Wouldn’t it be nice if taxpayers could somehow get a refund for government programs that didn’t work? Instead, the opposite tends to happen. Programs that fail to make a difference—like many of those that train workers for new jobs—endure indefinitely. Often, policy makers don’t even know which work and which don’t, because rigorous evaluation is rare in government. And competition, which punishes laggards in the private sector, is typically absent in the public sector. But there is some good news on this front. The idea goes by one of two names: pay for success bonds or social impact bonds.