Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2017 Jan 1.
Published in final edited form as: Soc Sci Med. 2015 Nov 22;148:25–33. doi: 10.1016/j.socscimed.2015.11.024

Cultural consensus modeling to measure transactional sex in Swaziland: Scale building and validation

Rebecca Fielding-Miller a,*, Kristin L Dunkle b, Hannah LF Cooper c, Michael Windle c, Craig Hadley c
PMCID: PMC4698209  NIHMSID: NIHMS741758  PMID: 26647365

Abstract

Transactional sex is associated with increased risk of HIV and gender based violence in southern Africa and around the world. However the typical quantitative operationalization, “the exchange of gifts or money for sex,” can be at odds with a wide array of relationship types and motivations described in qualitative explorations. To build on the strengths of both qualitative and quantitative research streams, we used cultural consensus models to identify distinct models of transactional sex in Swaziland. The process allowed us to build and validate emic scales of transactional sex, while identifying key informants for qualitative interviews within each model to contextualize women’s experiences and risk perceptions. We used logistic and multinomial logistic regression models to measure associations with condom use and social status outcomes. Fieldwork was conducted between November 2013 and December 2014 in the Hhohho and Manzini regions. We identified three distinct models of transactional sex in Swaziland based on 124 Swazi women’s emic valuation of what they hoped to receive in exchange for sex with their partners. In a clinic-based survey (n=406), consensus model scales were more sensitive to condom use than the etic definition. Model consonance had distinct effects on social status for the three different models. Transactional sex is better measured as an emic spectrum of expectations within a relationship, rather than an etic binary relationship type. Cultural consensus models allowed us to blend qualitative and quantitative approaches to create an emicly valid quantitative scale grounded in qualitative context.

Keywords: Swaziland, Southern Africa; HIV/AIDS; transactional sex; consensus modeling; mixed methods; scale validation

Introduction

The term “transactional sex” emerged approximately two decades ago to differentiate between sex work and relationships that focus on sexual-economic exchange but are not perceived as ‘commercial’ by either party (Chatterji et al., 2005; Cole, 2004; Dunkle et al., 2004b; Groes-Green, 2013; Leclerc-Madlala, 2003). In southern Africa transactional sex increases a woman’s risk of HIV by up to 50% and is significantly associated with intimate partner violence (IPV) (Dunkle et al., 2010; Dunkle et al., 2007; Dunkle et al., 2004b; Dunkle et al., 2006; Jewkes et al., 2010). Transactional sex is typically operationalized as “the exchange of sex for gifts or money” (Luke et al., 2011; Swidler & Watkins, 2007). To borrow epedemiological language, this definition is sensitive but not specific. Many relationships contain both some degree of economic dependence and an expectation of sex. While the risks inherent in transactional relationships are often clear in context, “the exchange of sex for gifts or money” could conceivably capture behaviors ranging from sex work to receiving an engagement ring (Brinig, 1990). Women whose sexual relationships are their primary source of economic support may identify the relationship as transactional or commercial, or they may reject those labels and the associated stigma (Cole, 2004; Dunkle et al., 2010; Groes-Green, 2013; Stoebenau, 2009; Stoebenau et al., 2013; Stoebenau et al., 2011). Rather than a binary measurement, transactional sex may be better operationalized as a characteristic present in almost all relationships as a matter of degree (Maganja et al., 2007; Stoebenau et al., 2011; Swidler & Watkins, 2007; Wamoyi et al., 2011).

The nature of transactional sex varies widely across geographical contexts. In southern Africa, women may use transactional sex to pay their school fees, acquire basic survival goods such as food or shelter, access fashionable commercial goods to improve their status amongst peers, or some combination of all of these (Groes-Green, 2013; Luke et al., 2011; Mojola, 2015; Stoebenau et al., 2011). While the act of receiving goods from a sexual partner does not pose health risks in itself, transactional relationships are likely to reflect strong gender and economic power imbalances, making it difficult for women to negotiate condom use or sexual encounters (Dunkle et al., 2004a). Transactional sex may also hurt or help a woman’s social standing, as these relationships exist within a network of social and financial obligations (Cole, 2004; Groes-Green, 2013, 2014; Stoebenau et al., 2013; Stoebenau et al., 2011; Swidler & Watkins, 2007; Wamoyi et al., 2011; Wamoyi et al., 2010). Peers or family may encourage women to charm their partners for more gifts and financial support, and those who receive nothing in exchange for sex can be mocked as “prostitutes” (Wamoyi et al., 2011, p.9) who have devalued themselves or failed to support their family (Groes-Green, 2013, 2014; Wamoyi et al., 2011). Access to fashionable goods may improve social status, however being seen as promiscuous may result in being cutoff from material help in times of need (Kaschula, 2011; Masvawure, 2010; Strebel et al., 2013).

Despite the importance and social normalcy of economic support from a partner, women must be careful that their relationships are perceived to be motivated by affection, rather than financial gain. Those who do not may be called materialistic, and risk being cut off from support in times of need (Bandali, 2011; Fielding-Miller et al., 2014; Kaschula, 2011; Miller et al., 2011; Stoebenau et al., 2011; Strebel et al., 2013). In a US study, fewer than 10% of women who reported initiating or staying in a relationship because of financial concerns agreed that they had ever exchanged sex for money (Dunkle et al., 2010). Measuring transactional sex based on what women actually receive from their partners, and how they weigh these items when considering their sexual obligations, would more accurately capture the degree of transaction inherent in the relationship than asking women to identify with a possibly stigmatized motive.

Social science and public health research has at its disposal a large toolkit of methodologies, drawing from a range of epistemologies, paradigms, and research traditions, to better understand the context, prevalence, and correlates of health behaviors and risks such as those associated with transactional sex (Creswell, 2013; Rosenthal & Rosnow, 1991; Schutt, 2004; Tebes, 2005). Etic perspectives allow researchers to compare concepts and behaviors across cultures using a universal “outside” definition, whereas emic perspectives emphasize the local understandings and meanings of a phenomenon (Harris, 1976). While neither emic nor etic perspectives are the exclusive domain of a particular research approach, qualitative research tends to approach research questions from an emic perspective, while quantitative projects more often utilize the latter. Similarly, while no method or research project draws exclusively from a single paradigm, in broad terms quantitative approaches tend to draw on a positivist tradition to measure prevalence and infer generalizable associations, while qualitative approaches originated from interpretivist traditions and typically utilize textual data to understand a research question from local, emic perspectives (Harris, 1976; Hennink et al., 2011; Schutt, 2004; Tebes, 2005).

The traditional operationalization of transactional sex stems from an international, etic, perspective that can be quantitatively compared across cultures. However it does not necessarily reflect women’s lived experiences and fails to capture the influence of economic considerations on sexual relationships as a matter of degree. Bridging the gap between the rich streams of qualitative and quantitative research on transactional sex requires a quantitative tool which measures these relationships in a way that captures women’s understanding of their own behavior, is grounded in the pressures of the social landscape, and is capable of measuring associations with both health behaviors and potential social risks and benefits.

