An interview with Johannes Haushofer on the state of the evidence for cash transfers
We’re pleased to introduce a new interview series for the Phenomenal World. We will be speaking with an array of academics and policymakers on the most ambitious yet tractable new ideas in the social sciences.
Johannes Haushofer is assistant professor of Psychology and Public Affairs at Princeton University. His work includes development economics, behavioral economics, psychology, and neurobiology. We spoke to him primarily about a neopolicy idea on which he has unique expertise: unconditional cash transfers (UCTs). Along with Jeremy Shapiro, he has led research on GiveDirectly’s UCT program in Kenya, and his work on short-term and long-term effects there has both provided the field with new evidence and set a course for deeper questions. Now Johannes is starting to work on a UBI pilot in a major US city, across the developing-nation/developed-nation divide.
We spoke to Johannes broadly about (1) where he sees the state of the evidence, (2) what conclusions can be drawn from developing nations to developed, and (3) his larger vision for the march of evidence regarding policies like this.
We’re grateful that Johannes took the time to speak with us and to inaugurate this interview series. Interviewing him was Michael Stynes, who leads JFI, as well as Sidhya Balakrishnan of JFI. This interview has been condensed and edited.
michael stynes: Johannes, first of all, thank you very much for speaking with us. Your work is incredibly important to the entire basic income/cash transfer research community and I’m happy that Sidhya and I will have the opportunity to talk through some of your work. We are particularly interested in the point in which a pilot intervention becomes viable policy. Here specifically I want to try to understand what the state of the evidence is in favor of unconditional cash transfers in the developing world, and how your research there might extend to the developed world. Can we start with an overview of the work you’ve done so far, particularly on cash transfers in Kenya?
johannes haushofer: The main completed study that I’ve done so far is a randomized controlled trial [RCT] on GiveDirectly’s UCT program in Kenya, that we finished maybe five years ago, and published two years ago. In that study, we delivered transfers that are on average $700, which is about two years of per capita consumption, to poor families in western Kenya. And we found pretty sizeable effects on outcomes like consumption, asset holdings, psychological well-being, and income. That piece of evidence is part of a larger body of evidence, which is that cash transfers do a bunch of good things. So they increase consumption, and other welfare outcomes we care about. There are lots of studies that make us think that. And there aren’t a lot of the negative effects that people were original worried about—temptation goods, conflicts, violence, and so on. We’re just now working on a paper that shows pretty large decreases in domestic violence, as a result of cash transfers.
I would say that the state of the evidence is: cash transfers do pretty good things. The 1.0 question for cash transfers is, “Is it better to get cash than to get nothing?” I think the answer to that is yes, it’s better to get cash. This isn’t surprising to many people, but it is surprising to some people who thought the poor are bad at handling money, and were going to blow it and make their lives worse.
ms: On the temptation goods question: did you find that to be context-dependent? What I mean is, you could imagine a situation where a recipient has two choices, a certain kind of leisure or a very attenuated labor market in which there is only one job, which could be, say, working in a mine, which is miserable. And so cash, in this otherwise pathological society, would lead to people defaulting to temptation goods because it and a miserable form of labor market participation are the only options. There is some world in which that is likely, I would imagine. Did you see any variation? It’s heartening to hear that overall, that’s not where money was spent.
jh: We didn’t see a lot of heterogeneity in the effects on temptation goods consumption. That’s true pretty much across the board: it doesn’t go up, if anything it goes down a little bit.
