From the article:
A year and a half ago The Economist wrote about a promising approach to cutting poverty in Bangladesh (“On their bikes”, January 27th 2018). RDRS, a charity, was offering small loans to more than 100,000 poor farmers on the condition that they migrated temporarily to a city for work. Everything seemed to be set fair. Smaller randomised controlled trials had shown that many men could be persuaded to move while the rice crop is growing, when there is not much work to be done at home. Although the migrants found only low-paid jobs, as rickshaw drivers, building labourers and the like, their fortunes had greatly improved. It looked like a true poverty cure.
…
Do-gooding schemes that work brilliantly in trials often fail when they are scaled up, says Justin Sandefur of the Centre for Global Development, a think-tank. Trials are often overseen by determined phd students. When large charities or government officials take over, as they must if a project is to be done at scale, much changes. Rules and regulations multiply; bad behaviour becomes more likely. Big schemes can attract hefty opposition.
One charity in Kenya had shown that hiring teachers on fixed-term contracts improved pupils’ test scores. So the government rolled out the contracts across the country. But a political backlash meant that the contracted teachers were promised trade-union representation, just like ordinary teachers. Not surprisingly, an evaluation by Mr Sandefur and others found that the government’s reform had no effect.