Skip to content
CyclepediaThe trading encyclopedia
ITENESPT-BRFRDE
Learning path Gold Professional operator

Market beta

Beta estimates the linear sensitivity of an asset's or portfolio's returns to a stated benchmark. It depends on the data, window and method and does not by itself measure total risk, causality or explained variation.

On this page

Who this page is for — Anyone who needs to quantify how the returns of a security, strategy or portfolio have moved with a benchmark, without treating a historical estimate as a permanent property.

Market beta is an estimated linear sensitivity. In its common form, it is the slope coefficient from regressing an asset's returns on the returns of a benchmark:

Rᵢ = α + β × Rᵦ + ε

When the regression includes an intercept, the same estimate can be written:

β = Cov(Rᵢ, Rᵦ) / Var(Rᵦ)

Rᵢ is the asset return, Rᵦ the benchmark return, α the intercept and ε the component not described by the linear relationship. Beta is dimensionless. An estimated beta of 1.2 means that, in the sample and specification used, a 1% benchmark change was associated on average with an approximately 1.2% asset change in the same direction. It does not guarantee that response in the next period.

Three linear sensitivities, three different coordinates Beta estimates a return slope; delta and DV01 are local value sensitivities Three linear sensitivities, three different coordinates Beta estimates a return slope; delta and DV01 are local value sensitivities estimated relation or change ≈ coefficient × factor movebenchmarkMarket betaunderlyingOption deltayield +1 bpDV01 Shared limitation A linear model: sample, large shocks, curvature and regime changes need morelenses. Cyclepedia · source-checked conceptual map
Three linear sensitivities, three factors and three different units: they cannot be added without a stated transformation.

Coordinates that make beta auditable

A number without coordinates cannot be reproduced. A report should state at least:

Coordinate Why it matters
Benchmark beta to a global index can differ from beta to a sector index
Return definition simple or logarithmic, total return or price only, hedged or unhedged currency
Frequency daily, weekly and monthly observations can produce different estimates
Window one year, three years or a selected regime contain different information
Currency and timing non-synchronous markets and FX conversion can alter covariance and variance
Method OLS, weights, outlier treatment and missing data affect the result
Estimation date beta changes as observations enter or the portfolio changes

The benchmark is not necessarily “the market” in an absolute sense. It is the variable chosen by the analyst. A multi-asset portfolio may require several factors: equities, rates, credit, currency or volatility. One beta compresses every other dependency into the residual.

Beta, correlation and R² are different

Correlation normalises covariance by the volatility of both series and ranges from −1 to +1. Beta also contains their volatility ratio:

β = ρᵢ,ᵦ × σᵢ / σᵦ

An asset can therefore have high correlation and beta below one, or moderate correlation and high beta. R² measures the share of sample variability accounted for by the regression; beta measures the slope. Saying that “beta explains half the move” confuses the two quantities. A beta of one can coexist with low R² when the residual is large.

The sign also needs context. Negative beta records an inverse historical linear relationship to the selected benchmark, not a certain hedge. A value near zero only rules out a strong sample linear relationship: nonlinear, conditional and stress exposures may remain.

From one position to a portfolio

For linear positions measured in the same currency against the same benchmark, beta-adjusted exposure can be approximated as:

Eᵦ = market value × β

Portfolio beta can then be computed as a signed-market-value-weighted average relative to the chosen reference value. This does not permit equity beta, option delta and DV01 to be placed in one additive column: they describe different shocks and units. An option first needs a coherent delta; a bond position needs duration or DV01.

Estimates also change with leverage, composition, regime and horizon. A rebalanced ETF, dynamic strategy or options portfolio does not necessarily have constant beta. Nonlinear payoffs require recalculation after shocks and full scenario valuations alongside the local measure.

What beta does not measure

  • maximum or expected loss;
  • issuer-specific, liquidity or counterparty risk automatically;
  • causality between the benchmark and the asset;
  • a curved, asymmetric or regime-dependent relationship well;
  • a certain forecast or a judgment on investment quality;
  • concentration, stress outcomes or other factor exposures.

Beta belongs beside exposure, gross and net views, dependencies and scenarios. It can help explain a first- order market component without becoming a complete risk model.

Common mistake — Calling an asset “high beta” as though it were immutable. The estimate belongs to a benchmark, sample and method; change one coordinate and the number may change as well.

Working checklist

  1. Define the asset, quantity, currency and benchmark.
  2. Align return calendar, time and frequency.
  3. State the window, return definition and estimation method.
  4. Report beta with its estimation uncertainty, R² and observation period.
  5. Keep beta-adjusted exposure separate from market value and loss risk.
  6. Test stability across alternative windows and regimes.
  7. Use scenarios when payoffs or dependencies are nonlinear.

Sources

Suggest an edit My proposals