CIRJE-J-71. Obinata, Takashi, "Permanent Earnings of Industrial Firms in Japan\Part I A Cross - Sectional Analysisi1j\", April 2002.

This paper investigates the value relevance of accounting earnings, with testing empirically the association between stock price levels and earnings of industrial firms in Japan. This research provides major five results as follows. First, we find the serious industry effect. The coefficients on earnings, the significance levels of regressions, and adjusted R2s depend whether the industry dummies are included or not in the regression. The industry effect differs according to fiscal year. This result implies that the relation between the reported earnings and the permanent earnings is different every fiscal year in each industry. Second, whether to deflate the variables included in the regression model determines whether to satisfy the basic presumption of OLS regression. The deflating method gives important effects to the empirical results. To deflate the variables by stock prices at the end of previous year is very effective for alleviating the heteroscedasticity due to size. Third, the relevance is high in the order of operating profits, ordinary income (earnings before taxes and extraordinary items) and net income. Though the difference in relevance between operating profits and ordinary income is little, the difference between ordinary income and net income is very large. This result shows the rationality of the conventional financial analysis technique for regarding ordinary income as the "normal earnings generating power" of the firm. On the other hand, it throws doubt on valuing ordinary income excessively because operating profits is more relevant indicator of permanent earnings than ordinary income. Forth, by controlling the loss (negative profits) firms, the relevance of earnings increases. However, the regression results weakly support the hypothesis that losses contain large noises, so losses are not value relevant. We find the years when losses have the same information contents as positive profits, and years when losses have the different information contents from that of positive profits. Fifth, book value of equity has the very limited relevance. We cannot observe the more relevance of book value of equity than earnings, excluding a few exceptional years. Though the information combined earnings with book value of equity is more relevant than the sole earnings information, the combined information is less relevant than the earnings information in which losses are modified. This result suggests that, for explaining the variation of stock prices, the book value of equity information is redundant and merely earnings information is sufficient.