Valuation across the cycle

Here is a puzzle that trips up almost every beginner. Two of the most reliable-sounding rules in investing are "buy cheap" and "avoid expensive." Yet across a full market cycle, cheap things keep getting cheaper for a long time, and expensive things keep getting dearer for a long time — often for years, and often to extremes that make no sense on the day they finally reverse. If cheap is good, why does it so often keep falling? If dear is bad, why does it so often keep rising?

The answer is that the cycle does not just move earnings up and down. It moves what people are willing to pay for those earnings — the multiple, the price you pay per rupee of profit. In a boom, optimism stretches the multiple wider and wider; in a bust, fear compresses it tighter and tighter. So the price of a share moves for two reasons at once: the earnings change, and the multiple applied to those earnings changes. The question this module answers is: how does the cycle distort valuation, and how do you read a multiple without being fooled by the stage you are standing in?

Most beginners are taught valuation as a snapshot: look up the price-to-earnings ratio — the share price divided by earnings per share — and judge cheap or dear against some fixed line, say "under 15 is cheap, over 30 is expensive." This works badly, and it works badly for a specific reason: both numbers in the ratio are moving with the cycle, and they move in ways that can flip the ratio's meaning entirely.

Consider the two forces. First, earnings themselves swing with the cycle, and for some businesses they swing violently. A steelmaker or a property developer earns enormous profits at the peak of demand and tiny or negative profits at the trough. Second, the multiple the market pays swings too — wide in confident times, narrow in fearful ones. Put these together and you get the trap at the heart of cyclical investing: a stock can show its lowest P/E at the most dangerous moment, because a normal price is being divided by a temporarily huge peak profit.

This is why valuation must always be read through the cycle, not against a fixed ruler. The investor Howard Marks built a whole discipline around the modest version of this: you cannot time the turn, but you can read roughly where you stand and lean accordingly. A multiple read without knowing where you are in the cycle is not information — it is a number that means opposite things at opposite stages. Learning to place it is the whole point.

Two distinct things stretch and compress across a cycle, and it helps to keep them apart.

The multiple expands and compresses with mood. In a confident market, investors extrapolate the good times forward and pay more per rupee of earnings — a process called re-rating. The same steady company that fetched 20x its earnings in a sober year can fetch 40x in a euphoric one, with no change in its actual profits. In a fearful market the reverse happens — a de-rating — and the multiple collapses even if earnings hold. This is the reflexive loop from the last module wearing a valuation costume: rising prices breed the optimism that pays higher multiples, which lifts prices, which breeds more optimism.

For cyclicals, the earnings themselves swing — and the low P/E lies. This is the sharper trap. When a commodity or cyclical business is at the peak of its demand cycle, its earnings are temporarily bloated. Divide a roughly normal price by a bloated peak profit and the P/E looks tiny — 4x, 5x — and screams "cheap." But that peak profit is not repeatable; the cycle will turn, earnings will shrink, and the flattering multiple will vanish. At the trough, the opposite: earnings are crushed, so even a low price divided by a tiny profit shows a high P/E that screams "expensive" — right when the stock may actually be cheap on any through-cycle measure.

The defence against both traps is the same idea in two forms: do not read a single snapshot. For cyclicals, look at earnings averaged across a whole cycle — good years and bad — so a peak profit cannot masquerade as normal. This through-cycle habit, sometimes formalised as a cyclically-adjusted multiple, strips out the stage you happen to be standing in. And for the multiple itself, compare it to the company's own history and to a sober, cycle-averaged level — not to a fixed rule and not to the euphoric multiple of the moment. Behind both sits the quiet gravity of mean reversion: multiples and cyclical earnings both tend, over long stretches, to pull back toward their own long-run averages — though, as ever, on no timetable you can predict.

Walk one cyclical through the trap. illustrative

A composite steelmaker sits at the peak of a construction boom. Steel prices are high, its plants are running flat out, and it is earning a record ₹100 per share. The stock trades at ₹500, so its P/E is 5x. A beginner runs a screen for "low P/E stocks," sees 5x, and concludes this is the cheapest quality name on the exchange. The story on every channel agrees: infrastructure is booming, demand is unstoppable.