Cultural consensus modeling is a systematic measurement approach that moves from interpretive, emic description to positivist, quantitative measurement within one study design (Dressler, 2005; Weller, 2007). The researcher uses rapid ethnographic methods to define the boundaries of a set of knowledge or behaviors shared by a group - a cultural domain -followed by quantitative analysis of numerical data generated in the ethnographic phase. The final product is a cultural consensus model (CCM), an emicly valid operationalization of a cultural domain that can be used in quantitative studies (Weller, 2007).

The consensus analysis process assumes that if informants answer a question about their culture (rather than their personal tastes) in a similar way, they do so because they are drawing on a shared cognitive domain, or realm of cultural knowledge. Cultural consensus analysis (CCA) is essentially an exploratory factor analysis that uses participants, rather than items, as variables of interest and identifies clusters of similar answer patterns. An answer key and competence score for each individual can then be generated. The answer key identifies clusters of similar answers and assumes that these answers are similar – and therefore emicly correct -- because participants who share knowledge of a domain answer similarly, while the answers of those without knowledge of this domain will be more scattered. Competence scores reflect the number of culturally correct answers each participant gave, and range from 0 – 1.00. Participants with a competence score of 1.00 are assumed to have perfect knowledge of a domain and would likely make reliable key informants (Hruschka & Maupin, 2013; Romney et al., 1986). Consonance, ie how much an individual aligns with the values or behaviors that relate to the domain, can be determined in a second sample of individuals by assessing whether or to what degree individuals endorse a value or enact a behavior identified. For further details see Romney, Weller, and Batchelder (1986).

When conducting CCA an eigenvalue ratio greater than 3.0 between participant answer clusters suggests that participants are drawing on a single dominant CCM. An eigenvalue ratio below 3.0 suggests that no single dominant CCM exists within the sample and participant answers are drawing from two or more CCMs.

Our objective was to build an emicly valid quantitative scale of transactional sex that was responsive to both the social and health consequences of transactional sex and based on concrete behaviors rather than subjective assessments of motive. To do this, we used cultural consensus modeling to first create a scale of transactional sex, and then validated our scale by measuring its association with social standing and condom use compared to the etic definition. We used an iterative series of research questions:

  1. What items do women hope to get in exchange for sex?

  2. How are these valued in exchange for sex?

  3. Are there distinct subgroups that value items differently?

  4. How do these distinct subgroups differ from one another?

  5. How does level of participation in a CCM affect social status and condom use?

Methods

Setting

Swaziland is a small absolute monarchy in southern Africa with a population of approximately 1.2 million (DHS, 2007). Two thirds of Swazis live on less than $1.25 a day and 25% are food insecure ("Swaziland loses AGOA benefits," 2014; Spaull, 2013; WFP, 2009). Swaziland has an adult HIV prevalence of approximately 31% and the Swazi government has highlighted gender inequality and transactional sex as drivers of the epidemic (Bicego et al., 2013; NERCHA, 2014).

Ethical considerations

Ethics approval was granted by the Emory University IRB, the Swaziland Scientific and Ethics Committee, traditional leadership at rural study sites, and head nurses at clinical sites. All participants provided written informed consent. Preliminary results were disseminated at the community and national level at the conclusion of fieldwork.

Study design

Drawing on the qualitative notion of research as a cyclical process, we collected and analyzed data iteratively (figure 1) (Creswell, 2013; Hennink et al., 2011). We used cultural consensus analysis to build and characterize transactional sex CCMs, in-depth interviews (IDIs) to learn about Swazi women’s perceptions of transactional sex and risk in their own relationships, and a clinic-based survey to measure the association between transactional sex CCMs, social standing, and condom use. Data were collected between November 2013 and October 2014.

Figure 1.

Figure 1

Open in a new tab

Order of study phases, with main research question and sample size.

What do women hope to get in exchange for sex?

To compile a list of items that Swazi women hoped to get in exchange for sex we discussed the etic definition of transactional sex with local colleagues and created a free-list question that preserved the intent but allowed participants to express the concept emicly: “What do Swazi women get, or hope to get, in exchange when they have sex with a man?”

We recruited a convenience sample of women from university, rural, urban, and peri-urban sites, as the literature suggested that women living in different areas may have different priorities. Swazi research assistants (RAs) and the first author approached women in public spaces (ie, public shopping areas in urban and peri-urban sites, and in roads and by a river in rural sites) and asked them to list as many items as they could think of in English or siSwati in response to the free-list question. The meaning was left to participants’ interpretation: an “item” could be emotional (love), tangible (a cell phone), or relational (marriage). Researchers purposively sought women with a range of education, age, and marital statuses, although there was no specific quota, as the literature suggests that these may influence what items women would hope to get. Free-list responses were condensed to a single master list based on frequency of mention, discussion with local colleagues and experts, and theoretical interest. Sample size was determined based on iterative checks for data saturation.

How are items valued in exchange for sex?

We used a rating activity to understand the importance of receiving an item in exchange for sex. A second convenience sample was approached in the same way and same locations as before. For each item women were asked, “On a scale of 1–5, how important is it to get this item in exchange when having sex with a man?” Researchers emphasized that the question related to Swazi women in general rather than the participant’s individual preference. Age, education, recruitment site, and marital status were recorded for each participant. After the rating activity we asked women if they were interested in participating in a longer personal interview. If they agreed we recorded their contact details and linked these to rating data using an anonymous unique ID.

Are there distinct transactional sex CCMs?

We used CCA, multiple regression quadratic assignment procedure (MRQAP), exploratory factor analysis (EFA), and bivariate tests of significance to identify distinct transactional sex CCMs, the most correct way of rating each item within a CCM, and individuals who were highly competent within CCMs.

How do CCMs differ from one another qualitatively?

IDIs with key informants (KIs) were used to elicited personal narratives of engaging in a particular transactional sex CCM. KIs were women who had participated in the rating activity, had high cultural competence scores within a CCM, and had agreed to a follow-up interview. A bilingual female Swazi RA trained in qualitative interview methods conducted all interviews in siSwati. Interviews were recorded, transcribed, translated, and analyzed using comparative analysis (Creswell, 2013; Hennink et al., 2011).

How does participation in a CCM affect social status and condom use?

We conducted an audio computer-assisted self-interview (ACASI) survey with pregnant women accessing antenatal care in one urban and one rural public clinic. Antenatal clinics were chosen because this project is nested within a larger study examining transactional sex and HIV risk. 95% of Swazi women experience at least one lifetime pregnancy, 97% of these access care at least once, and all pregnant women accessing antenatal care receive HIV screenings at all clinic visits (DHS, 2011; Ministry of Health, 2011). A female Swazi RA approached all women waiting in line for antenatal services. The survey took place in a private area with the RA present to assist women who were unfamiliar with laptops or had questions. The survey was translated into siSwati, back translated into English, and administered in siSwati.

