Reference
Indexes and Exchanges
What each index actually contains, how it is weighted, and when it is rebalanced. Three questions that decide what an index number means, and that most references skip in favour of the number itself.
An index is a rule, not a measurement. Someone decided which companies belong in it, how much each one counts, and when that changes, and every property of the resulting number follows from those three decisions rather than from the market. Two indices covering the same companies can disagree for a whole session because they weight differently, and neither is in error.
Three mechanics account for most of the confusion. Weighting decides influence: by capitalisation, by share price, or equally. Float adjustmentdecides whether locked-up shares count. Rebalancing decides when the list changes, and on those days the volume is mechanical, produced by funds obliged to trade, which is a genuine trap for any volume-based reading.
That last point is why this section sits in a volume reference at all. Breadth data is counted per exchange, and an exchange's list is full of things that are not operating companies; index rebalances generate enormous volume that carries no opinion. Knowing what is in a list, and when it changes, is a prerequisite for reading participation rather than an administrative detail.
The three weighting schemes
Almost every index in daily use follows one of three rules, and the rule determines what the number is capable of telling you.
| Scheme | Influence follows | Consequence |
|---|---|---|
| Float-adjusted capitalisation | Tradable market value | A handful of the largest members can carry the index while most of the list falls. The reason breadth data exists. |
| Price | Share price alone | A high-priced small company outweighs a low-priced large one, and a split changes a constituent’s influence without changing the company. |
| Equal | Nothing, every member counts the same | Behaves like a breadth measure expressed in returns, and requires regular rebalancing to stay equal. |
The comparison worth making regularly is a capitalisation-weighted index against its own equal-weighted version. When the weighted index outruns the equal-weighted one over weeks, the largest members are doing the work, the same conclusion the advance/decline line reaches by counting, arrived at through returns instead. When the equal-weighted version leads, participation is broad. Neither is a forecast, and both are more informative than the index level on its own.
Why an exchange list is not a list of companies
This is the point at which index mechanics stop being administrative and start distorting readings. Breadth data is counted per exchange, over every issue listed there, and a substantial part of that list is not an operating company.
Closed-end funds, preferred shares, exchange-traded products, warrants, units and multiple share classes of the same business all appear as separate issues and all count once in an advance/decline total. Many of them are interest-rate instruments in everything but name: they move together on bond news, and when they do, a breadth reading records a participation event that no equity investor experienced. This is why practitioners have preferred common-stock-only counts since at least the 1990s, and why a raw exchange count and a common-stock count can diverge from each other rather than merely from the index.
Two smaller effects belong with it. Multiple share classes double-count a single business, so a large company with two listed lines carries two votes in a measure designed to give one per company. And newly listed issues cannot set a 52-week extreme until they have a year of history, so a heavy listing period quietly shrinks both the new-high and new-low counts without anything changing among the companies already listed.
US large cap
Three indices with three different weighting schemes, which is why they disagree with each other on any given day.
- S&P 500Float-adjusted capitalisation weighting, committee selection, quarterly rebalance.
- Dow Jones Industrial AveragePrice-weighted and only thirty names, the oldest and least representative construction still in daily use.
- Nasdaq 100Modified capitalisation weighting, non-financial, with a special rebalance when concentration limits bind.
- S&P 100The largest hundred of the S&P 500, and the options-market reference.
Broad market and small cap
Where the constituent list itself is the useful object. These are the indices people actually look up rather than watch.
- Russell 1000 componentsThe large-cap segment of the Russell 3000.
- Russell 2000 componentsThe small-cap benchmark, and the annual reconstitution that moves it.
- Russell 3000 componentsThe parent universe both of the above are cut from.
- S&P 400 (mid cap)The middle band, frequently overlooked between the 500 and the 600.
- S&P 600 (small cap)Small cap with a profitability requirement, which makes it behave unlike the Russell 2000.
- Nasdaq CompositeThe one major index that selects nothing: every eligible security listed on the exchange is in it.
Exchanges
What is actually listed where. This matters most for breadth data, which is counted per exchange.
