The information content of earnings forecasts in Japan

Abstract

In Japan, four-fifths of listed firms have fiscal years ending March 31, and nine-tenths forecast earnings for the following year. We assess the information content of reported earnings (RE), management forecasts for the following fiscal year (MF), and security analysts' forecasts made in June and September of the ensuing year (AF). We use standard event-methodology to assess information content. Using the market model, we estimate alpha and beta for each stock from returns for the 48 months ending 12 months before the release of RE and MF, and 13 and 16 months prior to the release of the June and September AF. We restrict our sample to firms listed on the first section of the Tokyo Stock Exchange. that have March 31 fiscal years. We chose one-third of such firms at random, for a final sample of 161 firms. Managerial forecasts are superior to a naive, "no-change" forecast. The Wilcoxon sign rank test of absolute errors shows that MF is more accurate at the one per cent significance level. Analysts' forecasts in June are similar to MF; a slight apparent improvement is nonsignificant. This result is not surprising, as the time between MF (generally in late May) and June is small; 70 per cent of AF in June are identical to MF. After three additional months, however, AF in September are significantly more accurate than earlier forecasts.