Which Valuation MetricsStill Predict Returns?

12 valuation and quality factors tested independently on US equities from 2005 to 2026. Full decile analysis (D1 through D10) for two universes: market cap >$200M and >$5B. Annual rebalance with 12 monthly cohorts averaged to eliminate timing bias. Quarterly TTM fundamentals with availability lag.

Methodology

For each factor, every month we sort all eligible stocks into 10 equal groups (deciles) based on that metric. D1 contains the stocks ranked "best" by the factor (cheapest P/S, highest dividend yield, etc.). D10 contains the "worst" (most expensive P/S, lowest yield, etc.). We then measure the forward 12-month return of each decile. The 12 monthly cohorts are averaged to produce annual returns, eliminating start-date bias.

Universe: US domestic common stocks. Ex-financials for value metrics. All sectors for quality metrics.
Fundamentals: Sharadar quarterly TTM data with availability lag (avail_ym).
Period: 2005–2026 (21 years, out-of-sample).
A factor "works" if D1 consistently outperforms D10 — the spread is positive and stable.

The Scorecard

The spread between the cheapest/best decile (D1) and the most expensive/worst decile (D10) for each factor. Green = works. Red = dead or inverted. Grey = marginal.

Summary Table

Price-to-Sales WORKS

Revenue is the hardest fundamental to manipulate. Low P/S stocks buy $1 of revenue for less.

Price-to-Earnings DEAD

The most popular valuation metric. Low P/E stocks are "cheap" relative to earnings.

EBITDA / Enterprise Value DEAD

Capital-structure-neutral valuation. High EBITDA/EV = cheap operating earnings relative to total firm value.

Price-to-Free Cash Flow MARGINAL

Cash flow is harder to manipulate than earnings. Low P/CF = cheap relative to cash generation.

Price-to-Book INVERTED

The Fama-French value factor. Low P/B = cheap relative to net assets on the balance sheet.

Dividend Yield WEAK

High dividend yield as a value signal. D1 = highest yielding stocks.

Buyback Yield INVERTED

Net share repurchase yield. Positive = company shrinking share count.

Shareholder Yield INVERTED

Dividends + buybacks combined. Total cash returned to shareholders.

Return on Equity INVERTED

Profitability measure. High ROE = efficient use of shareholder capital.

Profit Margin INVERTED

Net income / revenue. High margin = highly profitable on each dollar of sales.

EPS Growth INVERTED

Year-over-year earnings growth rate. High growth = fast-growing earnings.

Accruals-to-Price MARGINAL

Earnings quality signal. Low accruals = cash-backed earnings (more reliable).

Conclusions

What works

• P/S — the only traditional value factor with a reliable spread (+9.6% at >$200M, +4.4% at >$5B). Revenue is the hardest fundamental to manipulate.
• Accruals — modest but consistent. Low accruals = higher earnings quality.

What died

• P/E — near-zero spread. The most popular metric has no predictive power.
• EBITDA/EV — zero spread at >$200M. SBC distortion and ZIRP killed it.
• Dividend Yield — thin positive spread but unstable base rates.

What inverted

• P/B — expensive stocks outperform cheap ones by 1.8% at >$200M, 3.8% at >$5B. Tangible assets are no longer the dominant form of corporate value.
• Buyback Yield, SH Yield — strongly inverted. The pre-computed Sharadar TTM measure may not match the shares-outstanding method used in academic literature.
• ROE, Profit Margin, EPS Growth — high-quality/high-growth stocks underperform. These metrics may be priced in or proxying for slow-growth value traps.

The common thread

The factors that survive are built on the hardest-to-manipulate fundamentals. Revenue (P/S) requires real transactions. Cash vs accruals is a direct measure of earnings reliability. Metrics that rely on more manipulable inputs — earnings, EBITDA, book value — have been arbitraged away or structurally broken by changes in how companies create value (intangibles vs tangible assets).