That axiom has loomed large over international development policy for decades. There has been a great deal of aid focused on teaching a man to fish — that is, providing people with skills instead of simple assistance — to lift up the global poor. From 2002 to 2012, the World Bank invested $9 billion in skills training programs.

Skills training programs take a lot of forms, but there are generally two kinds: programs aimed at individuals, which try to teach them everything they’ll need to take higher-paying local jobs, and programs aimed at business owners and prospective business owners, which try to teach them skills to run a business more efficiently and expand their operations.

Their objectives are laudable, but there’s just one problem: They largely don’t work. Participation rates in the programs aren’t very high. People who do participate often drop out, if the program lasts more than a few days, and unsurprisingly, it’s hard to teach important results in that time. For that matter, participants might be right to ignore the program or drop out, as research suggests that the programs don’t reliably increase income.

This isn’t to say every skills training program is ineffective. But even the programs that do show results often don’t stand up to cost-benefit analysis: The results they get are worse than if they just gave people the money that is spent on training them. That said, recent research has found cost-effective results for programs that take a combined approach: training and mentoring, plus direct grants of assets. Those programs, more than just pure skill-training approaches, look to be worth further study and investment going forward.

We’ve tried lots of skills training programs. Most of them don’t work.

Skills-focused programs have enjoyed the support of governments, private foundations, and individual donors, and they’ve been attempted all around the world for years. The World Bank has long been one of the most enthusiastic proponents of skills training programs.

Over the past decade, though, as more and more research has come in with less than promising results, the World Bank has started to moderate its enthusiasm. “Less than a third of training programs have positive results for earnings and employment and even those that are successful are costly, with returns that rarely justify the investment,” its website admits today. It’s sponsoring many skills training projects all the same, and many of them have grown dramatically in the past decade.

One popular focus for skills training programs is business training. The people who run small businesses in developing countries typically don’t follow “best practices” as they’re understood in the developed world, and program after program has set out to change that.

One of the most widely implemented business training programs is the International Labor Organization’s Start and Improve Your Business (SIYB) program. Since it launched in 1977, the program has trained more than 15 million participants, 10 million of those since 2005. The program offers both vocational training (teaching people how to do in-demand jobs) and business management training: teaching accounting, management best practices, financial literacy, and how to start a business. SIYB is the largest such program, but many programs run on a similar model.

A 2014 review of the research into business training interventions, including SIYB, found little evidence of an impact on earnings for micro-entrepreneurs (the group that includes most of the world’s poorest). The researchers also found that almost no micro-entrepreneurs hired an additional employee as a result of a business training program.

The program could still create benefits by increasing income, or reducing the risk of business failure, so additional research focused on looking for impacts on those metrics is needed. Most of the existing research found no effects, but the studies were small enough that even substantial effects could have gone undetected. Most businesses did implement some changes to their business processes as a result of their new training, but studies typically didn’t detect a gain in income as a result. (A few of the larger studies detected small gains.)

The results are similarly discouraging for a related class of programs, sometimes called “hand-holding” programs, where each trainee gets one-on-one follow-up assistance with their business projects. For example, in a program in rural Pakistan, recipients of loans and training got follow-up visits in which they could ask questions and discuss the concepts they’d learned in training. Researchers found this had no impact on any of the metrics of interest (income and assets, networking, decision-making, or outlook on life).

In a 2015 paper, Chris Blattman and Laura Ralston summarized the problem this way:

It is hard to find a skills training program that passes a simple cost-benefit test. After repeated studies of technical, vocational, and business skills training programs, most programs do not have positive impacts, especially on men. Those that do are often so expensive that costs far outweigh benefits. And most poor people turn these programs down or drop out.

There are lots of things that could potentially be interfering with the success of these programs. A major one seems to be a lack of assets. Blattman and Ralston found that “the poor seem to be held back by too little capital and an absence of cheap credit”:

When programs give capital to the poor, be it cash, tools or livestock, to small business owners, unemployed youth or ultrapoor women, we tend to see similar results: poor people expand the number and size of their businesses, and increase the profitability of work in their portfolio.

Another problem seems to be that we don’t know how to teach the skills that are actually useful for small businesses, making the best use of local conditions and local opportunities. Training programs don’t seem to be very good at persuading trainees to change their practices. The skills that matter most may be hard to teach.

Yet another problem is that it’s hard to reach business owners who’d benefit from training. The sort of business owner who pays attention to free training programs available in their area, signs up, attends the programs, and implements the changes is probably one who is already unusually motivated to adopt best practices for their business. They might have already adopted the procedural changes that make sense for them, and get little out of training programs. Some studies suggest that the business owners who’d benefit from training are the ones that won’t seek it out.

A combined approach using skills training and cash does better than cash alone

But not every attempt to teach skills has failed. “Graduation” programs are a promising new, multifaceted intervention aimed at the ultra-poor. Graduation programs involve giving people assets (such as livestock, money or equipment) as well as training and mentoring. Unlike microfinance programs, which don’t work very well, graduation programs don’t demand repayment of the initial grants. As Blattman and Ralston observe above, asset transfer programs seem to work better than pure training programs, and graduation programs include both elements.

Initial results seem promising. A 2015 paper found that in a large-scale randomized trial across six countries, a graduation program improved income and savings, food security, and well-being. (In one country, it failed because the chickens that people were given died of disease.) The evidence suggested that the researchers had hit on a skills training program that worked. That left the question of whether it worked any better than just giving the involved people cash.

Two studies published in 2017 suggest that it does. Both of them employed “cash benchmarking” — testing the effects of their program in comparison to a program where they just give the population cash. This lets us determine whether there’s any extra effect from the training, mentoring, and targeted asset grants.

The first study, conducted in South Sudan, found that the gains from the graduation program were more sustained than the gains from directly giving cash. The second study, from Uganda, found substantial gains to consumption and income — while recipients of cash spent it on reducing debt. That reflects reasonably robust evidence that the training and mentoring aspects of the program are actually adding value.

Graduation programs dodge a lot of the problems identified above that make most programs with a training component fail. They give people assets, which means that they’ll get results if those people are primarily limited by lack of assets as well as if they’re primarily limited by lack of skill. They may be able to reach more recipients, since the value of attending the program is more obvious.

That suggests that skills training might be worth pursuing, but not blindly. They seem to only get results when they’re one part of a big picture. Capital transfers — giving people money or tangible physical assets — seem to be an essential ingredient of success. Yet graduation programs in South Sudan and in Uganda outperformed purely giving people money.

To return to the metaphor, you need to make sure someone is healthy enough to fish and has access to fishing supplies, and that they’re being taught to fish by someone with relevant local expertise, and that there aren’t any good reasons they’re not already fishing. But skills training programs that narrowly target the poorest people globally, transfer them assets with no strings attached, and offer them mentoring and assistance may actually produce the sustained gains in prospects we’ve been looking for all along.