This is difficult work. We already have some evidence that, controlling for current
ROEs, historical accounting numbers have limited predictive power for future ROEs
(e.g., Fairfield et al. 1996; Dechow et al. 1999). First-level decompositions of ROE into
more detailed financial ratios do not appear to yield substantial gains. Further development
may require detailed industry-level anadysis (e.g.. Amir and Lev 1996) or the use
of alternative information sources (e.g., Tasker 1998). But ultimately these types of
research will advance the field