Correlation’s Fresh Look At Venture Capital Returns

Posted November 18th, 2013 in General Venture Capital, VC-backed Biotech Returns

The broad underperformance of the venture capital asset class during the past decade is widely discussed; there’s a frequently-cited assertion that U.S. venture returns trailed the overall public equity stock market in the 2000s.

Few if any published analyses have actually examined realized returns on a deal-by-deal basis using “apples to apples” cash investments in either venture-backed deals or public indices.  The data mavens at Correlation Ventures have just shed some much needed light on this question, and their analysis at least in part challenges the conventional wisdom.  They created a “theoretical S&P 500” investor who is assumed to have invested in the S&P500 index with the same amount of capital on the same days as the individual investments were made into venture-backed companies, and then sold their index positions on the same day as the venture-backed exit.  This approximates the methodology of “Public Market Equivalent” (PME) analyses.  One big difference is that most PME analyses include the “mark-to-market” paper valuations of unrealized, illiquid private companies; Correlation’s analysis focuses only on realized returns and “exits” – acquisitions, IPOs, and shutdowns.

[Update 11/19/13: To clarify questions about this dataset, here’s more information: From 2003-2012, this dataset captures 7976 realized deals across all venture sectors. This is an exit-year based analysis of individual deals, but includes the cash-in/cash-out invested over the life of that specific deal.  Three types of realization classifications were used: out of business, M&A, or IPO.  The distribution of outcomes: 39% went out of business, another 29% exited for less than their invested capital and hence lost money, and the remaining 32% were positive outcomes returning >1x, according to the Correlation database.  This set of realized deals therefore is one of the largest datasets available, and reflects a representative return distribution of deal outcomes seen across the industry (with loss ratios and winning percentages similar to other analyses).  Importantly, and repeated for purpose of clarity, this is not a “winners” only analysis, but reflects a large proportion of money losing deals (68% of the datapoints, in fact).  Lastly, since unrealized deal valuations do not correlate well with final realized deal outcomes, this analysis reflects an interesting snapshot on deal-level performance in venture.  While not a fund-level analysis, which is the ultimate measure for LP’s, it remains an informative and thought-provoking analysis of one of the largest venture industry datasets.]    

The chart below captures overall U.S. Venture Capital performance (blue line) relative to this theoretical S&P500 investor (dashed line).

Venture Capital vs SP500

The team at Correlation makes two conclusions:

Gross realized multiples have been much higher for U.S. venture investors than for an “apples-to-apples” S&P 500 investor in each of the past few years, and the average realized multiple for the past decade (2003-2012) was 36% higher for U.S. venture than for the S&P 500 investor….  Please note that when we estimate “net” returns (i.e., including fees and/or carry), U.S. venture has still significantly outperformed public equities in recent years.

Average realized multiples for U.S. venture have significantly increased during the past three years. This is consistent with our hypothesis that returns should increase as the significant overcapitalization of the industry that occurred in 1999 and 2000 exits the system.

As a healthcare and specifically biotech investor, I’m particularly interested in those cuts of the above analysis.  Appended below are those two charts.  Just like in the broader venture asset class, the realized returns in biotech and healthcare as a whole outperformed the cash-matched theoretical S&P500 investor.  This isn’t surprising given prior analyses highlighting biotech and healthcare’s above-average performance vs other venture sectors in the 2000s (here, here, here).

BioPharma vs SP500

Healthcare vs SP500

In 2010-2012, other venture sectors have contributed more to outperformance than biotech; this is most likely due to some very large social media and consumer realizations/IPO’s in that period.  Given the current excitement around the biotech IPO window, I would expect biopharma’s numbers to look quite strong when 2013 statistics are included.

