Schooling, Wealth, Risky Sexual Behaviour, and HIV/AIDS in Sub-Saharan Africa

Abstract

Economic growth and development have improved human health in many regions, while sub-Saharan Africa continues to lag behind. Economic theory and the existing empirical evidence suggest that development may not generate large reductions in the leading cause of adult mortality in the region, HIV/AIDS, and may increase risky sexual behaviour. We examine the association between schooling/material standard of living and HIV risk using data from more than 500,000 respondents in 32 sub-Saharan African countries. The results of our descriptive analysis suggest that the rapid increase in primary school completion without improvements in living standards or secondary school completion might not mitigate HIV transmission.

Acknowledgements

We thank seminar participants at the Population Association for America 2017 Annual Meeting, Alden Boetsch, Latifat Ibisomi, Jessica Leight, Dr. Richard Palmer-Jones, and an anonymous reviewer for many excellent comments. The PopPov Research Network and the Population Reference Bureau provided generous financial support. The findings, interpretations, and conclusions expressed in this paper are those of the authors and do not necessarily represent the views of the aforementioned individuals or agencies. All errors are our own.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

1. Increased male education could also reduce the price of risky sex through a reduction in search costs, again increasing quantity demanded.

2. A large literature (for example, Morris and Kretzmar 1997; Halperin & Epstein, 2004; Mah & Halperin, 2010; and Fox, 2014) has identified multiple concurrent partnerships as a key driver of the HIV/AIDS pandemic. Two other key risky sexual behaviours identified in the literature are not using a condom (for example, Weller & Davis, 2002) and engaging in transactional sex (for example, Baral et al., 2012).

3. Of the studies that follow, four studies use data from Malawi, two studies each use data from Uganda and Kenya, and one study each uses data from Botswana and Zambia.

4. Several studies (for example, Higgins, Hoffman, & Dworkin, 2010; Poulin et al., 2016) have documented and analysed the relative lack of attention to males in HIV policy and research.

5. The one exception to the focus on a single country is Burke et al. (2015) that uses data from 21 sub-Saharan African countries to examine the effects of weather shocks on HIV and finds that large negative weather shocks increase HIV risk.

6. Methodological differences likely explain the (partly) divergent HIV-education gradient results across countries. Fortson (2008) primarily reported regression results using pooled samples of females and males, estimated separately by country. De Walque (2009) primarily reported regression results estimated separately for females and for males, estimated separately by country and pooled across countries. In addition, the main regression specifications in Fortson (2008) did not simultaneously control for educational attainment and for household wealth, whereas the main regression specifications in De Walque (2009) simultaneously controlled for educational attainment and for household poverty (that is an indicator variable for earth floor, instrumented by ownership of other household consumer durables). Fortson (2008) allowed for non-linearities in the HIV-education gradient and De Walque (2009) restricted the gradient to be linear.

7. A final approach has been systematic reviews. In a review of articles from six countries, Hargreaves and Glynn (2002) found evidence that HIV prevalence was correlated with educational attainment, although they concluded that the relationship may be changing over time.

8. The is a sub-national fixed effect, for example the Northern region of Ghana.

9. We estimate heteroskedastic robust standard errors.

10. The countries and survey years are: Benin 2011, Burkina Faso 2010, Burundi 2010, Cameroon 2011, Comoros 2012, Congo 2011, Congo Democratic Republic 2013, Cote d’Ivoire 2011, Ethiopia 2011, Gabon 2012, Ghana 2014, Guinea 2012, Kenya 2014, Lesotho 2009, Liberia 2013, Madagascar 2008, Malawi 2010, Mali 2012, Mozambique 2011, Namibia 2013, Niger 2012, Nigeria 2013, Rwanda 2014, Sao Tome and Principe 2008, Senegal 2010, Sierra Leone 2013, Swaziland 2006, Tanzania 2010, Togo 2013, Uganda 2011, Zambia 2013, and Zimbabwe 2010.

11. Behavioural outcomes (even self-reported) may be more accurate measures of HIV risk than biomarkers (Corno & Aureo, 2016). The relatively low HIV transmission rate per HIV discordant unprotected coital act can lead to a large difference between latent HIV risk and actual HIV status.

12. The asymmetry by gender in measures of risky sexual behaviour results from the DHS only asking males about exchanging money for sex. Swaziland does not ask any respondents about exchanging money for sex.

13. This does not address concerns about relative over-reporting by males.

14. One concern about the risky behaviour results for females and males is that these behaviours may not be particularly risky for married and older individuals. Using data from six sub-Saharan African countries, De Walque (2007b) demonstrates that two-thirds of HIV-infected couples are sero-discordant (that is only one individual is HIV positive) and in approximately one-third of these it is the female that is HIV positive. This indicates that unprotected sex within marriage is still quite risky. Nonetheless, we provide additional evidence that our results reflect truly risky behaviour. We re-estimated Equation (1)(1) $SE{X}_{irc}=\gamma +\beta SE{S}_{irc}+X_{irc}^{{}^{\prime }}\mathrm{\Omega }+{\mu }_{rc}+v_{irc}$(1) using just unmarried individuals (and then just unmarried, younger individuals) and found patterns for the signs and statistical significance of the coefficient estimates nearly identical to those in Tables 3 and 4, with small-to-medium changes in magnitudes. These findings indicate that our results reflect differences in particularly risky behaviours.

15. Not all sub-Saharan African DHS countries for which we have sexual behaviour data implemented a HIV testing module. For those that implemented a HIV testing module, a subset of DHS respondents were asked to participate in the module. Approximately 80 per cent of those asked agreed to participate. In our full sample of 32 countries, the following 23 countries provide data from DHS HIV testing modules that can be linked to the main DHS data: Burkina Faso 2010, Burundi 2010, Cameroon 2011, Congo Democratic Republic 2013, Cote d’Ivoire 2011, Ethiopia 2011, Gabon 2012, Ghana 2014, Guinea 2012, Lesotho 2009, Liberia 2013, Malawi 2010, Mali 2012, Namibia 2013, Niger 2012, Rwanda 2014, Sao Tome and Principe 2008, Senegal 2010, Sierra Leone 2013, Swaziland 2006, Togo 2013, Zambia 2013, and Zimbabwe 2010.

16. Polygamy does not appear to be driving this result. Among unmarried males, 14 per cent report multiple partnerships in the past 12 months.

17. We examine the robustness of our results to an alternative measure of consumer durables ownership, the DHS wealth index, in Section 5.2. One reason not to use the wealth index in our primary specification is that it is constructed separately by country and may be difficult to compare directly across countries.

18. Eighty-five per cent of female respondents and 88 per cent of male respondents report knowing that being faithful reduces HIV risk. Seventy-four per cent of female respondents and 82 per cent of male respondents report knowing that condom use reduces HIV risk.

19. A few DHS also ask respondents whether they know that abstaining from sex is an HIV prevention technique, but this question was not asked in most of the countries in our sample.

Additional information

Funding

This work was supported by the PopPov Research Network [Ad hoc]; and Population Reference Bureau [Ad hoc].