This cash experiment cut child deaths in half. Here’s the catch.

It sometimes seems like the basic income wars will never go away. My first Vox piece on the idea of a government-provided guaranteed income came in the summer of 2014 — a simpler time, the Obama years. I wrote a big feature about it in 2017. Since then, we’ve had Andrew Yang’s presidential run, Covid-era stimulus checks, and massive progress in AI, all of which have made the idea feel more plausible.

We’ve also had some research findings that throw cold water on the concept, at least in the US. Three studies that gave out unrestricted cash to Americans during the pandemic found nulls on all the outcomes they tested: the cash didn’t improve health or self-reported well-being or even, in one study, how well people say they’re doing financially.

Two more recent, even bigger studies have backed that up. The Baby’s First Years Study, which began in 2018 and gave $4,000 a year to low-income American families with young kids for over four years, found no effects on child development outcomes at the four-year mark. No reduction in behavioral problems, no improvements in language ability — nada. Another study run by the group OpenResearch gave out $12,000 a year to families for three years. While it found some positive outcomes, like parents spending more on their kids, mostly it found null effects, too. Participants spent more because their incomes grew, but they also worked less, offsetting the income gain a bit.

But as I wrote back in 2017, “The biggest potential for basic income isn’t in the US but in developing countries.” A big new study from the charity GiveDirectly seems to back that up, finding that cash grants in Kenya not only reduced poverty but actually saved lives.

Even here, though, it’s important to be careful. Covering basic income experiments for more than a decade has taught me that readers love to hear about ways in which cash programs work and are less eager to hear about ways they fall short. That resulted in some big expectations that, in more recent US research, at least, have been dashed.

Some coverage of the GiveDirectly study has only focused on the good news (lives saved) and not limitations, like the fact that it didn’t save lives very cost-effectively. That’s a problem.

The headline finding of the GiveDirectly study from economists Michael Walker, Nick Shankar, Edward Miguel, Dennis Egger, and Grady Killeen is that a randomized experiment providing one-time cash grants of worth about $1,871 each to over 10,500 households in rural Kenya found the cash reduced infant and child mortality. (The actual value was exactly $1,000, but given that prices are lower for most things in Kenya, it could buy what Americans think of as about $1,871 in goods and services.)

The intervention here is larger than most countries’ cash programs. Kenya’s main national cash program grants 2,000 shillings (about $45) a month to its most vulnerable citizens — a tiny fraction of the $1,871 drop this study examined. And the effect size from the $1,871 is truly massive as well: a 48 percent reduction in deaths, mostly shortly after the cash was dispersed in 2015-2017.

A reduction that big from any intervention is eye-popping. From an intervention that isn’t even specifically meant to improve health or reduce deaths, it’s sufficiently impressive that you should be instinctively skeptical. Other, non-experimental studies have found reductions in mortality due to cash programs, but I’ve found only one other controlled experiment that found the same. It’s not a big evidence base.

What is the mechanism by which this might have happened? The study can’t say definitively, but it offers some clues. The cash drove a big increase in the share of mothers who delivered babies in hospitals as opposed to at home. If hospital-based delivery is safer than home delivery in these Kenyan villages, that could have caused some reduction in deaths. But it seems very unlikely to cause the entire reduction, especially given that other studies have not found that delivery in a hospital or other health facility reduces the odds of infant death.

But there are other ways the cash could have helped. Mothers bought more food, and a “food security” measure (reflecting how often children skipped meals or went to bed hungry) improved. That could plausibly drive greater survival. Among mothers who received cash, hours worked fell by half in the three months before and three months after childbirth, implying that the transfer functioned as a kind of paid parental leave program (fathers’ work hours didn’t fall significantly). That could, perhaps, reduce maternal stress in ways that reduce infant mortality.

When I asked economists not involved in the study to comment, they generally said it was a well-designed, credible experiment. But no study is perfect. Berk Özler, formerly a top economist at the World Bank now at the University of Otago and Stanford, noted that overall childbirths were 13 percent more common in villages getting cash. That raises the possibility of a selection effect. Perhaps the cash didn’t cause babies who would’ve been born anyway to be healthier but instead induced women likelier to have healthy babies anyway to get pregnant and give birth. For their part, the study’s authors conducted a number of tests and argue that this is unlikely to explain much of the results.

Perhaps the most significant caveat about the study, however, is that, while the effect on infant mortality was large, the cost-effectiveness of the cash program as a lifesaving tool isn’t impressive. The cash program cost $25.75 million, and, per the study results, saved 86 children’s lives, for a cost per life saved of $299,418. That’s very good by rich country standards. For comparison, Medicaid takes $5.4 million to save a life. But it’s not competitive with the most cost-effective ways to save lives in countries like Kenya. GiveWell estimates that Vitamin A supplementation can save a life for about $3,500, while malaria prevention meds can do the same for about $4,500.

GiveWell helped fund the cash/infant mortality study, in part seeking to improve its estimates of the cost-effectiveness of cash grants. Its conclusion was that the finding doesn’t change its rankings of charities too much and that most of the benefit from cash transfers comes from reducing poverty (which the Kenya experiment certainly did) rather than saving lives.

One really bad takeaway you could have from this study is “cash grants are all you need to save lives in the Global South.” The study authors and GiveDirectly itself are clear that cash needs other health infrastructure to work. “People who were further away from health facilities, the improvement wasn’t as great,” Dr. Miriam Laker-Oketta, a physician based at Uganda’s Makerere University and an advisor at GiveDirectly, told me in an interview. “What we need as a complementary intervention is infrastructure” like health facilities.

Another really bad takeaway, more from the US studies than this one, would be that cash has no health effects at all, even in the Global South. Among very poor, malnourished people who cannot afford regular medical care, there are strong intuitive reasons to think cash does something for their health. While lower monthly payments, like Kenya’s $45 a month, almost certainly won’t cause infant mortality to fall by half, they might help people’s health on the margin. See, for example, this study, which found improvements in height-for-age among children born to young cash transfer recipients in Malawi. Of course, whether or not cash is the best or most cost-effective way to promote child health is a different question — and one where cash does worse.

Writing about cash programs is a funny thing. For most topics, readers exhibit a negativity bias: They click more on bad news. But I’ve found that when I write about cash, big flashy headlines about all the good it does get lots of hits, while nuanced pieces about benefits and limitations don’t. (Bad omen for this piece, I guess.)

That’s a bad set of incentives, and I’m trying my best to resist them. People are working hard to understand what cash can and can’t do. It does them a disservice to only share the good things they learn.