Exxon earned $6.70 a share in 2025 with oil at $69. It expects about $12 this year with Brent near $100.
Same wells. Same refineries. Same management, same strategy, same costs. The earnings nearly doubled because a war closed much of a strait.
So when you divide the share price by "the earnings", which earnings? At $160.59 the shares are 24 times last year's figure and 13 times this year's. Those are two completely different investments, and the difference has nothing to do with the company.
The ratio is lowest when the danger is highest
This is the specific cruelty of cyclicals. At the top of a cycle earnings are at a maximum, so the price-to-earnings ratio is at a minimum — the business looks cheapest at the moment the next few years are most likely to disappoint.
At the bottom the opposite happens. Earnings collapse, the ratio becomes enormous or meaningless, and the screen filters the company out entirely — at exactly the point where buying it would have worked.
The signal runs backwards. Not weakly: reliably, in the same direction, every cycle.
| At the top | At the bottom | |
|---|---|---|
| Earnings | Maximum | Near zero or negative |
| P/E ratio | Looks cheap | Looks absurd |
| What a screen does | Surfaces it as value | Filters it out |
| What happens next | Earnings fall | Earnings recover |
| In our library | A gross margin of 85%, up from −12% | Combine units −51% in two years |
A cyclical looks cheapest at the top and impossible at the bottom. Any method that reads the last twelve months will buy high and sell low, reliably.
Divide by the cycle, not by the year
Graham's answer, and it has not been improved on: value a cyclical on the average earnings of a whole cycle, not on the earnings of any one year in it.
Applied to Exxon, the midpoint of $6.70 and $12 is about $9.35, which puts the shares at roughly 17 times — neither the comforting 13 nor the alarming 24. That number is dull, and it is the one you can act on.
Applied to SanDisk, our report produced three figures that are all correct: 8 times next year's expected earnings, about 30 times the average of a whole cycle, and 178 times the last cycle's. A 22-fold spread between the cheapest and dearest, with nobody disputing any fact.
| Exxon at $160.59, divided by… | Multiple | What you are assuming |
|---|---|---|
| 2025 earnings · $6.70 at $69 oil | 24.0× | A hundred-dollar Brent is temporary |
| Mid-cycle · ~$9.35 | 17.2× | Both kinds of year happen |
| 2026 expected · ~$12 at ~$100 | 13.4× | The war premium persists |
Count things, do not read prices
Mid-cycle earnings tell you what to pay. They do not tell you where in the cycle you are standing — and for that, the useful data is physical.
Our Deere report did not conclude the cycle had bottomed from revenue or margins. It concluded it from three counts:
- Combine unit sales across America fell 51% in two years. Units, not dollars — a collapse that deep in a replacement market creates its own recovery, because machines keep ageing whether or not anybody buys.
- Used high-horsepower inventory down nearly 40% year on year. The second-hand market clears before the new one does. Falling used inventory is the first genuine sign of a turn, and it appears long before it reaches an income statement.
- Model-year-2027 early-order programmes came in strong. Orders placed with money, for machines not yet built — the closest thing to a forward signal that exists in an industrial business.
Revenue and margin describe what prices did last quarter. Units, inventory and orders describe where the imbalance is now.
In a cyclical the second thing predicts the first, never the other way round — which is why an investor reading only the accounts is always a year behind someone counting machines.
Make sure you are valuing the right company
One more trap, and it caught our own screen. Deere's profile is what our report called a museum of category errors: a bankruptcy score in the grey zone, net debt of roughly five times earnings, a price of 54 times free cash flow, and a discounted cash flow value of $199 against a $649 share price.
Every one of those is an artefact of consolidating a $54 billion lending book into a tractor company. Strip the bank out and the manufacturer carries about $2.6 billion of net debt against $21 billion of equity — an entirely ordinary industrial balance sheet.
A captive finance arm does this to every manufacturer that has one. Before valuing the cycle, check you are valuing the factory rather than the factory plus a bank.
Four steps, in order
It is disclosed as a segment. Until you have done this, the leverage, the cash flow and every ratio built on them describe a company that does not exist.
Peak to peak, or trough to trough — usually seven to ten years. Divide today's price by that average. This is the number to hold in your head, and it will rarely be the exciting one.
Units sold, inventory on the ground, order books, rig counts, capacity under construction. These are all published, and they move before the accounts do.
By then it cannot anchor you. A single-digit figure at the top of a cycle is a description of the peak, and you will recognise it as one rather than as a bargain.
One whose earnings nearly doubled on a war, one at the top of its cycle, and one that has probably just bottomed.
Each one valued on the cycle rather than the quarter.
Frequently asked
Why does the P/E ratio not work for cyclical companies?
Because it is lowest at the top. At the peak of a cycle, earnings are at their highest and the multiple looks cheapest — which is precisely when the next few years are most likely to disappoint. At the bottom the ratio looks expensive or is meaningless because earnings are near zero. The signal runs backwards to what you want it to say.
What are normalised or mid-cycle earnings?
The average a business earns across a whole cycle, good years and bad together, rather than whatever it earned last year. Exxon earned $6.70 a share at $69 oil in 2025 and expects about $12 this year with Brent near $100 — the midpoint of roughly $9.35 is a far better basis for a long-term valuation than either end.
How do you know where you are in a cycle?
Watch physical quantities rather than prices or profits. Deere's combine unit sales across America fell 51% in two years and used high-horsepower inventory was down nearly 40% year on year — units sold and inventory on the ground tell you about the cycle, while revenue and margin tell you about last quarter's prices.
Is a low P/E on a cyclical always a trap?
Not always, but it is never evidence on its own. The same company can be 8 times next year's expected earnings, 30 times the average of a whole cycle and 178 times the last cycle's — all three correct. The single-digit number is always available at the top, which is why it should be the last thing you look at rather than the first.