Several recent studies have used U.S. analysts' forecasts to test for underreaction or overreaction to information in earnings announcements. These tests have provided mixed results. Evidence in Mendenhall (1991) is that analysts underreact. By contrast. results in De Bondt and Thaler (1990) showoverreactionby U.S. financial analysts to earnings announcements. The current study contributes to this topic by examining over/underreaction by Japanese financial analysts. Test results show that Japanese analysts do not overreact to earnings announcements, market to book ratios and sales growth. Instead. there is strong evidence that Japanese analysts underreact to earnings announcements and that their underreaction is more pronounced for firms with mostly permanent earnings. Our results also show that Japanese analysts display larger forecast biases for earnings reported under U.S. GAAP as opposed to Japanese GAAP. Finally, we find that US. analysts discount information in earnings announcements to a larger degree (relying to a greater extent on information in past prices instead) when compared to their Japanese counterparts. Further, in contrast to their Japanese counterparts, these analysts display no optimistic bias. The results above suggest that the impact of each country's unique culture and capital norms will have to be taken into account by policy makers in evaluating the feasibility of harmonization of accounting standards.