Price-to-Earnings Ratio (P/E)

Reviewed 12 August 2026 · Sourced from the SEC's investor education pages, Regulation G and the SEC staff's non-GAAP interpretations, FINRA Rule 2241, and the published work of Campbell, Shiller, Siegel, Asness and Easterwood

What the ratio is actually pricing

Two companies, same industry, same screener, same afternoon. One row says P/E 12, the other says P/E 34. Nothing on the screen says whether either number used earnings already reported or earnings an analyst expects — or whether those earnings are the audited ones in the filing or the flattering ones in the press release. Four defensible numbers exist for each company. The screener printed one and did not say which.

The SEC's investor education glossary puts it the same way: the ratio “is calculated by dividing the current stock price by the current earnings per share,” where “earnings per share are calculated by dividing the earnings for the past 12 months by the number of common shares outstanding.” The same fraction works a level up, as total market value over total profit, which is why market cap and P/E are two readings of one thing.

The framing that makes it click

A P/E is the price of one dollar of annual earnings. Not a rating, not a grade, not a signal — a price, in dollars per dollar-per-year.

Flip it and you get a yield

Almost everything else in finance is quoted as a percentage per year, so invert the ratio.

Northbay: $3.00 ÷ $60.00 = 5.0%, which is 1 ÷ 20. A P/E of 10 is a 10% earnings yield, 25 is 4%, 50 is 2%. The earnings yield is the form that lets a stock sit beside a bond, because both are finally percentages per year.

The numerator is a fact and the denominator is an opinion

This is the honest core of the page, and almost nothing written about P/E says it plainly.

The price is a fact. It came out of a continuous auction, it is timestamped, it is observable to the penny, and nobody's judgment produced it. Two people looking at the same stock at the same instant see the same price.

The earnings are not a fact in that sense. Earnings per share is the bottom line of an income statement prepared under US generally accepted accounting principles, and GAAP is a rulebook that requires estimates, not a measuring instrument. Inside it, honest management makes choices that move reported profit:

• Revenue recognition timing — when control transfers, how refunds are estimated, how a multi-year contract is spread across periods.
• Depreciation and amortization — a longer assumed useful life means a smaller annual charge and higher reported earnings from the identical asset.
• Reserves and allowances — expected credit losses, warranty reserves, inventory write-downs. Each is a forecast that lands in this period's profit.
• Write-off and impairment timing — not whether a bad asset is bad, but which quarter that gets recognized in.
• Capitalize or expense — a cost released over years lifts this year's earnings against the same cost expensed now.

None of that is fraud. Judgment inside GAAP is legal, disclosed in the accounting-policy notes, and unavoidable. But it means two honest managements running identical businesses can report different earnings, and therefore different P/E ratios at identical prices.

Two consequences. Cash is harder to flatter than profit — harder, not impossible — which is why cash-based measures sit beside earnings-based ones on any serious screen. Do not mistake harder for authoritative: free cash flow is itself a non-GAAP measure with no standard definition, which is the most important fact about it. Reading the income statement against the cash flow statement is the real skill, and it is what Stages 2 and 3 of Fundamental Analysis are for.

And the share count under EPS carries the same ambiguity as the share count under market cap. Diluted EPS uses a larger count than basic, so it is the smaller EPS and the higher P/E. That whole problem — issued, treasury, basic, diluted — is worked through on market capitalization and applies here unchanged.

Trailing, forward, and the one nobody labels

There are two common P/E ratios, they answer different questions, and the quoted number usually does not say which it is.

Analyst estimates lean optimistic, and this is documented

The forward P/E is only as good as the estimate under it, and the direction of the error is not random. Easterwood and Nutt, “Inefficiency in Analysts' Earnings Forecasts: Systematic Misreaction or Systematic Optimism?”, Journal of Finance 54(5), 1999, found that analysts “underreact to negative information, but overreact to positive information” — a pattern they describe as consistent with “systematic optimism in response to information.”

Regulators treat the tilt as real too. The SEC's investor publication on analyst recommendations warns that investors “should not rely solely on an analyst's recommendation” and should understand “the potential conflicts of interest analysts might face.” And FINRA Rule 2241(c)(2) requires a firm to print, in every research report carrying a rating, the percentage of its rated securities it assigns to buy, hold and sell, plus the share of companies in each bucket for which it has done investment banking in the previous 12 months. A disclosure rule of that shape does not get written unless the distribution is worth looking at.

Practical version: an unlabeled P/E is probably trailing and worth checking, and a forward P/E is a price divided by somebody's opinion about the future.

GAAP earnings, adjusted earnings, and why Regulation G exists

Most large companies report earnings twice on the same day: the GAAP figure that goes into the filing, and an “adjusted,” “pro forma” or “non-GAAP” figure in the press release, usually in a larger font and nearly always higher. The typical add-backs are stock-based compensation, restructuring, acquisition-related amortization, impairments, litigation settlements, and items labeled one-time that turn up again next year.

