Is the market expensive? Valuation without market timing

Every valuation page on every site implies the same promise: learn to read this number and you will know when to buy. The honest version is narrower. Valuation tells you something real about long-horizon expected returns and almost nothing about what happens next year — and the gap between those two statements is where most of the money gets lost.

What the evidence supports

Across long histories and many markets, starting valuation has a real relationship with returns over ten years and longer. Buy an index in its cheapest decile and your decade tends to be better than average; buy in its dearest decile and it tends to be worse. The relationship is noisy but it is there, and it is not controversial.

Over one to three years, that relationship is weak enough to be unusable. An expensive market can get more expensive for years at a stretch, and it regularly does. This is not a flaw in the measurement — it is the actual behaviour of the thing being measured.

So a valuation reading is a statement about the distribution of long-run outcomes, not a signal about the next move. Treating it as the second is the standard error.

Two things that look identical and are not

Market timing is going to cash when a P/E crosses some threshold and buying back when it crosses another. This requires two correct decisions, the second harder than the first, and it has a poor record. The failure mode is not usually selling too early — it is never buying back, because the conditions that felt safe enough to re-enter arrive long after the market has recovered.

Rebalancing is holding a target allocation and selling whatever has grown past its share. This also sells into strength, but it requires no view about the future, has a rule that fires on its own, and never takes you to zero equity. A valuation reading can inform the target — a case for 60% equity rather than 75% — without ever becoming a switch.

The difference matters because the first thing tends to get justified using the same chart as the second.

What a percentile actually means

When a page says an index sits in the 8th percentile of its own history, that is a factual statement about 3,579 stored daily closes: 8% of them were lower. It carries no claim that the number will rise.

Two cautions that apply to every such figure:

The history is short. Fifteen years of Indian index data covers roughly two full cycles. A "decade low" is a low against a small sample, and it is not the same class of evidence as a century of data.

The index is not constant. The Nifty 50 of 2012 and of today share a name and about half a constituent list. Comparing a P/E across that span is comparing two related but different earnings streams, and the further back you go the weaker the comparison gets.

The zones, and what they are not

Valuation bands — cheap, fair, slightly expensive, expensive — are a reading aid. They describe where an index has traded. They are not a recommendation, and the honest version of each is narrower than it sounds:

• Historically cheap does not mean it will not get cheaper. Most of the cheapest readings in any series happened on the way down, and felt terrible.
• Fairly valued means the number is unremarkable, which is the condition markets spend most of their time in. The correct response is usually nothing.
• Expensive does not mean a fall is due. Indices spend long stretches in their top decile, and the ones that left it quickly are heavily over-represented in memory.

The practical version

If you are running SIPs, a valuation reading should generally change nothing. The whole mechanism of a SIP is that it buys more units when prices are low without requiring you to notice, and interrupting it on a P/E reading converts an automatic advantage into a manual judgement call.

Where a reading is genuinely useful is a lump sum. Deploying a large amount into an index in its dearest decile is a different proposition from doing it in its cheapest, and staggering the entry over some months is a reasonable response to a high reading — not because the market will fall, but because it reduces how much the outcome depends on one date.

It is also useful as a check on a story. When a sector has run hard and the narrative is that it deserves to, the sector's P/E against the broad market — and against its own history of that ratio — is the cheapest available reality test. That is the reading most worth taking from a valuation page, and the one least often taken.

For the mechanics of the ratios themselves, see how to read an index P/E ratio.

Key takeaway

Starting valuation has a real relationship with ten-year returns and a weak one with next year's, so a reading belongs in a target allocation rather than an in-or-out decision. Keep SIPs running regardless; stagger a large lump sum when the reading is high; and use a sector's multiple against the broad market as a reality test on a story that has run. A market can sit in its top decile for years, and most of the cheapest readings on record happened on the way down.