Don’t teach a man to fish. Just give him the goddamn fish.

I write a lot about the benefits of fighting poverty by giving poor people cash. It’s supported by good evidence, it’s relatively easy, it respects the decisions of poor people as to how to spend it, and it avoids the central planning challenges of some other anti-poverty policies.

The most common response is something along the lines of “give a man a fish and you feed him for a day; teach a man to fish and you feed him for a lifetime.”

I’ve always hated that saying, not least because a healthy diet requires eating more than just fish, but it’s actually sort of helpful in this context. The main reason I prefer just giving someone a fish is that we really don’t know, especially in international development, how to teach people to fish.

In this context, “teaching to fish” means something like “getting a poor country on a sustainable long-term growth path.” There are plenty of theories of how to do this, but it’s hard to think of an aid program that actually did it. For years, the default approach of the World Bank and the International Monetary Fund was an approach that came to be known as the “Washington Consensus,” and emphasized trade liberalization, deregulation, fiscal austerity, privatization, floating exchange rates, and other free market reforms. There’s little evidence that those policies actually did much to promote growth or poverty reduction.

Critics of the Washington Consensus like Ha-Joon Chang have typically advised poor countries to abandon free trade and protect their “infant industries” as a way to produce export-based growth. But empirical evidence suggests that doesn’t typically work either. Economist Jeffrey Sachs thought he’d figured out how to break poor communities out of poverty traps, and set about creating “Millennium Villages” implementing his ideas. Subsequent evaluations found some positive results, but none suggesting that the intervention set off permanently higher growth rates. Paul Romer, the most respected theorist of economic growth of his generation, thought that setting up “charter cities” that had strong property rights but were free of poor countries’ broken and corrupt institutions could work. But he stopped hawking the idea after attempts to set up cities in Madagascar and Honduras both floundered.

So the move in development economics of late has been away from grand theories of how countries grow and toward smaller, rigorously tested interventions in the provision of health, education, and other social services. But even there, there are hiccups. An intervention that works in Kenya won’t necessarily work in Bangladesh. An intervention tried just among schoolchildren won’t necessarily work with a different population. As this delightful graphic from Stanford economist and AidGrade founder Eva Vivalt shows, there’s just a huge amount of variation in empirical results about the effects of different aid policies:

We can be pretty sure some things — like supplying free anti-malarial bednets and providing deworming pills, or cash — work. But for a lot of important interventions, we just don’t know. And the stuff that really matters (political stability, trustworthy institution, good health infrastructure) isn’t that amenable to randomized evaluation.

We know more about how to tackle poverty within rich countries. Programs to help people move out of disadvantaged neighborhoods with concentrated poverty appear to be quite effective in raising the lifetime incomes of children whose families move. We’re getting better at knowing what kinds of schools work for kids growing up in poverty. But one of the best-supported ways we’ve figured out to help people is giving them cash through programs like the Earned Income Tax Credit. And the one time a proposal to just give every citizen cash every year was implemented in a North American city, the city got healthier, saw more teenagers go to school and graduate, and, with the exception of new moms and teenagers, didn’t work any less.

We shouldn’t stop trying to figure out ways to solve bigger structural problems. But the fact remains that we don’t really know how to teach people to fish. We do, however, know how to give people fish — and we know it leaves them a lot better off as a result.

CORRECTION: An earlier version of this article rated that Jeffrey Sachs “couldn’t get Millennium Villages to work.” Subsequent evaluations have shown some positive results from the Millennium Villages project, and the piece has been updated to reflect that.