Interview with Angus Deaton: Critiques of Cosmopolitan Prioritarianism and Randomized Control Trials

David A. Price of the Richmond Fed carries out an interview titled “Angus Deaton: On deaths of despair, randomized controlled trials, and winning the Nobel Prize” (Econ Focus: Federal Reserve Bank of Richmond, Fourth Quarter 2023, pp. 18-22). 

Here are a few of Deaton’s comments that caught my eye:

On his shift from “cosmopolitan prioritarianism” to “domestic prioritarianism”

If you try to find out what an economist believes philosophically, they will say it’s utilitarianism. … And so there’s a widespread belief in economics that poorer people deserved our attention more than less poor people, because an extra dollar given to someone who is really poor would do more good than an extra dollar going to someone who already had plenty. Philosophers nowadays call that “prioritarianism,” meaning people who have the lowest level of well-being are the ones who deserve the most at the margin.

The other dimension is “cosmopolitan,” meaning you apply this idea across the whole world, without paying attention to national boundaries. Many do seem to embrace cosmopolitan prioritarianism, in which you metaphorically line up everybody in the world from worst off to best off and you prioritize the people at the bottom — without regard to where they are.

I certainly believed this for a long time, and I spent many years consulting for the World Bank where this view was strongly held. But I now think it’s wrong for a number of reasons … One of them is that national boundaries really do matter. … We accept obligations for other people in our country, which we don’t accept for other people outside the country. So whether we like it or not, we’re locked in this tangle or this system of reciprocal obligation … Our fellow countrymen, whether we care for them because we feel like them or not, we have a responsibility in terms of our taxes and welfare systems, such as they are, and so on. So that’s part of it.

The other part is my suspicion, and this is deeply controversial, that some of the poorest people in America are every bit as poor in terms of overall well-being as the people in Africa or India or wherever the aid agencies like to hold up in front of us. And again, that’s not just money. It’s living in a functional society with societal supports. For instance, if you read some of the ethnographic literature about the Mississippi Delta, there are horrible things going on there in people’s lives. I don’t know how to estimate those in terms of numbers, because we don’t have very good tools for that. But I do challenge the idea that there’s no global poverty in America. So I am increasingly drawn to a form of domestic prioritarianism in which I worry a lot about others in my country who have the least.

On his doubts about the research methodology of randomized control experiments:

In the old days, we used to say here’s a regression and here’s a bunch of regression diseases. There’s a bunch of randomized controlled experiment diseases, too, which can get in the way.

People seem to think if you randomize — if you have two groups picked at random and one gets the treatment and one doesn’t — they say the only difference between the two groups is the treatment. But it’s dead wrong. When I used to teach this class, I would say, if I pick one of you at random with my eyes shut, and I pick another one with my eyes shut, does that make you identical? Of course not. You could argue that’s a large-sample or small-sample thing: If you pick a million people at random, then on average, they’re going to be the same in the two groups. And that’s true. But we don’t know how big it really has to be. And a lot of the experiments are pretty small. So it could be that the two groups you’re looking at are different at random but still different.

The other thing is that randomization can’t control for things that are the same in the two groups. That’s the external validity issue. One of my co-authors in the field of randomized controlled trials, the philosopher Nancy Cartwright, has an example that I like to give. There is famous work that Ed Miguel and Michael Kremer did on worms and deworming. They gave deworming pills in Kenya, and the kids who got the deworming pills did much better in school. Nancy lives in Oxford, and she said, “I have my granddaughter living with me and she’s not doing very well in school, so now I know what I should do, which is I should give her deworming pills, right?” But somewhere between Kenya and Oxford, the pills stop working.

So then, why and where? Of course, what’s on the line is there has to be worms or there has to be lack of sanitation or people are not wearing shoes or something, which is never in the experiment, because everybody in the experiment doesn’t have shoes. Or everybody in the experiment is walking around in an unsanitary field or something, and that’s not what you get in Oxford, so it’s not going to work there. But you have to know what these conditions are if you’re actually going to use those results. So sometimes these little experiments are not much more than anecdotes. You don’t really know what to take away from them. To paraphrase Bertrand Russell, you need a deeper view of the structure of reality.