.7% (15)
Often	6.2% (47)	5.6% (32)	8.2% (15)		6.4% (41)	5.0% (29)	18.2% (12)
Emotional IPV	26.2% (190)	25.6% (140)	28.9% (50)	0.429	21.6% (137)	20.9% (119)	28.6% (18)	0.125
Physical IPV	26.4% (197)	27.6% (156)	22.9% (40)	0.232	17.9% (113)	17.1% (97)	25.4% (16)	0.087
Sexual IPV	10.2% (75)	9.1% (51)	13.9% (24)	0.082	4.0% (25)	4.1% (23)	3.2% (2)	0.746
Had sex with main partner	96.3% (725)	96.3% (550)	96.2% (175)	0.902	83.5% (533)	84.5% (484)	75.4% (49)	0.114
Correct condom use at last sex with main partner	38.4% (291)	39.1% (225)	36.3% (66)	0.404	29.3% (180)	30.9% (171)	14.8% (9)	0.010
Depression (significant symptomology)	20.5% (155)	20.2% (116)	21.4% (39)	0.687	13.7% (88)	13.4% (77)	16.7% (11)	0.470
Problem drinking	4.6% (35)	3.7% (21)	7.7% (14)	0.019	24.5% (157)	24.7% (142)	22.7% (15)	0.731
Drug use	6.1% (46)	4.9% (28)	9.9% (18)	0.007	38.2% (245)	36.8% (212)	50.0% (33)	0.041

^aRefers to girls’ partners who have at least completed high school or boys’ partners who have at least a primary level schooling

On average, girls were dating boys who were approximately three years older than themselves and many boyfriends had not completed high school. Most girls reported that their relationship was mutually a main partnership with their boyfriend. Average relationship equity tended toward the lower, more controlling end of the scale. Girls’ average communication scores suggested good communication skills and relatively open SRH communication. Although the majority of girls had sex with their partner, only 38.4% (n=291) used a condom the last time they had sex with their partner. Two thirds of the girls reported that they never or only rarely quarrelled with their partner and average relationship conflict indicated lower levels of conflict. A considerable proportion of girls (n=286, 39.78%) reported experiencing IPV within their relationship.

A large proportion of girls (n=575, 80.0%) reported a positive (“excellent” or “good”) assessment of their current relationship. The positive assessment group differed significantly from the negative assessment group in that they had better communication skills, better educated partners, and more mutual main partnerships. Reports of relationship equity, relationship conflict, quarrelling, and all forms of IPV victimisation were similar in the positive and negative assessment groups. More girls in the negative assessment group reported problem drinking and drug use in comparison to the positive assessment group; whereas proportions of depressed girls were the same in both groups.

Overall, boys reported being approximately one year older than their girlfriends and most of these girlfriends had completed some high school grades. Most boys reported their relationship was a mutual main partnership and that they had sex with this partner, but less than a third reported using a condom the last time they had sex with this girlfriend. Quarrels were reportedly absent or rare in most boys’ relationships and average relationship conflict tended toward the lower end of the range. Average relationship equity tended slightly toward the lower, more controlling end of the scale. Boys’ average communication scores suggested that although they were fairly open in discussing SRH with their partners, they generally reported quite poor communication skills. In terms of IPV perpetration, overall 199 boys (31.4%) reported perpetrating some form of IPV against their current or recent partner.

Most boys (n=576, 89.7%) reported a positive assessment of their current relationship. In comparison to those negatively assessing their relationship, boys in the positive assessment group had better educated partners, more mutual main partnerships, quarrelled less with their partner, and more of these boys used a condom correctly the last time they had sex with their girlfriend. Boys in both positive and negative assessment groups reported similar levels of relationship equity, communication skills and openness, relationship conflict, IPV perpetration, and sex. Fewer boys in the positive assessment group reported ever using drugs in comparison to those in the negative assessment group; depression and problem drinking were similar in both groups.

