Roy's Safety-First Ratio

Builds onSharpe Ratio — if this page feels steep, start there.

SF = \frac{R_p - R_L}{\sigma_p}

$SF$: the safety-first ratio: how many 'units of wobble' separate you from disaster
$R_p$: the portfolio's expected return
$R_L$: the threshold (L = 'level'): the return below which the investor is in real trouble
$R_p - R_L$: the cushion: expected return above the survival floor
$\sigma_p$: the volatility — each unit of sigma is one 'typical-sized' surprise

Why it must be true

Not every investor benchmarks against the risk-free rate — many have a hard floor: the return below which they're in trouble (a pension's funding requirement, a bank's capital minimum). Roy's criterion asks: how many standard deviations of cushion sit between the expected return and disaster?

It is structurally the Sharpe ratio with the risk-free rate swapped for the threshold $R_L$ . The portfolio with the highest safety-first ratio has the smallest probability of breaching the floor (under normality) — pick that one.

The derivation

The probability of falling below the threshold $R_L$ depends on how far $R_L$ sits below the mean, measured in standard deviations:

P(R < R_L) = P\!\left(Z < \frac{R_L - R_p}{\sigma_p}\right)

Minimizing that probability means maximizing the standardized distance:

SF = \frac{R_p - R_L}{\sigma_p}

When to reach for it

An investor with a hard floor — a pension funding requirement, a covenant, a survival threshold — choosing among portfolios by shortfall risk.

Listen for

“cannot tolerate returns below …”“minimum acceptable / threshold return”“probability of falling short”

Back-of-the-envelope

Estimate it in your head first — then the calculator only confirms.

≈
Identical mechanics to Sharpe with the floor swapped in: (return − floor) ÷ σ. If you can estimate Sharpe in your head, you can do this.
≈
Cushion intuition: SF ≈ 1 means disaster is one 'typical wobble' away; SF ≥ 2 is comfortable. Use that scale to sniff-test the answer.

Traps in applying it

✗Using the risk-free rate where the investor's own threshold RL belongs.
✗Reversing the numerator — RL − Rp ranks portfolios backwards.
✗Reading the ratio as a return forecast rather than a ranking device.

Limits & criticisms

The link from the ratio to an actual shortfall probability requires normality — with skewed or fat-tailed returns the ranking can mislead. It uses only mean and σ, and the threshold itself is a judgement call: two trustees with different floors can rank the same portfolios oppositely, both correctly.

Where it came from

A. D. Roy published "Safety First and the Holding of Assets" in 1952 — the same year as Markowitz's "Portfolio Selection." Markowitz himself later wrote that Roy deserved an equal share of the credit for portfolio theory. Roy's angle was different: real investors fear disaster (falling below a survival threshold) more than variance itself.

Today his criterion drives shortfall-risk analysis for pensions and endowments — any investor with a hard floor (a funding requirement, a covenant) is a safety-first investor whether they know it or not.

Drill this formula Back to the library