Fraudtracker

Screener Famous frauds Circular AI AI bubble risk Valuation Earnings mgmt Forensic-accounting screener · SEC EDGAR

Famous accounting frauds — and how this screener would have caught them

For each case: the mechanism, the metrics that would have fired, and an honest read on whether a quantitative screen alone could have flagged it. Cases ordered by scale and notoriety.

Enron (2001)

$74B market-cap collapse · Houston, energy/trading · CEO Skilling, CFO Fastow convicted

The mechanism

  • Mark-to-market accounting on long-dated energy contracts — booked NPV of decades of future profits in year-one revenue.
  • Special-Purpose Entities (Raptor, LJM, Chewco) held losing trades and underperforming assets off the parent balance sheet.
  • Backed SPEs with Enron stock; when stock fell, structures unwound, forcing losses back onshore.

What our screen would have flared

  • CFO − NI gap deeply negative every year — earnings far exceeded operating cash.
  • TATA elevated — recurring high accruals, classic Sloan signal.
  • AQI rising — soft "investment" assets ballooning relative to PP&E.
  • SGI high (revenue tripled 1997–2000) without proportional cash growth.
  • Beneish M would have read above −1.78 in 2000.
Detection verdict: partial. The screens would have raised concerns, but Enron's worst frauds (the SPEs) were *off* the consolidated statements. The on-book numbers looked manipulated; the truly damaging items were invisible. A screener picks the scent up; only forensic accountants reading footnote 16 of the 10-K caught the smoke.

WorldCom (2002)

$11B fraud · Telecom · Largest U.S. bankruptcy at the time · CEO Ebbers convicted (25 years)

The mechanism

  • Reclassified $3.8B+ of "line costs" (operating expenses paid to other carriers) as capital expenditure.
  • Operating costs got depreciated over 5–10 years instead of expensed immediately.
  • Result: artificially inflated EBITDA, inflated PP&E, suppressed depreciation rate.

What our screen would have flared

  • DEPI > 1 — depreciation rate falling relative to a ballooning PP&E base. The single most diagnostic metric for this fraud.
  • AQI rising — capitalized costs sitting in non-current assets.
  • TATA elevated — earnings divorced from cash flow.
  • CFO − NI gap negative and worsening each year.
  • LVGI rising — the company had to keep borrowing to fund cash shortfalls.
Detection verdict: caught. This is the textbook Beneish case. Every component of M except SGI moves in the wrong direction. Anyone running a quarterly Beneish on WCOM in 2000–2001 would have seen the score deteriorate every period.

Wirecard (2020)

€1.9B fictitious cash · German payments processor · CEO Braun arrested · Auditor EY missed it for years

The mechanism

  • Claimed €1.9B sat in escrow accounts at Asian "trustee banks". The cash didn't exist.
  • Inflated revenue from third-party acquirers in Philippines, Dubai, Singapore — counterparties either fake or vastly overstated.
  • Margins, ROA looked exceptional vs. peers (Adyen, Stripe).

What our screen would have flared

  • SGI with simultaneously surging cash balance is suspicious — payments processors carry float, but Wirecard's growth in cash outpaced economic plausibility.
  • CFO − NI gap looked good *because* the cash was fictitious — paradoxically, Wirecard's numbers screened clean on the standard Beneish components.
  • Goodwill/TA was elevated from acquisition spree (Hermes, GI Card Services, Citi prepaid).
Detection verdict: missed. Wirecard is the classic counterexample to "just run Beneish". When the fraud is fabricated *cash and counterparties* rather than aggressive accruals, the standard ratios look fine — sometimes better than peers'. Detection here required (a) reading the auditor scope notes, (b) FT investigative reporting, (c) short-seller forensic work on counterparties, and (d) margin comparisons to industry. Quantitative screens alone would not have caught it.

Luckin Coffee (2020)

$310M fabricated sales · NASDAQ-listed Chinese coffee chain · COO Liu fired, company delisted, refiled 2022

The mechanism

  • Fabricated transactions through related-party "vendors" who placed bulk orders that were never fulfilled to real customers.
  • Paired with circular cash flows: company paid the vendor, vendor paid back as "revenue".
  • Inflated reported same-store sales growth and average ticket.

