Sam Carter and Joe Huston: The science behind Universal Basic Income

Link post


In this discussion from EA Global 2018: San Francisco, Sam Carter and Joe Huston discuss universal basic income: what we know about it, what we’d like to know about it, and what research we’re working on to bridge the gap.

A transcript of their talks, including their answers to audience questions, is below. We have lightly edited the talks for clarity. You can also watch them on YouTube and read them on effectivealtruism.org.

Sam’s Talk

We’re really excited to be here this afternoon to talk about Universal Basic Income, or UBI. We think it’s important to talk about this, because UBI is a big idea that has the potential to transform how we, as a community of effective altruists, donors, infomentors and researchers, think about doing good in the world.

I’ll start us off today by introducing UBI and the debate about its effectiveness, and discussing how existing evidence from RCTs can help inform how we think about this debate. Then I’ll hand it over to Joe, who will talk about some ongoing UBI research, and wrap up with some thoughts and questions about what this means for all of us as effective altruists.

If you’ve kept up with the development economics news over the past year or so, you’ve probably heard a lot of buzz about UBI. UBI is a cash-transfer that is, as the name implies, universal, meaning that all people in a given area receive it. It is not targeted at specific populations, and there are no conditions placed on how the money may be used. The transfer is regularly recurring, delivered over the long term, and sufficient to meet basic needs.

Around the world, there are studies taking place to better understand the various iterations of UBI, and their potential impacts. So, what explains the widespread interest in UBI today?

On the one hand, in many high-income countries, like the US, the economy is being rapidly transformed by technology. Automation is getting cheaper and better every day. According to one study by economists at MIT and Boston University, each robot that came into the world between 1993 and 2007 reduced the number of available jobs by 5.6. Automation has already displaced workers, and likely will continue to, especially in certain industries. This makes UBI a potentially attractive policy option, for those who are concerned about the wellbeing of workers left behind by automation and technology advances.

On the other hand, while the world has made large strides in reducing poverty, more than 700 million people still live on less than $1.90 per day. Especially in low income countries, where the extreme poor live, UBI is seen as a policy option to help bring adults and children up to the poverty line, and ensure that their basic needs are met.

In high-income and low-income countries alike, there are often concerns about the effectiveness of existing social safety net programs. People wonder if current programs are as effective as they could be. Programs might stigmatize recipients, for example, or be ineffectively delivered. And, with the proliferation of the gig economy and alternative work arrangements, employer-based approaches to social welfare, like many of those that we have in place today, may not be suitable in the future.

Like with any potentially disruptive new approach, there are supporters and detractors of UBI. Supporters argue that UBI might be more efficient than existing social programs, that its unconditional nature offers recipients flexibility and autonomy, and that new technology makes it more feasible than ever to distribute basic income transfers.

On the other hand, detractors raise concerns about UBI being too costly to be a reasonable solution, and about the ways recipients might use unconditional cash. Also, despite new technology, the infrastructure to receive payments may not exist in all contexts, especially those where the poorest people live.

There are valid points on both sides of this debate. How can we, as effective altruists, decide where to put our weight as a community, and what more do we need to know?

Let’s look at some existing evidence. First, let’s consider the concern that receiving cash with no conditions attached might lead to a reduction in work. On average, there is evidence that it doesn’t lead to such a reduction. Next, let’s consider concerns that cash given with few or no strings attached will be squandered. In Kenya, researchers tested a program implemented by GiveDirectly, which transferred a lump sum of money to poor households. They found that it led to increases in both economic and non-economic wellbeing, and did not lead to increases in spending on temptation goods like alcohol or tobacco.

What’s important to note, however, is that these outcomes were measured less than a year after the transfer was given, and if we’re trying to understand cash transfers as a way to reduce poverty, and if we’re trying to think about them as a way to inform our opinions about UBI, we should probably consider longer-term effects.

Fortunately, we have some research on that as well. Researchers in Uganda and Sri Lanka have found that unsupervised cash grants, distributed in a business setting, had positive impacts on business investment, profits, and household income after 4 or 5 years. In Uganda, the grants were given to groups of young people who had submitted business plans. In Sri Lanka, grants were given to micro-enterprise owners. If we want to understand the relevance of cash transfers for UBI, however, we might want to think about cash outside of the context of business.

