Risk Premia vs Anomalies

Value stocks have beaten growth stocks over long stretches of history. So has a portfolio that shorts companies with abnormally high accruals. Both facts look identical in a backtest — a positive average return with a t-stat above 2 — but they can come from two entirely different places, and telling them apart is the difference between a strategy that survives being discovered and one that dies the moment a paper is published about it.

Two stories for the same-looking return

A risk premium is compensation for bearing a risk that other investors are unwilling or unable to bear. The value premium's leading academic explanation is that value firms are more exposed to distress and economic downturns — they're cheap because the market is pricing in a real chance of trouble, and the extra return is the reward for holding that risk. If this story is right, the premium should persist indefinitely, because the underlying risk doesn't go away just because people notice it. It will still show up in bad years, by construction, since that's precisely when the compensated risk hits.

An anomaly is a return pattern that comes from a market friction or a behavioral bias rather than genuine risk-bearing — something the theory says shouldn't earn a premium at all. The accruals anomaly is the textbook case: firms that use aggressive accruals to inflate current earnings tend to underperform afterward, and the leading explanation is that investors are fooled by the earnings number and don't look through to the lower-quality cash flows underneath (see The Sloan Accruals Anomaly). Nothing about "high accruals" is a compensated risk — it's a mispricing that persists only as long as enough investors keep making the same mistake and not enough capital arbitrages it away.

Worked example. Sloan (1996) found firms in the lowest accruals decile outperformed the highest accruals decile by roughly 10% annually in the original sample period. After the paper's publication and wide dissemination through the 2000s, follow-up studies found the long-short spread had shrunk by more than half, consistent with quant funds systematically shorting high-accrual names once the signal became well known — exactly what should happen to a friction-based anomaly once smart money can trade against it, and exactly what should not happen to a genuine risk premium, since selling insurance on distress risk doesn't stop being risky just because more people do it.

What this means in practice

Allocators size positions differently for the two. A fund treats a risk premium as something to hold through drawdowns, because the drawdown is the risk being compensated. An anomaly gets sized down as its spread narrows and capacity fills, and desks watch published academic capacity estimates and crowding proxies (13F overlap, short-interest concentration) as early warning that the edge is thinning.