Radar Perene / Archive / science

Sector rotation: what changes when the sector itself is too small to rotate

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

Science

Sector rotation is one of the most elegant ideas in applied finance: capital would migrate between sectors according to the phase of the economic cycle — from one group of industries to another, in a choreography that repeats with every turn. The classic image is a wheel with ten or eleven sectors, each holding dozens of companies, turning through the cycle. It is a good idea. And it was formulated for a market that has the whole wheel.

Sector rotation strategies assume that, for each phase of the cycle, there exists a set of investable, diversified, liquid sectors to migrate between. In a market where several "sectors" are two or three companies — and some chapters of the wheel simply have no listed representative — the imported strategy does not degrade gracefully: it changes nature. It stops being allocation across industries and becomes the selection of half a dozen stocks under another name.

The contrast is not rhetorical. The sector taxonomy used in developed markets distributes thousands of companies across eleven sectors; the Brazilian exchange distributes a few dozen liquid names across the same labels. The label survives the journey; the content does not.

A wheel with missing spokes

Three things break when the wheel shrinks. The first is coverage: phases of the cycle that, in the manual, call for exposure to a specific sector find in Brazil a sector with one dominant representative — or none with real liquidity. The migration the model prescribes has no local destination. The second is granularity: with few names per sector, "rotating into sector X" and "concentrating everything in company Y" become the same operation, and all the idiosyncratic risk that sector diversification was supposed to dilute comes back through the window — the problem documented in sector baskets in a concentrated market.

The third break is statistical, and it is the least discussed: testing whether rotation works requires watching the cycle turn many times. Economic cycles last years, as Track 1 recalls in economic cycles: describing is not predicting — and a market history with few usable decades contains very few complete turns. The Brazilian researcher who wants to validate the wheel has, at the same time, fewer sectors, fewer companies per sector and fewer observed cycles. The three scarcities multiply.

The study the house keeps on the bench

The house maintains its own study, still in draft, on sector rotation in the Brazilian market — precisely because the question "what remains of rotation when the wheel is small?" has no ready answer in the literature. The study's design and its preliminary findings stay on the bench, as is house rule for undeposited work: a draft is not a result, and a preliminary result announced is the raw material of promises that age badly.

What the stage of the work already allows us to say is the nature of the problem. This is not about calibrating the American strategy to local data — swapping parameters, adjusting windows. It is about deciding whether the object of the strategy exists here. That question precedes optimization, and the house's experience with imported formulas teaches that it is almost always skipped: the literature arrives with its prestige, the local data is forced into it, and the bill for the mismatch arrives years later, in the series that does not behave as the manual said.

What the reader can use right away

Even without the study's results, the structural question is transferable. Faced with any sector rotation proposal applied to Brazil, three checks come before any performance statistic: how many investable sectors the proposal actually distinguishes, how many liquid companies each contains, and how many complete cycles the test period covers. Proposals that do not answer all three with numbers are testing the packaging, not the strategy.

None of this concludes that sector rotation "does not work" in Brazil — that conclusion would require exactly the careful study that sits on the bench. The claim that already stands is more modest and more useful: the imported version of the strategy does not describe the market it is meant to be applied to, and any honest validation has to begin by rebuilding the object, not copying the recipe.

Frequently asked questions

Why does the house mention a study whose results it does not publish?

Because the stage is part of the honesty: the study exists, it is in draft, and drafts generate no public claims. When it matures under the house's criteria, it will be deposited with a permanent identifier, like the others. Announcing a preliminary finding would be asking credit for work that can still change.

Is sector rotation invalid in every small market?

The classic formulation assumes a structure that small markets do not have — that is composition arithmetic, not a performance verdict. Whether some adapted version preserves value is an empirical, open question, and it is what justifies a study of one's own instead of direct importation.

Doesn't the problem go away with sector ETFs?

The vehicle does not create what does not exist: a sector fund over three liquid companies carries the same concentration with one more layer of packaging. The relevant question remains the composition, not the format.

Where does cycle reading fit into this?

As a prerequisite: rotation assumes identifying the cycle's phase in real time, which is already hard — describing cycles is more honest than predicting them. The fragility of the Brazilian wheel adds to that difficulty; it does not replace it.

---

Continue the trail: Small sample, big conclusion: the risk of generalizing from scarce data in Brazil →

House reading: today's reading is in the Diário; the cycle turns the archive recorded, in the Atlas.

Assessing the viability of sector rotation for a specific portfolio — with the wheel that actually exists, not the manual's — is a conversation the house has case by case.

Characters: Method