Factors associated with relationship quality assessment

Table 2 presents the models of factors associated with relationship assessment for girls and boys. Among girls, better relationship quality was associated with having a more educated partner, communicating better about SRH matters with their partner, and being in a relationship where both partners are one another’s main partner. Factors from girls’ backgrounds also influenced relationship assessment. Specifically, girls from higher SES backgrounds and those exposed to more trauma in childhood were less likely to assess their relationship positively.

Table 2.

Factors associated with positive relationship assessment among girls and boys

	Girls			Boys
	OR	95% CI	p	OR	95% CI	p
Age (years)	1.145	0.926–1.414	0.237	1.172	0.909–1.464	0.383
SES	0.846	0.710–0.995	0.018	1.116	0.935–1.357	0.320
Childhood Trauma	0.842	0.710–0.990	0.055	1.071	0.777–1.367	0.671
Communication about Sexual and Reproductive Health	1.245	1.067–1.420	0.018
Mutual main partnership	3.670	2.488–5.633	0.000	6.986	3.993–12.065	0.000
Quarrelling with partner
Never				Reference group
Rarely				0.991	0.472–1.884	0.981
Sometimes				0.629	0.264–1.524	0.273
Often				0.276	0.099–0.653	0.009
Partner Education	2.069	1.380–3.142	0.001

Among boys, only two relationship factors were associated with relationship quality assessment: quarrelling often with a partner decreased the likelihood of a positive relationship assessment, and a mutual main partnership increased the likelihood of a positive relationship assessment. In contrast to the girls, boys’ background factors did not significantly influence their current relationship assessment. For both boys and girls, relationship equity, communication openness with partner, relationship conflict, intimate partner violence, having sex with partner, and age difference between partners were not associated with relationship quality assessment.

Health outcomes

Table 3 presents the adjusted odds ratios and confidence intervals for the association between girls’ and boys’ relationship assessment and various health outcomes. Among girls, relationship quality was not associated with correct condom use at last sex with the main partner, depressive symptomology, or problem drinking. A decreased likelihood of drug use was associated with girls’ positive relationship assessment. Among boys, positive relationship assessment was associated with correct condom use with the main partner and a decreased likelihood of drug use. There was no association with depressive symptomology or problem drinking.

Table 3.

AORs of Positive relationship assessment associations with other health outcomes

OUTCOMES	Girls			Boys
	OR	95% CI	p	OR	95% CI	p
Correct condom use at last sex with main partner (vs. incorrect or no use)	1.169	0.857–1.613	0.400	2.492	1.216–5.291	0.015
Depression (vs. no depression)	0.934	0.651–1.415	0.752	0.764	0.372–1.501	0.464
Problem drinking (vs. no problem drinking)	0.583	0.241–1.111	0.164	1.061	0.555–2.045	0.845
Drug use (vs. no drug use)	0.521	0.280–0.881	0.053	0.518	0.338–0.991	0.020

Discussion

Although girls’ and boys’ concepts of good relationships shared one common value, relationship primacy, their models were different. For girls, a good relationship was one that was a mutual main partnership with an educated boyfriend in which there was good, open communication. For boys, a good relationship was one that was a mutual main partnership with very little quarrelling. Girls and boys who believed they were in good relationships were less likely than those in poor relationships to report ever using drugs and boys in these relationships were additionally more likely to report using a condom at last sex with their main partner.

The different models of relationship quality suggest that what girls’ and boys’ values in relationships differs. This difference is unsurprising given the gendered nature of young people’s relationships (Harrison et al. 2001a, Harrison et al. 2001b, O’Sullivan 2005, Harrison 2008, Jewkes et al. 2010, Gevers et al. 2012). Qualitative research has found that young people ideally desire monogamy but, for a variety of reasons, often have multiple, hierarchical partnerships (Gevers et al. 2012, Jewkes and Morrell 2012). This study’s findings suggest that teenagers value being the most important, or main, partner.

