Efficient Frontier

The efficient frontier is the set of feasible portfolios with the highest modeled expected return for each risk level under stated inputs and constraints.

The efficient frontier is the set of feasible portfolios that offers the highest modeled expected return for each level of modeled risk, or equivalently the lowest modeled risk for each expected return. A frontier is valid only for its stated asset universe, expected returns, covariance estimates, constraints, currency, horizon, and risk measure.

The frontier is not a list of portfolios guaranteed to perform well. It is a boundary produced by a model. Change the inputs or constraints and the boundary changes.

Key Takeaways

  • The feasible set contains every portfolio permitted by the model and constraints.
  • The efficient frontier contains only nondominated portfolios on the upper minimum-variance boundary.
  • The global minimum-variance portfolio is the lowest-risk point; the lower branch below it is inefficient when higher expected return is preferred.
  • A portfolio below the frontier is dominated under the model, not necessarily unsuitable in the real world.
  • Short-sale, leverage, turnover, liquidity, tax, and concentration rules reshape the frontier.
  • Estimation error can make an apparently precise frontier unstable and difficult to implement.

Risk-Return Space

In a standard mean-variance chart:

  • horizontal axis: portfolio standard deviation (\sigma_p)
  • vertical axis: portfolio expected return (E(R_p))

For weights (\mathbf{w}), expected returns (\boldsymbol{\mu}), and covariance matrix (\boldsymbol{\Sigma}):

$$ E(R_p)=\mathbf{w}^{T}\boldsymbol{\mu} $$
$$ \sigma_p=\sqrt{\mathbf{w}^{T}\boldsymbol{\Sigma}\mathbf{w}} $$

Each permissible weight vector maps to one point. Together, those points form the feasible set.

Constructing the Frontier

For a target expected return (\mu^*), solve:

$$ \min_{\mathbf{w}} \quad \mathbf{w}^{T}\boldsymbol{\Sigma}\mathbf{w} $$

subject to:

$$ \mathbf{w}^{T}\boldsymbol{\mu}=\mu^* \quad\text{and}\quad \mathbf{1}^{T}\mathbf{w}=1 $$

plus practical constraints such as:

$$ 0\leq w_i\leq w_{i,\max} $$

Repeating the optimization across target returns traces the minimum-variance boundary. The efficient portion begins at the global minimum-variance portfolio and proceeds toward higher expected return and higher modeled risk.

Worked Example: Two Risky Assets

Assume:

InputAsset AAsset B
Expected return8%4%
Volatility12%6%
Pairwise correlation0.250.25

For long-only portfolios with weights summing to 100%, selected combinations are:

Weight in AWeight in BExpected returnVolatility
0%100%4.0%6.00%
12.5%87.5%4.5%5.81%
25%75%5.0%6.00%
50%50%6.0%7.35%
60%40%6.4%8.14%
75%25%7.0%9.49%
100%0%8.0%12.00%

The weight of Asset A in the two-asset global minimum-variance portfolio is:

$$ w_A^{GMV} = \frac{\sigma_B^2-\operatorname{Cov}_{AB}} {\sigma_A^2+\sigma_B^2-2\operatorname{Cov}_{AB}} $$

With (\sigma_A^2=0.0144), (\sigma_B^2=0.0036), and covariance 0.0018:

$$ w_A^{GMV} = \frac{0.0036-0.0018} {0.0144+0.0036-2(0.0018)} = 0.125 $$

The minimum-variance portfolio therefore holds 12.5% in A and 87.5% in B. It has expected return of 4.5% and volatility of approximately 5.81%.

The B-only portfolio has lower expected return (4.0%) and higher risk (6.0%) than the minimum-variance portfolio, so B-only lies on the inefficient lower branch. Portfolios from the minimum-variance point toward higher A weights form the efficient branch in this restricted two-asset example.

Dominance

Portfolio X dominates Portfolio Y under mean-variance assumptions if:

  • X has at least as high expected return and lower risk, or
  • X has higher expected return and no more risk

with at least one strict improvement.

In the example, the 25% A / 75% B portfolio has the same 6.0% volatility as B-only but expected return of 5.0% rather than 4.0%. It therefore dominates B-only under the model.

