VC Fund Performance - Some History

I’ve had my head down in venture fund performance data for the past week. I love data. And so I thought I’d share some of it with you.

This is Venture Economics data. It shows money out over money in; distributions over paid in capital in the vernacular of venture fund performance wonks. I decided to stop at vintage year 1997, because I only want to focus on fully distributed funds and most venture funds have a 10 year life. I recognize that the data gets ugly a few years later but I don’t want to focus on that just yet. I plan on doing a series of posts on this topic.

Vc_fund_payouts

There are several things that are interesting to me in this chart. First, like all asset classes, the early years of VC were the best. That’s when the market was inefficient and there was more demand for venture than supply. By the early 80s, the market stabilized and so did fund performance. For most of the 80s and 90s, the best funds (top quartile) paid out between 2x and 3x the paid in capital. The median fund paid out between 1x and 2x. The best funds really started to outperform significantly in the 90s when they started paying out 3x and up to 5x in the 1996 vintage year (the best year for venture since the very early days).

But the most amazing thing is the numbers for the top fund of the vintage. You can really hit it out of the park in the venture capital business. The very best funds of every vintage generally pay out at least 5x and sometimes more than 10x.

Like all asset classes, venture is all about manager selection and timing.

Next up, some data on the bad years (1999 to 2003).

#VC & Technology

Comments (Archived):

  1. rob

    Nice Data. I think it would make more sense if you looked at it on a log scale.

    1. fredwilson

      Interesting. Why do you think so?

  2. a former VC

    Fred,This should be a very interesting series of posts since data is like scripture – one can quote it to support the desired conclusion. A really telling thing would be to understand whether the same firm is able to be the top firm (or near the top) in a vintage more than once (kind of like the Yankees winning 26 World Series). Is there any way to predict success or are we just chimpanzees pounding away on typewriters hoping to produce Shakespeare?

    1. Giordano

      Or, like we say in Italy, data is like your balls: you can stretch it in any direction you want. Erm….

    1. fredwilson

      My tumblog is messed up. I am aware of it and working on fixing itFred

      1. joshwa

        thanks, fred… sorry to be bugging you about it!

  3. VC Data Junky

    Fred, enjoy the blog, long time reader, first time commenter. Had to jump on a top near to my heart …To the “former VC” commenter’s point there is lots of information about serial persistance – the general observation being that it is particularly valid in the VC world (unlike the mutual fund world). Paul Kedrosky has written a bunch about the topic (http://paul.kedrosky.com/ar… which you can see in this post or just search the phrase on his blog. Research by a-shall-remain-unamed-large-consulting-firm-that-everyone-knows showed that a top quartile previous fund was 45% likely to produce a top quartile next fund (and 73% likely to produce a fund above the median).Fred, an interesting discussion on these numbers is the constant question of how biased are the reports? If you look in the Thomson data at reporting funds in a given vintage vs numbers of funds purported to be raised it has been declining over time and for many key years was not much higher than 35-40% (last time I checked for funds from 1980 to 1997 the average was 37% of funds raised, by number not dollar, were in the returns data). So, does this mean the sample is biased or, at minimum, not truely representative? Some argue the data is on the high side as only “good” funds would report, but others, including Thomson, argue the data is a representative sample as they get data from LPs as well as funds. Interesting question.You should also look at pooled average vs the top quartile and median over time. I believe it makes clear the arguement that top funds have the most dollars.

