Abe Othman, head of data science at AngelList, just published a research paper about early-stage venture returns (mostly seed) investing that had some surprising (!) findings. Q&A below:
In a blockbuster conclusion, you found that seed stage investors should put money into every credible deal – this is a surprising finding!
Yes, we found that investors could benefit from indexing as broadly as possible at the seed stage, by putting money into every credible deal, because any selective policy for seed-stage investing—absent perfect foresight—will eventually be outperformed by an indexing approach.
We did do some simulations, not in the publication, on what indexing at seed actually looks like over human timescales. Over a ten-year investment window, indexing beats 90-95% of investors picking deals, even when those investors have some alpha on deal selection. So the idea that there are some terrific seed investors that soundly beat indices is not inconsistent with our results.
I guess the tough part is defining and then finding ‘credible deals’?
I think this is the central question that the paper brings up. Based on this report, the definition of credible deals is “historical winning seed deals on AngelList”, which obviously does not provide any forward guidance. One might advance a definition that would be something like “Is there an attainable version of this company that could raise Series A from a VC?”. One of the things that my team has worked on at AngelList is building machine-learning models that assess whether deals are “credible” or not.
The reason that there isn’t a magical money-printing machine for seed investing is that assessing whether a seed deal is credible or not is very challenging. Seed deals often look like two people, a slide deck, and six months of part-time work. The company may not even be founded yet. Is that a “credible” deal?
I think one of the things the paper implies is that people that are good at finding and assessing credible seed deals will make a lot of money. I think it’s possibly the most important role that can be played in the economy going forward.
I think one of the things the paper implies is that people that are good at finding and assessing credible seed deals will make a lot of money”
In your paper you mention a unique shape to returns at the seed stage called an unbounded mean power law. Can you tell us a bit more about what this means?
What we find is, as you describe, very unique. One way to contextualize it is this: imagine picking investments at random. Under virtually every other possible distribution of returns, picking more investments does not change expected returns. But with this kind of power law, picking more investments actually increases expected returns. This is very unintuitive!
I think another way to contextualize this is from the other side: Regret. How bad do you feel if you miss a terrific deal? At Series D, you’re missing out on perhaps a 5x return; you should have done that deal, it would have made you money, but things are still OK; maybe you invested that money in a deal that did 2x instead.
At Seed, you may be missing out on something like a 5000x return that would completely and totally change your life. You get written up in the Financial Times about how stupid you were to pass and how bad you feel and how jealous you are of people that did the deal.
You paper mentioned your focus was on winning investments – how did you handle the losing investments or companies that it’s not clear if they’re doing/did well or not?
Obviously having fewer losing investments is better, but the results about broadly indexing increasing expected returns at seed holds whether you have 10% or 40% or 50% or 60% or 90% losing investments in the pool you draw from.
If an investor has a crystal ball that can perfectly identify losing investments, then the investor should avoid those. But one of the provocative things about an unbounded mean power law is that if the crystal ball is a little bit cloudy, then eventually on a long enough time span you will miss a winning investment and that winning investment will do so well that you will have wished you had indexed.
You mention that companies grow faster early which may be a reason indexing early stage seed rounds is a compelling strategy. Can you tell us more?
It’s actually two factors: the first is that companies tend to grow faster earlier, which we provide empirical support for with our data. Our results suggest that each successive year of a company’s existence is worth less and less from the perspective of compounding investment growth. For instance, we can expect a company to grow about as much in years three and four, or in years four, five, and six, as it does in its first yearThe second is that earlier investments come earlier than later investments. So taken together, winning early-stage investments get more time to compound at higher growth rates.
The two core factors of the paper are pretty standard conventional wisdom. It’s only when we fit them to the AngelList data that we saw these provocative conclusions emerge.
You go one to mention that, since companies are waiting longer to go public a very small set of investors are capturing a large % of the wealth creation from startup investing, particularly at the seed stage. What should regulators do to improve access to these returns?
Right, one of the more interesting results in the paper is that the unbounded mean power law at the seed stage only emerges after a little over five years of startup lifetime, which is a threshold companies may not have been hitting in private markets last cycle but that they are certainly hitting in this cycle. I would link this concept to the general unease that many market participants: regulators, VCs, LPs, and public market investors, have all felt about startups staying private longer. There is a general sense that by the time companies go public their growth has been “used up”: You can contrast the performance of Uber’s seed round and Uber’s IPO.
On the issue of exposing retail investors to VC, it is not a novel idea; there are publicly traded venture capital firms in Europe. You can take a look at Draper Espirit, and in their annual report you can see the level of disclosure that is required of their portfolio companies. European VCs can get away with this because their venture markets are much less mature and companies simply have fewer options. On the other hand the thinking has been in the US that no good early-stage company would ever agree to take investment with a firm that would require such disclosure when they could just take money from someone that doesn’t impose that disclosure burden on them. So there is an unfortunate tension between opening up the asset class (which I support) and investor disclosure (which I also support).
An informal description of one of the ideas we have had that would be consistent with existing disclosure rules and investor protections is a fund-of-microfunds model: a publicly traded investment fund that invests in lots and lots of early-stage microfunds, but then the microfunds would serve as a firewall to the disclosure of actual company valuations. That is, the disclosure would stop at the microfund level, e.g., “we hold a partnership interest in New Seed Venture Fund L.P. with a current market value of $367,000”, but not disclosing the valuations of the underlying investments in that partnership interest. So it wouldn’t be a total black box, but it also wouldn’t be adversarial to the microfunds when they are trying to source deals.
I should say that I’m not a securities lawyer and do not have a full understanding of the issues at play. Our response to a recent SEC Concept Release, on the other hand, is written by securities lawyers and should be considered AngelList’s position on these matters.