Charts · Heatmaps
Sector Heatmaps and Classification
Before a sector map can be drawn, somebody has to decide which sector each company is in. That decision is a judgement, providers disagree about the interesting cases, and moving one company changes two blocks.
The decision that comes before the data
A stock heatmap needs two numbers per company: a size and a change. A sector heatmap needs a third thing that is not a number at all, an assignment. Somebody has to say which block each company belongs in, and nothing in the price data answers it.
Several classification providers do this, with different schemes and different rules, and they agree about most companies. The disagreements are concentrated exactly where it matters: a retailer that sells mostly online, a manufacturer with a financing arm larger than some banks, a media business that owns its distribution. Two maps built from the same prices and different classifications are different maps.
That asymmetry is the practical point. A company is a small share of a large sector and a large share of a small one, so a reclassification barely moves the block it leaves and can dominate the block it joins. Sector maps are therefore least reliable exactly where they look most interesting, in the small blocks, where one member decides the colour.
Value-weighted or equal-weighted
Sector blocks are conventionally value-weighted, which means a sector’s colour is the colour of its largest members. In a sector with two enormous companies and forty small ones, that is a statement about two companies.
The equal-weighted version answers a different question, what the typical company in the sector did, and the two disagree regularly. Neither is wrong; they are the sector-level version of the index-against-breadth distinction that the whole breadth section is about. What is wrong is reading a value-weighted block and concluding something about most of the companies in it.
When a block moves as one
The most common reading of a sector map is that a whole block is one colour, and it is a legitimate thing to notice: the companies respond to the same input, which is what a sector is.
It also has a consequence that recurs across this reference. A sector-wide breadth count on such a day is close to uninformative, a unanimous energy block on a day when crude moved is one event counted many times, not broad participation. The oil pageworks through it in detail, including the case that ought to complicate every energy block: refiners earn a spread rather than a price, so a falling crude price can help them while it damages producers. A single sector colour averages that away.
Sector, industry, or sub-industry
| Level | What it groups well | Where it fails |
|---|---|---|
| Sector | A glance-level view: eleven or so blocks, readable at once, enough to see rotation. | Groups businesses whose relationship is looser than it looks, refiners with producers, payments with software. |
| Industry | Companies that genuinely respond to the same input, which makes an unanimous block meaningful. | Too many blocks to read as a picture; better as a table. |
| Sub-industry | Precise comparisons, the level at which "these two are competitors" is true. | Blocks with two or three members, where the average is one company and the map implies a group. |
The practical rule that follows: if a block looks internally contradictory, suspect the level of the grouping before the market. A sector that contains businesses on opposite sides of the same price is doing what its definition asks and telling you nothing.
What to check before quoting a sector figure
Four questions, and none takes more than a moment. They are the sector-level version of the checks every measure on this site ends with.
Whose classification? Providers disagree about the borderline companies, and the borderline companies are frequently large. A sector figure without a stated scheme is not reproducible.
Weighted how? A value-weighted block is a statement about its largest members; an equal-weighted one is a statement about the typical member. Both are useful and they answer different questions.
How many members? A block of three companies has an average that is really one company, and the map gives it the same visual authority as a block of sixty. This is where a sector map most often implies a group where there is a name.
Was there one input? If a commodity price or a rate expectation moved, a unanimous block is that one event counted many times. Checking the day's news before the map is the cheapest step in the whole procedure, and it is the one that most often turns an apparent rotation back into a single headline.
The definitions themselves move
Classification schemes are revised, and the revisions are occasionally large, whole groups of companies have been moved into new or renamed sectors. Each change is announced, and none of them is visible in a performance series afterwards.
So a sector’s long-run return history can span two different definitions of that sector, with the discontinuity invisible in the numbers. It is the same class of problem as the composition drift that runs through the breadth comparability page: the measure is computed correctly at every point, and what it measures has changed underneath it. Anyone comparing sector performance across a decade should establish that the sector meant the same thing at both ends of the comparison, and in several well-known cases it did not.
Frequently asked questions
What does a sector heatmap show?
The same two encodings as a stock heatmap — area for size, colour for change — grouped into blocks by sector. The grouping is what makes it useful and it is also the part that is a judgement rather than a measurement: somebody had to decide which sector each company belongs to, and that decision is not derivable from the prices.
Who decides which sector a company is in?
A classification provider, and there are several in use with different schemes and different rules. They agree about most companies and disagree about the interesting ones, a retailer that sells mostly online, a manufacturer with a large financing arm, a media business that owns a distribution network. Two maps built from identical prices and different classifications will not show the same picture.
How much difference can one company make?
Enough to change a sector’s sign, which the figure on this page computes. Moving one mid-sized company between two sectors shifts both sector averages, and because sector blocks are value-weighted, the effect on the smaller sector is much larger, a company that is a modest share of a big sector can be a substantial share of a small one.
Do the schemes change over time?
Yes, and the changes are announced and occasionally large, whole groups of companies have been reclassified into new or renamed sectors. That matters for any historical comparison: a sector’s long-run performance series may span two different definitions of the sector, and the discontinuity is not visible in the numbers. Anyone comparing sector returns across a decade should check whether the definition survived it.
Should the blocks be value-weighted or equal-weighted?
It depends on the question and the two answers frequently disagree. A value-weighted sector figure says what happened to the money in that sector; an equal-weighted one says what happened to the typical company in it. In a sector with two enormous members and forty small ones (technology, most of this century) the two can point in opposite directions on the same day, and only one of them is a statement about most of the companies.
Why does a whole sector often move as one block?
Because the companies in it respond to the same input, which is the honest reason a sector map is worth looking at. It also means a sector-wide breadth reading is close to uninformative: on a day when the oil price moves, an energy block will be almost unanimous, and that unanimity is one event counted many times rather than broad participation. The Toronto and oil pages work through the same problem at a market level.
Is the sector block the right unit at all?
Sometimes an industry or a sub-industry is the better one. Sector is the coarsest level of most schemes, and it groups businesses whose relationship is looser than it appears, refiners and producers sit inside energy and can move in opposite directions on the same news, which the oil page sets out. If a sector block looks internally contradictory, the classification level rather than the market is usually the reason.
What is a sector map good for?
Seeing that one group is doing something the others are not, which is a legitimate and useful glance. What it cannot support is a claim about how many companies participated (the blocks are size-weighted, so counting is impossible), or a comparison of two sectors’ magnitudes by eye, since area is read inaccurately. Both of those need a table or a breadth count beside the map.