Beneish M-Score: Formula & Example

Table of Contents

    The Beneish M-Score helps investors decide where to investigate the quality of a company’s reported earnings. It combines changes in receivables, margins, assets, sales, depreciation, expenses and leverage with a measure of accruals. The useful result is a set of questions to take back to the financial statements, rather than an accusation based on one number.

    This guide defines an identified eight-variable version, calculates it from two years of hypothetical statements, and explains its cutoff and limitations. The worked approach follows Appendix A of Beneish, Lee and Nichols’s 2013 paper, including its cash-flow-based accrual measure and year-end asset denominator. It is intended for analysis of nonfinancial operating companies, with a U.S. reporting perspective.

    What Is the Beneish M-Score?

    Messod Beneish developed the model to distinguish companies with characteristics associated with earnings manipulation from other companies. Its inputs can reflect possible reporting distortions or business pressures that warrant closer attention. The original paper’s public abstract explicitly cautions that an apparent distortion may instead have another structural explanation.[1]

    A Screen for Further Research

    A flagged score means the company meets a specified screening rule. It does not prove fraud. An unflagged score does not establish that its accounts are reliable, that the business is healthy, or that its shares are attractively valued.

    Use the score within fundamental analysis. Before forecasting earnings or comparing valuation multiples, investigate whether the reported inputs describe the business consistently and whether unusual movements have supportable explanations.

    The Eight-Variable Beneish M-Score Formula

    The following equation uses the coefficients in the authors’ 2013 Appendix A presentation of the unweighted probit model. “Unweighted” names the estimation version; the equation still has a different coefficient for each input. Here, DSRI corresponds to the paper’s DSR, LVGI to LEVI, and TATA to its Accruals variable.[2]

    Eight-Variable Score
    M = −4.84 + 0.920 × DSRI + 0.528 × GMI
    + 0.404 × AQI + 0.892 × SGI + 0.115 × DEPI
    − 0.172 × SGAI + 4.679 × TATA − 0.327 × LVGI
    A unitless score. Retain the intercept, coefficient signs and specified variable definitions.

    The raw score is not a percentage. A result of −1.5 is higher than −2.5. Although the model was estimated using a probability framework, this negative fitted score is not itself a company’s probability of fraud. A reliable present-day probability interpretation would require appropriate validation for the population and data being analyzed.

    Define the Eight Inputs Before Calculating

    Let t mean the current fiscal year and t−1 the preceding year. Use comparable annual flows and the corresponding year-end balances. In the table, Sales means net sales, AR means net receivables, CA means current assets, PPE means net property, plant and equipment, TA means total assets, and D means PP&E depreciation expense. CFO is cash flow from operating activities.

    Input definitions for the version used in this guide
    Input Calculation What a higher input describes
    DSRI: Days’ Sales in Receivables Index (ARt / Salest) / (ARt−1 / Salest−1) More receivables relative to sales than in the prior year.
    GMI: Gross Margin Index GMt−1 / GMt, where GM = (Sales − COGS) / Sales A lower current gross-margin fraction when both margins are positive.
    AQI: Asset Quality Index [1 − (CAt + PPEt) / TAt] / [1 − (CAt−1 + PPEt−1) / TAt−1] A larger share of assets outside current assets and net PP&E.
    SGI: Sales Growth Index Salest / Salest−1 Higher sales growth; 1.20 means a 20% increase.
    DEPI: Depreciation Index DRt−1 / DRt, where DR = D / (D + PPE) A lower current depreciation rate under this definition.
    SGAI: SG&A Expense Index (SG&At / Salest) / (SG&At−1 / Salest−1) More selling, general and administrative expense per sales dollar.
    LVGI: Leverage Index LEVt / LEVt−1, where LEV = (Current liabilities + Long-term debt) / TA Higher leverage using this particular liability measure.
    TATA: Total Accruals to Total Assets (Income before extraordinary itemst − CFOt) / TAt A larger earnings-minus-operating-cash-flow amount relative to current year-end assets.

