The Accuracy and Bias of Equity Values Inferred from Analysts' Earnings Forecasts

Abstract

We evaluate the extent to which unbiased and accurate estimates of equity value can be derived from three multi-period accounting-based valuation models using consensus analysts' earnings forecasts over a four-year horizon. The models are: (a) the earnings capitalization model, (b) the residual income model without a terminal value, and (c) the residual income model with a terminal value that assumes residual income will grow beyond the horizon at a constant rate determined from the expected residual income growth rate over the forecast horizon. Our analysis is based on valuation errors that are calculated by comparing estimated prices to actual prices. We find that, on average, analysts' earnings forecasts convey information about value beyond that conveyed by current earnings, book values and dividends. Each of the models that we used has valuation errors that decline monotonically as the horizon increases, implying that earnings forecasts at each horizon convey new value relevant information. We cannot find a clear advantage to using firm specific growth rates instead of a constant rate of 4% across all sample firms. In addition, only 17% of the imputed growth rates could be used in terminal value calculations. The residual income model with a terminal value shows the best performance on average, but it values more accurately only 48% of our sample firms. The earnings capitalization model and the residual income model without a terminal calculation value more accurately 18% and 13% of the sample firms, respectively. The remaining 21% of firms are more accurately valued using only reported current earnings and book values of equity. Thus, different models are appropriate for different firms. The conditions under which given models work best relate to ex-ante growth indicators such as the current book-to-market, earnings-to-price, the present value of the expected residual income over the forecast horizon, the growth rate in expected earnings, and firm size, but not to industry membership. In all models estimated prices are, on average, downward biased and inaccurate and they explain at best 70% of the variation in market prices. We examined the quality of the earnings forecasts and the quality of the GAAP earnings as two possible reasons for the biased and inaccurate results. Our tests provide evidence consistent with both of these reasons. Thus, we conclude that the poor model performance is due to information missing from the forecasts and to the practice of conservative accounting.