Failure is acceptable

The Shell Foundation has recently published 'Aid industry reform and the role of enterprise'. (Disclosure: I once worked for Shell.) The report - really an extended op-ed - offers an interesting contrast between what an entrepreneur has to do to get a loan, and what the aid industry needs to do.

There's now a pretty strong consensus that official development assistance not only failed to boost economic growth in poor countries, but on the whole appears to have reduced it.

Really? The report neglects to mention any dissenting voices (with one notable exception) but that's not the same as a consensus. The real complaint should be not that aid doesn't work, but that the aid industry hasn't tried hard enough to show which aid works and which aid doesn't.

But now here's an interesting point, about World Bank projects in Africa in the 1990s:

Between 65 and 70% of these projects failed...what board would allow the existing management to continue running the show when 65% of the projects they were responsible for were judged to have failed!

100% success rates are usually a sign that someone is using his own, convenient definition of success. Some of the corporate governance scandals of recent years have resulted from obsessively hiding failures.

In fact, 100% success is not really desirable. In the United States itself (leave aside Africa), about 60% of all new businesses fail to make a profit. Venture capital firms would expect to fund a large majority of failures, yet still achieve spectacular returns.

The truth is that failure is part of dynamic economic growth. The market does a great job of weeding out failures and rewarding successes, quickly. Large public and private sector institutions, including the World Bank, naturally struggle to do that. We shouldn't be looking at the percentage of projects that fail: what matters is that we learn from failures and don't allow them to bleed away cash, while building quickly on successes. Michael Klein and I have described one way in which this might be done, even in a bureaucratic setting.