Concentration Risk in Portfolios: Measuring It by Position, Sector and Factor

Concentration risk is the exposure a portfolio carries because too much of its outcome depends on one thing: one position, one sector, one balance sheet type, or one underlying driver. It is measured by looking at how unevenly exposure is distributed rather than by counting holdings, and the three levels at which it is usually measured, position, sector and factor, frequently give different answers about the same portfolio.

What it measures

Concentration is a property of the weight distribution. If a portfolio’s capital is spread evenly, no single outcome dominates. If it is lumpy, a small number of outcomes decide most of the result.

Several standard summaries exist, and each compresses the distribution differently.

Largest position weight. The simplest measure. It answers the blunt question of how much rides on the single biggest holding.

Top-N weight. The combined weight of the largest five or ten positions. It is easy to communicate and is the measure most often quoted in mandates and fund documents.

Herfindahl index. Square each holding’s weight, expressed as a decimal fraction, and add the squares. The index rises as concentration rises. Squaring is what does the work: it makes large weights count disproportionately, so the measure notices lumpiness that a simple count cannot.

Effective number of holdings. One divided by the Herfindahl index. This is the most intuitive of the family, because it converts the index back into a number of positions. It answers: how many equally weighted holdings would produce this much concentration? A portfolio holding sixty names with a lopsided weight distribution can have an effective number in the teens. The gap between the actual count and the effective number is often the single most revealing concentration statistic a portfolio produces.

The same arithmetic applies at every level. Group holdings by sector and compute the effective number of sectors. Group by factor exposure and do the same. Nothing changes except what the weights are attached to.

How to read it

Position concentration is necessary but never sufficient. Position-level measures catch the obvious case where one holding dominates. They are blind to the more common case, where no individual position is large but many positions do the same thing. A portfolio of thirty equally weighted names looks perfectly diversified at position level and can still be a single undiversified bet.

Sector concentration catches the next layer, imperfectly. Grouping by sector reveals whether the portfolio is leaning heavily on one part of the economy. The limitation is that sector labels are administrative categories, not behavioural ones. Two companies in the same sector classification can have almost nothing in common economically, and two in different sectors can share the same customer, the same commodity input, or the same funding market. Sector concentration is a useful proxy for shared drivers and a poor substitute for identifying them.

Factor concentration is where the real bet usually hides. Factor exposure asks what characteristics the portfolio is loaded on: valuation level, size, momentum, profitability, leverage, liquidity, and so on. A portfolio can be spread across many names and many sectors and still be overwhelmingly a bet on one characteristic, for example small size or high momentum. That exposure only becomes visible when holdings are scored on the characteristic and the weights are aggregated. This is the work described in factor exposure analysis.

Read concentration alongside correlation. Concentration measures how much exposure sits where. Correlation measures whether the exposures move together. Neither is complete alone. A portfolio with modest position weights and high average pairwise correlation is effectively concentrated even though no weight measure says so, which is why the two are usually read as a pair with the correlation matrix.

Distinguish deliberate concentration from accumulated concentration. A portfolio can become concentrated without anyone deciding to concentrate it. Winners grow, weights drift, and a position that was sized at three percent becomes nine percent through performance alone. That is drift rather than decision, and it is the reason concentration measures are usually tracked over time rather than checked once. The mechanics are covered in measuring portfolio drift.

Watch the non-obvious axes. Concentration is not only about names and sectors. Portfolios can concentrate in liquidity, where a large share of capital sits in positions that would take many days to exit; in market capitalisation band; in promoter or group affiliation, where several holdings belong to the same business house; in geography of end demand, such as export exposure to one region; and in a single input cost or currency. Each of these is a valid grouping for the same arithmetic, and each occasionally turns out to be the axis that mattered.

The recurring pattern is that concentration is discovered on the axis nobody was measuring.

What it does not tell you

It does not tell you the right level. Concentration measures describe a portfolio. They contain no view about whether that level is appropriate, because appropriateness depends on mandate, horizon, liability structure, and constraints that live outside the data. The measurement and the judgement are separate activities and should stay separate.

It does not measure magnitude of risk. A ten percent weight in a stable, low-volatility holding and a ten percent weight in a volatile one contribute very differently to portfolio outcomes, yet every weight-based concentration measure treats them identically. Weight is exposure, not risk. Combining weight with volatility gives risk contribution, which is a different and often more informative view.

It does not detect shared drivers by itself. The Herfindahl index computed on sector weights only knows the sector labels it was given. If the true common driver cuts across sectors, for example an interest-rate sensitivity shared by lenders, real estate and consumer finance, the sector measure will report comfortable diversification. Concentration measurement is only as good as the grouping chosen.

It says nothing about what happens under stress. Exposure that looks distributed in normal conditions can collapse into one exposure in a sell-off, when correlations rise and liquidity thins. Static concentration measures do not anticipate this. Stress testing does, which is the point of stress testing a portfolio.

It is a snapshot, and snapshots can be timed. A concentration figure computed on one date reflects that date’s weights. Portfolios reported at period ends may look different from how they were held through the period. A series of measurements is more honest than any single one.

It does not capture concentration inside a holding. A single company can itself be concentrated, deriving most of its revenue from one customer, one product, one plant, or one regulator. That look-through concentration never appears in a portfolio weight table, and it is a genuine source of exposure. Reading it requires the business analysis described in segment analysis.

It does not measure income concentration. Portfolio yield frequently comes from far fewer holdings than portfolio value does, a point developed in portfolio dividend yield. A portfolio can be well distributed by capital and highly concentrated by income.

A practical routine

Compute concentration on at least three groupings rather than one, and expect them to disagree. Track effective number of holdings, effective number of sectors, and the portfolio’s aggregate exposure to whatever characteristics the process cares about. Watch the series rather than the point. And whenever a grouping shows comfortable diversification, ask what grouping was not tested, because that is where the concentration usually is.

Related reading

• Portfolio Metrics Explained: the hub connecting the portfolio-level statistics.
• Correlation Matrix in Portfolios: the co-movement half of the concentration question.
• Factor Exposure Analysis: finding the characteristics a portfolio is actually loaded on.
• Diversification: How Many Stocks?: what the holding count does and does not buy.
• Measuring Portfolio Drift: how concentration accumulates without a decision.

This article is educational. Altys Labs is not a registered research analyst or investment adviser, and nothing here is investment advice or a recommendation to buy, sell, or hold any security.