What our screen would have flared

  • SGI off the charts — claimed 5.5× revenue growth.
  • DSRI rising — receivables grew much faster than economic activity.
  • Beneish M deep into manipulation territory in 2018 and 2019.
  • Piotroski low — losses, weak CFO, deteriorating liquidity.
  • CFO − NI gap mixed — some of the "circular" cash actually showed up as CFO.
Detection verdict: caught. A Muddy Waters report (anonymous, Jan 2020) compiled the same signal types this screener formalizes — outsized growth, suspiciously rising AR, margins inconsistent with comparable chains. Beneish would have screamed.

Bausch Health / Valeant (2015)

$90B → $7B equity wipe · Pharma roll-up · Channel stuffing through Philidor specialty pharmacy

The mechanism

  • Acquired pharma assets, immediately raised list prices 5×–10×.
  • Routed sales through Philidor, a "captive" specialty pharmacy that wasn't disclosed as a related party — let Valeant book sales without insurer pushback.
  • Goodwill grew from $4B (2010) to $19B (2015) as the roll-up accelerated.

What our screen would have flared

  • Goodwill / TA > 0.55 — extreme acquisition reliance.
  • LVGI elevated — debt-funded M&A.
  • DSRI rising — Philidor inventory effectively held receivables outside the consolidated entity.
  • AQI rising — soft assets dominating the balance sheet.
  • Beneish M > −1.5 in 2014–2015.
Detection verdict: caught (with caveats). The roll-up signature was unmistakable in the metrics. The Philidor relationship itself was off-book and required investigative work (Citron Research short report, October 2015), but the underlying earnings-quality deterioration was visible quarters in advance.

Tyco International (2002)

$600M looting + accounting manipulation · CEO Kozlowski convicted

The mechanism

  • "Spring-loading" — pressured acquired companies to defer revenue and accelerate expenses pre-close, so post-close earnings looked higher.
  • Reported acquisition costs as one-time non-recurring charges to mask ongoing operational deterioration.
  • Plus straightforward executive theft (unauthorized loans, art purchases, the $6,000 shower curtain).

What our screen would have flared

  • SGI high during M&A spree.
  • Goodwill / TA rising rapidly.
  • AQI rising — acquired intangibles bloating soft assets.
  • SGAI volatile — masking ops costs as restructuring charges.
  • CFO − NI gap mildly negative.
Detection verdict: partial. The roll-up signature was visible. But the spring-loading itself only shows up indirectly — earnings look smoother than peers, but no single metric screams. The personal looting wasn't an accounting issue at all.

Lehman Brothers (2008)

$639B bankruptcy · Investment bank · Repo 105 transactions hid leverage

The mechanism

  • "Repo 105" — moved $50B of repo financing off the balance sheet at quarter-end by structuring transactions as "true sales" under a UK legal opinion no U.S. firm would issue.
  • Reported leverage ratios that understated actual leverage by 200–300 bps every quarter-end.
  • Held mortgage-backed assets at marks that didn't reflect 2007–2008 deterioration.

What our screen would have flared

  • LVGI rising (between quarter-ends).
  • AQI rising — "Level 3" assets growing as MBS markets froze.
  • The screener's metrics are tuned for non-financial companies; banks need different screens (T1 capital trends, repo book mix, Level 3 ratio).
Detection verdict: missed. Beneish, Altman, and Dechow are calibrated on industrial firms. Banks game leverage with structured repo and Level 3 marks — different toolkit required. This case is a reminder that *which* screen you use depends on industry.

Satyam Computer Services (2009)

$1B fictitious cash · India's largest corporate fraud · Chairman Raju confessed in a letter to the board

The mechanism

  • Inflated cash balances by $1B for years using fake bank statements.
  • Inflated revenue and operating margins to keep stock price up so chairman's family could borrow against shares.
  • Margins claimed (~24%) far above Indian IT services peers (~18%).