Women were asked to provide a pseudonym for up to 5 partners in the past year. For each partner, women were asked which of the list items they had received in the last year and if they had used a condom the last time they had sex. To assess the social benefits or consequences of transactional sex, women were asked to place themselves on the MacArthur scale of subjective social status, an internationally validated measure of community social status (Brown et al., 2008; Hu et al., 2005; Operario et al., 2004; Yip & Adler, 2005). To assess community perceptions of promiscuity, women were asked if they knew or thought that they had ever been called a nasty name that implied mercenary or promiscuous sexuality with a list of examples. Terms for the latter item were based on qualitative data and included local equivalents of “slut” and “gold digger.” Other measures included socio-economic indicators, lifetime number of sexual partners, and age of sexual debut. HIV status was recorded from test results of that day’s clinic visit.

We asked if women had been motivated to have sex with their most recent partner because of poverty, hunger, for basic groceries, or for the sake of her children. Women who answered affirmatively to any of these were coded as having engaged in ‘etic transactional sex.’

To assess how strongly women conformed to a transactional sex CCM we built a weighted scale based on the culturally correct rating for each item within a CCM. Items were weighted according to the value established by CCA. A participant’s consonance with a CCM was equivalent to the total weighted value of each item she had received from her most recent sexual partner. Participants who had received more items that were rated more highly were considered more consonant with the CCM.

We examined the bivariate associations between CCM consonance and subjective social status, and CCM consonance and condom use at last sex. For each CCM we built a logistic regression model to assess the adjusted odds ratio (aOR) of condom use at last sex by CCM consonance, and a multinomial logistic regression model to assess the adjusted relative risk ratio (aRRR) of subjective social status according to CCM consonance. To test the sensitivity of our scale against a more traditional, binary definition predicated on motive we built a final logistic regression model examining the aOR of condom use or high social status for women who self-identified engaging in etic transactional sex.

Results

What do Swazi women hope to get in exchange for sex?

47 women were recruited for the free-list activity. Mean age was 28.0 (SD 8.9, range 18 – 40) with mean education of 11.1 years (SD 1.9, range 6 – 13). Participants named 177 unique items. These were condensed into a single master list of 31 items based on frequency of item mention, discussion with local colleagues, and theoretical interest (table 1).

Table 1.

Items that a woman may hope to receive from her partner, with weighted value by group

Aspirational Inkhosikati University
Item Score Item Score Item Score
*fun night out 4.5 *basic food 3.9 love 4.2
*phone 4.4 *alcohol 3.4 marriage 4.1
*airtime 4.1 *phone 3.2 *provide for children 3.9
*clothes 3.9 *clothes 2.9 label clothes 3.9
*toiletries 3.7 *airtime 2.7 *fashionable clothes 3.7
*hairstyle 3.7 *cosmetics 2.6 expensive clothes 3.7
*restaurant 3.6 *fashionable clothes 2.5 *nice lifestyle 3.7
*alcohol 3.6 *toiletries 2.5 enjoyment 3.6
label clothes 3.6 *hairstyle 2.4 *phone 3.6
*basic food 3.6 *provide for child 2.2 *clothes 3.6
expensive clothes 3.6 *provide for family 2.1 *rent 3.5
*takeaway 3.6 marriage 2.1 *shoes 3.5
enjoyment 3.5 label clothes 2.0 *hairstyle 3.3
*fashionable clothes 3.5 sexual satisfaction 2.0 sexual satisfaction 3.3
sexual satisfaction 3.4 *nice lifestyle 1.9 *job 3.2
*rent 3.4 *fun night out 1.9 *provide for family 3.0
*shoes 3.4 enjoyment 1.8 *car 2.9
*jewelry 3.3 *rent 1.7 status 2.8
*transport 3.2 love 1.7 *fun night out 2.7
status 3.1 status 1.6 *basic food 2.6
*place to sleep for the night 3.0 *jewelry 1.6 *jewelry 2.6
*nice lifestyle 2.8 school fees 1.5 *cosmetics 2.5
*cosmetics 2.8 *takeaway 1.3 school fees 2.5
*job 2.2 *job 1.3 *restaurant 2.4
*provide for family 2.1 *place to sleep for the night 1.2 *transport 2.1
love 1.8 expensive clothes 1.2 *airtime 2.0
*car 1.2 *shoes 1.0 *toiletries 1.8
*provide for child 1.2 *car 0.8 *alcohol 1.2
school fees 1.1 *transport 0.7 children 0.8
children 0.8 *restaurant 0.6 *place to sleep for the night 0.7
marriage 0.7 children 0.5 *takeaway 0.5
Open in a new tab
*

Retained in ACASI consonance scale

Are there distinct CCMs that value items differently?

77 women rated the 31 list items. Mean age was 27.4 (SD 8.1, range 18–57) and mean years of education were 10.9 (SD 2.5, range 0 – 13).

To identify emic CCMs we used UCINET software to perform a cultural consensus analysis on rating data (Borgatti S.P., 2002). Our initial eigenvalue ratio for the full sample was 1.37, with an average competence score of 0.31 suggesting that the 77 participants were drawing from more than one transactional sex CCM (Borgatti S.P., 2002; J. Hruschka, Sibley, LM., Kalim, N., Edmonds, JK., 2008).

To identify clusters of participants who rated items similarly, we conducted exploratory factor analysis (EFA) with the full sample (n = 77) in Stata 13, using participants as the variables of interest (StataCorp, 2013). We constrained Stata to 4 total factors and assigned group membership based on the EFA group in which an individual had the highest factor loading score. We then created four separate UCINET data sets according to EFA results. Group size ranged from n=12 to n=27 (see table 2). We conducted a second round of CCA in UCINET with each group data set to determine if the four groups identified by EFA were drawing on distinct transactional sex CCMs.

Table 2.

Identified models of transactional sex with descriptive demographics

Full sample Aspirational Inkhosikati University
n 77 12 19 17
Eigenvalue ratio NA 3.61 3.03 4.55
Age
mean (SD)		 27.42 (8.14) 27 (7.27) 31.44† (11.43) 24.41 (3.24)
Education*
mean (SD)		 10.90 (2.48) 11.33 (2.19) 10.21 (2.37) 12.29 (1.36)††
Married
% (n)		 29.87 (23) 16.67 (2) 47.37 (9) 11.76 (2)
Rural*
% (n)		 55.84 (43) 41.67 (5) 68.42 (13) 29.41 (5) †
University** %
(n)		 32.47 (25) 50.00 (6) 15.79 (3) 64.71 (11)††
Open in a new tab

Significant ANOVA:

*

p< .05

**

p<.01

Significant ttest or chi-square vs. other consensus groups:

†

p< .05

††

p<.01

Three groups had eigenvalue ratios greater than 3.0, indicating distinct transactional sex CCMs. The sample size and average competence score for each of these three groups were sufficient to assume with 95% confidence that participants had classified 95% of items correctly (S. C. Weller & Romney, 1988). The fourth had an eigenvalue ratio of 2.59, suggesting these participants were not drawing from a unique CCM. Participants from this fourth group were more likely to have been recruited from peri-urban sites with high work migration, making it likely they were drawing on multiple cultural models. Based on IDI and demographic data, we named the 3 significant groups “inkhosikati” (a traditional Swazi term denoting a respected senior woman), “aspirational,” and “university.” We elaborate on each of these group descriptors below.

