Risk-Adjusted Performance

What risk-adjusted performance measures

Risk-adjusted performance relates investment return to a measure of risk. A raw return shows how much an investment gained or lost over a period; it does not show how much variability or market exposure accompanied the result. A risk-adjusted measure can help compare alternatives with different risk profiles, but the answer depends on the selected measure, benchmark, time period, and data conventions. No single ratio proves that an investment is suitable or that a manager has skill.

Begin with the decision. If the question is how much excess return was earned per unit of total variability, the Sharpe ratio is a common measure: (portfolio return − risk-free rate) ÷ standard deviation of portfolio returns. If the question is excess return relative to systematic market exposure, Treynor ratio uses beta in the denominator. Jensen’s alpha compares actual return with a model-implied return, often using CAPM. The ratios are related but answer different questions; do not choose a denominator merely because the calculation looks familiar.

The Sharpe ratio uses standard deviation, which measures dispersion of returns around their mean. Its numerator is the portfolio’s excess return over a chosen risk-free return for the same period. Suppose a portfolio earns 8%, a matching-period risk-free rate is 2%, and the portfolio’s annualized standard deviation is 12%. The Sharpe ratio is (8% − 2%) ÷ 12% = 0.50. It means 0.50 units of excess return per unit of measured total variability under those conventions, not a 50% probability of profit.

Units and periods must match. Do not subtract an annual risk-free rate from a monthly portfolio return and divide by a standard deviation measured over a different horizon. If monthly data are annualized, use a consistent method and disclose the observation frequency. Standard deviation annualization often uses a square-root-of-time convention that assumes stable conditions; it is a modeling convention, not a guarantee that observed risk behaves that way.

The Treynor ratio replaces standard deviation with beta: (portfolio return − risk-free rate) ÷ portfolio beta. Beta estimates sensitivity to a chosen market benchmark. Suppose the portfolio earns 9%, the corresponding risk-free rate is 3%, and beta is 1.2. Treynor ratio is 6% ÷ 1.2 = 5 percentage points of excess return per beta unit. The metric focuses on systematic market exposure and may be less informative for a concentrated or poorly diversified portfolio whose specific risk is substantial.

Compare the common measures

Jensen’s alpha asks whether the portfolio return exceeded or fell short of the return predicted by a selected model. Under a simple CAPM setup, expected return is risk-free rate + beta × (market return − risk-free rate). If the risk-free rate is 2%, market return is 8%, and portfolio beta is 1.1, the model-implied return is 2% + 1.1 × 6% = 8.6%. If actual return is 9.4%, the estimated alpha is 0.8 percentage points. This result is relative to the chosen model and inputs; it is not proof of persistent skill.

A downside-sensitive measure can answer a different investor question. The Sortino ratio replaces total standard deviation with downside deviation relative to a chosen target or minimum acceptable return. It treats upside volatility differently from harmful shortfall. An investor focused on withdrawals may care about falling below a spending threshold. The target must be stated, and downside deviation depends on the observations and method selected. Do not compare Sortino results that use different targets as if their denominators were identical.

Worked comparison: Fund A earns 10% with a 15% standard deviation and a 2% risk-free rate. Its Sharpe ratio is 8 ÷ 15 = 0.53. Fund B earns 8% with a 9% standard deviation under the same assumptions. Its ratio is 6 ÷ 9 = 0.67. B has the higher historical excess-return-per-unit-of-total-variability measure, even though its raw return is lower. That does not by itself establish that B is better for a client; the period, holdings, fees, benchmark, and future risk still matter.

The risk-free input is also an assumption. A short-term government yield may be used as a proxy, but the maturity and period should correspond to the return observations. Using a risk-free rate that is too high or too low changes the numerator and can change rankings. A displayed ratio without the benchmark and risk-free convention is difficult to interpret or reproduce.

Check assumptions and data quality

Risk-adjusted measures inherit weaknesses in their data. Standard deviation uses historical observations and can understate risks that did not occur in the sample. Beta can change as holdings, leverage, or market conditions change. A benchmark may not capture the portfolio’s actual risks. Private or illiquid assets can have infrequent or smoothed valuations that make measured volatility appear lower than economic risk. Treat a calculated ratio as a summary of selected data, not as a complete risk inventory.

A ratio can also be distorted by the observation period. A short favorable interval may produce a striking Sharpe ratio that disappears when the market regime changes. A crisis period may dominate the standard deviation. Compare equal and relevant periods, and inspect the underlying return series where practical. Annualized figures can make a short sample look comparable to a long record, but the uncertainty in the estimate does not disappear through annualization.

