Syndicate content

Education, fertility and HIV: It’s complicated

Markus Goldstein's picture

An interesting, recently revised working paper by Duflo, Dupas and Kremer looks at the effects of providing school uniforms, teacher training on HIV education, and the two combined. This paper is useful in a number of dimensions – it gives us some sense of the longer term effects of these programs, the methodology is interesting (and informative), and finally, of course, the results are pretty intriguing and definitely food for thought. 

So we’ve talked in a number of the previous posts about the lack of longer term results. Here is a paper which tracks kids over seven years, and seven important years for the fertility-HIV-education nexus, as they go through and finish school.   So this is cool. But, as you might imagine, following a bunch of Kenyan teenagers over time is going to be tricky, particularly when they leave school.

This is where one of the first areas of interesting methodology come in. More on the data collection in a minute, but when the time for the follow-up survey came, they only had 55% who were found in the first round of tracking.   This necessitated a second round of “intensive” tracking, where they took a sub-sample of those who weren’t yet found and went after them with gusto. And they found a lot of these folks.   One interesting thing: boys are easier to find than girls.   In the first round of tracking, they found 59% of the boys against 51% of the girls. In the intensive tracking this gap was smaller: they found 77.5% of girls and 84% of boys. But why are girls harder to find?   Part of their explanation is that boys are more likely to be home with mom but the other part is that girls are harder to track. Interesting…Has anyone had similar experiences? Any other explanations for this? 

Another interesting aspect of their methodology is that they bring to bear a range of data collection tools. While the tracking/follow-up survey gives them end-point data, in the intervening time they did “roll-call” data where their teams showed up at schools (unannounced) and asked about attendance, fertility, marital status, enrollment, and a couple of other things. They did this by reading the sample list to all upper-grade students present that day. This gives them a bunch of data points on key variables before the richer, longer survey.   And, of course, this being HIV work, they also collected biomarker data with the endline.

One of the other reasons I like this paper is that it nicely makes the point that an oblique tool may work better.   A lot of times when working with project teams, I get the question: “why is that question on your survey? It has nothing to do with my project.” And indeed, that outcome is not anywhere in the framework of the project design, the stated objectives, etc.   But sometimes, an intervention which is obliquely related to a given (important) problem may be incredibly effective. Here you have free school uniforms (goal: staying in school) and improved HIV education (goal: don’t get HIV kids).  So which works better for HIV? Which works better for staying in school? And which works better for fertility (which is not generally aim #1 for either intervention)?     

The answer is not straightforward – at least for the HIV and fertility questions (it’s kind of what you would expect for the enrollment). So free school uniforms keep girls (and boys) in school and reduce the likelihood that girls get pregnant.   They also shift girls towards younger partners and reduce earlier marriage. But they don’t have an impact on HSV2 infection (HSV2 is the STI they use to capture risky sexual behavior – the HIV rates in their sample are quite low).

Training for teachers in the government HIV curriculum shows some promise in terms of teachers being more likely to mention HIV and students more likely to mention abstinence and faithfulness (important messages from the curriculum).   But in terms of other indicators, there isn’t much of an effect – there is no significant effect on drop-outs and no significant impacts on teen marriage or childbearing. It did, however, reduce the number of out of wedlock pregnancies.   And, the kicker, it had no significant impact on HSV2 infection. 

But where things get really complicated is in the interaction of the two programs (the design here is four groups – control, each intervention alone, and combined). First, when the programs are combined, the reduced drop-out effect of the uniforms is no longer significant. And away goes the effect on pregnancy as well. But the combined program did have an impact on HSV2 infection rates – reducing them by 14-20% relative to the control group – for girls. No effect on HSV2 for boys.   They put forth a model which uses a distinction between committed and casual relationships to try to reconcile these results.   In the end, when you take these results all together they suggest we have further work to do to understand the interplay between specific school HIV education campaigns and the ability of children to stay in school and what these will do to both education and sexual behavior decisions. 

Comments

Submitted by Olivier on
I haven't read the paper and might be off target but one explanation could be that girls might be harder to find because they got married and went to their husband's family.

Olivier -- my guess this is getting at part of the story -- they make a comment about boys being more likely to still be living at home. If you start with an initial population of teenagers the same ages, and men marry younger women (and say everyone moves out -- i am not sure about the case in these communities), then mechanically you will end up with more boys at home than girls...But definitely the fact as to whether the wife moves to the husbands family/village or vice versa could be part of the story too.