Randomizing cash transfers: practical equipoise in social science experiments

When is it ethical to randomly determine the provision of a treatment to people in an experiment? Researchers designing such experiments often appeal to the principle of equipoise, which means  “that there is genuine uncertainty in the expert medical community over whether a treatment will be beneficial.” This principle captures a straightforward ethical logic: we know that ARVs are effective in treating AIDS, so it would be unethical to design an experiment that denies some AIDS patients access to ARVs. In contrast, until recently we did not know that starting ARVs immediately had health benefits for people with HIV who do know yet have AIDS. The experiment that proved that this was the case was entirely ethical to conduct.

This approach works well for medical trials, because doctors have the joint social roles of studying which treatments are effective and actually providing people with the treatments that work. Social scientists tend to be in a very different situation. First, relative to medical research, we are substantially less sure what the impacts of various interventions will be.** Second (despite what people might think about economists in particular) we just don’t have that much influence. There are many policies and programs that economists know to be effective that have little or no traction in terms of actual policies. An example that I am particularly fond of is the idea that if you want to maximize uptake and use of a product, you should give it away. The notion that people won’t “value” things they get for free is so popular that two economists actually had to go out and randomize prices for insecticide-treated bednets to prove that giving them away would work for fighting malaria. And still, I see people charging money for them.

The bednet study is a great example of what I am dubbing practical equipoise. I personally had no doubt, prior to that study, about which direction the demand curve for bednets sloped. Almost any economist would probably have felt the same way, so “clinical equipoise” did not hold. Yet policymakers felt strongly that user fees for the nets were a great thing. Practical equipoise is a situation where the science is in but policymakers don’t agree with it; while the scale of knowledge is not balanced, we can try to use persuasion and facts to balance the scale of actual practice. In this situation, randomized experiments are an ethical imperative – they benefit people who would otherwise have lost out, and also can help promote universal access to beneficial treatments in the future.

Practical equipoise holds for many policies that social scientists want to study, and particularly for cash transfers. There is a growing body of evidence that just giving money to poor people carries large benefits and few costs, leading prominent economists to advocate that evaluations of social programs should include a cash transfer arm as a basis of comparison. This is something we should probably do in general, and should definitely do as a way to evaluate giving people food handouts. Recent research by IFPRI comparing food vs. cash for helping the hungry has found that cash does much better; while their own discussion of their results has tried to take a moderate tone, the fact is that outcomes are similar for cash at a much lower cost. But people – especially the elites who make decisions about these things – remain broadly unconvinced that it is safe to give cash to poor people who need food. We are ethically obligated as scientists to push for randomized testing of cash transfers as an approach to food crises: we are highly confident they work, and need to do what we can to convince policymakers that we are right.

*Medical research is also focused on treatments that directly impact people’s well-being in an objectively-measurable way, which is relatively uncommon in social science research.

**A prominent example is the effect of moving kids from housing projects to richer ones, which was found to be surprisingly small and negative for some outcomes in analyses that focused on older children and people who volunteered to move. Recent work examining younger kids and on people who do not voluntarily sign up for housing vouchers has shown that there are in fact substantial benefits for those groups.

Leave a Reply

Your email address will not be published. Required fields are marked *