Its not clear exactly why these findings appear different than the analysis by Kauffman in 2012 (report here, discussed here); the authors of that report share data showing an underperformance relative to public markets of a large basket of venture funds in their portfolio.  I haven’t explored the underlying methodological differences between this Correlation Ventures’ analysis and the Kauffman report, but a big difference is evident on the surface related to realized vs unrealized returns.  This Correlation analysis focuses on realized exits in the year they exited on a deal-by-deal basis, vs the Kauffman analysis’ focus on venture funds themselves and therefore in a vintage year (birth year) manner, which at least in the past decade must include a large number of unrealized “paper” investments (companies that are still alive).  Realized returns have beaten unrealized returns for exits in the 2000s (see first figure of post here), so this certainly explains some of the more positive Correlation findings.  Importantly, LP’s don’t just care about realized returns – the drag on a fund from unrealized zombie portfolio companies is presumably significant for both of these analyses to be true.  Furthermore, most bubble vintage funds of 1999-2001 have struggled to escape the craters of that period; any exit analysis focused on 2003 and later exits implies by definition that those companies were either ‘born’ after or were successful survivors of that period.

Thanks again to the Correlation Venture’s team for sharing their data with the world.  For about their firm and data-driven approach to venture capital, check them out here.

This entry was posted in General Venture Capital, VC-backed Biotech Returns and tagged , , , , . Bookmark the permalink.
  • Stan Fleming, Forward Ventures

    Bruce, If your numbers are correct it would be good news for the bio-venture community. However, before celebrating, I have some questions.
    The Correlation numbers appear to be counter-intuitive, given the institutional investment community’s down-grade of the sector, particularly bio-venture. The reported S&P returns seem to under-estimate the index’s performance. Looking at the mid-point of 2004 to 2012, I see about a 70% gain for the S&P versus 1.2X on the graph.
    Kauffmann and other LP-type observers typically look at dollar-in-dollar-out returns at the fund level. The drag on bio-venture (and the pharmaceutical industry) has been the cost of losses. I assume that the Correlation data has some way of looking at all investments made in given period and then matching that number against the realized returns, but I am not sure how one does that without creating a success bias. The multiple that this study shows looks more like one that matches dollars returned to individual investments as reported by groups like HBM. How does Correlation count write-offs? Mark-to-market accounting and the tax man often require that VCs maintain moribund companies on the books and not realizes losses long after they are dead for practical purposes. A ten-year reporting period means that less than half of the portfolios studied would qualify as “mature” by bio-pharma standards and as a result could still be carrying large future losses.

  • LifeSciVC

    Stan, thanks for your comments and apologize for the lack of clarity in my post. I’ve updated it above to shed more light on the dataset. Few things:

    1. This is an exit year analysis tracking the cash-in/cash-out for both a theoretical S&P investor and the individual exits, aggregated at the ‘exit’ year level. So the S&P line is made up of all the cash-in and cash-out data points mirroring the deals in the analysis – so your concern about “missing performance” of S&P isn’t valid. 2. This is not a fund-level analysis, which includes unrealized “active” deals; its an exit-year analysis of individual deals and all of their prior analyses. While Fund-level analyses are ultimately the key for LP’s, this analysis sheds like on the distribution of outcomes from over 7900 deals – which is a big dataset – vs the S&P. 3. You mention a common misconception – “the drag on bio-venture (and the pharmaceutical industry) has been the cost of losses”. Relative to other sectors, including information technology, biopharma has historically had a lower capital-adjusted loss ratio. This data has been shared previously on this blog and elsewhere and is a frequently asserted but factually incorrect perception. 4. Regarding the losers and mark-to-market, see the note I added above. 39% of the dataset include out-of-business deals, another 29% include exits at <1x. This is in line with other distribution analyses in venture when the dataset is weighted to technology sector in particular. Lower loss ratios exits in biopharma alone (68% in all venture here, closer to 40% in biopharma in prior analyses we've reported on). As an exit year analysis, companies found to be moribund are moved into the exit or out-of-business category as I've been told by Correlation. 5. Since this is a deal level analysis, its not a reflection on venture vintages. I agree that lots of 2003-2012 vintage venture funds still have unrealized losses/winners in there; but as an exit-year analysis, the 2003 datapoint includes deals that were invested in over the prior decade presumably, etc…
    Hope that's helpful. The reality is LP's ultimately only care about fund-level returns – I get that. And they've not been great, especially as the sector continues to recover from the 1999-2002 period of massive losses. But I think analyses like this one from Correlation are helpful in that they take a huge dataset and shine light on distributions in an interesting way. By ignoring the unrealized valuations (which don't correlate well with eventual outcomes), it creates a much more homogenous study population to examine. The more data we can use to create more transparency around the asset class, the better. Allows us to focus on the real issues and not the misperceptions and myths about the sector.

