Failure Isn’t Enough

April 15, 2011 Editor 0

The notion that innovation and failure go hand-in-hand has become popular of late. It begins with the argument that many important innovations involved taking risks with potentially large costs but, most critically, completely unknown upsides. In his soon-to-be published book, Adapt, Tim Harford recounts the story of Mario Capecchi who leveraged safe NIH projects to fund a highly speculative attempt to make specific changes to the DNA of mice. Any rational evaluation at the time (1980) would have put that project in the realm of science fiction. Capecchi proved everyone wrong and we all reaped the rewards in terms of a sea change in cancer research.

There is a case to be made that Capecchi is the exception that proves the rule when it comes to failure in innovation. Had Capecchi failed and wasted that money, his career would have been shattered. For fascinating reasons that I’ll leave to Harford to explain, Capecchi may be the type of person who ignored those risks. But the implication is that innovation — particularly radical innovation — does not happen enough and that if we could soften the failure consequences for individual researchers, more risky experiments would take place.

This version of the role of failure is akin to the saying “you have to break a few eggs to make an omelet.” Cultures and systems that do not overly punish failure will generate more people willing to undertake risky experiments. As the economist Nathan Rosenberg wrote some time ago, one of the hallmarks of capitalism is its tolerance of failure (see here: PDF) in a way the Soviet command did not. Alongside this comes a diversity in organizational forms that can select experiments (see the account by Scott Stern). But to be sure, by allowing more failed experiments, the goal is not to see greater funds spent on failed projects but instead to increase the ratio of total value generated from all experiments relative to the total funds and resources expended. That is, allowing more failure should increase the overall rate of return.

The difficulty with this argument is that it does not tell us what mechanism we should use to tolerate more failure. One mechanism involves directly lowering the costs of experiments. This can happen through more liberal grants (something funders like the Howard Hughes Medical Institute or MacArthur Foundation engage in) or by removing impediments (as the Obama administration has proposed for small business entrepreneurship). These mechanisms can encourage more participation and diversity in innovation but, at the same time, it is not clear the marginal projects will be the speculative ones that failure is all about. What is more, to the extent that lowering costs shields individuals from bearing losses, there is concern that while more experiments may occur, the pressure to make them work will be diminished.

An alternative is to target the upside and make sure that the rewards from successful experimentation are higher. While some policies, such as removing the ability of incumbent firms to use market power to harm entrepreneurial entry, do just that, there is only so much reward available. In some situations, organizations have established prizes to motivate experimentation on a dimension. But even there one needs specification of what a “solution” looks like to actually award the prize. The problem Capecchi faced, for example, was that he was innovating in an area that people were unconvinced was a solution for anything.

Interestingly, Harford puts forward a stronger thesis than simply the notion that failure accompanies innovation. He argues that it is a necessary ingredient. That you can’t succeed without having failed first. He points to the U.S. experience in Vietnam and how that eventually translated through personal connections to finding a better strategy in Iraq. The idea is that one innovator’s failure yields important information as to where to experiment next. That is, failure brings with it learning. Specifically, without information on failure, it is easy to believe that continued success is based on skill rather than luck. As game theorists Drew Fudenberg and David Levine showed, your false beliefs, left unchallenged, can be self-confirming.

Of course, having more experiments, and failed experiments at that, will provide a necessary ingredient for that task. But it isn’t sufficient. Someone needs to learn about the failure and then learn from it — something that took too long for Iraq. When it is the same person who fails and tries again, that path is easy. But it’s really very hard to learn from someone else’s experience of failure. Indeed, for all we know, we already have plenty of failure. We just haven’t learned enough from it.

Joshua Gans is an economics professor at Melbourne Business School and a visiting researcher at Microsoft Research. All views here are his own.

Go to Source

Categories: Innovation

Tags: argument, capecchi, failure, Innovation