Alpha appears in every reporting document, but it is rarely calculated with rigor and even less frequently interpreted correctly. This guide covers the calculation of non-annualized alpha: its exact definition, applicable formulas, attribution methods and the errors that systematically distort its reading.
01 — What is performance alpha?
Alpha measures the performance gap between a portfolio and its benchmark, once the level of risk is taken into account. In the context of active management, it is the primary metric used to evaluate what a manager truly adds compared to passive index exposure.
In practice, the term covers two different realities that must be distinguished from the outset:
- Gross alpha (or differential alpha): simple arithmetic difference between the portfolio’s performance and that of the benchmark over the same period.
- Jensen’s alpha: residual outperformance after correcting for systematic risk (beta). This is the “clean” version of alpha, derived from the Capital Asset Pricing Model (CAPM).
An alpha can be positive, negative, robust over the long term or simply accidental. Reading it correctly requires knowing which calculation method was used, over what time horizon and against which benchmark.
02 — The fundamental formula for gross alpha
Gross alpha is the direct difference between the two performances over the same period:
α = Rp − Rb
- α: performance alpha for the period considered
- Rp: total portfolio return over the period (in %)
- Rb: total benchmark return over the same period (in %)
Calculating returns: net asset value or market value
A portfolio’s return over a given period is derived from the net asset values at the beginning and end of the period, adding any interim distributions:
Rp = (NAVt₁ − NAVt₀) / NAVt₀
- NAVt₁: net asset value at the end of the period
- NAVt₀: net asset value at the beginning of the period
- Note: for distribution funds, add dividends and coupons paid during the period to the numerator.
Example — Gross alpha over 18 months
A European equity fund gains +14.8% between January 1, 2023 and June 30, 2024. Its benchmark, the Stoxx Europe 600 total return, gains +11.2% over the same period.
α = +3.6% — Gross alpha of +3.6 percentage points over 18 cumulative months. This figure is not annualized: it represents the effective outperformance over the entire period, not an annual rate.
03 — Jensen’s alpha: integrating systematic risk
Gross alpha says nothing about the risk that generated it. A manager overweighting assets more volatile than the benchmark will show positive gross alpha in a bull market through a purely mechanical effect, without this reflecting any added value in stock selection.
Jensen formalized in 1968 a measure that corrects this bias by integrating systematic risk via the CAPM:
αJ = Rp − [Rf + βp × (Rb − Rf)]
- αJ: Jensen’s alpha for the period
- Rp: portfolio return over the period
- Rf: risk-free rate over the period (e.g., €STR or short-term government bond yield)
- βp: portfolio beta relative to the benchmark
- Rb: benchmark return over the period
- [Rf + βp × (Rb − Rf)]: theoretical expected return given the risk assumed
The bracketed term is the portfolio’s theoretical expected return given its market risk exposure. Jensen’s alpha is the gap between the actual return and this theoretical return.
Calculating beta over the period
Beta is obtained by linear regression of portfolio excess returns on benchmark excess returns:
βp = Cov(Rp − Rf, Rb − Rf) / Var(Rb − Rf)
The calculation requires a series of periodic returns (weekly or monthly) over the analysis period, not just beginning and end-of-period performance figures.
Zero risk-free rate assumption — In practice, the risk-free rate is difficult to measure consistently across variable periods. A common simplification is to set Rf = 0, which reduces Jensen’s alpha to αJ = Rp − βp × Rb. This assumption is particularly justified in a low or negative rate environment (2015-2022). Otherwise, use the €STR rate or the 3-month government bond yield for the relevant period.
Example — Jensen’s alpha over 18 months
Same fund: Rp = +14.8%, Rb = +11.2%, cumulative Rf = +2.1%, measured β = 1.15.
Theoretical expected return = 2.1% + 1.15 × (11.2% − 2.1%) = 12.57%
αJ = +2.23% — Gross alpha was +3.6%. Adjusted for risk (beta 1.15), Jensen’s alpha drops to +2.23%. The 1.37 percentage point gap reflects the manager’s additional risk-taking, not selection skill.
⚠️ Warning: gross alpha ≠ Jensen’s alpha — A beta of 1.30 in a rising market mechanically produces positive gross alpha. This is leveraged exposure, not active management. Jensen’s alpha isolates this bias. Always specify to the client which measure is being used.
04 — Why calculate non-annualized alpha?
Annualizing alpha is one of the most widespread errors in financial reporting. Unlike raw performance, alpha cannot be directly transposed to an annual basis without losing its meaning, or even distorting it.
The fundamental distinction
A return can be annualized via the CAGR. But alpha is a difference of returns, and the difference of two compounded rates is not the compounded rate of their difference:
(1 + Rp)^(1/n) − (1 + Rb)^(1/n) ≠ (Rp − Rb)^(1/n)
The difference between the annualized CAGRs of the portfolio and the benchmark is not equal to the annualized gross alpha. These two methods produce different results. Neither is universally correct.
When to use non-annualized alpha
Cumulative non-annualized alpha is particularly appropriate in the following situations:
- Short periods (< 1 year): annualizing a 3 or 6-month alpha introduces significant mathematical distortion.
- Comparison between mandates of different durations: non-annualized cumulative alpha over the exact mandate duration is the most faithful measure of what the client actually received.
- Monthly or quarterly reporting: each period’s alpha is presented as a gross value, then aggregated using geometric linking over successive periods.
- Fixed-term fund analysis (private equity, real estate funds, target-date funds): performance over the total fund life takes priority over annualized performance.
