Radar Perene / Archive / science

From 12 to 82 stocks: why American-style market breadth does not fit Brazil

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

Science

American market-analysis manuals devote entire chapters to a family of indicators called market breadth: how many stocks rise against how many fall, what fraction of the universe trades above its moving averages, how many names print new highs. The premise of these indicators is so fundamental that the manuals do not even bother stating it: a market with hundreds or thousands of liquid stocks is assumed, where the "average opinion" of the names is a stable statistic. Transplant the formula to a market whose liquid universe fits on a short list, and it keeps computing — numbers come out, charts get drawn. They just no longer measure what they claimed to measure.

The classical market-breadth methodology presupposes a universe large enough that the behavior of a few names cannot dominate the statistic — and that premise, invisible in the manuals, is false in Brazil for most of the available history. That is the structural conclusion of the draft study the house maintains on the breadth of the Brazilian market, and the evidence supporting it is public: the count of the universe itself.

The universe the formula assumes and the universe that exists

The study's survey, run on public market data, sizes the problem without any sophisticated technique. In numbers: the universe of Brazilian stocks meeting minimum liquidity criteria for this kind of metric went from 12 names in 2000 to 82 today. Not 500, not 3,000 — 82, after more than sextupling.

The two ends of the count condemn naive importing, each in its own way. At the back end, a "market breadth" computed over 12 stocks is not a statistic of the market — it is nearly a roll call, where each name answers for an enormous fraction of the total and any idiosyncratic corporate event becomes a "signal" of the whole. At the front end, 82 is better, but still far from the scale at which the classical indicators were designed and validated. And between the two ends lives a problem more treacherous than either: the universe changed size along the series. A breadth reading from 2003 and one from 2024 are not the same metric on different dates — they are metrics of different resolutions wearing the same name. Comparing the two as if comparable is the kind of error that gives no warning when it happens.

Why the formula breaks in silence

What breaks is not the arithmetic — it is the interpretation. In the manuals, breadth works as a thermometer of participation: an index rising on wide breadth rises "with the market along"; an index rising on narrow breadth is being carried by a few. The reading presupposes that "few" and "many" are statistically distinguishable categories — that there is enough mass for the difference to mean something beyond noise.

In a short, concentrated universe, that distinction collapses from two sides at once. On the counting side: with a few dozen names, the border between "wide" and "narrow" participation moves with the entry or exit of a handful of stocks — the indicator becomes sensitive to exactly what it should ignore. On the weighting side: the Brazilian market concentrates a disproportionate share of value and liquidity in a few issuers, so the count of how many names rise can diverge completely from what the index does — not as the informative refinement the literature celebrates, but as a permanent artifact of concentration. The draft study documents episodes where this fragility showed up in practice; the taxonomy of those failures, and what the house changed because of them, is content of the study and stays there. The principle, however, needs no episodes: no sophisticated correction can give a statistic the mass its universe does not have.

The familiar relative: small sample under another name

None of this is Brazilian exoticism — it is a special case of a problem this series has already treated in sample size in financial series. Market breadth is, at bottom, a cross-sectional statistic whose sample is the day's universe of stocks. In the United States, that sample has thousands of observations; in Brazil, dozens. All the discipline demanded of conclusions drawn from short series — suspicion of extremes, wide margins, humility in comparisons — applies identically to statistics drawn from short universes. The difference is that a short series at least announces its length; the imported formula hides its own behind a consecrated name.

Hence the posture of the house's study: before asking "what does Brazilian market breadth indicate?", ask "from what universe size onward does that question make sense — and since when does Brazil reach it?". It is an inversion of order that changes the final product: instead of importing the indicator and interpreting the outputs, one first sizes the terrain and designs the metric it can carry.

What transfers to other small markets

The 12→82 count is Brazilian, but the argument holds no passport. Any emerging market with a liquid universe in the dozens — and there are many — faces the same trap when importing metrics designed for American scale: indicators that compute without protest and inform less than they appear to. The diagnostic script the house's study applies to Brazil — counting the eligible universe over time, locating the borders where the metric changes nature, mapping the effect of concentration — transfers to any of them. That transfer, applied to a specific market or index, is methodological consulting work; the diagnosis of principle is here, public.

Frequently asked questions

Where do the numbers 12 and 82 come from?

From the house's draft-study survey of public market data: the count of Brazilian stocks meeting minimum liquidity criteria for breadth metrics, in 2000 and today. The exact eligibility criteria are part of the study.

So breadth indicators are useless in Brazil?

Not useless — unimportable. The argument's conclusion is that the metric must be designed for the universe that exists, with interpretation sized to the available mass, rather than transplanted with another market's calibration.

Doesn't the growth from 12 to 82 solve the problem over time?

It eases one part and creates another: even if the universe keeps growing, the historical series will still cross stretches of different resolutions — and long comparisons will keep demanding the care this article describes.

Which concrete episodes motivated the study?

The study documents dated failures of the imported metric and the corrections adopted, and that material belongs to the draft, not to the article. What can be stated in public is the structural diagnosis — the scale premise Brazil does not meet.

---

Continue the trail: How many stocks are enough to measure "the market" as a whole →

House reading: how today's market splits between a few large names and many small ones is daily reading in the Diário; the episodes where concentration spoke loudest, in the Atlas.

Sizing this diagnosis for another small market — or for a specific index — is bench conversation the house has with those who seek it out.