How do CCMs differ demographically?

To identify distinct demographic characteristics within groups, we first used a multiple regression quadratic assignment procedure via double dekker semi-partialling (MRQAP) in UCINET (Borgatti S.P., 2002) on the full 77 participant sample. This is a matrix regression that determines the amount and significance of answer variability explained by demographic characteristics that participants have in common as compared to participants who do not share that characteristic.

We assessed variability in list item ratings according to the following matrices of binary variables: Participant recruited in rural area, participant not recruited in rural area, participant married, participant not married, participant had grade 12 education, participant had no grade 12 education, participant was over 25, and participant was not over 25. Being recruited from a rural area accounted for 11% variability in participant’s rating of list items (p = 0.001), while not being recruited from a rural area explained 6% of variability (p = 0.02), implying that being from a rural area has a larger effect on women’s tendency to rate items similarly than being from a non-rural area. Being married explained 4.8% of variability in answer patterns (p = 0.04), while women who were not married did not rate items in significantly similar ways. Neither age nor education significantly explained variance in item rating patterns in the full sample.

We next compared demographic characteristics across the 4 identified groups. Place of recruitment and education level were significantly different across groups in ANOVA analyses. In chi-square and t-tests women in the inkhosikati group were on average 5.3 years older (p = 0.02) and those in the university group had 1.8 more years of education (p < 0.01) and were more likely to have been recruited at an urban or university site (table 2).

How are items valued in exchange for sex?

CCA generated an answer key of the culturally correct rating of exchange items within each CCM, along with participant competence scores. Answer keys reflect which items were considered most important within each transactional sex CCM and are shown in table 1.

How do CCMs differ qualitatively?

We recruited in-depth interview participants based on competence scores within each of the three significant CCMs.

We interviewed 3 aspirational, 3 inkhosikati, and 5 university informants who were part of the previous rating sample, had agreed to a follow-up interview, and provided contact details. All emphasized the importance of both love and financial demonstrations of affection in their relationships. Inkhosikati relationships reflected those of married, older, more rural women with slightly less formal education. Marriage conferred both social and economic stability, and married women were highly respected in their communities. Inkhosikatis linked a partner’s ability to provide financial support with the courtship process and family approval. They felt a man should provide for his family and that marriage improved social status for both partners. Women in the aspirational group, who were more likely to be urban and have at least a secondary education, were predominantly characterized by the aspiration to either transform their relationship into marriage, or to mirror a marriage relationship as closely as possible. These informants were practical about their partner’s economic role in their lives. They felt it was his responsibility to provide for them and gifts were expected when they met for sex, especially if marriage was impossible because he already had a wife or primary partner. University informants were predominantly young women recruited from the residential university site. They linked love to gift giving but were not immediately interested in marriage, and were more likely to declare an independent ideology in which they did not need a man, but still liked to receive gifts from a boyfriend as a sign of affection. Some relationships were primarily motivated by economic gain. These could be affectionate but were also likely to be hidden from family and all but certain friends. Family and friends were more likely to approve of a relationship if money and high status items were shared and the man had the potential for a strong economic future. When asked “what does it mean to be a respected Swazi woman?” across all CCMs, participants agreed that feminine respectability was closely linked to sexual behavior and financial resources. According to participants, “materialistic” women were condemned, and participants suggested that “loose” women would have difficulty accessing financial assistance from their community in times of need. All informants agreed that according to traditional Swazi culture, all women should aspire to marriage, and marriage represented the height of feminine respectability (table 3). Further details on qualitative findings can be seen elsewhere (Fielding-Miller et al., 2015).

Table 3.

In-depth interview quotes, perspectives on different relationship models and the notion of feminine respectability in Swaziland

Aspirational Inkhosikati University What does it mean to be a
respectable Swazi woman?
Quote You develop a mentality that you’re
going to get what you’re having sex
for… It is what I am in need of that
particular time, what you have both
agreed on, what he promised to
give…maybe it’s airtime, maybe we
agreed that you would buy me that
airtime when we sleep together…
You don’t sleep with someone you do
not love; you only sleep with someone
you love.		 He would give me money just so I
could be able to buy myself
toiletries or clothing… we would
sleep together, it would be my way
of saying thank you. It was basically
that …At home they did not want
to understand, because he has a
poor family background, so they
felt like I had no future with him.
But I loved him and today he is my
husband. And they accepted him --
everything is fine.		 I do it because I love this person and he
promised me heaven and earth… The fact
that he loves me and he promised to give me
everything I need and mostly he loves
me…that he will love me till the end and one
day we will get married and be husband and
wife… I don’t have that mentality of being
dependent on other people… I do ask him for
some money if I need to do my hair… He
knows that every month he has to give me
money to do my hair and buy myself anything
I need.		 She is respected by the way she dresses up
and the manner in which she talks to
people…It has to be in a respectable way like
she doesn’t dress in clothes which will reveal
her body and she has to look so smart … a
woman has to know her status and if she is
still playing some games then she has to stop
doing so…the cheating games, having more
than one boyfriend and if she is married
having secret lovers, you have to be straight.
-Nobuhle
- Thembisile
36 years old, grade 11 education, not
married.
Consonance score: 0.77		 - Sibongile
28 years old, grade 9 education,
married.
Consonance score 0.78		 - Nolwazi
21 years old, university student, not married.
Consonance score 0.81		 20 years old, grade 12 education, not
married.
Quote I started this relationship not based on love
but I wanted his money… the allowance we
get here is not enough to get us through the
whole year. He is understanding. He can see
that I’m a single parent and I am still at
school, he helps me a lot when I tell him my
problem. The only problem I have is that I
have to have sex with him whereas I don’t
love him at all… my friends they were
supportive of this relationship because they
also get to have a share.		 I have to respect myself and respect those
that I live with so that… when I am in need
of something they can help me…. You have
to know other people and humble yourself
before other people… if you are known in
the community that you have money, of
course they will want to get closer to you,
because they know that if they need help
they will get it from you.
-Zamokuhle 28 years old, grade 9 education
-Thandeka
19 years old, university student, not married.
Consonance score 0.76
Open in a new tab

How does CCM participation affect social status and condom use?

For the ACASI survey we recruited 406 pregnant women. Each participant was assigned a score assessing her consonance with each CCM, determined by how many items she had received from her partner and how valuable these items were according to each weighted CCM scale. The original 31 items were further collapsed into 22 items to account for redundancy (“fashionable” vs “label” clothes) and non-material items (“love” or “marriage”).

