Beta and alpha, explained
By Investor Wisdom Daily Team·
Two managers, one good year
Section titled “Two managers, one good year”Two hypothetical managers: Priya’s fund beat the market by a wide margin last year; Sam’s roughly matched it. Priya looks like the better picker, until you learn her fund tends to move about one and a half times as much as the market, and the market rose. Part of her lead was simply extra market risk in a rising year.
Beta and alpha are the two numbers that separate those effects. Beta asks how strongly a return series moves with a benchmark. Alpha asks what is left over after you allow for that. In the main example below, a hypothetical stock earns 0.625 percentage points a month more than its beta “explains”, which is about 0.625 × 12 = 7.5 points over a year if you simply add up months. This lesson shows where that number comes from, and why it still might be nothing but luck.
The idea in one paragraph
Section titled “The idea in one paragraph”A benchmark is the yardstick you compare against, often a broad stock market index. Beta measures sensitivity: if the benchmark moves 1%, how much has the asset typically moved with it? A beta of 1 moves in step, 1.5 moves half as much again, and 0.5 moves half as much. Mathematically, beta is the covariance of the asset’s returns with the benchmark’s (how much they move together) divided by the variance of the benchmark’s returns (how much the benchmark moves on its own). It is also the slope of the straight line that best fits a scatter chart of the asset’s returns against the benchmark’s. Jensen’s alpha is the asset’s actual average return minus the return the Capital Asset Pricing Model (CAPM) would expect for that beta: risk-free rate + beta × (benchmark return − risk-free rate). R-squared says what share of the asset’s ups and downs the benchmark explains, from 0 to 1.
The conventions in this lesson
Section titled “The conventions in this lesson”- Returns are hypothetical monthly returns in percent. No real security or fund is used.
- Risk-free rate: an assumed 3% a year, or 3 ÷ 12 = 0.25% a month. This is a teaching assumption, not a current rate. The Sharpe and Sortino lesson explains where real-world risk-free proxies come from.
- Sample statistics, dividing by n − 1. Beta is the same either way, because the n − 1 cancels out of the ratio.
- Real beta reports usually use far more data, for example five years of monthly returns (60 pairs). Six pairs are for learning the mechanics only.
Worked example 1: beta by hand
Section titled “Worked example 1: beta by hand”The benchmark (call it Market M) and Stock S returned, month by month:
| Month | Market M | Stock S | M − mean (1) | S − mean (2) | Product | (M − mean)² |
|---|---|---|---|---|---|---|
| 1 | 2 | 3.5 | 1 | 1.5 | 1.5 | 1 |
| 2 | −1 | 0 | −2 | −2 | 4 | 4 |
| 3 | 3 | 6 | 2 | 4 | 8 | 4 |
| 4 | 1 | 0 | 0 | −2 | 0 | 0 |
| 5 | −2 | −2.5 | −3 | −4.5 | 13.5 | 9 |
| 6 | 3 | 5 | 2 | 3 | 6 | 4 |
| Total | 6 | 12 | 0 | 0 | 33 | 22 |
- Means. Market: (2 − 1 + 3 + 1 − 2 + 3) ÷ 6 = 1%. Stock: (3.5 + 0 + 6 + 0 − 2.5 + 5) ÷ 6 = 2%.
- Deviations. Subtract each mean from its own column.
- Multiply the paired deviations and add them. 1 × 1.5 + (−2) × (−2) + 2 × 4 + 0 × (−2) + (−3) × (−4.5) + 2 × 3 = 33. A positive product means both were above average, or both below, in the same month.
- Covariance. 33 ÷ (6 − 1) = 6.6.
- Market variance. Square the market deviations and add: 1² + (−2)² + 2² + 0² + (−3)² + 2² = 22. Then 22 ÷ 5 = 4.4.
- Beta. 6.6 ÷ 4.4 = 1.5. Because both steps divide by 5, you can skip them: 33 ÷ 22 = 1.5.
The regression view. Plot the six months as dots, market return across, stock return up, and fit the best straight line. Its slope is the same 1.5, and its intercept is the stock mean minus beta × the market mean: 2 − 1.5 × 1 = 0.5. So the fitted line is “stock ≈ 0.5 + 1.5 × market”. Spreadsheets give these directly: =SLOPE(stock, market) and =INTERCEPT(stock, market), or =COVARIANCE.S(…)/VAR.S(…) for beta.
What beta 0.5, 1 and 1.5 imply
Section titled “What beta 0.5, 1 and 1.5 imply”| Beta | Market falls 10% | Market rises 8% | Plain reading |
|---|---|---|---|
| 0.5 | about −5% | about +4% | Half as sensitive as the benchmark |
| 1 | about −10% | about +8% | Moves in step, on average |
| 1.5 | about −15% | about +12% | Amplifies benchmark moves |
| −0.5 | about +5% | about −4% | Tends to move the opposite way |
Each cell is beta × market move, for example 1.5 × (−10) = −15. Company news adds scatter around these figures.