Now read it through the cycle instead. Ask the awkward question: is ₹100 per share a normal year or a peak year? Across the last full cycle, this company earned ₹100 at the top, ₹40 in a middling year, and lost money at the bottom — call it ₹35 through the cycle on average. On that normalised profit, ₹500 is not 5x; it is over 14x. The "cheap" 5x was an illusion created by dividing a normal-ish price by a peak profit that will not last.

Then the cycle turns, as cycles do. Construction slows, steel prices fall, and the company's earnings drop to ₹20 next year. The stock falls to ₹350 — but now the P/E, ₹350 divided by ₹20, is 17.5x. On the screen it now looks expensive, right when it may be closer to fairly valued than it was at the flattering 5x. The beginner who bought at 5x "cheap" is sitting on a 30% loss, baffled that a low-P/E stock fell. The multiple did exactly what cyclical multiples do: it was lowest at the top and highest near the bottom.

This is the inversion of valuation: the same low P/E is a bargain signal for a steady grower and a peak-danger warning for a cyclical. The number cannot be read without knowing which kind of business, and which stage of the cycle, produced it.

Reading valuation through the cycle protects you from a great deal. It also cannot do several things, and pretending otherwise creates its own losses.

It cannot tell you when the multiple reverts. Mean reversion is a long-run gravity, not a schedule. A stretched multiple can stretch further for years, and a compressed one can stay compressed long past the point of logic. Knowing a multiple is extreme tells you the neighbourhood, never the date. Selling because "40x must revert" can leave you watching it run to 60x — the market can stay irrational far longer than your patience holds.

It cannot tell you that cheap is safe or dear is doomed. Some cheap things are cheap because they deserve to be — a business in genuine, permanent decline. Some dear things are dear because the growth is real and durable. A low multiple is not a floor and a high one is not a ceiling; each is a question, not an answer.

And it cannot, by itself, tell you that "this time is different" is false. Occasionally it truly is — a genuinely new kind of business does sometimes deserve a genuinely new multiple. But the four most expensive words in investing earn their reputation because the story is usually excusing the price rather than describing durable cash flows. The honest test is never the confidence of the narrative; it is whether the extra multiple shows up as growing, durable cash — or only as the justification for what people are already paying.

Valuation traps are among the most expensive in investing precisely because the numbers look so objective.

1. Buying a cyclical on its low P/E. The lowest multiple often marks the top, because peak earnings sit in the denominator. The screen says "cheap"; the cycle says "peak." Always ask whether the E is a normal year or a boom year.

2. Judging a multiple against a fixed ruler. "Under 15 is cheap" ignores that a fair multiple depends on the business, its growth, and the mood of the moment. Compare to the company's own history and a sober, cycle-averaged level instead.

3. Mistaking a re-rating for earning power. When a steady company's price doubles because its multiple went from 20x to 40x, no extra rupee was earned — the cycle simply agreed to pay more. That premium is borrowed from sentiment and can be recalled.

4. Assuming reversion is imminent. A stretched or compressed multiple can stay that way for years. Extreme tells you the neighbourhood; it never tells you the hour, and betting the hour is timing in disguise.

5. Letting a story retire the arithmetic. "The old valuation rules don't apply here" is the sound a price makes when the cash flows can no longer justify it. Sometimes it is true — but only the durable cash flows, never the confidence of the telling, can show that.

Gather it into a single habit. Whenever you meet a valuation multiple, run three questions before you let it mean anything. Whose earnings am I dividing by — a normal year or a peak (or trough) year? For a cyclical, replace the snapshot profit with a through-cycle average so a boom cannot masquerade as normal. What is this multiple relative to the company's own history and a sober, cycle-averaged level — not a fixed ruler and not the euphoric number of the day? And how much of this multiple is the business's cash flows, and how much is the cycle's mood — earning power, or a re-rating that can be handed back?

Do that, and the two forces the cycle uses to fool you — swinging earnings and a stretching multiple — become things you can see through rather than be caught by. You will not be able to time when a dear multiple deflates or a cheap one recovers; that gravity has no clock. But you will stop reading a peak-earnings 5x as a bargain, stop mistaking a mood-driven re-rating for genuine progress, and stop being surprised when cheap gets cheaper and dear gets dearer for longer than seems possible. That is the whole of the skill: not predicting the multiple, but refusing to be fooled by the stage of the cycle that produced it.