Financial analysts' earnings forecasts in Japan show a steady improvement in accuracy as the year progresses. Moreover, echoing results in the United States, sell-side forecasts in Japan are overly optimistic. Although no single security house emerges clearly as the most accurate forecaster, Japanese houses produce more accurate forecasts than Western houses operating in Japan. The improved accuracy appears to result from a better gauge of company information at the end of the fiscal year and is not restricted to Japanese houses with investment banking links to the firm being forecast. Overly bullish forecasts seem to pervade both Japanese and Western houses, but if anything, Japanese houses are less optimistic than their Western counterparts. The evidence suggests a blend of informational and incentive effects. Incentives for sell-side analysts may contribute to overly optimistic forecasts and may be most pronounced for a security house that is a firm's main investment bank. In searching for accurate forecasts, using last year's forecast errors offers no significant advantages in picking a best forecaster for the current year. Moreover, a consensus average forecast does not yield significantly smaller mean forecast errors than a strategy of picking a single forecaster at random. This result suggests that forecasts in Japan contain a great deal of common, highly correlated information. Knowing more than a single forecast is still valuable because higher disagreement among analysts is a signal of increased uncertainty in forecasting. In addition, consensus forecasts provide a small forecasting advantage, especially in situations in which uncertainty is high.