For a less business-centered context, we can turn back to the study that researchers did of GiveDirectly’s program in Kenya. The researchers, Johannes Haushofer and Jeremy Shapiro, released longer-term results from that cash transfer. They’ve generated a lot of discussion.

Overall, the results of that study were largely ambiguous, mostly due to some methodology challenges. There were different attrition rates between the households who received transfers and those who didn’t, among other methodological challenges. But even though the results were pretty ambiguous, they’re still an important input into the broader evidence landscape around cash transfers and UBI. And, luckily, GiveDirectly is doing many other RCTs and many other studies whose results will continue to inform this discussion.

When GiveDirectly and the researchers started the study, they designed the evaluation to answer some really important policy questions. They assigned households to three different groups, and that allowed them to figure out what happens to people who receive cash, people whose neighbors receive cash, and people whose communities are not exposed to cash at all. After looking at the three-year survey results, the researchers found that most of those positive impacts that I mentioned after 9 months had disappeared.

Households who had received the cash transfers did have more assets, but the impacts on consumption, investment, and happiness were no longer apparent. And there’s also some suggestive evidence that households whose neighbors received cash, but who didn’t receive cash themselves, experienced negative spillover effects, but again, this evidence is subject to some methodological questions, and we’ll want to have more research on this question in the future.

Joe is going to tell you about some of the efforts that are currently ongoing to get more evidence on these questions in a little bit. But before he does, I want to return to one of the motivations behind UBI, which is reducing poverty.

We’re not yet sure if basic income will lead to a sustainable increase to economic or non-economic wellbeing of the households living in extreme poverty. But while we wait for results, there is already evidence on one approach, which is targeted to households living on less than $1.90 a day, that can sustainably improve livelihoods. It’s called the graduation approach, and it was initially designed and tested by BRAC, which is an NGO in Bangaladesh.

The graduation approach gives a transfer of a productive asset, which is meant to be the core of a small business. It also includes other complementary services, including coaching and consumption support, usually in the form of cash transfers. This is an actively managed program, in which households receive two years of support. JPAL-affiliated researchers have evaluated the program in Bangaladesh and in 6 other countries around the world, and found that overall the program had positive impacts on most measures of economic and non-economic wellbeing, both when the program ended, and a year later, after all the program support had ended. And in two sites, where it’s been tested four years later—so, 7 years after the asset was transferred originally—there’s evidence that consumption, income, assets, and psychosocial well-being all continued to improve. This suggests that the changes caused by this approach have lasting power.

Obviously, a program that involves all of these components is more expensive and more hands-on than delivering cash alone. So, researchers and implementers alike are really interested in figuring out whether the same good can be delivered in a less intensive version, or with fewer components.

In Ghana, researchers tested just the asset transfer alone, and in Uganda, researchers tested a cash transfer that was roughly equal to the cost of delivering a streamlined version of the graduation approach. And in both cases, they found that this sort of capital-infusion alone did not have the same positive effects on consumption, investment, or asset ownership. So, there seems to be something about the full package of the program that impacts households differently than capital alone can.

So what does this mean, going back to the question of basic income? One interpretation is that households who are living in extreme poverty in lower and middle income countries face a number of pervasive market failures, like limited access to capital, limited information, and low trust in market institutions. The full package of graduation seems to be able to mitigate some of these market failures, in a way that cash alone may not be able to. So, in these contexts, we still need more evidence on whether cash transfers that are sufficiently large, or delivered for a sufficiently long duration, can overcome some of the barriers that existing evidence—that I just mentioned—have not been able to.

It also raises interesting questions about cash transfers in high income countries like the US. Here, the market failures I just mentioned are likely to be less pervasive, and less complex than they are in the countries where we’re trying to reduce extreme poverty. But of course, we still need more research across contexts, to understand what impact something like a UBI will have on the household to receive it.

Joe’s Talk

I’ll start with saying that I’m really glad to be here. I originally found GiveDirectly through the effective altruism community, so conversations like these are very near and dear to my heart.

A core piece to start out with, when we’re thinking about “How would a universal basic income be different from some of the other things we already know about cash transfers” gets into what’s unique about a universal basic income—that particular structure of cash transfer.

First, it’s universal. Whole communities are receiving the cash transfer, relative to targeting specific income levels or specific levels of vulnerability. So, you have a lot of variation in the types of recipient. It’s not just business owners or people in extreme poverty. Instead, you have some variation in terms of whatever the communities look like.

