Skip to main content
PMC home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2017 Feb 9.
Published in final edited form as: Field methods. 2015 Aug 13;28(2):112–132. doi: 10.1177/1525822X15595343

Let’s Talk about Sex, Maybe: Interviewers, Respondents, and Sexual Behavior Reporting in Rural South Africa

Brian Houle 1,2,✉, Nicole Angotti 2,3,4, Samuel J Clark 4,5,8, Jill Williams 2,4, F Xavier Gómez-Olivé 4, Jane Menken 2,4, Chodziwadziwa Kabudula 4, Kerstin Klipstein-Grobusch 4,6, Stephen M Tollman 4,7,8
PMCID: PMC5300069  NIHMSID: NIHMS808912  PMID: 28190977

Abstract

Researchers are often skeptical of sexual behavior surveys: respondents may lie or forget details of their intimate lives, and interviewers may exercise authority in how they capture responses. We use data from a 2010–2011 cross-sectional sexual behavior survey in rural South Africa to explore who says what to whom about their sexual lives. Results show an effect of fieldworker age across outcomes -- respondents report “safer”, more “responsible” sexual behavior to older fieldworkers; and an effect of fieldworker sex -- men report more sexual partners to female fieldworkers. Understanding fieldworker effects on the production of sexual behavior survey data serves methodological and analytical goals.

INTRODUCTION

Researchers are often skeptical of sexual behavior surveys: respondents may lie (see Gribble et al. 1999 on reporting bias) or forget details of their intimate lives (see Graham et al. 2003 on recall bias), and interviewers may exercise authority in how they capture responses (see Randall et al. 2013 on interviewer power). Respondents might manage their “presentation of self” (Goffman 1959) during surveys by minimizing the gap between their own behavior and the perceived social values of the interviewer (Hewett, Mensch and Erulkar 2004). Concerns about reporting bias extend to myriad topics, such as political knowledge (Davis and Silver 2003) and economic attitudes (Groves and Fultz 1985). Given social expectations of what is considered “right”, sexual behavior reporting is likely subject to bias (Wellings et al. 2006).

Methodologists have examined influences on sexual behavior reporting, including interviewer characteristics such as sex (e.g., McCombie and Anarfi 2002), age (e.g., Ford and Norris 1997), and ethnicity (e.g., Becker, Feyisetan and Makinwa-Adebusoye 1995), as well as familiarity of the interviewer to the respondent (Weinreb 2006). Some studies used data collection techniques that obviate concerns about reporting sensitive topics to a live interviewer, including audio-assisted computer self-interviewing (ACASI) (e.g., Oloo et al. 2012), though it may be less effective with older and less educated participants (Reichmann et al. 2010). Methodological inquiries on collecting sensitive data are important in a setting like South Africa, where understanding risky sexual behaviors is critical for addressing high HIV prevalence (see Gómez-Olivé et al. 2013), and where our understanding of fieldworker influences on survey data is scant (for exceptions, see Bignami-Van Assche, Reniers and Weinreb 2003).

We use data from a 2010–2011 cross-sectional study on sexual behavior conducted in rural South Africa by 10 local fieldworkers (or “interviewers”). We test assumptions about social desirability bias to explore who says what to whom about their sexual lives. Based on our knowledge of the setting and relevant literature, we hypothesize that interviewer sex and age will impact reporting: 1) men and women will report more sexual partners to same, rather than opposite-sex interviewers; and 2) respondents will not reveal more sexual partners to older interviewers. With high HIV prevalence and wide-scale social marketing efforts promoting HIV prevention, we assume it is socially desirable to report fewer sexual partners and sexual behaviors that demonstrate “safe” or “responsible” sexual decision-making.

Setting

The Agincourt Health and Socio-Demographic Surveillance System (“AHDSS”) is a former apartheid homeland area in the Bushbuckridge sub-district of Mpumalanga Province in the northeast of South Africa. The AHDSS has been monitoring vital events annually since 1992 (Kahn et al. 2012). The 2011 population was around 90,000 people in 28 villages (420 km2 area). All villages in Agincourt have water provided through neighborhood taps, at least one primary school, and most have electricity and a secondary school (Kahn et al. 2012). Median years of education range from 6 – 11, and the most common jobs include construction for men and retail and domestic work for women (Blalock 2014). Labor migration is high, especially among men ages 35–50, of whom 60% live outside the study area for 6+ months per year (Kahn et al. 2012). The main ethnic group is amaShangaan and Christianity is the most prominent religion.

