NATIONAL BUREAU OF ECONOMIC RESEARCH
NATIONAL BUREAU OF ECONOMIC RESEARCH

Conference on Entrepreneurship and Economic Growth

Supported by the Ewing Marion Kauffman Foundation
Manuel Adelino and David T. Robinson, Organizers
October 14-15, 2016


Learning in Entrepreneurship

Most new ventures fail. So why do people become entrepreneurs? One explanation is that entrepreneurs are excessively over-confident or have other cognitive biases. Another is that they have non-standard preferences, in particular that they value non-pecuniary benefits of entrepreneurship, like being one's own boss or doing social good.

These explanations are to some degree at odds with canonical models of firm dynamics, in which entrepreneurs enter an industry and incumbent managers exit in response to new information about the net present value of the enterprise. The models contain an explicit or implicit learning process that is in general rational and homogenous.

Sabrina T. Howell studies this learning process at a microeconomic level in an effort to shed light on the puzzle of entry into entrepreneurship. She uses novel data from nearly a hundred new-venture competitions to show that negative feedback increases the chances a venture is abandoned. Further, learning in the sense of improvement predicts subsequent financing and employment.


Access the Paper on Which This Presentation Was Based


The overall responsiveness of the entrepreneurs to new information is consistent with a third explanation for the puzzle of low returns: that entrepreneurship is a largely rational process of experimentation. If the decisions to enter and abandon an entrepreneurial venture are viewed as real options, the high variance of the outcome distribution increases the option value, rather than decreases standard expected returns. Further, conditional on entering, the venture's path is often one of experimentation itself. The entrepreneur conducts a series of experiments to see if his idea will work. Venture capital investors typically contract on specific outcomes of these experiments, funding the firm in stages depending on whether they are successful. An emerging literature has developed this "entrepreneurship as experimentation" view.

The experimentation view requires founders to value and act on their abandonment option. In particular, they should be very responsive to new information. Finding that entrepreneurs do not readily learn from feedback would support a conclusion that they are excessively over-confident, risk-loving, or garner large non-pecuniary benefits.

Howell uses novel data on 4,328 new ventures participating in 96 competitions in 17 states between 1999 and 2016. The competitions offer a window into the earliest stages of entrepreneurship. They are one- or two-day programs in which founders present their technologies and business models to a panel of judges. The presenters are either deciding whether to pursue a new venture idea or have recently founded a venture. The judges participate in order to source deals, clients, or job opportunities, or to "give back" to the entrepreneurial ecosystem. The most comprehensive early stage financing database for recently founded startups indicates that competitions are an important piece of the startup ecosystem. The CB Insights database contains 15,850 firms that got their first VC investment between 2009 and 2016, of which 2,298 received cash prizes from a new venture competition or competitive accelerator.

To study learning in competitions, the key ingredient is the signal that ventures receive about their future prospects from aggregated rankings by judges. Therefore, the first step in the analysis is to establish that the competitions generate valuable, informative signals. If the judges cannot predict success, rational founders have nothing to "learn" from their feedback.

In a regression discontinuity design, Howell shows that, conditional on win status, rank robustly predicts measures of success like subsequent financing, employment, and survival. The figure shows this visually for the financing outcome.

Having established that rank is an informative signal, Howell examines learning in the sense of a founder responding to especially negative feedback by abandoning his venture. Within the sample of losers, she estimates the effect of a very low rank with knowledge of that rank, relative to a very low rank without such knowledge, and finds that receiving this negative, structured feedback increases by about 15 percent the probability that the venture is ultimately abandoned. This provides initial evidence in favor of the experimentation view.

A second measure of learning is in the sense of improvement across rounds within a competition, much as an educator might measure student learning as improvement in test scores over a semester. Learning across rounds strongly predicts subsequent financing, employment, and survival. This suggests that successful entrepreneurs adapt to new signals, also consistent with the experimentation view.

Ventures could improve across rounds by changing the idea or by changing how they sell the idea. To explore the learning mechanism, Howell uses criteria scores. These are aggregated to form the judge ranks used elsewhere in her analysis. Examples of criteria are the technology/product and the attractiveness of the target market. The strongest predictor of success is a higher team (leadership quality) rank. A higher technology/product rank also predicts success. Learning, however, is only relevant to venture outcomes for the financial and presentation criteria. Adjustments to the financing plan and the pitch are most useful, therefore, rather than dramatic pivots to a new idea.

Heterogeneity may help explain some of the evidence contradicting rational models of entry and exit. First, clean energy and "social impact" entrepreneurs, who likely derive large non-pecuniary benefits, tend to learn less. Second, there is evidence that elite founders are overconfident. Founders who received any degree from the top entrepreneurship universities or a top ten MBA learn less on average, and the former are much less likely to abandon their ventures in the face of negative feedback. This does not seem to reflect higher-quality ventures learning less; for example, ventures with previous financing learn more. As the effect of winning on financing is larger for elite school founders, there is no obvious rational reason for them to learn less.

 
Publications
Activities
Meetings
NBER Videos
Themes
Data
People
About

National Bureau of Economic Research, 1050 Massachusetts Ave., Cambridge, MA 02138; 617-868-3900; email: info@nber.org

Contact Us