A Michigan accounting student was able to identify a more effective way to use an old investment strategy involving accruals to yield better returns, according to a new study.
Seven years ago a student named Nader Hafzalla at the University of Michigan posed what turned out to be a fateful question to accounting professor Russell J. Lundholm after sitting through his class on financial statement analysis.
The question had to do with a venerable trading strategy of buying stocks of companies with low accruals while short-selling shares of firms with high accruals. The accruals could be non-cash accounting items such as changes in accounts receivable or accounts payable or inventory write-offs.
Hafzalla's question was why the strategy involved dividing the amount of accruals by the amount of assets. Why not divide accruals by earnings rather than by assets?
Lundholm was not impressed by the question. "I told him that the basis for the strategy's success was that accruals were less permanent than cash and that it didn't matter whether you scaled them by earnings or assets," he recalled.
But Hafzalla, a Ph.D. candidate, was not convinced, and a few weeks later he came back with some surprising news: scaling accruals by earnings, he reported, produced a group of low-accrual firms that was vastly different from the sample obtained the traditional way.
What added to the excitement was that the low-accrual firms were now, on average, substantially larger than the firms grouped in the traditional way.
“The small size of low-accrual firms, as traditionally scaled, had been tagged as a weakness of the model for a whole number of reasons, including low liquidity, high transaction costs, and high arbitrage risk,” Lundholm said. “But with earnings instead of assets as the denominator, firms classed as low-accrual were no longer so small. And so we just went exploring.”
The results of that exploring are published in the current issue of the American Accounting Association journal The Accounting Review, and they turn out to be dramatic. Employing corporate data spanning 19 years, the authors—Lundholm, hafzalla (now deceased), and Matt Van Winkle of Voyant Advisers, LLC of San Diego—compare the results computed via the traditional method and via the new method for both operating accruals and total accruals.
For both operating and total accruals the new method yields significantly better returns, with the sharpest difference being seen for operating accruals (net income minus cash from operations); there, the traditional model yields an annual return that is about 6.5 percent above than that of a portfolio of similarly-sized firms, and the new model produces an abnormal annual return about 11.7 percent above that of similarly sized firms.
In other words, the new method, which, Lundholm said, can be implemented by individual investors as well as firms, outperforms the traditional strategy by about 80 percent, even as both models substantially outperform the market. Equally impressive, returns from the new model exceed those of similarly sized firms by 10 percent or more in 15 of 19 years, compared to only four of 19 years for the traditional model.
The results, Lundholm said, seem already to be causing a ripple in the investing world—or at least that is the impression he has on the basis of presentations he has made when invited by investment conferences. "As to whether or not they're using it, that's a closely guarded secret. All I can say is that they sounded intrigued."
The findings are based on data of 81,526 firm years, with financial firms excluded, from the period 1989-2008, a sample the authors characterize as "representative of the nonfinancial market as a whole for this time period." Companies were organized into 10 portfolios on the basis of annual accruals divided by annual earnings (new method) or annual accruals divided by firm assets (traditional method). Low-accrual firms were those in the lowest decile in the previous year and high-accrual firms were those in the highest decile.
The trading strategy for both the traditional and new method was the same: buy and hold for one year shares of firms in the lowest decile, sell short borrowed shares of firms in the highest decile and buy back a year later.
What do companies in the lowest and highest decile, as defined by the new method, look like? In the words of the study, "the typical firm in the [lowest decile] has large positive cash from operations, but then accrues income back down to something much closer to zero." In contrast, "a typical firm [in the highest decile] has large negative cash from operations, but accrues income up to something near zero."
“In low-decile firms, this year's earnings may be close to zero, but a large part of the reason is a big negative-accrual component. Next year comes along, and the cash-flow piece of earnings (which was positive) remains, because it's highly persistent, but meanwhile the accrual part trends toward zero, resulting in an unexpected increase in earnings,” said Lundholm.
“As for firms in the high-accrual group,” he added, “they tend to have big negative cash from operations and big positive accruals. Think of some firm that's being pumped up illegitimately; it may have a lot of inventory that should be written off but isn’t, so basically management is delaying bad news. Or management may not write off enough for bad-debt expense, so that receivables grow artificially. The explanation for these kinds of maneuvers may be evil management, but that doesn't always have to be the case; it could be that management made honest estimates that turn out to be overly optimistic.”
According to the study, going long on the low-accrual firms produces an average annual size-adjusted return of about 5.5 percent and going short on the high-accrual group produces an average annual size-adjusted return of about 6.2 percent, the two combining for a total abnormal return of about 11.7 percent.
In sum, “We show that these extreme combinations of cash from operations and accruals are exactly the combinations that produce the most extreme differences between a sophisticated income forecast—one that distinguishes between flows and accruals—and a naive forecast.”
All too many investors, the study suggests, fail to make that important distinction, being afflicted by a “fixation on earnings.” As Lundholm put it, “Open any newspaper or business magazine, and you're not likely to read about accruals or even cash flows but about earnings. If people are even a little bit unsophisticated about the fact that all earnings aren't created equal, then that will be enough of a wedge to constitute a market inefficiency.”
A major challenge of the study was to demonstrate that problems which skeptics have found with the traditional method, whether valid or not, are not problems with the new method, an undertaking that occupies the better part of the paper. "Since Richard G. Sloan identified the accruals anomaly in a paper in The Accounting Review in 1996, there have been at least 100 papers questioning the method or tweaking the method or attacking the whole approach,” said Lundholm. “There are a lot of skeptics to convince."
But why, finally, should changing the denominator of a variable, key though that variable may be, yield such a dramatic improvement in results? Lundholm admits to not having a definitive answer. He surmised that the answer rests in the intuition Nader Hafzalla had seven years ago— namely, that the way of scaling accruals should match the basic reality that accruals are a component of earnings.
Hafzalla died in 2006 while in the fourth year of his Ph.D. program, and the paper is dedicated to his memory.
The study, entitled "Percent Accruals," appears in the January/February issue of The Accounting Review, published six times a year by the American Accounting Association, a worldwide organization devoted to excellence in accounting education, research and practice.