Ha Nakekela Study

In 2010–2011, we collected data on sexual behavior in the Ha Nakekela study (“We Care” in XiShangaan). The sample included 7,662 individuals permanently residing in the site, selected by sex and age (15+) stratified random sampling from the 2009 Census data.

All sampled participants were visited in their homes by a fieldworker and invited to participate in the study. There were 5 male and 5 female fieldworkers (ages 28–44) (mid-study, one male fieldworker left and was replaced by a female of a different age group). Fieldworkers were assigned to villages and households randomly by the field supervisor and met the respondent privately. If the interviewer knew the respondent, the interview was assigned to another fieldworker. The home visit lasted approximately 45 minutes and included written consent to participate (assent for minors), two risk behavior surveys (for chronic diseases and sexual behavior), and collection of biomarkers for HIV and cardiometabolic risk. The study received ethical approvals from University of the Witwatersrand Human Research Ethics Committee and the Mpumalanga Provincial Research and Ethics Committee.

DATA AND METHODS

We use data from the sexual behavior survey, which 4,684 participants from the 7,662 sampled answered. The remainder was ineligible (469), not located (2,156) or refused participation (353). The survey included respondents’ sexual practices over the past 2 years. For each sexual partner reported, interviewers asked about the nature of the relationship (e.g., casual partner; spouse), duration, and details about behavioral sexual risk (e.g., knowledge of partner’s HIV status, condom use, drug/alcohol use). Additional survey questions included lifetime number of sexual partners, HIV testing history, previous diagnosis for STIs, etc.

We modeled four outcomes to explore the effects of fieldworker and respondent characteristics on sexual behavior reporting. The subject of these measures are widely invoked in social marketing efforts to curb HIV transmission, and we believe highly sensitive and prone to social desirability (Edwards, Thomsen and Toroitich-Ruto 2005):

  1. Lifetime sexual partners: What is the total number of sexual partners you have had in your life?

  2. Condom use: Did you use a condom the last time you had sex with this partner?

  3. Discussing HIV: At the time you first had sex with this partner, had you ever discussed HIV with him/her?

  4. HIV testing: Have you ever been tested for HIV?

We modeled lifetime sexual partners using negative binomial regression; and condom use, discussing HIV, and HIV testing using logistic regression (with correlation among partners of the same respondent, when applicable). Our approach was to first model the outcome using respondent covariates we expected would influence the outcome but were not the focus of our research question about fieldworker effects. We next included fieldworker characteristics to assess their impact on sexual behavior reporting, and finally tested interactions between respondent and fieldworker characteristics.

For each outcome we first fit a base model with no fieldworker effects, including covariates of respondent’s age (mean centered) (see Aiken and West 1991), sex (male or not), village (indicators for each village excluding the referent), migration history (having resided outside study area for 6+ months in previous Census year), and quintiles of the respondent’s 2009 household socio-economic status. For partner data we also included the partner’s age (mean centered) and type of partner (girlfriend/boyfriend, casual, or anonymous; referent living together).

We then included fieldworker effects for age (dichotomized as <35 and 35+ years old) and sex (male or not). Next we tested interactions between respondent and fieldworker characteristics by first interacting respondent and fieldworker sex; then respondent and fieldworker age; and finally fieldworker age and respondent sex, and respondent age and fieldworker sex. Interactions that improved overall model fit (using likelihood ratio tests of the nested models) were included in the final models. There was little variation on other fieldworker characteristics: all had completed secondary school, were Xishangaan-speakers from the study site (per AHDSS job requirements), and, to the best of our knowledge, predominately Christian. We believe this to be a study strength: since the fieldworkers were similar on other characteristics, we are likely testing actual fieldworker (age and gender) effects. We can only assume, however, about reporting based on our results. Causal inference would require methodological strategies such as randomization or experimental designs.

We assessed model sensitivity to the fieldworker age cut-off by testing +/− 1 year differences. The results were not substantively different. Davis and colleagues note that one problem with studying interviewer age effects is that studies often dichotomize age in ways that obscure important generational differences (Davis et al. 2010). Our fieldworker age range (28–44) gives us variation for comparison, but does not categorize fieldworkers unrealistically. Our “older fieldworker” age category starts at age 35, which may seem young in some parts of the world. In Agincourt, however, life expectancy is 55 years for males, 62 years for females, and more than 50% of first births occur before age 20 (Kahn et al. 2012, Williams et al. 2013). We also assessed sensitivity to individual fieldworkers by testing all models with an indicator for each fieldworker, allowing us to detect if an outlying fieldworker(s) was influencing results. All models incorporated sampling (inverse probability) weights.

