Accounting Changes and Earnings Predictability.

Abstract Prior studies of the adoption of LIFO, SFAS No. 34, and APBO No. 18 document a significant positive association between security analysts' forecast errors and the current year earnings effect of changes in accounting method. Examination of a sample of largely unstudied and diverse accounting changes, using different sources of forecasts, allows evaluation of the pervasiveness and robustness of these earlier findings. Prior work is extended by considering a variety of voluntary and mandatory changes, the extent of prior disclosure of information regarding the change, the nature of forecast revisions during the year of the change, and by comparing the bias and dispersion of forecasts in change years to that in non-change years. Tests using a firm as its own control in a matched-pairs design control for industry and firm-specific factors. Consistent with prior work, the results of this study suggest that analysts do not fully revise their forecasts for the current year's earnings effect of changes in accounting method. A positive, but generally insignificant, association between forecast errors and the earnings effect of changes is reported. Both forecast errors and the dispersion of forecasts are greater in the year of an accounting change than in non-change years, particularly in the absence of prior information regarding the change. A significant negative association is reported between the revision in analysts' forecasts and the impact of an accounting change on income. This observed relation is consistent with managers adopting accounting changes with an income smoothing motivation.