Prior research documents an anomalous negative price-earnings relation when a simple earnings capitalization model is estimated for loss-making firms. Collins et al. (1999) suggest that the model is misspecified due to the omission of book value of equity. However, results from previous studies are confusing. We try to enrich prior literature by focusing on analysts' forecasts. In particular, we assess the role of earnings and book value in valuing loss firms using several measures based on the information provided by analysts. We hypothesize that the role of accounting figures depends on whether the loss firm is supported or not by investors. According to this argument, we construct several measures of investor support based on analysts' forecasts, and then test the value relevance of accounting information depending on the degree of support. Our results confirm the usefulness of the notion of 'investor support'. For those loss firms that are expected to liquidate, we find that the inclusion of book value of equity in the model removes the negative sign on the earnings coefficient. However, for those loss firms that are expected to reverse current losses, we find that the coefficient on earnings remains negative despite the inclusion of book value. Easton and Sommers (2003) verify the existence of a scale effect when price-level regressions are run, either with undeflated data or with per share data. Pinnuck and Lillis (2007) provide evidence that firms that are reporting losses are actually exercising the abandonment option, and therefore at least for some firms that equity value is mainly derived from the liquidation value.