Radar Perene / Archive / science

How many stocks are enough to measure \"the market\" as a whole

◦ Index methodology v2.2 (working papers with DOI). See the methodology.

Science

"The market went up." The sentence is said thousands of times a day, and almost no one asks how many companies it takes for it to mean anything. If three stocks account for half of an index's move, "the market went up" describes the mood of three boardrooms. If five hundred stocks participate at diluted weights, the same sentence describes something genuinely collective. Between one case and the other there is no visible line — and it is precisely because the line is invisible that it deserves to be searched for with method.

The minimum number of stocks required for an aggregate metric to be stable is not a universal constant: it is a property of each market, one that shifts as the universe grows, concentrates or dilutes — and it must be estimated, not assumed. That is one of the central questions of the house's draft study on the breadth of the Brazilian market, and the answer imported from the manuals does not serve, because it was computed for a different universe.

"Enough" is a statistical word, not a rhetorical one

What does "enough" technically mean? An aggregate metric — the fraction of names rising, participation above moving averages, any cross-sectional statistic — is stable when it stops depending on any individual name: remove one stock from the calculation, and the reading should not change category. Below a certain universe size, that property does not exist. Each name carries too much weight, and the metric does not measure the whole — it measures the sum of half a dozen corporate biographies. Above a certain size, adding names changes almost nothing: the statistic has saturated, and the extra universe is welcome redundancy.

Between instability and saturation lies a transition zone — and the title's question lives in it. The decisive point is that this zone does not sit in the same place in every market: it depends on how many names exist, how value is split among them, and how much the names move together. A universe where everything rises and falls as a block saturates later — each new name repeats the previous ones and adds little independent information; a heterogeneous universe saturates earlier. The American answer to the question, silently embedded in the classical indicators, was computed — when it was — over thousands of names with moderate concentration. Nothing in it travels for free.

The Brazilian case: the question changes its answer over time

In Brazil, the question gains an extra layer of difficulty, documented in the previous article of this trail: the eligible universe went from 12 names in 2000 to 82 today. That does not merely mean the market was small and grew — it means the answer to "how many are enough?" traveled through the historical series in motion. There were stretches when the entire universe sat below any defensible stability threshold, and more recent stretches where some aggregate metrics begin to hold. The border between those stretches is not decorative: it separates the period when a breadth statistic describes the market from the period when it describes noise wearing an indicator's name.

The house's study estimates where that border passes for the metrics it examines — and the numerical threshold resulting from that estimate is content of the draft, not of this article. What can be stated in public is the shape of the answer: a threshold exists, is estimable, moves with the market's structure and, in the Brazilian case, is crossed within the historical series itself. Whoever computes breadth over the full series without marking that crossing is splicing two different instruments and reading the result as one.

Concentration pushes the threshold upward

Counting names does not close the account, because names do not weigh equally. In a market where a few issuers dominate value and liquidity — the Brazilian case throughout the series — the nominal count overstates real diversity: 82 names of which a handful answers for a disproportionate share of the movement behave, for statistical purposes, like a universe well smaller than 82. It is the distinction the sampling literature draws between the size of a sample and its effective size, discussed in this series in sample size in financial series: correlated or unequally weighted observations count for less than they appear to.

The practical consequence is uncomfortable and worth saying plainly: the sufficiency threshold of a concentrated market sits higher than the raw count suggests. The Brazilian market needs more names than a hypothetical balanced market of the same size to sustain the same metric — and that extra demand is exactly what imported formulas, calibrated on another structure, never charge for.

A question every market should ask itself

The usefulness of this question does not end with Brazil or with broad indices. It applies, with the same mechanics, to any cut: does a sector with nine listed companies support a "sector breadth" metric? Does a thematic index with fifteen components measure a theme or fifteen stories? The answer never comes from the formula — it comes from examining the universe: how many, at what weights, moving how much together. It is the examination the house's study performs for the market as a whole, and it is replicable for any subset one wants to measure. The next article in this trail steps exactly one rung down that scale: what happens when the cut is a sector, and the whole sector fits in half a dozen names.

Frequently asked questions

So what is the minimum number?

The threshold the house's study estimates for Brazilian metrics is content of the draft. The honest public answer is the method of the answer: the number is an estimable property of each market, a function of count, concentration and co-movement — not a textbook constant.

Why don't American manuals discuss this threshold?

Because in the universe where they were written, sufficiency was abundant — thousands of names made the question academic. A premise never needs declaring where it is never violated; it only becomes visible where it fails.

Is the Ibovespa, with its portfolio, "enough" to measure the market?

Different questions. A weighted index measures the performance of a rule-defined portfolio; a breadth metric asks how many participate in the move. This article concerns the second — and its answer neither validates nor invalidates the first.

Does this analysis apply to crypto or other assets?

The mechanics — count, concentration, co-movement, effective threshold — are agnostic to asset class. What changes is the input data and each universe's structure; the examination must be redone, not reused.

---

Continue the trail: Sector baskets in a concentrated market: when a few companies decide the whole sector →

House reading: the daily participation of names in the market's movement is in the Diário; the episodes when a few names decided the whole, in the Atlas.

Sizing this question for another market, a sector or a specific index — how many components are enough, given their weights and co-movement — is an exercise the house's bench conducts on consultation.