    1. just.a.guy

      I have also looked at a lot of venture returns data and there are biases all over the place. Mainly sample selection bias through self-reporting. For example, if you look at venture returns multiple data on a by-deal basis, they’re hard to square with the aggregate returns data… because while the IPOs are clearly reported, and large M&A deals are as well, smaller M&A deals or ones that wouldn’t inspire pride or awe are simply not reported.It’s also worth noting that the 2X line that the median funds flirt with indicates about a 14% annualized return given a few reasonable assumptions (5 yr holding period for avg. investment, …). That rhymes with what Thomson would say is the 20-year average for venture returns… 15.7%. That is only 3-4% better than the NASDAQ or S&P over the long haul (12.5% and 11.7%). I would argue that given the variability of venture returns by fund and vintage, the LPs who invest in venture capital aren’t really being very well compensated for the amount of risk they’re taking.Finally, the more interesting picture than cash-on-cash return that Fred is presenting, or IRRs over time which you often see… is *dollar weighted* returns. That is, for the average dollar that has been invested in venture capital, what has been the return (both absolute and in annualized percentage terms).What you will find if you scour the available data and look at it through the dollar-weighted lens is that of all venture capital dollars invested from about 1990 – 2003 or so… that for every $1 that has gone into the asset class from LPs, just 60 cents has come back out. This is a surprising figure until you realize that somewhere around half of all venture capital ever raised was raised from 1998-2001. It is going to be tough for the NVCA to spin performance numbers once 1997 and 1998 drop out of the 10-year window… my guess is they’ll go back to the 20-year number to show how great a business this is.Long story short, the business is a tough one, has been nasty for a while, and unless you are Sequoia or one of a handful of other remarkable performers, it’s tough sledding. At a lot of firms, there are many guys who have NEVER seen a carried interest distribution… and the whole thing has turned from a value creation and carried interest bet into a 2% management fee income stream. Ugly.

      1. fredwilson

        i want to get to that ugly period. i don’t buy the argument that you can’t make money in venture though. i spent time yesterday with an LP who has a 20 year portfolio and their average cash on cash return is close to 2x. not stellar and probably in the range of the 15.7% IRR that you mention. but they are happy with the asset class and are eager to increase their exposure to it.

        1. Steven Kane

          fred, i’m confused — a 20 year portfolio that produces a 2X is only genertiung an IRR of 3.6% (not “range of 15.7%”). Or am I missing something?

          1. fredwilson

            The average duration of cash flows in a vc fund is less than five yearsFred

    2. fredwilson

      i will post the same numbers from Cambridge Associates who simply publish the data from all the funds that their clients are investors in. that’s biased too, because Cambridge theoretically gets their clients in the better funds. but it doesn’t suffer from self reporting bias.

  4. bsiscovick

    It would be very interesting to see how VC annualized returns compare to other alternative asset classes? Where are investors getting the best bang for their buck?I know plenty of comparisons exist, has anyone seen some interesting ones to share with the group?

    1. VC Data Junky

      I have some quick comparisons within private equity, don’t have real estate, etc. numbers handy.10/20 year returns by categoryseed stage VC 2 / 11 %early stage VC 39 / 22 %later stage VC 9 / 14 %all buyouts 8 / 13 % (outlier is small buyouts at 4/24)mezz 6 / 8 %Horizon returns at even 10 years are bad for VC, right now the big bubble bump is priced into the returns. The 2009 10-year horizon will begin a long string of pain as high-priced investments (and marked up public stock) are brought in at those inflated values.Be curious on real estate, especially given the current US status vs other places like Canada where it keeps on going (at least so far).

  5. Ethan Herdrick

    The big jump of the vintages’ top funds’ returns (purple line) could just be due to an increase in the number of funds. If there are more data points, the outliers should become more extreme. But was there really a boom in number of funds in 1992 and 1993? Seems a few years early.

    1. VC Data Junky

      Actually, the number of funds reporting is in big buckets that differ from what you might think: pre ’83 roughly 20 funds per year, 83-89 roughly 50 per year, 90-92 roughly 20 per year, 93-96 roughly 40 per year, with 97 being the beginning of the explosion. The only years with vastly larger reporting numbers are 99 and 2000 with over 100 funds per year reporting.

      1. Ethan Herdrick

        OK, thanks for the info. So that big spike in the ‘Top Fund’ line around 1992-1993 is due to other reasons.

        1. VC Data Junky

          I think it was simply the bubble. A 92-93 fund would have been making initial commitments as late as 95-96, so nice mature companies took very healthy exits. I know the fund I’m with, our 93 fund is our highest returner.Another factor is we are looking at TVPI (more or less same as DPI for funds > 10 years old), not IRR. The 95-96 funds had much higher IRRs, but lower absolute $ returns (what VCs care about, we don’t get paid of IRR).