What our screen would have flared

  • Like Wirecard: fictitious cash makes the standard accruals screens look good.
  • Beneish M would have read clean.
  • Industry-relative comparison would have flagged margin outlier — not in our current screener.
Detection verdict: missed. Same lesson as Wirecard: when the lie is on the cash line itself, the cash-vs-earnings sanity checks don't work. Industry-relative outlier detection is the missing layer.

HealthSouth (2003)

$2.7B inflated earnings over 7 years · CEO Scrushy charged (acquitted on accounting, convicted on bribery)

The mechanism

  • "Contractual adjustments" — booked plug entries to make every quarter hit Wall Street estimates.
  • Padded fixed assets and inventory to balance the fake income.
  • Earnings consistently exceeded analyst consensus by exactly $0.01–0.02 for 14 consecutive quarters.

What our screen would have flared

  • TATA and Accruals/TA elevated.
  • CFO − NI gap persistently negative.
  • AQI rising — fake fixed-asset entries.
  • DSRI rising — fake AR entries to balance the books.
Detection verdict: caught. HealthSouth is a textbook Beneish case — every metric moves consistently the wrong way for years. The pattern of perfectly hitting estimates is itself a signal a screener can flag with consensus data.

Olympus Corporation (2011)

$1.7B hidden investment losses over 13 years · Japanese optics maker · CEO Woodford fired for asking questions

The mechanism

  • Hid 1990s securities losses by paying inflated "advisory fees" on later acquisitions to offshore funds that absorbed the losses.
  • $687M in fees to advisors on a $2B Gyrus deal — a 35% advisory fee, vs typical 1–2%.
  • Wrote down acquired assets shortly after purchase as "goodwill impairment".

What our screen would have flared

  • Goodwill / TA rising sharply post-Gyrus.
  • AQI rising — soft assets bloated.
  • Recurring large goodwill impairments would show up as one-time hits — but the screen would have flagged the rising goodwill base.
Detection verdict: partial. The numbers showed acquisition-fueled bloat. But what made Olympus catchable was the *implausibility* of the advisory fee ratios — that requires reading 8-K disclosures, not running a Beneish.

Parmalat (2003)

€14B fictitious assets · Italian dairy · Founder Tanzi convicted

The mechanism

  • Forged a Bank of America letter claiming €3.95B in cash held in a Cayman Islands subsidiary.
  • Maintained 90+ shell entities globally, double-counting and round-tripping cash.
  • Issued bonds to repay older bonds — a Ponzi-style refinancing pattern.

What our screen would have flared

  • LVGI rising — debt issuance pace was unsustainable.
  • Goodwill / TA elevated from acquisition spree.
  • AQI rising — soft assets dominated.
  • But fictitious cash, like Wirecard/Satyam, made standard ratios look healthier than reality.
Detection verdict: partial. Pattern of perpetual debt issuance to a healthy-looking balance sheet is the canonical "too good to be true" trap. The leverage signal is detectable; the fictitious cash isn't, until someone calls Bank of America.

What this list teaches

  1. The screen catches accruals-based fraud reliably. WorldCom, HealthSouth, Luckin, Valeant, classic Beneish wins.
  2. The screen misses fictitious-cash fraud. Wirecard, Satyam, Parmalat all looked clean or *better* than peers because the lie was on the cash line itself.
  3. Off-balance-sheet fraud is a footnote problem, not a screen problem. Enron's SPEs, Lehman's Repo 105 — visible only to readers who interrogate the consolidation perimeter.
  4. Industry-relative comparison is the missing layer. Satyam looked clean against itself, but its margins were outliers vs. Indian IT services peers. Adding sector cohorts would catch that class.
  5. Banks need a different toolkit. Beneish/Altman are calibrated on industrials; financial firms need Level 3 ratio, T1 capital trends, repo book mix.

The screener in this app is a first-pass filter, not an oracle. A green M-Score doesn't mean a company is honest — it means the public numbers don't show the textbook accruals signature. Use the findings as a reading list, not a verdict.