Each of the 3 distinct CCM answer keys were used to construct weighted scales. If in the ACASI survey a woman reported receiving a cell phone, basic groceries, and basic clothes from her partner her score would be 11.85 (3.57+4.36+3.92) out of a possible 92.12 in the aspirational model, 10 (3.93 + 3.17 + 2.90) out of a possible 58.4 within the inkhosikati CCM, and 9.79 (2.58+3.61+3.60) out of a possible 87.77 in the university model. The participant would then be considered 13%, 17%, and 11% consonant with each model, respectively, suggesting that her actions – or the type of items she received from her partner – were most consistent with the inkhosikati transactional sex CCM. We converted each CCM scale into a standardized Z-score to compare analyses across groups. Each participant had 3 unique scale scores corresponding to the 3 unique scales.

ANC survey demographics are shown in table 4. Mean participant age was approximately 25 years, and fewer than half (47%) of participants had ever held employment. Thirteen percent of participants reported engaging in ‘etic’ transactional sex in the past 12 months. Thirty-four percent of participants were living with HIV.

Table 4.

ANC Survey Sample Demographics

n mean (SD) range % (n)
Consonance Score 401
   “Inkhosikati”
      raw 11.41 (8.24) 0.00 – 40.71
      standardized 0 (1) −1.38 – 3.56
   Aspirational
      raw 16.23 (12.60) 0.00 – 67.03
      standardized 0 (1) −1.29 – 4.03
   University
      raw 13.97 (10.61) 0.00 – 57.78
      standardized 0 (1) −1.32 – 4.13
“Etic” Transactional sex 401 12.97 (52)
Condom use at last sex (yes) 403 42.68 (172)
Social status quartiles 357
   1–2 31.37 (112)
   3–4 19.33 (69)
   5–6 27.17 (97)
   7–10 22.13 (79)
Ever called a nasty name 394
   no 75.13 (296)
   yes 24.87 (98)
Lifetime sexual partners 372 2.75 (5.14) 1–87
Age 406 24.54 (4.99) 18–42
Rural 403 54.09 218
Education 406
   none 3.45 (14)
   primary 24.14 (98)
   secondary 64.53 (262)
   University or technical college 7.88 (32)
Ever done work to earn money 405 53.33 (216)
HIV status 375
   positive 34.13 (128)
   negative 65.87 (247)
Open in a new tab

In bivariate t-tests, higher transactional sex CCM consonance scores were associated with lower condom use for aspirational and inkhosikati scales, but not the university model. Women who reported using a condom at last sex scored 0.2 units lower on both the inkhosikati and aspirational scales (p=0.04, both). In simple linear regression women placed themselves 0.32 rungs higher on the social status ladder for each standardized increase in the inkhosikati scale (p = 0.02), 0.43 rungs higher for each increase in the aspirational scale (0.002), and 0.38 rungs higher per standardized unit in the university scale (p = 0.006). Women who had engaged in etic transactional sex in the last 12 months placed themselves 0.89 rungs lower (p = 0.03).

Results for logistic and multinomial logistic regression models are shown in table 5. For every increase in standardized CCM Z-score, women were approximately 25% less likely to have used a condom at last sex across all scales.

Table 5.

aOR of condom use at last sex and aRRR of social status by CCM consonance quartile. Validated against etic definition of transactional sex

Inkhosikati consonance
(Z-score)		 Aspirational consonance
(Z-score)		 University consonance
(Z-score)		 Etic Definition
aOR (95% CI) aOR (95% CI) aOR (95% CI) aOR 95% CI
Logistic Regression 1
Condom use at last sex 0.73 (0.58 – 0.93) 0.76 (0.60 – 0.96) 0.76 (0.60 – 0.97) 0.78 (0.40 – 1.52)
Multinomial Logistic Regression 2 aRRR (95% CI) aRRR (95% CI) aRRR (95% CI) aOR 95% CI
Subjective social status
    Quartile 1 Ref Ref Ref Ref
    Quartile 2 1.17 (0.80 – 1.70) 1.23 (0.83 – 1.82) 1.23 (0.83 – 1.82) 0.80 (0.31 – 2.07)
    Quartile 3 1.36 (0.96 – 1.94) 1.44 (1.00 – 2.06) 1.49 (1.04 – 2.15) 0.39 (0.14 – 1.14)
    Quartile 4 1.41 (1.00 – 2.00) 1.51 (1.05 – 2.17) 1.52 (1.06 – 2.19) 0.89 (0.35 – 2.24)
Open in a new tab
*

Controlling for woman’s age, HIV status, education, having ever earned own income, lifetime number of sexual partners, and ever called a nasty name

Subjective social status was collapsed into a categorical variable by quartile to account for non-normality. Women who placed themselves in the highest status quartile scored 0.41 units higher on the inkhosikati consonance scale than women who placed themselves in the lowest quartile, although the lower end of the confidence interval overlapped 1.00. According to the aspirational consonance scale, women were marginally more likely to rate themselves in the 3rd quartile of social status as the number of items received from a partner increased (aRRR: 1.44, 95% CI: 1.00 – 2.06), and 50% more likely to place themselves in the top quartile of their community for each unit increase in items received (95% CI: 1.05 – 2.17). The university consonance scale was significantly correlated with an increased likelihood of placing oneself in the top 3rd and 4th quartiles (aRRR 1.49, 95% CI: 1.04 – 2.15, and aRRR 1.52, 95% CI: 1.06 – 2.19).

The binary etic definition of transactional sex had no association with condom use at last sex or subjective social status.

Discussion

We identified three dominant CCMs through which Swazi women conceptualize sexual-economic exchange. These had three distinct relationships to social status and condom use, which would not have been apparent using only the traditional measure. Measuring transactional sex using emic scales based on CCMs created an operationalization that was more responsive to health behavior and social effects than the etic binary measurement. Rather than a subjective declaration of her motives, each CCM scale relied on what women had actually received from a partner, weighted in a way that was relevant to her social context. Conducting IDIs with culturally competent KIs from each CCM provided contextual data on the reasons behind our quantitative findings based on the experiences of women whom we know to be experts within a model.

Understanding transactional sex as a scale rather than a binary status variable promotes its conceptualization as a spectrum that manifests differently in different relationship types, rather than a categorical one-dimensional identity. This has implications for interventions that seek to minimize relationship risk, but describe relationship characteristics that women may not identify or agree with, versus those that seek to support women who want to exit relationships with aspects they consider harmful (Baird et al., 2012; Dunbar et al., 2014; Tawfik & Watkins, 2007).

