3.4 Portfolio Mathematics and Safety-First

Portfolio Mathematics and Safety-First

Portfolio math rests on one insight: investors hold combinations of assets, so the risk and return of the whole portfolio matter more than the stand-alone properties of any single holding. A volatile stock can still improve a portfolio if its returns offset other holdings at the right times.

Expected return is linear; risk is not

Portfolio expected return is a weighted average: E(Rp) = w1 E(R1) + w2 E(R2) + ... + wn E(Rn). Weights normally sum to 1; a short position carries a negative weight, and leverage can push total long weights above 1. Portfolio variance is more involved because returns move together. For two assets:

Var(Rp) = w1^2 sigma1^2 + w2^2 sigma2^2 + 2 w1 w2 Cov(1,2)

Covariance measures co-movement in squared return units; positive covariance raises portfolio variance, negative covariance lowers it. Correlation standardizes covariance to the -1 to +1 range: rho(1,2) = Cov(1,2)/(sigma1 sigma2), so you can substitute Cov(1,2) = rho(1,2) sigma1 sigma2 directly into the variance formula. A correlation of +1 means perfect co-movement and no diversification benefit (portfolio standard deviation equals the weighted average of the two). A correlation below +1 delivers diversification, and a correlation of -1 can be combined in specific weights to drive portfolio risk to zero.

The headline exam idea: expected return is linear in weights, but risk is not, because covariance enters; that nonlinearity is the entire source of diversification.

Building covariance from scenarios

Covariance can be estimated from a scenario table. For each economic state, compute each asset's deviation from its own expected return, multiply the paired deviations, weight by the state's probability, and sum. A positive result means above-average returns for the two assets tend to coincide.

Safety-first and shortfall risk

Roy's safety-first criterion focuses on the chance of failing to meet a minimum acceptable return, RL. The safety-first ratio is:

SFRatio = [E(Rp) - RL] / sigma_p

Under a normality assumption, the portfolio with the highest safety-first ratio has the lowest probability of falling below the threshold, because the ratio equals the number of standard deviations between the expected return and the floor. Notice the parallel to the Sharpe ratio, [E(Rp) - Rf]/sigma_p: safety-first simply replaces the risk-free rate with the investor's minimum acceptable return.

Shortfall probability is estimated with a z-score: z = (RL - E(Rp))/sigma_p. If expected return is 10%, standard deviation is 8%, and the threshold is 2%, then z = (2 - 10)/8 = -1.0, and under normality the shortfall probability is the area below -1.0, about 16%.

Exam procedure

Safety-first is the right lens when avoiding a target breach (spending needs, liability payments, capital preservation) outranks maximizing the average. It does not replace full portfolio analysis because it assumes normality and compresses risk into a single standard deviation. Build every problem in layers to avoid mixing formulas: (1) compute each expected return, (2) compute covariance or correlation as required, (3) compute portfolio variance and take the square root for standard deviation, and (4) apply the safety-first or shortfall formula last.

Keep signs intact on short positions, and remember that the weights inside the variance formula are squared while the cross term carries the factor of two.

Worked two-asset example

Let Asset A have an expected return of 10% and standard deviation of 20%, and Asset B 6% and 12%, with a correlation of 0.30, held at weights 0.60 and 0.40. Expected return is 0.60(10%) + 0.40(6%) = 8.4%. The covariance is 0.30 x 0.20 x 0.12 = 0.0072. Portfolio variance is 0.60^2(0.20^2) + 0.40^2(0.12^2) + 2(0.60)(0.40)(0.0072) = 0.0144 + 0.002304 + 0.003456 = 0.02016, so portfolio standard deviation is sqrt(0.02016) = 14.2%.

Note this is well below the weighted average of the two standard deviations, 0.60(20%) + 0.40(12%) = 16.8%: the 2.6 percentage-point reduction is the diversification benefit, and it exists purely because the correlation is below +1. Had the correlation been +1, the portfolio standard deviation would equal that 16.8% weighted average exactly.

Connecting to later topics

The same machinery reappears in Portfolio Management as the efficient frontier and the capital allocation line, where the safety-first ratio's structure (excess return over standard deviation) becomes the Sharpe ratio. Recognizing that link helps on integrated item sets. For the exam, watch three recurring traps: forgetting to square the weights, dropping the factor of two on the covariance term, and confusing covariance (in squared return units, unbounded) with correlation (unitless, bounded between -1 and +1).

When a question gives correlation rather than covariance, multiply correlation by both standard deviations before inserting it into the variance formula.