How to Spot Discretionary Accruals in Earnings

Every dollar of GAAP net income is not a fact. A meaningful slice of it is an estimate — a judgment call about the future, dressed up as a number for the quarter that just closed. That's true at a company doing nothing wrong. It's also true at a company doing something very wrong. Same GAAP. Same statement format. Very different reality underneath.

Accrual accounting exists because cash timing is a poor proxy for economic performance. A company that ships $10 million of product in December and collects the cash in February hasn't had a bad December, and matching the revenue to when it was earned rather than when the check clears is supposed to make the income statement more informative, not less. A pure cash-basis statement would be harder to manipulate in some ways, but it would also make a capital-intensive business with lumpy collections look wildly more volatile than its actual economics — which is exactly why GAAP requires accruals in the first place. But that same flexibility — deciding when revenue is earned, how fast an asset wears out, how much a pending claim will eventually cost — creates room to move. Most of that room gets used the way GAAP intended it to be used. Some of it gets used for something closer to earnings management, and from the outside the two can look identical until you know where to check.

I'll walk through the line between normal accruals and discretionary ones, where discretionary accruals actually live inside a set of financial statements, the operating-cash-flow-vs-net-income screen that works as a practical first pass, and two companies — one an outright fraud, one a legitimate business whose reserve estimates ran too generous for too long — that show what it looks like when the discretion goes too far.

Normal Accruals vs. Discretionary Accruals

Total accruals are simply the gap between net income and cash from operations — everything in reported earnings that isn't cash. Richard Sloan's 1996 study on the accrual anomaly split that gap into two pieces, and the distinction still holds up as the most useful lens available to an outside investor working from public filings alone.

Normal accruals, sometimes called nondiscretionary accruals, move mechanically with the business. Receivables rise when sales rise. Payables shift with purchasing volume. Depreciation follows a schedule set years earlier. None of that requires much fresh judgment quarter to quarter, and none of it is inherently suspicious — a growing company will show growing accruals almost by construction.

Discretionary accruals are the residual: the portion of total accruals that isn't explained by the normal relationship between accruals and business activity. It's where management's estimates, not the business's actual cash-generating activity, are doing the work.

Discretionary accruals: the component of a company's total accruals attributable to management's accounting choices and estimates — reserve levels, capitalization decisions, useful-life assumptions — rather than to the mechanical growth or contraction of the underlying business.

Where Discretionary Accruals Actually Live

Three places account for most of it, and each one shows up in a different part of the filing — the footnotes, the MD&A's critical accounting estimates, or the revenue-recognition policy note, depending on which lever management is pulling.

Reserve and Allowance Estimates

Bad debt allowances, warranty reserves, insurance loss reserves, litigation accruals — each requires management to estimate a future cost today and book it against current-period earnings. A warranty reserve, for instance, is typically set as a percentage of revenue derived from historical claims experience; nudge that percentage down a notch and the current quarter's expense drops with it, with no change to the actual products in the field. Set the reserve too low and earnings look better now at the cost of a future correction. Set it too high in a good year and you've built a cushion to release into a weaker one later, smoothing what would otherwise be a lumpier earnings path. Either direction is a discretionary accrual at work, and neither one requires misstating a single invoice.

Capitalization Judgment Calls

Whether a cost gets capitalized onto the balance sheet or expensed immediately through the income statement changes current-period earnings without changing the cash actually spent. Software development costs, certain engineering spend, and capitalized interest on long construction projects all involve a threshold judgment — has the project reached technological feasibility, does this spending create a future economic benefit — that reasonable accountants can and do disagree on. Two companies spending identical cash on the same kind of internal software project can report meaningfully different current-year earnings depending on where each draws that line, with the difference unwinding only gradually as the capitalized asset amortizes.

Revenue Timing at the Margin

Under ASC 606, revenue recognition itself involves estimates: variable consideration, the timing of when a performance obligation is satisfied, how a contract's transaction price gets allocated across multiple deliverables. None of that requires anything as crude as booking a shipment that never happened. A few days' difference in when a quarter's marginal contracts get recognized is enough to turn a miss into a beat, repeatedly, without ever crossing into what an auditor would flag as fraud — auditors set materiality thresholds precisely because judgment calls this small are expected to occur inside any real set of books.