Second, it’s a particular amount: it’s basic. It’s sufficient to cover basic needs, which differs from some of what we already know about other sorts of cash transfer studies, which might be smaller, or big, one-time capital grants. It’s a particular size of cash transfer.

Third, it’s an income. It should be something you can rely on for your entire life, to provide a permanent cash floor, so that you’re always up to a certain standard of living. In contrast to one-time grants, like what Haushofer and Shapiro studied, or grants targeted towards a particular life stage -like a pension, or support for people who have kids in secondary school—a universal basic income should follow you for your entire life, and provide long-term cash support.

So what do we know about that type of cash transfer? Well, from what Sam told you, there are a lot of studies of cash transfers broadly. Literally over 100 studies on cash transfers all over the world. And more particularly, there is a handful of studies that get grouped in specifically about universal basic income.

In the 60s and 70s, there was a wave of experiments in the US and Canada that tested a variant of a universal basic income called a negative income tax. From those—they typically weren’t universal; they were targeted towards the poor. The experiments also weren’t universal because there was a lottery-like system, choosing people randomly to participate in the studies. So you didn’t get as much information on community-level effects. The studies were large, however, and usually they were basic: They provided a meaningful level of support. But they weren’t typically long-term—usually just a few years.

What you saw from those studies is similar to what you see in a lot of the cash studies. First, people got money, and so they were mechanically less poor. And then you saw that flow through to more broad-based improvement. You saw educational attainment increase. In Canada, you saw hospitalization rates fall pretty markedly. One thing that’s different from the Developed World studies that doesn’t show up in the Developing World studies, is that you did see modest reductions in work effort. But where those showed up, they were showing up in populations where you might be more okay with people working less. They showed up in teenagers, who worked less and went to school more, and young mothers, just after giving birth.

Otherwise, there’s a handful of studies in Namibia and Maja Pradesh in India, that looked at universal cash transfers. But, in Namibia it wasn’t a randomized controlled trial, and didn’t last for very long: only a couple of years. And in India, where they did use that experimental approach, it similarly didn’t last very long.

Then, recently, as you’ve probably seen in the news, there’s been an explosion of studies, basically all over the world. GiveDirectly has launched one in Kenya, there’s one in Finland, there’s one in the Netherlands. There’s been demonstration projects, such as one called My Basic Income in Germany. And there’s one coming up in Stockton, California as well.

Also, there’s been two larger randomized controlled trials that are coming onboard. First, in Ontario, done by the provincial government there. And second, by YCombinator in California. Because these studies are expensive, researchers have typically made cost-benefit calculations to prioritize what they’re studying. And so, almost none of these has been universal, in the sense of looking at whole communities. The YCombinator one isn’t looking at whole communities, but they are looking at some income variation, and so they are doing their best to balance what they look into.

Another issue is that typically, cash transfer studies haven’t gone that far out in terms of long-term, which might give you a question about whether that matters or not in interpreting them.

GiveDirectly took the approach of trying to design a study that would complement the existing body of research, including the body of research about cash transfers in general. So, our study is randomized at the village level. Whole communities will be treated differently, with about 44 villages receiving 12 years of monthly payments—about $23 - sized to sort of meet the poverty line in rural Kenya. That group of villages will be compared to villages receiving 2 years worth of basic income payments, and a group of villages receiving one-time cash grants of about $500 per adult. These one-time payments are roughly equal to the sum of the 2 years worth of payments elsewhere. Also, all three groups will be compared to a control group.

The research is being led by development economists, like Abhijit Banerjee, who founded JPAL, and Tavneet Suri, who is also director at JPAL, as well as Alan Krueger, who is the former chair of the Council of Economic Advisors for President Obama. We hope to bridge the policy worlds across the developing and developed world.

The actual surveying itself is being done by Innovations for Poverty Action, and we should have results out early next year, which is actually when we’ll have an explosion of results out from GiveDirectly, Canada, and Finland too.

With studies like these, you want to look at a broad set of outcome variables. And so we’re looking at both individual level outcome variables, in terms of earning or spending or assets, occupations and time-use, gender relations and risk-taking. At the individual level, how do those things evolve when UBI is introduced? We’re also looking at community level effects, because there’s something unique about this type of cash transfer, that everyone in society—or at least most of the people you’re seeing day-to-day—are also receiving that cash transfer. We’re looking at things like community level economic effects, access to health, education or water facilities, how things like community or political engagement evolve and how crime levels change.

