The Accruals Anomaly (Sloan 1996) — Verified Numbers, and a Documented Death

Sloan (1996) showed that the accrual component of earnings is less persistent than the cash-flow component, and that stock prices behave as if investors do not know this: firms with high accruals subsequently underperform, firms with low accruals outperform (https://www.cuhk.edu.hk/acy2/workshop/June2009Wasley/1996TAR%29.pdf). The original hedge-portfolio return is 10.4% in the first year, not the ~20% an earlier version of this page carried. The effect is real, was large, was stable for 30 years — and has since been documented as decayed to the point of being no longer reliably positive (https://doi.org/10.1287/mnsc.1110.1320).

What Sloan (1996) actually measured

Published in The Accounting Review 71(3):289-315 (https://doi.org/10.2308/tar-9608042309) — not the Journal of Accounting Research, as this page previously claimed.

Sample

40,679 firm-year observations, NYSE and AMEX, covering the 30 fiscal years from 1962 to 1991, from Compustat with returns from CRSP. The funnel is stated explicitly: 71,732 NYSE/AMEX firm-years on the Compustat tapes, reduced to 53,322 by requiring the data to compute accruals, then to 42,120, then to 40,679 by requiring the following year's income (https://www.cuhk.edu.hk/acy2/workshop/June2009Wasley/1996TAR%29.pdf).

Accruals definition — a correction that matters. Sloan does *not* define accruals as "net income minus operating cash flow". He uses a balance-sheet formula: *the change in non-cash current assets, less the change in current liabilities (exclusive of short-term debt and taxes payable), less depreciation expense, all divided by average total assets*. Cash flow is then the residual: earnings minus accruals, where earnings is income from continuing operations over average total assets (https://www.cuhk.edu.hk/acy2/workshop/June2009Wasley/1996TAR%29.pdf). The cash-flow-statement definition became common later; using it is a different signal from the one these numbers were measured on.

Persistence (the mechanism, H1)

pooled, overall earnings persistence is 0.841 (t=303.98). Split into components, the accrual coefficient is 0.765 against a cash-flow coefficient of 0.855, with equality rejected at F=614.01. On decile ranks the gap is wider: 0.565 vs 0.838, F=4894.24. The accrual coefficient is the smaller one in 86% of industries in levels and 99% in ranks (https://www.cuhk.edu.hk/acy2/workshop/June2009Wasley/1996TAR%29.pdf).

The trading-strategy numbers, with the right t-statistic on the right row

Decile portfolios on accrual magnitude, returns cumulated from four months after fiscal year-end, averaged over the 30 annual observations (https://www.cuhk.edu.hk/acy2/workshop/June2009Wasley/1996TAR%29.pdf):

Size-adjusted returns

- Year t+1: lowest-accrual decile +4.9% (t=2.65), highest-accrual decile −5.5% (t=−3.98), hedge 10.4% (t=4.71). - Year t+2: lowest +1.6% (t=1.17), highest −3.2% (t=−2.25), hedge 4.8% (t=3.15). - Year t+3: lowest +0.7% (t=0.55), highest −2.2% (t=−1.61), hedge 2.9% (t=1.64) — not significant.

Jensen alphas (same paper, different column): hedge 10.4% (t=4.42) in year one, 4.8% (t=2.41) in year two, 3.8% (t=1.62) in year three.

Read that pair carefully. The hedge return is 10.4% in *both* the size-adjusted and the Jensen-alpha column, but the t-statistics differ — 4.71 and 4.42 — because they are different tests. Quoting "10.4% (t=4.42)" as the size-adjusted result is exactly the failure mode this wiki keeps finding: a true number welded to the wrong source because the two sat next to each other. See What Counts as an Edge Here: The Evidence Bar This Wiki Applies to Every Technique.

Stability

the hedge return was positive in 28 of the 30 years. The only losing years were 1966 (−19.5%) and 1981 (−2.2%). Sloan uses that better-than-90% hit rate to argue against a risk-based explanation, supported by the hedge portfolio's beta of 0.02 (long decile beta 1.25, short decile 1.23). Beyond year three, abnormal returns from years 4 through 10 were statistically insignificant in each year (https://www.cuhk.edu.hk/acy2/workshop/June2009Wasley/1996TAR%29.pdf).

Post-publication decay — now sourced, and it is the headline

Green, Hand & Soliman (2011), *Management Science* 57(5):797-816 (https://doi.org/10.1287/mnsc.1110.1320), state it directly in the abstract: "we document that the hedge returns to Sloan's ... accruals anomaly appear to have decayed in U.S. stock markets to the point that they are, on average, no longer reliably positive" (abstract read at https://ideas.repec.org/a/inm/ormnsc/v57y2011i5p797-816.html).

