The P/E ratio and everything it fails to tell you

What the ratio actually measures

The price-to-earnings ratio is one number divided by another. Take the market price of one share, divide it by the earnings attributable to one share over some period, and you have it. A share trading at 60 against earnings of 4 per share gives a P/E of 15.

There are three ways to read that result, and they are all the same arithmetic viewed from different angles.

The first is the literal one. You are paying 15 units of price for each unit of annual earnings.

The second is the payback framing. If earnings stayed frozen at 4 forever and were somehow all returned to you, it would take fifteen years to get your money back. Nobody believes earnings stay frozen, which is precisely why this framing is useful. It forces the question of what has to change for the number to make sense.

The third is the reciprocal, and it is the most underrated. Flip the ratio and you get the earnings yield. A P/E of 15 is an earnings yield of about 6.7 per cent. A P/E of 40 is 2.5 per cent. A P/E of 8 is 12.5 per cent. Expressed this way, the multiple sits on the same scale as every other yield you might be considering, and comparisons become concrete rather than atmospheric. "Forty times earnings" sounds abstract. "Two and a half per cent, before any growth" is a sentence you can argue with.

That is the entire mechanism. Everything else in this article is about what the two inputs quietly assume.

Three different numbers wearing the same name

When someone quotes a P/E, they are quoting one of at least three calculations, and the gap between them is often larger than the gap between two different companies.

Trailing

The trailing P/E divides the current price by earnings already reported, usually the last four reported quarters or the last full financial year. Its virtue is that the denominator is audited or at least filed, and you can trace it to a document. Its weakness is that it is a rear-view mirror. If something material changed after the reporting date, a factory closure, a large acquisition, a currency move, the denominator describes a company that no longer exists in that form.

Forward

The forward P/E divides price by an estimate of future earnings, typically the next twelve months or the next full year. It is more relevant in principle and more fragile in practice, because the denominator is now someone's forecast. Two things follow. First, a forward P/E is only as good as the forecast inside it, and forecasts cluster. Second, forward multiples almost always look lower than trailing multiples, because most forecasts assume growth. A company that looks expensive on trailing earnings and reasonable on forward earnings has not become cheaper. It has been assumed to become cheaper.

Adjusted or normalised

The third version uses a company-defined earnings figure, described as adjusted, underlying, core or normalised. Accounting standards define how earnings per share is presented, including the requirement to show both a basic and a diluted figure that accounts for options and convertible instrumentsSource^source. They do not define "adjusted earnings". That is a management construction, and it is where the argument lives.

Adjustments are not automatically illegitimate. Stripping out a genuinely one-off legal settlement or the accounting noise from a single disposal can produce a more representative number. The problem is the recurring exception. If a company excludes restructuring charges every year for six years, restructuring is not an exception, it is an operating cost with a euphemism attached. Supervisors in several markets expect these alternative measures to be reconciled back to the audited figures and not presented more prominently than them, precisely because the adjustments are the contested partSource^source.

Before comparing two P/E ratios, confirm they are the same species. Comparing one company's adjusted forward multiple with another's trailing statutory multiple is not analysis, it is a category error.

The denominator is an opinion

Price is close to a fact. At any moment there is a bid and an ask, and a trade either happened or it did not. Earnings is a different kind of object. It is the output of a long chain of judgements, each defensible, each with a range.

The main judgements that move reported earnings include:

• **Depreciation and amortisation assumptions.** How long a machine, a building or an acquired customer list is assumed to last determines how much cost lands in this year rather than the next eight.
• **Revenue recognition timing.** When a multi-year contract is judged to have been earned changes which year the profit appears in, without changing a single cash flow.
• **Provisions and impairments.** Expected credit losses, warranty provisions and asset write-downs are estimates about the future recorded as costs today.
• **Capitalisation choices.** Development spending treated as an asset flatters this year's earnings and burdens future years with amortisation. Treated as an expense, it does the reverse.
• **Tax.** Effective tax rates move with geography, incentives and one-off settlements, and a lower rate raises earnings per share without any operating improvement.

None of this makes earnings meaningless. It makes earnings a modelled quantity rather than an observed one, which matters when you are dividing a market price by it and treating the answer as precise to one decimal place.

Because listed companies file periodic statements with their securities regulator, you can always go back to the filed document rather than relying on a summary figure of unknown constructionSource^source.

Worked example, two companies on the same multiple

Both companies below are hypothetical. Both trade at exactly 15 times last year's earnings. Assume both have 100 million shares and a share price of 60, so both have a market value of 6 billion and both reported earnings of 400 million, or 4 per share.

**Company A** makes industrial fasteners. Revenue has grown roughly in line with regional construction for a decade. Operating margin has sat between 11 and 13 per cent every year for ten years and was 12 per cent last year. Capital spending runs close to depreciation. Cash from operations has tracked reported profit closely.

**Company B** makes a commodity chemical. Its margin has ranged from 3 per cent in bad years to 18 per cent in good ones, and last year was 18 per cent. It has spent heavily on a new plant, so capital expenditure ran at roughly double depreciation, and it carries meaningful debt.

Now normalise. Suppose Company B's ten-year average margin is 9 per cent, half of last year's. Applying that mid-cycle margin to the same revenue base cuts earnings from 400 million to roughly 200 million, or 2 per share. At an unchanged price of 60, the multiple on mid-cycle earnings is 30, not 15.