Fees alter performance measures. If one fund’s return is net of expenses and another’s is gross, their Sharpe ratios are not directly comparable. The risk-free rate and dispersion should be calculated from returns on the same basis. A low-cost portfolio may deliver a lower gross excess return but a better client outcome after expenses. State whether the returns include advisory charges, fund expenses, trading costs, and reinvested distributions, as appropriate to the comparison.

Apply the measure to a client decision

A ratio is not a suitability test. A high historical Sharpe ratio does not mean the investment is appropriate for a client who needs immediate liquidity, cannot tolerate loss, or already has concentrated exposure. The client’s time horizon and ability to bear risk remain central. Nor does a lower ratio necessarily make an investment unsuitable: a diversifying holding can have a role even if its stand-alone risk-adjusted measure appears less attractive.

Worked application: a client compares a broad equity fund with a short-term bond fund. Their Sharpe ratios are 0.55 and 0.70 over the same five-year period. The bond fund’s higher ratio may reflect lower volatility and a positive excess return, but the figures do not show expected return over the client’s horizon or how either fund behaves in the client’s full portfolio. The adviser should examine portfolio context, correlation, withdrawal needs, expenses, and downside scenarios before recommending either holding.

If two measures disagree, explain what they emphasize. A portfolio may have a strong Treynor ratio because it earned excess return per unit of beta while having considerable company-specific volatility, producing a weaker Sharpe ratio. A portfolio with downside protection may score differently under Sortino than Sharpe. This is not necessarily a calculation error. The measures use distinct definitions of risk; identify the relevant one and discuss the risk omitted by the chosen denominator.

A practical review starts with the return series and dates, then checks whether income is included and whether returns are gross or net. Match the risk-free rate and benchmark period. Select the measure that fits the question, calculate with consistent units, and describe the denominator in plain words. Finally, review holdings, leverage, liquidity, concentration, and the client’s objective. If the input data or benchmark is weak, say so rather than presenting a precise ratio as definitive.

A repeatable calculation review

For exam calculations, preserve the arithmetic. Convert the inputs to consistent units, subtract the risk-free rate before dividing for Sharpe or Treynor, and use the stated beta or standard deviation. For Jensen alpha, calculate the model-implied return first, then subtract it from actual return. Keep percentages as percentage points or convert all values to decimals consistently. Interpret the answer as a historical or model-based comparison under the question’s assumptions.

Risk-adjusted performance is most useful as one part of a disciplined comparison. It can make raw return less misleading by relating performance to a stated risk proxy. It cannot eliminate estimation error, forecast future returns, or replace client-specific analysis. A clear report identifies the formula, period, benchmark, risk-free proxy, fee basis, and material data limits so the reader can understand both the calculation and its boundaries.

Look at the underlying return path

Two portfolios can have the same average return and standard deviation while reaching those summary statistics through very different paths. One may have frequent small changes; another may include a sudden large loss followed by a recovery. A Sharpe ratio compresses that history into a quotient and does not tell the client when losses occurred or whether a withdrawal was due during the decline. Review drawdowns and liquidity alongside the ratio when those details matter to the decision.

The benchmark affects beta and alpha. A growth-oriented fund compared with a broad market index may appear to have unexplained alpha if the benchmark does not capture its style exposure. A small-cap portfolio compared with a large-cap index may also have a misleading beta. Select a benchmark that is relevant to the mandate and document the choice. If the comparison changes after a more appropriate index is used, explain that the original result reflected a benchmark mismatch rather than a sudden change in the portfolio’s underlying return history.

Diversification also changes how a ratio should be used. An asset with a low stand-alone Sharpe ratio may reduce overall portfolio volatility when its returns do not move closely with the rest of the holdings. Conversely, a holding that looks attractive alone can add concentration or liquidity risk to a client’s existing exposure. Evaluate the total portfolio, not just the stand-alone score. The client’s other assets, debts, spending needs, and restrictions can change which risks are material.

A fair side-by-side review should specify whether returns are time-weighted, how external flows are handled, whether dividends are reinvested, and whether reported performance is net of all relevant costs. A fund with a return series that omits transaction costs may show a higher ratio than the client would experience. If the records use different start dates or valuation frequency, do not force a numeric ranking without first addressing that data problem. Sometimes the defensible conclusion is that the evidence is insufficient for a clean comparison.

A ratio is also not a forecast of how the portfolio will behave in a new market regime. A strategy that benefited from falling interest rates may have a strong historical result and a different sensitivity when rates rise. Before extending a historical ratio into a recommendation, ask which holdings or economic conditions drove the record and whether those conditions are likely to persist. Keep the historical calculation separate from any forward-looking scenario or capital market assumption.