RESULTS

Respondent and fieldworker sample descriptives are presented in Table S1. The sample was balanced by sex, with a mean age of 41 years. Almost 60% of the sample had a previous migration history. Additional information presents average responses to sexual behavior survey questions. For fieldworkers, 45.5% were male and ages ranged from 28–44, with an average among those under 35 of 29.6 years and 39.2 years among those 35+.

Lifetime Sexual Partners

Table S2 (a) shows the results of the base negative binomial regression for number of lifetime sexual partners, without fieldworker effects. Including age-squared (p < 0.001) and age-cubed (p < 0.001), interacted with sex (p < 0.001), significantly improved model fit.

Figure S1 shows the predicted number of lifetime sexual partners by sex and age, averaging across the other covariates. There is a general age effect – one accumulates more sexual partners throughout the life course. It starts to decline past the mid-40s, indicating either recall bias (older respondents may not accurately remember their total number of lifetime sexual partners) or a cohort effect (older people have had less sexual partners than the generations that proceed them). Across all ages, males report a higher number of lifetime sexual partners.

Table S2 (b) shows the results of the negative binomial regression including the fieldworker’s sex and age. Interacting respondent and fieldworker’s sex significantly improved model fit (p < 0.001); further interacting respondent and fieldworker’s age significantly improved model fit (p < 0.001); finally, interacting respondent’s sex and fieldworker’s age, as well as fieldworker’s sex and respondent’s age, significantly improved model fit (p = 0.001) and constituted the final model.

Figure 1 shows the predicted number of lifetime sexual partners by respondent and fieldworker’s sex and age. For all respondents, having an older fieldworker reduced the number of reported lifetime sexual partners, though the effect is stronger for male respondents than for females. Moreover, for male respondents, having a male fieldworker decreased the number of reported lifetime sexual partners (β=−0.296). Altogether, male respondents report more sexual partners to female fieldworkers than male fieldworkers, but report fewer sexual partners to older fieldworkers than younger ones.

Figure 1.

Figure 1

Open in a new tab

Predicted Number of Lifetime Sexual Partners using Average Marginal Effects (Table S2B), by Respondent and Fieldworker's Age and Sex, Agincourt, South Africa, 2010 – 2011.

Condom Use

Table S3 (a) shows the results of the base logistic regression on the probability of condom use at last sexual intercourse, without fieldworker effects. Including age-squared significantly improved model fit (p < 0.001).

Figure S2 shows the predicted probability of using a condom at last sexual intercourse by sex and age, averaging across the other covariates. Reported condom use declined with age. Lower reported condom use for females may reflect childbearing intentions; higher reported condom use for males may reflect condom use with additional sexual partners, though these sex differences are not significant (95% CI: 0.82, 1.26).

Table S3 (b) shows the results of the logistic regression including the fieldworker’s sex and age. Interacting respondent and fieldworker’s age significantly improved model fit (p = 0.011).

Figure 2 shows the predicted probability of condom use at last sexual intercourse by respondent and fieldworker’s age. For all respondents, and across the life course, having an older fieldworker increased the odds of reported condom use at last sexual intercourse (OR: 1.4). This effect is strongest for older respondents: as respondents become older, they are more likely to report condom use to older fieldworkers (OR: 1.028).

Figure 2.

Figure 2

Open in a new tab

Predicted Probability of Condom Use at Last Sexual Intercourse using Average Marginal Effects (Table S3B), by Respondent's and Fieldworker's Age, Agincourt, South Africa, 2010 – 2011.

Discussing HIV

Table S4 (a) shows the results of the base logistic regression on the probability of discussing HIV with a partner, without fieldworker effects. Figure S3 shows the predicted probability of discussing HIV with a partner by sex and age, averaging across the other covariates. Reported discussing HIV with a partner declined with age.

Table S4 (b) shows the results of the logistic regression including the fieldworker’s age. Interacting respondent and fieldworker’s age significantly improved model fit (p < 0.001). Sensitivity tests of outlier fieldworkers suggested one with significantly different results compared to the others, who we thus omitted from the final model.

Figure 3 shows the predicted probability of discussing HIV with a partner by respondent and fieldworker’s age. Having an older fieldworker and a male fieldworker increased the probability of reporting having discussed HIV with a partner (OR: 1.649; OR: 1.427 respectively).