Nothing has been discovered about the future. No forecast has been made. All that happened is that the denominator was replaced with a figure more representative of the business across a cycle, and the same headline multiple turned out to describe two situations roughly twice as far apart as the ratio suggested.

The lesson is not that cyclical companies are bad. It is that a P/E computed on a peak-margin year and a P/E computed on a typical year are not comparable quantities, even when they are numerically identical.

What the ratio ignores entirely

The P/E ratio compares the value of the equity with the earnings that belong to the equity. It says nothing about the capital structure that sits behind both.

Consider two more hypothetical companies, each with a market value of 6 billion and earnings of 400 million, so each on 15 times.

• **Company C** holds 2 billion of net cash and no debt.
• **Company D** carries 2 billion of net debt.

Strip out the cash and Company C's operating business is being valued at 4 billion. Add the debt and Company D's operating business is being valued at 8 billion. On enterprise value, the second is twice the first, while the P/E ratio reports them as identical. The ratio cannot see this, because the balance sheet never enters the calculation.

The same blindness applies to several other items:

• **Dilution in waiting.** Options, convertible bonds and performance shares increase the future share count. The diluted earnings per share figure captures part of this, which is one reason the standards require it to be presentedSource^source, but instruments issued after the reporting date will not be there.
• **Minority interests and associates.** Consolidated earnings may include profits that partly belong to other shareholders, or exclude the economics of significant stakes held at cost.
• **Pension and lease obligations.** These are fixed claims on future cash that behave like debt without always being labelled as such.
• **The quality of the earnings.** Two companies can report identical profit while one collects cash promptly and the other funds growing receivables.

The cheap trap and the expensive trap

A low multiple is not cheapness

A low P/E is a statement about expectations, not about value. The market is not offering a discount out of generosity. It is pricing something, and your job is to find out what. Common explanations include earnings expected to fall, a business in structural decline, heavy debt raising the risk that equity holders absorb losses, governance concerns, capital that is trapped in a jurisdiction or subsidiary, or accounting that is not trusted.

Sometimes the pessimism is wrong, and that is where returns come from. But "it is on eight times" is the beginning of the research, not the end of it.

A high multiple is not expensiveness

Symmetrically, a high P/E can be rational. A multiple embeds expected growth, the durability of that growth, how much capital must be reinvested to achieve it, and the risk attached to the whole stream. A business that can grow earnings substantially for many years while reinvesting little deserves a higher multiple than one that must spend heavily to stand still. Arithmetically, a company growing earnings at a rate that doubles them in five years is on half its current multiple in five years at an unchanged price.

The honest way to use a high multiple is to invert it. Ask what growth rate, sustained for how long, would be needed to justify the price at a return you would accept. Then ask whether anything you know about the industry, its competition and its capital needs makes that plausible.

Where the ratio stops working

Some cases are not edge cases so much as territory where the tool does not apply.

• **Loss-making companies.** With negative earnings the ratio is undefined or meaningless. Data providers usually display a blank or a nonsensical figure.
• **Cyclicals.** The ratio inverts. Cyclical companies look cheapest at the peak, when earnings are inflated, and most expensive at the trough, when earnings are depressed but the recovery is closest. Reading a cyclical P/E naively gets the timing exactly backwards.
• **Banks and insurers.** Earnings depend heavily on provisioning judgements and on leverage that is intrinsic to the business model. Price to book value alongside return on equity carries more of the analytical weight, and capital adequacy matters more than either.
• **Asset-heavy or investment-holding companies.** Where value sits in property or in stakes in other businesses, reported earnings can be dominated by revaluations that have nothing to do with operations.
• **Companies mid-restructuring, mid-merger or recently listed.** The reported year describes a transition, not a run rate.
• **Heavy serial issuers of shares.** Earnings per share can fall while total earnings rise, or vice versa, purely through share count.

Five questions to ask of any P/E

This is an interrogation, not a scoring system. It takes about ten minutes with the filed statements open.

1. **Which E is this?** Trailing, forward or adjusted, statutory or company-defined, basic or diluted. Write it down. Half of all bad multiple comparisons die here.
2. **Is this E representative?** Compare last year's operating margin with the average of the last five to ten years. If the gap is large, recompute the multiple on the average margin and use both numbers.
3. **What is the balance sheet doing to this?** Express net cash or net debt as a percentage of market value. If it is more than a small fraction, the P/E is describing a different thing than you think, and enterprise-value based measures will serve you better.
4. **What does this multiple imply?** Invert it into an earnings yield. Ask what growth would have to follow for that yield to reach a return you would accept, and over how many years.
5. **Compared with what?** Compare against the company's own history, against genuine peers computed the same way, and against the plain yield alternatives available to you. A multiple with no comparison set is a number without a sentence around it.

If a question cannot be answered from published documents, record that as an unknown rather than filling it with an assumption.

What belongs next to it

The P/E is a summary statistic, and summaries are most useful when surrounded by the things they compress. In practice the companions that add the most information are cash conversion, which tests whether the earnings are turning into money; return on capital, which tests whether growth is worth having; and some enterprise-value measure, which restores the balance sheet to the picture.

Used that way, the ratio does real work. It is fast, it is universally available, it is comparable within a homogeneous group, and tracking its movement over time is a clean way to watch expectations change. What it cannot do is tell you whether something is cheap, because cheapness is a statement about value received for price paid, and the ratio contains no information about the value side beyond a single accounting year that may or may not be typical.