Worked example 2: Jensen’s alpha and R-squared
Section titled “Worked example 2: Jensen’s alpha and R-squared”Jensen’s alpha for Stock S, with the 0.25% monthly risk-free rate:
- Market excess return: 1 − 0.25 = 0.75%.
- Beta times that: 1.5 × 0.75 = 1.125%.
- CAPM expected return: 0.25 + 1.125 = 1.375% a month.
- Alpha: actual minus expected, 2 − 1.375 = 0.625% a month.
Note that this is not the regression intercept of 0.5. The intercept compares raw returns; Jensen’s alpha compares returns above the risk-free rate. They are linked: 0.5 − 0.25 × (1 − 1.5) = 0.625. When you read an “alpha”, check which version it is.
R-squared. Square the stock deviations and add: 1.5² + (−2)² + 4² + (−2)² + (−4.5)² + 3² = 55.5. That is the stock’s total variation. The part the market explains is beta² × the market’s total: 1.5² × 22 = 49.5. What is left, 55.5 − 49.5 = 6, is stock-specific. R-squared = 49.5 ÷ 55.5 ≈ 0.8919, so about 89% of this stock’s variation lines up with the market. (Same answer from 33² ÷ (22 × 55.5).) With R-squared that high, the beta is a fairly tidy description of this sample.
Worked example 3: a low R-squared and a negative beta
Section titled “Worked example 3: a low R-squared and a negative beta”Fund L returned 1.5, 2, 4, −3, −0.5, 2 against the same market. Its mean is (1.5 + 2 + 4 − 3 − 0.5 + 2) ÷ 6 = 1%. Its deviations from 1 are 0.5, 1, 3, −4, −1.5, 1, and the cross-products with the market deviations add to 1 × 0.5 + (−2) × 1 + 2 × 3 + 0 × (−4) + (−3) × (−1.5) + 2 × 1 = 11. Beta: 11 ÷ 22 = 0.5. Expected return: 0.25 + 0.5 × 0.75 = 0.625, so alpha = 1 − 0.625 = 0.375% a month.
That alpha sounds good, but check R-squared. Total variation: 0.5² + 1² + 3² + (−4)² + (−1.5)² + 1² = 29.5. Explained: 0.5² × 22 = 5.5. R-squared = 5.5 ÷ 29.5 ≈ 0.1864. The market explains under a fifth of Fund L’s movement, so its beta and alpha are shaky; another benchmark might fit far better.
Asset N returned 0, 1.5, −0.5, 0.5, 2, −0.5: mean (0 + 1.5 − 0.5 + 0.5 + 2 − 0.5) ÷ 6 = 0.5%. The cross-products add to 1 × (−0.5) + (−2) × 1 + 2 × (−1) + 0 × 0 + (−3) × 1.5 + 2 × (−1) = −11, so beta = −11 ÷ 22 = −0.5. It tended to rise when the market fell. CAPM’s expected return is 0.25 + (−0.5) × 0.75 = −0.125%, below the risk-free rate, because an asset that cushions market falls is treated as insurance. Alpha: 0.5 − (−0.125) = 0.625%. A modest 0.5% a month shows the same alpha as Stock S. (Asset N has no stock-specific noise, so its R-squared is exactly 1; real negative betas are rarely that neat.)
| Measure (monthly) | Stock S | Fund L | Asset N |
|---|---|---|---|
| Mean return | 2% | 1% | 0.5% |
| Beta | 1.5 | 0.5 | −0.5 |
| CAPM expected return | 1.375% | 0.625% | −0.125% |
| Jensen’s alpha | 0.625% | 0.375% | 0.625% |
| R-squared | ≈ 0.8919 | ≈ 0.1864 | 1 |
Why alpha is not skill by itself
Section titled “Why alpha is not skill by itself”- Luck. Change one number: suppose Stock S had lost 3% in month 4 instead of 0, because of one piece of bad company news. Month 4 had an average market return, so beta stays 1.5, but the stock’s mean drops by 3 ÷ 6 = 0.5, to 2 − 0.5 = 1.5. Alpha falls to 1.5 − 1.375 = 0.125%. One month erased four-fifths of the “outperformance”. Six months of alpha mostly measures noise.
- Period choice. Start and end dates decide which booms and busts are inside the window. A manager or an advertisement can pick the window that looks best.
- Benchmark choice. Alpha is only “excess” against the benchmark you picked. A small-company fund measured against a large-company index can show alpha that is really just its small-company exposure. Fund L’s low R-squared is a warning sign of a poor fit.
- The model is simple. CAPM uses one factor, the market, so returns from other traits (company size, value, momentum) show up as “alpha”.
