Week 35: Advanced Financial Statement Analysis — DuPont, Working Capital, and Distress Models

---

1. Why This Is Important

Week 8 taught you to read the three statements. Week 19 taught you the capital structure they sit on. Week 20 taught you to trust cash over earnings. Week 21 taught you to discount the cash. This week is the operator's wrench set — the small number of compact ratios and scoring models that professional analysts actually use, every quarter, for every name they cover, to answer four very practical questions:

Where does this company's return on equity actually come from? A

17% ROE at JPMorgan is a fundamentally different animal from a 17% ROE at Ford. One is built on a 12× balance sheet and a 4% margin, the other on a 7× balance sheet and a 3% margin doing very different work. The DuPont decomposition splits ROE into its three (or five) drivers so you can tell margin businesses from leverage businesses from turnover businesses, and notice when one of those legs starts to wobble.

Is the working capital a tailwind or a tap on the brake? A

manufacturer that ships product before it gets paid, holds inventory for ninety days, and pays suppliers in thirty is bleeding cash even when the income statement looks fine. The cash conversion cycle — DSO + DIO − DPO — turns that into a single number you can track quarter to quarter and compare across competitors.

Are the earnings being managed? The Beneish M-score combines

eight ratios into one probability that a firm is manipulating earnings. It will not catch every fraud (Wirecard's was a different class of lie), but Beneish flagged Enron in 1997, three years before it imploded. As an investor you do not need a perfect detector. You need something that fires often enough to make you actually open the 10-K when the score lights up.

Is the company a going concern, or is it walking toward Chapter 11?

The Altman Z-score, published by Edward Altman in 1968 and barely modified since, predicts bankruptcy risk over the next two years with roughly 80–90% accuracy on the original test set. Its threshold zones — distress below 1.81, gray 1.81–2.99, safe above 2.99 — are crude, and that is the point. Crude rules survive regimes; finely tuned ones do not. Alpha is rare, but avoiding negative alpha is cheap, and a $0 ten-line spreadsheet is one of the cheapest sources.

The chart below shows what DuPont looks like when you put five very different businesses next to each other. Apple is a margin machine with serious leverage from buybacks. JPMorgan is a leverage machine with thin margins and almost no asset turnover. Ford is a turnover machine with both thin margins and meaningful leverage. The same 13–17% ROE is being manufactured three different ways, and the path to a 0% ROE is different for each.

---

2. What You Need to Know

2.1 DuPont Decomposition — Three-Factor and Five-Factor

The original DuPont formula, named after the analyst department at DuPont Corporation in the 1920s, is identity arithmetic:

$$ \text{ROE} \;=\; \frac{\text{Net Income}}{\text{Equity}} \;=\; \underbrace{\frac{\text{NI}}{\text{Sales}}}_{\text{Net Profit Margin}} \;\times\; \underbrace{\frac{\text{Sales}}{\text{Assets}}}_{\text{Asset Turnover}} \;\times\; \underbrace{\frac{\text{Assets}}{\text{Equity}}}_{\text{Equity Multiplier}} $$

The two middle ratios cancel algebraically. What survives is a description of how a company gets to its ROE. Margin × Turnover is operating efficiency; Equity Multiplier is leverage. A high ROE built on margin is a brand or moat business. A high ROE built on turnover is a logistics or scale business. A high ROE built on leverage is a financial business — or, when leverage rises in a non-financial, an early warning sign.

The five-factor extension splits margin into three pieces — tax burden, interest burden, and operating margin — to isolate where profitability is being squeezed:

$$ \text{ROE} \;=\; \frac{\text{NI}}{\text{EBT}} \times \frac{\text{EBT}}{\text{EBIT}} \times \frac{\text{EBIT}}{\text{Sales}} \times \frac{\text{Sales}}{\text{Assets}} \times \frac{\text{Assets}}{\text{Equity}} $$

Tax burden (NI/EBT) and interest burden (EBT/EBIT) are between zero and one. Each is a leak. The five-factor version is what you reach for when an ROE is moving and you cannot tell whether it is the operating business, the tax code, the cost of debt, the asset base, or the buyback that is doing the work. It almost always turns out to be more than one.

