Detecting earnings manipulation
Beneish M-Score
The Beneish M-Score is an 8-variable mathematical model developed by Professor Messod Daniel Beneish at Indiana University in 1999. It measures the degree of distortion in a company's financial statements and produces a single score that indicates the likelihood of earnings manipulation.
What Is the M-Score?
Unlike traditional financial ratios that measure performance, the M-Score is specifically designed to detect whether a company has manipulated its reported earnings. It works by comparing year-over-year changes in financial statement items that tend to distort when management inflates revenue, understates expenses, or otherwise manipulates reported income. The model was originally validated against companies that were later found by the SEC to have engaged in earnings manipulation.
The 8 Variables
Measures whether receivables and revenues are in or out of balance. A large increase in DSO suggests revenue may have been inflated. DSRI = (Receivables_t / Revenue_t) / (Receivables_{t-1} / Revenue_{t-1}).
Measures deterioration in gross margins. Companies with worsening margins are more likely to manipulate earnings. GMI = Gross Margin_{t-1} / Gross Margin_t. A GMI > 1 means margins are declining.
Measures the ratio of non-current assets other than PP&E to total assets. An increase suggests excessive capitalization of expenses. AQI = [1 - (Current Assets_t + PP&E_t) / Total Assets_t] / [1 - (Current Assets_{t-1} + PP&E_{t-1}) / Total Assets_{t-1}].
Measures sales growth rate. While growth itself is not manipulation, high-growth companies face more pressure to meet expectations and are statistically more likely to manipulate. SGI = Revenue_t / Revenue_{t-1}.
Measures whether the rate of depreciation has slowed. A declining depreciation rate may indicate that useful life estimates have been revised upward to reduce expenses. DEPI = Depreciation Rate_{t-1} / Depreciation Rate_t.
Measures disproportionate changes in SG&A relative to sales. An increase may signal administrative inefficiency or deliberate expense manipulation. SGAI = (SGA_t / Revenue_t) / (SGA_{t-1} / Revenue_{t-1}).
The single most powerful variable. Measures the gap between reported earnings and cash flow. High accruals strongly correlate with manipulation. TATA = (Net Income - Cash Flow from Operations) / Total Assets.
Measures the change in leverage. Increasing debt levels create incentive and pressure to manipulate earnings. LVGI = Total Debt_t / Total Debt_{t-1}.
The Formula
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×LVGINote the coefficient on TATA (4.679) — total accruals has by far the largest weight. A company with large paper profits that don't convert to cash will score higher on this variable alone.
Thresholds
The company is unlikely to be a manipulator. Most healthy companies fall well below this threshold.
Inconclusive — warrants closer inspection of accruals, receivables, and cash flow quality.
The company's financial statements show patterns consistent with earnings manipulation. Does NOT mean fraud is proven — it means the pattern matches known manipulators.
How We Use M-Score
- *We calculate the M-Score for every stock we grade, using the most recent 10-K filing and the prior year for comparison.
- *An M-Score breach (> -1.78) is a critical flag — it can push a stock to Grade D or F by itself, regardless of other checks.
- *An M-Score in the grey zone (-2.22 to -1.78) is a watch item that contributes to the overall grade.
- *We always present the raw M-Score value and each variable in our reports, so you can see exactly which component drove the result.
- *The M-Score is Check F1 in our 18-point screening framework.
Limitations
- *The model was built on manufacturing-era financial data (1982-1992). Some modern business models (SaaS, marketplace, asset-light) can trigger false positives.
- *Companies with large stock-based compensation may have high accruals that inflate TATA without any manipulation.
- *Rapid organic growth (SGI) is weighted positively, meaning fast-growing companies score higher even if growth is legitimate.
- *The model cannot detect fraud that occurs below the financial statement level (e.g., customer fabrication, channel stuffing within normal-looking DSO).
- *M-Score is backward-looking — it uses last year's financial data and cannot predict future manipulation.
- *Financial companies (banks, insurance) have fundamentally different balance sheets that make the model unreliable.
Historical Accuracy
In Beneish's original study, the model correctly identified 76% of companies that were later found to have manipulated earnings, while producing a false positive rate of approximately 17.5%. The model notably flagged Enron as a likely manipulator in 1998 — three years before the fraud was publicly exposed.
Sample Reports with M-Score Flags
These reports show real M-Score breach cases from our analysis: