Earnings quality · Manipulation detection · Accounting fraud screen
Beneish M-Score: Earnings Manipulation Formula, Thresholds & Examples
The Beneish M-Score is a statistical model that uses eight financial ratios to flag potential earnings manipulation. Developed by Professor Messod Beneish in 1999 using SEC enforcement cases, it produces a single score that separates likely manipulators from non-manipulators. A score above −1.78 warrants investigation before trusting the reported numbers.
The Beneish M-Score detects potential earnings manipulation using eight accounting ratios combined into a weighted score. 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*LVGI. A score above -1.78 suggests possible manipulation; a score at or below -2.22 indicates low manipulation risk.
What the M-Score Measures
Professor Messod Beneish trained the model on a sample of firms subject to SEC Accounting and Auditing Enforcement Releases (AAERs) between 1982 and 1992. Each of the eight variables captures a distinct symptom of manipulative accounting: inflated receivables, deteriorating margins, aggressive asset classification, high accruals, or rising leverage.
The model is a manipulation risk screen, not a verdict. A score above −1.78 does not prove fraud; it means the financial statements have characteristics statistically associated with manipulation. Enron scored −1.89 in the year before its collapse — just inside the grey zone — and WorldCom scored above −1.78 for three consecutive years before restatement. Use the M-Score alongside the Piotroski F-Score (financial strength) and the Altman Z-Score (bankruptcy risk) for a complete picture of accounting quality and financial health.
Because the model requires two consecutive years of financial data to compute index ratios, it cannot be applied to companies in their first year of public reporting. Accuracy in Beneish's original study was approximately 76% — meaning roughly one in four flagged companies was a false positive. Always read the 10-K and scrutinize notes before drawing conclusions from a high M-Score.
The 8 Variables Explained
TATA has the highest coefficient (4.679) in the model, reflecting strong academic evidence that accruals — the gap between reported earnings and earnings per share supported by cash — are the most reliable manipulation signal. When TATA is elevated, treat the other variables as corroborating rather than independent evidence.
Interpretation Thresholds
High manipulation risk. Financial statement characteristics strongly resemble firms subject to SEC enforcement. Enron scored −1.89 before collapse.
Ambiguous signal. Some manipulation characteristics present. Investigate high-coefficient variables (TATA, DSRI) and read the notes to the financial statements before concluding.
Low manipulation risk. Accounting ratios are inconsistent with the manipulation profile in Beneish's training sample. Original study showed ~76% accuracy at this threshold.
The −1.78 threshold maximizes detection accuracy at the cost of more false positives. Some practitioners use the stricter −2.22 cutoff to reduce false alarms in screening large universes. The Beneish M-Score calculator shows the exact score and all eight component ratios for any ticker, letting you determine which variables triggered the flag.
Worked Example: Company XYZ (Hypothetical Mid-Cap)
Starting from two years of financial statements, we compute each index ratio by dividing the current year figure by the prior year figure (or by total assets for TATA). All inputs are simplified to illustrate the method. Year 1 figures shown in parentheses.
- DSRI: (Receivables/Revenue) Year 2 / Year 1 — ($48M/$320M) / ($38M/$290M)
- 1.52
- GMI: Gross Margin Year 1 / Year 2 — 42.1% / 38.6%
- 1.09
- AQI: (1 - (CA+PP&E)/TA) Year 2 / Year 1 — 0.31 / 0.27
- 1.15
- SGI: Revenue Year 2 / Year 1 — $320M / $265M
- 1.21
- DEPI: (Depreciation/(Dep+PP&E)) Year 1 / Year 2 — 0.148 / 0.132
- 1.12
- SGAI: (SG&A/Revenue) Year 2 / Year 1 — (58/320) / (44/265)
- 1.09
- TATA: (Net Income - CFO) / Total Assets — ($28M - $14M) / $410M
- 0.034
- LVGI: (LT Debt/TA) Year 2 / Year 1 — (140/410) / (108/370)
- 1.17
- M = -4.84 + 0.920(1.52) + 0.528(1.09) + 0.404(1.15) + 0.892(1.21) + 0.115(1.12) - 0.172(1.09) + 4.679(0.034) - 0.327(1.17)
- −1.65
- M-Score (−1.78 threshold: Grey / Likely Manipulator boundary)
- −1.65
At −1.65, Company XYZ falls above the −1.78 threshold, flagging as a likely manipulator. The primary drivers are DSRI = 1.52 (receivables growing well ahead of revenue), TATA = 0.034 (accruals exceeding the 0.031 threshold), and DEPI = 1.12 (slowing depreciation rate). SGI at 1.21 and LVGI at 1.17 add further pressure. The next step would be reading the revenue recognition note in the 10-K and comparing the allowance for doubtful accounts year-over-year.
Limitations of the Beneish M-Score
- Calibrated on manufacturing firms. The original training sample consisted primarily of manufacturers and industrial companies. Financial institutions, banks, insurance companies, and SaaS businesses have structurally different balance sheets where many ratios (especially TATA and AQI) lose their interpretive meaning.
- Requires two years of data. All eight variables are computed as year-over-year index ratios. First-year public companies, spin-offs, and companies that have undergone significant M&A restructuring may produce misleading scores because the prior-year baseline is incomparable.
- High SGI is a false-positive trap. Legitimate high-growth companies — particularly early-stage technology firms — routinely score above 1.60 on SGI. Because SGI carries a positive coefficient (0.892), fast organic growth mechanically pushes the M-Score higher even without any manipulation. Always contextualize SGI against the company's growth stage and industry.
- Screen, not verdict. Beneish's original study reported ~76% accuracy, implying roughly one in four flagged companies is a false positive. The M-Score is best used as a first-pass filter to decide where to focus due diligence, not as a standalone conclusion. Pair it with a review of the earnings quality checklist and direct 10-K analysis.
- Looks backward, not forward. The model uses reported financials which may themselves be manipulated. By definition, the most sophisticated manipulation is designed to avoid detection by models like this one. Sophisticated channel stuffing, related-party transactions, or off-balance-sheet structures may not appear in the eight ratios.
Beneish vs Altman vs Piotroski
The three models are complementary. Use all three together when evaluating financial statement quality, especially for value or distressed investments where accounting integrity matters most.