Misstatement screen

What it measures¶

The Dechow F-score estimates how much a company's accounts resemble those of companies later found to have misstated them. It combines seven figures, each with a published weight:

• accruals — profit not backed by cash — measured across the whole balance sheet;
• the change in money owed by customers;
• the change in inventory (goods in stock);
• "soft" assets — those that are neither cash nor property and equipment;
• the change in cash sales;
• the change in return on assets;
• whether the company issued new debt or shares in the year.

The weighted sum is turned into a probability of misstatement, and that probability is divided by the share of all companies that misstated in the study (0.37%). An F-score of 1.0 is therefore average risk; 2.0 is twice the average. The check asks: is F below 1.0?

Why it is on the page¶

It was built from the largest published set of regulatory enforcement cases and looks at different traces than the Beneish score — new financing, soft assets, a fall in return paired with rising accruals — so the two screens catch different patterns.

Thresholds, their source and the population they were fitted on¶

• Source: Dechow, Ge, Larson and Sloan (2011), "Predicting material accounting misstatements", Model 1.
• Fitted on: 2,190 enforcement actions of the US securities regulator, 1982–2005. Out of its sample it classified about 63% of companies correctly at the 1.0 line.
• Evidence grade: peer-reviewed and replicated.
• Basis: absolute.

Long-term investments and preferred shares are not stored, so the accrual term leaves them out; "issued new debt or shares" reads the debt-issued line and the share count, whichever is stored.

A worked example¶

Apple, fiscal year 2025 (statements stored on 25 October 2026): F = 1.31 — 31% above the average resemblance — failed, and the Earnings quality tile's reason sentence named it: Misstatement screen (1.31) misses < 1.0, absolute. The same year Apple passed the Beneish screen, which shows the two models reading different traces.

What “No verdict” means here¶

The score needs two consecutive years of many balance-sheet lines, including short-term investments and property and equipment. When one is missing it is not evaluable, the missing line is named — for Meta Platforms in fiscal 2025, short-term debt FY2025 — and the check leaves the area's count.

What the verdict is and is not¶

An F-score above 1.0 says the accounts look somewhat more like those of companies that misstated than the average does. Most companies above 1.0 have misstated nothing: at this line the model is right a little under two times in three. It is not a finding of wrongdoing, not a forecast, and not an instruction to trade the shares.