For Apple in FY2024 the three-factor reads roughly NPM 24% × Turnover 1.07 × Equity Multiplier 6.4 → ROE ≈ 165%. That equity multiplier is not leverage in the bad sense — it is the consequence of $725B of buybacks since 2013 (Week 19). Apple has shrunk the denominator faster than the numerator. ROE in that situation is no longer a measure of business quality; it is a measure of how aggressively the company has returned capital. Use ROIC (Week 21) for the cleaner read.

2.2 Cash Conversion Cycle and Working Capital Efficiency

The cash conversion cycle (CCC) is how many days a dollar of input spending stays trapped in the business before it comes back as a dollar of cash from a customer:

$$ \text{CCC} \;=\; \text{DSO} + \text{DIO} - \text{DPO} $$

• DSO (days sales outstanding) = Receivables / Revenue × 365.

How long customers take to pay you.

• DIO (days inventory outstanding) = Inventory / COGS × 365.

How long product sits before being sold.

• DPO (days payables outstanding) = Payables / COGS × 365.

How long you take to pay suppliers.

A negative CCC — Apple, Costco, Amazon — means suppliers finance your inventory. You are running on float. A positive CCC means you are financing the supply chain, which is a working-capital tax that grows with revenue. Watch the trend more than the level. Three rising quarters of DSO is one of the most reliable advance warnings of a revenue-recognition problem.

A second working-capital ratio every analyst eventually internalises is the accruals ratio:

$$ \text{Accruals ratio} \;=\; \frac{\Delta\text{Working Capital} - \Delta\text{Cash}} {\text{Average Total Assets}} $$

This is the Sloan (1996) accruals anomaly in summary form (see Week 20, image/week20_accruals_anomaly.png). High positive values mean the income statement is leading the cash flow statement — the firm is booking revenue and profit faster than the cash arrives. Sloan's quintile spread averaged ~9% per year, and it has not inverted in three decades.

2.3 The Beneish M-Score — Earnings Manipulation Detector

Daniel Beneish's 1999 paper combined eight one-year-change ratios into a single probit-style score. The mnemonic: DSRI, GMI, AQI, SGI, DEPI, SGAI, LVGI, TATA. You will rarely compute it by hand — your data provider does it — but the intuition matters:

Variable	What it captures
DSRI Days Sales in Receivables Index	Receivables growing faster than sales
GMI Gross Margin Index	Margin deterioration year-over-year
AQI Asset Quality Index	Non-current non-PPE assets rising (capitalised costs)
SGI Sales Growth Index	Aggressive growth tempts management
DEPI Depreciation Index	Slowing depreciation (longer useful lives)
SGAI SG&A Index	SG&A rising faster than sales
LVGI Leverage Index	Rising leverage
TATA Total Accruals to Total Assets	The Sloan anomaly piece

The composite is:

$$ M \;=\; -4.84 + 0.92 \cdot \text{DSRI} + 0.528 \cdot \text{GMI} + 0.404 \cdot \text{AQI} + 0.892 \cdot \text{SGI} + 0.115 \cdot \text{DEPI} - 0.172 \cdot \text{SGAI}

• 0.327 \cdot \text{LVGI} + 4.679 \cdot \text{TATA} $$

A score above −1.78 is classified as "likely manipulator". Beneish back- tested it against 74 known manipulators and flagged ~76% with a 17% false-positive rate. It famously fired on Enron in 1997 and 1998, on WorldCom in 1999, on Valeant in 2014. It missed Wirecard, because Wirecard simply made up the cash balance — there was no earnings manipulation footprint, just an outright lie. The lesson: M-score is a screening filter — a tool for looking at the right numbers, not a verdict. When it lights up, you read the 10-K. When it doesn't, you still read the 10-K, just with less urgency.

2.4 The Altman Z-Score — Bankruptcy Prediction

Edward Altman, NYU, 1968. Multiple discriminant analysis on 33 bankrupt and 33 non-bankrupt manufacturers. The original public-firm formula:

$$ Z \;=\; 1.2 \, A + 1.4 \, B + 3.3 \, C + 0.6 \, D + 1.0 \, E $$

Term	Definition	Interpretation
A	Working capital / Total assets	Short-term liquidity buffer
B	Retained earnings / Total assets	Cumulative profitability
C	EBIT / Total assets	Operating productivity of assets
D	Market value of equity / Total liabilities	Market-tested solvency cushion
E	Sales / Total assets	Asset turnover

The cutoffs:

• Z > 2.99 — "safe" zone. Bankruptcy in next 2 years rare.
• 1.81 ≤ Z ≤ 2.99 — gray zone. Watch list.
• Z < 1.81 — distress zone. Materially elevated bankruptcy

probability.