    These constructions follow Exhibit A1, with leverage clarified by the paper’s Appendix B.[2] GMI and DEPI run prior year divided by current year. Reversing them changes their meaning. The other change indices run current divided by prior; TATA is a current-year accrual level, not a year-over-year index.

    An index of 1 means that its underlying quantity or ratio is unchanged. It does not establish an industry norm or a safe company. DSRI compares receivables-to-sales ratios rather than requiring a separate days-sales-outstanding calculation.

    Use the Right Statement Lines

    • Assets: Use net PP&E, not gross original cost. CA already includes current receivables; do not add AR again in AQI. AQI’s residual can contain legitimate intangible assets and other noncurrent assets.
    • Depreciation: Use PP&E depreciation consistently. Exhibit A1 separates amortization from the combined depreciation-and-amortization amount. The rate denominator is current depreciation plus net PP&E, not accumulated depreciation plus net PP&E.
    • Leverage: Include current liabilities plus long-term debt. Replacing that sum with only interest-bearing debt changes the variable. Avoid counting current debt twice when mapping disclosures.
    • Signs and units: Enter COGS, SG&A and depreciation as positive expense magnitudes. Retain the actual sign of income and CFO. Match currency units throughout.

    The income statement and comprehensive income guide explains the earnings subtotals. For this calculation, neither comprehensive income nor EBITDA is an interchangeable substitute for the specified income input.

    Which Accrual Method Does This Version Use?

    The original 1999 model estimated accruals from balance-sheet changes. The 2013 authors instead use income minus operating cash flow. These are different constructions and should not be silently mixed in a calculation or comparison.[2]

    This guide follows Exhibit A1’s current year-end total-assets denominator. The same paper’s Appendix B prose alternatively describes average assets; that denominator is not used here. Record the choice when comparing a vendor score or another worked example. The historical income label also requires careful mapping to the actual statements; our example assumes no extraordinary or discontinued items, so its supplied net income equals the specified income input.

    The statement of cash flows provides CFO. Use operating cash flow rather than free cash flow, investing cash flow, or the total change in cash. The SEC’s statement-reading guide explains how operating cash flow reconciles earnings with noncash items and operating asset and liability movements.[3]

    The Fitted Signs Matter

    SGAI and LVGI have negative coefficients in this equation. Holding every other input fixed, a larger SGAI or LVGI lowers the score. Their economic rationale as possible signs of business pressure does not make their fitted contributions positive. Likewise, the coefficients do not show that increasing leverage or expenses causes a company to become more trustworthy.

    How to Interpret the M-Score Cutoff

    The 2013 presentation uses M greater than −1.78 to flag a potential manipulator. Its discussion ties that choice to treating the cost of missing a manipulator as roughly 20 to 30 times the cost of falsely flagging another company. It is a historical screening objective, not a universally optimal dividing line.[2]

    The specified screening rule
    Unrounded score Result under this rule Appropriate response
    M > −1.78 Flagged. Validate the inputs and investigate the underlying changes.
    M ≤ −1.78 Not flagged. Continue ordinary reporting-quality analysis; the result does not rule out manipulation.

    Do not silently replace this cutoff with −2.22 while describing the same decision rule. A different cutoff changes which companies are flagged and needs its own documented rationale. Compare the unrounded score with the chosen threshold, especially near the boundary.

    A Beneish M-Score Example From Two Years of Statements

    Consider Cedar Components, a fictional nonfinancial manufacturer. The selected statement inputs below are hypothetical USD millions. Both fiscal years cover twelve months, use the same reporting basis and business boundary, and have year-end balance-sheet amounts. There are no extraordinary or discontinued items. Depreciation covers PP&E only; the table is an input extract rather than complete financial statements.

