Adjusting in this way for the selection bias of firms that go bankrupt, the mean return on VC investments is 57 percent per year, still very large but less dramatic that the 700 percent mean before correcting for selection bias.
Venture capital (VC) investments carry more risk than most investments in the broad public market and their returns are much more modest than commonly thought, according to a new paper by NBER Research Associate John Cochrane. He concludes that VC investments are not dramatically different from publicly listed small growth stocks.
Estimates of the returns to VC investments can be highly misleading because they typically reflect only those firms that have initial public offerings or are acquired by another company. Private companies are more likely to go public when they have achieved a good return. Those that do not achieve a good return are more likely to stay private or go bankrupt. Therefore, ignoring those companies that stay private only counts the winners; it induces an upward bias in the measure of expected returns for potential investors.
In The Risk and Return of Venture Capital (NBER Working Paper No. 8066), Cochrane includes those companies that stay private -- the losers as well as the winners-- so as to more accurately estimate the returns on VC investments. His analysis is based on 17,000 financing rounds in 8,000 companies, representing $114 billion of VC dollars, between 1987 and 2000.
Before controlling for the selection problem, Cochrane finds very large average returns among companies that go public or are acquired. The average return is almost 700 percent. Returns in this sample are also very volatile, with a standard deviation of 3,300 percent. Underlying these averages, however, there are a few companies with astounding returns, and a much larger fraction with modest returns. About 15 percent of companies that go public/are acquired achieve returns greater than 1,000 percent; yet 35 percent of the companies achieve returns below 35 percent; and 15 percent of the companies deliver negative returns. The most probable return is only about 25 percent.
Cochrane then estimates how the probability of going public or being acquired increases as the value of the firm increases and the point at which companies go bankrupt, in order to estimate the overall underlying average return, volatility, and sensitivity to movements in the stock market (beta) of VC investments.
Adjusting in this way for the selection bias of firms that go bankrupt, the mean return on VC investments is 57 percent per year, still very large but less dramatic that the 700 percent mean before correcting for selection bias. VC investments are still extremely volatile, with an annual standard deviation of about 100 percent. This is much greater than the roughly 10 percent standard deviation for the S&P-500 in the same period, but similar to the volatility of small publicly traded NASDAQ stocks. The "beta" is close to one, indicating that VC investment returns move up and down one-for-one with the stock market as a whole.
The high volatility is necessary to explain the occasional spectacular successes. Only very volatile investments can occasionally attain 1,000 percent returns. The high average return is explained by the high volatility. If an investment has an even chance of doubling or halving in value, it has a 25 percent mean return. For each dollar invested, you could make a dollar, or lose 50 cents. The larger the volatility, the greater this effect. More directly, VC investments derive their large average returns from a very small chance of a huge payoff. Therefore, enjoying this average return without enormous risk requires a very diversified portfolio. The market also went up substantially in this period, so a 57 percent return would not be that surprising with a beta of 2 to 3; the estimated beta of one implies that investors received an extra reward for holding the poorly diversifiable risks of venture capital in this period.
Cochrane finds that although typical health/biotech investments did better than typical information technology (IT) investments, the higher volatility for IT gives it a larger chance for occasional spectacular successes and thus a larger arithmetic mean return.
Cochrane also finds that second, third, and fourth rounds of VC financing are successively less risky than the first, as one might have guessed. They have progressively lower volatility and therefore lower mean returns. The betas of successive rounds also decline dramatically, from near one for the first round to near zero for fourth rounds, reflecting lower risk in the form of lower sensitivity to market conditions.
In closing, Cochrane cautions that his data sample ends in June of 2000, and most of the positive returns come from the late 1990s. As our sample extends to the NASDAQ decline and the wave of failed venture capital projects, the mean return estimates may decline, and the beta estimates may rise.
-- Andrew Balls