Many readers of this blog told us they were inspired by the first story in our series on #nostringscash aid — about a ground-breaking experiment in Kenya to test the benefits of giving poor people a steady stream of cash in place of traditional aid.
But some questioned the ethics of studies like this.
At issue is a plan by the U.S. charity GiveDirectly to give an extra $22 a month to every adult in as many as 200 impoverished villages across Kenya. The payouts start this autumn and continue for the next 12 years. Over that time GiveDirectly will track how incomes, health and other markers of well-being in those villages stack up against a comparison group of as many as 100 similar villages that won't get the cash aid. (GiveDirectly already has a pilot version of the monthly payouts underway in one village, which was the focus of our story.)
Readers wondered: Is it moral for experimenters to bestow a benefit on one group of people and not another? And what are the risks of unintended negative consequences — creating lasting income inequalities between villages, for instance, or even fueling tensions between the residents?
These kinds of issues are hardly limited to GiveDirectly's experiment. Virtually any scientist who does experiments involving human beings inevitably confronts ethical trade-offs. We spoke with several experts on ethics and poverty research — as well as with GiveDirectly's director Paul Niehaus — about how they navigate the challenge.
Edward Miguel says the potential for unintended consequences is "a serious issue." He is an economics professor at University of California Berkeley who has done extensive field research on poverty programs in Africa. This work includes an ongoing study of GiveDirectly's original version of its cash aid program — which, over the last six years, has provided hundreds of thousands of people with relatively modest one-time cash grants of about $1,000. To date, Miguel notes, he and others have not found evidence of serious jealousies or conflict in affected communities. But he thinks the more extensive and long-term nature of GiveDirectly's steady income model could make it more divisive.
Miguel de Figueiredo, a law professor at University of Connecticut, agrees. De Figueiredo has grappled with ethical issues in the course of doing his own field work researching topics ranging from voter behavior to the impact of criminal law in poor countries. And he notes that it's only since the early 2000s that a growing number of researchers began setting up field experiments to study poverty solutions. The vast majority of these experiments involved much shorter time frames, probably because this type of work is still so new. In offering people monthly cash for 12 years "GiveDirectly is going into uncharted territory," says de Figueiredo.
Edward Miguel adds that the issue isn't just that paying out a steady income could produce larger disparities in wealth. It's also that "this isn't a setting where, when someone else does extremely well, no one cares." He says decades of sociological and anthropological studies across rural East Africa suggest there's a strong egalitarian strain in the culture: When someone comes into money there's an expectation that he or she will share it among relatives and community members — and a sense of resentment if they don't.
He points to a recent experiment involving rural Kenyans that illustrates just how powerful this social pressure can be. The researchers, Pamela Jakiela and Owen Ozier, gave subjects a small amount of money that they could use to win money through some games and told them if they wanted to keep their earnings secret from the rest of the group they would need to pay a premium. Almost a third chose to pay the premium, presumably because they figured if their winnings were made public they'd lose even more of it to relatives who would demand a share.
But Miguel says this mindset doesn't mean GiveDirectly should stop its experiment but rather that it should actively monitor for signs the study is fostering tensions. Paul Niehaus, a cofounder and director of GiveDirectly, says the charity will do just that. "I think we should have every opportunity to pick up on [rising tensions] if it's going on," he says. He adds that GiveDirectly is committed to publicizing the results on a continuing basis and will not hesitate to alter course if there is evidence the experiment is causing significant harms of any kind. "It's just the right thing to do," he says.
But Niehaus also says that of all the possible unintended results that worry him, the prospect of tensions due to jealousies actually ranks very low. That's because, in the case of its one-time cash grants, GiveDirectly found that people living near those who got the money often indirectly benefited as grantees quickly spent their newfound wealth — hiring neighbors to help them add onto their home, for example, or buying goods from local businesses.
In fact, when it comes to unintended downsides, it's this kind of spillover that concerns Niehaus far more. Increased spending by the recipients could drive up local inflation, he notes. If many people in the villages getting cash started buying an item, he says, that could bid the price up. Then it could become expensive for people in villages that aren't getting the cash — and "that would hurt people," he says.
Apart from the risk of unforeseen harms, what about the morality of giving only some villages in the experiment cash and others nothing?