Dominance depends on the chosen risk measure and constraints. A portfolio can appear dominated in mean-standard-deviation space yet provide liquidity, inflation sensitivity, liability matching, tax characteristics, or tail protection omitted from the model.

Global Minimum-Variance and Tangency Portfolios

Global Minimum-Variance Portfolio

The global minimum-variance portfolio has the lowest modeled variance among all feasible risky portfolios. It does not require expected-return estimates when only variance is minimized, although its construction still depends on covariance and constraints.

Tangency Portfolio

When a risk-free rate is introduced, the tangency portfolio maximizes modeled excess expected return per unit of volatility:

$$ \max_{\mathbf{w}} \quad \frac{\mathbf{w}^{T}\boldsymbol{\mu}-R_f} {\sqrt{\mathbf{w}^{T}\boldsymbol{\Sigma}\mathbf{w}}} $$

The result depends on the risk-free rate, eligible assets, and borrowing, lending, and short-sale assumptions. It should not be called the one best portfolio for every investor.

Efficient Frontier Versus Capital Market Line

FeatureRisky-asset efficient frontierCapital allocation or market line
AssetsRisky assets onlyRisk-free asset plus a selected risky portfolio
Shape in mean-standard deviation spaceCurvedStraight under stated assumptions
Key portfolioGlobal minimum-variance and other efficient risky portfoliosTangency portfolio
Main sensitivityMeans, covariance, and constraintsAlso risk-free borrowing and lending assumptions

Under CAPM assumptions, the capital allocation line through the market portfolio is the Capital Market Line. It is not correct to say that every point on that line is part of the original risky-asset frontier; most points combine a risk-free asset with the tangency portfolio.

How Constraints Change the Frontier

An unconstrained frontier can include concentrated long and short positions. Practical rules can require:

  • long-only weights
  • issuer or asset-class maximums
  • minimum liquidity or cash
  • leverage and gross-exposure limits
  • duration, credit, factor, or currency ranges
  • turnover and transaction-cost penalties
  • tax realization limits
  • prohibited securities or jurisdictions

Tighter constraints generally shrink the feasible set. The constrained frontier can lie below or to the right of the unconstrained frontier, but may produce a portfolio that is more stable, understandable, and implementable.

Input Sensitivity

The frontier can move materially when:

  • expected returns change slightly
  • covariance is estimated from another period
  • correlations rise during market stress
  • a new asset is admitted or removed
  • short selling or leverage is permitted
  • costs and taxes are introduced
  • the base currency or horizon changes
  • a downside or tail-risk measure replaces variance

A chart should therefore state the input date, data frequency, lookback window, return convention, currency, constraints, and whether costs are included.

Common Mistakes

  • Describing the frontier as a forecast of actual portfolio returns.
  • Calling every diversified portfolio efficient.
  • Including the lower minimum-variance branch as efficient when higher return is preferred.
  • Comparing frontiers based on different assets, dates, or constraints.
  • Selecting a portfolio solely because it appears on the frontier.
  • Ignoring estimation error and displaying false precision.
  • Treating the capital market line as identical to the risky-asset frontier.
  • Assuming the unconstrained solution is investable.
  • Forgetting that variance treats gains and losses symmetrically.

Authoritative Context

The Nobel Prize’s 1990 economics prize explanation describes Markowitz’s operational theory of portfolio selection under uncertainty. Federal Reserve research on the inverse covariance matrix defines the risk-return efficiency frontier as portfolios of minimum variance conditional on expected return.

FAQs

Is every portfolio on the minimum-variance boundary efficient?

No. The portion below the global minimum-variance portfolio has lower expected return for risk that can be matched or improved by another feasible portfolio. The upper branch is efficient under the standard preference for more expected return and less variance.

Does a portfolio on the efficient frontier guarantee better performance?

No. The frontier uses estimated expected returns and covariance. Realized returns, risks, costs, taxes, liquidity, and correlations can differ materially.

Why does the efficient frontier change over time?

Expected returns, volatility, covariance, eligible assets, market conditions, and practical constraints change. The frontier is conditional on those inputs rather than permanent.

Educational Use

This article provides general financial education. It is not personalized investment, portfolio-construction, quantitative, tax, legal, accounting, or fiduciary advice.

Browse Investing