In honor of Labor Day here in the US, I want to talk about a recent nutrition paper by Emla Fitzsimons, Bansi Malde, Alice Mesnard and Marcos Vera-Hernandez. This paper, “Household Responses to Information on Child Nutrition,” is one with a twist – they look not only at nutrition outcomes, but they also try and figure out where these might be coming from – and hence also look at labor supply.
The setting: Malawi, where 48% of the children under 5 are stunted – which is the second highest rate in sub-Saharan Africa. Fitzsimons and co. are working in the Central Region of Malawi, in a district which looks slightly (but not majorly) worse off than the average.
The intervention: Fitzsimons and co. are working with a mother and child research and development project where volunteers nominated by local leaders are trained to provide information and advice to mothers about infant (less than six months old) feeding. These counselors cover a population of around 1000 people where they seek out pregnant women – visiting them at home once before birth and four times afterwards. The gist of the information they provide is to encourage breastfeeding and then, as the child is weaned, the importance of a varied diet, with a focus on more nutritious food.
The evaluation design: So this is an RCT, where they took the district in which they are working and divided it into 48 parts, each of about 8,000 people. Then they took the center of each “part” and worked outwards until they had about 3000 people in what they call the study area or zone. This leaves them with a buffer area between their zones (or treatment/control areas). This is kind of cool, since it really insulates against spillovers (and remember, this is an information intervention – so spillovers/contamination will be more likely). While this idea makes a lot of sense from an internal validity standpoint, it made me wonder a bit what this does for external validity. Anyhow, they select 12 of these zones for treatment, and 12 for controls (the other 24 are in another experiment).
The data: The first piece of data they had is a baseline census – which gives them a bunch of basic characteristics. From this they draw a sample of women who are to be surveyed 3.5 and 4.5 years after the intervention started (which was in July of 2005). In the first of these surveys they have pretty significant attrition – losing about 1/3 of the sample. The reason for this makes me think it might be good to set up a currency hedge for fieldwork: “This was lower than planned due to an unexpected drop of the British Pound against the Malawi Kwacha resulting in data collection having to be cut short.” This is an issue we all face, and here it cut down on their ability to track folks down. Their second survey round doesn’t suffer from this problem – they find 91% of the folks who they interviewed in the first survey round. At any rate, they provide some analysis in the paper to try to allay concerns that may come from this attrition.
The surveys cover consumption, with a particular emphasis on child consumption, expenditures, health, anthropometrics, labor supply and some other areas. One interesting side note is that these data are coming a good amount of time after the intervention was started – giving us some time to see how these early childhood interventions may or may not bear fruit.
Results: Households spend more on food, particularly proteins and fruits and vegetables, which are 1/3 higher than control areas. Among kids under six months, there is a drop in the consumption of water and non-mother’s milk, which suggests that these kids are being more exclusively breastfed. Kids over six months are consuming a more diverse diet, including more protein rich foods and staples (but they, as with kids everywhere, are steering clear of the veggies).
So where does this food come from? Fitzsimons and co. find that it’s the men going to work here. Men overall are more likely to have a second job (a 6.6 percentage point increase against an average of 12% in the control areas). But to strengthen this argument, Fitzsimons and co. focus in on fathers versus non-fathers – and here they find significant effects on the likelihood of having a second job (now 8 percentage points) and in hours worked (significant at the ten percent level). Nothing happens with mothers or women in general.
Children’s health also improves – height increases by 0.20 standard deviations of the WHO reference population. There is also a meaningful drop in diarrhea in children under six months, which lines up with the result that these kids are consuming less water. These results leave us with two (non-exclusive) potential channels: improved growth from lower diarrhea in the early months or from a better diet after breast feeding ends.
So what do we take away from this? Fitzsimons and co. have clearly added to the literature that information matters for nutrition (particularly in this counseling form) and with the twist that the information can get dads out there working more. Now, as they point out, their effects on height are only about half that you would see from an intervention where actual food was provided. But then this raises questions of cost – and it would be nice to see what this intervention (which was driven by volunteers after all) costs relative to more intensive interventions.