  • dvhwgumby

    I really don’t get this analysis AT ALL. Fund level return is the only reasonable metric since LPs don’t get to invest in individual deals (and especially not only in the ones which, in retrospect, were winners) any more than index investors get to chose who goes into the index. The LPs get a piece of EVERY deal, and presumably the GPs are picking only ones they think prospectively will be winners.

    You can’t cherry pick the winners retrospectively; the survivorship bias will obviously pretty much guarantee you an excellent return, just like a mutual fund or hedge fund who ignores fees will show an astonishing return.

    The correlation guys seem pretty smart but this report doesn’t show them at their best.

  • Stan Fleming

    Thanks, Bruce. The Correlation database is potentially a valuable tool for understanding the dynamics of our business and providing an independent perspective on performance. To see the implications of these findings, it is important to understand how and why these results differ from those of other sources like Kauffman. Both Kauffman and Cambridge report consistent results at the fund level that are well below those shown here. On the other hand the HBM per-deal exit data are closer to the Correlation numbers. The clinical success ratios reported by pharma would seem to track closer to the fund-level analyses. The venture capital loss ratio may be favorable, but the same can’t be said about pharma. While venture can run trials for less cost than pharma, I don’t buy the idea that small biotech clinical teams are smarter than their pharma brethren and have a higher probability of success. After all, we recruit our teams from pharma; they are us.
    If the 7900 deals constitute an accurate surrogate for the larger bio-venture portfolio and the industry were able to provide the return ratios you show, bio-venture should be stronger than it is today. Perhaps those multiples are driven by the extraordinary performance of a small minority of funds, which would be typical of venture in general and carry important implications for the future of the industry. However, that would appear to be inconsistent with quartile analyses by vintage.
    I would like to know more about the Correlation data and what it says about the venture landscape today. How do we translate the Correlation results into a stronger institutional investment hypothesis for bio-venture? Does Kauffman have to be wrong in order for Correlation to be right, or can both be right (which would leave us back in the institutional penalty box–good deal performance but weak fund results)? Unless and until we can explain the difference in the two perspectives, the institutional investors are going to believe Kauffman.

  • LifeSciVC

    Stan, thanks again for your engagement on this. Importantly, as noted in the post, the 7900 deals represent ALL the venture-backed realizations of the 2003-2012 period, not just biopharma. There were about 750 biopharma realizations. I think the Kauffman analysis and the Correlation analysis can both be true. Kauffman is fund-level analyses which include lots of in-process unrealized deals and addresses the ultimate returns from its limited basket of VC funds. The Correlation analysis focused on realized deals themselves, and exits – 7900 of them from 2003-2012. Its very different. Kauffman highlights how unrealized deals are often uncorrelated with ultimate realized returns, which in some ways makes the Correlation analysis even more interesting. I’d like to believe that when all is said and done on the deals funded in the mid-2000s that the upward curve we are seeing on realized returns in 2009-2012 is a leading indicator of success going forward on a fund-level. Reality is our asset class (and “returns” in the class) was overly distorted by the huge amount of capital raised in 1999-2001 that has taken nearly a decade to work through. With that albatross behind us, and capital flows more rational, returns in the asset class should improve and the data from Correlation suggests that’s happening for realized returns in the past 3-5 years. That’s a good sign. Hopefully more to come.

    Bruce L. Booth, D.Phil.