The rule that makes this checkable

Regulation G, 17 CFR § 244.100, requires that whenever a company publicly discloses a material non-GAAP financial measure it must also give “a presentation of the most directly comparable financial measure calculated and presented in accordance with Generally Accepted Accounting Principles” and “a reconciliation (by schedule or other clearly understandable method)” between the two. It also forbids a presentation that “contains an untrue statement of a material fact or omits to state a material fact necessary in order to make the presentation of the non-GAAP financial measure…not misleading.” Congress ordered the rule: Sarbanes-Oxley § 401(b), codified at 15 U.S.C. § 7261(b).

The SEC staff has gone further. Its non-GAAP interpretations state at Question 100.04 that an adjustment changing “the recognition and measurement principles required to be applied in accordance with GAAP would be considered individually tailored and may cause the presentation of a non-GAAP measure to be misleading.” Question 102.10 covers undue prominence — putting the non-GAAP figure first, styling it in bold or larger type, or discussing it without a similar discussion of the GAAP figure.

One note, because it is the largest add-back at most technology companies. Stock-based compensation is a real cost. Adding it back does not make it vanish; it reappears as shares, enlarging the count in the denominator of EPS over time. A P/E excluding it is measuring a company that pays its people in something free, and no such company exists.

When the ratio means nothing, which is often

P/E gets quoted as though it always has content. It does not. Four ordinary situations leave the number undefined or actively misleading, and between them they cover a large share of listed companies.

1. Negative earnings: there is no P/E

If a company lost money over the period, EPS is negative and so is the ratio. A negative P/E is not a high P/E and not an expensive stock. It is a fraction that has stopped meaning anything — “the price of one dollar of earnings” presupposes a dollar of earnings.

Careful data providers print a blank, an em dash or N/A rather than a number — the discipline this site applies everywhere, since on the Markets pages a metric that cannot be computed renders as an em dash, never as zero. If you see a negative P/E, the honest reading is “this company is not profitable,” and you go look at revenue, cash burn and the balance sheet instead.

2. Near-zero earnings: the ratio explodes

As the denominator approaches zero the ratio approaches infinity, so a company scraping past break-even generates a spectacular P/E carrying no information at all.

Our own arithmetic, on a $40 price. The shape is the point: between EPS of five cents and minus five cents the P/E travels from 800 to nothing to −800, and the business is identical in all three rows.

3. Cyclical businesses: the P/E looks lowest at the top

This one costs real money, because the failure runs backwards from intuition. In a cyclical business — steel, chemicals, shipping, homebuilding, autos, energy — earnings swing hard with the cycle. At the peak, earnings are highest, so the denominator is largest, so the P/E is lowest exactly when the earnings are least sustainable. At the trough, the reverse.

That reversal is not a trick of the example. It is what one year of earnings does to a fraction when earnings are volatile, and it is the reason the cyclically adjusted ratio below exists.

4. Asset-heavy and asset-light earnings are not the same object

Depreciation is subtracted before you reach net income, so how capital-intensive a business is changes what “earnings” represents. A steel mill's depreciation charge stands in for machinery that wears out and has to be replaced with cash; a software company's largest cost is people, and its profit sits much closer to cash generation. Comparing the P/E of a mill with the P/E of a software business compares two differently constructed denominators, which is why sector-relative comparison is the only version that holds up. Earnings are also struck after interest, so a heavily indebted company's profit is a levered number and its P/E can look modest while the whole enterprise, debt included, is priced quite differently — the enterprise value point on market capitalization.

There is no good P/E, and a low one is not a discount

The most common question about this ratio is what number is good. The honest answer is that no authority publishes one. No statute, regulator, accounting standards board or exchange defines a fair, cheap or expensive multiple. Every threshold you have been quoted is a convention, true only in the sense that enough people repeat it. This page will not print one, because printing it would dress a habit up as a rule.

What survives scrutiny is comparison rather than a level, and all three versions are shaky:

• The same company against its own history — weakening as the business changes. A company that swapped hardware revenue for subscriptions is not the company whose ten-year average you are quoting.
• Close competitors, on the same earnings definition — which means checking that both figures are trailing or both forward, both GAAP or both adjusted.
• The sector — useful as context, and draggable a long way by a few very large members or one company with near-zero earnings.

A low P/E is as often a warning as a bargain

The value trap is the standard failure. A low multiple is not a discount from an inattentive market; it is usually a forecast that earnings are going to fall. If the denominator drops next year, today's low P/E was measuring profit the company is about to stop making. This site's Markets · Fundamental page says it in a line: “a low P/E usually means the market expects earnings to fall, not that the stock is cheap. Cheap and falling look identical on a screener.”

