Alpha vs Beta: What You Are Actually Paying For

Reading-order note. Numbered #3 and drafted fourth, immediately after CAPM, because the alpha-beta split is a direct consequence of the model: alpha is defined as whatever the model does not explain. All figures come from the hub's canonical parameter set and are illustrative teaching values. This article is a cost argument and a measurement problem: it does not claim that skill does not exist or that active management cannot work, names no fund or manager, and endorses no fee level.

The alpha-beta distinction is how that claim gets tested. Beta is return that comes from bearing market risk: available to anyone, in unlimited quantity, at close to zero cost. Alpha is return the risk exposure does not explain, the part attributable to the manager rather than the market. The two components have wildly different prices, and most published performance figures do not separate them.

Separating the two, and why raw outperformance misleads

The decomposition is mechanical once you have CAPM. Take the return, subtract what the asset's market exposure alone should have produced, and whatever remains is alpha. The number that looks like skill in a marketing document is usually the raw excess over an index, and raw excess mixes the two components together: a fund that simply holds more market risk than its benchmark will beat that benchmark in rising markets without any skill whatsoever, because it took more of the thing everyone gets paid for. On the worked example below, a fund beats the market by 2.0 percentage points and exactly half of that is beta: it was carrying more market exposure than the index it was measured against. Only the other half is a candidate for skill. Which makes the first practical rule straightforward: an outperformance figure quoted without the risk taken to produce it is not evidence of anything. This is also why which model you use to define "explained" matters enormously. Alpha measured against CAPM is the residual after market exposure. Measured against a model that also accounts for company size, valuation, momentum and other documented return patterns, the same fund's alpha frequently shrinks or disappears, because what looked like skill was exposure to a return pattern that could have been bought systematically. Alpha is not a property of a fund; it is a property of a fund relative to a model, and a manager can be genuinely skilled at harvesting something that has since become purchasable at low cost. Factor models are Pillar 28's territory.

The cost arithmetic, which is where the argument actually lives

The canonical fee pair in this pillar is 0.90% for an active fund and 0.10% for an index fund, a gap of 0.80 percentage points a year. That sounds negligible and is not. On $10,000 over 20 years at a 9.0% gross return, the index version reaches $55,024.69 and the active version $47,480.35. The difference is $7,544.34, which is 13.7% of the final outcome, consumed by a fee gap of eight tenths of one percent. The mechanism is that the fee compounds against you every year while the return compounds for you, so the gap widens with time rather than staying proportional. Now put the two halves together, which is the article's central point. Suppose a manager genuinely generates 1.0 percentage point of alpha: real skill, correctly measured, not luck. Charging 0.90% against that alpha, the investor keeps 0.10 percentage points and the manager keeps 0.90. The skill was real; nine tenths of its value went to the person who supplied it. This is the honest form of the argument: not that alpha is impossible, but that it must exceed the fee before the investor sees a cent of it, and that the fee is charged with certainty while the alpha is not. There is also a structural point about the aggregate, developed properly in Active vs. Passive Investing: since all investors collectively hold the market, active investors as a group must earn the market return before costs and less than it after, an accounting identity rather than an empirical claim. That identity constrains the average and says nothing about any individual manager, which is precisely why it is an argument about costs rather than about competence, and why the article linked above carries the qualifications that belong with it.

The problem that makes alpha nearly unverifiable

Even where alpha exists, distinguishing it from luck takes far more data than anyone has. A fund's returns wobble around its benchmark, and that wobble, tracking error, is the noise against which a small alpha signal must be detected. The arithmetic is unforgiving. For a manager producing 1.0 percentage point of alpha with 4% tracking error, establishing that the result is not chance at conventional confidence takes roughly 64 years of track record. At 6% tracking error, about 144 years. A larger alpha helps enormously: 2 points at 4% tracking error needs about 16 years, and 3 points about 7. But modest genuine skill is, in a strict statistical sense, undetectable within a career, let alone within the three- and five-year windows on which funds are actually marketed and selected. Three consequences follow, and none of them is cynical. A strong five-year record is weak evidence of skill, because across thousands of funds a great many will produce one by chance alone. Manager changes reset whatever evidence had accumulated. And the survivorship problem compounds it: the records available to inspect belong disproportionately to funds that survived, a distortion Survivorship Bias in Data: Why Dead Companies Matter covers directly and this pillar's closing articles develop. The reasonable conclusion is not that skill is a myth; it is that confidence about identifying it in advance should be much lower than the industry's marketing implies, and that a fee is a certainty weighed against a possibility.