    Hypothetical selected inputs for Cedar Components (USD millions)
    Statement input 20X1: prior year 20X2: current year
    Net sales 1,000 1,200
    Cost of goods sold 600 780
    Net receivables 150 240
    Current assets 500 600
    Net PP&E 400 440
    Total assets 1,000 1,200
    PP&E depreciation expense 50 40
    SG&A expense 100 132
    Current liabilities 200 260
    Long-term debt 200 280
    Net income / specified income input 100 130
    Operating cash flow 90 50

    1. Calculate the Eight Variables

    First compute each index from the underlying statement amounts. Retain full precision through the final sum; the six-decimal displays below are for readability.

    From statement inputs to M-score variables
    Variable Substitution Result
    DSRI (240 / 1,200) / (150 / 1,000) 1.333333
    GMI [(1,000 − 600) / 1,000] / [(1,200 − 780) / 1,200] 1.142857
    AQI [1 − (600 + 440) / 1,200] / [1 − (500 + 400) / 1,000] 1.333333
    SGI 1,200 / 1,000 1.200000
    DEPI [50 / (50 + 400)] / [40 / (40 + 440)] 1.333333
    SGAI (132 / 1,200) / (100 / 1,000) 1.100000
    LVGI [(260 + 280) / 1,200] / [(200 + 200) / 1,000] 1.125000
    TATA (130 − 50) / 1,200 0.066667

    Receivables rise from 15% to 20% of sales, while gross margin falls from 40% to 35%. TATA uses the 80 / 1,200 earnings-cash gap, entered as approximately 0.066667. Entering 6.6667 instead would mistakenly inflate that input one hundredfold.

    2. Apply the Coefficients and Add the Intercept

    Hypothetical score contributions, displayed to six decimals
    Term Multiplication Contribution
    Intercept Fixed −4.840000
    DSRI 0.920 × (4/3) 1.226667
    GMI 0.528 × (8/7) 0.603429
    AQI 0.404 × (4/3) 0.538667
    SGI 0.892 × 1.20 1.070400
    DEPI 0.115 × (4/3) 0.153333
    SGAI −0.172 × 1.10 −0.189200
    TATA 4.679 × (1/15) 0.311933
    LVGI −0.327 × 1.125 −0.367875
    Cedar’s Calculated Score

    Adding the intercept and all eight contributions using unrounded values gives M = −1.4926464286, displayed as −1.493. Rounded table terms may produce a small summation difference.

    Because −1.4926464286 > −1.78, Cedar is flagged under the specified screening rule. This result identifies a reason to investigate; it does not establish that Cedar’s hypothetical accounting is improper.

    3. Check What Changes the Result

    For a mathematical sensitivity, keep all inputs fixed except current-year CFO. Compare the original CFO of 50 with CFO of 130, equal to the supplied income. This changes only TATA in the equation; it is not a forecast or a fully reconstructed alternative set of statements.

    Hypothetical sensitivity to the current-year cash-flow input
    20X2 CFO (USD millions) TATA M-score Above −1.78?
    50 (130 − 50) / 1,200 = 0.066667 −1.493 Yes.
    130 (130 − 130) / 1,200 = 0 −1.805 No.

    Removing the accrual contribution lowers the score by 4.679 × (80 / 1,200) = 0.3119333333, to −1.8045797619. Crossing the line changes the screen’s label, while the receivables, margin and other questions still exist. CFO itself also deserves scrutiny: operating decisions and classification choices can influence its presentation.[4]

    What If an Input Is Zero, Negative or Missing?

    Do not force a score when the required ratios are undefined or lose their usual meaning. A zero prior-year receivables-to-sales ratio prevents ordinary DSRI division; a zero current gross margin prevents GMI; and a zero prior leverage ratio prevents LVGI. AQI also fails when the prior residual asset share is zero. Missing information is not a reported zero.

    Very small denominators can create enormous ratios. Negative margins, sign changes, or inconsistent asset classifications can produce finite numbers with misleading interpretations. For example, GMI is not a straightforward “margin deterioration” indicator when margins cross zero. Record the problem and investigate the underlying statements rather than applying the ordinary cutoff confidently.