Figure 3.

Figure 3

Open in a new tab

Predicted Probability of Discussing HIV with a Partner using Average Marginal Effects (Table S4B), by Respondent and Fieldworker's Age, Agincourt, South Africa, 2010 – 2011.

HIV Testing

Table S5 (a) shows the results of the base logistic regression on the probability of reporting ever taking an HIV test, without fieldworker effects. Including age-squared and age-cubed significantly improved model fit (p < 0.001). Further interacting respondent’s sex and age (p < 0.001) and age-squared (p = 0.001) significantly improved model fit and constituted the final model.

Figure S4 shows the predicted probability of ever taking an HIV test by sex and age, averaging across the other covariates. Women report a higher probability of ever taking an HIV test than men (likely because of antenatal testing) until after ages 50+. For men, the probability of ever taking an HIV test increases with age until age 40, and then declines.

Table S5 (b) shows the results of the logistic regression including the fieldworker’s age and sex. Having an older fieldworker increased the probability of reporting ever taking an HIV test, while having a male fieldworker decreased the probability, though neither effect was significant.

DISCUSSION

Our results show an effect of fieldworker age and sex across sexual behavior outcomes. Respondents report fewer sexual partners and more “responsible” sexual behavior with their partners to older interviewers. We also find, contrary to our hypothesis, that men report having more sexual partners to female interviewers. Our results show that men being interviewed by men and older fieldworkers look different than their counterparts being interviewed by women and by younger fieldworkers. This has implications for analyzing sexual behavior data, especially when the results may be significant but the (fieldworker) effects are obscured (Weinreb 2006).

Our results yield implications for future research, which we discuss alongside our study limitations. First, we can only make assumptions about respondent reporting. To prove it, we would need data not currently available. Future studies might test for interviewer effects via means of randomization or experimental design to explore unmeasured or unrecognized patterns in respondents’ reporting to different categories of interviewers. Analysts might also control for interviewer effects in their models, and collect additional data on interviewer characteristics that could factor into future analyses. Additionally, it is unknown if missing respondents would vary systematically in their reporting of sexual behavior, though we introduced the study to sampled participants as one about diseases in the community that need chronic care, which may help minimize the potential for refusal bias.

Second, we do not have enough variability on interviewer characteristics to know if we are detecting other (unobserved) interviewer effects, such as community reputation, familiarity with the respondent’s family, or perceptions of interviewer attractiveness (Jæger 2013). This is both a strength and weakness: while unknown interviewer characteristics may be obscured, our interviewers are similar on other characteristics such as ethnicity, education, and religion, suggesting we are likely identifying actual fieldworker (age and gender) effects. We also do not exploit the extent to which “role restricted interviewers’ effects”-- the influence of an interviewer’s conduct on responses (e.g., particular greetings, the way they pose questions) (Bignami-Van Assche, Reniers and Weinreb 2003)-- affect our results. We tried to address this by using an indicator for each fieldworker when modeling each outcome. In one outcome we found an outlying fieldworker: responses to this fieldworker indicated a significantly higher probability of discussing HIV with a partner than other fieldworkers. We thus omitted that interviewer from the model. Fuller attention to role restricted interviewer effects requires greater data on the interview process itself. Finally, due to the small sample of fieldworkers, we were unable to explore additional interactions, such as fieldworker sex and fieldworker age, or respondent and fieldworker home village. Future research could examine the influence of these characteristics.

Third, our analysis of age must be interpreted conservatively. Perceptions of one’s age are hard to decipher and we do not know how respondents’ perceive the age of the interviewer. We also recognize that respondents over age 44 are older than our oldest interviewer. Given data limitations, we tried to account for it in our models: when it improved model fit, we included an interaction between the age of the respondent and the age of the interviewer. Our focus here, however, is on the significance of social categories of age, not of age differences between respondents and interviewers.

Finally, our results suggest greater attention be paid to the influence of interviewer embeddedness in the setting on study results. We did not include a measure for this in our analysis (we do not have it). We assume that all interviewers in our study are “insiders”, as a requirement for their employment with the AHDSS is that they live in the study area. This does not capture the degrees of “insider-ness” studied by Weinreb (2006), for example, who operationalized insiders as those who knew the respondents’ family and strangers as those who did not, and found insiders do a better job of collecting reliable data on sensitive topics such as family planning and AIDS (Weinreb 2006). We believe our study engages the debate about the possible pitfalls of interviewer insider-ness by demonstrating how other social characteristics of interviewers – age and gender – influence sexual behavior reporting where interviewers and respondents live in the same setting.