Two warnings. First, Altman calibrated the model on US manufacturers with public equity. The variants — Z' for private firms, Z'' for non-manufacturers and emerging markets — change the coefficients and cutoffs. Use the right one. Second, Z is a noisy point estimate; the trend matters more than any single reading. GE's Z drifted from 2.6 in 2012 down to 1.2 by 2018, three years before the dividend cut and the breakup. The trend line was the signal; the absolute level was just the noise.

The chart below shows three trajectories. GE 2010-2024 (the deteriorate-and-recover curve), Ford 2018-2024 (perpetually parked in the gray zone, which is about right for a cyclical), and Apple 2020-2024 (deeply safe, the shape of a brand business with a clean balance sheet). Same model, same cutoffs, completely different stories.

2.5 Putting It Together — The Two-Page Health Check

In practice, here is what a compact diligence sheet looks like for a new name:

DuPont 3-factor and 5-factor for the trailing 5 years. Are any

of the legs trending in a direction the management story does not explain?

CCC for the trailing 8 quarters. Direction matters more than

level. Compare to two named competitors.

Accruals ratio for the trailing 5 years. Persistent positive

reading is a yellow flag.

Beneish M-score, latest fiscal year. Above −1.78 is a yellow

flag. Above −1.0 is a red flag.

Altman Z-score, latest 5 years. Below 1.81 in two consecutive

years is a position-size question. Trending down through the gray zone is a research question.

None of this is alpha generation in the new-edge sense. It is the opposite: it is the negative-alpha filter. Most amateur portfolios underperform not because they failed to find the next Apple but because they held a Bear Stearns or a Valeant or a GE in the 2017 stretch when a five-minute screen would have asked them to think twice. The interactive lab at the end of this lesson lets you pick a preset firm or punch in your own numbers and watch the Z-score band classification flip in real time.

**Horace's view — this work is defensive, not offensive, and the textbook quietly misframes it.** The orthodox CFA framing puts DCF, quality screens, and the Z-score on the alpha-generation side of the ledger — as if a careful analyst, armed with these tools and a clean spreadsheet, can systematically pick winners that the rest of the market has missed. My own experience says that is the wrong way to read this material. The market is right far more often than I am. Genuine, articulable mispricings — the rare moments where I can point to exactly where the crowd has it wrong and why — are not what DCF produces; DCF produces a fair-value range that mostly agrees, within a wide band, with where the stock already trades. The honest read is that this entire toolkit is *negative-alpha filtering*: it stops you being on the wrong side of an asymmetric mispricing that the rest of the market has already priced in.

That reframing changes how you spend your time. If you treat financial-statement work as alpha generation, you grind harder on the marginal name, looking for an edge that probably isn't there and overtrading when you talk yourself into one. If you treat it as the negative-alpha filter — as the screen that decides which names never enter the portfolio in the first place — you spend the same hours much more productively. The real edges in this course are elsewhere: macro, sector rotation, structural flow mispricings, the occasional tax or instrument-structure advantage. The bottom-up toolkit is what keeps you from owning the next Valeant, not what finds you the next Apple. Worth doing carefully. Not worth mistaking for the source of returns.

---

3. Common Misconceptions

"High ROE is always good." A high ROE built entirely on a rising

equity multiplier is leverage in disguise. Decompose it before you admire it.

"DuPont is just bookkeeping arithmetic." It is identity

arithmetic, but the change in each factor is information. A margin compressing while turnover and leverage hold steady tells you exactly which line of the income statement to investigate.

"A negative cash conversion cycle is a goal." It is a

consequence of supplier power and customer payment terms. You cannot will it into existence; chasing it through aggressive payable- stretching can break supplier relationships and collapses in recessions.

"The Beneish M-score is a fraud detector." It detects the

footprint of accounting manipulation — receivables and accruals and margin changes. Frauds that bypass the books entirely (fake cash balance, fake invoices, related-party transactions) leave no M-score footprint.

"Altman Z below 1.81 means bankruptcy." It means *elevated

probability* of bankruptcy in the next two years, not certainty. Plenty of firms live in the distress zone for years and emerge. Plenty of firms in the gray zone go to zero. Use it as a sizing and research signal, not an exit signal in isolation.

