A share trades at ₹1,200 and earned ₹12.30 last year. It is on 97.6 times earnings.
Somebody will tell you that is expensive. Somebody else will point out that a business with a long enough runway can carry a multiple in that region for years on end, and that the people calling it expensive have been calling it expensive the whole time.
Both remarks are fair and neither is analysis. A multiple is not a verdict. It is a sentence about the future, and the only useful response is to work out what the sentence says and decide whether you believe it.
What is actually inside the number
For a business in a steady state, the multiple a company deserves rises with its growth rate, rises with the return it earns on the capital it reinvests, and falls with the return investors require.
Two of those three are properties of the business and the third is a property of the market. That is the whole of it.
The middle term is the one most often forgotten. A company earning 30% on reinvested capital needs to plough back far less to grow at a given rate than one earning 12%, so more of its profit reaches the operator, so it deserves a higher multiple at the same growth. Two companies both growing 20% with the same multiple can be worth very different amounts.
Which means a high multiple is a statement that the market expects strong growth at good returns for a long time. Whether that expectation is reasonable is the analyst's question. Whether the number is large is not.
Trailing, forward, and the gap between them
The trailing multiple divides today's price by earnings actually reported over the last twelve months. Its virtue is that the denominator is a fact.
The forward multiple divides by expected earnings for the coming year. At ₹1,200 with expected earnings of ₹15.50, the forward multiple is 77.4 against a trailing 97.6.
Markets price the future, so the forward figure is generally the more relevant one, and it carries somebody's forecast inside it. When that forecast is consensus, the multiple embeds the average sell-side view, which is documented to be too optimistic early in a downturn and too cautious after one.
The gap between the two tells you how much growth is assumed. A wide gap means the price depends on the earnings arriving.
Which earnings, exactly
The denominator is where most of the damage is done, and four questions decide it.
Basic or diluted? A company earning ₹615 crore with 50 crore shares reports basic earnings of ₹12.30. If outstanding options and convertibles would add 3 crore shares, diluted earnings are ₹11.60, and the multiple rises from 97.6 to 103.4. Use the diluted figure by default. Where a company's release quotes only the basic number and the two are far apart, that was a choice.
Attributable to whom? For a group, the numerator of earnings per share must be profit attributable to the parent's owners, after the share belonging to minority holders in partly-owned subsidiaries. In Indian groups with large partly-owned operating companies, that gap can exceed a fifth of the reported number.
Standalone or consolidated? For most Indian groups the operating businesses sit in subsidiaries, so the consolidated figure is the one to use, and screeners mix the two routinely.
Reported or adjusted? A company reporting ₹615 crore after an ₹180 crore exceptional charge earned ₹795 crore before it. Which belongs in the multiple depends entirely on whether the charge recurs, and the test is arithmetic: add up every item labelled exceptional across five or six years. If the total is large and consistently negative, those charges are a cost of running this business.
The four situations where it misleads badly
Cyclicals. This is the classic and it has ended more portfolios than any other single mistake in relative valuation. A steel or cement company at the top of its cycle reports peak earnings, so the trailing multiple is at its lowest exactly when the stock is most dangerous. At the bottom, earnings collapse and the multiple looks absurd, which is usually the opportunity.
The fix is to use mid-cycle earnings. A producer earning ₹1,200, ₹400, ₹800, ₹1,600 and ₹600 crore over five years has a mid-cycle figure of about ₹920 crore, and the multiple computed on that number is the one worth comparing. The better version is built rather than averaged: normal capacity utilisation multiplied by a normal spread, defended against the industry's own history.
Companies with a lot of debt. Earnings are measured after interest, so a leveraged company's earnings are a small residual on top of a large obligation. Small changes in operating profit produce large changes in earnings, and the multiple swings for reasons that have nothing to do with the business. An enterprise-value multiple is the right lens where capital structures differ.
Companies with unusual tax positions. A firm enjoying a tax holiday reports earnings that will fall when the holiday ends, without anything happening to the business. The tax reconciliation note discloses it and usually names the expiry.
Companies with large non-operating income. A company reporting ₹840 crore of pre-tax profit of which ₹210 crore is treasury income is really an operating business plus a bond portfolio. Applying one multiple to the total values the bond portfolio at the same rate as the business.
Turn it upside down
Divide earnings by price instead, and you get the earnings yield. A stock on 20 times has a 5.0% earnings yield.
The inversion changes how the number reads. A yield can be compared directly against a government bond yield, against the company's own cost of debt, or against any other investment, and the comparison exposes what the multiple implies about the required return.
At a 7% ten-year government bond and a 5% earnings yield, the equity is priced below the risk-free rate on current earnings, which is only rational if those earnings are going to grow substantially. That sentence is more useful than "the P/E is 20".