Our findings suggest that studies intending to measure the prevalence or associations of transactional sex may be better served if they adopt two main modifications: 1) Ask participants about items actually received from a partner, rather than relationship motive, to address the potential for social desirability bias resulting from participants’ reluctance to admit to stigmatized behavior; and 2) Consider measuring transactional sex as a scale reflecting a spectrum of sexual-economic entanglement, rather than a binary relationship category. Further research is needed to determine the potential of developing a simplified list of items received that can be used across the region, or the possibility of developing regional, rather than country-specific, item based scales for cross country comparisons.

Measuring transactional sex using 3 different scales highlighted the divergent social risks and benefits across relationship models. All women must walk a careful tightrope to manage social and sexual reputations, financial needs, and financial obligations (Cole, 2004; Groes-Green, 2013; Kohler et al., 2007; Swidler & Watkins, 2007). Higher consonance with a transactional sex CCM was significantly associated with higher social status for aspirational and University models, with a more pronounced effect in the University model. This finding is consistent with ethnographic studies that have explored the role of consumer goods as symbolic capital among University students in particular, and some subsets of urban women in general (Cole, 2004; Masvawure, 2010). Conversely, the insignificant association between higher inkhosikati consonance and social status can be explained by our qualitative findings, which suggest that that within these models social status is derived from the relationship itself – marriage – rather than the individual items received from a partner. For all women, however, being perceived as promiscuous or materially driven can severely compromise social status. KIs confirmed that financial support or a high status partner may increase a woman’s social standing, but being seen as ‘loose’ or a ‘gold digger’ can harm her.

Increased gifts or support from a partner decreased condom use to a similar degree in all 3 relationship models, while the binary measure was not sensitive to condom use. While this may be a result of uneven power dynamics, as previous research suggests (Miller et al., 2011), our qualitative data suggests that financial support and gifts within a relationship are just as likely to demonstrate love and serious intent as economic coercion, and some women may choose to eschew condoms out of trust, affection, or even a desire for mutual pleasure, rather than a lack of agency.

The use of a cross-sectional clinic-based sample to validate our scale presents several limitations. The link between transactional sex consonance and condom use at last sex may be bi-directional: More gifts may decrease women’s desire to use condoms with a partner, decreased condom use may increase a partner’s gift giving or financial support, or both cases may be simultaneously true. A clinic-based sample may not be generalizable to the full population. Comparisons with a national household sample (SHIMS, 2012) and national ANC sentinel surveillance data (provided privately by Swazi Ministry of Health) suggest that women without primary or tertiary education may have been undersampled, along with women over the age of 45 (analyses not shown). While asking participants to report condom use at last sex is likely a more accurate measure than condom use frequency overall, the very high proportion of pregnant women reporting condom use at last sex suggests that this measure was subject to social desirability bias. The clinic sample may also not necessarily match the convenience sample from which we derived our scales, however, average age and education were approximately similar to the freelist and rating convenience samples. Additionally, the eigenvalue ratio and average competence scores suggested that our rating sample was drawing from a known cultural model, and the derived scales were sensitive to both health behaviors and social correlates, suggesting that despite the disparate samples we have managed to capture and measure salient cultural domains. Despite these shortcomings, triangulating our qualitative and quantitative findings allows us to suggest possible causal links between transactional sex, condom use, and social status, and to report the link between transactional sex and decreased condom use with some confidence.

Conclusion

We measured transactional sex using three distinct emic scales that weighted items differently depending on the priorities of different relationship models. We found that the association between condom use and transactional sex, as well as between social status and transactional sex, varied depending on relationship model, and that all 3 scales were more sensitive to condom use and social status than a more traditional definition. While our scale is specific to Swaziland, the methods used were rapid and inexpensive, and are easily transferrable to other cultural settings. Similar approaches have been used globally to identify competing models of health knowledge, measure the health effects of cultural consonance, and define the process and effects of acculturation from one model to another (Brewis, 2006; Broesch & Hadley, 2012; Hruschka, Sibley, Kalim, Edmonds, 2008; Sweet, 2010). Consensus modeling can create an emicly grounded measure of a construct that is epidemiologically meaningful but difficult to operationalize. Because this approach bridges the divide between interpretivist and positivist approaches, CCMs are in important tool for areas of research, such as transactional sex, in which emic constructions of meaning and quantifiable risk surveillance are equally necessary.

Research Highlights.

  • Transactional sex relationships represent a spectrum of sexual-economic obligation

  • Cultural consensus modeling identified 3 emic Swazi scales of transactional sex

  • Our emic scales were more sensitive to health outcomes than an etic binary

  • Women’s social status increased with increasing emic scale scores for all 3 models

Acknowledgements

We thank profusely the 559 Swazi women who shared their time, wisdom, and experiences with us over the course of this project. We thank Mphumi Ncongwane, Nonhlanhla Mazibuko, Cynthia Vilakati, and Samkeliso Sikhosana for their insights and skills as research assistants and translators. We would also like to thank Blythe Terrell for her assistance with research sites and local introductions. This study was supported by the Swaziland National Emergency Response Council on HIV/AIDS. Financial and material support was provided by a US Fulbright Grant, the Health Community Capacity Collaborative, Swaziland, and the Center for AIDS Research at Emory University (P30AI050409). Support in preparing this manuscript was provided by NIDA grant T32 DA 023356.