Geometric linking of alphas across sub-periods
Once a period is broken down into sub-periods, alphas cannot simply be added. Geometric linking is required:
αcumulative = [(1+Rp,1)×…×(1+Rp,n)] − [(1+Rb,1)×…×(1+Rb,n)]
Example — Geometric linking vs arithmetic addition
| Period | Portfolio return | Benchmark return | Periodic alpha |
|---|---|---|---|
| Q1 | +5.2% | +4.0% | +1.2% |
| Q2 | −3.1% | −4.5% | +1.4% |
| Q3 | +8.4% | +6.8% | +1.6% |
| Q4 | +2.0% | +3.5% | −1.5% |
| Arithmetic sum | +2.7% — incorrect | ||
| Geometric linking | +13.2% | +9.9% | +3.3% — correct |
The arithmetic sum of quarterly alphas gives +2.7%, while the geometric cumulative alpha is +3.3%. This 0.6 percentage point gap stems from the compounding effect of interim returns. It widens with volatility and the number of sub-periods.
05 — Performance attribution: going beyond aggregate alpha
A portfolio’s aggregate alpha tells us whether outperformance exists, not where it comes from. The Brinson, Hood and Beebower attribution model (1986) allows this alpha to be decomposed according to its actual sources.
“Asset allocation accounts for an average of 91.5% of the variation in portfolio performance over time. Stock selection and market timing together explain only 8.5%.”
— Brinson, Hood & Beebower, Determinants of Portfolio Performance, Financial Analysts Journal, 1986
The Brinson model: three sources of alpha
α = Allocation + Selection + Interaction
- Allocation = Σ (wp,i − wb,i) × (Rb,i − Rb) — Value added by overweighting/underweighting decisions across benchmark segments
- Selection = Σ wb,i × (Rp,i − Rb,i) — Value added by stock selection within each segment
- Interaction = Σ (wp,i − wb,i) × (Rp,i − Rb,i) — Joint effect of allocation and selection within the same segment
Comparison of alpha calculation methods
| Method | Description | Advantages | Limitations |
|---|---|---|---|
| Gross alpha | Direct performance difference | Simple, transparent | Ignores risk taken |
| Jensen’s alpha | CAPM regression residual, incorporates beta | Corrects for systematic risk | Assumes CAPM validity |
| BHB attribution | Decomposes alpha into allocation, selection, interaction | Identifies alpha sources | Requires segmented benchmark |
06 — The 5 classic pitfalls in reading alpha
-
Confusing gross alpha with Jensen’s alpha — A fund with beta 1.4 in a market rising 15% will show positive gross alpha without the manager having done anything special. This is a consequence of amplified exposure, not stock selection. Jensen’s alpha neutralizes this effect.
-
Annualizing short-term alpha — Multiplying a 6-month alpha by 2 to obtain an annualized figure is mathematically incorrect. Over short periods, the distortion introduced can reach several percentage points. Alpha should be presented over its actual duration, without extrapolation.
-
Summing periodic alphas arithmetically — The sum of quarterly alphas does not equal the annual alpha. The compounding effect produces a systematic gap that widens with volatility. Aggregating sub-periods requires geometric linking.
-
Choosing an inappropriate benchmark — Comparing an emerging market equity fund against a domestic index produces a meaningless figure. Three types of benchmarks are available: a market index (Stoxx 600, MSCI World…), a model portfolio representing the target allocation, or a peer group. The latter is often the most relevant: comparable funds share the same management style and a similar beta, making alpha directly interpretable as a skill differential.
-
Interpreting a statistically insignificant alpha — A positive alpha over 1 or 2 years may simply reflect luck. Establishing statistical significance via a t-test generally requires 5 to 10 years of track record. Over short horizons, a positive alpha does not allow conclusions about manager skill.
07 — Practical use for financial advisors
In fund selection
Non-annualized alpha over a complete market cycle (5 to 7 years including both bull and bear markets) is the most relevant indicator for evaluating the consistency of active management. Practical reference points:
- Gross alpha net of fees > 0 over a complete market cycle: the minimum condition to justify active management fees.
- Positive and statistically significant Jensen’s alpha over 5 years: a signal of genuine management skill, not a beta effect.
- Information ratio (α / tracking error) > 0.5: indicative threshold for consistency of outperformance relative to active risk taken.
In portfolio construction
For a professional managing a multi-fund allocation, BHB decomposition makes it possible to distinguish what stems from asset class allocation decisions (strategic bet) from what stems from fund selection within each sleeve (tactical selection).
08 — Conclusion: alpha as a tool requiring rigor
Alpha is the primary measure of value added by active management. But a raw figure without context can equally reflect genuine selection skill, a simple beta effect in a bull market, or a benchmark chosen after the fact.
Three rules suffice for rigorous use:
- Specify the method: gross alpha or Jensen’s alpha. These two measures do not say the same thing.
- Do not annualize alpha over a partial period. Cumulative alpha over the exact observation period is the only correct presentation.
- Aggregate using geometric linking, not arithmetic addition, across successive sub-periods.
For the financial advisor, alpha is also a tool for questioning managers: a positive alpha must be explained by allocation, stock selection or management style — not simply displayed. Brinson attribution tools allow you to go beyond the aggregate figure to understand where outperformance comes from and to what extent it is likely to be repeated.
On EnvestBoard — EnvestBoard uses Jensen’s alpha as the reference measure in its historical simulation module: it automatically corrects for the beta effect and produces a measure of outperformance net of systematic risk. For market cycle analysis, the platform applies BHB decomposition (Brinson-Hood-Beebower) to identify the portion of alpha attributable to asset class allocation versus selection within each sleeve. EnvestBoard gives you access to more than 4,000 possible peer groups.
