Treynor Ratio: Excess Return, Estimated Beta, and Comparable Risk Units

Calculate and audit ex-post or ex-ante Treynor ratios with matched total returns, risk-free rates, market-proxy regressions, beta uncertainty, explicit annualization, and disciplined treatment of near-zero or negative beta.

For educational purposes only; not investment advice. Investing may result in loss.

Direct answer

The Treynor ratio divides a portfolio’s average or expected excess return by its estimated sensitivity to a stated market proxy:

Treynor ratio = portfolio excess return ÷ portfolio beta

For ex-post periodic data, portfolio excess return = mean(R_P,t − R_f,t). For an ex-ante calculation, both the numerator and beta must be forward-looking estimates for the same horizon and portfolio. The ratio uses beta as the systematic-risk denominator, so it is most defensible for a sufficiently diversified portfolio and a relevant market proxy. It has return units: a monthly ratio and an annual ratio are numerically different conventions.

A higher ratio for two genuinely comparable, positive-beta portfolios means more measured excess return per unit of estimated market exposure. It does not prove skill, forecast return, or capture total risk. At beta = 0 the ratio is undefined; near zero it is unstable; with negative beta, signs and rankings lose the usual positive-beta interpretation.

Seven-step reproducible workflow

Freeze the evaluation claim. Identify portfolio, account, composite or share class; mandate and diversification; decision being supported; currency; gross, net or after-tax basis; sample or forecast horizon; and whether the result is ex post or ex ante. Compare only portfolios for which the same beta-based question is economically relevant.
Build matched return inputs. Use synchronized total returns with the same dates, frequency, time zone, distributions, corporate actions, external-cash-flow treatment and currency. Match each portfolio observation with a risk-free return for the same currency and period. A quoted annual Treasury rate is not automatically a realized one-month risk-free return; tenor, yield convention and holding-period treatment matter.
Choose the market proxy and estimate beta. Define y_t = R_P,t − R_f,t and x_t = R_M,t − R_f,t, then fit y_t = alpha + beta × x_t + epsilon_t. With an intercept, beta = Σ[(x_t − x̄)(y_t − ȳ)] ÷ Σ(x_t − x̄)². The policy benchmark and CAPM market proxy may differ. Because the theoretical market portfolio is unobservable, performance conclusions are joint with proxy and model choice.
Audit the regression rather than accepting beta as fixed. Report sample dates, observations, beta standard error and confidence interval, alpha, residuals, R-squared, outliers and stability across rolling windows, frequencies and sensible proxies. Least squares can be sensitive to outliers; stale prices, nonsynchronous trading, leverage, derivatives, nonlinear payoffs and changing holdings can distort a constant linear beta.
Construct a matched numerator. For ex post, use the arithmetic mean of the same periodic portfolio excess returns used in the beta regression unless another fully specified convention is intended. For ex ante, use expected portfolio return minus a horizon-matched expected risk-free return and pair it with forecast beta. Do not divide historical excess return by forecast beta, or forecast return by historical beta, and label gross and net returns separately.
Compute, annualize and handle the denominator. Calculate T_periodic = mean portfolio excess return ÷ beta. Beta is dimensionless and is not multiplied by the number of periods. Under an arithmetic convention, annualized excess return may be q × periodic mean excess return; under a geometric convention, compound portfolio and risk-free wealth separately before subtracting. State units and do not mix these conventions. Treat zero beta as undefined and show sensitivity or decline to rank when beta is near zero or negative.
Interpret with complementary evidence. Sharpe ratio divides excess return by total-return standard deviation and therefore captures residual volatility that Treynor omits. Jensen’s alpha is the fitted regression intercept, not excess return divided by beta. For the same regression and nonzero beta, Treynor = mean market excess return + alpha ÷ beta. Review alpha uncertainty, Sharpe, drawdown, concentration, tail loss, liquidity, leverage, costs, taxes, capacity and the portfolio’s hedge role before reaching a conclusion.

An ex-post Treynor ratio is a sample statistic conditional on realized returns, a risk-free proxy, a market proxy, a frequency and an estimated beta. An ex-ante ratio is a forecast conditional on expected returns and a risk model. Neither is a universal score, and the CAPM assumption that only systematic risk is priced does not make omitted credit, duration, liquidity, volatility, currency or nonlinear risks disappear.