- Costs and survivorship. Alpha before fees can vanish after them, and high-alpha lists often omit funds that closed.
As the SEC’s Invest Wisely guide puts it, “The past success of a particular investment is no guarantee of future performance.”
Limits of beta
Section titled “Limits of beta”- It looks backward and drifts. Beta comes from a past window; a company that changes its business, debt or size can have a different beta next year.
- It depends on settings. Data frequency and window length change beta; compare only betas built the same way.
- It is not total risk. A low-beta stock can still be very volatile on its own news. Beta covers only market-linked risk, which diversification cannot remove; the risk and diversification lesson covers the part it can. For total spread, see standard deviation in the Sharpe ratio lesson.
- It assumes a straight line. Many assets react more to falls than to rises. A single slope averages that away.
Try it by hand (or in a spreadsheet)
Section titled “Try it by hand (or in a spreadsheet)”Work it by hand first, then check it with the alpha and beta calculator, which computes beta, R², Jensen’s alpha, the regression intercept and tracking error from two return series you paste. Use four hypothetical months: market 1, 3, −2, 2 and asset 2, 5, −3, 4.
- Means: (1 + 3 − 2 + 2) ÷ 4 = 1 and (2 + 5 − 3 + 4) ÷ 4 = 2.
- Deviations: market 0, 2, −3, 1; asset 0, 3, −5, 2. Cross-products: 0 × 0 + 2 × 3 + (−3) × (−5) + 1 × 2 = 23. Market squares: 0² + 2² + (−3)² + 1² = 14.
- Beta: 23 ÷ 14 ≈ 1.6429. Check it with
=SLOPE(B1:B4, A1:A4). - Alpha at 0.25% a month: 2 − (0.25 + 1.6429 × 0.75) ≈ 0.5179.
What to notice: add 1 to the asset’s return in the month where the market deviation is 0 (the first month). Beta does not move; only alpha rises. Then try =RSQ(B1:B4, A1:A4) to see how well the line fits.
Common mistakes
Section titled “Common mistakes”- Reading beta as total risk. Beta measures co-movement only. Check volatility and drawdowns too.
- Ignoring R-squared. Fund L’s 0.375% alpha rests on a line that explains under 20% of its movement.
- Mixing alpha versions. The regression intercept and Jensen’s alpha differ unless beta is 1 or the risk-free rate is 0. Compare like with like.
- Treating a negative beta as a free hedge. It is measured over one window and can flip sign; its expected return may also be low.
- Annualising six months of alpha and calling it a track record. The luck example shows how fragile a short sample is.
Real-world context
Section titled “Real-world context”Stock quote pages often show beta labelled with a window such as “5Y monthly”. Fund fact sheets and portfolio reports often list beta, alpha and R-squared together, usually over three or five years against a named benchmark index. Check the benchmark, period, data frequency and whether returns are before or after fees.
Quick check
Section titled “Quick check”- An asset has a beta of 1.2. The market falls 5% in a month. What move does the market-linked part suggest?
Answer to question 1
About 1.2 × (−5) = −6%. Company-specific news can push the actual result higher or lower.
- A fund returned 9% in a year when the market returned 10% and the risk-free rate was 2%. Its beta is 0.8. What is its Jensen’s alpha?
Answer to question 2
Expected return: 2 + 0.8 × (10 − 2) = 8.4%. Alpha: 9 − 8.4 = 0.6 percentage points. It trailed the market but beat what its beta implied.
- The sum of cross-products is 12 and the sum of squared market deviations is 8. What is beta?
Answer to question 3
12 ÷ 8 = 1.5.
- A fund shows a large positive alpha with an R-squared of 0.15. What should you conclude?
Answer to question 4
Very little. The benchmark explains only 15% of the fund’s movement, so the alpha may reflect a poor benchmark choice or noise rather than skill.
Key takeaways
Section titled “Key takeaways”- Beta = covariance with the benchmark ÷ benchmark variance, which is also the slope of the best-fit line.
- Jensen’s alpha = actual return − (risk-free rate + beta × (benchmark return − risk-free rate)).
- R-squared tells you how much to trust the beta and alpha: low R-squared, weak fit.
- Positive alpha over a short window can come from luck, the chosen period or the chosen benchmark.
- All figures here are hypothetical and backward-looking; this is education, not advice about any security.
Next steps
Section titled “Next steps”- Sharpe and Sortino ratios: return per unit of total or downside risk, the companion to this lesson.
- Risk and diversification: why stock-specific risk can be diversified but market risk cannot.
- Asset allocation basics: how the mix of assets sets a portfolio’s overall market exposure.
Sources
Section titled “Sources”- U.S. Securities and Exchange Commission, Invest Wisely: Advice From Your Securities Industry Regulators (accessed November 19, 2026): past success is no guarantee of future performance.
Key terms
For educational purposes only. This material does not provide investment advice.