Tuesday, June 4, 2013

Rainfall, sex, and HIV

The Economist tell us that times are tough for sex workers in the UK (HT: Sol).  A sluggish economy and has dried up demand for paid sex, and high unemployment has increased the supply of people willing to sell it.  As a result, the street price for sex has plummeted, and women appear to be offering increasingly risky -- and presumably better compensated -- services in response.  

This relationship between economic conditions and sexual behavior is something that is increasingly studied by economists and other social scientists.  The goals in this research are both to understand the determinants of sexual behavior, as well to understand how these changes in behavior affect key health outcomes.  A recent focus has been on the role of economic conditions in the HIV epidemic in Africa.  Over 2 million people continue to become newly infected with HIV every year, and the vast majority of these new infections occur in Sub-Saharan Africa through unprotected heterosexual sex.  So if economic conditions shape sexual behavior in Africa (as they appear to in England), then it is plausible that this could in turn shape the trajectory of the HIV epidemic.  

Such a finding would really matter for policy.  If changing economic conditions are found to affect rates of HIV infection -- for instance,  if economic downturns result in a lack of alternative employment that in turn causes women to enter the sex trade, and that this in turn increases the number of people infected with HIV -- then this would suggest policy approaches for HIV prevention that are pretty different from current efforts to expand access to antiretrovirals, to increase male circumcision, and to change behavior through information campaigns about safe sex.   However, as the Economist story suggests, it's not at all clear how the economic effects would play out in terms of overall disease transmission.  If tough economic conditions lead to an inward shift in the demand curve for sex and an outward shift in the supply curve, the equilibrium price of sex will fall (as in the story) but the effect on the equilibrium quantity of sex could go up or down, depending on how the demand and supply curves are sloped and which shifts more.  If riskier (e.g. unprotected) sex is being transacted, that will clearly matter for infection risk too.  

Research in Africa has increasingly documented the first part of the story, that economic conditions shape sexual behavior.  For example, Jon Robinson and Ethan Yeh somehow persuaded a sample of Kenyan women to keep detailed diaries of their sexual activity, and document that these women increase their supply of risky (and better compensated) sex when faced with income shortfalls.  Similarly, Nick Wilson shows that when economic times are good in Zambian mining towns, rates of transactional sex decline.  Nevertheless, these papers do not tell us how these changes in behavior translate into to actual HIV outcomes, which for many is the outcome of interest. 

In a recently updated working paper, Erick Gong, Kelly Jones and I explore how local variation in rainfall -- a oft-used proxy for rural incomes -- affects rates of HIV infection across sub-Saharan Africa. To do this we match gridded weather data to information on the HIV status of over two hundred thousand individuals across 19 African countries (as gleaned from the Demographic and Health Surveys).  We then compare whether rates of HIV prevalence are higher in locations where recent growing season rainfall had been particularly low (relative to average), as compared to nearby locations where growing-season rainfall had been closer to that location's normal rainfall.

We find that exposure to a single "rainfall shock" increases local HIV prevalence by 11%, where a shock is defined as a year in which growing season rainfall is at or below the 15th percentile of historical realizations at a given location.  This is a large effect, and it passes a number of "sniff" tests:  the effect is larger when we adopt a more severe definition of "rainfall shock", it is larger in rural areas where incomes should be more sensitive to rainfall, and it is larger in areas where baseline HIV prevalence is high (i.e. where there is actually HIV to pass around).  Running a similar analysis at the country level on a different HIV dataset and across all African countries, we find that rainfall shocks explain up to 20% of the variation in HIV prevalence at the country level.  

While we believe these results are unlikely to be spurious, they are clearly very "reduced form":  we are effectively regressing HIV on rainfall at either the individual or country level.  A reasonable concern would then be that variation in rainfall could be affecting all sorts of other things (beyond the market for transactional sex) that might also affect HIV.  For instance, perhaps negative rainfall shocks lead to more civil war and in turn that war-related sexual violence increases HIV rates, or perhaps rainfall shocks induce migration and that in turn spreads HIV, or perhaps rainfall shocks lower individuals' nutritional status and this directly affects their susceptibility to HIV infection.  

Using available data in the DHS and evidence from the literature, we do our best to show that these alternative stores are probably not what's going on, and instead that rainfall affects HIV through its effect on how individuals cope with changes in income.  In particular, we show that self-reported measures of sexual activity (being sexually active, having sex with multiple partners or a non-spouse partner) increase in response to rainfall shocks.  We also show that changes in HIV rates are concentrated among women working in agriculture, who would likely suffer the largest income drop in response to a rainfall shock, and among men not working in agriculture, who would be less likely to suffer the drop in purchasing power that would limit their ability to purchase sex.  

Overall, we interpret our results as further evidence that poor people have a lot of trouble coping with negative income shocks, and that in the absence of the sorts of financial tools available to people in wealthier countries (savings, credit, insurance), they sometimes have to resort to behavior that helps maintain incomes in the short run but could be very costly in the longer run.  In terms of HIV, this again implies approaches to stemming the HIV epidemic that are different than current policy focuses.  Our results add further impetus to efforts aimed at reducing farmers' sensitivity to weather (e.g. through drought tolerant varieties), and aimed at providing tools (e.g. savings or insurance) to help them smooth consumption when their incomes fall.  Our findings imply that the failure to provide these tools could have large and previously unrecognized effects on the HIV epidemic.  

The paper is here.  Comments very welcome!

No comments:

Post a Comment