The Sloan Accruals Anomaly

Two companies close the year with $100m of net income each. The first collected $130m of cash from customers. The second collected $40m and booked the rest as receivables and unsold inventory. The income statement calls both of them $100m earners. A year later the first is usually still earning that much, and the second usually is not.

Richard Sloan measured what "usually" is worth in a 1996 paper that became one of the most replicated results in accounting. The finding has two halves, and the second is the interesting one. First: the accrual part of earnings decays faster than the cash part. Second: share prices behave as though it did not.

Splitting earnings in two

Every reported profit breaks into exactly two pieces, with nothing left over:

$$\text{Earnings} = \text{Cash from operations} + \text{Accruals}.$$

Cash from operations is money that actually moved. Accruals are the bridge to reported profit — revenue booked before the customer paid, inventory built and not yet sold, depreciation subtracted with no cheque written. They are not a scandal; they are the point of accrual accounting, which exists so a period's performance is not hostage to when invoices settle.

Measure the accrual half straight off the cash flow statement, scaled so firms of different sizes compare:

$$\text{Accrual ratio} = \frac{\text{Net income} - \text{Cash from operations}}{\text{Average total assets}}.$$

In words: how much of this year's profit never showed up in the bank. Sloan's original measure used balance sheet changes instead — non-cash working capital less depreciation — but Hribar and Collins showed in 2002 that acquisitions and currency translation move those balances with no accrual behind them. Use the cash flow statement.

How much faster does the accrual half decay

Sloan regressed next year's return on assets on this year's two components separately. Across US firms from 1962 to 1991:

$$\text{ROA}_{t+1} = \alpha + 0.765 \times \text{Accruals}_t + 0.855 \times \text{Cash}_t.$$

A dollar of cash earnings carries about 85.5 cents into next year; a dollar of accrual earnings carries about 76.5 cents. He then backed out the coefficient prices were implicitly using. It came out near 0.911 for accruals — the market was treating accrual dollars as more durable than even cash dollars, when they are in fact the flimsier half.

Working the numbers

Both firms from the opening report $100m of net income on average total assets of $1,000m — a 10% return on assets each.

Firm B, the accrual-heavy one, reconciles like this: net income 100, add back depreciation 55, subtract a $70m rise in receivables, a $40m inventory build and a $5m fall in payables. Cash from operations is $100 + 55 - 70 - 40 - 5 = 40$, so its accrual ratio is $(100 - 40)/1000 = 6\%$.

Firm A collected $130m, so its accrual ratio is $(100 - 130)/1000 = -3\%$ — profit understates the cash it took in.

Now forecast next year with Sloan's coefficients, working in return-on-asset units. Firm A: $0.765 \times (-0.03) + 0.855 \times 0.13 = -0.023 + 0.111 = 0.088$, or 8.8% of assets, $88m. Firm B: $0.765 \times 0.06 + 0.855 \times 0.04 = 0.046 + 0.034 = 0.080$, or 8.0%, $80m.

A forecaster who saw only "both earned 10%" and applied the pooled persistence of 0.841 would pencil in 8.4%, or $84m, for each. The split says Firm A beats that by $4m and Firm B misses it by $4m — an $8m spread between two firms that look identical on the income statement.

The trade, and what became of it

Sloan sorted US stocks each year into ten buckets by accrual ratio, bought the lowest decile and shorted the highest, rebalancing four months after fiscal year end so the filings were already public. The hedge portfolio returned about 10.4% a year size-adjusted over 1962–1991, positive in 28 of 30 years, and most of that return clustered around subsequent earnings announcements — what you would expect if the market was being surprised rather than paid a risk premium.

The edge has since faded. Green, Hand and Soliman found the US spread largely gone after 2003, crowded out by hedge fund capital: published, screened on, traded away. The diagnostic use survives — accruals still predict earnings, and still feed the earnings-quality leg of most quality factor definitions.

Common pitfalls

• Using the balance sheet definition on an acquisitive firm. A purchase adds receivables and inventory with no accrual behind them, and the screen flags the buyer.
• Forgetting depreciation is an accrual too. It enters with a negative sign, so a capital-intensive firm can score low accruals for uninteresting reasons.
• Sorting on the level when the change is the signal. Persistently high accruals are structural; a jump this year is news.
• Rebalancing monthly. The signal updates once a year with the filing; faster turnover adds cost, not information.