    A negative income figure or negative CFO does not automatically invalidate TATA when total assets are positive. Keep the actual signs. By contrast, zero sales or nonpositive total assets can defeat the conventional constructions. Do not replace invalid denominators with 1, add an arbitrary tiny number, or quietly cap an extreme result.

    The 2013 study capped variables at annual sample-based 1st and 99th percentiles, a procedure called winsorization. This single-company teaching calculation uses raw inputs and no invented percentile caps. Reproducing a research or vendor score requires matching its data definitions and outlier treatment as well as its equation.[2]

    How to Investigate a Flagged M-Score

    Begin with the company’s annual filing, statements and notes. The following is an analytical workflow, not an audit procedure or a finding about management:

    1. Validate the extract. Check issuer, fiscal-year dates, currency units, original versus restated figures and line definitions. Confirm that an annual income amount was not paired with quarterly CFO. Separate information available at the screening date from later corrections.
    2. Trace receivables and revenue. Examine payment terms, customer concentration, allowances, collections and receivables sales. Ask whether acquisitions or a change in sales mix explain the increase. Use the revenue recognition guide for the underlying accounting questions.
    3. Explain margin and asset changes. Compare pricing, input costs and business mix. Identify what entered AQI’s residual assets, including acquisitions and capitalized costs; a larger residual is not automatically an overstated asset.
    4. Read depreciation estimates. Check asset additions/disposals, useful lives, residual values and methods. A lower rate can reflect asset composition as well as a changed estimate.
    5. Reconcile earnings, cash and financing. Follow working-capital movements, noncash items, debt changes and covenants. Compare management’s explanation with note disclosures and subsequent periods.
    6. Seek corroboration and preserve uncertainty. Read the auditor’s report, control disclosures and relevant amendments. Our audit opinions guide explains assurance distinctions. Record the supported explanation, unanswered questions and evidence needed next.

    A rapid expansion can legitimately create sales growth, more receivables and cash tied up in operations. An acquisition can alter asset mix and comparability. These are possible explanations to test against evidence, rather than automatic excuses or proof of manipulation. The original screening caution and broader reporting-quality framework both support that distinction.[1][4]

    Beneish M-Score vs. Earnings-and-Cash-Flow Analysis

    Related analyses answer different questions
    Approach Useful question What remains to investigate
    Eight-variable M-score Does a defined combination of statement characteristics cross the selected screening threshold? Input validity, structural causes, population fit and supporting evidence.
    Accrual or earnings-CFO comparison How much does reported earnings differ from operating cash generation? Normal working capital, noncash expenses, classifications and the source of the gap.
    Detailed filing analysis What policies, estimates and transactions explain the reported amounts? Whether explanations are supported and what external or later evidence is needed.

    The score’s TATA input overlaps with accrual analysis, but the other seven variables add different information. Neither approach by itself estimates intrinsic value, predicts a bankruptcy date, or establishes legal fraud.

    Common Mistakes When Using the Beneish M-Score

    • Mixing versions. Record the eight-variable equation, accrual method, asset denominator, outlier policy and cutoff together.
    • Reversing GMI or DEPI. Both divide the prior ratio by the current ratio.
    • Calling every higher input a positive contribution. SGAI and LVGI have negative fitted coefficients here.
    • Substituting familiar ratios. Total-debt-to-assets and debt-to-equity are not interchangeable with this LVGI construction; EBITDA is not the TATA income input.
    • Rounding before classification. Retain full precision, and document an input problem instead of hiding it with a rounded score.
    • Using the answer as a verdict. A flag needs investigation, and an unflagged result cannot close an otherwise unresolved reporting concern.

