When the index lies about the market
A cap-weighted index can rise while most of its constituents fall. Breadth is how you find out which kind of day you just had.
An index is a weighted average, and in a cap-weighted index the weights are very unequal. A handful of the largest constituents can carry the whole number while the majority of the list goes the other way.
Which means the index can close up on a day when most of the market fell, and nothing on your index chart will tell you.
Three measures, in increasing usefulness
Advances versus declines. The simple count: how many constituents rose, how many fell. Crude, instantly interpretable, and the right first question. The running total — an advance-decline line — is the classic form.
Percentage above a moving average. What share of constituents are above their own 50-day or 200-day average? This is better than the daily count because it measures participation in a trend rather than in one session. A rising index with a falling percentage-above-average is the specific pattern worth noticing: the move is narrowing.
Equal-weight against cap-weight. Compare an equal-weighted version of the index with the cap-weighted one. When the equal-weighted version lags, the gains are concentrated in the large names. This is the cleanest single measure of narrowness, because it holds everything else constant and varies only the weighting.
What breadth is good for
Qualifying a breakout. An index breaking to a new high with broad participation is a different proposition from one doing it on three names. Both look identical on the index chart. The first has more ways to keep going; the second needs its three names to keep cooperating.
Noticing a rotation before it shows up in the index. Breadth often deteriorates while the index is still making highs, because money leaves the majority before it leaves the leaders. That sequence is common enough to be worth watching for.
Explaining a frustrating day. If your stock-level trades are losing while the index rises, breadth usually explains it immediately. You were not wrong about the market; you were in the wrong two-thirds of it. That is a sizing and selection lesson, not a thesis lesson, and it is worth distinguishing.
What it is not good for
Timing. This is the big one. Breadth divergences can persist for months. A narrowing market is a market with a particular fragility, not a market about to turn, and traders who sell narrowness as a signal spend long stretches being early and calling it being right.
Precision. There is no threshold. "Fewer than 40% above their average" is not a level at which anything happens. Breadth tells you about the character of a market, and character is a context for other decisions rather than a decision itself.
The useful framing is the same as every other reading in this module: internals change how much you trust a view, and how large you take it. They do not generate the view.
The data trap in breadth, and it is a familiar one
To compute breadth you need to know which instruments are in the index. That list changes — constituents are added and removed, companies merge, get renamed, and occasionally cease to exist.
Which means a breadth series computed over today's constituent list, applied backwards through history, is survivorship-biased in exactly the way the first module warned about. You are measuring the past participation of the companies that made it to the present. The names that were dropped — often because they fell a long way — are absent from your historical breadth, and their absence makes past breadth look better than it was.
Two practical defences:
Use the constituent list as of each date, not the current one, when computing any historical breadth series. This is more work and it is the difference between a measurement and a flattering reconstruction.
Identify constituents by a stable identifier rather than a ticker. Tickers get reassigned and renamed; an identifier like an ISIN follows the security itself. A breadth series that joins on ticker symbols will silently attach one company's history to another's after a rename, and nothing in the output will look wrong.
That is a mundane plumbing point with a large effect, and it is the reason the same instrument-identity discipline that matters for order routing also matters for a statistic you might think of as purely analytical.
What to take from this chapter
- A cap-weighted index can rise while most of its constituents fall. Every index-derived indicator describes the weighted average, not the market.
- Three measures, increasingly useful: advances versus declines, percentage above a moving average, equal-weight against cap-weight.
- Use breadth to qualify a breakout, to notice a rotation early, and to explain why your stock trades disagreed with the index.
- Do not use it for timing. Divergences persist for months, and there is no threshold at which something happens.
- A historical breadth series built on today's constituent list is survivorship-biased. Use the membership as of each date.
- Join constituents on a stable identifier, not a ticker. A rename silently grafts one company's history onto another's.