“Cambridge Associates, an advisor to institutions that invest in venture capital, says that only about 20 firms – or about 3 percent of the universe of venture capital firms – generate 95 percent of the industry’s returns, and the composition of the top 3 percent doesn’t change very much over time.”

Anyone who has looked at Cambridge data over time can see that the distribution of VC returns is a *highly* persistent power law, as Andy notes.

Felix Salmon presents the Kauffman data here which shows the same power law.

The existence of a persistent power law is a signature of path dependence.  Path dependence is often driven by preferential attachment. Path dependence is everywhere in economics.  As Douglass said in his Nobel lecture:

“… path dependence [explains]”one of the remarkable regularities of history. Why do economies once on a path of growth or stagnation tend to persist?”

Specifically regarding venture capital and path dependence, being around very smart people with sound judgment who work hard and like to network is self-reinforcing which produces the power law distribution.   The power law distribution represents Matthew effects happening in multiple dimensions (skills, networks, judgment, deal flow, etc.).  The more skillful, wise and successful you get the more skillful, wise and successful you get [repeat with nonlinear impacts].

The people at Santé Fe Institute believe that this is the best paper that explains why the distribution of VCs returns follows a power law. I have issues with the paper since there are other factors at work, but it is directionally correct on the issue of networks and network effects:

“..better-networked VC firms experience significantly better fund performance, as measured by the proportion of investments that are successfully exited through an IPO or a sale to another company.”

Skill plays a huge role (some of which is an outcome of some people winning what Warren Buffet calls “the genetic lottery”) but so does luck. On skill vs. luck take a look at the fantastic work of Michael Mauboussin, especially his wonderful new book The Success Equation.  Also, Felix Salmon on the secret to success in the arts.

Luck and skill are tightly linked.  If you get lucky early in life, you get to hang around smarter people, which makes you smarter and that process self-reinforces.  If you can get connected to great networks of people those network connections similarly self-reinforces.  As just one example, it is a powerful determinant of success in VC to be able to invest only in” referred deals” sent in by smart people with sound judgment.  The highest quality startups with the best talent are attracted to past success of a VC and that self-reinforces success.  Importantly,  exposure to smarter/better startups over time actually makes VCs smarter and have better pattern recognition skills and judgment.

Why do LPs still invest in “marginal VCs” as shown by the Cambridge data given the persistence of the power law in VC returns that Andy notes above?  Marginal VC funds are raised from LPs because pension funds and universities have fake ≈ 8% return assumptions that ZIRP makes radically absurd.   The marginal LPs need some justification that they will deliver magical returns via the VC asset class even though the Cambridge data says it won’t happen unless they get into VCs in the top 5%.  The LPs as a result invest in marginal VCs.  In other words,i t is a way for the investment committee of the pension or endowment to defer the pain and make the huge underfunding problem the job of someone else in the future.

Persistently successful VCs raise relatively larger funds, but have found through a discovery process that fund size is inversely correlated with performance if the size of a given fund exceeds a certain threshold. Smart VCs know that this threshold on fund size changes over time as economics conditions change.  That small funds under perform relative to larger funds is easily explainable by marginal VCs not being able to raise as much money since their historical returns are negative.   The supply of LPs who are desperate for support of their ≈ 8% assumed rates of return and will invest in marginal bottom two quartile VCs is limited.

Marginal VCs in the lower two quartiles will disappear more quickly when pension funds are forced to face reality about their return assumptions.   VCs in the second quartile will do better financially when that happens.  When LPs stop funding marginal VCs is unclear since psychological denial is a persistently powerful force in the course of human events.