What PEG does and does not do
Price to earnings divided by the expected growth rate in percentage points. A company on 28 times growing at 22% has a PEG of 1.27.
The rule of thumb that one is fair value has no theoretical basis. It ignores the cost of capital entirely, ignores how long the growth lasts, and ignores the return at which the growth is earned, which is the variable that decides whether growth is worth anything at all.
Treat it as a screening shorthand. It is not a valuation.
The comparison that matters more than the level
A multiple is only meaningful against something: the company's own history, its peers, or the market.
Against its own history over a decade tells you whether today's expectation is higher or lower than the one the market has usually held, and any large deviation needs a reason.
Against peers requires the peer set to be genuinely comparable, meaning similar growth, similar returns on capital and similar cyclicality, rather than merely sharing an industry label. Five to eight names, median rather than mean, and computed identically for all of them.
Against the market tells you what premium the business commands, which is a statement about perceived quality and durability.
When to reach for something else
EV/EBITDA where capital structures differ or where depreciation policies diverge enough to make earnings incomparable. It is the workhorse for capital-heavy sectors and it hides exactly what it excludes, so read it beside capital expenditure to sales.
Price to book against return on equity for banks and lenders, where the balance sheet is the business and earnings depend on provisioning judgements.
Free cash flow yield where you want a measure that is harder to manipulate than earnings, averaged across a cycle because capital spending is lumpy.
Price to embedded value for a life insurer, where accounting earnings badly misdescribe a business whose value is created at the point of sale and recognised over thirty years.
Cyclically adjusted earnings for an index or a mature cyclical, though applying a ten-year average to a company that has quadrupled produces a number that means nothing.
The two-minute routine
Compute the multiple on diluted, consolidated, attributable earnings, adjusted for anything genuinely non-recurring.
Compute it forward as well as trailing, and note which forecast you used.
Put it beside the company's own ten-year range and beside a peer median.
Then say, in one sentence, what growth and what return on capital the multiple implies, and whether the business has ever delivered them.
If that sentence cannot be written, the multiple has not been understood, and quoting it is a description of a calculation rather than a view about a company.
What a multiple looks like across a decade
The single most useful chart in relative valuation is not a peer comparison. It is a company's own multiple, plotted for ten years, with its earnings on the same panel.
Three patterns appear and each means something different.
A multiple that has stayed inside a band while earnings compounded is a business the market has consistently believed in, and the returns to shareholders came from the earnings rather than from a re-rating. That is the healthiest shape, because it does not depend on anyone paying more for the same rupee of profit.
A multiple that expanded steadily while earnings grew slowly is a re-rating, and re-ratings are borrowed. Whatever the market decided to pay more for can be decided again in the other direction, and the fall is usually faster than the rise.
A multiple that collapsed while earnings held is either an opportunity or a signal that the market has stopped believing the earnings. Distinguishing those two is what the accounting work is for.
Ten years of data, one chart, and it reframes the question.
The Indian complications
Three of them, and all three affect the denominator.
Standalone against consolidated. For most Indian groups the operating businesses sit in subsidiaries, so the consolidated figure is the right one. Screeners mix them, and a company whose standalone earnings are a fraction of consolidated will appear on a wildly different multiple depending on which was used.
Attributable profit. Consolidated profit includes the share belonging to minority holders in partly-owned subsidiaries. Earnings per share must use profit attributable to the parent's owners, and in groups with large partly-owned operating companies that difference can exceed a fifth of the headline.
Tax transitions. A company that moved to a lower corporate tax regime reports a step change in earnings with nothing happening to the business, which makes any multiple spanning the transition year misleading in both directions.
None of these is subtle once you know to check. All three are invisible on a screener.
The question the multiple is standing in for
Every valuation multiple is a compressed answer to the same question: how much am I paying for a rupee of this company's future profit, and how confident am I that the rupee arrives?
The multiple compresses that into one number and throws away the reasoning. Which is fine as shorthand between people who have done the work, and useless as a substitute for doing it.
So the discipline is to expand it back out. State, in a sentence, what growth and what return on capital the current multiple implies. Then say whether this business has ever delivered them, and what would have to be true for it to keep doing so.
If that sentence cannot be written, the number has not been understood.
What to say instead of quoting the number
The useful sentence about a multiple is never the multiple. It is the expectation the multiple encodes, stated in terms a person who runs a business could argue with. A company on 60 times forward earnings is being priced for something specific: a rate of growth, at a level of return on capital, sustained for a number of years. Work out what that combination is, write it in one sentence, and hand it to somebody who knows the industry. They will not have an opinion about whether 60 times is expensive, because nobody does. They will have a very quick opinion about whether a business of that kind can grow at that rate for that long at those returns, and that opinion is the analysis you were looking for. The multiple was only ever the shorthand.
A high multiple is a claim, not a price.
A low one is a different claim.
Both need the same work.