One of the perennial issues facing the venture industry is whether it's a closed club (i.e. the same VCs generate the top returns through preferential access to the best investments) or whether it is a truly competitive industry (i.e. a new entrant has the same chance of success as an established firm). This issue is often called serial persistence - namely, do the same funds remain at the top of pile fund after fund after fund.
"Common wisdom" (please note the quotation marks) dictates that serial persistence is very strong and this results in LPs using prior performance as the main indicator of future performance. Given the impact on capital allocation practices and the dramatic drop off in performance between top quartile performance and the rest, you would imagine that VCs, LPs and academics would be concentrating time and energy to answer this question. However, as far as I can tell, when I discuss this issue with colleagues in the industry few are aware of the evidence to support or disprove this "common wisdom", but rely instead on a intuitive belief. It's not all "head in the sand" behavior though. There have been some germane posts on this issue notably by Bill Burnham and Paul Kedrosky which inspired me to hunt for evidence on both sides of the argument
This hunt for quantitative analysis and reasoned logic led me to two studies which came to diametrically opposed conclusions. Steve Bird, a GP at Focus Ventures, is the author of a recent paper which claims very strong serial persistence ("churn of the Top 50 firms over the course of 15 years was only 16%"). Steve Kaplan and Antoinette Schoar of the University of Chicago authored a paper which claims only moderate serial persistence (that the likelihood of a top tercile fund remaining in the top tercile is only 48%, when you would expect it to be 33% if it was truly random). Given the conclusions are so different and thus the implications (Steve Bird recommends that if an LP cannot invest in one of these top 50 firms, then they should avoid the asset class entirely), I decided to dig into these studies to try and understand the divergence. I asked myself three questions?
1. What are the Inputs Being Measured?
Steve Bird used a data set of 8,000 venture financings over a 24 year period. However he then zoomed in on only two time periods - the PC Boom 1983 to 1987 and the Internet Boom from 1997 to 2001.
Steve Kaplan and Antoinette Schoar focused on funds, not financings, as the unit of analysis including those that 1) have been officially liquidated 2) whose returns are unchanged for 6 quarters 3) whose reported unrealized capital is less than 10% of committed capital and 4) who had greater than $5M under management in 1990 dollars resulting in a clean data set of 580 venture funds from 1980 to 2001.
2. What is the Output Being Measured?
Steve Bird chooses "percentage of IPO value created" as his output to be measured, rather than the more traditional LP metrics of IRR. This results in him only counting some of the winners (he ignores M&A value creation) and none of the losers
Steve Kaplan and Antoinette Schoar use IRR and PME (Public Market Equivalent which is a form of IRR adjusted for public market performance over the same period). This results in all the winners and all the losers being included.
3. What's the Author's Starting Point?
Steve Bird is a GP at Focus Ventures, a venture firm specializing in expansion stage capital for IT companies. So he is a participant in the venture industry, is well aware of the common or conventional wisdom and has a vested interest in the analysis. This leads me to believe that Steve could be guilty of subconsciously choosing the date to fit the thesis (nota bene: I do not have first hand knowledge of Steve and have no wish to impune his integrity. I am merely commenting on the well documented universal human condition of accepting data that fits an already held belief and rejecting data that challenges same said belief.)
Steve Kaplan and Antoinette Schoar are academics and not to my knowledge participants in the venture industry. As such they pose questions and let the data lead them to the answers.
So in conclusion, after reviewing the data sets, methodologies and potential for objectivity I place greater weight on Steve Kaplan and Antoinette Schoar's conclusions that while serial persistence exists in the venture industry, it is not nearly as strong as either Steve Bird or "common wisdom" claims.
Let the debate continue.
Comments