Among girls, having a more educated partner may indicate not just educational attainment, but also greater age and a greater likelihood of him earning money. These factors increase his status; the girls themselves would gain peer status based on their partner’s characteristics and these status rewards may influence how much girls like and value their relationships especially in a context were peers play a significant role (Gevers et al. 2012, Jewkes and Morrell 2012). The general preference for older partners among girls has been described in qualitative research and also found to intersect with other characteristics older teenage girls find desirable in a partner, namely sexual and romantic competence, money, maturity, and respectability despite the potential risks of these older partners also being controlling or violent (Jewkes and Morrell 2012). However, this pattern is not a reference to the ‘sugar daddy’ phenomenon discussed in other research with young women (Silberschmidt and Rasch 2001, Jewkes and Abrahams 2002, Kaufman and Stavrou 2004, Luke 2005) because boyfriends were, on average, approximately three years older than the girls in this study. Of concern is previous research that found that having a partner three or more years older and having an educated partner are significant risks for HIV among these young women perhaps because these men can use their higher status and increased mobility to get more partners and thus have larger sexual networks (Jewkes et al. 2006a). Although education may also improve skills and knowledge that improve relationships, it is concerning that a feature that adolescent girls value in a partner may also be a significant risk factor to their health and well-being; therefore, interrogating relationship quality concepts in interventions is recommended.

Open sexual communication has been associated with overall relationship satisfaction among young adults in the USA (Montesi et al. 2011) and was similarly found to be associated with positive relationships among girls in this study. Personal characteristics such as confidence and good SRH knowledge may influence girls’ abilities and likelihood to discuss SRH matters with partners and the ability to discuss and negotiate SRH matters may signify a more supportive and open partner and result in safer sex practices.

Boys seem particularly averse to quarrelling with their partners. Patriarchal norms may explain this association; that is, men may expect to be unchallenged leaders in their relationships with women and they may interpret quarrelling as a challenge to their authority (Jewkes and Morrell 2012). Indeed, boys in this study reported relatively low levels of gender equity in their relationships. Further, other research with this population has noted that most boys had conservative gender attitudes (Shai et al. 2012). Given the link between gender inequity and IPV and sexual risk behaviour (Jewkes 2002, Jewkes et al. 2010, Shai et al. 2012), gender transformative interventions to improve boys’ attitudes and perceptions of gender relations and building their conflict resolution skills are needed.

It is notable that IPV did not play a significant role in girls’ or boys’ relationship assessment. This trend may be owing to violence being fairly common and suggests social acceptance of such violence (Jewkes and Morrell 2012). This finding does not suggest that teens like or want such violence, but perhaps they are unable to challenge widely accepted IPV nor the patriarchal norms supporting it and they have little access to recourse. The social importance and value of dating relationships among teenagers may also contribute to acceptance of IPV.

Open, assertive communication skills and relationship equity are often promoted in intervention programmes (e.g., Foshee et al. 1998, Jewkes et al. 2006b, Jewkes et al. 2011) especially in efforts to prevent IPV. Yet, there was no significant association between these factors and positive relationship quality assessment. Nevertheless, the role of these skills and attitudes in IPV prevention are important.

Having sex with a partner was not a significant factor in both girls’ and boys’ models of relationship quality, similar to the finding in the USA (Welsh et al. 2005). Yet, studies have described the importance of sex in adolescent intimate relationships (Harrison 2008, Selikow et al. 2009, Gevers et al. 2012). Qualitative research findings have discussed the desire, among young women, for good sex (Jewkes and Morrell 2012) so it is possible that although sex itself does not influence relationship assessments, sexual pleasure may play a role. A majority of participants in this sample were having sex with their partners so the lack in variance may have influenced this finding.

The analysis of associations between relationship assessment and health outcomes is limited by the inability to establish the direction of the association. That is, poor mental health and risk behaviour may lead to poorer relationships or a poor relationship may lead to or exacerbate mental health problems and risk behaviour. Drug use among young people in this community is particularly uncommon for girls and socially disapproved for boys, so drug-using teens may be marginalised and vulnerable in many ways that may impact on their intimate relationships.

