Concentration Is a Hidden Factor

Buy a broad index fund and you are told, reasonably, that you own the market: hundreds or thousands of companies, the very picture of diversification. The arithmetic of how that index is built quietly says otherwise. Because the largest indices weight their members by market capitalisation, a small number of the biggest companies can come to dominate the whole, and by early 2025 a handful of large US technology firms accounted for an outsized share of the major benchmarks. The passive investor who believed they held the market in fact held something narrower and more opinionated: a concentrated position in one story about technology, growth, and the future of computing. Concentration had become a factor, and nobody had chosen to buy it.

Section 01Breadth by count, concentration by weight

The trap is the difference between two kinds of breadth. Breadth by count is the number of names you hold; breadth by risk is the number of genuinely independent bets those names represent. Capitalisation weighting drives a wedge between them. When the few largest constituents carry a large share of the total weight, the index behaves, for risk purposes, far more like those few names than like its long tail. A thousand holdings with most of the weight in a dozen is, when something moves that dozen, close to a dozen-name portfolio wearing the costume of a diversified one.

This is not an abstract worry. In its May 2025 Financial Stability Review, the European Central Bank warned that strong market concentration, along with exposures to a handful of large firms, mostly US-based technology companies, continues to expose global markets to risks arising from shocks to these entities. The supervisory language and the portfolio language are describing the same object from different angles: a market whose apparent diversity rests on a narrow base.

Section 02Concentration as a factor

Figure 5 follows the decomposition. The broad index narrows to its dominant cap weights, which sit overwhelmingly in one sector, which in turn loads on a small set of shared factors: growth, the long-duration cash flows that make those valuations sensitive to interest rates, and a collective expectation about AI capital spending. Because the big names share these factors, they tend to move together, so a shock to any one of the shared factors is a shock to all of them at once. The diversification implied by the count of holdings is largely illusory; the effective number of independent positions is far smaller, and it is that smaller number that governs how the portfolio behaves under stress.

The live demonstration came in January 2025, the subject of this volume's first issue. A single model release repriced the shared premise that frontier AI requires ever more hardware, and the move passed almost directly into the index because so much of the index was, in substance, the same bet. That episode was not a freak. It was the hidden factor revealing itself, exactly as a factor does: invisible while quiet, decisive when moved. Concentration is not a separate risk sitting beside market risk; it is a lens that magnifies whichever shock happens to strike the crowded exposure.

effective_constituents   N_w  =  1 / sum_i ( w_i^2 )   # inverse Herfindahl

if a few weights are large:   N_w  <<  number_of_holdings

portfolio_variance  ~=  (big weights)^2 * var(shared_factor)  +  ...
                    => dominated by a few names loading one factor

Section 03A portfolio problem and a stability problem

What makes concentration more than a private portfolio matter is that the same exposure is held, in much the same form, across the system, and often through vehicles that can be forced to sell. The ECB's May 2025 review is direct about the danger: financial markets, particularly equity markets, remain vulnerable to sudden and sharp adjustments due to persistently high valuations and risk concentration. It noted that the higher than expected US tariffs announced on 2 April 2025 injected significant volatility into markets and triggered a major sell-off in riskier assets, with magnitudes not seen since the early stages of the COVID-19 pandemic. A concentrated, richly valued market had found its trigger.

The review then adds the amplifier. Declining liquid-asset holdings and significant liquidity mismatches in some open-ended investment funds, together with procyclical flow dynamics, could amplify adverse market shocks and tip an orderly correction into a disorderly one. Put the pieces together and concentration is revealed as a hidden factor at the level of the system, not only the portfolio: a shared exposure, richly priced, held through vehicles that must sell into weakness. That is the same loop this series meets again in its work on crowding, arrived at from the direction of the index rather than the strategy.

The label says diversified. The variance says concentrated.

The intellectual distinction is between constituent breadth, weight breadth, and risk breadth. The first counts securities; the second measures how capital is distributed; the third asks how many independent economic exposures remain after covariance is considered. A capitalisation-weighted index can score well on the first and poorly on the other two. Concentration is therefore not a forecast of correction. It is a statement about transmission: when a shared premise is repriced, more of the portfolio moves together than the number of holdings would suggest.

This research is analysis and commentary for general information. It is not investment advice, an offer, or a solicitation, and it contains no price forecasts. Regulatory descriptions summarise the cited sources; the interpretation is the author's.