Portfolio theory suggests that making investments in uncorrelated assets could substantially lower the risk that investors must shoulder to achieve their desired investment returns. Unfortunately, one well-known complication with portfolio optimization is that it is very challenging to predict the future correlations between asset classes. For instance, Russell Investments data suggests that over the past 30 years, the correlation between stocks and bonds has shifted from modestly positive to modestly negative in most years but with occasional sharp reversals.
Invesco’s whitepaper The Case for Venture Capital features an astonishing result casually tucked away in a table on the second page: that the correlation between venture capital returns and large-cap equity returns is actually slightly negative, at -0.06. This was a surprising result because we would expect to see a positive correlation. For example, startup exit values such as IPO prices are highly dependent on public market valuations.
Fortunately, AngelList Venture makes so many early-stage investments—we participated in the seed rounds of more than 1,500 startups in 2019—that we can examine the recent correlation between public markets and early-stage venture using our internal data. The below figure shows the gross returns1 of the AngelList seed portfolio on a monthly basis against the performance of the Nasdaq Index from January 2015 through March 2020. Results from the first quarter of 2020 are shown in pink.
The correlation between these two time series is 0.0, totally consistent with Invesco’s original finding.
An obvious rebuttal is that it is unrealistic to expect simultaneous changes between public and private markets. Instead, a more accurate model of the world could be that the public markets do well or poorly, and that performance is then reflected gradually, perhaps over the next twelve months, in the private markets as companies get priced up or down when they go out to raise money. So what we really should be interested in is the lagged correlation between the index and our portfolio, which we should expect to be modestly positive over a near-term window. That turns out to not be the case either:
In this plot, the blue line is the lagged correlation between the monthly percent changes in the Nasdaq index and the gross performance of our seed portfolio. The orange lines represent the critical correlation values for significance (shown at p < 0.01, in an effort to accommodate the multiple hypothesis testing here): values between the orange lines are not statistically significant. Through 20 months none of the lagged correlations are statistically significant, even at the p < 0.05 level, and even fishing through multiple hypotheses.
We can think of several possible explanations for why early-stage venture might be uncorrelated with the public markets.
First, in the immediate aftermath of the recent stock market crash, every liquid asset—stocks, bonds, gold, oil, Bitcoin—saw their asset prices collapse in what appeared to be a flight to dollars. This increases correlation between these asset classes because all of them are moving simultaneously downwards. Early-stage venture could be insulated from some of these effects because of its illiquidity; even if investors wanted to sell off their positions, they would be unable to. This forces them to treat early-stage venture capital as a distinct asset class in the short term.
Additionally, early stage venture returns are generally driven by a few outliers that grow tremendously over a long period - Uber was over a 2,500X return for investors, but it was 10 years between the early stage financing and the end of lockup for those investors. If startup valuations are bets on what will happen many years later in the market that could provide insulation from the fluctuations of short-term market conditions.
Looking Longer Term
The AngelList Venture investment portfolio is only seven years old. To learn more about long term trends, we looked at venture capital benchmarks published by Cambridge Associates. We see only a very weak correlation between annual vintage TVPI performance and public market equivalent indices:
Similarly, indices with very long time horizons do not appear to be correlated with public market equivalent indices:
These results may align with conventional investing wisdom. As we have mentioned, venture investors typically expect a small number of breakout companies each year to drive the vast majority of venture returns. This smaller sample may have more sources of variance and thus a lower correlation with a large public market index. Moreover, the value of these breakout companies is likely driven by idiosyncratic, company-specific factors. Whether or not a startup can bring an emerging technology to market and dominate their industry is usually a bigger driver of value than the health of the market or economy.
In short, we find no evidence for correlation between early-stage venture and public markets looking at both AngelList and long-term performance data. It’s never the wrong time to build the future.
1 Gross returns do not reflect the fees and carry that would have been charged to investors. We use gross returns here to include the fund investments that make up the majority of recent AngelList seed investments and the impossibility of accurately dividing their portfolio-level fees to the individual investment level. Informally, the effect of looking at net returns would be to shift all of the scatterplot points down a bit, which would have limited effects on correlation.