Concentration risk is the risk that a large exposure to one investment, issuer or shared source of loss makes a portfolio unusually vulnerable to that exposure going wrong. A list of different securities can still conceal the same economic bet. For a quant, the practical task is to identify what dominates the portfolio's losses, rather than simply count its holdings.
Concentration can be deliberate, or it can emerge as one holding outgrows the rest. Funds can overlap with direct holdings, and different companies can depend on the same industry, region or financing conditions. FINRA's explanation discusses these routes, including concentration in investments that are difficult to sell.
A large holding makes an isolated loss matter more
Consider a synthetic AUD 100,000 portfolio of three stocks, with no borrowing. Values and weights are measured at the beginning of one holding period. There are no trades, external cash flows, distributions or costs during the period, so the price changes below are also simple total returns in AUD.
| Holding | Beginning value | Beginning weight | Stipulated return | Ending value |
|---|---|---|---|---|
| A | AUD 60,000 | 60% | −30% | AUD 42,000 |
| B | AUD 20,000 | 20% | 0% | AUD 20,000 |
| C | AUD 20,000 | 20% | 0% | AUD 20,000 |
| Portfolio | AUD 100,000 | 100% | −18% | AUD 82,000 |
Only A is hit, but its loss removes AUD 18,000 from the whole portfolio. With fixed beginning weights and simple returns ,
This is a conditional scenario, not a forecast or a probability of loss. An equal-thirds allocation to the same stocks would lose 10% under that isolated A shock. But if all three stocks fall 30%, both allocations lose 30%. Reducing a large individual position helps against that position's isolated shock; it need not reduce exposure to a common shock.
Original synthetic scenarios: fixed beginning weights and one-period simple returns in the same currency. The loss bars share a common magnitude scale. Shocks are stipulated, with no probability assigned.
A concentration score measures the allocation
For a fully invested long-only portfolio, with and , a useful weight-concentration measure is the Herfindahl–Hirschman index, or HHI:
Use decimal weights in this convention. The reciprocal, often called the effective number of holdings, translates the score into an equal-weight equivalent. For the example,
The portfolio contains three stocks, but its weight concentration is greater than that of three equal holdings. Equal thirds gives and . More generally, : equal weights attain the lower bound, while putting everything in one holding attains the upper bound. Kritzman and Turkington, pp. 3–4, use the reciprocal squared-weight sum to measure holdings concentration and distinguish it from factor concentration.
Effective holdings are not independent bets. HHI contains no returns, volatilities or covariances. Identical weights produce identical HHI even if the assets' joint behaviour is completely different. A higher HHI therefore does not, by itself, prove that one portfolio has higher total volatility than another.
Grouping also changes the question. An issuer-level HHI requires aggregating positions in the same issuer before squaring; a sector-level HHI uses sector weights. Splitting one economic exposure into extra account lines should not create an apparent improvement in diversification.
The bounds and equal-weight interpretation above apply to nonnegative weights that sum to one. For a long–short portfolio, net weights can hide offsetting positions and leverage. Squaring net weights, or normalising absolute exposures by gross exposure, answers different questions; neither automatically inherits this long-only interpretation.
Look through the names to the shared exposure
Three different businesses might all suffer when financing becomes scarce. An investor might also own employer shares while relying on the same employer for salary. These are different routes to dependence on one adverse event. Fund holdings, collateral, counterparties and the ability to exit can matter alongside the visible security weights.
A useful concentration review combines several views:
- Position and group weights: identify the largest names, connected issuers and overlapping fund holdings.
- Shared drivers: measure factor exposures, including exposures that cross sector labels.
- Risk contributions: examine how much of a specified risk measure comes from each position or group. Capital weights and risk contributions need not match.
- Joint loss scenarios: use stress testing to examine common shocks, together with liquidity risk if many positions might need to be sold together.
This is also why position limits are useful but incomplete. They restrict how much capital can depend on one name; they do not ensure the remaining names depend on different drivers. Elton and co-authors discuss both the danger of concentrated allocations from noisy return forecasts and the trade-offs of security-level upper bounds in Modern Portfolio Theory and Investment Analysis, Chapter 10, p. 212.
To move from a holdings list to a model of shared drivers, Risk Models: APT and Factor Structure develops the aggregation of security exposures into portfolio factor exposures and the split between factor and specific variance. The full lesson requires membership; its public page provides a preview.
Sources
- FINRA, Concentrate on Concentration Risk, 15 June 2022: concentrated holdings, correlated exposures, fund overlap, weight drift and illiquid investments.
- Mark Kritzman and David Turkington, The Fallacy of Concentration, version 6 October 2025, pp. 3–4: holdings concentration, effective number of assets and its distinction from factor concentration. The numerical scenarios here are original; no empirical results from that paper are used.
- Edwin J. Elton, Martin J. Gruber, Stephen J. Brown and William N. Goetzmann, Modern Portfolio Theory and Investment Analysis, 9th edition, Wiley, 2014, Chapter 4, pp. 56–57 and Chapter 10, p. 212: covariance and diversification; forecast-driven concentration and position limits.