The Sloan accruals signal — why cash-flow quality leads reported earnings

Richard Sloan's 1996 paper Do Stock Prices Fully Reflect Information in Accruals and Cash Flows about Future Earnings? is the single most-cited accounting anomaly in the quantitative investing literature. Thirty years later it still produces measurable alpha. This post explains what accruals actually are, why the gap between cash flow and reported earnings predicts the next twelve months of returns, and how DeepVane wires this into both factor scoring and pattern detection.

The accounting behind the signal

Reported earnings is a constructed number. Revenue gets recognised when goods or services are delivered, regardless of when cash arrives. Expenses get accrued when incurred, regardless of when cash is paid. The gap between the two — called accruals — is the difference between net income and operating cash flow, normalised by total assets:

accruals = (net_income − operating_cash_flow) / total_assets

High positive accruals mean reported earnings are running ahead of cash generation. Low (negative) accruals mean cash generation is running ahead of reported earnings. Sloan 1996 found that this single ratio predicts the cross-section of returns: firms in the bottom accruals decile (cash-richer than reported earnings suggest) outperformed firms in the top decile by roughly 10% annualised over the following year.

Why the anomaly exists

Investors treat earnings as the headline number. A company reporting strong earnings but weak cash conversion looks fine on a P/E screen; the accruals component of that earnings is less persistent — it reverses as cash collections catch down to the original recognition. Markets systematically under-weight this persistence difference, so the cash-heavy cohort under-prices and the accrual-heavy cohort over-prices.

The anomaly doesn't disappear because the mechanism isn't a mistake — it's embedded in accounting standards. GAAP demands accrual accounting; cash-basis accounting exists only on the secondary statement. Retail investors look at the first statement and under-weight the second.

Why it still works in 2026

The original effect decomposes into three parts:

1. Mechanical reversal. High accruals genuinely reverse as working capital accounts normalise, typically within 2-4 quarters.
2. Earnings quality. Companies that choose accrual-heavy recognition often do so because cash generation is weak, which correlates with business-model fragility.
3. Fraud risk. Extreme accruals are a statistical fingerprint of revenue manipulation. Enron, WorldCom, and most post-SOX fraud cases showed distinctive accrual spikes before restating.

All three still operate. Post-SOX enforcement reduced the frequency of extreme-fraud cases but didn't eliminate the quality component — the middle two quintiles still separate reliably on accruals.

How DeepVane implements it

The score_accruals factor in APEX is calibrated so that the bottom accruals decile (cash-richer) maps to high scores near 85-90, and the top decile (accrual-heavy) maps to low scores near 10-15. The factor uses the universe cross-section as its denominator rather than a static threshold — so the scoring stays meaningful as the distribution shifts through economic cycles.

Beyond the standalone factor, accruals appears in three of our fifteen confluence patterns:

• VALUE_TRAP — low value + low quality + bad accruals. LSV 1994 cheap-low-quality underperforms; Sloan adds the confirming accruals leg. Strongest negative pattern we track.
• QUALITY_CRACK — high reported quality + bad accruals + rolling momentum. Early warning for quality deterioration 1-2 quarters before GAAP numbers show it.
• PRICE_AHEAD_OF_FUNDAMENTALS — rising price + accruals deterioration + weak NLP tone. Divergence warning, typically mean-reverts within a quarter.

What to watch when applying the signal

Accruals is robust but not bulletproof. Failure modes to know:

1. Industry-specific adjustments. Financials and utilities have structurally different accrual patterns from industrials or tech. DeepVane applies sector-aware benchmarks so a bank with large loan-loss accruals doesn't register as a VALUE_TRAP the way a retailer with identical accrual ratio would.
2. One-time charges. A legitimate restructuring or impairment charge shows up as extreme negative accruals. Check the 8-K — if the charge is explained and contained, the pattern won't follow through.
3. Growth distortion. Fast-growing companies with rising working capital have structurally higher accruals without any manipulation. Our factor applies a growth-adjustment so a high-growth SaaS company with expanding AR isn't mis-tagged.

Where to see this live

Every ticker's score_accruals appears on the individual stock page factor breakdown. The pattern pages (VALUE_TRAP, QUALITY_CRACK, PRICE_AHEAD_OF_FUNDAMENTALS) list current tickers matching each pattern — most of the matches fire because accruals is contributing the confirming leg.

Sloan is in good company. The APEX composite combines accruals with eleven other academic factors (see How the DeepVane APEX score is calculated). The accruals component isn't the entire edge — it's one repeatable-alpha slice, shrunk Bayesian with eleven others into a more reliable composite than any single factor could deliver alone.