Earnings quality & red flags
What it measures¶
This area asks whether the published accounts show the patterns that published studies link to profit that was overstated — by optimistic accounting, by error, or by manipulation. Its central idea is accruals: the part of profit that has not (yet) arrived as cash, such as a sale booked before the customer pays. Some accruals are normal; large ones have historically been a warning. The area runs four checks:
| Check | The question | Basis |
|---|---|---|
| Accruals | Is net income minus operating cash flow within 10% of average assets? | absolute |
| Manipulation screen | Is the Beneish M-score below −2.22? | absolute |
| Misstatement screen | Is the Dechow F-score below 1.0? | absolute |
| Cooking-the-books count | Are no more than 2 of Montier's 6 warning flags raised? | absolute |
Why it is on the page¶
Every other area reads the published figures as they stand. This one asks whether they can be taken at face value. The four models were built by studying companies that were later caught by regulators or saw their profit fall back, and they look for the traces those companies left in their accounts.
Thresholds, their source and the population they were fitted on¶
| Check | Line | Source and fitted population | Evidence grade |
|---|---|---|---|
| Accruals | within ±10% of average assets | Sloan (1996): US companies 1962–1991, ranked into tenths; the 10% line is a practitioners' line | peer-reviewed and replicated / practitioner convention |
| Manipulation screen | M below −2.22 passes; −2.22 to −1.78 reads "possible"; above −1.78 caps the overall | Beneish (1999): 74 companies charged by the US securities regulator, 1982–1992 | peer-reviewed and replicated |
| Misstatement screen | F below 1.0 | Dechow and colleagues (2011): 2,190 enforcement actions, 1982–2005 | peer-reviewed and replicated |
| Cooking-the-books count | 2 or fewer of 6 flags | Montier (2008): US and European companies, 1993–2007 | practitioner convention |
Counted once. The four models share one ingredient — accruals carry Beneish's largest weight and open Dechow's model. Counting each as a separate verdict on the same fact would weigh accruals four times, so the accrual construct is counted once here and never again in Profitability & returns.
The cap. A Beneish M-score above −1.78 keeps the band's overall word at Mixed at most (see the cap).
A worked example¶
Apple, fiscal year 2025, on the statements stored on 25 October 2026:
- Accruals — passed. Net income minus operating cash flow was 0.1% of average assets.
- Manipulation screen — passed. M of −2.32, below −2.22.
- Misstatement screen — failed. F of 1.31, above 1.0.
- Cooking-the-books count — failed. 3 of Montier's 6 flags were raised.
All four could run and two passed: 2 of 4 checks passed · 50% · Mixed –.
Honeywell International, the same year, shows the other state. Its accruals passed (−2.3% of assets) and none of Montier's flags was raised, but no net property, plant and equipment line is stored for fiscal 2025, which both the Beneish and the Dechow models need. Two checks could run, so the tile reads No verdict — net property, plant and equipment FY2025 with 2 of 4 checks could run — and because this is a core area, Honeywell's overall figure is withheld.
What “No verdict” means here¶
The tile shows No verdict when fewer than three checks could run. The Beneish and Dechow models each need two consecutive fiscal years of many balance-sheet lines (receivables, net property and equipment, short-term debt), so a company whose stored statements lack one of them often reaches only two checks. The missing line is named.
Earnings quality is a core area: without a verdict here, the overall figure is withheld.
What the verdict is and is not¶
A failed check here is a flag that says "look closer", not a finding of wrongdoing. The Beneish line wrongly flagged about one in seven honest companies in its own study, and fast-growing companies trip it often for innocent reasons. A pass is not a certificate that the accounts are sound. The verdict is not an instruction to trade the shares.