The most common analytical error in sector work is not a wrong number. It is a right number computed with the wrong tool. Price-to-earnings is applied to a steel producer at the top of the cycle, enterprise value to EBITDA is applied to a bank, price-to-sales is applied to a low-margin distributor as if it were a software firm. Each of those calculations is arithmetically correct and analytically worthless, because the multiple was never designed to describe that kind of business.
A multiple is a compressed statement about three things: how fast a business can grow, how much capital it must consume to grow, and how certain that growth is. Choosing a multiple means choosing which part of the business you are willing to look at and which part you are agreeing to ignore. This article works through what each of the main multiples encodes, where it breaks, which sector it belongs to, and the specific trap that comes with it — including the cyclical earnings trap, which catches more experienced investors than any other single mistake in Indian sector analysis.
A multiple is a compressed discounted cash flow
Every multiple is shorthand for a full valuation. If you assume a business grows at a steady rate forever, the value of its equity is the cash it returns to shareholders, discounted at the return investors require. Rearranging that relationship gives you the multiple the business deserves.
That rearrangement is the whole game. It tells you that a higher multiple is justified by faster durable growth, by a lower required return (less risk), and by needing less capital to fund each rupee of that growth. Nothing else. When you see two companies in the same sector trading at different multiples, exactly one of those three variables is being priced differently — or the market is wrong about one of them.
This is why 'expensive' and 'cheap' are not analytical statements on their own. A multiple is an answer to a question. You have to know what question was asked before you can say whether the answer is high.
Justified P/E = payout ratio × (1 + g) ÷ (cost of equity − g) | Justified P/B = (ROE − g) ÷ (cost of equity − g)- g = the sustainable long-run growth rate of earnings
- ROE = return on equity; the profit generated per rupee of shareholders' funds
- Cost of equity = the return an investor requires for carrying this business's risk
- Payout ratio = the share of earnings returned as dividends or buybacks
Note — Read the justified P/B identity carefully: when a company's return on equity equals its cost of equity, the formula collapses to a price-to-book of 1. Everything above book value has to be earned by out-returning the cost of capital.
The three places P/E breaks
Price-to-earnings is the default multiple because earnings per share is the number every company reports. It is also the multiple that fails most often, and it fails in three distinct ways.
- Cyclicals — the denominator swings
- In metals, cement, sugar, shipping, chemicals and refining, earnings are the difference between two large and volatile numbers: realisation and input cost. A modest move in the spread produces an enormous move in profit. So the earnings figure you divide by is not a stable base — it is a snapshot of where the cycle happens to be.
- Lenders — leverage is the product
- For a bank or an NBFC, borrowing is not a financing decision that sits alongside the business; borrowing is the business. Deposits and wholesale funds are raw material. Interest expense is cost of goods sold. Any multiple that tries to strip out the effect of debt — which is what enterprise-value multiples do — deletes the operation itself.
- Asset-heavy and loss-making — the E is unusable
- A company midway through a large capital expenditure programme carries heavy depreciation and interest before the asset earns anything, so reported earnings understate the business. A genuinely loss-making company has a negative denominator, which makes the ratio meaningless rather than merely misleading.
The cyclical earnings trap, in detail
This deserves its own section because the mechanism is counter-intuitive and the loss it causes is large.
Work through the sequence. At the trough of a commodity cycle, prices are low, plants run below capacity and fixed costs are spread over few tonnes. Earnings collapse towards zero. Price falls too, but not to zero — the market knows the assets still exist. Dividing a modest price by a tiny earnings number gives a very high P/E. The stock screens as expensive.
Now run the cycle forward. Demand recovers, capacity utilisation rises, realisation improves, and because most of the cost base is fixed, almost every extra rupee of revenue falls to profit. Earnings do not double — they multiply. Price rises as well, but analysts extrapolate cautiously, so price rises by less than earnings. Dividing a much higher price by a hugely higher earnings number gives a very low P/E. The stock now screens as cheap, at the exact moment the earnings are least sustainable.