Worked examples

Period matching and annualization. A portfolio earns 3.0000% in a quarter, the matched quarterly risk-free return is 1.0000%, and beta is 0.8000. Quarterly Treynor ratio is (3.0000% − 1.0000%) ÷ 0.8000 = 2.5000%. Under an arithmetic annualization, (2.0000% × 4) ÷ 0.8000 = 10.0000%. If those quarterly returns were constant and compounded separately, portfolio annual return would be 1.03⁴ − 1 = 12.5509%, risk-free annual return 1.01⁴ − 1 = 4.0604%, and geometric annual Treynor ratio (12.5509% − 4.0604%) ÷ 0.8000 = 10.6131%. Subtracting an annual rate directly from a quarterly return would be period mismatch.
Five-period beta regression. Market excess returns are −2.0000%, −1.0000%, 0.0000%, +1.0000%, +2.0000%; portfolio excess returns are −2.1000%, −1.2000%, +0.4000%, +1.2000%, +2.7000%. Their means are x̄ = 0.0000% and ȳ = 0.2000%; Σ[(x_t − x̄)(y_t − ȳ)] = 12.0000 percentage-points squared and Σ(x_t − x̄)² = 10.0000 percentage-points squared, so beta = 12.0000 ÷ 10.0000 = 1.2000. The intercept is alpha = 0.2000%, residuals are +0.1000%, −0.2000%, +0.2000%, −0.2000%, +0.1000%, and R-squared = 1 − 0.1400 ÷ 14.5400 = 0.9904. Periodic Treynor ratio is 0.2000% ÷ 1.2000 = 0.1667%; high fit does not eliminate beta or alpha uncertainty in five observations.
Near-zero and negative beta. With excess return 2.0000%, beta 0.0500 gives 2.0000% ÷ 0.0500 = 40.0000%, while beta 0.0200 gives 2.0000% ÷ 0.0200 = 100.0000%; at 0.0000 the ratio is undefined. Separately, excess return −2.0000% divided by beta −0.2000 equals +10.0000%. The positive quotient does not turn a loss into superior standalone performance; a negative-beta position may be insurance whose value depends on stress behavior and portfolio context.
Fees and portfolio aggregation. A portfolio holds 60.0000% in sleeve A and 40.0000% in sleeve B. A has gross return 10.0000% and beta 1.2000; B has gross return 6.0000% and beta 0.3000; the matched risk-free return is 3.0000%. Under the stated linear, fully invested assumptions, portfolio gross return is 60.0000% × 10.0000% + 40.0000% × 6.0000% = 8.4000% and portfolio beta is 60.0000% × 1.2000 + 40.0000% × 0.3000 = 0.8400, so gross Treynor ratio is (8.4000% − 3.0000%) ÷ 0.8400 = 6.4286%. A 0.8000 percentage-point fee drag gives net return 7.6000% and net ratio (7.6000% − 3.0000%) ÷ 0.8400 = 5.4762%. The weighted average of sleeve ratios is not the portfolio ratio because division is nonlinear.

Calculation and evidence checklist

Identify the portfolio, account, composite, share class, mandate and measured investor experience.
Confirm that a beta-based comparison is appropriate for the portfolio’s diversification and objective.
Match portfolio, market and risk-free total returns by date, frequency, currency and time zone.
State gross, net-of-product-expense, net-of-advisory-fee and after-tax return bases separately.
Match the risk-free proxy’s currency, period, tenor, yield convention and holding-period treatment.
Distinguish the portfolio’s policy benchmark from the market proxy used to estimate CAPM beta.
Preserve price, total-return, distribution, corporate-action and external-cash-flow consistency.
Estimate beta from synchronized portfolio and market excess returns with an intercept.
Report observations, regression window, frequency, alpha, beta, standard error, confidence interval and R-squared.
Inspect residuals, outliers, heteroskedasticity, autocorrelation and nonsynchronous pricing.
Test rolling windows, subperiods and sensible alternative market proxies for beta stability.
Review leverage, derivatives, options, stale marks and nonlinear or regime-dependent exposures.
Keep historical ex-post inputs separate from forward-looking ex-ante return and beta forecasts.
State arithmetic or geometric numerator convention and annualization in explicit return units.
Do not annualize beta or mix an annual risk-free rate directly with a shorter-period return.
Treat zero beta as undefined and near-zero beta as numerically unstable.
Do not rank negative-beta portfolios by the ordinary positive-beta higher-is-better rule.
Recalculate portfolio return and beta from positions rather than averaging sleeve ratios.
Compare Treynor with Sharpe, Jensen’s alpha, drawdown, tail risk, liquidity, fees, taxes and capacity.
Archive data vintages, proxy identifiers, formulas, code, rounding and independent recomputation.

Common misconceptions

“Treynor ratio is an annual return forecast.” It is a conditional risk-adjusted statistic with return units, not a promised return.
“Beta is a fixed, complete measure of risk.” It is an estimated linear loading on a chosen proxy and omits residual and nonlinear risks.
“A very high ratio with low beta proves skill.” A small or unstable denominator can dominate the result and make ranking meaningless.
“Treynor, Sharpe and Jensen’s alpha are interchangeable.” They use beta, total volatility and a regression intercept, respectively.
“Portfolio Treynor is the weighted average of sleeve Treynor ratios.” Portfolio returns and beta aggregate first; the ratio is calculated afterward.

Authoritative sources

Treynor, How to Rate Management of Investment Funds - characteristic-line and fund-performance framework in the Wiley republication of Treynor’s work.
Sharpe, Capital Asset Prices - equilibrium relationship among expected return, systematic risk and the market portfolio.
CFA Institute, Measures of Risk-Adjusted Return - Treynor and Jensen definitions, beta interpretation, diversification and limitations.
GIPS Standards Handbook for Firms - consistent input data, total returns, benchmarks and fair performance presentation.
U.S. Treasury Daily Par Yield Curve Rates - official daily par-yield observations and maturity labels; a yield still requires period-matched return treatment.
Federal Reserve H.15 Selected Interest Rates - selected rate series, frequencies, units, conventions and source notes.
NIST Linear Least Squares Regression - least-squares model form, parameter estimation, intervals, strengths and outlier sensitivity.
NIST Residual Analysis - model validation, limits of R-squared alone and graphical residual diagnostics.