    ATLAS VENTURE | 25 First Street, Suite 303 | Cambridge, MA 02141 | P: +1 (617) 588-2636 | C: +1 (917) 302 8571
    @LifeSciVC | Blog: | EA: Ann –

  • LifeSciVC

    Dvhwgumby, I don’t argue with the premise that the ultimate measure relevant to LP’s is fund-level returns, but as a leading indicator of changes in the return profile of the asset class, trends in realized returns are an interesting and provocative datapoint. 7900 realized deals – winners, losers, and kick saves – are all included here. Its not a analysis of winners alone – in fact, only 32% of the deals included over 2003-2012 in this analyses were positive returns. The point of this analysis is to not change the vintage-based perspective of returns for LP’s, but to highlight that the realized return trend is more positive than perceived, changing for the better, and may be a leading indicator of better vintage level outcomes in the future. The Kauffman report itself highlighted that unrealized returns were irrelevant – they didn’t correlate with ultimate realized outcomes. So focusing on such a large dataset of realizations is instructive in thinking about how the return dynamics in venture may be evolving. And also that the myths and realities of venture might be wrong (e.g., the myth that biotech somehow trails other sectors in the 2000s is just not borne out with data). I clearly didn’t frame the analysis properly and my update above tried to frame it more clearly.

  • dvhwgumby

    Thanks for responding, Bruce. I guess I am still skeptical.

    It’s all about whether this change really is a harbinger of a larger trend or not. It’s a good point that there are almost 750 data points, but that’s 750 from an n of what (or even the broader number of 8000 from an n of what)?

    Perhaps you can find a smart kid out of Stanford GSB or the math department who wants to crunch a more complete dataset (if they can even get it). A macro (not fund level) analysis would be fascinating enough! I suspect it would also be depressing, though, like reading one of Ioannidis and Panagiotou’s papers on biomarkers.

    What this really says is that Life Science investing is a very difficult job on top of being in a very difficult discipline. I don’t think it’s that great for entrepreneurs either, who don’t even get a portfolio benefit (and take so much capital that they don’t get a big payout even with a successful exit). Neither of these facts bodes well for the long term.

  • Mr.Booth, thanks for this. This is your second recent article about research from Correlation Ventures (the former one being from October 3). When I go to Correlation Ventures, I can’t see the reports. Instead, CV’s news webpage links to your Forbes articles. Are the reports themselves publicly available or does CV just share them with co-investors or other people known to the partners?

  • Investors are 2 problems identified. First, good things can if not adventurous. Second, if you accept the inevitable risk you will have great success or failure is extremely

  • I want to accept these conclusions. I really do. However, I have struggled with them for weeks. I see two problems.

    The first one is easy to identify: The S&P 500 is a poor benchmark. The question posed: “How did biotech (or HC generally) do versus the broad market?”, encompasses at least two risks (sector and liquidity) that can be easily decomposed.

    I do not have access to the database so cannot replicate the cashflows in the dollar-weighted analysis discussed above. However, the time-weighted results of the S&P 500 versus the NASDAQ Biotech Index (NBI) from December 30, 2002 to November 29, 2013 show an approximately 100 percent return to the S&P 500 and a 350 percent return to the NBI without dividends reinvested. I suspect that if I reinvested the dividends and copied the dollar-weighing of Correlation Ventures’ deal database, that the VC returns would not outperform the NBI.

    So, it’s not clear that taking the liquidity and manager risk of investing in VC deals beats passively replicating the biotech index.

    Second, while I accept that IRRs of unrealized deals are untrustworthy, simply ignoring them cannot be done, especially for recent years. This is because the most winning companies in the portfolio surely get sold first. That is, the 2012 exit year surely includes some rock-star deals from 2007-2009, but those vintages will also have unrealized investments that are zombies, which will be sold at a loss in a few years. So, the “hockey stick” in the last five years of the chart is surely statistical artifact. This timing differential between winners and losers will always be a problem unless unrealized investments are included, even if the valuation is flawed.

  • MB

    Bruce, this is an interesting post and data.

    It seems an advantage in investing in the S&P500 over
    the long term is that there is a good level of predictability about its fundamental
    returns. For instance, we would expect the total fundamental return of the S&P500
    to be equal to real GDP growth + inflation + dividend yield.

    Whereas in biotech there does not seem to be the same level
    of predictability in estimating the fundamental returns that investors can get
    from the sector. Over time with innovation, we may expect there to be new
    discoveries and good returns from these discoveries but along the way there
    will also be a high number of failures.