Skip to content
Criée

Markowitz’s efficient frontier

A mix’s risk depends on correlations: some portfolios beat others on both counts.

Lesson 89 min

By the end of this lesson

  • Calculate the expected return of a mix of two assets
  • Explain how correlation reduces the risk of a mix
  • Recognise a dominated portfolio and an efficient portfolio
In this lesson2

Mixing two investments gives a return that is the average of the two, but a risk that is often smaller than the average of the risks. Harry Markowitz turned this into a method for choosing a portfolio. What you will learn: calculate the return and risk of a mix of two assets, and recognise efficient mixes, those no other mix beats.

Two dimensions: return and variance

The 1990 Nobel Prize press release sums up Markowitz's contribution, published in 1952 in the article "Portfolio Selection": under certain conditions, choosing a portfolio comes down to balancing two quantities, expected return and variance; and the portfolio's risk depends not only on each asset's variance but also on the covariances between them; what matters for an asset is not its risk in isolation but its contribution to the risk of the whole (Royal Swedish Academy of Sciences, 16/10/1990).

For two assets A and B, with a share w invested in A:

  • Expected return = w × return of A + (1 − w) × return of B: a simple weighted average.
  • Variance = w² × σA² + (1 − w)² × σB² + 2 × w × (1 − w) × ρ × σA × σB, where ρ is the correlation. Risk is the square root of this variance.

Writing RAR_A and RBR_B for the expected returns, σA\sigma_A and σB\sigma_B for the standard deviations, ρ\rho for the correlation, and RpR_p and σp\sigma_p for the mix's expected return and standard deviation:

Rp=w RA+(1−w) RBσp2=w2σA2+(1−w)2σB2+2 w (1−w) ρ σAσBR_p = w \, R_A + (1 - w) \, R_B \qquad \sigma_p^2 = w^2 \sigma_A^2 + (1 - w)^2 \sigma_B^2 + 2 \, w \, (1 - w) \, \rho \, \sigma_A \sigma_B

The third term does all the work: the lower ρ, the smaller it is, and the calmer the mix.

Two fictitious investments, correlation 0.2Fictional

A: expected return 6%, standard deviation 8%. B: expected return 12%, standard deviation 20%. Correlation 0.2.

  1. Half and half, return: 0.5 × 6 + 0.5 × 12 = 9%.
  2. Variance: 0.25 × 64 + 0.25 × 400 + 2 × 0.25 × 0.2 × 8 × 20 = 16 + 100 + 16 = 132.
  3. Risk: √132 = 11.49%. The average of the two risks would have given 14%: the mix gains 2.51 points of calm.
  4. The calmest mix: the share of A that minimises variance is (σB² − ρσAσB) ÷ (σA² + σB² − 2ρσAσB) = (400 − 32) ÷ (64 + 400 − 64) = 368 ÷ 400 = 92%.
  5. This mix returns 0.92 × 6 + 0.08 × 12 = 6.48% for a risk of 7.84%.

Surprise: 92% in A and 8% in B earn more than A alone (6.48% against 6%) and are less risky (7.84% against 8%). Holding A alone is a poor choice for everyone.

Minimum-variance portfolio

The mix of assets with the smallest possible risk. No other mix of the same assets is calmer.

Efficient frontier

The set of portfolios offering the highest expected return for their level of risk. It starts at the minimum-variance portfolio and rises towards the most profitable asset; portfolios below it are dominated.

In plain wordsSplitting a field

Imagine a farmer who splits his field between two crops: one likes rain, the other copes with drought. Whatever the season, one makes up a little for the other, so the total harvest varies less than either crop alone. He can try every split, from all of one to all of the other. Some are poor choices: another split carries the same risk with a better expected harvest. The efficient frontier is the list of splits that no other beats. Except that it rests on estimates: get the seasons wrong and you get the frontier wrong.

With a correlation of + 1, does the half-and-half mix of A and B have a risk below, equal to or above 14%?

The frontier in motion

Vary the correlation: at + 1, the curve is a straight line from A to B; the lower the correlation, the more it bends to the left, towards less risk; at − 1, it touches zero risk. The green line is the efficient frontier, the red dotted line the dominated part.

The efficient frontier of two investments
05101520681012AB

Horizontal: risk (standard deviation, %). Vertical: expected return (%).

Chosen portfolio

9.00% / 11.49%

return / risk

Minimum variance

92.0% in A

6.48% / 7.84%

Verdict

Efficient

no other mix offers more for this risk

Fictitious investments: A, expected return 6%, standard deviation 8%; B, 12% and 20%. Risk of the mix: square root of w²σA² + (1 − w)²σB² + 2w(1 − w)ρσAσB.

The OpenStax textbook shows the same effect on real US shares: a single share with a 51.90% standard deviation falls to 30.43% when mixed half and half with a weakly correlated share (0.35), but only to 49.11% with a highly correlated one (0.87) (OpenStax, 15.2).

The frontier is computed from expected returns, standard deviations and correlations that must be estimated, often from the past: an error in these figures shifts the whole curve. The Nobel press release also notes that, because returns are correlated in practice, risk cannot in general be eliminated entirely, however many securities are held. The Investing section calls what remains systematic risk. Choosing a point on the frontier depends on your horizon and risk tolerance: this calculation is not investment advice.

From memory: write the variance of a mix of two assets, and say which term depends on the correlation.

Your turn

30% in A (return 5%) and 70% in B (return 10%). The mix's expected return is...

130% in A (return 5%) and 70% in B (return 10%). The mix's expected return is...

A mix's risk is lowest when the correlation between the two assets is...

2A mix's risk is lowest when the correlation between the two assets is...

A portfolio is dominated when another mix offers...

3A portfolio is dominated when another mix offers...

According to the 1990 Nobel press release, the risk of an asset that matters for Markowitz is...

4According to the 1990 Nobel press release, the risk of an asset that matters for Markowitz is...

Read next: Académie royale des sciences de Suède, communiqué du prix 1990 (Markowitz, Miller, Sharpe)

A question? Ask your teacher, or an AI assistant, giving it the title of this lesson.

Share:WhatsAppLinkedIn

Going further