Beneish M-Score
- The Beneish M-Score uses eight financial ratio variables to detect earnings manipulation in reported financials.
- A score above -1.78 signals a higher probability of manipulation and deserves closer scrutiny.
- The model is probabilistic, not a verdict, and financial institutions are excluded from its original sample.
What is the Beneish M-Score?
The Beneish M-Score is a statistical model that uses eight variables from financial ratios to detect possible earnings manipulation. It estimates the probability of manipulation, acting as a mathematical tripwire for companies cooking the books.
It was built by accounting professor Messod Beneish and has proven useful to short sellers and forensic accountants for decades. The goal is simple, to see if reported earnings look engineered.
How the formula works
The M-score combines eight financial ratio variables into one number, from receivables growth to asset quality to the accrual share of total assets. The formula is -4.84 + 0.92*DSRI + 0.528*GMI + 0.404*AQI + 0.892*SGI + 0.115*DEPI - 0.172*SGAI + 4.679*TATA - 0.327*LVGI.
This yields a single score that combines all eight variables into one read. A high score means the financial ratios resemble companies caught stretching the truth in the model's sample.
What the eight variable inputs measure
You can read the eight inputs as four quiet warning signals. The first watches revenue quality through receivables growth. The second tracks margin pressure. The third scans asset and depreciation quality. The fourth measures how much reported income leans on accruals, the accounting estimates that bend with judgment.
Each index rises when a company's numbers look stretched. Rising day-sales-in-receivables means sales are booked faster than cash arrives. A falling gross-margin index hints at softening demand. A jump in total accruals flags earnings built on timing rather than real cash flow.
The rest complete the read. Sales growth marks firms expanding faster than their peer group supports. Selling and general expenses reveal hidden cost creep. And LVGI, the debt input, asks whether a company has quietly borrowed more to flatter its returns.
The -1.78 threshold and limitations
The magic number is -1.78. Below it, the company is unlikely to be a manipulator. Above it, the probability of manipulation is high, meaning treat it as a red flag and dig into the underlying numbers.
The model is probabilistic, not a verdict, so it cannot detect manipulation with total accuracy. Financial institutions are excluded from the original sample because banks and insurers make money differently, so the ratio signs do not behave the same way.
It also has blind spots. A high score flags possible manipulation, not proven fraud, so dig deeper. The model was built on restated financials from the 1990s, and a savvy manager can still slip past it. Use it as one tool, never the whole answer.
Track record and aggregate use
The M-score has a real track record. Cornell students flagged Enron as a manipulator in 1998, years before the collapse, using this very model, catching a firm most analysts trusted.
Researchers also apply aggregate M-scores to whole markets. A 2023 paper found the score hit a 40-year high in early 2023 across listed firms, and such market-wide readings have tied high manipulation scores to coming recessions.