The accounting profession requires that firms disaggregate net income
into specific components. Despite the widespread assumption that earni
ngs disaggregation is important for assessing firm profitability, ther
e is little empirical evidence that the classification scheme actually
improves profitability forecasts, We analyze accuracy improvements in
out-of-sample forecasts of one-year ahead return-on-equity (ROE) to e
xamine the predictive content of earnings disaggregations, The results
demonstrate that the classification scheme prescribed by the accounti
ng profession does increase the predictive content of reported earning
s. We find forecasting improvements from earnings disaggregation. Thes
e improvements go beyond separating extraordinary items and discontinu
ed operations from the other components of earnings. Further disaggreg
ation of earnings (into operating earnings, non-operating earnings and
taxes, and special items) improves forecasts of ROE one year ahead.