Beta explained: what it measures, how it's computed, and why it can mislead

Beta is the slope of the regression line between a stock's returns and the market's returns. A beta of 1.0 means the stock moves one-for-one with the market: when the S&P 500 rises 1%, the stock tends to rise 1%. A beta of 1.5 amplifies market moves by 50%; a beta of 0.5 dampens them.

The metric is simple, well-defined, and widely used. The complications are empirical: the elegant theory that gave us beta — the Capital Asset Pricing Model — made specific predictions about returns that haven't held up, which is why beta's role in modern quant scoring is more nuanced than its textbook presentation suggests.

What the formula actually says

β = Cov(stock returns, market returns) ÷ Var(market returns)

In English: how much the stock and market move together, scaled by how much the market moves on its own. Computed from historical returns — usually three to five years of monthly or weekly data.

Different providers use different lookback windows and frequencies, which is why a stock's “beta” on Yahoo Finance, FMP, and Bloomberg can differ slightly even at the same moment. QScoring uses the beta as reported by FMP, computed against ~5 years of price history — long enough to be statistically stable, short enough to reflect the current business profile.

How to read it

β = 1.0 — the stock is “the market” in volatility terms
β > 1.0 — amplifies market moves; “high beta”
β < 1.0 — dampens market moves; “low beta”
β < 0 — moves opposite the market; rare, often utilities or precious-metals miners

The more interesting question is what to do with that information.

Where the textbook story breaks

The Capital Asset Pricing Model (Sharpe 1964, Lintner 1965) made a specific prediction: high-beta stocks should earn higher returns to compensate investors for the additional volatility. Investors are risk-averse; risk needs a price.

The empirical reality has been very different. High-beta stocks have, on average, delivered worserisk-adjusted returns than low-beta stocks over multi-decade periods. This is the “low-volatility anomaly” — possibly the most well-replicated finding in factor research that contradicts the textbook. Frazzini and Pedersen's 2014 paper “Betting Against Beta” is the canonical modern treatment.

The implication: high beta isn't a free ticket to higher returns. If anything, investors who are forced to lever up by buying high-beta stocks (because they can't use leverage directly) bid those names up to overpriced levels.

How QScoring uses it

Beta is one of two inputs in the QScoring risk factor, alongside 60-day annualized realized volatility. Stocks with beta closer to 1.0 score higher than stocks with extreme betas in either direction. This implements the spirit of the low-vol anomaly — preferring stocks that don't over-amplify market noise — without trying to chase pure low-vol exposure.

See the risk section of the methodology for the full mapping curve and the regime-change weakness inherent in any historically-computed beta.

Real example

High-beta names in our universe tend to be growth-tier semiconductors and cyclical industrials. AMD, TSLA, and CAToften run beta > 1.5. Low-beta names tend to be mature consumer staples and utilities —KO, PG, and JNJ typically run beta < 0.7. Mega-cap tech like MSFT and AAPL sits near 1.0 — they essentially are the index in many ways.

Common mistakes

Treating beta as forward-looking.It isn't. Beta is computed from historical returns and assumes business profile and capital structure stayed roughly stable. A company that just took on huge debt or pivoted into a new business has a backward-looking beta that says nothing about the next year.
Assuming high beta = high return.CAPM says it should; the data says it hasn't. Don't buy high-beta stocks just because you're looking for “more upside.”
Ignoring the standard error.A beta computed from 5 years of monthly data has meaningful confidence intervals. A reported beta of 1.2 might really be “between 0.95 and 1.45 with 95% confidence.” Treat single-decimal beta readings with appropriate skepticism.
Comparing betas across providers without checking methodology. Different lookback windows produce materially different numbers. FMP, Yahoo, and Bloomberg can disagree by 20%+ on the same stock at the same moment.