That is the trap. In a cyclical business, a low P/E is a signal about where you are in the cycle, not a signal about value. The old rule of thumb — that cyclicals are dangerous at low P/Es and interesting at high ones — is a description of this arithmetic, not a piece of contrarian folklore.
The fix is to stop dividing by a single year's earnings and start dividing by something that survives the cycle.
- 1
Estimate mid-cycle earnings
Take revenue or volume at a normal level of capacity utilisation and apply a margin drawn from the average of a full cycle, not the last four quarters. The output is a mid-cycle earnings figure you can divide into.
- 2
Cross-check with price to book
Book value in a commodity business is anchored to physical assets, so it moves far less than earnings. Price-to-book compresses the cycle out of the denominator and gives you a steadier reference point.
- 3
Anchor to replacement cost
Work out roughly what it would cost to build the same capacity today. If a company's enterprise value sits well below the cost of building its plants, new supply is uneconomic — which is itself a statement about the cycle.
- 4
Read the spread, not the profit
Most commodity industries have one headline spread the market actually trades — the gap between output price and the main input cost per unit. Track that series directly; it leads reported earnings.
Watch out — Never run a low-P/E screen across a commodity sector without also looking at where capacity utilisation, spreads and margins sit relative to their own long-run range. The screen will hand you peak-cycle earnings every time.
Price to book — why it is the standard for lenders
For a bank or an NBFC, the balance sheet is the business. Almost all of its assets are financial claims, carried at values that are far closer to fair value than a factory is. Its equity is the buffer that lets it lend at all, and regulation ties how much it can lend to how much capital it holds. So book value is a real, meaningful economic quantity — which is exactly what it is not for a software firm whose value sits in people and contracts.
The correct way to use price-to-book is never on its own. Pair it with return on equity. The justified price-to-book identity says the multiple should rise as return on equity rises above the cost of equity. So a lender consistently earning a high return on equity should trade above book, and one earning less than its cost of equity should trade below it. When you see a high price-to-book alongside a mediocre return on equity, or the reverse, you have found the question worth asking.
The trap with price-to-book for lenders is that book value is an accounting estimate, and the estimate depends on how honestly bad loans have been recognised. If provisions are thin, book value is overstated and the ratio flatters the bank. Which is why the ratio is only as good as the asset-quality disclosure underneath it.
- Gross and net non-performing assets (GNPA and NNPA) — how much of the loan book has stopped performing, before and after provisions.
- Provision coverage ratio — the share of bad loans already written down. A thin ratio means book value has further to fall.
- Restructured and stressed-but-standard exposures — loans not yet classified as bad but modified to avoid it.
- Capital adequacy — how much regulatory capital sits above the required minimum, which determines whether growth needs fresh equity.
- Adjusted book value — book value less the portion of bad loans not yet provided for. This is the number the ratio should really use.
Pro tip — Enterprise value multiples are meaningless for lenders. Enterprise value adds debt back to market capitalisation on the assumption that debt is financing. For a bank, debt is inventory.
EV/EBITDA — what it fixes and what it deliberately ignores
Enterprise value divided by EBITDA is the standard tool for capital-intensive operating businesses — telecom, cement, utilities, hotels, ports, hospitals. It fixes two real problems. Because enterprise value includes debt, it lets you compare two companies with completely different capital structures. And because EBITDA sits above interest, tax and depreciation, it lets you compare operating performance across different tax regimes, different depreciation policies and different vintages of asset.
What it ignores is the point. EBITDA — earnings before interest, tax, depreciation and amortisation — deliberately excludes the cost of the assets that produce it. For a business that must keep spending heavily just to stand still, that exclusion is enormous. A tower company, a cement plant and a hospital all consume capital continuously; EBITDA quietly pretends they do not.
There is also an accounting wrinkle worth knowing in India. Under Ind AS 116, operating leases are capitalised: the rent that used to sit in operating cost now appears partly as depreciation and partly as interest. That mechanically raises reported EBITDA for lease-heavy businesses such as organised retail, quick-service restaurants and hotels, without changing a rupee of cash. Comparing a lease-heavy company's EV/EBITDA against an owner-operator's, or against its own pre-transition history, without adjusting for this is a straightforward error.
