CAIA® alpha and beta: what explains a fund return?

On this page

A fund's return can be separated into a risk-free component, compensation for exposure to a chosen market factor, and a residual called alpha under that model. Beta tells you how strongly the fund tends to move with the chosen market; alpha is the part left after the specified risk adjustment. For a CAIA® candidate, the crucial qualifier is under that model: change the benchmark, add a relevant factor or allow beta to change over time, and the measured residual may change too.

Start with the benchmark and the period

A raw return does not say how much was earned for taking systematic risk. A fund that rises when the broad market rises may have delivered a large nominal gain while adding little beyond its market exposure. Conversely, a modest positive return in a weak market could be meaningful after risk adjustment. The CAIA Level I curriculum companion for 2024 explicitly calls for candidates to calculate beta, forecasted alpha and realised alpha, and to discuss model errors and changing beta. That is why the exam asks for a decomposition rather than a judgement based on the headline return alone.

Beta is a relationship with a specified factor, usually estimated from a series of paired fund and market returns. It is not a permanent property stamped on a strategy. A beta close to one means the fund has tended to move roughly with the chosen market in the chosen sample; a beta above one means larger average sensitivity to that factor, not an assured result in every future period. A beta near zero describes limited *linear* sensitivity to that particular factor. It does not establish that a fund is risk-free, uncorrelated with every other factor, or protected in a crisis.

The common single-factor calculation is compact, but the labels must stay precise. Let the fund return be the observed return, the risk-free rate represent the time-value component, the market return be the chosen benchmark's return, and beta be the sensitivity to that benchmark. Then the benchmark-adjusted return is the risk-free rate plus beta times the market's excess return. Alpha is the fund return minus that benchmark-adjusted return. The CAIA discussion of Jensen's alpha presents this form and shows why the benchmark and fee treatment matter.

Work an invented example from gross return to residual

Take a fully illustrative one-year example with invented round numbers. Suppose a fund earns 12%, the risk-free rate for the period is 2%, the chosen market earns 10%, and the fund's estimated beta is 0.75. The market's excess return is eight percentage points. Multiplying by beta gives six percentage points of market-linked compensation. Add the two-per-cent risk-free component and the model benchmark is 8%. Subtract that from the fund's 12% result: realised alpha under this model is four percentage points. The pieces reconcile to the observed 12%: two from the risk-free rate, six attributed to market exposure and four as the residual.

Do not call the six-point piece a fee-free investment you could necessarily replicate exactly. In this teaching example it is a model attribution for one period. Transaction costs, leverage, imperfect tracking and the availability of an investable proxy would matter in a real comparison. Also do not call the four-point residual proven manager skill. It includes everything the model did not explain. Good judgement, random outcomes, exposure to an omitted factor and measurement error could each contribute. The arithmetic identifies a question for investigation; it does not resolve that question by itself.

Now hold the fund and market returns fixed, but imagine an analyst revises the beta estimate from 0.75 to 1.00 using a better-fitting sample. The market-linked component becomes eight points, the adjusted benchmark becomes 10%, and alpha falls to two points. The fund did not earn a different return; the allocation of that return changed because the model input changed. If a candidate reports only the alpha number and omits the beta and benchmark used to obtain it, the number cannot be interpreted properly.

Forecasted alpha is a claim made before the return

An analyst may expect a fund to earn more than its risk-adjusted benchmark before a period begins. That difference is forecasted alpha. After the period, observed returns permit a realised residual calculation. The two calculations answer different questions: one records an expectation under assumptions; the other describes an outcome under a selected model. Actual outcomes also contain shocks and idiosyncratic noise, so a realised positive alpha does not by itself confirm that the forecast came from repeatable skill.

Use a second invented comparison. Before a period, suppose the analyst expects a fund return of 11%, expects the market to return 9%, assumes a 2% risk-free rate and beta of one. The expected model benchmark is 9%, leaving forecasted alpha of two percentage points. Suppose the market actually returns 5% and the fund returns 8%. Using the same beta and the realised market return, the realised model benchmark is 5%, so realised alpha is three points. The realised residual exceeded the forecasted one by one point. That extra point may reflect luck or an omitted exposure; it is not sound to label it skill simply because it has a positive sign.

The testable habit is to mark which inputs are expected and which are observed. Combining an expected market return with an observed fund return produces a number, but it is not a clean forecasted or realised alpha. The same discipline applies to fee treatment: a gross fund return and a net fund return will produce different residuals against the same benchmark. Specify which return the question gives you, and make like-for-like comparisons.

Why alternative investments make the split difficult

The label alternative does not tell you whether a return driver is alpha or beta. A rules-based exposure to a broad factor can be a beta driver even if packaged in an unusual vehicle. An active strategy may aim for returns beyond priced exposures, but it may still carry market, credit, liquidity or other systematic risks. The curriculum companion distinguishes alpha drivers from beta drivers and asks candidates to consider passive beta exposure, product innovation, changing betas and imperfect models. These are features of the strategy and the attribution method, not of its name alone.

Suppose the illustrative fund in the earlier calculation also earns a premium for holding less liquid assets. A single-market-factor calculation may put that premium into alpha because it has no separate place to record it. Adding a relevant liquidity factor may move part of the four-point residual into factor compensation. That adjustment does not prove the manager added no value; it says that a richer benchmark explains more of the observed return. The direction is the exam point: omitting a positively rewarded exposure can overstate apparent alpha, while an overstated beta can understate it when the market excess return is positive.

Betas may also drift. A manager can vary leverage or market exposure deliberately, or a strategy can become more market-sensitive under stress. An average beta estimated in calm conditions may then be a poor description of the next period. A constant-beta model can assign the effect of this changing exposure to alpha. To interrogate a result, examine the observation window, the regime covered, the benchmark's relevance and whether a nonlinear payoff might be hidden behind a simple slope. A single regression coefficient is a summary, not a full account of the strategy.

A practical decision sequence for candidates

Read the question in this order: identify the fund return and whether it is forecast or realised; identify the same-period risk-free and market figures; calculate or use the stated beta; compute the risk-adjusted benchmark; then subtract once to obtain alpha. After the arithmetic, state exactly what the residual can and cannot show. If the question asks about a strategy rather than a number, decide whether its return is primarily a passive exposure to a priced factor or an active attempt to add return beyond that exposure. Then ask whether the evidence could be confounded by omitted factors or changing beta.

The CAIA programme page is a starting point for placing this foundation in the broader syllabus. The how-it-works page can support a study routine in which you practise the calculation and then explain its limitations in plain language. The most common exam mistake is to treat alpha as a directly observed amount of skill. Alpha is a model-dependent residual; a claim about skill requires stronger evidence than one favourable return.