PEG ratio formula: 1,018, none negative

On 2026-09-13 the PEG ratio formula returned a value for 1,018 of the 5,518 US listings in this product's pool, and not one of those 1,018 is negative. A further 3,989 listings are refused, each with a reason recorded on the row. Three hundred and thirty-four of the values sit below one, which is the region Lynch's rule was written about.

Six days earlier the same field held 674 values, of which 290 were negative and 96 rounded away to zero, leaving 288 on which the rule could be asked at all. What changed is not the market. It is where the two inputs come from.

Both legs are ours now

Until 11 September 2026 the stored PEG preferred a data vendor's published trailing ratio. Neither of its two inputs was read from a filing here, the periods were the vendor's, and the numerator was not necessarily the price to earnings ratio this site publishes for the same company.

The formula now divides two figures this product computes:

The worked example is Apple. Its price to earnings ratio is 44.36 and its earnings per share compounded at 17.74 percent a year over the five fiscal years to the newest one on file, so its PEG is 2.50. Both numbers expand into their own inputs and both come from statements you can open.

The base year is found by its own fiscal year rather than by counting rows. A company with a gap in its filed history has a row five places back that is not five years ago, and compounding across it would advertise a five-year rate measured over six or seven.

It refuses instead of returning a number nobody can use

A price to earnings ratio divided by a negative growth rate is a negative number, and a negative PEG sorts below every company that passed Lynch's test on any screen that treats lower as better. The ratio no longer produces one. Where a leg of the formula is missing or undefined, the row stores no value and records why.

Three of those seven are the cases that used to become a negative PEG. Two more are the cases that used to round to nothing, because a rate compounded off a base year near zero is a very large number with no meaning attached to it.

3M is the plain example. Its price to earnings ratio is 27.27 and its earnings per share compounded at minus 8.49 percent a year over the window, so the row reads not meaningful with "earnings shrank over the window" attached. Under the old method it would have read about minus 3.2, and minus 3.2 would have sorted ahead of every company in the table below.

The count of listings the rule can be asked on went up

The pool grew from 5,257 listings to 5,518 between the two readings, which accounts for none of that. Every one of today's 1,018 values can be compared with Lynch's line, because the ones that could not be are now refusals rather than numbers.

Two listings still round to 0.00 at the two decimal places the figure is stored at. The smallest stored value is 0.00019 and the largest is 354.92.

The shape of the 1,018

The median is 1.55. Three hundred and thirty-four listings sit below one, 6.1 percent of the pool, and the distribution no longer piles up at the bottom of that range: 89 of the 334 are in the region just under one that Lynch was describing, against 18 on the old reading.

Two hundred and fifty-nine of the 334 are worth a billion dollars or more and 104 are worth ten billion or more, so this is not a small-company artefact.

By sector, the 334 fall like this: financial services 110, consumer cyclical 48, industrials 41, technology 36, real estate 22, healthcare 22, consumer defensive 16, communication services 12, energy 11, basic materials 9, utilities 7.

What to do with a ratio built this way

Read the two inputs, not the quotient. This has not changed and it is still the main thing. A price to earnings ratio and a five-year growth rate are both figures you can check against a filing. Their quotient inherits the instability of both and reports neither.

Read the refusal as information. "The latest year lost money" is a fact about the company, and it is more useful than a number that hides it. On 2026-09-07 it was the reason recorded on 1,687 listings, 30.6 percent of the pool that day.

A high PEG can be the honest answer. Procter & Gamble prices at 21.52 times earnings against a five-year rate of 3.48 percent, so its PEG is 6.19. Nothing is broken there. The ratio is saying that the growth behind the price is slow, which is what it is for.

What this leaves out

Every count is a single reading of the production database taken on 2026-09-13, after the recompute that finished at 03:03 UTC that morning, over the 5,518 active equity listings in the pool. The counts move nightly. The live figures are at /coverage.

The 2026-09-07 row in the comparison table is the reading this page carried before it was rewritten. It was a true reading of what was stored on the morning of 2026-09-07, on a pool of 5,257 listings, and it is kept so the change is visible rather than quietly removed.

The growth rate is measured over exactly five fiscal years of filed earnings per share, so it says nothing about what earnings will do next. Lynch's rule was written about a growth rate he expected, and a rate read out of five filings is a different quantity wearing the same name.

Counts are per listing, not per company. A company with two share classes appears twice.

Of the 5,518 listings in the pool on 2026-09-11, 5,207 carried an indicator row at all. Within those, 1,018 held a PEG and 3,989 held a stated reason; 200 held neither, which is a gap in the record rather than a judgement about those companies.

Rounding to two decimal places is done at storage, so the two zeros are an artefact of the stored precision as well as of the underlying value.

Nothing here is a view on any company. A PEG below one is an arithmetic result, not a conclusion.