- Enterprise value (EV)
- Market capitalisation plus total debt plus minority interest, less cash and cash equivalents. It is what it would cost to acquire the whole operating business, debt included.
- What EV/EBITDA answers
- How much am I paying for a rupee of operating cash generation, regardless of how the business is financed or taxed.
- What it hides
- Capital expenditure, working capital, the real economic cost of depreciation, and the interest burden of the debt sitting inside the enterprise value.
- The fix
- Look at EV to EBIT as well, which puts depreciation back in, and at free cash flow after maintenance capital expenditure. If the two tell different stories, the capex is the story.
Price to sales — the last resort, and its danger
When a business has no earnings — because it is early-stage, because it is investing ahead of revenue, or because margins are temporarily suppressed — price-to-sales is often the only ratio that computes. That is its entire justification. It is used because nothing better exists, not because it is informative.
The danger is that revenue carries no information about whether profit is achievable. A distribution business and a software business can report identical revenue and end up with wildly different profits, so the same price-to-sales ratio means completely different things for the two. Applying a peer's price-to-sales to a company with a structurally different margin is not comparison; it is a category error.
The only defensible way to use it is to convert it into an implied earnings multiple. Divide the price-to-sales ratio by the net margin you believe the business can sustain at maturity. That gives you the price-to-earnings ratio you are implicitly paying. Now you have to defend an actual assumption — the margin — instead of hiding behind a ratio.
Implied mature P/E = (Price ÷ Sales) ÷ sustainable net margin- Sustainable net margin = the net profit margin you believe the business can hold at scale
- State the margin assumption explicitly — it is doing all the work in this calculation
- If the implied P/E looks absurd, the price-to-sales ratio was hiding it, not solving it
Replacement cost and capacity multiples for commodity businesses
For businesses whose output is a commodity, value is ultimately anchored to the cost of creating new supply. Nobody builds a new plant unless the economics of building beat the economics of buying, so the cost of new capacity sets a long-run reference for the value of existing capacity.
This produces a family of per-unit multiples that practitioners use alongside the accounting ratios: enterprise value per tonne of cement or steel capacity, enterprise value per megawatt of generation, enterprise value per hotel room, enterprise value per subscriber in telecom, enterprise value per acre in land-heavy businesses. Each divides what the market is paying by a physical unit of capacity.
The signal they generate is about supply. When enterprise value per unit sits well below replacement cost, building new capacity destroys value, so rational operators stop building. Supply growth stalls, and eventually demand catches up — which is how commodity cycles turn. When it sits well above replacement cost, new capacity is profitable to build, capital floods in, and future supply is being created that will compete with the existing asset.
The trap is that capacity is not the same as earning power. A tonne of capacity in a location with cheap limestone, captive power and short freight distance to its market is worth far more than the same tonne stranded far from demand. Per-unit multiples treat them as identical.
Example — If listed capacity in an industry is being valued below what a new plant costs to build, that is a statement about future supply discipline, not a claim that the shares are mispriced. The two are often confused.
The mapping table — sector type, multiple, reason, trap
The table below is the practical summary. Read the last column as carefully as the second: knowing what a multiple hides is more useful than knowing what it shows.
