Treynor Ratio

The formula in one line

Treynor ratio  =  (R_p − R_f)  /  β_p

Where R_p is the portfolio's return, R_f the risk-free rate, and β_p the portfolio's beta — the slope coefficient from a regression of portfolio excess returns on market excess returns. The numerator is identical to the Sharpe ratio's. The denominator is what changes: β instead of σ.

Why divide by beta instead of sigma

Sharpe and Treynor split on a single question: what risk does the market actually compensate? CAPM's answer: only systematic risk. Idiosyncratic risk vanishes in diversification and earns no premium. If you're going to combine this fund with thirty others, only the part of its volatility that's correlated with the rest of your book actually contributes to portfolio risk.

For an undiversified fund holding 10 concentrated bets, total σ is dominated by firm-specific vol. Sharpe (which uses σ) punishes that volatility heavily. Treynor (which uses β) ignores it — because in the institutional investor's full book, that firm-specific noise will average out. The right comparison depends on context:

Context	Right metric	Why
Whole portfolio evaluation (retiree's total nest egg)	Sharpe	No further diversification will happen; total σ is what the investor lives through
Single fund to add to an existing diversified portfolio	Treynor	Idiosyncratic vol will average out at the broader-portfolio level
Pension-fund manager evaluation	Treynor	Pension funds hold dozens of managers; each is a component, not the whole
Hedge-fund seeding (single fund)	Either, with caution	Non-linear payoffs make β unstable; check both metrics

Worked example: Treynor for three funds

Three actively managed equity funds over 5 years. Market return 10%, risk-free 3%, market vol 16%.

Fund	R_p	β	σ	Sharpe	Treynor
Diversified growth (40 stocks)	12%	1.20	18%	(12−3)/18 = 0.50	(12−3)/1.20 = 7.5%
Concentrated biotech (5 stocks)	14%	0.80	40%	(14−3)/40 = 0.275	(14−3)/0.80 = 13.75%
Index fund (S&P 500)	10%	1.00	16%	(10−3)/16 = 0.44	(10−3)/1.00 = 7.0%

Notice the disagreement. Sharpe says the diversified growth fund is best (0.50), then the index (0.44), then biotech (0.275). Treynor says the biotech is best (13.75%), then growth (7.5%), then index (7.0%). The biotech's high total σ (40%) crushes its Sharpe — but its β is just 0.80, so Treynor says its market-correlated risk is low.

Which ranking is right depends on what you do with the fund. If biotech is your entire portfolio, you'll experience its 40% volatility — Sharpe is right to penalize it. If biotech is one of 30 funds in a diversified institutional book, its 40% firm-specific volatility will mostly diversify away — Treynor's ranking better reflects the marginal contribution to portfolio risk.

Jack Treynor and the pre-Sharpe CAPM

Jack Treynor finished a Harvard MBA in 1955 and joined Arthur D. Little. In 1962 he circulated an unpublished manuscript, "Toward a Theory of Market Value of Risky Assets," containing an early CAPM derivation — predating Sharpe's 1964 published paper. Modigliani read it. Sharpe read it. But Treynor never submitted his version to a journal. The Treynor ratio paper "How to Rate Management of Investment Funds" appeared in the Harvard Business Review in January-February 1965, six months before Sharpe's mutual-fund Sharpe-ratio paper.

The historiography of CAPM is complicated. Sharpe, Lintner, Mossin, and Treynor all derived equilibrium pricing under mean-variance preferences in 1964–1966. Treynor never received the Nobel — Sharpe did, in 1990. Treynor's ratio remains the workaday metric for institutional portfolio construction, especially in fund-of-funds and pension-allocation contexts where Sharpe overstates risk for components.

Treynor vs Sharpe vs Jensen — when to use which

Metric	Risk measure	Sign of skill	Best for
Sharpe	σ (total)	Higher slope vs R_f baseline	Stand-alone portfolios; total wealth evaluation
Treynor	β (systematic)	Higher excess return per β unit	Funds inside a diversified portfolio
Jensen's α	(absolute, R_p − [R_f + β·ERP])	Positive number above CAPM line	Manager skill identification; cross-sectional ranking
Information Ratio	Tracking error	Higher α per unit of tracking error	Benchmark-tied active management
Sortino	Downside deviation	Higher slope vs target return	Asymmetric / option-like strategies
Appraisal ratio	α / σ(ε) (idiosyncratic vol)	Higher α per residual vol	Stock-picking skill isolation

The Treynor ratio is the line from (0, R_f) to (β_p, R_p) in beta-return space, just as Sharpe is the line from (0, R_f) to (σ_p, R_p) in vol-return space. Plotted on the Security Market Line (CAPM's β-return space), every fund's Treynor ratio is its slope above R_f. CAPM equilibrium predicts all funds plot on a single line — the SML — with identical Treynor ratios. Funds above the line are alpha-generators with higher Treynor than the market.

