Beneish M-Score
The Beneish M-Score is a statistical model developed by Indiana University professor Messod Beneish in 1999 that uses eight financial ratios derived from accounting data to estimate the probability that a company has manipulated its reported earnings, with scores above -1.78 suggesting a higher likelihood of manipulation.
Messod Beneish developed the M-Score by studying a sample of companies that had been identified by the SEC as having engaged in earnings manipulation and comparing their accounting patterns to a control group of non-manipulators. He identified eight financial ratios that showed statistically significant differences between manipulators and non-manipulators, then combined them using discriminant analysis weights into a single score.
The eight components address different manipulation mechanisms. The Days Sales Receivable Index (DSRI) measures whether receivables are growing faster than sales — a red flag for premature revenue recognition. The Gross Margin Index (GMI) measures gross margin deterioration, which pressure-tests management to manage reported earnings. The Asset Quality Index (AQI) measures changes in non-current, non-physical assets as a proportion of total assets, capturing capitalization of expenses that should be expensed. The Sales Growth Index (SGI) measures revenue growth, as high-growth companies face greater pressure to sustain momentum through accounting choices. The Depreciation Index (DEPI) measures declining depreciation rates relative to assets, suggesting asset life extension to reduce expenses. The Sales, General and Administrative Expenses Index (SGAI) flags disproportionate cost growth relative to revenue. The Leverage Index (LVGI) captures increasing leverage that heightens manipulation incentives. Finally, Total Accruals to Total Assets (TATA) directly measures the ratio of accrual-based earnings to asset base.
The benchmark threshold of -1.78 classifies companies as potential manipulators above this score. In Beneish's original study, the model correctly identified approximately 76 percent of manipulating companies. A famous post-hoc application found that Enron's M-Score before its collapse was well above -1.78, suggesting the model would have flagged manipulation risk years before the accounting fraud became public.
Investors and forensic accountants use the M-Score as one component of a broader accounting quality review rather than a definitive fraud detector. False positives — non-manipulating companies flagged as potential manipulators — are relatively common, particularly in periods of rapid organic growth when receivables and revenue naturally expand simultaneously. The model performs best when used to prioritize deeper investigation rather than as a standalone sell signal.