Can I imagine groups that would increase their temptation goods consumption if we gave them cash? Maybe if you had a pretty serious addiction, you would increase it, but at the moment the evidence is not really there.
sidhya balakrishnan: Evans and Popova found that cash transfers actually reduce temptation goods consumption.
ms: So you have the study and attendant paper from 2016. And you’ve recently just completed another study.
jh: Yes, now we’re doing a general equilibrium project, with Paul Niehaus, a founder of GiveDirectly, and Edward Miguel and Michael Walker at UC Berkeley. There, we’re interested in price effects mainly: is it the case that when you give these large cash infusions, no one is really better off, it’s just that prices go up? The answer to that seems to be no. We haven’t finished writing the paper, but it looks like prices don’t change a lot.
ms: Is this based on your first cash transfer? Is this a basic income study?
jh: No, this is a new study. It’s a much larger project, that is much better powered to detect price effects, that we weren’t so well powered to detect in the first study. In the first study, we did look at price effects and didn’t find any, but that was just a small number of villages. Now we have 650 villages, and $11 million in transfers, which corresponds to about 15% of GDP in these villages. That’s now a magnitude where you can compare it to a very large government stimulus.
ms: In this new study, does every person in the village receive it, or a select group?
jh: Everybody who is eligible receives it. GiveDirectly has now moved to a model where they give to everyone who meets the eligibility criteria, which in our new study was having a house with a thatched roof. We do study within-village spillover effects, but those are on ineligibles.
ms: Do you see variation in price effects on different kinds of goods, e.g., landlords do increase rent but shops don’t increase prices?
jh: We haven’t broken it down yet into very small, granular categories. I can say that at the coarse categorical levels of durables, non-durables, and temptation goods, we don’t see effects. We haven’t looked yet at, for example, does fish become more expensive, does meat? In first study, we saw a decrease in the price of fish, for whatever reason, though that could have been just a fluke. In this new study, we don’t know for the individual categories quite yet.
ms: So you have the 2016 study, and this new study on price effects. What else are people doing on unconditional cash transfers?
jh: So, as I mentioned before, in cash transfers 1.0, the question is, “Is cash better than nothing?” Now, the next important question is, “Is cash better than other stuff we could be doing with the cash instead?” There was just a really cool study by Craig McIntosh and Andrew Zeitlin where they compare GiveDirectly transfers to a USAID program in Rwanda. They find that neither the nutrition program nor the cash transfers at the same magnitude have a lot of impact, but much larger cash transfers do have impact. So that’s a really great first step in the direction we need to go in. We need to see: where is cash useful, and for what is it useful? I’m involved in a study in Kenya just now where we’re giving either cash, or psychotherapy, or both, or nothing, to a group of people who either start out depressed or healthy. The idea is to see: which of the two is more effective? Are they most effective in combination? Which groups of people is each intervention most effective for?
ms: This reminds me of other interesting work you did--your paper called “Depression for Economists.” Did that inspire this study? Could you explain that a bit?
jh: That paper was a reflection of my increased interest in depression over the last few years. That in turn comes from my original interest, which was about psychological mechanisms that perpetuate poverty. I originally had this thought that maybe poverty leads to stress, and stress impairs your ability to make economic decisions, which leads to more poverty, and so on. That second link, the link from stress to economic decisions, I think it looks pretty weak at the moment. The lab evidence is kind of there, but it’s not like stress has really large effects on patience or related variables. So I looked for other mediating psychological variables, that could be a psychological poverty trap, a true trap, in the strong sense. Depression seemed plausible, so I got interested in that. It’s definitely a consequence of poverty, and very likely a negative contributor to labor market outcomes.
ms: You have a specific definition of depression in the paper. It’s actually a microcosm, in certain respects, of economic depression—a belief in diminished returns.
jh: Exactly. We started by asking, what does depression even mean? That’s when I went and read the diagnosis criteria in the DSM-5, and those originally proposed by Beck. That’s how psychiatrists diagnose depression. When I read through this list, the criteria seemed to be related, to a surprising degree, to economic primitives, and specifically beliefs about the returns to effort. Depression is characterized by pessimistic beliefs about whether your efforts will amount to anything, whether it’s even worth getting out of bed in the morning, whether you’re just a burden on others... The thrust of that paper was to see how much mileage we can get if we use the language of economics in understanding the symptoms of depression. The claim is that many of these symptoms actually look like pessimism about the returns to effort. So far that’s just a theoretical exercise, to reconceptualize depression a little bit, for those of us who have economics training. But we’re going to try to do some empirical work to see what really changes in depression. Do beliefs about returns to effort truly change? That would be the prediction.
