Roy’s Safety‑First Criterion (SFRatio) is a simple, single‑period risk management rule that ranks investments or portfolios by the probability they will fall below a user‑specified minimum acceptable return (the threshold). Introduced by A. D. Roy in 1952, the criterion directs investors to choose the portfolio that minimizes the probability of returns dropping below that threshold — equivalently, the portfolio that maximizes the SFRatio.
Key takeaways
– SFRatio = (expected portfolio return − minimum required return) / portfolio standard deviation.
– SFRatio is effectively a z‑score: higher SFRatio → lower probability of falling below the minimum return.
– If returns are normally distributed and the minimum required return equals the risk‑free rate, the SFRatio equals the Sharpe ratio.
– Limitations: single‑period view, reliance on mean and standard deviation (sensitive to non‑normal returns and tail risk), and does not capture investor utility for returns above the threshold.
Formula
SFRatio = (r_e − r_m) / σ_p
where
– r_e = expected return on the portfolio (mean return),
– r_m = investor’s minimum required return (threshold),
– σ_p = standard deviation of portfolio returns.
How Roy’s criterion is used (intuition)
– SFRatio is a standardized distance (in standard deviations) between the portfolio’s expected return and the minimum acceptable return.
– A positive SFRatio means the expected return is above the threshold; negative means it’s below.
– The SFRatio converts to the probability of underperforming the threshold using the normal cumulative distribution function (CDF):
P(return < r_m) = Φ((r_m − r_e) / σ_p) = Φ(−SFRatio) = 1 − Φ(SFRatio),
where Φ is the standard normal CDF. For example, an SFRatio of 1.0 implies about a 15.9% chance of falling below the threshold (Φ(−1) ≈ 0.1587).
Step‑by‑step: How to calculate the SFRatio
1. Define the analysis horizon and the minimum required return (r_m).
• r_m can be a “must‑have” nominal return (e.g., minimum cashflow needed), a real/inflation‑adjusted target, or the risk‑free rate depending on your objective.
2. Estimate the portfolio expected return (r_e).
• Use historical average returns, forward‑looking capital market assumptions, or scenario/Monte Carlo averages over the chosen horizon.
3. Estimate the portfolio standard deviation (σ_p).
• For a multi‑asset portfolio, compute σ_p from asset volatilities and the covariance matrix:
σ_p = sqrt(w' Σ w) where w are weights and Σ is the covariance matrix.
• If working with annualized returns from monthly data, ensure consistent annualization.
4. Compute SFRatio = (r_e − r_m) / σ_p.
5. (Optional) Convert SFRatio to probability of underperformance:
• P(return < r_m) = Φ(−SFRatio). Use a standard normal table or software to get Φ.
6. Compare SFRatios across candidate portfolios. The optimal choice (by Roy’s rule) is the portfolio with the highest SFRatio (lowest probability of falling below r_m).
Worked numerical example
Assume three portfolios and an investor threshold r_m = 5%:
– Portfolio A: r_e = 12%, σ_p = 20% → SFRatio = (12% − 5%) / 20% = 7% / 20% = 0.35
• Probability return < 5% ≈ Φ(−0.35) ≈ 36.3%
– Portfolio B: r_e = 10%, σ_p = 10% → SFRatio = (10% − 5%) / 10% = 0.50
• Probability return < 5% ≈ Φ(−0.50) ≈ 30.9%
– Portfolio C: r_e = 8%, σ_p = 5% → SFRatio = (8% − 5%) / 5% = 0.60
• Probability return < 5% ≈ Φ(−0.60) ≈ 27.4%
By Roy’s criterion, Portfolio C is preferred because it has the highest SFRatio (lowest chance of failing to meet the 5% threshold).
Practical steps and best practices for investors
1. Choose the threshold (r_m) that reflects a real constraint (cash needs, liability, survival, minimum spending). Be explicit whether r_m is nominal or real and whether it includes taxes/fees.
2. Use an appropriate data and horizon:
• Match return and volatility estimates to the investment horizon (e.g., monthly returns scaled to annual if appropriate).
3. Use robust return and volatility estimates:
• Combine historical, forward‑looking, and scenario approaches.
• Use shrinkage or factor models for covariance estimation if data are noisy.
4. Account for non‑normal returns and tail risk:
• If returns exhibit skewness or fat tails, the SFRatio’s normal‑approximation probability can be misleading. Consider Monte Carlo simulation, bootstrapping, or modeling with distributions that capture skew/kurtosis.
5. Consider multi‑period adjustments:
• For independent returns over T periods, expected excess scales linearly and volatility scales with sqrt(T): SFRatio_T = (r_e,T − r_m,T) / σ_p,T. But be cautious—assumptions of independence and stationarity matter.
6. Integrate with optimization and constraints:
• Instead of naive selection, maximize SFRatio subject to investment constraints (weights, liquidity, regulations). This is analogous to mean‑variance optimization but with an objective targeting the threshold.
7. Stress test and scenario analysis:
• Check performance under adverse market scenarios and assess sensitivity to r_m and vol estimates.
8. Combine with other risk measures:
• Use SFRatio along with CVaR, drawdown analysis, and utility‑based approaches to form a fuller picture.
When Roy’s SFRatio is appropriate
– When the investor’s primary concern is avoiding outcomes below a clear minimum target (e.g., survival, meeting liabilities).
– When a single‑period, mean–variance style approximation is reasonable and return distributions are not extremely non‑normal.
Limitations and cautions
– Single‑period metric: ignores path dependency and multi‑period consumption needs.
– Mean–variance based: relies only on mean and standard deviation; ignores skewness and tail leptokurtosis.
– Can favor low‑volatility, low‑return portfolios that may have unacceptable long‑term opportunity cost.
– Sensitive to the choice of r_m; small changes in the threshold can change portfolio ranking.
– If returns are not approximately normal, converting SFRatio to a probability is misleading without simulation.
Relation to the Sharpe ratio
– The Sharpe ratio = (r_e − r_f) / σ_p, where r_f is the risk‑free rate.
– If an investor sets the minimum required return r_m equal to the risk‑free rate r_f, then SFRatio and Sharpe ratio are identical numerically. Conceptually, Roy’s criterion frames the choice in terms of avoiding an unacceptable shortfall, whereas the Sharpe ratio is usually presented as a reward‑for‑volatility measure.
Extensions and alternatives
– Use scenario‑based or simulation‑based probabilities to better capture non‑normality.
– Consider downside risk measures like Sortino ratio (focuses on downside deviation) and CVaR (Conditional Value at Risk).
– Multi‑period target‑based measures: goal‑based investing frameworks or utility maximization with downside constraints.
References and further reading
– Roy, A. D. (1952). “Safety First and the Holding of Assets.” The Quarterly Journal of Economics, 66(1), 95–106.
– Investopedia. “Roy’s Safety‑First Criterion (SFRatio).”
Summary
Roy’s Safety‑First Criterion is a compact, intuitive way to compare portfolios when the investor’s primary objective is to minimize the chance of falling below a specified minimum return. It is easy to compute and interpret (a standardized distance from the threshold), but it relies on mean‑variance assumptions and is sensitive to the chosen threshold and to non‑normal return behavior. Use it as a practical screening tool, but complement it with simulations, stress tests, and other risk measures before making final portfolio decisions.