In this paper we analyze the links between the informational relevance of earnings components for valuation and for forecasting abnormal earnings. We show that forecasting irrelevance implies valuation irrelevance. However, except in the case where aggregate earnings is the sufficient earnings construct for valuation and forecasting, valuation irrelevance does not imply forecasting irrelevance. Additionally, we show that to be irrelevant for valuation and forecasting, an earnings component need not be unpredictable. Our analysis has implications for the design and interpretation of empirical tests of informational relevance. It also has potential implications for practical investment analysis. Generally, value estimates based on aggregate earnings are expected to be less precise than value estimates based on knowledge of earnings components and their relation to other accounting items.