A valuation ratio is a measurement choice
A value screen asks how much accounting value or earning power an investor receives for a market price. It does not establish what a company is worth. Book-to-market, earnings yield and enterprise-value ratios compare different claims on the business. A study must specify the numerator, denominator, units, accounting scope and observation dates before testing which ratio appears most successful.
Kenneth French's variable definitions describe a book-to-market construction that pairs prior-year fiscal book equity with prior December market equity for June formation. That is a particular research convention, not permission to attach the newest balance sheet to every historical price. The same page describes book-equity adjustments. A simple shareholders' equity field should therefore be labelled a proxy unless its mapping reproduces the intended definition.
Work through two apparently cheap companies
Consider two fictional companies with market equity of $400 million. Company A has book equity of $200 million and trailing earnings of $24 million. Company B has book equity of $320 million and earnings of $8 million. B ranks cheaper on book-to-market, 0.80 versus 0.50; A ranks cheaper on earnings yield, 6% versus 2%. The disagreement is useful information, not a calculation error.
Before combining these signals, inspect what generates the difference. A write-down can reduce both assets and earnings; a share repurchase can shrink equity without improving operations; acquired goodwill can change book comparisons. These are possible explanations, not facts about the fictional firms. Record which accounting adjustments the research actually makes, rather than explaining every unfavorable outcome after observing its return.
Build an auditable cross-section
Construct each ranking from the eligible securities and information available on the formation date. Align consolidated earnings with the relevant equity claim, avoid double-counting multiple share classes, and make currencies consistent. Negative book equity should not quietly enter the cheapest bucket. Negative earnings are likewise economically different from a very high positive earnings yield; retain an explicit exclusion or separate treatment.
Use a declared outlier policy and publish how many observations it changes. A percentile rank reduces the influence of extreme magnitudes but discards information about distance. A standardized score retains distance but depends on the distribution and clipping rule. Compare sector exposures before calling the result stock selection: an aggregate value portfolio may primarily overweight industries with structurally different accounting.
What a credible result would need
Freeze one primary definition and treat alternatives as recorded sensitivity checks. A useful evaluation asks whether conclusions survive reasonable accounting lags, removal of very small companies, and conservative turnover costs. Compare like-for-like total returns against an eligible-universe benchmark. Neither a high valuation spread nor a long historical factor series supplies evidence that today's selected securities are mispriced.
Keep the claim narrower than the calculation. This example explains why two valuation measures can disagree; it does not recommend either company or estimate a premium. A publication should disclose unavailable filings, restatement handling and denominator mismatches alongside results. Without historical availability evidence, describe the screen as a retrospective accounting exercise, not a point-in-time investment process.
Inspect the example
| Fictional issuer | Market equity | Book equity | Trailing earnings | Book / market | Earnings yield |
|---|---|---|---|---|---|
| A | $400m | $200m | $24m | 0.50 | 6.0% |
| B | $400m | $320m | $8m | 0.80 | 2.0% |
Sources & reproduction
Primary references support the definitions and method. Numerical examples are illustrative unless a source record is explicitly identified.