Footnotes

Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

References

  1. Baird SJ, Garfein RS, McIntosh CT, Oezler B. Effect of a cash transfer programme for schooling on prevalence of HIV and herpes simplex type 2 in Malawi: a cluster randomised trial. Lancet. 2012;379:1320–1329. doi: 10.1016/S0140-6736(11)61709-1. [DOI] [PubMed] [Google Scholar]
  2. Bandali S. Exchange of sex for resources: HIV risk and gender norms in Cabo Delgado, Mozambique. Culture Health & Sexuality. 2011;13:575–588. doi: 10.1080/13691058.2011.561500. [DOI] [PubMed] [Google Scholar]
  3. Bicego GT, Nkambule R, Peterson I, Reed J, Donnell D, Ginindza H, et al. Recent patterns in population-based HIV prevalence in Swaziland. Plos One. 2013;8:e77101. doi: 10.1371/journal.pone.0077101. [DOI] [PMC free article] [PubMed] [Google Scholar]
  4. Borgatti SPEMG, Freeman LC. Ucinet for Windows: Software for Social Network Analysis. Harvard MA: Analytic Technologies; 2002. [Google Scholar]
  5. Brewis A, Gartin M. Biocultural construction of obesogenic ecologies of childhood: Parent-feeding versus child-eating strategies. American Journal of Human Biology. 2006;18:203–213. doi: 10.1002/ajhb.20491. [DOI] [PubMed] [Google Scholar]
  6. Brinig MF. Rings and promises. Journal of Law Economics & Organization. 1990;6:203–215. [Google Scholar]
  7. Broesch J, Hadley C. Putting culture back into acculturation: Identifying and overcoming gaps in the definition and measurement of acculturation. The Social Science Journal. 2012;49:375–385. [Google Scholar]
  8. Brown RA, Adler NE, Worthman CM, Copeland WE, Costello EJ, Angold A. Cultural and community determinants of subjective social status among Cherokee and White youth. Ethnicity & health. 2008;13:289–303. doi: 10.1080/13557850701837302. [DOI] [PMC free article] [PubMed] [Google Scholar]
  9. Chatterji M, Murray N, London D, Anglewicz P. The factors influencing transactional sex among young men and women in 12 sub-saharan African countries. Social Biology. 2005;52:56–72. doi: 10.1080/19485565.2002.9989099. [DOI] [PubMed] [Google Scholar]
  10. Cole J. Fresh contact in Tamatave, Madagascar: Sex, money, and intergenerational transformation. American Ethnologist. 2004;31:573–588. [Google Scholar]
  11. Creswell JW. Qualitative inquiry and research design : choosing among five approaches. Los Angeles: SAGE Publications; 2013. [Google Scholar]
  12. DHS. Swaziland Demographic and Health Survey. Mbabane, Swaziland: Central Statistics Office, Macro International; 2007. [Google Scholar]
  13. DHS. Uganda Demographic and Health Survey. Demographic and Health Survey. Kampala, Uganda: Uganda Bureau of Statistics, Measure DHS, ICF International; 2011. [Google Scholar]
  14. Dressler B, Balieiro dos Santos. Measuring cultural consonance: Examples with specia reference to measurement theory in anthropology. Field Methods. 2005;17:331–355. [Google Scholar]
  15. Dunbar MS, Dufour MSK, Lambdin B, Mudekunye-Mahaka I, Nhamo D, Padian NS. The SHAZ! Project: Results from a Pilot Randomized Trial of a Structural Intervention to Prevent HIV among Adolescent Women in Zimbabwe. Plos One. 2014;9:20. doi: 10.1371/journal.pone.0113621. [DOI] [PMC free article] [PubMed] [Google Scholar]
  16. Dunkle Jewkes, Brown Gray, McIntryre Harlow. Transactional sex among women in Soweto, South Africa: prevalence, risk factors and association with HIV infection. Social Science & Medicine. 2004a;59:1581–1592. doi: 10.1016/j.socscimed.2004.02.003. [DOI] [PubMed] [Google Scholar]
  17. Dunkle K, Wingood GM, Camp CM, DiClemente RJ. Economically Motivated Relationships and Transactional Sex Among Unmarried African American and White Women: Results from a US National Telephone Survey. Public Health Reports. 2010;125:90–100. doi: 10.1177/00333549101250S413. [DOI] [PMC free article] [PubMed] [Google Scholar]
  18. Dunkle KL, Jewkes R, Nduna M, Jama N, Levin J, Sikweyiya Y, et al. Transactional sex with casual and main partners among young South African men in the rural Eastern Cape: Prevalence, predictors, and associations with gender-based violence. Social Science & Medicine. 2007;65:1235–1248. doi: 10.1016/j.socscimed.2007.04.029. [DOI] [PMC free article] [PubMed] [Google Scholar]
  19. Dunkle KL, Jewkes RK, Brown HC, Gray GE, McIntryre JA, Harlow SD. Transactional sex among women in Soweto, South Africa: prevalence, risk factors and association with HIV infection. Social Science & Medicine. 2004b;59:1581–1592. doi: 10.1016/j.socscimed.2004.02.003. [DOI] [PubMed] [Google Scholar]
  20. Dunkle KL, Jewkes RK, Nduna M, Levin J, Jama N, Khuzwayo N, et al. Perpetration of partner violence and HIV risk behaviour among young men in the rural Eastern Cape, South Africa. AIDS. 2006;20:2107–2114. doi: 10.1097/01.aids.0000247582.00826.52. [DOI] [PubMed] [Google Scholar]
  21. Fielding-Miller R, Dunkle K, Shai NJ, Hadley C, Windle M, Cooper HLF. She is doing the wrong things for the right reasons: Respectability and risk in transactional relationships. (manuscript in preparation) 2015 [Google Scholar]
  22. Fielding-Miller R, Mnisi Z, Adams D, Baral S, Kennedy C. “There is hunger in my community”: a qualitative study of food security as a cyclical force in sex work in Swaziland. BMC Public Health. 2014;14:1–10. doi: 10.1186/1471-2458-14-79. [DOI] [PMC free article] [PubMed] [Google Scholar]
  23. Groes-Green C. "To put men in a bottle": Eroticism, kinship, female power, and transactional sex in Maputo, Mozambique. American Ethnologist. 2013;40:102–117. [Google Scholar]
  24. Groes-Green C. Journeys of patronage: moral economies of transactional sex, kinship, and female migration from Mozambique to Europe. Journal of the Royal Anthropological Institute. 2014;20:237–255. [Google Scholar]
  25. Harris M. History and significance of emic-etic distinction. Annual Review of Anthropology. 1976;5:329–350. [Google Scholar]
  26. Swaziland Ministry of Health. Improving the Quality of Maternal and Neonatal Health Services in Swaziland: A Situational Analysis. UNICEF, WHO, UNFPA: Ministry of Health World Bank; 2011. [Google Scholar]
  27. Hennink MM, Hutter I, Bailey A. Qualitative research methods. London; Thousand Oaks, Calif.: SAGE; 2011. [Google Scholar]
  28. Hruschka Maupin. Competence, Agreement, and Luck: Testing Whether Some People Agree More with a Cultural Truth than Do Others. Field Methods. 2013;25:107–123. [Google Scholar]
  29. Hruschka J, Sibley LM, Kalim N, Edmonds JK. When there is more than one answer key: Cultural theories of postpartum hemmorhage in Matlab, Bangladesh. Field Methods. 2008;20:315–337. [Google Scholar]
  30. Hu P, Adler NE, Goldman N, Weinstein M, Seeman TE. Relationship between subjective social status and measures of health in older Taiwanese persons. Journal of the American Geriatrics Society. 2005;53:483–488. doi: 10.1111/j.1532-5415.2005.53169.x. [DOI] [PubMed] [Google Scholar]
