Empirical Evidence on the Effects of Measurement Error in Non-GAAP Earnings Research

Abstract

This study examines how measurement error in earnings expectations affects prior evidence regarding (1) investors’ preferences for GAAP versus non-GAAP earnings and (2) the quality of non-GAAP reporting in meet-or-beat settings. Prior research on non-GAAP reporting computes earnings surprises for both GAAP and non-GAAP earnings relative to analysts’ non-GAAP earnings forecasts. As a result, GAAP earnings surprises are subject to measurement error due to the use of a misaligned earnings expectation. Many studies highlight this measurement error problem and caution that evidence of an investor preference for non-GAAP earnings might simply be due to the mechanical error in GAAP surprises, which creates statistical bias in favor of non-GAAP earnings. Utilizing newly available GAAP forecasts, we find that the traditional GAAP earnings surprise is comprised of 60% measurement error, on average. Nevertheless, we find that the impact of this form of measurement error on inferences regarding investors’ preferred earnings metric is small, and provide evidence that the reason the impact is small is that the components measured with error have low persistence. Next, we examine how measurement error influences inferences regarding the use of non-GAAP exclusions to meet-or-beat analysts’ forecasts. Contrary to prior evidence, after we correct for measurement error, we find that the non-GAAP disclosures of benchmark-beating firms are of lower quality than those of other non-GAAP reporters.