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

• In Event: Analyst Forecasts

Abstract

This study re-examines the inference that investors find non-GAAP earnings to be more informative than GAAP earnings. Due to the unavailability of a reliable GAAP earnings expectation, prior researchers suggest that measurement error might incorrectly induce the conclusion that investors prefer non-GAAP earnings. As a result, they call for additional research to examine this potential explanation for prior evidence. Using recently available GAAP earnings forecasts in I/B/E/S, we find that measurement error significantly biases the GAAP earnings surprise, since 55% of the GAAP forecast error in prior studies was correctly forecasted by analysts, but incorrectly classified as a surprise. Once we abstract from this measurement error concern, we find strong evidence that non-GAAP earnings are more informative to investors than GAAP earnings. We also re-estimate our analyses in settings where investors’ preference for GAAP versus non-GAAP earnings has been inconclusive in prior research. The results illustrate how these settings are particularly influenced by measurement error and how correcting for measurement error results in consistent evidence regarding investors’ preference for non-GAAP earnings. Finally, although investors find non-GAAP earnings to be more informative relative to aggregate GAAP earnings in assessing firm performance, our evidence indicates that investors find incremental information content in non-GAAP exclusions relative to non-GAAP earnings. Thus, contrary to inferences that managers and analysts only exclude value-irrelevant items, we find that investors find earnings exclusions to be informative. Overall, our results provide evidence that prior studies correctly raise concerns regarding the effect of measurement error on GAAP earnings surprises. However, we also provide evidence that inferences regarding investors’ preference for non-GAAP earnings are not an artifact of measurement error.

Authors

• Theodore E Christensen, Brigham Young University

• Kurt H Gee, Stanford University

• Benjamin C Whipple, The University of Georgia