A low P/E is not a discount. It is a forecast, and often a correct one

In 2011 you could buy Nokia at about eight times earnings.

It was the largest phone maker on the planet and had been for fourteen years. It had more cash than most governments, a distribution network nobody could replicate, and a brand that in large parts of the world simply meant "phone". By every valuation screen in existence it was one of the cheapest large companies you could own.

The screens were reading the price correctly. They were reading the earnings wrong, because those earnings were about to stop existing.

This is the mechanism behind almost every value trap, and it is worth being precise about, because the fix is not "ignore valuation". Valuation matters enormously. The fix is a specific correction that keeps the useful part.

A multiple is not a price tag

Price-to-earnings is one number divided by another, so it is tempting to read it the way you read a shelf: eight times is cheap, forty times is dear.

It is not a price. It is a summary of what the market currently believes about how durable that E is.

A business trading at eight times earnings is not on sale. It is a business the market expects to earn less next year than it did this year, and the market is often right, because it can see the same things you can and has usually seen them first. It is not right always, which is the only reason value investing works at all. But the low multiple is a forecast, and treating it as a discount means systematically buying whatever the market is most pessimistic about, and hoping to be right about the exceptions.

The trap closes quietly. The multiple was eight because earnings were falling. Earnings keep falling. The multiple stays at eight and the price halves. Nothing ever re-rated. The denominator moved.

That is Nokia. It is also, in a slightly different key, Tattooed Chef, which I owned personally and which went to zero.

The same error, running the other way, costs more

Screening on a low multiple does not only buy you bad companies. It excludes good ones, permanently, and that side of the ledger is bigger.

Take a business compounding earnings at 30% a year, trading at forty times. Against a company growing at 2%, the multiple alone says it is five times more expensive. But the earnings you are paying forty times for will not be these earnings for very long, and the arithmetic of that is not intuitive: at 30% growth, earnings roughly double every two and a half years.

Screen on P/E and you exclude that company. Not occasionally. Structurally, on every single run, for the entire compounding life of the business, which is exactly the period during which owning it would have mattered.

That is not a filter with a false-negative rate. That is a filter pointed at the wrong thing.

The first correction, and why it is not enough

Dividing the multiple by the growth behind it, which the textbooks call PEG, fixes both directions at once. A high multiple attached to real growth reads as reasonable. A low multiple attached to no growth still reads badly, which is the outcome the raw ratio gets backwards.

It is better. It is not sufficient, for two reasons that both matter.

Growth measured over one window is mostly noise. One weak quarter, one pandemic, one large acquisition, and the growth rate you divided by describes an accident rather than a business.

And it still runs on reported earnings, which is the deeper problem, because reported earnings are an opinion. Enron was named America's most innovative company by Fortune six years running, and the sixth of those awards came in 2001, the year it went bankrupt. Nobody at Fortune was corrupt. They were doing what a reasonable person does with a company whose reported numbers are extraordinary, which is to be impressed. The numbers were extraordinary because Enron had worked out that if you book the entire projected lifetime profit of a twenty-year contract in the quarter you sign it, every quarter looks spectacular, right up until you run out of contracts to sign.

Any ratio with earnings in the denominator inherits whatever the accounting did.

What we do instead

Two changes, and both of them are about the denominator rather than the price.

Charge every business for the capital it uses, at what that capital actually costs it. Not only the interest on its debt, but what shareholders expect for carrying the risk of owning it, built from that company's own balance sheet and its own risk rather than a single house rate applied to everyone.

This is not a subtlety. A company earning 9% on capital that costs 11% is destroying value every year while reporting a healthy profit, and a company earning 7% on capital that costs 5% is creating it. Return on capital cannot tell those two apart. Charging the real cost can, and it is where our measure, RTEP, starts.

And treat research as the asset it obviously is. Accounting rules make a company expense R&D in the year it is spent, which means the businesses reinvesting hardest in their own future look the least profitable right at the moment they are building the thing that will matter. We capitalise it and amortise it over its useful life, so a company is not punished in the numbers for building something.

The result is a profitability figure that moves when the business changes and sits still when only the accounting does.

Valuation still counts. It is just not the whole vote

None of this means price stops mattering. Paying too much for a good business is one of the two ways to lose money in equities, and it is the one that feels responsible while you are doing it.

It means valuation is one of five pillars rather than the filter that runs first and throws everything else away. Quality asks whether this is a business worth owning for a decade. Economic profit asks whether it earns anything after paying for its capital. Value asks whether the price is sane for the growth you are actually buying. Momentum asks whether the market has started to agree. Smart money asks what the people with the best information are doing.

Nokia in 2011 passed the value screen and failed the quality trend badly. A single-metric filter cannot express that. Five can, and when they disagree with each other, the disagreement is the useful part rather than a defect in the display.

Charlie Munger was once asked how a private investor should compete with the professionals. He said it was not supposed to be easy, and that anyone who finds it easy is stupid.

He meant it kindly. The point was not that you cannot do this. It was that anything which feels frictionless is usually a thing you have not understood yet. A screen that hands you a clean, confident, obvious answer in four seconds should worry you more than one that hands you a mess.

Ours produces a mess fairly often. That is the honest version.

One habit, which costs nothing

If you take one thing from this and never touch our product, take this.

Before you buy anything, write down in two minutes, in plain language, why you are buying it. Not a model and not a target price. One or two sentences that a reasonable person could disagree with. Peter Lynch made himself do this, and his rule was that if he could not write the sentence, he did not buy.

The value is not in the writing. It is that in eighteen months, when the position is down 30% and you are deciding whether this is the moment to hold or the moment to admit it, you can go back and read what you actually believed at the time, rather than what you have since talked yourself into believing you believed.

I did not do this with Tattooed Chef. If I had, I would have had to write something like "a well-known investor is enthusiastic about plant-based food", looked at it in my own handwriting, and understood immediately that it was not a reason.

Do it in a notes app. Do it on paper. There is a field for it on the watchlist here, but that is genuinely not the point.

Educational and proprietary. This explains what our research does, not the exact formula behind it, and it is not personalised investment advice. See the full disclaimer.