Implementation in 10 lines of Python

import numpy as np
import statsmodels.api as sm

def treynor_ratio(portfolio_returns, market_returns, rf_per_period, periods_per_year=252):
    """portfolio_returns, market_returns: arrays at same frequency
       rf_per_period: scalar or array
       Returns: annualized Treynor ratio."""
    p_ex = np.asarray(portfolio_returns) - rf_per_period
    m_ex = np.asarray(market_returns) - rf_per_period
    # OLS regression for beta
    X = sm.add_constant(m_ex)
    beta = sm.OLS(p_ex, X).fit().params[1]
    if beta == 0:
        return np.nan
    mean_excess = p_ex.mean()
    return mean_excess * periods_per_year / beta

One implementation gotcha: the annualization differs from Sharpe's. Sharpe multiplies by √T (because vol scales with √T). Treynor multiplies by T directly (because returns scale linearly with T, and beta is unit-less). A daily Treynor of 0.0003 annualizes to 0.0003 × 252 = 0.076 — meaning 7.6% excess return per unit of β.

Where Treynor breaks down

• Negative beta strategies. A gold-mining fund with β = −0.3 and positive R − R_f produces a negative Treynor ratio, which the formula treats as bad. But a negative-beta fund with positive excess return is excellent insurance — its CAPM-predicted return is below R_f. Use Jensen's alpha here.
• Unstable beta. Hedge funds with non-linear payoffs (vol arbitrage, merger arbitrage) have betas that swing widely with market regime. A 2-year sample beta of 0.4 can become 1.8 during a crisis. Treynor computed on one window often misranks during stress.
• Concentrated portfolios. When β doesn't capture the relevant risk (high firm-specific vol, jump risk), Treynor systematically understates risk. Hedge funds in single-stock concentration positions or distressed-debt portfolios are common offenders.
• Roll's critique applies. Like CAPM, Treynor depends on a market proxy. Different proxies (S&P 500, MSCI World, total wealth including human capital) give different betas and thus different Treynor rankings.
• Survivorship bias. Hedge funds that blow up disappear from databases; reported cross-sectional Treynor distributions are upward-biased by 0.5–1.5 percentage points in academic studies.

Real-world applications

• Pension-fund manager selection. Yale Endowment, CalPERS, Norway sovereign wealth all rank candidate equity managers by Treynor over 5-year windows. The implicit assumption: the new manager joins a portfolio of dozens of others, so idiosyncratic vol diversifies away.
• Fund-of-funds construction. Multi-strategy hedge-fund allocators rank component funds by Treynor (or its closely related appraisal ratio) rather than Sharpe, because the FoF combines low-correlated managers.
• Strategic asset allocation. Mean-variance optimization on asset-class returns implicitly uses Treynor-like reasoning: covariance with the portfolio (β-equivalent) drives weights, not standalone vol.
• Mutual-fund star ratings (older versions). Treynor was historically used alongside Sharpe in Morningstar's earlier ratings; modern Morningstar uses risk-adjusted return measures closer to Sharpe.
• Academic studies. Fund-performance research uses both Sharpe and Treynor as standard ranking measures. Discrepancies between the two diagnose diversification levels.
• Institutional reporting. Bloomberg's Portfolio Risk module reports Treynor alongside Sharpe by default for portfolio-level dashboards on the PORT function.

Common pitfalls

• Using Treynor for a stand-alone portfolio. If the fund is the whole investment, σ matters — not β. Treynor ignores firm-specific risk you'll actually experience.
• Quoting unannualized Treynor. A "Treynor of 0.01" is meaningless without frequency context. Always annualize before quoting.
• Trusting β from a short window. 1-year monthly β is only 12 observations — noise dominates. Use 3+ years of weekly data minimum.
• Negative β confusion. The sign of Treynor flips for β < 0 strategies; use Jensen's α or the appraisal ratio for any negative-β fund.
• Mismatched market proxy. Beta computed against S&P 500 differs from β against Russell 3000 or total-wealth proxies. Be consistent within a comparison set.
• Confusing levered and unlevered β. A 1.5× leveraged S&P fund has β = 1.5 and the same Treynor as unlevered S&P. The number doesn't reward leverage on its own.