The Four Sources of Alpha
What this covers
Patrick talks with Michael Mauboussin about how excess return divides into behavioral, analytical, informational, and technical sources. Each source has its own conditions for exploitation. The conversation moves between market structure and company-level valuation, and Mauboussin keeps returning to two questions: who is on the other side of the trade, and can the investor afford to stay in it?
The behavioral discussion starts by rejecting the syllogism that irrational individuals make irrational markets. It then treats over-extrapolation as the best-established bias, one that even baseball general managers show. Crowds are smart only under certain conditions, and extreme sentiment pulls disciplined investors toward the herd. The informational section separates data from information. It uses the Regulation FD episodes to show how selective disclosure once created an edge, and it treats the credit-analyst exemption and its 2010 reversal as a contested case. The 3Com and Palm trade shows a clear insight defeated because Palm shares could not be borrowed. The technical section covers forced selling, margin calls, and Geanakoplos's leverage cycle. It also cites evidence that fund flows account for up to a third of reported hedge fund alpha.
The analytical turn is the most practical. Mauboussin explains how the Warranted Multiple Framework ties EV-to-EBITDA to incremental returns and growth. That explains why a 5 percent estimate cut can justify a 15 to 20 percent price drop for a high-return grower. He also argues that EBITDA should be split into its components, because accounting changes distort amortization. The closing material is about judgment itself: small samples mislead, Brier score feedback improves calibration, and Benign Myths may be worth more to an organization than their rigor.
Mauboussin argues that alpha comes from four sources (behavioral, analytical, informational, and technical), each defined by who is on the other side of the trade, and that edge is only capturable when investors overcome principal-agent constraints, implementation frictions, and their own decision-making habits.
- Individual irrationality does not imply market irrationality, so behavioral edge is real but hard to exploit because the crowd pulls the investor toward the error.
- Opportunities are lumpy in time, and they are only exploitable by investors who have the committed capital to act when they appear.
- Information is not the same as data, and whether information becomes edge depends on implementation costs, attention limits, and regulatory access.
Behavioral mispricing, especially over-extrapolation, is the largest but hardest-to-exploit edge.
- Behavioral mispricing is probably the largest source of excess return, plausibly more than 25% of the error in markets, but it is also the most emotionally difficult to exploit because the investor must take the contrarian side against crowds.
“I would suspect that behavioral is the biggest one. So something more than 25%”
- Overconfidence is not a market-level bias when traders are on both sides of trades, because one overconfident buyer and one overconfident seller cancel out in aggregate price formation.
“I can be overconfident as a seller and a buyer, and we offset each other”
- Over-extrapolation, the tendency to project recent trends too far into the future, is the best-established behavioral bias and underlies the value factor and momentum; it creates prices embedding unlikely expectations that are probabilistically better traded against.
“this idea of over extrapolation... I think that's the backbone for things the value factor and so forth, and even momentum”
- Investors tend to give money to managers who have performed well and withdraw from underperformers, and fired managers often outperform their replacements over the following 24 months.
“people tend to give money to managers who've done well, and they tend to take it away from managers who have not done well.”
Analytical edge comes from scorable, Bayesian forecasts with feedback that improves calibration.
- Mauboussin's course is organized into four building blocks: capital markets and market inefficiency, valuation (translating prices into value), competitive strategy, and decision-making, with decision-making now the most important because it improves the quality of thinking in general.
“It basically is broken into four chunks.”
- Analytical edge depends heavily on being a good Bayesian, meaning updating views effectively as new information arrives rather than clinging to prior beliefs.
“this idea of updating our views effectively. We call it being a good Bayesian.”
- An investor should lay out the thesis for a stock, state where it deviates from market belief, and assign probabilities to specific measurable signposts such as a quarterly sales threshold, creating a scorable forecast that enables intermediate evaluation and Bayesian updating.
“I should be able to assign probabilistic forecasts to certain signposts. So I believe that their sales are going to be higher than the market believe.”
- In probabilistic fields such as sports and poker, top practitioners systematically document their decisions and reasoning, which is second nature to them, and critical peers can help make that documentation useful.
“in a lot of other probabilistic fields, the really best people do document it.”
- A Brier score of zero means a perfect forecast, while values near two mean consistently wrong forecasts, so a forecaster with a score near two is informative only in the inverse sense, since reversing their calls would be correct.
“zero means you've nailed it... the scale of two, let's say, is, you're wrong about everything.”
- When Brier scores are tracked and fed back to decision-makers such as meteorologists or medical professionals, they become better calibrated, showing that structured feedback improves probabilistic judgment.
“whenever Brier scores are kept, and that feedback is given back to decision-makers... they get better at this. They get better calibrated.”
- Jim Collins's company research may not meet high methodological standards, but managers feel better and are motivated after interacting with him, so the net effect can be positive even when the science is weak.
“The research may not be exactly the highest-quality science, but it's got a lot of other attributes that make people feel good and get them motivated”
Every claimed edge must be tested by who is trading opposite and why.
- Excess return, or alpha, can be decomposed into four sources: behavioral (investor errors), analytical (superior processing of the same data), informational (better or earlier information), and technical (forced buying and selling that is unrelated to fundamentals), and each source has different conditions for exploitation.
“So Behavioral Analytical Informational and Technical.”
- Every candidate source of edge should be tested by asking who is on the other side of the trade and why that counterparty is trading, and whether the investor has a specific reason to believe they are in a better position than that counterparty.
“A very important question to pose every single time. Who's on the other side? Why do I think I have edge?”
- Early access to information cannot be eliminated because there is always an original source, but regulators try to force universal dissemination of material information, which shrinks the edge from being first.
“there's got to an original source information somewhere, right? So this can't, by definition, go away.”
- Central banks and similar credit-market actors may not be trading on fundamental value, which creates distortions in credit prices and a possible source of opportunity.
“there may be a number of actors in the credit markets, especially central banks and so forth, that may not be completely... They're not looking at fundamental value.”
Accounting and valuation conventions, plus capital allocation choices, shape how returns should be read.
- Being much bigger generally hurts investment results, but scale can provide option value that lets a manager become the last capital standing in a crisis, as Klarman, Soros, and Buffett were able to be when LTCM collapsed.
“being much bigger tends to be deleterious to results like that's roughly true. But this is a case where it can give you the option value”
- The amortization component of depreciation and amortization has swung dramatically over the last 30 to 40 years mainly because of changes in accounting rules, especially acquisition accounting for intangibles, so investors who ignore accounting can miss the real economics.
“the pattern of amortization has swung wildly in the last 30 or 40 years, mostly because of accounting changes.”
- Bill Stiritz of Ralston Purina was a skilled buyer of his own stock, repurchasing it when cheap and pausing when dear, which exemplifies the intrinsic-value buyback approach described in William Thorndike's book The Outsiders.
“Bill Stiritz was a genius with this. He was really good about buying it cheap, and laying off when it was dear.”
- EV/EBITDA was rarely used in the 1980s, is now second only to P/E as a valuation metric, and probably originated around 1973 to 1974 with John Malone at Tele-Communications Inc., according to William Thorndike's account.
“I think P/E is still the number-one thing people use, but EV-to-EBITDA is now right behind it”