Then, we’ll look at both individual and community level effects relative to the differences in type of cash transfer. How it varies between the 12 year group of villages, and the 2 year group, which should help inform how we should think about other studies, that are relatively shorter term, as well as the structure of how cash grants achieve different goals or different outcomes versus recurring stream payments.

And then, because it’s universal, we have a rich set of data of different types of people receiving a cash transfer, and then seeing if they spend it differently. How do different starting-income groups spend cash differently? How about different age groups, or genders? This information should help inform broader policy questions beyond just those about UBI.

So, how should the effective altruist community think about all this? I think there’s a few different types of questions the community should try to answer:

The first one is what will be the direct impact of these cash transfers? For this, we have rich data on other types of cash transfers, from literally all over the world, where, if you give cash to people who don’t have it, they are immediately a little bit less poor, because they have cash, and you start to see that flow-through into broader wellbeing: things like increased consumption, or increased assets and earnings. You also see it show up more indirectly, like reduced stress or improved psychological wellbeing.

In Malawi, you saw women whose families received small, recurring payments get pregnant later, marry later, and have lower rates of HIV, because they had a little bit more security in society. And so, part of the bet on effective altruism, is a bet on how it’ll differ from other types of cash transfers. Which we can maybe dig into in the Q&A.

The second answer we should anticipate, is what the broader policy impact of this type of study will be. And this applies to any of the—I think it’s 12 - studies that GiveDirectly is doing. To give you one example of what the policy potential of our universal basic income study, this graph is taken from an estimate by the government of India, where they modeled what might happen if they replaced existing subsidy programs with a modest universal basic income, essentially universally. It shows what that change would do to poverty in India.

What they estimated was that it would take poverty from about 22% to 0.45%. Literally bringing over 100 million people out of poverty, basically immediately. That doesn’t necessarily mean that that’s what we should do: there’s an opportunity cost for this type of spending. There might be other effects of the cash transfer that are worth studying. But it gives you a sense of the potential scale of policy implications of this type of research, beyond the 21,000 people who are receiving cash from GiveDirectly in the shorter-term.

The third question for the effective altruism community to try to predict is more foundational. We want to look at the process that we mostly apply, at least for effective altruism, to help existing poor people today. Where it looks like the following process:

  1. Choosing a set of outcomes that we think are important, such as earnings, health, or how long people live;

  2. Choosing goods and services that we think would improve those outcomes, whether it’s a graduation program or a deworming pill, or a goat;

  3. Choosing providers who we think can deliver those goods and services as best they can.
    I think a lot of what the effective altruism community is focused on is doing this process better than how aid and development have traditionally done it before.

What I think is helpful about cash transfers is they pose a more foundational question of who should make these choices, and what types of outcomes are important. We want to examine how you value earnings, or not getting rained on, or being a little bit extra healthy. How you value improvements in the lives of your children. The existing status quo is, donors and funders start with a pool of money, and almost by virtue of that, they become the choosers for how it gets used. But cash transfers literally put the budget in charge of the people we’re trying to help, and lets them make those choices.

I think that switch of power is important, in part just because of practical reasons. There’s a tale of two track records where the poor have over 100 studies backing up the quality of their decision making when they control their own budgets. And then, the funders, implementers, have the track record of whatever you think about the last 50 years of development history.

Now, it’s a genuinely tricky problem to figure out what types of outcomes matter. In part, this is because people are very, very different. We see GiveDirectly recipients buy an incredibly diverse array of things. We’ve seen solar panels, livestock, roofs, and even someone deciding to start a band. There’s an incredible diversity in humans. And that’s amplified by the fact that most of the decision making is done by people in San Francisco, New York, DC or London. And it’s on behalf of people who are living in places like rural Western Kenya. Doing the relevant calculations—which we can try to do better, as effective altruists—is inherently very, very difficult.

Which is not to say that the entire development aid portfolio should shift towards recipient-chosen, or recipient-decided, but existing portfolios are extremely skewed the other way. Basically, any way you cut it, whether you look at humanitarian aid, or institutional aid given by governments, or US international charitable giving, basically the whole pots of money meant to help the poor are decided by the rich. This is an unintuitive result, given what we’ve talked about before. So I think one of the most helpful things that cash transfers can do, with UBI being an example of this, is force the effective altruism community to answer that more foundational question of who should decide which outcomes matter, and what types of goods and services best achieve those outcomes.

