The Beneish M-Score
The Beneish M-score (Beneish, 1999) is a probit-style model that combines eight financial ratios into a single score estimating the probability that a firm has manipulated its earnings. The eight indices capture deteriorating fundamentals and rising accruals: receivables (DSRI), gross margin (GMI), asset quality (AQI), sales growth (SGI), depreciation (DEPI), SG&A (SGAI), leverage (LVGI), and total accruals to total assets (TATA). The common rule of thumb is that a score above roughly −1.78 flags a higher likelihood of manipulation. Crucially, the M-score is an early-warning probability, not proof of fraud. A high score means "investigate", never "guilty".
Why it matters
Beneish asked a simple question: do firms that later turn out to have manipulated earnings look different beforehand? They do. Their receivables balloon, margins erode, accruals swell. The M-score bundles eight such symptoms into one number that behaves like a manipulation thermometer. A high reading says the patient has a fever worth examining, not that the diagnosis is settled. Treated as a screen it is genuinely useful, including in the famous case where students flagged Enron with it; treated as a verdict it is dangerously overconfident.
Formulas
Worked examples
A firm's indices give DSRI = 1.8 (receivables ballooning), GMI = 1.3 (margins eroding), SGI = 1.5 (rapid sales growth), and a high TATA (large accruals). The computed M-score is −1.2. What does this tell you?
Because −1.2 is above the −1.78 threshold, the model places the firm in the higher-probability-of-manipulation zone. The drivers are intuitive: receivables and accruals are swelling while margins fall, the classic profile of aggressive revenue recognition. The correct conclusion is that the firm warrants a deeper forensic look, not that manipulation is proven. The M-score has raised the prior; investigation must confirm or dismiss it.
Wirecard, the German payments group, reported large and fast-growing receivables and cash held in trust by overseas partner banks. In June 2020 it admitted that roughly €1.9 billion of that cash, about a quarter of its balance sheet, did not exist, and it filed for insolvency days later. How does this case fit the logic of an M-score-style screen?
The profile that drives a high M-score was present for years: ballooning receivables (a high DSRI) and asset balances that grew faster than verifiable cash, the accounting fingerprint of inflated revenue and fabricated assets. A screen built from these indices is designed to flag exactly this pattern as an elevated probability of manipulation, prompting investigation. The lesson is twofold. Quantitative red flags can surface a fraud the market is ignoring, and confirmation still required forensic work, since the fictitious cash was the core of the deception.
Common mistakes
- ✗A high M-score proves the firm committed fraud. The M-score is a probabilistic early-warning flag; it indicates elevated likelihood and a need to investigate, never legal proof of manipulation.
- ✗The model has eight inputs but a fixed cut-off that is exact. The −1.78 threshold is a rule of thumb from the original sample, and some users add a −2.22 grey-zone boundary; scores near either should be read as a continuum of probability, not a hard pass or fail.
- ✗A score below the threshold guarantees clean reporting. Some manipulators score low and some honest firms score high; the model has false positives and false negatives like any screen.
- ✗The M-score replaces reading the statements. It is one quantitative input within a broader quality-assessment process, useful precisely because it directs attention.
Revision bullets
- •Beneish M-score: eight ratios combined into a manipulation probability
- •Indices: DSRI, GMI, AQI, SGI, DEPI, SGAI, TATA, LVGI
- •Rule of thumb: M above about −1.78 flags higher manipulation likelihood
- •Conservative grey zone sometimes used: −2.22 (unlikely) up to −1.78
- •TATA (accruals) and DSRI (receivables) carry large weights
- •It is an early-warning probability, not proof of fraud
Quick check
The Beneish M-score combines eight financial ratios to estimate
A firm scores M = −1.0, above the −1.78 threshold. The correct interpretation is
Connected topics
Sources
- Beneish, M. D. "The Detection of Earnings Manipulation." Financial Analysts Journal 55, no. 5 (1999): 24-36.Original eight-variable probit model and the manipulation-probability interpretation of the M-score.
- CFA Program, Financial Reporting QualityCFA Institute. "Financial Reporting Quality." CFA Program Curriculum, Financial Statement Analysis. CFA Institute.Presents the Beneish M-score as a quantitative screen for the probability of earnings manipulation.