Alpha（阿尔法超额收益）
Alpha is a model-dependent estimate of return not explained by a selected benchmark and risk model over a specified period. It can reflect skill, omitted risk factors, leverage, data choices, fees, luck, or estimation error; positive alpha is not proof of manager skill.
Formula (Jensen's Alpha)
Alternatively, simplified: Alpha = Portfolio Return − Benchmark Return. The Jensen's version adjusts for risk using beta.
Why It Matters
Alpha is the central concept in the active vs. passive investing debate. If a fund manager generates positive alpha consistently after fees, active management is worthwhile. However, the SPIVA (S&P Indices Versus Active) reports consistently show that over 15-year periods, approximately 90% of actively managed U.S. stock funds underperform their benchmark — meaning most deliver negative alpha. This is the strongest argument for passive index investing: why pay 1-2% annual fees for negative alpha when an index fund delivers the market return for 0.03%?
Yet alpha does exist. Legendary investors like Warren Buffett generated estimated alpha of +10-12% annually in their early decades. The key insight is that alpha tends to be persistent only among a tiny minority of managers, and identifying them in advance is extremely difficult. Small-cap and emerging market strategies show more potential for alpha because these markets are less efficient — information isn't as quickly or accurately reflected in prices. Understanding alpha helps you evaluate whether your investment strategy (or your fund manager) is genuinely adding value or just riding the market's coattails while charging fees.
Real-World Example
Consider the ARK Innovation ETF (ARKK), managed by Cathie Wood. In 2020, ARKK returned approximately 150% while the S&P 500 returned 18%. ARKK's alpha was massive — roughly +130%. However, in 2021-2022, ARKK lost over 70% while the S&P 500 declined about 18%. The negative alpha over those two years wiped out most of the 2020 outperformance. By 2024, ARKK's cumulative alpha since inception was negative, trailing the S&P 500.
Contrast this with Apple (AAPL) stock over the past decade. Apple has consistently generated positive alpha against the S&P 500, averaging approximately +5-8% annually from 2014-2024. An investor who simply held Apple instead of the S&P 500 would have dramatically outperformed. But this is survivorship bias in action — for every Apple, there were dozens of tech stocks that underperformed. Alpha is easy to identify in hindsight and extremely hard to predict.
Common Mistakes
- Ignoring fees when calculating alpha: A fund may return 11% vs. the benchmark's 10%, but if the expense ratio is 1.5%, the true alpha is -0.5%. Always use net returns for alpha calculation.
- Using the wrong benchmark: Comparing a small-cap growth fund to the S&P 500 (large-cap blend) inflates or deflates alpha. Always match the appropriate benchmark for the investment style.
- Chasing recent alpha: Past alpha does not predict future alpha. A fund with +5% alpha over 3 years may simply have been lucky or in a favorable style cycle. Require at least 5-10 years of data.
- Confusing beta-driven returns with alpha: If your portfolio returns 25% by holding 2x leveraged S&P 500 (beta = 2.0) while the market returns 12%, that's not alpha. It's just amplified market risk. Jensen's Alpha corrects for this.
Pro Tips
Use Information Ratio alongside alpha: Information Ratio = Alpha ÷ Tracking Error. It measures how consistently you generate alpha. A high IR (above 0.5) means alpha is reliable, not just a lucky quarter or two.
Decompose alpha sources: Is your alpha coming from stock selection, sector allocation, market timing, or factor exposure? Use attribution analysis to understand what's working and whether it's repeatable or accidental.
Frequently Asked Questions
What does alpha mean in investing?
Alpha is the intercept or residual return from a specified performance model. Its meaning depends on the benchmark, factor model, return frequency, fees, and estimation period. A positive estimate does not by itself prove skill, and a negative estimate does not by itself prove poor decision-making.
How is alpha different from beta?
Beta estimates sensitivity to a selected benchmark or factor. Alpha is the portion of return left unexplained by the chosen model. Changing the benchmark or adding factors can materially change both estimates.
Can alpha persist?
Historical alpha may not persist. Statistical uncertainty, changing exposures, fees, taxes, capacity, survivorship bias, and market regimes can affect future results. Evaluate confidence intervals and out-of-sample evidence rather than relying on one point estimate.
Should alpha be measured before or after fees?
State the convention explicitly. Net-of-fee alpha is generally more relevant to an investor's realized experience, while gross alpha may be used to study a strategy before implementation costs.