ms: Would that be a lab study?
jh: We’re doing that in the context of an existing psychotherapy RCT in India. You can’t induce depression, so the way to study it is to alleviate it and to see if that has effects on the variables you care about. There’s an existing study that randomly administered psychotherapy, or not, to depressed people, run by Vikram Patel and colleagues, and we are currently following up on this study with a team that consists of Gautam Rao from Harvard, Frank Schilbach and Pierre-Luc Vautrey from MIT, and Jonathan de Quidt from Stockholm to understand to study the effects of that treatment on psychological and economic outcomes; beliefs about returns to effort, time and risk preferences, and so on. Many of these measures are collected in a lab-in-the-field type setting.
ms: How would cash intervene in depression, intervene in people’s beliefs in a way that would stimulate a change? Versus psychotherapy? For example, could it be the case that cash could give you a boost in your mood--which might be enough to help someone over a hurdle, to get them seeking help? I could imagine that cash could get someone increased access to different kinds of labor market opportunities--you’d now have money to take a bus to do an interview further away. Another way to frame this is: especially if you’re impoverished, there could in fact be low returns to your effort, in many contexts, making a belief in low returns to effort an accurate representation of your circumstances. And then the cash, in such a situation, might raise you to a new income level at which your returns to effort are in fact higher.
jh: I think that’s the best answer; cash could enable you to make investments that raise (or even just remind you of) your own productivity, and then potentially change those pessimistic beliefs. Now, this isn’t a watertight argument, because our point of departure was that, if you suffer from depression, you misattribute the things that happen to you; e.g., a negative shock is attributed to your abilities, a positive shock is attributed to luck. In that case, a cash transfer wouldn’t necessarily translate into less pessimistic beliefs about returns. But it might get you back to sampling the world. In the economic model in our paper, one of the mechanisms by which people get stuck in a trap in is that they stop learning about potential returns. So when you get a negative shock, you may transiently have pessimistic beliefs, but over time, if you’re still sampling the world, you learn the truth. But if the shock is large enough, it could lead you to a point where you stop sampling the world. And if the cash transfer is large enough to make you do that again, then you could over time learn your true returns again. This could be a mechanism that acts in concert with what you were saying--now your returns are higher just because the cash makes them so.
ms: To return to cash, let’s review where we said the state of the evidence lies. You introduced a distinction between 2 counterfactuals: Cash 1.0: where the counterfactual is no cash, and Cash 2.0, where the counterfactual is comparably expensive interventions. You mentioned initial work there. I’d like to introduce another distinction, between one-off transfers that are meant to spark a new economic trajectory, and longer-term transfers such as basic income. Is that a relevant distinction in developing country contexts? Does that complicate the evidence you described?
jh: Yes, I think it is different in the sense that these long-term transfers, if they are truly unconditional basic income that you get forever, would represent a change in your permanent income. That means you should adjust a whole bunch of things, including consumption levels, for the rest of your life, and so that makes different predictions from these one-off transfers. If you take the strictest model of the permanent income hypothesis, then a single, one-time cash transfer wouldn’t change your consumption much at all -- it’ll change it by the returns on the transfer, which in these contexts are not zero, but they’re not super large either. But it’s still interesting to ask what are the effects of these short-term transfers, because those are feasible from a policy perspective. And it turns out that the effects are actually larger than many of us expected. Even in terms of long-term outcomes, there’s some evidence to suggest that they’re useful. Asset holdings are still higher a couple of years later. Other evidence, e.g. recent work by Chris Blattman and colleagues, suggests more modest long-term returns. And on the really long term results, the jury is still out.
ms: Let’s stick with the developing-world context for the moment more, before we try to transfer the evidence to the developed-world context. At what point, from your view, which I acknowledge is not a policymaker’s view, does evidence cross the threshold to be able to claim that cash transfers are a good policy for a wide variety of policy objectives, poverty reduction, wellbeing goals, and so forth? At what point can you say, “This is likely to be an effective policy?” If we aren’t there yet in the developing context, what else needs to be proved?
