Financial metric · Composite scores
Beneish M-Score
A score built from eight year-over-year financial ratios that estimates the likelihood that a company's earnings have been manipulated. The higher the score, the more red flags.
How to calculate it
M = −4.84 + 0.920×DSRI + 0.528×GMI + 0.404×AQI + 0.892×SGI + 0.115×DEPI − 0.172×SGAI + 4.679×TATA − 0.327×LVGIDSRI receivables to sales index; GMI gross margin index; AQI asset quality index; SGI sales growth index; DEPI depreciation index; SGAI SG&A index; TATA total accruals to total assets; LVGI leverage index. Each index is this year ÷ last year, so 1 means no changeA score above −1.78 suggests a higher likelihood of manipulationHow to read it
- The heaviest weights are on accruals (profit not backed by cash), receivables and sales growth. Revenue growing fast, receivables growing faster and cash not keeping up drives the score up.
- When a company crosses the cutoff, find which index contributes most, then go back to the statements for an explanation. Often there is a legitimate one.
Common pitfalls
- It measures statistical suspicion, not proof. In Beneish's 1999 study the model misclassified a meaningful share of normal companies as suspicious and missed some actual manipulators.
- Fast-growing companies and those that just made large acquisitions naturally show high SGI and AQI, and are often flagged falsely.
- Some practitioners use −2.22 as the cutoff (from an earlier version of the model) instead of −1.78; results depend on the choice.
- It does not apply to banks and insurers.
Where to find it in a 10-K
Two consecutive 10-Ks: revenue, cost of revenue, SG&A, depreciation, receivables, current assets, PP&E, total assets, debt and operating cash flow.
U.S. vs. China reporting
The coefficients were estimated on U.S. companies that faced SEC enforcement for earnings manipulation. They are a reasonable screen for U.S. stocks but should be treated as a rougher guide for A-shares, where common manipulation methods (related-party transactions, fund occupation by controlling shareholders) differ.