Recently, state level freedom of information (FOIA) acts have been used as tools to extract information about venture capital investments and returns from public limiteds such as state pension funds and university endowments. I'll get around to talking about the economic implications of this later, but for now, let's look at some of the returns that have been reported, and how to interpret them. Here's a pretty well organized performance report from the Unversity of Michigan that I'll use as an example. Get that in a second browser window before reading further.

The Fund Type column shows VC (venture capital), PE (private equity), and BUY (buyout). My comments are going to apply to the first only, since it's where I have first-hand experience.

Now look one column right, to IRR. IRR is the internal rate of return of the fund. That's the annualized rate of return that fits the actual capital draw schedule of the fund, and the eventual sequence of pay-outs to the limiteds. When the fund is still in progress, the computation assumes that it's paid out immediately at the current valuation. When I say 'fits' I mean it in a mathematical sense: the IRR formula makes an implicit assumption that the rate of return is constant through the time period being modeled. Now that may be a good assumption if we're talking about CDs or T-bills, but it doesn't fit the reality of a venture fund. Nevertheless, we use it because it provides a consistent benchmark and the math is tractable. If you'd to play around with IRRs, there's a canned formula in Excel, so fire it up and construct your own cash flow series.

Now before we move on, scan up and down that IRR column. Jumps around a bit, eh? That's the picture of a high variance asset class. How does an investor deal with a high variance situation? If they're fools, they try and pick one or two winners. If they are professionals, they average. You'll notice the UM portfolio has at least five different VC funds in every complete year from 1996 onward. They are pros.

Now let's talk about the sources of that variance. We can break it into three components: variance among funds, variance among sectors, and variance of the overall market. If you compare the IRRs among the VC funds of the same year, you'll see there is plenty of variance there. It does look like UM has been kicking out the low performers over time, but as they say "Past performance is no guarantee of future returns." So you'll notice, for instance, that while Matrix V, vintage 1997, was a thing of beauty even in an awesome year, Matrix VI of 2000 is a dog so far.

Unfortunately, there's not enough info here to talk about sector variance, that is, biotech vs. comms vs. IT vs. blends, so we'll pass over it. Shift one column right, over to Benchmark. This is overall market variance data, in the form of the average of IRRs of reporting funds opened in each year. The analogy with wines is direct: some years are amazing, some are best forgot. Over a long period of time, this pattern is roughly cyclic. This time series isn't long enough to show the last serious down cycle, but the effect of the exaggerated Internet up cycle is pretty obvious on vintages 1995-7.

There is also a systematic and predictable variance of the IRRs within each vintage, in the years after the funds' inception. This is the "J-curve effect" cited in the Disclaimer at the end of the table. Simply put, the J-curve means that the value of a venture fund almost always goes down, and its IRRs are negative, in the first few years, before both hopefully climb. Ergo, you really can't compare IRRs of funds in their first 4 or 5 years to those which are more mature. The Disclaimer has a pretty good explanation of why this happens, but here's my version:

- Dogs die before winners run. We all make mistakes in investments. The best managers will recognize those mistakes and kill them off or exit at a loss quickly. The winning investments (in a normal market) take years to achieve uprounds and eventually exit at a profit.

- Management fees do not depend on current return. Management fees are typically 2.5% per annum. That's actually your baseline, not 0%. The capital gains in out years will hopefully obscure the management fees in both the spreadsheet and the limiteds' minds, but not so in the first few years.

- Market cycles can exaggerate or minimize the J effect in any vintage. You might have noticed a down draft in the public markets in the last few years; that's reflecting in private valuations. Sector effects can make this even more dramatic. Notice ComVenture V of 2000: -47% IRR. And remember that's the simple, not compounded, rate. Ouch! Sure enough, it's a fund specialized on communications equipment.

- Remember I mentioned that the IRR computation involved assuming liquidation at the current fund valuation? That valuation is just the sum of your cash with all of the current private company investments. As the Disclaimer says:
There is no industry standard for valuing private companies, and valuations of similar companies may vary from fund to fund based on a subjective measure of value.

There is no clearing auction market for this stuff, unlike publicly traded securities, and you can get up a good argument among VCs regarding what should affect valuations and when changes should be recognized. In the long run, the truth will out, when a willing buyer must be found for the next round or the exit. But in the first few years of a fund, the inherently uncertain nature of the value of an early stage, private company means that the variation among funds is often an artifact of their valuation policies. So it might be that Matrix VI is just more conservative on their valuation policy, not really a canine.

Well that was short and fun, wasn't it? But if you've suffered through this entire post - and worked the exercises - you now know more than you will likely learn from the

*schadenfreude* merchants in the mainstream press.