Do Angel Investors in

California Outperform

the Rest of the World?

(Hint: Yes. By a lot.)

I’ve been working on figuring out the attributes that might be correlated with successful Seed-stage Angel investors. But, before one can make much progress on any serious investigation of this question, the definition of “successful” is a big hurdle. It is a surpisingly complicated issue for a number of reasons:

  1. It takes a very large number of observations to separate luck and skill in a statistically significant manner, when you look at the returns data. This is particularly true given the Power Law nature of the returns from investments into startup companies.
  2. Getting a bias-free set of data that has realized and unrealized returns associated with Seed-stage Angel investments is difficult. This may well be an insurmountable problem.
  3. Vintage year of the investment is the single largest determinant of success of any given startup investment. Unfortunately, it is only knowable several years down the road whether or not a given vintage year turned out to be a good one or bad one. In the current environment, where the S&P 500 has gone up an average of 20% per year for each of the previous six years, the tailwind at the back of every Angel investment done in recent memory is very strong. So, we might be tempted to draw our consideration universe from a pool that goes back very far in time in an attempt to include good and bad vintage years. Unfortunately, this would require going back a couple of decades, and it is hard to find consistently defined and measured data sets over that long a period of time. And, more fundamentally, the nature of Seed-stage Angel investing has changed materially over the last couple of decades, making comparison between the decades highly suspect, if not outright nonsense.
  4. Ideally, we like to use results associated with eventual actual cash returns on investments. But, given an average time to liquidity of seven or eight years, basing an analysis on actual returns dooms us to consider only those investments made long ago — in a period of time very different from the present.

Useful Proxy: Did a Series A Round Happen?

All the difficulties outlined above can more or less be dealt with by using the binary success outcome of “did a subsequent Series A financing happen for this startup after this Seed-stage Angel financing?” Of course, this ignores the successful acquisitions that sometimes happen before a Series A can occur. But, these acquisitions are modest in number as percent of all Seed investments, and even more modest as a percent of all returns generated from Seed investments.

Probably the biggest problem with this simple yes/no binary success metric is that it ignores the huge difference in the likely success trajectory of a company that did a small Series A financing at a modest valuation from a middling investor and a company that did a large Series A financing at an eye-popping valuation from a top-tier investor. This issue is addressable with data that is fairly readily available, and may be worth the hassle of incorporating it into this type of analysis. That being said, there is a lot to be said in favor of using the simple binary “Did a subsequent Series A Round Happen” as a Seed-stage investment success metric, including:

  1. The data is generally available in a relatively complete form for the recent past. I would like to acknowledge the good folks at CB Insights for the data used in the analysis below.
  2. While the fraction of companies that see a subsequent next round happen is clearly a function of the investment environment, the binary “Did a Series A Happen?” success metric doesn’t change anywhere near as much as a measure of step ups in valuation would in a hot market.
  3. It only takes a year and a half for most Series A financings to happen, so we don’t have the big time lag that waiting for ultimate exit returns would require.

California-based Angel Performance

The CB Insights data is well structured for dichotomizing populations of investors, constructing the populations of companies they have invested in, then performing subsequent analyses on those populations of companies. In this case, I tested the simple question:

Do Seed-stage Angel investors based in California outperform Seed-stage Angel investors based elsewhere?

The analysis here covers Seed-stage investments made from 1/1/2010 to 1/1/2014, to allow time for subsequent Series A rounds to happen. The tables below summarize the results of these queries, and a statistical significance test of the question of whether or not the averages of the two populations are actually different.

Companies that have a California-based Angel investor a much more likely to have a subsequent Series A financing.
A Difference in Means test shows a very high confidence in the assertion that the populations have different Average values.

The main takeaways from the two tables above are:

  1. California-based Angel investor dramatically outperform Angel investors in the rest of the world.
  2. The difference is statistically valid with a high degree of confidence.

A Difference in Means over nine Standard Errors is astronomically high. It means that it is much more likely that I have screwed up the analysis, or the underlying data has biases in it, than there is a chance that this outcome was reached due to statistical noise. Barring errors, we can be just about certain of the assertion that companies who have a California-based Angel investor in their Seed round are more likely to do a subsequent Series A round. Importantly, since the Standard Error is inversely related to the square root of population counts, we have a lot of statistical headroom for additional analyses that split the Investors into smaller populations. This is great because there are many additional questions to resolve!

As to the causal issues associated with the relative lack of success of Seed-stage Angel investors outside of California, I remain surprised at the performance persistence that California has. I would have thought the trends of the last decade or two would have led to a democratization of the startup investing opportunities out there, reducing the California advantage to something much more modest than we find here. But, the data is pretty clear.

I feel one final caution is in order. We’ve taken “Did a Series A happen?” as a proxy for what we really want to measure — “Did this Seed investment generate a good return?” — because what we really want to measure takes too long to realize and has outcomes with a Power Law distribution, which are noteworthy for how hard they are to analyze. But, this does not lend any support to (nor contradict) the notion that it is wise for Seed-stage Angel investors to optimize primarily for investments that will get a Series A financing.