Investor frameworks
The Beneish M-Score — detecting manipulated earnings
Eight ratios that together flag a higher-than-usual probability that reported earnings have been manipulated. It is a screening signal about accounting, not a fraud verdict.
What it measures
The M-Score combines eight year-over-year drift ratios into a single number. Lower is safer. A score above roughly −2.22 places a company in the range where manipulation is more likely than the base rate — which is a statement about probability, not a finding.
The eight components each ask whether something moved in a direction that manipulation would produce:
| Component | The question it asks |
|---|---|
| Days sales in receivables | Are sales being booked faster than cash arrives? |
| Gross margin | Has the margin deteriorated, creating pressure to flatter results? |
| Asset quality | Is a growing share of assets neither current nor plant? |
| Sales growth | Is growth fast enough to strain the reporting? |
| Depreciation rate | Has the depreciation rate slowed, lifting reported profit? |
| SG&A | Are overheads rising disproportionately to sales? |
| Leverage | Has debt increased, tightening covenant pressure? |
| Total accruals to assets | How much of profit is not cash? |
Why drift, not levels
Notice that every input is a change, not a level. That is the design.
A company with permanently high receivables may simply sell to slow-paying customers. A company whose receivables suddenly stretch relative to sales has changed something, and the change is what carries information. Beneish fitted the model on companies later found to have manipulated earnings and asked which ratios moved in the year before discovery.
What a high score is, and is not
It is a prior, not evidence. The model was fitted on a small sample of known manipulators, and most companies it flags have not manipulated anything: rapid growth, an acquisition, or a genuine change in business mix produce the same drift.
Read a high M-Score as an instruction to look at the filings rather than a conclusion drawn from them. Its most useful pairing is with the Altman Z-Score, because the two describe different failure modes — Altman asks whether the company can survive, Beneish asks whether the numbers describing it can be trusted. A company weak on both deserves considerably more scrutiny than one weak on either.
The limits worth stating
The model dates from 1999 and was fitted on US filers under the accounting standards of that period. Financial companies and REITs sit outside its intended domain — their balance sheets make several of the ratios mean something different. And a manipulator sophisticated enough to know the formula can, in principle, avoid tripping it.
It is a smoke alarm with a known false-positive rate, which is worth having precisely because the alternative is no alarm.
Check yourself
5 questions. Nothing is recorded unless you are signed in, and nothing here affects anything else.
Sources
- Beneish (1999), The Detection of Earnings Manipulation, Financial Analysts Journal (opens in a new tab) — The original eight-variable model and the sample of manipulators it was fitted on.
- Schilit, Financial Shenanigans (opens in a new tab) — The qualitative companion — what the ratio drift looks like in the filings themselves.
- Sloan (1996), Do Stock Prices Fully Reflect Information in Accruals and Cash Flows? (opens in a new tab) — Why the accrual component of earnings is the part that reverses.
Further reading
- Earnings management (opens in a new tab) — Why accrual drift and revenue quality matter for the M-score.
- Accrual accounting (opens in a new tab) — How non-cash accruals feed manipulation-detection models.