How to read an index P/E ratio

A single stock's P/E is a price divided by that company's earnings. An index P/E is the same arithmetic done across a whole basket — and almost everything that makes it confusing comes from the differences between those two sentences.

What NSE actually publishes

NSE computes a P/E, a P/B and a dividend yield for every index it maintains, once a day, after the close. Three things about that are worth knowing before you read one:

It is consolidated, trailing twelve months. The denominator is the sum of the last four reported quarters of consolidated earnings across the constituents. Not a forecast, not this quarter annualised. So the number moves for two unrelated reasons: the index level changed, or a company reported and the trailing window rolled forward.

There is no intraday value. The P/E you see quoted during market hours is someone dividing a live index level by yesterday's earnings figure. NSE itself publishes one number per trading day, and that is the number with a defensible denominator.

It is free-float weighted, like the index. A company that is 70% promoter-held contributes 30% of its earnings and 30% of its market capitalisation. This is why an index P/E and a simple average of its constituents' P/Es are different numbers, sometimes by a lot.

Why the absolute number tells you very little

"The Nifty is at 22" is not a fact you can act on until you know what 22 has meant. Three reasons the raw figure travels badly:

Indices are not comparable to each other. A bank index and an IT index have structurally different P/Es because they have structurally different growth, capital intensity and accounting. Nifty Bank spending most of its life below the Nifty 50's P/E is not a decade-long bargain, it is what a leveraged balance sheet trades at.

Indian P/Es are not comparable to American ones. Different constituent mix, different growth rates, different tax treatment, different accounting standards. The comparison is made constantly and means almost nothing.

The same index is not comparable across regimes. An index whose composition has shifted from commodities to software over fifteen years is not the same earnings stream, even under the same name.

What survives all three objections is comparing an index to its own history. That is why valuation pages quote a percentile: not "22", but "the 41st percentile of every close since 2012".

The tail that breaks the average

Index P/Es have a distribution shape that catches people out. The denominator can approach zero — one quarter in which the constituents collectively earn almost nothing sends the ratio to a number with no useful meaning. Real example from NSE's own published series: Nifty Realty has a stored daily close with a P/E of 23,307.

A single print like that destroys any statistic built on a mean. The average of that series is 100.6 with a standard deviation of 1,069 — so "the average minus one standard deviation", a perfectly standard way to describe a cheap reading, comes out at minus 968. There is no such thing as a negative P/E band.

The fix is to stop using the mean. Percentiles are positions in the sorted series, so they cannot leave the range the index has actually traded in, and one absurd day moves the tenth percentile by one observation instead of by nine hundred points. When you see valuation bands quoted as a 10th/90th percentile rather than a standard deviation, that is what they are avoiding.

The same effect explains why a high index P/E is so often an earnings event rather than a price event. In 2020 and 2021, index P/Es across the world hit records while prices were well below their own peaks — the denominator had collapsed and was recovering on a twelve-month lag.

P/B and dividend yield, and why all three

Price-to-book compares the index to the accounting value of its constituents' net assets. Book value does not swing on one quarter's profit, so P/B is the steadier of the two ratios and the more useful one for banks and other balance-sheet businesses, where earnings are volatile and assets are the business.

Dividend yield reads backwards. For P/E and P/B, lower is cheaper. For yield, higher is cheaper — the same dividends are being bought for a lower price. Any tool that colours a dividend-yield chart on the same scale as a P/E chart has it inverted, and a decade-high yield will show as "expensive".

The three disagreeing is information, not noise. An index that looks cheap on earnings but expensive on dividends is usually telling you about payout ratios — companies retaining more of a growing profit — rather than about valuation.

What to do with it

Read it as a description of where an index sits in its own history, not as an instruction. A market can stay in its top decile for years, and the historical record of P/E as a short-horizon timing signal is genuinely poor. The next guide covers what the evidence actually supports doing with a valuation reading — which is less than most people expect, but not nothing.

Key takeaway

An index P/E is a consolidated trailing-twelve-month figure published once a day, and it is only interpretable against that same index's own history — percentiles, not the absolute number, and not another index or another country. Because the denominator can approach zero, any statistic built on the mean breaks on the tail; percentile bands cannot leave the range the index has actually traded in. And dividend yield reads backwards from the other two: high is cheap.