Equity Valuation Spread

This page tracks the equity valuation spread between the European and US markets, that is, the distance between the most expensive stocks and the cheapest ones. It also tracks the price paid for three classic investment factors: value, quality, and momentum. Data is updated weekly.

What a valuation spread measures

Most market indicators answer the question "are stocks expensive." This one answers a different question: "are stocks expensive relative to one another."

The difference isn't cosmetic. Two markets can show the same average valuation and have nothing else in common. In the first, every stock trades at roughly the same price. In the second, a handful of names trade at extreme multiples while the rest goes nowhere. The first is an expensive market, the second is a dislocated one, and they don't behave the same way afterward.

That's what the valuation spread captures. It divides the sales multiple of the most expensive decile by that of the cheapest decile. When it rises, dispersion widens. When it falls, the market tightens up.

This isn't a statistical curiosity. Adam Zaremba and Mehmet Umutlu showed in 2019, in a study covering global equity markets, that the valuation spread is a robust predictor of a strategy's future returns. In other words, the price paid today for a given approach tells you something about what it will deliver next. Obvious for individual stocks, far less commonly applied to strategies themselves.

One fact stands out in our own series and deserves to be flagged. Over twenty years, the cheapest decile has barely moved, in either zone. All of the spread's widening comes from the top of the distribution. There's a valuation floor the market never breaks through, whatever the fashion of the moment, and a ceiling that, by contrast, breathes enormously.

Benjamin Graham would have liked this observation. His margin of safety rests on a simple idea, that an investor's protection comes from the price paid rather than the quality of forecasts, a thesis Seth Klarman later picked up and extended half a century on. What this floor shows is that the entry price into neglected stocks hasn't drifted in twenty years, even at the 2021 peak. Euphoria doesn't remove the pool of discounted stocks, it just makes it harder to see, since everything else has taken off in the meantime.

One caution, though: don't confuse the two ideas. A low multiple isn't a margin of safety, at best it's the starting point. The company still has to be worth more than what you're paying for it, and no single ratio tells you that on its own.

This type of measure isn't new. AQR publishes valuation spreads periodically in its research notes, and Research Affiliates tracks relative factor valuations quarterly, mostly on US markets. What was missing was a freely accessible version, covering Europe as much as the US, and updated weekly instead of four times a year.

The three factors tracked

A factor is a ranking rule. All stocks are sorted by a given criterion, then the top 10% is compared to the bottom 10%. The question here isn't whether the factor works, it's how much it costs today relative to its own past.

Value, measured by FCF/EV

Free cash flow yield divides the cash a company actually generates by its enterprise value, that is, its market capitalization plus net debt. A high yield signals a discounted stock.

We use enterprise value rather than market cap alone, because a heavily indebted company looks cheap only until you account for what an acquirer would have to repay. This ratio is the subject of a detailed backtest and a live, continuously updated ranking (FR) of the top forty European stocks.

Quality, measured by the Piotroski F-Score

Joseph Piotroski proposed a nine-point score in 2000, each point a binary test covering profitability, financial structure, and operating efficiency. Five of those nine points reward improvement over the prior fiscal year, not a level achieved.

That detail explains part of the results shown above. The F-Score also picks up cyclical companies in recovery, structurally discounted, whereas a quality factor built on gross profitability would select durably profitable companies, which trade at a premium instead. The F-Score is the subject of a full backtest.

Its all-or-nothing nature rules out splitting it into deciles: we compare companies scored eight or nine to those scored two or below, two groups whose size shifts with the cycle.

Momentum, on a twelve-minus-one-month basis

Momentum ranks stocks by their past performance. We use the standard academic definition, twelve-month performance excluding the most recent month. That exclusion isn't a whim: over the very short term, prices tend to reverse whatever they just did, and including the last month would blur the signal. Momentum and the main financial ratios are covered in our reference article on fundamental analysis (FR).

The unit of measure: price to sales

All three factors are measured with the same ratio, price divided by revenue. That choice isn't incidental. Revenue is never negative, unlike earnings or a cash flow figure. It's therefore the only ratio that stays interpretable at both ends of the distribution, including for money-losing companies, which populate precisely the most expensive decile. We've dedicated a backtest in Switzerland and France (FR) to it.

How to read these numbers

The valuation spread reads both ways, which is what makes it useful. A wide spread signals that the stocks driving the market higher are expensive relative to its floor, which calls for caution. But it's also, historically, the point at which strategies tilted toward discounted stocks have gone on to perform best. The same observation therefore feeds both worry and opportunity, depending on the horizon.

For the factors, what a high number means depends on the factor itself. High momentum reflects a premium paid for recent gains. A low value number, by contrast, reflects an unusually deep discount.

One last caution. A number's position in its own history and its absolute level are two separate pieces of information, and they can point in opposite directions. A market can sit low in its own history while still, in absolute terms, running hotter than its neighbor. That's why levels are always shown alongside historical positions.

Method and limitations

Data comes from Portfolio123. The universe excludes stocks too illiquid to be realistically tradable. The two zones are treated separately and never aggregated. Since the published ratios are ratios between two figures expressed in the same currency, they don't depend on any exchange-rate conversion, which makes the comparison between zones immediate.

We publish decile medians rather than capitalization-weighted aggregates. A median describes the typical stock in the group, an aggregate describes the typical dollar invested. The latter is the institutional convention, the former is far more robust to extreme values, which matters when working precisely on the tails of a distribution.

Deciles aren't sector-neutralized. That might seem questionable for a ratio as dependent on business type as price to sales, but we checked: the sector composition of both markets has barely shifted in twenty years, and what little it has shifted pushes in the opposite direction of what a composition-effect hypothesis would require. Each series is also compared to its own history, so a stable bias cancels out.

One limitation worth knowing. Companies without usable revenue can't be assigned a sales multiple, so they're excluded from the calculation. These are precisely the most speculative names, biotechs with no marketed product, recently listed shells, and they would sit at the expensive end of the distribution. Every published spread is therefore understated, never the reverse. The share affected runs around one stock in ten in the US, somewhat less in Europe, and widens during IPO waves.

Accounting data is taken in its preliminary form, the version published by the company at the time of the announcement, rather than the final version filed later. Both are faithful historical figures with no look-ahead bias; they differ only in the date the information became available. The measured difference between the two runs around one percent.

The European quality factor starts in 2011, not 2007. Before that date, European accounting coverage was too patchy for an F-Score to be reliably computed across the market, which would have distorted the composition of the lowest decile.

The valuation spread is recalculated every week. Factor valuations are recalculated monthly, and immediately whenever the spread moves sharply, so that everything stays consistent during fast-moving markets. Update dates appear under the charts.