  31. Jewkes RK, Dunkle K, Nduna M, Shai N. Intimate partner violence, relationship power inequity, and incidence of HIV infection in young women in South Africa: a cohort study. Lancet. 2010;376:41–48. doi: 10.1016/S0140-6736(10)60548-X. [DOI] [PubMed] [Google Scholar]
  32. Kaschula. Using people to cope with the hunger: social networks and food transfers amongst HIV/AIDS afflicted households in KwaZulu-Natal, South Africa. AIDS Behav. 2011;15:1490–1502. doi: 10.1007/s10461-011-0006-z. [DOI] [PubMed] [Google Scholar]
  33. Kohler HP, Behrman JR, Watkins SC. Social networks and HIV/AIDs risk perceptions. Demography. 2007;44:1–33. doi: 10.1353/dem.2007.0006. [DOI] [PubMed] [Google Scholar]
  34. Leclerc-Madlala S. Transactional sex and the pursuit of modernity. Social Dynamics-a Journal of the Centre for African Studies University of Cape Town. 2003;29:213–233. [Google Scholar]
  35. Luke N, Goldberg RE, Mberu BU, Zulu EM. Social Exchange and Sexual Behavior in Young Women's Premarital Relationships in Kenya. Journal of Marriage and Family. 2011;73:1048–1064. doi: 10.1111/j.1741-3737.2011.00863.x. [DOI] [PMC free article] [PubMed] [Google Scholar]
  36. Maganja RK, Maman S, Groves A, Mbwambo JK. Skinning the goat and pulling the load: transactional sex among youth in Dar es Salaam, Tanzania. Aids Care-Psychological and Socio-Medical Aspects of Aids/Hiv. 2007;19:974–981. doi: 10.1080/09540120701294286. [DOI] [PubMed] [Google Scholar]
  37. Masvawure T. 'I just need to be flashy on campus': female students and transactional sex at a university in Zimbabwe. Culture Health & Sexuality. 2010;12:857–870. doi: 10.1080/13691050903471441. [DOI] [PubMed] [Google Scholar]
  38. Miller Bangsberg, Tuller Senkungu, Kawuma Frongillo, et al. Food Insecurity and Sexual Risk in an HIV Endemic Community in Uganda. Aids and Behavior. 2011;15:1512–1519. doi: 10.1007/s10461-010-9693-0. [DOI] [PMC free article] [PubMed] [Google Scholar]
  39. Mojola SA. Material girls and Material love: Consuming femininity and the contradictions of post-girl power among Kenyan schoolgirls. Continuum. 2015:1–12. doi: 10.1080/10304312.2015.1022949. [DOI] [PMC free article] [PubMed] [Google Scholar]
  40. NERCHA. The Extended National Multisectoral HIV and AIDS Framework (eNSF) 2014–2018. Mbabane, Swaziland: 2014. [Google Scholar]
  41. Operario D, Adler NE, Williams DR. Subjective social status: Reliability and predictive utility for global health. Psychology & Health. 2004;19:237–246. [Google Scholar]
  42. President Obama removes Swaziland, reinstates Madagascar for AGOA Benefits. Washington DC: Office of the United States Trade Representative Executive Office of the President; 2014. [Google Scholar]
  43. Romney AK, Weller SC, Batchelder WH. Culture as consensus - a theory of culture and informant accuracy. American Anthropologist. 1986;88:313–338. [Google Scholar]
  44. Rosenthal R, Rosnow RL. Essentials of Behavioral Research: Methods and Data Analysis. New York: McGraw-Hill Publishing Company; 1991. The Nature of Behavioral Research; pp. 3–25. [Google Scholar]
  45. Schutt RK. Investigating the Social World: The Process and Practice of Research. Thousand Oaks: Pine Forge Press; 2004. Evaluation Research; pp. 310–336. [Google Scholar]
  46. Spaull N. Poverty & privilege: Primary school inequality in South Africa. International Journal of Educational Development. 2013;33:436–447. [Google Scholar]
  47. StataCorp. Stata Statistical Software: Release 13. College Station, TX: StataCorp LP; 2013. [Google Scholar]
  48. Stoebenau K. Symbolic capital and health: The case of women's sex work in Antananarivo, Madagascar. Social Science & Medicine. 2009;68:2045–2052. doi: 10.1016/j.socscimed.2009.03.018. [DOI] [PubMed] [Google Scholar]
  49. Stoebenau K, Nair RC, Rambeloson V, Rakotoarison PG, Razafintsalama V, Labonte R. Consuming sex: the association between modern goods, lifestyles and sexual behaviour among youth in Madagascar. Globalization and Health. 2013:9. doi: 10.1186/1744-8603-9-13. [DOI] [PMC free article] [PubMed] [Google Scholar]
  50. Stoebenau K, Nixom S, Rubincam C, Willan S, Zembe Y, Tsikoane T, et al. More than just talk: the framing of transactional sex and its implications for vulnerability to HIV in Lesotho, Madagascar and South Africa. Globalization and Health. 2011:7. doi: 10.1186/1744-8603-7-34. [DOI] [PMC free article] [PubMed] [Google Scholar]
  51. Strebel A, Shefer T, Potgieter C, Wagner C, Shabalala N. 'She's a slut … and it's wrong': Youth constructions of taxi queens in the Western Cape. South African Journal of Psychology. 2013;43:71–80. [Google Scholar]
  52. Swaziland HIV Incidence Measurement Survey (SHIMS) First Findings Report. Mbabane, Swaziland: ICAP, PEPFAR, Swaziland Ministry of Health; 2012. [Google Scholar]
  53. Sweet E. "If your shoes are raggedy, you get talked about": Symbolic and material dimensions of adolescent social status and health. Social Science and Medicine. 2010;70:2029–2035. doi: 10.1016/j.socscimed.2010.02.032. [DOI] [PubMed] [Google Scholar]
  54. Swidler A, Watkins SC. Ties of dependence: AIDS and transactional sex in rural Malawi. Studies in Family Planning. 2007;38:147–162. doi: 10.1111/j.1728-4465.2007.00127.x. [DOI] [PubMed] [Google Scholar]
  55. Tawfik L, Watkins SC. Sex in Geneva, sex in Lilongwe, and sex in Balaka. Social Science & Medicine. 2007;64:1090–1101. doi: 10.1016/j.socscimed.2006.10.002. [DOI] [PubMed] [Google Scholar]
  56. Tebes JK. Community science, philosophy of science, and the practice of research. American Journal of Community Psychology. 2005;35:213–230. doi: 10.1007/s10464-005-3399-x. [DOI] [PubMed] [Google Scholar]
  57. Wamoyi J, Fenwick A, Urassa M, Zaba B, Stones W. "Women's Bodies are Shops": Beliefs About Transactional Sex and Implications for Understanding Gender Power and HIV Prevention in Tanzania. Archives of Sexual Behavior. 2011;40:5–15. doi: 10.1007/s10508-010-9646-8. [DOI] [PubMed] [Google Scholar]
  58. Wamoyi J, Wight D, Plummer M, Mshana GH, Ross D. Transactional sex amongst young people in rural northern Tanzania: an ethnography of young women's motivations and negotiation. Reproductive Health. 2010:7. doi: 10.1186/1742-4755-7-2. [DOI] [PMC free article] [PubMed] [Google Scholar]
  59. Weller. Cultural consensus theory: Applications and frequently asked questions. Field Methods. 2007;19:339–368. [Google Scholar]
  60. Weller SC, Romney AK. Systematic data collection. Sage; 1988. [Google Scholar]
  61. WFP Swaziland Annual Report. Mbabane, Swaziland: WFP Swaziland Office; 2009. [Google Scholar]
  62. Yip W, Adler N. Does social standing affect health and happiness in rural China. Abstract presentation at 2005 International Health Economics Association World Congress; Barcelona, Spain. 2005. [Google Scholar]

RESOURCES