The experimental method that GiveDirectly plans to use — randomly assigning its subjects into two otherwise identical groups, one of which gets the cash and the other that does not — is called a randomized controlled trial. And it's considered the gold standard of research because it's the most reliable way of ensuring that any observed differences in outcomes are actually caused by the factor being tested rather than unrelated factors.
By definition a randomized controlled trial requires those running it to treat the subjects of the experiment unequally. In this case some get the benefit of cash. Some do not.
Deborah Hellman, a professor of law at University of Virginia who specializes in bioethics, says that from an ethical standpoint there are two ways to think about this.
"The first is to say, 'Well the people who don't get the money aren't any worse off than they would have been if the researchers had never arrived.' This is an argument for why it's okay to do this."
But Hellman says there's another school of thought holding that "once you involve yourself in the situation" — specifically in the lives of the people in an experiment — "you are morally entangled with it. You now have obligations that you never would have had in the first place."
One of those obligations, she says, is to do what's best for the individuals participating in the experiment. But in a randomized controlled trial, researchers often end up sacrificing an individual participant's interest for the greater good of the experiment's accuracy.
For a stark illustration, Hellman points to a controversial experiment in the late 1990s to test a method of reducing transmission of HIV from pregnant woman to their unborn children. At the time the state-of-the-art method — a prolonged course of treatment with the anti-retroviral zidovudine (AZT) — was considered too expensive and impractical to offer in poor countries. But there was evidence that a much shorter, and therefore more affordable, course of treatment with AZT, while less effective than the longer course, could also cut down on mother-to-child transmission to some degree.
There was no doubt that giving pregnant women the shorter course was better than doing nothing. But it wasn't clear by how much. And for governments of countries where HIV/AIDS was running rampant this was a pressing question. If the short course made a substantial difference it might be worth spending money to deliver it to many women. If not, it probably didn't make sense to divert scarce resources from other life-saving health spending in the budget — buying malaria drugs, for example.
So researchers, in many cases funded by the U.S. government, set up randomized controlled trials in nearly a dozen low income countries — including Kenya — in which HIV-positive pregnant women coming into health clinics were randomly assigned to two groups: the first got the short-course of AZT treatment, the second got a useless sugar pill.
From a scientific standpoint this set-up was the best way to determine just how much better the short course was than doing nothing. And arguably millions of lives depended on knowing that answer with accuracy: The more precise the findings, the more effectively governments could deploy their health spending.
The problem, says Hellman, is that for individual participants there was no doubt that getting the treatment was better than getting the sugar pill. "It was raising a classic conflict between your obligation to the people who could benefit [in society] versus your obligation to the person who is sitting right before you," she says. (Ultimately the experiments were halted once evidence emerged that the short course was indeed quite effective.)
When it comes to GiveDirectly's experiment, Hellman points out some instructive differences: The stakes are lower when you are giving people money rather than a potentially life-saving drug. Also, says Hellman, in a case where people are actively coming to your health clinic, "you're interacting in a much more direct way."
GiveDirectly's Niehaus notes another distinction: The reason that GiveDirectly is not giving the monthly payouts to the expected 100 or so villages in the control group is not because it's trying to make the experiment as accurate as possible. It's because it doesn't have the money.
Seen in this light the experiment is a sort of side-benefit to the main objective — a way of making a virtue out of the vice that not everybody can be covered. Given that Niehaus can't provide every single poor person in Kenya with the payouts, he argues, isn't the most ethical course to at least make the effort to measure how much of a difference cash aid makes for those getting it compared to those who do not? After all, if researchers find that cash aid has a big impact that kind of evidence "increases the chances that some of the other villagers might get a decent income in the future too" if governments and other charities are inspired to give out more cash aid.
What's more, says Niehaus, GiveDirectly's use of a lottery to decide which villages get the payouts isn't just ensuring that the data from its randomized controlled trial is as accurate as possible, it's distributing the cash aid in the fairest possible way.
"The reality is almost any anti-poverty program doesn't reach all the people that you'd want it to. And the question then is what is the most ethical way of determining who gets it and who doesn't," he says.
"So on those grounds I actually think that experiments are better because at least everybody has the same chance up front of being part of it as opposed to it being an arbitrary bureaucratic decision or potentially subject to politics."
And cash aid has an added ethical advantage, he says. Any time aid groups intervene to offer help in a poor country they are in some sense playing God with people's lives, he believes. But with cash aid, instead of making the decision of what form the intervention should take on behalf of poor people, you are at least allowing them to make their own decisions.