Interviewer characteristics – such as age and sex -- cannot be ignored in data analysis on sensitive topics like sexual behavior, and likely other value-laden topics. Their influences, however, may be culturally determined and thus important to analyze systematically for each setting (McCombie and Anarfi 2002). While it is beyond the scope of this paper to explore the theoretical underpinnings of these social interactions, our study shows the need for increased attention to the influence of interviewer characteristics on the types of information that different categories of respondents report, and consequently, the data their interactions in the field research process produce.

Supplementary Material

Supplemental (Tables S1-S5 & Figures S1-S4)

Acknowledgments

We gratefully acknowledge funding from the National Institute on Aging (R24AG032112), National Institute of Child Health and Human Development (K01 HD057246; R01 HD054511), Wellcome Trust, UK (085477/Z/08/Z), William and Flora Hewlett Foundation and Anglo American Chairman’s Fund. This project also received funding and administrative support from the University of Colorado Population Center, funded by the Eunice Shriver National Institute of Child Health and Human Development (NICHD R21 HD051146). We also thank the Medical Research Council and University of the Witwatersrand for continued support; the Ha Nakekela field team and participating communities; and Vusimusi G. Dlamini, Stephane Helleringer, Alexander Weinreb and the anonymous reviewers for their helpful comments.

REFERENCES

  1. Aiken Leona S, West Stephen G. Mulitple Regression: Testing and Interpreting Interactions. Thousand Oaks: Sage Publications; 1991. [Google Scholar]
  2. Becker Stan, Feyisetan Kale, Makinwa-Adebusoye Paulina. The Effect of the Sex of Interviewers on the Quality of Data in a Nigerian Family Planning Questionnaire. Studies in Family Planning. 1995;26(4):233–240. [PubMed] [Google Scholar]
  3. Bignami-Van Assche Simona, Reniers Georges, Weinreb Alexander A. An Assessment of the Kdicp and Mdicp Data Quality. Demographic Research. 2003;S1(2):31–76. [Google Scholar]
  4. Blalock Casey. Labor Migration and Employment in Post-Apartheid Rural South Africa. PhD, Sociology, University of Colorado Boulder; 2014. [Google Scholar]
  5. Davis Darren W, Silver Brian D. Stereotype Threat and Race of Interviewer Effects in a Survey on Political Knowledge. American Journal of Political Science. 2003;47(1):33–45. [Google Scholar]
  6. Davis RE, Couper MP, Janz NK, Caldwell CH, Resnicow K. Interviewer Effects in Public Health Surveys. Health Education Research. 2010;25(1):14–26. doi: 10.1093/her/cyp046. [DOI] [PMC free article] [PubMed] [Google Scholar]
  7. Edwards Mildred, Thomsen Sarah C, Toroitich-Ruto Cathy. Thinking Aloud to Create Better Condom Use Questions. Field Methods. 2005;17(2):183–199. [Google Scholar]
  8. Ford Kathleen, Norris Anne E. Effects of Interviewer Age on Reporting of Sexual and Reproductive Behavior of Hispanic and African American Youth. Hispanic Journal of Behavioral Sciences. 1997;19(3):369–376. [Google Scholar]
  9. Goffman Erving. The Presentation of Self in Everyday Life. New York: Anchor Books; 1959. [Google Scholar]
  10. Gómez-Olivé Francesc Xavier, Angotti Nicole, Houle Brian, Klipstein-Grobusch Kerstin, Kabudula Chodziwadziwa, Menken Jane, Williams Jill, Tollman Stephen, Clark Samuel J. Prevalence of Hiv among Those 15 and Older in Rural South Africa. AIDS Care: Psychological and Socio-medical Aspects of AIDS/HIV. 2013 doi: 10.1080/09540121.2012.750710. [DOI] [PMC free article] [PubMed] [Google Scholar]