jh: That line is crossed when there is solid, probably randomized controlled trial or some other good identification strategy evidence, that the policy works. By works, I mean, it affects outcomes that policymakers care about. For cash transfers, we have pretty good evidence that for short and medium run they do have those positive effects. Then, there’s a second question, which we talked about earlier: if you’re choosing from a menu of options, which one should you choose? And that’s a question that scientists can’t really speak to--it hinges on what you, as a policymaker, prioritize. It depends on what variables you care about more than others. Now obviously, if one of these interventions costs as much as the other but dominates on every outcome measure, that’s a no-brainer. But often there are tradeoffs, and those are tradeoffs that policymakers need to make. So do you choose the one that’s better on education and worse on consumption, or vice versa?
ms: Or even, for instance, how much do you as a policymaker understand your constituents to value autonomy? Is that an independent value?
sb: I have a segue from that. Where do you stand on conditionality of cash? My own perspective is that it works differently for different research questions.
jh: Again, it depends on what you want to achieve as a policymaker. If you really care about health and education outcomes, maybe you want to make it conditional. The downside of conditionality transfers is that you often can’t reach the very very poor, because they have trouble fulfilling even the very mild conditions that you put on the transfers.
sb: In addition to that, the arbitrary means test can leave out people just marginally above it.
jh: Yes—I would put a lot of weight on the extremely poor and making sure they’re not excluded. That biases me a little against conditionality. But that’s also partly true because the effects on the outcomes that you’re conditioning on are there, but the welfare gains are not so massive or important that we can’t afford not to have them, and that leaving out the ineligibles or the people who can’t meet the conditions is justified. So, if I were a policymaker, I would have some combination of both, and try to get the truly poor people with unconditional transfers, then for people who can afford to meet conditions. That seems like a useful combination.
sb: Likewise, do you have a strong perspective on in kind versus cash?
jh: It’s been a while since I caught up on that literature. My sense is that often in-kind transfer programs are at least partly extramarginal. And so, from that perspective, and also from the perspective that when you just give money there’s no good evidence that people are throwing it away and buying stuff that no policymaker would want them to buy--I don’t see a strong argument for in-kind transfers. Unless you’re in a setting where markets aren’t functioning well, for example, if you can’t actually buy things, or there’s price-gouging.
ms: To follow up on that: is the burden of proof on the in kind side or the cash side? It seems like you ought to have, as a policymaker, a strong reason for thinking that your constituents can’t represent their preferences well.
jh: Right, but where the burden of proof lies depends on where you start as a policymaker. If you’re in the world where cash has always been this preposterous idea, “Of course people are going to waste cash, we should be giving out rice instead,” then obviously the burden of proof is on cash transfers. For economists, the burden of proof is on the in-kind intervention; we would ask, “Why would we expect in-kind to be better, ever?”
ms: To the last theme: the transferability of evidence from developing contexts to the developed world. In the last 18 months, I’ve seen a number of new pilots in the United States, and Ontario and Finland had significant PR events, let’s say. Having conducted most of your research in the developing world, how do you think about applying that to the developed context? I understand that, from an academic perspective, obviously there are serious limitations to that extrapolation. In an expanded view, how do you think about that, and where can evidence from the developing world play a role in a developed country’s UBI research, in pilots or policy?
jh: Cash transfer studies from the developing world can provide motivating evidence and hypotheses. There are two things I would say: the first is that there are definitely cultural differences, but at the same time, people are people, and many of the economic behaviors that are relevant are probably fairly universal. From that perspective I’d be optimistic about knowledge transferring. The second thing to say: That’s true a forteriori, in the sense that, in order for you to think cash transfers would work differentially well here versus a developing country, you’d have to think the population you’re targeting is somehow different in one place versus another. I could see somebody saying, “Well, in Kenya everybody’s poor, so you’ll get a mix of very able and some less able people when you give cash transfers.” In contrast, for programs in the US they might say, “The people who know how to handle money have already made it, so you’re only getting the bottom of the distribution, the people who don’t know how to handle money.” That would be an argument that would lead you to be skeptical about whether cash transfers would work here. But I’m not a believer in that argument, because my sense is that there’s enough randomness here, in terms of starting conditions, socioeconomically, in terms of race and ethnicity, to make it the case that there are plenty of poor people who happen to be poor by bad luck, not by some “deficiency”. And so I’m optimistic.
