Beneish M-Score

$$-4.84+0.92\cdot DSRI+0.528\cdot GMI+0.404\cdot AQI+0.892\cdot SGI+0.115\cdot DEPI-0.172\cdot SGAI-0.327\cdot LVGI+4.679\cdot TATA$$

Eight-index probabilistic earnings-manipulation detector (Beneish 1999). M > -1.78 flags potential manipulator; ≤ -1.78 reads as clean. Built from YoY indices on receivables, gross margin, asset quality, sales growth, depreciation, SG&A, leverage, and total accruals.

Direction: Lower better Category: forensic

Inputs (XBRL fallback chains):

• current_assets: AssetsCurrent
• net_income: NetIncomeLoss -> ProfitLoss
• operating_cash_flow: NetCashProvidedByUsedInOperatingActivities
• ppe_net: PropertyPlantAndEquipmentNet
• sga_expense: SellingGeneralAndAdministrativeExpense -> GeneralAndAdministrativeExpense

Returns None when: Returns None when prior-year anchor isn't reachable or any of the index inputs is non-positive (the ratios assume positive endpoints).

Source code

Implementation: beneish_m_score in src/eqtytrk/metrics/ratios.py (or valuation.py for multiples).

See also

For full methodology — decomposition, thresholds, when each score lights up vs the others, citations to original papers — see the Forensic screening essay.