Where does portfolio selection collide?

Markowitz treated a portfolio's expected return as the weighted average of its assets' returns, and its risk as the variance of that return, which depends on how the assets move together and not on each one alone. An investor should hold only portfolios that give the most expected return for a given variance, the set he called the efficient frontier. Because assets rarely move in lockstep, combining them lowers variance for the same return, and that is the exact meaning diversification carries here. His 1952 paper Portfolio Selection and his 1959 book of the same name turned the old advice about eggs and baskets into a constrained optimization over asset weights. The referent is a specific mean-variance calculation over a known set of assets, not a general habit of spreading bets.[1][2]

Unit

A set of asset weights on the mean-variance efficient frontier

Recovery means

Not applicable

Measured by

Expected returns, variances, and covariances of assets fed into a quadratic optimization of the portfolio weights

The mean-variance rule is only as good as the numbers put into it, and expected returns in particular are estimated with large error. Small changes in those estimates swing the optimal weights wildly and pile them onto a few assets, so the optimization tends to amplify the mistakes in its inputs, a problem sometimes called error maximization. DeMiguel, Garlappi, and Uppal's 2009 study Optimal Versus Naive Diversification compared many optimized strategies against the plain rule of splitting money equally across the assets, the 1/N portfolio, and found that across most datasets none of the optimized rules reliably beat equal weights out of sample. The meaning of diversification is still Markowitz's here, but the finding is that the elegant frontier can perform worse than a rule that ignores it. This is the same construct as the first field, measured honestly, and it shows the frontier is fragile even at home.[3]

Unit

Out-of-sample performance of an estimated portfolio versus the equal-weight rule

Recovery means

Not applicable

Measured by

Backtested certainty-equivalent return, Sharpe ratio, and turnover compared across optimized portfolios and the 1/N rule

In corporate strategy a firm's portfolio is the set of businesses or product lines it owns, and diversification means expanding into activities beyond its original line. The claimed benefit borrows the finance intuition that a spread of businesses is safer than one, but the mechanism argued in practice is different: shared resources, market power, and internal capital markets rather than the covariance of returns. Rumelt's 1982 study Diversification strategy and profitability sorted firms by how related their businesses were and found that those diversifying into related activities tended to outperform those spreading into unrelated ones. So here more diversification is not simply safer, which turns the finance reflex on its head. The meaning of the word is a corporate structure, not an optimized set of weights, and no efficient frontier is ever computed.[4]

Unit

The set of related and unrelated businesses a firm operates

Recovery means

Not applicable

Measured by

Firms placed in diversification categories by relatedness, then compared on accounting profitability and growth

Ecologists borrowed the finance term to name why a diverse community can be steadier than a single-species one: if many populations rise and fall somewhat independently, their combined abundance varies less than any one alone, just as a mixed bag of assets varies less than a single stock. Doak, Bigger, Harding, Marvier, O'Malley, and Thomson's 1998 paper The Statistical Inevitability of Stability-Diversity Relationships in Community Ecology showed that much of the observed link between diversity and stability follows from this averaging alone, before any special biology is invoked. The shared idea is the variance of a sum shrinking as you add weakly correlated parts, which is the same covariance argument Markowitz used. But the parts are populations, not assets, no one chooses the weights to optimize anything, and stability is read off as the variability of total abundance over time. The word portfolio is a deliberate metaphor here, and treating an ecosystem as if it were an optimized investment would misread it.[5]

Unit

The temporal variability of total community abundance summed across many populations

Recovery means

Not applicable

Measured by

Variance or coefficient of variation of aggregate abundance, compared with that of the individual populations