Alpha

Alpha is the difference between an investment's actual return and the return that would be expected given the market risk it took. If a fund carried a certain amount of market risk (its beta) and the market delivered a certain return, the capital asset pricing model predicts what the fund should have earned. Whatever it earned above that prediction is positive alpha; whatever it fell short by is negative alpha. Zero alpha means the result was exactly what its risk would predict, no better and no worse.

Alpha is therefore the standard way of asking whether a result reflects skill or merely exposure. A fund that returned 12% in a year the market returned 12% did not necessarily do anything clever; it may simply have matched the market's risk and the market's reward. Alpha strips out the reward that came from bearing market risk and reports only what is left, which is the part a manager can plausibly claim credit, or blame, for.

The hard problem with alpha is telling skill from luck. In any period, some funds post positive alpha and some post negative alpha, and pure chance guarantees that a share of managers will beat their risk-adjusted benchmark even if none has any skill at all. Distinguishing a manager who is genuinely adding value from one who got a good draw requires a long record, and even then the signal is faint. The published evidence is sobering: across long horizons, the majority of actively managed funds underperform their benchmarks after fees, which means aggregate alpha net of costs is negative. This is not an accident. Active management is close to a zero-sum game before costs (one investor's outperformance is another's shortfall), so once fees are subtracted the average active dollar must lag. This is the empirical heart of the case for low-cost index investing.

Alpha is also fragile in a way that flatters managers. It is measured relative to a chosen benchmark and a chosen risk model, and a manager can appear to generate alpha simply by taking a risk the benchmark does not capture. If a fund loads up on small companies or cheap "value" stocks, and those factors happen to pay off, a single-factor model that only knows about overall market risk will score the extra return as alpha, when it was really a reward for bearing a different, identifiable risk. This is why more elaborate factor models exist: they try to strip out the returns that come from known factor exposures, so that what remains as alpha is smaller and harder to fake. Much of what once looked like manager skill dissolved once these factors were accounted for.

One naming point avoids confusion. A separate, unrelated use of the word has spread through the advice industry: "advisor's alpha," a marketing term for the value an advisor may add through steps such as tax-efficient placement, disciplined rebalancing, and keeping a client invested through downturns. Whatever the merits of that idea, it is a different concept from the statistical alpha defined here, which is strictly about risk-adjusted investment return relative to a benchmark. The two share a Greek letter and nothing else.

Start with what a security's risk predicts. Suppose a short-term Treasury yields 4%, the broad market returns 10%, and a fund has a beta of 1.2. The capital asset pricing model predicts the fund's return as the risk-free rate plus beta times the market's excess return: 4% + 1.2 × (10% − 4%) = 4% + 7.2% = 11.2%.

Now compare the prediction with reality. If the fund actually returned 13%, its alpha is 13% − 11.2% = +1.8 percentage points: it earned 1.8 points more than its market risk alone would justify. If instead it returned 9%, its alpha is 9% − 11.2% = −2.2 points, meaning it underperformed what its risk predicted even though it made money.

The lesson in the second case is the point of the whole measure. A fund can post a positive absolute return and still have negative alpha, because it earned less than an investor could have gotten by simply taking the same amount of market risk through a cheap index fund. Alpha asks not "did it make money?" but "did it beat what its risk should have produced?" All figures are illustrative.

Pros

• It separates skill from market exposure, so it credits a result to decisions rather than to a rising tide.
• It makes managers and strategies comparable on a risk-adjusted basis rather than on raw return.
• Consistently positive alpha, sustained over a long record, is genuine evidence of value added.

Cons

• Skill and luck are extremely hard to tell apart over any short record, and chance alone produces positive alpha for some managers.
• Most actively managed funds show negative alpha after fees, so paying for alpha usually buys underperformance.
• It depends on the benchmark and risk model chosen; hidden factor bets can masquerade as alpha under a simple model.
• It is backward-looking, and past alpha is a weak predictor of future alpha.