APEX factor · Accruals
Accruals measure how much of reported earnings came from accounting estimates versus actual cash. High-accrual earnings tend to reverse — receivables that don't get collected, inventory that gets written down, deferred revenue that never materialises. Stocks with high accruals systematically underperform the next year.
Two companies report identical earnings of $100m. One collected $100m in cash. The other collected $40m in cash plus $60m in receivables, deferred billings, and inventory build-up. Both look identical on the income statement. Twelve months later: the cash-rich one continues earning $100m, the accrual-rich one writes down $40m of those phantom assets and earns $60m. Sloan showed this is not an exception — it's the rule. The market under-prices cash quality of earnings, and the pricing error reverses on a roughly one-year cycle as the accruals either turn into cash or turn into writedowns.
total_accruals = net_income - operating_cash_flow // base Sloan (1996) working_cap_acc = Δ(receivables + inventory) - Δ(payables) // Richardson 2005 discretionary_acc = working_cap_acc residual to peer industry // Dechow-Sloan-Sweeney acc_score = (total_accruals + discretionary_acc) / total_assets accruals = -1 · z_score(acc_score) // sign flipped: low accruals = bullish
Sign flip is the key step — high accruals predict negative future returns, so we negate so that the factor reads bullish-when-high like the others. We compute over trailing four quarters to smooth out one-off charges. The discretionary component (the part not explained by industry working-capital norms) carries more weight than total accruals because it isolates accounting choice from operational cycle.
Accruals come from XBRL-tagged income-statement and cash-flow lines on SEC EDGAR. Recompute happens after every 10-K and 10-Q ingestion, plus a daily 06:00 UTC sweep that catches restated filings. The factor lives in apex_factor_scores per ticker per filing-date and forward-fills until the next quarter. Sector-relative residualisation matters here too — utilities have structurally different working-capital cycles than software, and the cross-sectional comparison must respect that.
Accruals' strongest interaction is with Quality. Sloan-low-accruals + Novy-Marx-high-quality is a textbook compounder — the QUALITY COMPOUNDER confluence pattern fires bullish on this combination. High-accruals + low-quality + falling momentum triggers QUALITY CRACK bearish. The third interaction is with Value: a value-cheap stock with high accruals is the canonical value trap (Lakonishok-Shleifer-Vishny 1994 mechanism), so the VALUE TRAP pattern fires when the two combine.
Accruals breaks down for companies with structural working-capital cycles — large infrastructure projects, biotech with milestone payments, consumer-goods retailers with seasonal inventory. The factor reads a Pfizer with $20bn of late-stage trial accruals as bearish even when the trials are succeeding. We mitigate two ways: (a) sector-relative residualisation, which lets a biotech with peer-relative-low accruals score normally; (b) the pharma vertical's separate Phase 3 LLR table overrides the accruals signal for clinical-stage tickers. Honest limitation: we don't yet apply this carve-out beyond pharma — capital-intensive industrials remain a known noisy zone.
Every ticker page shows the per-factor decomposition. The Accruals score is one of twelve composing the 0–100 APEX composite.