ms: And not just deficiency in navigating the labor market, because even if you there were a deficiency specifically in handling the labor market, you could presumably have no deficiency in managing your own consumption. In which case it would still be a benefit. It would have to be a pretty robust thesis of deficiency.
jh: Yes, it would have to be that you’re a bad decision maker in any area of life.
ms: Interesting—this is a higher degree of confidence than I thought you were going to give. That’s exciting for our work.
sb: We’ve been wondering a lot about transferring evidence from developing to developed countries. Those two contexts are quite different as far as systems, markets, and institutions are concerned. They’re systematically different. Welfare systems are different. If a cash policy has to become a policy in the developed world, it’s more a replacement of an existing welfare system, whereas in the poorest of the poor countries, you’d be building it from scratch: it is the welfare system. We see differences also in the way that cash transfers are administered, in pilots or experiments. In developing countries, it’s nonprofits or academic researchers. In developed countries, it’s increasingly public-private partnerships, with city and state governments. I think that would change the game quite a bit, in terms of what kinds of evidence we’d be able to say are similar. They are operating in different institutions and markets.
jh: Yes, I totally believe that that’s plausible. Tessa Bold and colleagues just published a paper on teacher hiring and contracts in Kenya, where they show that the same program, administered by an NGO versus the government, has completely different effects. I’m sympathetic to the point of view that how things get implemented is super important. And the partner organization, or implementing organization, is only one way in which that can matter. The institutional context more broadly is super important. I guess what that means: if you’re translating a policy into a new area, you have to make some educated guesses about how the institutional context is going to be different, and adjust accordingly.
sb: That’s extremely hard to do. First, you rarely get effects sizes from studies, they don’t support it. Second, even if you want to do preliminary power calculations, you have to hypothesize the effects, which is hard when you don’t know how a public-private partnership of cash works in developed contexts. So, it complicates your experimental design quite a bit.
jh: Absolutely, and I think there’s a blind spot in this literature: the attention to implementation context. There’s a whole area waiting to be opened up there, which is partly done by this paper. That should be studied more.
ms: Ending this line of questioning, I’ll ask for a summary. At the highest level, where does the evidence stand? It sounds like at the micro and meso, and even the macro level, which you’ve begun to looking at, confidence in its efficacy, even when the counterfactual is in-kind or conditionality, is improving significantly. Is that fair to say?
jh: I think that’s right. There’s pretty good evidence that it does good things, and I think open questions are: Is it more effective than other stuff? What are the medium- and long-term results? And how important are the spillovers?
ms: And it’s a closed question that it does not increase spending on temptation goods?
jh: Closed is a strong word, but there’s pretty solid evidence now that there’s not a lot of that.
ms: Do you have any last words you’d like to add?
jh: There are two things I’d like to say in the cash transfer debate that get brought up in comments online and elsewhere. One, there is a sense in which, if we think the poor are adults who should be making their own decisions, we shouldn’t even necessarily be studying what they do with the money, because they, by definition, know best what to do with it. That’s worth having in the back of our minds, as we opine on whether they’re using transfers wisely. In these conversations around, are these transfers spent on good things or bad things, we ought to keep in mind that there’s an argument that we shouldn’t care. Two, these are randomized experiments, and I want to emphasize that that we’re always very careful in our work to treat people as respectfully as possible. Sometimes the public misperceives this, and thinks that we are treating people as guinea pigs. But that isn’t true, everyone joins voluntarily, and they get a token of appreciation for their time, even when they’re in the control group.
ms: Thank you for everything, this has been fascinating.