Their attributed mechanism is arbitrage, not statistical illusion: the decay "stems in part from an increase in the amount of capital invested by hedge funds into exploiting it, as measured by hedge fund assets under management and trading volume in extreme accrual firms", with a decline in the size of the mispricing signal itself playing a "(weaker) role" (https://doi.org/10.1287/mnsc.1110.1320).

This is the diagnostic pattern described in Out-of-Sample vs Post-Publication Decay: The Two Numbers That Tell You If a Premium Is Real: an effect that survived its out-of-sample window and then faded after publication reads as a genuine mispricing competed away, not as an overfit. What this page does not claim: that the accruals premium fell by the specific ~26%/~58% figures McLean & Pontiff (2016) report as averages across 97 predictors (https://doi.org/10.1111/jofi.12365). Those are cross-sectional averages; Green, Hand & Soliman give no percentage decay for accruals, and none should be inferred.

Transaction costs — what is known and what is not

Sloan (1996) reports no transaction-cost numbers at all. The paper contains no cost estimate, no breakeven threshold, and no net-of-cost return; searching the full text for transaction, trading or bid-ask cost figures returns nothing. All the returns above are gross. What Sloan does offer is a qualitative caveat in his conclusion: "The information acquisition costs and processing costs associated with implementing the strategy outlined in this paper in real time are non-trivial. Moreover, the returns to exploiting the strategy are potentially limited by price pressure effects" (https://www.cuhk.edu.hk/acy2/workshop/June2009Wasley/1996TAR%29.pdf).

The paper usually cited for the cost question is Mashruwala, Rajgopal & Shevlin (2006), *Journal of Accounting and Economics* 42(1-2):3-33 (https://doi.org/10.1016/j.jacceco.2006.04.004). Its DOI resolves but its full text is paywalled and was not read for this page, and RePEc publishes no abstract for it (https://ideas.repec.org/a/eee/jaecon/v42y2006i1-2p3-33.html). This page therefore carries no figure from it — no cost assumption, no breakeven level, no net return. Its title states the finding it is cited for (idiosyncratic risk and transaction costs as limits to arbitrage); anything more precise must be read from the paper. Use Transaction Cost Accounting — the arithmetic that separates a real edge from a paper one and Computing a Strategy's Transaction-Cost Threshold — the four-step check that decides whether a documented edge is tradable to compute the breakeven yourself against the 10.4% gross figure above.

What does NOT work

- Do not trade the 10.4% as a live expectation. It is a 1962-1991 gross, size-adjusted number, and the effect it measures has been documented as no longer reliably positive (https://doi.org/10.1287/mnsc.1110.1320). - Do not use the cash-flow-statement accrual definition and expect these numbers. Sloan's results are measured on the balance-sheet formula above. - Do not quote a net-of-cost verdict from this page. Nobody's cost figure has been read; the gross number is all that is established. - Claims removed from the previous version of this page, none of which survived contact with the source: an unsourced hedge return of roughly 20% per year (the paper says 10.4%); a vague "t-statistic ~4.0 or above" (the paper gives 4.71 and 4.42, for different tests); the definition of accruals as net income minus operating cash flow; a phantom "Sloan and Watts (2002)" citation that resolves to no real paper; the journal attributed as Journal of Accounting Research; and a Piotroski (2000) citation given with a title and journal that do not exist — the real paper is "Value Investing", Journal of Accounting Research 38 (https://doi.org/10.2307/2672906), and it is not about the accruals anomaly.

Per What Counts as an Edge Here: The Evidence Bar This Wiki Applies to Every Technique, this page now clears the bar for effect size, significance, sample and persistence, and clears it for post-publication survival with a negative answer. It does not clear it for cost survival.

Related

- What Counts as an Edge Here: The Evidence Bar This Wiki Applies to Every Technique — the evidence bar; this page passes on magnitude and decay, fails on net-of-cost. - Out-of-Sample vs Post-Publication Decay: The Two Numbers That Tell You If a Premium Is Real — the decay framework this anomaly is a textbook instance of: real, then arbitraged away. - Transaction Cost Accounting — the arithmetic that separates a real edge from a paper one — compute the breakeven against 10.4% gross rather than trusting a verdict. - The Size Factor (Small-Minus-Big) — a textbook case of post-publication decay this wiki uses as a yardstick — parallel case of an anomaly whose decay and liquidity problems are documented. - What Counts as an Edge Here: The Evidence Bar This Wiki Applies to Every Technique — the checklist that catches "right number, wrong column" errors like the 4.71/4.42 pair above.