Q&A

Question: Negative spillovers when some people in the village get money and others don’t: what does that look like? How does somebody’s life get worse in that situation?

Sam: Researchers have done a little bit of work to try to understand what this means. There’s one paper that they’ve released, which finds that there might be some negative implications on people’s wellbeing. So, people might feel more stressed or depressed, if their neighbor is receiving cash and they aren’t. In the 2018 paper, which is the 3 year results that I mentioned, the researchers posit that households that didn’t receive cash were selling off assets, so they ended up owning fewer assets. Households who did receive cash, as I mentioned, had more assets at the end.

Joe: Yeah. I think the 3 year study is one that’s genuinely hard to interpret, because of the methodological things which we can dig into. It’s a tricky study to synthesize and know what to do about. In 2014, we launched a study specifically geared to answering this question in a high-quality way: The question of how non-recipients of cash get impacted by being near recipients of cash. We randomized the concentration of cash transfer delivered in different regions. That way, we can really highlight how people are affected. That study will have results out around August or September this year.

I think the other question for the effective altruism community is how those results, in part, can inform research design, about how you should approach individually randomized studies where people are very close to each other. For GiveDirectly’s programming, we’ve moved very, very far away from hand-picking individuals from communities. In general, we’ve moved much more towards enrolling almost everyone. That’s literally the case for the UBI study, but even for our day-to-day programming, we’re moving much towards saturating villages. And so it’s a little bit harder to know, without seeing the other study, which will look across villages as well, how to draw the right conclusions from that 3 year study for what we do today.

Question: A clarification question. When you mention the graduation program, the concept of asset transfer, I’m picturing a cow. Is that the right image for me to have in mind?

Sam: Yeah, so the graduation program is always tailored to whatever context where it’s being implemented. Usually it’s livestock, but it can also be supplies to open up a small shop and sell berets, or makeup, or cigarettes, whatever makes sense for the community.

Question: What are the challenges of running a study that’s going to go on for 12 years? Things are going to change in the world around you. I mean, it’s always hard to hold things constant, but, in that span of time, it gets even more difficult. How do you think about those challenges?

Joe: Yeah, it’s tricky. You have to guess what you think your follow-up call center will cost in 2029, for example. You have to figure out how to manage the cash between Kenyan banks and US banks, to maintain a standard payment across those 12 years. You have to think about how to increase the transfer’s size with inflation. It’s a bunch of really hard questions, that we’re trying to be pretty paranoid about, in terms of playing out different scenarios. If GiveDirectly got shut down in Kenya for whatever reason, could we continue payments if we wanted to? Things like that. We have a bunch of specific answers to those little questions, but it required a decent amount of work.

The nice thing is once you get past this hurdle—we’ve now enrolled 21,000 people—it’s a steadier state. The picture looks like: once a month we send out payments, and people receive text messages, and then once every few months we have our call center call them. So that’s a much lighter level of engagement, that we have to keep up for the next decade or so. So that piece of it is a lot easier than the initial lift to get all those people enrolled.

Question: When you think about the tradeoffs between a negative income tax and the universal basic income, why give the universal basic income to everyone? For those that don’t need it, it seems wasteful. How about the universal the non-universal negative income tax, as an alternative?

Joe: One thing I should note is, you can achieve pretty similar distributions of incomes post-tax, as in post-transfers, with either negative income tax or UBI. Because with a UBI, if you give Bill Gates his $12,000 a year, you end up taxing it back. And so you can end up with the same distributions for both settings. I think the tradeoffs become in part political: what’s more palatable? Sometimes it’s easier to say, let’s give cash to the poor. Other times, it might be easier to get people on board with universal support, that’s, everyone’s dividend is part of being in society. Other people also argue, which I’m looking forward to seeing, that there’s something helpful about the cash support being universal, so that it’s not stigmatizing. That it’s not just the dole for the poor. If it’s less that and more a universal floor for everybody, you might think that in part politically it’s easier to maintain, but also maybe it has different effects. That people don’t view it as support they get for being poor, but rather, not deserving to starve as part of the societal bargain more generally. And that may have different sorts of effects in terms of how UBI is spent. Or maybe it will be spent more communally. Those are among the things we’ll try to test.

No comments.