- NYSEAnd why its issue count includes a great deal that is not an operating company.
- NYSE American (AMEX)The smaller list, and its very different composition.
- Toronto Stock ExchangeThe Canadian reference, heavily weighted to resources and financials.
Beyond equities
Not every benchmark in this section is a list of shares. Where the underlying is a commodity, one name covers several different constructions, and which one is meant decides what the number means.
- Oil indicesProducers, services, refiners or the crude price itself, four things carry the name, and one can rise on the news that sends another down.
Frequently asked questions
Why do the Dow and the S&P 500 disagree on the same day?
Because they weight differently. The S&P 500 weights by float-adjusted market capitalisation, so a company’s influence follows its size. The Dow weights by share price, which is close to arbitrary. A $400 stock moves the Dow four times as much as a $100 stock regardless of which company is larger. On a day when high-priced constituents move against large ones, the two indices can point in opposite directions without either being wrong.
What is float adjustment and why does it matter?
Only shares actually available to trade are counted. Stock held by founders, governments or cross-holding companies is excluded, so a business whose shares are largely locked up carries less index weight than its total market value implies. Without the adjustment an index would be pushed around by shares that cannot be bought, which was a real distortion before the major providers adopted float weighting in the 2000s.
Why does the Russell reconstitution matter?
Because it is a single annual event that reassigns thousands of issues between the Russell 1000 and 2000 at once, and every fund tracking those indices has to trade the changes. The result is one of the highest-volume sessions of the year, in which volume says nothing whatever about opinion; it is mechanical. Any volume-based reading taken on reconstitution day is measuring index maintenance, not participation.
Does the S&P 600 behave like the Russell 2000?
Less than their descriptions suggest. Both are small-cap benchmarks, but S&P applies an earnings requirement for inclusion and the Russell does not, so the S&P 600 systematically excludes unprofitable small companies that the Russell 2000 holds. That single rule has produced persistent differences in their returns and volatility, and it is why the two are not interchangeable as "small cap".
Are index constituent lists free to obtain?
The current membership of the major indices is published by the providers and is easy to find; complete historical membership, with the dates of every addition and removal, is a commercial product. This asymmetry has a consequence worth knowing about. Any study built from a current list applied to past data has quietly excluded every company that was dropped, which biases the result upward, the survivorship problem, and the reason a backtest on "the S&P 500 today" over twenty years is not a test of the index at all.
What is the difference between an index and an exchange?
An index is a selected list with a weighting rule; an exchange is a venue where securities are listed and traded. The Nasdaq Composite is close to the boundary — it contains essentially everything listed on the exchange rather than a selection — which is exactly why it behaves differently from the Nasdaq 100. Breadth data is counted per exchange rather than per index, so the distinction decides what any advance/decline number actually covers.
Why is a price-weighted index still in use?
History and recognition rather than merit. The Dow Jones Industrial Average dates from the 1890s, when adding share prices and dividing was the only calculation available to a newspaper, and it survives because it is quoted everywhere and understood as a landmark. It is not a good measurement of the American market (thirty names, weighted by an accident of share price), and everyone who works with it professionally knows that, which is a reasonable definition of a convention.
How much volume does a rebalance actually generate?
Enough to make the session unrepresentative. On the annual Russell reconstitution, and to a lesser degree on quarterly S&P rebalances and index-related expiries, funds tracking the affected indices must trade the changes at the same reference prices, which concentrates a very large amount of activity into the close. The share counts are real and the interpretation is not: none of that volume carries an opinion about anything. Any volume comparison spanning such a date should exclude it or state that it does not.
Does the number of listed issues change over time?
Substantially, and it undermines any long history of raw breadth counts. The number of issues listed on the American exchanges has risen and fallen over the decades with waves of listings, mergers, delistings and the growth of funds and preferred issues that are not operating companies. A count of 200 new highs means something different on a list of 3,000 than on one of 1,500, which is the argument for the bounded, percentage-based forms of every breadth measure whenever the comparison reaches back more than a few years.