Positive relationship assessment was associated with condom use at last sex among the boys. Consistent condom use within this study group was found to be rare and condom use patterns were associated with patterns in young men’s gender attitudes, IPV perpetration, sexual risk-taking, and SES (Shai et al. 2012). Of concern, inconsistent condom users, compared to consistent and never users, had a high-risk profile in terms of IPV perpetration and risky sexual behaviour (Shai et al. 2012). It is likely that consistent users were more likely to report condom use at last sex in this study; therefore, these boys are likely more equitable and less violent and this demeanour may impact positively on their relationships. This finding further suggests that boys can be happy with their relationship and use condoms which contrasts with qualitative research findings that suggests condoms are often perceived as decreasing the love and trust between partners (Gevers et al. 2012, Shai et al. 2012). It is also possible that boys in positive relationships have better social and relationship skills including safer SRH practices. Interventions should challenge ideas about condoms restricting relationship quality; indeed, condom use may indicate valuing sexual health in the relationship, which may contribute to a more positive relationship.

There are several limitations to this study. The data are cross-sectional and thus we cannot make conclusions on causality or investigate changes in relationship assessment. It is possible that some of the associations between relationship characteristics or behaviour and relationship assessment are bi-directional. For example, young people who have better social and relationship skills are likely to engage in more open, less quarrelsome relationship behaviours which contribute to a positive relationship and the assessment itself may motivate adolescents to engage in continuous health-promoting behaviours. Poor relationships may contribute to depression and substance use, but people with depression or substance use problems may also have cognitive biases and skill deficits that contribute to a negative relationship.

Another limitation is the homogenous study sample (black, isiXhosa-speaking South Africans) and so generalisability to the broader population may be limited. In addition, the study relied on a non-probability sample of volunteers who knew the intervention content and which intervention arm their cluster was assigned to and some girls were prevented from participating by their strict parents (Jewkes et al. 2008). Personal characteristics (e.g., extroversion, social anxiety) and relationship characteristics (e.g., a partner who disapproved of participation or a relationship that was being conducted in secret) may also have influenced participation. Very conservative communities, and some conservative families within amenable study communities, declined participation in the study suggesting that the sample may be under representative of people with very conservative attitudes to gender and adolescent relationships. The developmental differences between 15/16 year olds and 17/18 year olds may impact on their relationship conceptualisations and constructions; however, an age stratified analysis was not possible. The large majority of the sample were 17/18 year olds and therefore the findings presented here are likely more reflective of older adolescents. This study included both peri-urban and rural, male and female young people in the large sample and similarities between qualitative reports from this study area and other studies (Gevers et al. 2012) suggests that there may be several similarities in young people’s ideas about intimate relationships in different areas of the country.

Self-report measures in any study run the risk of biased or inaccurate reporting and this risk may increase when measuring SRH and IPV behaviour. Participants’ anonymity and confidentiality were assured in this study to allay anxieties. The extent of violent and other risk behaviour reported in this sample suggest that participants were generally comfortable enough to report these behaviours and, therefore, inaccuracies in the data are more likely to reflect an underestimation. The exploratory analyses were limited by the variables available in the Stepping Stones study dataset and there may be other factors that influence adolescents’ assessments of their relationship quality. Further, there may be nuances in the relationship factor variables that are not captured by the measurement tools used in this study. In particular, the relationship quality assessment relied on a response to one, face-valid item. Nevertheless, as noted earlier, this measurement approach allowed for a unique analysis of factors that influence teenagers’ relationship assessments.

Conclusions

These results have given insight into factors important to young people themselves in their dating relationships. Some of the factors influencing young people’s relationship assessments may pose some risk (e.g., having educated partners or avoiding discussions for fear of beginning to quarrel) and the lack of influence of important issues (e.g., IPV) suggest that interventions should encourage teens to critically evaluate their personal and group concepts of good and poor relationships. For example, challenging norms and attitudes accepting of IPV and promoting gender equitable attitudes and relationships may impact on adolescents’ relationship expectations and subsequent assessments. In addition to promoting healthier, pro-social attitudes and norms, interventions should equip young people with skills to develop and maintain good, healthy relationships and how to deal with unsatisfactory or poor relationships. These skills will help young people to use their dynamic relationship assessments to guide their behaviour and choices within their relationships.