Sunday, January 13, 2013

Why Entrepreneurs Fail: On Average Correct, but Overconfident most of the Time

Autobiographies by successful entrepreneurs often depict their business success as being "against all odds," accomplished only through extraordinary effort and sometimes also luck at critical junctures. They are right. Successfully starting an entrepreneurial venture is against the odds, as least as far as we can tell from the statistics that are available. Although the numbers differ by nation, the 50-5 rule that more than 50% of all new ventures are gone in 5 years is pretty good (perhaps a bit optimistic). A venture that closes within its first five years has probably lost money. On the upside, the ventures that make it past the 5-year mark are highly likely to survive the next five year.

These statistics have made people wonder whether entrepreneurs are overly confident, given the poor odds of success. In a new article in Organization Science, Robin Hogarth and Natalia Karelaia show that entrepreneurs are indeed overly confident, but this is – or could be – a result of them being right on average. This statement is not as paradoxical as it seems. The point is that even when our judgments are right on average, as they often are, they are not exactly right.  There is some error, plus or minus. I may assess my chances of success at some task as too high at one time, and too low at another, but in the long run it averages out. You and I make look at the same task at the same time and make too high and too low judgments of success, but on average we may be right.  (We are not always right on average, but for the sake of argument, let's assume we are.)

Now suppose the task in question was to start a new venture. You and I would only do this if we thought our chances of success were good. On average we are right. But that means that if we are both looking at the same opportunity at once, the one who is more confident of success will move. If I am looking at the same opportunity more than once, I will move when I am more confident of success. Suppose the value of the venture is exactly zero, or even negative. Because we only enter when we are confident of success, those who become entrepreneurs will on average be overconfident of success even if all potential entrepreneurs are – all the time – on average correct.

It is a neat model that helps us understand overconfidence in entrepreneurs. It also happens to be a rediscovery. In a 1984 article in Administrative Science Quarterly, J. Richard Harrison and James G. March made a model of decision makers that are on average correct about the value of different alternatives before making a choice. They take the highest-value alternative. as we all do. But if there is some error in their judgments, even if the error is unbiased, the act of choosing the alternative with the highest value will usually make them disappointed in their choice, because the alternative that seems to have highest value is also the one most likely to be overestimated. This is the same insight as Hogarth and Karelaia, except it is more general: instead of choosing between doing nothing and entering a business, the Harrison and March model is a choice between a number of different alternatives.

Rediscovery or not, it is important to be aware of this trap. The way we make choices sets us up for disappointments, even when we are on average right. So entrepreneurs can be on average right, but also too confident of entering new businesses. And scientists can be on average right, but also too confident of having new findings. 


  1. Interesting post and interesting blog overall Henrich -- thanks for sharing this. One of the questions I'd have is to what extent do "unobservables" (for the researcher) explain success, as opposed to pure luck? Our data sets typically include a limited set of variables, and entrepreneurs have a larger set of variables at their disposal. It may be possible that we observe such high rates of failure conditional on what social scientists observe -- but does that mean that nobody can predict success or failure?

    Of course your mechanism, as well as pure luck or failure, are part of the explanation because otherwise we would see larger success rates. If entrepreneurs had access to all of the potential information on their venture's success, failure rates would be attributed to a lack of skill rather than a lack of information.

  2. Amine, that is a nice point. I think that "unobservables" are important for the researcher and for the entrepreneur. We can try to observe more of the unobservables and as a result fail less. Harrison and March would make that recommendation because signal-to-noise ratio is a part of their model. We teach entrepreneurship that way, because exploring better before you commit is a key lesson. There will still be errors in the evaluation because judging the success of an entrepreneurial venture is genuinely hard, and these models tell us not to be too quick to blame failure on inability.