Purpose – This paper seeks to examine the predictive ability of disaggregated earnings in forecasting future profitability conditional on keiretsu affiliation in the context of Japan. Design/methodology/approach – Industry-adjusted future profitability measure is regressed on current profitability measures and on interaction terms of keiretsu affiliation and profitability measures. Out-of-sample forecasting test is conducted by regressing future profitability on current profitability measures for keiretsu-affiliated and independent firms. Findings – Disaggregated earnings improve out-of-sampling forecasting for firms not affiliated with keiretsu networks. For the keiretsu-affiliated firms industry adjustments actually diminish the out-of-sample forecasting accuracy. Research limitations/implications – The boundary between keiretsu-affiliated and independent firms is not so obvious. Originality/value – Examining industry competitiveness in Japan using accounting information is a significant contribution to the literature.