| Business type | Primary multiple | Why it fits | The trap |
|---|---|---|---|
| Banks, NBFCs | P/B with ROE | Leverage is the product; book is the capital base | Book is overstated if provisioning is thin |
| Cyclicals — metals, cement, sugar | EV/EBITDA, P/B | Earnings swing too hard to divide by | P/E is lowest exactly at the earnings peak |
| Capital-intensive — telecom, utilities | EV/EBITDA, EV per unit | Capital structures differ too much for P/E | EBITDA ignores the capex that sustains it |
| Consumer staples, franchise businesses | P/E, free cash flow yield | Earnings are stable, cash-backed and repeatable | Durability may already be fully priced in |
| IT services, asset-light exporters | P/E, EV/EBIT | Little debt, real earnings, low capital intensity | Currency and one client can move the earnings |
| Loss-making or pre-margin | P/S with a target margin | There is no earnings figure to divide by | Revenue says nothing about whether margin arrives |
| Life insurance | Price to embedded value | Profit emerges over decades, not quarters | Embedded value rests on long-dated assumptions |
| Holding companies, conglomerates | Sum of the parts | Consolidated earnings blend unlike businesses | The holding-company discount can persist for years |
A premium multiple can be entirely rational
The justified-multiple identity says something uncomfortable for value-oriented investors: a higher multiple is the correct price for a better business, and paying it is not a mistake.
Consider what drives the number. A business that earns a high return on incremental capital needs to reinvest less to grow by a given amount, so more of its profit is genuinely free. A business with a long runway can compound that advantage for many years before growth fades. A business with predictable demand has a lower cost of equity because investors require less compensation for uncertainty. All three push the justified multiple up, and all three are real economic characteristics, not sentiment.
The mirror image is the harder lesson. A low multiple is not a thesis. For a cheap stock to make money, something has to change — earnings inflect, capital allocation improves, a structural drag is removed, the market re-rates the sector. If you cannot name the specific thing that closes the gap, you have not identified value; you have identified a business the market has decided is worth less, and it may be right. A multiple can stay low for a decade while book value erodes underneath it.
Watch out — Cheapness is a starting observation, not a conclusion. Write down the specific mechanism that would cause the multiple to re-rate. If you cannot, the discount is an opinion about the business, not an opportunity.
Compare against the right benchmark: its own history, its own sector
Comparing a company's multiple to the broad market index is almost always meaningless. The index is a weighted blend of unrelated business models with different capital intensity and different growth. There is no reason a cement plant and a software exporter should trade at the same multiple, so the comparison generates noise.
There are only two comparisons that carry information. The first is against the company's own history: plot the multiple over several years and see where today sits within its own range. If it is at the top of its range, ask what has improved — return on capital, growth visibility, balance sheet — to justify it. If nothing has, you are paying more for the same business. The second is against its own sector peers, adjusted for the differences you can actually identify: growth rate, return on capital, leverage, and the quality of the earnings.
One caution on the historical comparison. A multiple range is only meaningful if the business is still the same business. If a company has changed its mix — moved from commodity products to specialty, from lending to fee income, from services to products — its historical multiple range describes a company that no longer exists.
- Plot the multiple over a full cycle, not the last three years, and mark where today sits in that range.
- Note whether the business mix has changed enough to invalidate the historical range.
- Build a peer set from the same industry, then adjust for growth, return on capital and leverage before comparing.
- Check whether the whole sector has re-rated. A company cheap against its peers inside an expensive sector is not cheap.
- Cross-check every multiple with a second one that uses a different denominator. Agreement is information; disagreement is where the work is.
What a multiple can never tell you
A multiple is a price expressed as a ratio. It is a summary of what other people currently believe, and it will always be a lagging description of consensus rather than a forecast.
It cannot tell you whether the accounting underneath it is honest. It cannot tell you whether the promoter will allocate the next decade of cash flow well or badly. It cannot tell you whether a competitor is about to commission capacity that removes the pricing power the multiple assumes. It cannot tell you whether a regulator is about to change the economics of the industry. Every one of those questions is answered by reading filings, annual reports, capacity announcements and policy documents — not by comparing ratios.
So use the multiple for what it is good at: telling you what is being assumed, and forcing you to state whether you agree. When you find a multiple you think is wrong, the useful next step is never to buy or sell on it. It is to write down which of the three variables — growth, return on capital, or risk — you think the market has mispriced, and then go and find evidence for or against that specific claim.
Key points
Justified P/E = payout ratio × (1 + g) ÷ (cost of equity − g). Justified P/B = (ROE − g) ÷ (cost of equity − g). Enterprise Value (EV) = market capitalisation + total debt + minority interest − cash and equivalents. Implied mature P/E = (Price ÷ Sales) ÷ sustainable net margin.
