At today’s office hours, Bob Lipe presented this working paper with Lail and Yi.  The upshot of the paper is that we need to be very careful when we think about the standard-setting implications of the accrual anomaly.  It is tempting to say that accrual standards must be a problem because (as the Richardson, Sloan, Soliman & Tuna paper LLY cite argues) investors react to accruals too much, leading to later reversals, and do so more for less reliable accruals.  But this paper points out that RSST include earnings and accruals as independent variables, which means that accruals are actually counted twice (since they are included in earnings).  This makes it hard to interpret the significant reversal in future returns associated with the accrual terms.  LLY include cash flows and accruals, so it is much easier to interpret the meaning of the coefficient on the accrual items–which show no reversals at all.

I think the paper has an important lesson for empirical researchers.  Writing a paper to address standard-setting issues is quite different from writing a paper that simply seeks to show market inefficiencies.  If I am a money manager (like, say, Sloan, Soliman and Richardson have all been at some point), it is enough to show that there is a robust inefficiency that can serve as the foundation of a profitable trading strategy.  But if I am a standard setter (and remember, Bob Lipe worked at FASB last year), I need to know exactly why the market is behaving inefficiently, and which particular bits of information are at issue.  LLY conclude that the so-called accrual anomaly is actually more of a ‘cash flow’ anomaly.  If all you want to do is make money, I am not sure this really matters–you can still sort on accruals and show returns.  But if you want to justify a claim like ‘standard setters should move closer to cash flow accounting,’ you need a more careful research design that clearly pins the blame for the anomaly on the accruals.  LLY implement the design and conclude that accruals are not to blame.