    Limitations of the Beneish M-Score

    Historical Evidence Has a Defined Scope

    The original estimation used 1982–1988 data and a 1989–1992 holdout. The 2013 extension excluded financial-services companies and imposed size and data-availability screens. Those studies do not supply a universal current fraud probability for every industry, country or accounting framework.[2]

    Financial firms require different analysis. Banks and insurers have business models and statement structures that differ from the operating-company relationships used here. Do not apply the ordinary formula and cutoff to them merely because some fields can be populated.

    Both false flags and missed cases matter. In the authors’ selected retrospective table of high-profile cases, Enron is flagged and WorldCom/MCI is not. These are the paper’s historical classifications, not fresh calculations in this guide or a general success-rate estimate.[2] Our Enron scandal case study examines the transactions and obligations that a screening number cannot explain.

    Comparability and behavior change. Acquisitions, reorganizations, fiscal-calendar changes, accounting policies and evolving reporting techniques can alter the inputs. An asset-light company’s normal asset mix can differ markedly from a manufacturer’s. A recent score should be examined with the company’s own history, comparable businesses and original disclosures.

    The underlying data can be wrong or incomplete. The model relies on reported amounts and cannot independently verify hidden transactions. A smooth numerical answer does not eliminate missing information. It is also no substitute for valuation, liquidity analysis or an assessment of investment risk.

    Frequently Asked Questions


    It is flagged under the historical rule used in this guide’s identified eight-variable version. Check the inputs and investigate the related statements and notes. The cutoff reflects a tradeoff between missed manipulators and false flags; crossing it does not prove fraud.


    Yes. Seven inputs compare the current and preceding year. The cash-flow-based TATA used here needs current-year income, CFO and year-end assets. A different accrual construction can require additional balance-sheet inputs, so identify the version before collecting data.


    They may use different model variants, accrual methods, asset denominators, field mappings, reporting periods, restatements or outlier adjustments. Compare those choices before attributing the difference to arithmetic. A different cutoff can also change the label even if the raw score is identical.


    An undefined required ratio prevents the ordinary calculation. Record which input fails and examine its cause. Do not silently fill it with zero or one. Negative or near-zero margins and other unusual denominators can also undermine interpretation even when division produces a finite number.


    The presentation used here excludes financial firms. Their balance sheets, revenues and operating relationships require industry-specific analysis. Populating the equation with available bank fields does not validate its ordinary cutoff or probability interpretation.


    No. A lower score may avoid this screening flag, but it does not establish trustworthy reporting, financial strength or an attractive price. Investment analysis still needs evidence about the business, cash generation, obligations, valuation and risks.

    Sources and Methodology

    Sources accessed November 15, 2026. Model definitions and the selected cutoff are historical; the worked company and sensitivity are hypothetical. This guide follows the 2013 paper’s Appendix A/Exhibit A1 construction, with year-end assets for TATA and raw, untrimmed example inputs. It does not reproduce every empirical data-processing step or recalculate an actual issuer’s score.

    1. Messod D. Beneish, The Detection of Earnings Manipulation, September 1999, public publisher abstract. Screening purpose and the need to distinguish manipulation from structural explanations; premium full text was not accessed.
    2. Messod D. Beneish, Charles M.C. Lee and D. Craig Nichols, Earnings Manipulation and Expected Returns, Financial Analysts Journal 69(2), March/April 2013. Printed pp. 60–61; Appendix A/Exhibit A1, pp. 76–77; Appendix B/Table B1 and relevant notes, pp. 78–80. Author-coauthored paper hosted by the University of Delaware.
    3. U.S. Securities and Exchange Commission, Beginners’ Guide to Financial Statement, February 4, 2007. Statement relationships, operating cash-flow reconciliation, footnotes and MD&A.
    4. CFA Institute, Financial Reporting Quality, 2026 public refresher introduction and summary. Reporting versus results quality, accounting choices, estimates and CFO considerations.
    Disclaimer

    This article is for educational and informational purposes only and does not constitute investment, accounting or legal advice. Screening results are not findings of fraud or recommendations to buy, sell or short a security. Review original disclosures and consult qualified professionals where appropriate.