Example — A commodity producer reports its best-ever profit after a period of strong realisations, and its P/E falls to the lowest level in years because earnings rose faster than the share price. A screen ranks it as one of the cheapest names in the market. Recompute the same P/E using mid-cycle earnings — volumes at normal utilisation, margins at the average of a full cycle — and the multiple is no longer low at all. The stock did not become cheap; the denominator became temporarily large.
Pro tip — Never quote one multiple in isolation. Compute a second one that uses a different denominator — earnings against book, or EBITDA against free cash flow after maintenance capex. When two multiples built on different denominators disagree, the disagreement is the analysis. Investigate why before you form a view.
Warning — Applying the wrong multiple produces a confident, precise and completely wrong answer, which is far more dangerous than no answer. Enterprise-value multiples on a lender, price-to-earnings on a peak-cycle commodity producer, and price-to-sales on a business whose margin structure you have not stated are the three errors that recur most often. Check that the multiple matches the business model before you compute anything.
Frequently asked questions
Why does a cyclical stock look cheapest when it is most risky?
Because the earnings denominator peaks before the price does. At the top of a commodity cycle, high realisations and full capacity utilisation push profit up faster than the share price, since the market discounts that the profit is temporary. Dividing a high price by an even higher earnings figure produces a very low price-to-earnings ratio. The multiple is telling you where you are in the cycle, not what the business is worth.
Why can't I use P/E for banks?
You can compute it, and it is a reasonable cross-check, but it is not the primary lens. A lender's earnings depend on how aggressively it recognises bad loans, so reported profit can be managed through provisioning in a way that a manufacturer's cannot. Price-to-book paired with return on equity is more robust because book value is the regulated capital base the lender actually operates on. Enterprise-value multiples, by contrast, are meaningless for a bank — they treat debt as financing when for a bank it is raw material.
What is the difference between P/E and EV/EBITDA?
P/E looks only at equity: it divides the share price by earnings after interest and tax, so it is affected by how much debt a company carries and what tax rate it pays. EV/EBITDA looks at the whole business: enterprise value includes debt and excludes cash, and EBITDA sits above interest, tax and depreciation. That makes EV/EBITDA better for comparing companies with different capital structures, and worse for any business where capital expenditure is a large ongoing cost, because EBITDA ignores it entirely.
How do I calculate enterprise value?
Start with market capitalisation — share price multiplied by shares outstanding. Add total debt, both short and long term. Add minority interest if subsidiaries are consolidated but not wholly owned. Subtract cash and cash equivalents, since an acquirer would get that cash. The debt, cash and minority interest figures come from the most recent balance sheet in the company's filings, so the number is only as current as the last reported quarter.
Is a stock with a lower P/E than its sector always the better value?
No. A discount to the sector is a question, not an answer. It usually reflects something the market has identified — lower return on capital, higher leverage, weaker growth, governance concerns, or a business mix that is structurally less attractive. The analytical task is to work out which of those is being priced and whether you disagree with the market's assessment. If you cannot name a specific reason the gap should close, the discount is likely to persist.
How do I compare a company's valuation to its own history correctly?
Plot the multiple across a full business cycle rather than the last few years, and mark where the current reading sits within that range. Then check whether the company is still the same business — if the revenue mix, capital intensity or return on capital has changed materially, the historical range describes a company that no longer exists. Finally, check whether the entire sector has re-rated, because a stock at the middle of its own range inside a sector at the top of its range is not neutral.
What does EV per tonne or EV per megawatt actually tell me?
It tells you what the market is paying for one physical unit of capacity, which you can then compare against roughly what that unit would cost to build today. If the market value sits well below replacement cost, building new capacity is uneconomic, so supply growth should slow. If it sits well above, new capacity is profitable to build and future competition is being created. The limitation is that it treats all capacity as identical, when location, input access and freight distance make some units far more profitable than others.