One of the things I’ve learned over the years is that there is a persistent dualism in conducting research balance—a between reliable preexisting findings and questioning preconceived notions.
New research often builds on old findings. Those findings themselves come off the back of other discoveries and so on. “It takes a village.” This unbroken chain drives beneficial compounding effects for innovation and new understanding. Sometimes though, starting from scratch can bring in new insights. Jeff Bezos had no previous formal experience in the retail book business; Elon Musk had no experience in car manufacturing. In the same manner, it always behooves the investment analyst to “trust, but verify”. Sometimes old paradigms, upon closer inspection, are ripe for disruption.
Access to quantitative analytics are available freely, which is to say that almost anyone can generate some really appealing empirical evidence that has no bearing on reality. It is more important than ever to start from first principles when tackling research questions. I am fully embroiled in one of those right now, probably more so than in quite some time on one topic.
A few months ago, my friend Steven Wood of Greenwood Investments and I were discussing markets, managers, and investment philosophy. Our conversations are usually full ranging because he is a hard core Graham and a Dodd guy. I’m a quant, or at least pretend to be one in my day job. His investment theses are deep, award winning, and enlightening.
Somehow we landed on the topic of manager performance—clearly a sore spot for the industry. I floated the idea that there are two characteristics, regardless of asset class—equities, bonds, VC, PE, derivative, etc., that dictate investment portfolio performance. I have not yet in my career come across a unified theory of investment performance. There may be one, I just haven’t found it (and no, MPT doesn’t count because it doesn’t work!).
These characteristics, which I refer to as dimensions, are pretty intuitive at first glance. But, proving out the logic has taken an ungodly amount of time given the requisite trips down rabbit holes for probability, statistics, information theory, performance attribution, option pricing, and a whole lot of mathematical pyrotechnics. The other day I left my notebook on my home desktop. My wife opened it up expecting some juicy personal diary intel, but was met with pages of Greek variable scribble-scratch.
At first, I posited two dimensions of performance—consistency and magnitude. Consistency has to do with how often the manager generates winning bets. Magnitude is the ratio of wins to losses. Anyone with experience as a trader is probably laughing right now, because these are such basic concepts. Through just these two dimensions, one can relatively easily break down the return profile of most portfolios. Here are two extreme portfolios that fall on either ends of the spectrum—insurance and venture capital.
Insurance portfolios include lots of small positions. Those bets frequently win, but have capped upside (the premium collected), and unknown and infrequent downside. Insurance companies are writers of out of the money put options on unknown, but estimate-able risks. Venture portfolios fall at the other end of the spectrum. They are effectively portfolios of out of the money call options on the future success of some investable idea. Downside is known, frequent, and quantifiable, while upside is unknown and infrequent.
What Steven and I realized is that a key third dimension is that of manager conviction. If a manager invested in a stock that shoots the moon, but its only 1% of the portfolio, its contribution to performance may be small. Conversely, if a losing bet carries a weight of, say, 50% in the portfolio, that would be very very bad for performance. All else equal, two managers could select the exact same investments, but the manager that overweights good investment outcomes will do much better than the manager that overweights bad outcomes.
To complicate matters, this element of conviction is not static. When facts change, so should new information be incorporated into portfolios. More technically, this is known as Bayesian Inference. For example, venture investing commonly features follow-on investment opportunities. A VC manager makes an investment; the company achieves some set of pre-determined objectives, and then raises more money to achieve the next set of objectives. If the venture investor doesn’t follow-on by putting in more money, their investment can get diluted. If they do follow-on, the investment represents a greater allocation in their overall fund. Remember from above that conviction can result in vastly different performance of portfolios with the exact same underlying investments. This is a critical decision for VC’s. Some are very much for it. What should a VC do?
This is the dimension I have spent the most time researching. Conviction is ultimately about portfolio construction choices and whether the manager’s wins correlate with position weights in the portfolio. Look for upcoming posts, and a paper, that attempts to tie all these concepts together.