What Happened When We Gave to Whole Villages

by Joy

“How do you decide whom to give to?” is usually the first question I get when I describe GiveDirectly’s model. It’s also a question we think about a lot internally. Effective targeting is a core feature of the product we offer donors, and a moral imperative.

In 2013, we began to experiment with a new approach to targeting: giving cash transfers to all households in a poor, rural village. This “saturation” approach was much more inclusive than our standard approach of enrolling only households living in thatch and mud homes: 9 in 10 households were eligible with saturation compared to 5 in 10 with thatch only. (Households living in homes made fully of permanent materials like cement and iron sheets were excluded.)

We thought saturation might reduce tension and conflict in the community, costs, and/or gaming, while continuing to reach extremely poor people. Starting in mid-2013, we randomly assigned 19 villages to saturation and 18 to thatch only. After the first transfer, we administered a follow-up survey with more than 90% of recipients and conducted focus groups in six villages in order to get quantitative and qualitative measures of tension and conflict.

The data, though not conclusive, surprised us.

First, it’s not clear that saturation villages experienced less tension and conflict. We looked at several indicators in the follow-up survey, including the percentage of recipients who heard complaints in the community, the nature of those complaints, and observed crime or violence. Only about half the indicators were lower in saturation villages.

Second, the focus groups indicated that tension across both types of villages was low – mostly rumors and some awkwardness. And nearly everyone, when presented with the situation we face, said to prioritize households living in thatch homes.

Third, saturation did not significantly reduce costs or gaming. In saturation villages, enrollment costs were < 5% lower and the percentages of households falsely posing as eligible during enrollment were comparable (4.3% vs. 3.6%).

All this suggests that targeting thatch-only does not actually cause significant tension or conflict, as we’d feared – nor does it cause significantly different levels of tension than saturation. Given that, we’ll likely continue targeting poorer families living in thatch and mud homes. But we’ll also pilot the inclusion of some iron-roof households who are extremely poor. Although we know iron-roof households are on average less poor than thatch, we also know that there are exceptions. We’re considering a few approaches to identifying them, such as applying additional objective criteria (e.g., elderly, widowed, or visibly disabled) and accepting community-based nominations of a fixed number of “special” cases.

So how do we decide whom to give to? By continually experimenting and refining our approach based on data and feedback from recipients.