  11. Graham Cynthia A, Catania Joseph A, Brand Richard, Duong Tu, Canchola Jesse A. Recalling Sexual Behavior: A Methodological Analysis of Memory Recall Bias Via Interview Using the Diary as the Gold Standard. Journal of Sex Research. 2003;40(4):325–332. doi: 10.1080/00224490209552198. [DOI] [PubMed] [Google Scholar]
  12. Gribble James N, Miller Heather G, Rogers Susan M, Turner Charles F. Interview Mode and Measurement of Sexual Behaviors: Methodological Issues. Journal of Sex Research. 1999;36(1):16–24. doi: 10.1080/00224499909551963. [DOI] [PMC free article] [PubMed] [Google Scholar]
  13. Groves Robert M, Fultz Nancy H. Gender Effects among Telephone Interviewers in a Survey of Economic Attitudes. Sociological Methods & Research. 1985;14(1):31–52. [Google Scholar]
  14. Hewett PC, Mensch BS, Erulkar AS. Consistency in the Reporting of Sexual Behaviour by Adolescent Girls in Kenya: A Comparison of Interviewing Methods. Sexually Transmitted Infections. 2004;80(suppl 2):ii43–ii48. doi: 10.1136/sti.2004.013250. [DOI] [PMC free article] [PubMed] [Google Scholar]
  15. Jæger Mads Meier. Hello Beautiful? The Effect of Interviewer Physical Attractiveness on Interview Success. The Danish National Centre for Social Research, Unpublished Working Paper No. 02/2013. 2013 [Google Scholar]
  16. Kahn Kathleen, Collinson Mark A, Xavier Gomez-Olivé F, Mokoena Obed, Twine Rhian, Mee Paul, Afolabi Sulaimon A, Clark Benjamin D, Kabudula Chodziwadziwa W, Khosa Audrey, Khoza Simon, Shabangu Mildred G, Silaule Bernard, Tibane Jeffrey B, Wagner Ryan G, Garenne Michel L, Clark Samuel J, Tollman Stephen M. Profile: Agincourt Health and Socio-Demographic Surveillance System. International Journal of Epidemiology. 2012;41:988–1001. doi: 10.1093/ije/dys115. [DOI] [PMC free article] [PubMed] [Google Scholar]
  17. McCombie Susan C, Anarfi John K. The Influence of Sex of Interviewer on the Results of an Aids Survey in Ghana. Human Organization. 2002;61(1):51–57. [Google Scholar]
  18. Oloo Isaiah, Gust Deborah A, Shinde Sanjyot, Mc Lellan-Lemal Eleanor, Pals Sherri L, Chege Wairimu, Gumbe Anne, Thomas Timothy K, Mills Lisa A, Chen Robert T. Effect of Gender of the Recorded Voice on Responses to Sensitive Sexual Behavior Questions: Use of Audio Computer-Assisted Self-Interview (Acasi) in Kisumu, Kenya. Field Methods. 2012;24(4):367–381. [Google Scholar]
  19. Randall Sara, Coast Ernestina, Compaore Natacha, Antoine Philippe. The Power of the Interviewer. Demographic Research. 2013;28(27):763–792. [Google Scholar]
  20. Reichmann William M, Losina Elena, Seage George R, III, Arbelaez Christian, Safren Steven A, Katz Jeffrey N, Hetland Adam, Walensky Rochelle P. Does Modality of Survey Administration Impact Data Quality: Audio Computer Assisted Self Interview (Acasi) Versus Self-Administered Pen and Paper? PloS one. 2010;5(1):e8728. doi: 10.1371/journal.pone.0008728. [DOI] [PMC free article] [PubMed] [Google Scholar]
  21. Weinreb Alexander A. The Limitations of Stranger-Interviewers in Rural Kenya. American Sociological Review. 2006;71:1014–1039. [Google Scholar]
  22. Wellings Kaye, Collumbien Martine, Slaymaker Emma, Singh Susheela, Hodges Zoé, Patel Dhaval, Bajos Nathalie. Sexual Behaviour in Context: A Global Perspective. The Lancet. 2006;368(9548):1706–1728. doi: 10.1016/S0140-6736(06)69479-8. [DOI] [PubMed] [Google Scholar]
  23. Williams Jill R, Ibisomi Latifat, Sartorius Benn, Kahn Kathleen, Collinson Mark, Tollman Stephen, Garenne Michel. Convergence of Fertility Levels between South Africans and Mozambicans in Rural Northeast of South Africa from 1993–2009. Global Health Action. 2013;6(19236) doi: 10.3402/gha.v6i0.19236. doi: http://dx.doi.org/10.3402/gha.v6i0.19236. [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supplemental (Tables S1-S5 & Figures S1-S4)
NIHMS808912-supplement-Supplemental__Tables_S1-S5___Figures_S1-S4_.docx (137.4KB, docx)

RESOURCES