There was an article I found interesting in the WSJ yesterday. It details a study by Stanford law professor Joseph Grundfest, who was previously a board member of the SEC, and Nadya Malenko, a doctoral candidate at the Stanford Graduate School of Business. The study looked at reported EPS in nearly a half-million earnings reports over a 27 year period and specifically the first digit of EPS not reported; a tenth of a cent. Though companies don’t report EPS to the tenth of a cent, the number is easily calculated and Ms. Malenko obvously developed a software application to do just that for hundreds of thousands of earnings reports.
If earnings are not managed, earnings reports at the one-tenth of a cent level should be evenly distributed among ten possible reported amounts: zero to nine-tenths of a cent. They aren’t though. 4/10ths, 3/10ths and 2/10ths are all underrepresented in the survey while 5/10ths and amounts greater than 5/10ths are overrepresented.
The authors coined the term “quadrophobia” to desribe companies that have an aversion to reporting EPS amounts ending in .4, .3 and .2 cents per share. What causes this ailment? The authors believe it is earnings management and note that 1/10th of a cent is, on average, only $31,000 in additional quarterly after-tax income among the survey companies. When the opportunity to round EPS up to the next penny presents itself, managers find the $31,000, $62,000 or $93,000 necessary to do so a statistically significant portion of the time.
The more interesting part of the study is that quadrophobia is a leading indicator of later earnings restatements and charges of accounting violations. The WSJ article cites Dell, which never reported earnings at 4/10ths of a cent from its 1988 IPO until its restatement of earnings in 2007. The WSJ pegs the probability of 76 straight earnings reports with not a single one ending in 4/10ths of a cent occurring randomly at 1 in 2,500.
Obviously, the study is based on the idea behind the earlier studies that revealed backdating of stock option grants. Certain variables should be random unless the outcome is intentionally altered. Where outcomes that should be random are not, they are probably influenced by an outside agent that interferred with the data. In the case of stock option grants, those issuing stock options to themselves and their colleagues did so disproportionately on dates when the stock was at the bottom of a trough. In the case of EPS, companies round the number up more frequently than they round it down.
What does this study teach us? The people who calculate and audit EPS don’t believe it is accurate to the penny or 1/10th of a cent. Investors probably shouldn’t either.