ConvexPi

Portfolio Selection

Harry Markowitz

The Journal of Finance · 1952 · 4528 citations

Factor Zoo
Community wiki✎ Edit⟲ History

Portfolio Selection


Source: Markowitz, H. (1952) · Journal of Finance 7(1), 77–91 · doi:10.2307/2975974


TL;DR

Founds modern portfolio theory. Markowitz rejects "maximize expected (discounted) return" — which

never favors diversification — and replaces it with the expected return–variance of return (E–V)

rule. What matters for a portfolio is not each asset's risk in isolation but its covariance with

the rest; combining imperfectly correlated assets reduces variance for a given return. The set of

portfolios with maximum return at each variance is the efficient frontier.


Problem it solves

How should an investor choose among securities when the future is uncertain? Markowitz shows the

"maximize discounted anticipated return" rule implies putting all funds in the single highest-value

security and "never implies that there is a diversified portfolio which is preferable" — contradicting

observed, sensible diversification. He therefore casts portfolio choice as a trade-off between a

desirable thing (expected return) and an undesirable thing (variance).


The method

  • Portfolio expected return = Σᵢ Xᵢ μᵢ; portfolio variance = Σᵢ Σⱼ Xᵢ Xⱼ σᵢⱼ (= wᵀΣw), where σᵢⱼ is
  • the covariance of returns i and j (σᵢⱼ = ρᵢⱼ σᵢ σⱼ).

  • Constraints: weights sum to one (ΣXᵢ = 1); short sales excluded (Xᵢ ≥ 0).
  • An E–V efficient portfolio gives maximum expected return for its variance (and minimum variance
  • for its return). Tracing these out yields the efficient frontier; the rational mean-variance

    investor holds a frontier portfolio. Markowitz illustrates the relations geometrically.


    Assumptions & inputs

    Requires the vector of expected returns and the full covariance matrix as "relevant beliefs about

    future performances" (Markowitz's stage one, observation/experience, supplies these; the paper treats

    stage two, the choice given beliefs).


    How to use it

    The conceptual and mathematical basis for the CAPM, risk budgeting, and essentially all quantitative

    portfolio construction; establishes covariance/diversification as the heart of risk management.


    Limitations & pitfalls

  • Mean-variance optimization is extremely sensitive to estimation error in expected returns and
  • covariances (DeMiguel et al. show naive 1/N often beats it out of sample).

  • Variance penalizes upside symmetrically and ignores fat tails / higher moments; inputs are hard to
  • estimate; the no-short-sale form is a special case.


    Key references

  • Markowitz, H. (1952) — Portfolio Selection — Journal of Finance
  • Sharpe, W. (1964) — Capital Asset Prices — Journal of Finance
  • DeMiguel, V., Garlappi, L. & Uppal, R. (2009) — Optimal Versus Naive Diversification — Review of Financial Studies


  • Provenance: verified/generated from the paper's full text.


    Community-maintained wiki — anyone can suggest an edit or view its revision history. Not peer-reviewed; verify claims against the original paper.

    Wiki last updated: June 23, 2026