UNUSUAL PATTERNS IN REPORTED EARNINGS

Carslaw [1988] provides evidence that New Zealand firms may round up earnings when they are just below reference points denoted by Nx 10k. He observes more zeros and fewer nines than expected by chance in the second-from-left-most digit for a sample of reported earnings numbers. This study examines COMPUSTAT firms to determine whether reported earnings for U.S. firms follow similar patterns. It also reexamines the expectation model used by Carslaw. While U.S. earnings numbers deviate less from expectations, relative to Carslaw's sample, a number of other interesting patterns are observed. Firms reporting losses exhibit the opposite patterns (fewer zeros and more nines). Analysis of quarterly earnings data reveal similar, though considerably smaller, deviations from expected frequencies. Examination of per share earnings (EPS) suggests that rounding behavior is more prevalent in EPS numbers than it is in earnings numbers. Using an alternative model of expected frequencies, unusually high proportions of EPS numbers divisible by ten cents and five cents are observed for firms reporting profits, but no such deviations are observed for firms reporting losses. Again, quarterly EPS data exhibit the same patterns observed for annual EPS numbers. C ARSLAW [1988] documents unusual patterns in reported earnings of New Zealand firms. Specifically, there are more zeros and fewer nines than would be expected by chance, in the second-from-left-most digit. He interprets this evidence as being consistent with investors having cognitive reference points of Nx 10k. Similar to the wellknown "$1.99" pricing phenomenon, earnings of $1,900,000 may be perceived as being much lower than $2,000,000, relative to other $100,000 differences (say, between $2,100,000 and $2,000,000). Therefore, firms round-up earnings just below Nx 10k (with a nine in the secondfrom-left-most digit) and report earnings numbers that just exceed such reference points (with a zero in the second-fromleft-most digit). This pattern is observed for both owner and manager-controlled domestic firms. The only group not exhibiting this pattern are firms controlled by foreign owners, for whom foreign currency denominated earnings could potentially be important. Managers have incentives to round earnings for two general reasons; one relates to firm valuation and the other is based on the use of accounting numbers in various contracts. If reported earnings affect firm value as perceived by various I thank Sidney Browne, Bill Kinney, Safwan Masri, and two anonymous reviewers for their helpful comments, and Won Choi for his research assistance. Jacob K. Thomas, Columbia University. Manuscript received May 1988. Revisions received December 1988 and May 1989. Accepted June 1989.