CIRJE-J-82. Obinata, Takashi, "Permanent Earnings of Industrial Firms in Japan\Part II Analysis by Industry (1)\", September 2002.

This paper investigates, by industry and by period, the value relevance of earnings of industrial firms in Japan. The richness of its relevance is determined by the closeness of relation between expected permanent earnings of a firm and reported earnings, i.e. the persistence of earnings. The persistence of earnings is also affected by such a competitive condition as the endurance of firmfs competitive advantage in the industry. Therefore, the industry effect in an economy (country or region), which is characterized by the composition of industry and the environments in each industry, may affect the value relevance of earnings in that economy. In addition, the industry effect is not necessarily fixed for a long time. In order to controlling the changeable nature of industry effect, this research partitions sample years into three periods. In this way, this paper directly examines the changes in the relevance of earnings by period in each industry. For investigating the factor affecting the relevance of earnings and book value of equity, we apply some meta analysis to the estimated results by OLS. Major evidence and results provided by this research are as follows.

First, whether earnings is value-relevant differs much across industries and among periods. This shows the existence of industry effect. Second, when the factor indigenous to losses (negative earnings) is separated by introducing a dummy variable of loss firms, earnings is consistently value-relevant during all sample periods. It is not observed that the bubble economy or the following recession eliminates the relevance of earnings. Third, the effect by the factor indigenous to losses differs across industries and among periods. The larger the ratio of loss firms in sample is, the larger that effect becomes. This implies that investors donft evaluate loss of a firm so strictly when the industry that it belongs to is under depression. Fourth, we cannot find the fact that the relevance of book value of equity has been recently increasing. Contrarily, the number of industry, where the book value of equity is value-relevant, has been decreasing. Fifth, in the multiple regression including earnings and book value of equity, it is more likely that the coefficient on book value of equity is significant when the factor indigenous to losses significantly affects the relevance of earnings. This is evidence that information on book value of equity complements earnings information by controlling the factor contained losses.