Indexes · Price-weighted
The Dow and Its Price Weighting
Thirty companies weighted by the price of one share each, a method chosen in 1896 because it could be done by hand. It is the only major index built this way, and everything unusual about the Dow follows from it.
One decision, made in 1896
Every other index in this section weights its members by market value: influence follows size. The Dow weights by share price. A member whose stock trades at six hundred dollars moves the index many times as much as one trading at thirty, whatever the two companies are worth.
That is not a theory about measurement. It is what could be computed by hand in 1896 — add the prices, divide by the number of members — and it required no share-count data, which was not conveniently available. The method survived because the index became famous, and the famous number could not be redefined without breaking its own history.
| Member | Share price | Market value | Weight by price | Weight by value |
|---|---|---|---|---|
| High-priced insurer | $620 | $180bn | 44.1 % | 3.6 % |
| Industrial conglomerate | $210 | $340bn | 14.9 % | 6.8 % |
| Telecom | $22 | $190bn | 1.6 % | 3.8 % |
| Retail giant | $95 | $780bn | 6.8 % | 15.6 % |
| Bank | $48 | $620bn | 3.4 % | 12.4 % |
| Software major | $410 | $2,900bn | 29.2 % | 57.9 % |
The split problem
The sharpest illustration is a corporate action that changes nothing. Suppose the high-priced insurer splits its stock four for one: every holder ends up with four times as many shares at a quarter of the price, and the business is untouched.
Its weight in a price-weighted index falls from 44.1 % to 16.5 %, roughly a quarter of what it was. In a value-weighted index the weight would not move at all, because the company is worth the same.
So a decision about share denomination, made for reasons of marketing and liquidity, changes how much a company matters to the index. That is the clearest available evidence that price weighting measures something other than the market, and it is the reason no index designed after the twentieth century uses it.
What the divisor is for, and what it costs
The index is not the sum of the thirty prices divided by thirty. It is that sum divided by a divisor maintained by the index provider and adjusted whenever a member splits, is replaced, or pays a special distribution.
The purpose is continuity: without it, the four-for-one split above would take a large slice out of the index level overnight for no market reason. The divisor absorbs the change so the series stays comparable, which is exactly right and is the same mechanism the S&P 500 page describes.
The cost is that the level becomes unreadable. It cannot be reconstructed from the member prices without the current divisor, its relationship to any dollar amount is arbitrary, and comparing the level across decades compares two scalings. This is the same class of error as reading the level of a cumulative volume line: the direction and the changes are meaningful, the number is a construction.
Why it usually tracks the broad market anyway
Given all of the above, the Dow’s long-run correlation with a broad value-weighted index is high enough that the two rarely tell different stories over a year. That fact is often produced as a defence of the method, and it is worth being precise about what it actually shows.
It shows that the largest American companies dominate both measures, the same conclusion the S&P 100 page reaches from the other direction. A thirty-company committee selection of large industrials, however oddly weighted, ends up holding much the same businesses that a value-weighted index is dominated by. The agreement is a property of market concentration, not evidence that weighting by share price is sound.
Where the two do diverge is over shorter periods and around index changes. A substitution replaces one member out of thirty and can move the index’s composition noticeably in a single step, and a high-priced member having a bad month affects the Dow more than its size warrants. Those are the occasions when quoting the Dow and quoting the market are different acts.
What it is good for
Its history. The series runs further back than any other widely quoted equity index, and for questions about the very long run there is often nothing else, which is why the crash record and the bear-market page both use it. That value is real and it is independent of the weighting method.
For anything else, a broad float-weighted index answers the question better. And for the question this site is mostly about, how many companies took part in a move, no weighted index of thirty members can help at all, which is what the breadth section exists for.
Frequently asked questions
How is the Dow weighted?
By share price. Each of the thirty members contributes in proportion to the price of one share, so a company whose stock trades at six hundred dollars carries many times the influence of one trading at thirty, regardless of which is the larger business. Every other index in this section weights by market value, which is why the Dow behaves differently from all of them.
Why would anyone weight by price?
Because in 1896 it was the only practical method. The original average was the sum of the member prices divided by their number, which is something a person can compute by hand daily and which needs no share-count data at all. It is an artefact of what was possible rather than a view about how to measure a market, and it survives because the index became famous before anyone had reason to change it.
What is the divisor?
A number the index is divided by, adjusted whenever a member splits its stock, is replaced, or pays a special distribution, so that the index level does not jump for a reason unconnected to the market. It does its job well and it has a consequence: the level is no longer a sum of prices divided by thirty, and it cannot be reconstructed from the member prices without knowing the current divisor. The number in the headline is a scaled quantity, not a price.
Does a stock split change a company’s influence?
Yes, and this is the clearest demonstration that price weighting measures the wrong thing. A four-for-one split leaves the business identical and quarters the share price, which quarters the company’s weight in the index. Nothing happened to the company; its influence fell by three quarters. The divisor prevents the index level from jumping, and it cannot restore the weight.
How much does the choice of method matter?
Enough to change which companies drive the number. The figure on this page computes both weightings for six hypothetical members and shows the gap: a high-priced but mid-sized company is substantially over-weighted relative to its size, and the largest business in the list is substantially under-weighted. Over a long period the two indices are broadly similar because the same large companies dominate the market anyway, which is a statement about the market rather than a defence of the method.
How are the thirty members chosen?
By a committee, with no published quantitative rule. The stated aim is representation of American industry outside transport and utilities, which have their own averages, and changes are infrequent. As with the S&P indices, membership is not reproducible from any formula, but here the committee is smaller, the count is fixed at thirty, and the price-weighting method means that a substitution can shift the index’s composition noticeably in one step.
Is it still worth following?
As a very long continuous series, yes: it is the oldest widely quoted equity index and its history is genuinely useful for that reason alone. As a measure of the American market it is inferior to a broad float-weighted index in every respect, thirty companies, weighted by an accident of arithmetic. The sensible position is to read it as a piece of financial history that remains in the headlines, and to use something else when the question is what the market did.
What about the breadth data underneath it?
A breadth count across thirty companies is a poor measure of anything: the sample is too small for a ratio to be stable, and one issue moves the count by more than three per cent. Breadth measures need a full exchange list, which is why every breadth page on this site works from exchange-wide counts rather than from index members. For the Dow, the useful version of the same question is simply how many of the thirty rose, read as a count rather than a ratio.