"These models work the same for banks and insurers." They do

not. Banks have a balance sheet that is mostly financial assets; asset turnover is meaningless and Z-score variants are required. Use ROTCE and tangible book trends instead.

"You need a Bloomberg terminal to do this." You need an SEC

EDGAR account, which is free, and a calculator. Every ratio in this lesson can be computed from a 10-K in under thirty minutes.

**"If the M-score and Z-score both look fine, the company is

safe."** They are necessary, not sufficient. Wirecard was fine on both. So was Bernie Madoff's "fund". The models score *what is in the books*. They do not audit whether the books are real.

---

4. Q&A Section

Q: Should I use three-factor DuPont or five-factor? A: Start with three-factor for screening across many names. Move to five-factor when one of the three is moving and you need to know which of tax, interest, or operating margin is doing the work. Five-factor is also essential for cross-country comparisons because tax burdens diverge sharply across jurisdictions.

Q: How often should I recompute these ratios? A: Each fiscal quarter when the 10-Q drops, plus a clean run through the 10-K every January. Working capital ratios in particular are volatile quarter-to-quarter; trend over 4-8 quarters matters far more than a single point.

Q: What is the typical M-score for a clean blue-chip? A: For mature, slow-growing US large-caps the M-score sits around −2.5 to −3.0. Apple, Microsoft, Coca-Cola, JPMorgan all run comfortably below the −1.78 threshold. Aggressive-growth names routinely score between −2.0 and −1.5 simply because the SGI (sales growth index) is elevated; that is a feature of the model, not necessarily a manipulation signal.

Q: Which of the four models is most useful? A: Honestly, the cash conversion cycle. It is the most robust to reporting style, easiest to compute, hardest to fake without leaving a trail elsewhere, and most directly tied to the operating reality. The M-score and Z-score are scoring composites; CCC is a measurement of an actual physical fact about the business.

Q: Can I use the Z-score for an ETF or a fund? A: No. Z-score requires a single corporate balance sheet. For a fund or ETF you would aggregate the holdings — Bloomberg and similar tools report a weighted-average Z-score for index ETFs, but the interpretation is loose. The model was designed for single-firm distress prediction, not portfolio risk.

Q: What is the right benchmark for DSO and DIO? A: The named competitor in the 10-K's own competitive section, plus the industry median. Direction matters more than level. A retailer with DIO of 90 and a competitor at 60 should worry; a retailer whose own DIO went from 60 to 90 over four quarters should worry more.

Q: Why does AAPL's equity multiplier look so extreme? A: Buybacks. Apple has retired roughly $725B of equity since 2013 (see Week 19 chart). The denominator of ROE has shrunk faster than the numerator. The "ROE" reading north of 150% is real arithmetic but not a meaningful measure of business quality at this point. Use ROIC or return on tangible assets instead.

**Q: Does the M-score still work after 25 years of academic publication?** A: Less well than at publication, but it has not collapsed. The mechanical relationships it captures — receivables growing faster than sales, gross margin deteriorating, accruals running positive — remain the actual physical signatures of earnings management. Public disclosure of the formula has raised the cost of crude manipulation, which is itself a form of efficacy.

Q: How does this fit into Horace's four-tranche framework? A: Mostly Tranche 2 (factor / quality sleeve) and Tranche 3 (active / single-name alpha). Tranche 1 (passive index) does not need any of this; you bought the basket. Tranche 4 (cash and dry powder) does not need any of this either. The middle two tranches are where the "look at cash, look at quality" alpha sources live, and these are the ratios that operationalise that look.

Q: I have a stock with a low Z-score. Should I sell? A: Not on the basis of the score alone. First check the trend (is it deteriorating, stabilising, or recovering?). Second check whether the company is being correctly classified — banks, REITs, insurance, and asset-light tech do not behave like 1968 manufacturers. Third look at the bond market: if credit spreads on the company's debt are not widening, the bond market does not believe the equity model. The combination of a deteriorating Z trend and widening credit spreads is the actionable signal, not either alone.

Q: Can these ratios be gamed by management? A: Yes, all of them, partially. Working capital can be window-dressed at quarter-end (factoring receivables, paying suppliers slowly). Margins can be smoothed by changing inventory accounting. Even bankruptcy probability can be lowered for a quarter by a debt-for- equity swap that drives up book equity. The defence is the trend, not the point. A company that consistently games one ratio will eventually break another.