The Operating-Cash-Flow-vs-Net-Income Screen

The practical starting point for an individual investor doesn't require reconstructing management's estimates line by line. It requires pulling the cash flow statement next to the income statement and asking one question, repeated across several years: does operating cash flow track net income, or does net income keep outrunning it?

Total accruals equal net income minus cash flow from operations. Scale that figure by average total assets and you get an accruals ratio that's comparable across companies and over time. A ratio that's small and stable is unremarkable. A ratio that's large, positive, and persistent across multiple years — net income consistently ahead of operating cash flow, year after year — is close to the profile Sloan's original research tied to weaker forward earnings and, disproportionately, to later restatements. Sloan's decile-sorted portfolios showed the highest-accrual firms underperforming the lowest-accrual firms by a wide margin over the following year, which is a stronger statement than "watch out" — it's a measurable, persistent mispricing that the market took years to start pricing in. That specific gap gets fuller treatment in a piece on cash flow and earnings divergence.

None of this proves anything on its own. A capital-intensive company mid-expansion can show a wide gap for perfectly good reasons — working capital build ahead of growth, or the timing of a large contract's deferred-revenue recognition. So the screen tells you where to look harder, not what you'll find when you get there. One wide year is noise. Three or four in a row, without an obvious operational explanation, is a pattern worth pulling the footnotes over.

Two Companies, Two Kinds of Discretion

Waste Management and the Depreciation Estimates That Weren't

Waste Management is the clearest historical case of discretionary accruals crossing into fraud, and it's clear precisely because the mechanism was accrual estimation, not phantom revenue. Between 1992 and 1997, the company overstated pretax earnings by a cumulative $1.7 billion — restated via a 10-K/A filed in February 1998, at the time the largest restatement in U.S. corporate history — almost entirely by manipulating depreciation. Management repeatedly extended the assumed useful lives of garbage trucks and containers and inflated their expected salvage values, understating depreciation expense quarter after quarter. The SEC's litigation release notes the company assigned dumpsters a useful life of 15 to 20 years against an industry norm closer to 12. Reserves got the same treatment in reverse: the company under-accrued some reserves and improperly reversed others directly into income, turning a balance-sheet cushion into a same-quarter earnings boost.

None of that shows up as one suspicious line item in any single year's filing. It shows up as a pattern — reported earnings consistently ahead of what the business was actually generating in cash, sustained for the better part of a decade before the estimates could no longer absorb the gap.

General Electric's Insurance Reserves

Contrast that with a case that wasn't fraud at all. GE Capital's long-term-care insurance book had gone years without originating new business, but the reserves backing old policies still depended on actuarial assumptions about claims and interest rates that GE, like every insurer carrying long-duration liabilities, had to keep re-estimating. GE's fourth-quarter and full-year 2017 10-K disclosed a $6.2 billion after-tax GAAP charge tied to a comprehensive review of those reserves, and laid out a plan to contribute roughly $15 billion in additional statutory reserves over the following seven years, through 2024.

I'm less sure than I'd like to be about how much of that gap reflects reserve assumptions that were simply too optimistic for years, versus a genuine, non-culpable shift in long-term-care claims trends and low interest rates squeezing every insurer in the category at once. The honest answer is probably some of both, and untangling the split from outside the company is hard. But the shape of the disclosure is the same shape as Waste Management's, minus the intent: reserve estimates that flattered earnings for years, followed by a large, unavoidable catch-up once the estimates could no longer hold.

Put the two side by side and the lesson isn't "distrust every reserve." It's that a reserve estimate is a claim about the future dressed up as a number, and claims about the future are exactly the kind of thing that deserves a second look when they've been running one direction for a long time.

Key Takeaways

• Total accruals split into normal accruals, which move mechanically with business activity, and discretionary accruals, the residual driven by management's estimates and judgment calls.
• Reserve estimates, capitalization decisions, and revenue-timing judgments are where discretionary accruals concentrate — none of the three requires anything as crude as fabricating a sale.
• The operating-cash-flow-vs-net-income gap, scaled by average total assets into an accruals ratio, is the practical starting screen: persistent and widening is worth investigating, one wide year usually isn't.
• Waste Management shows discretion crossing into fraud; GE's insurance reserves show the same accounting mechanism producing a large correction without any allegation of intent — the 10-K reveals the pattern either way, and the screen doesn't tell you in advance which kind you're looking at.