And a high P/E can be perfectly rational

The denominator is one year of earnings. If that year is a small fraction of what the business will earn later, a large multiple of it is not automatically an overpayment.

CAPE, PEG, and why an index P/E is a different animal

Three variations you will meet. Each fixes something and each adds a problem.

CAPE, the cyclically adjusted P/E

Robert Shiller, who developed the measure with John Campbell, addresses the cyclical problem above by replacing one year of earnings with a long average. In their update of “Valuation Ratios and the Long-Run Stock Market Outlook” the method is stated directly: “we smooth earnings by taking an average of real earnings over the past ten years.” Real means inflation-adjusted, so the ten years sit in comparable dollars — the inflation adjustment is necessary work, not decoration. Shiller publishes the underlying monthly US series, back to 1871, free.

Two things to hold. It is an index-level measure, built for the market as a whole rather than one stock. And its record as a timing tool is contested. Jeremy Siegel, “The Shiller CAPE Ratio: A New Look,” Financial Analysts Journal 72(3), 2016, argues that changes in how GAAP earnings are computed biased CAPE's recent readings toward pessimism — note what that argument is: a dispute about the denominator, the theme of this whole page. Campbell and Shiller themselves declined to over-claim, writing of the January 2000 reading: “We do not find this extreme forecast credible; when the independent variable has moved so far from the historically observed range, we cannot trust a linear regression line.”

PEG, and what it actually divides

The idea is reasonable: a multiple should be read against how fast earnings are growing. The convention that a PEG below 1 marks something “reasonable” is exactly that — a convention with no derivation and no body publishing it. This site's own metric reference lists PEG second and states the trap in a line: it is “entirely dependent on a growth forecast, and forecasts are wrong.”

Be blunt about the structure. PEG divides an accounting output by a forecast. The numerator carries every judgment inside the P/E's denominator, and the result is divided by a growth rate that is itself unspecified — historical or projected, over one year or five, from whose estimate. It does not repair the P/E's weakness; it multiplies it by a second one. The popular attributions of the ratio — to Peter Lynch's One Up on Wall Street (1989), and earlier to Mario Farina in 1969 — we could not verify against a primary source, so both sit in the Unverified row below.

The P/E of an index is not the P/E of a stock

An index has no earnings per share. Somebody has to decide how to combine hundreds of members, so an index P/E depends on the aggregation method as much as on the market.

The aggregate method is the one index providers document. MSCI's fundamental data methodology sums constituent market values over summed constituent per-share figures, and applies an adjustment factor where earnings are negative specifically to “avoid increasing losses artificially in P/E and P/CE calculations at index level.” Turn Member B's $20 billion of earnings into a $20 billion loss and the aggregate becomes $1,000B ÷ $10B = 100.

And because broad indices weight by float-adjusted market capitalization, an index P/E is dominated by its largest members whichever method is used — an index ETF tracking it inherits that concentration exactly.

What is sourced here and what is not

This page mixes rules with habits, and the difference matters more than the content. Every claim above, sorted by what stands behind it.

What trips people up

• Comparing two P/E ratios built on different earnings. Trailing against forward, or GAAP against adjusted, is not a comparison. It is the most common error with this ratio and it is invisible on a screener.
• Treating a low multiple as a discount. It is usually a forecast that earnings will fall. Cheap and deteriorating look identical in a P/E column.
• Reading a negative P/E as an expensive stock. A loss-making company has no meaningful P/E at all. Go look at revenue, cash and the balance sheet instead.
• Quoting a P/E on a company earning almost nothing. At a $40 price, five cents of EPS is a P/E of 800 — arithmetic, not information, and a penny either way moves it hundreds of points.
• Buying a cyclical on a low trailing multiple at the top of its cycle. The multiple is lowest exactly when earnings are least sustainable — 10.0 at the peak and 50.0 at the trough above, with the ordering reversing against average earnings.
• Comparing across industries. Capital-intensive and asset-light businesses have differently constructed denominators; depreciation and interest are both subtracted before you reach the number you divide by.
• Taking a forward P/E as a fact. It is a price divided by an estimate with documented optimism in it. A 25% miss turned a stated 15.0 into a real 18.75 above.
• Trusting a headline “adjusted” multiple without opening the reconciliation. Regulation G requires that table to exist, and it explained a P/E of 12 against a P/E of 20 above.
• Applying stock-level intuition to an index P/E. The number depends on the aggregation method — 20.0, 27.5 or 17.5 from the same two companies — and one big loss-making member distorts it enormously.
• Waiting for the ratio to give a verdict. P/E is a price, and prices do not come with instructions.

The figures on this page are checked against the source that publishes them, and dated. Published rates move after the release named above — the linked source always carries the current number. This page explains a term; it does not recommend a product.