On Empirically Capturing 'Other Information'

In this study, we examine the performance of two different measures of 'other information' found in the accounting literature. One is based upon last year's valuation error, once the impact of accounting variables on value is controlled for. The other is based upon using one-year-ahead consensus analyst forecasts. We consider both measures from a theoretical perspective and find that they suffer from different problems which make it difficult to rank them in the absence of empirical evidence. We adopt the criterion of addition to explanatory power to evaluate the measures empirically. The two measures are also likely to impose (possibly different) sample selection biases, as well, in terms of both the numbers of firm-years lost and the characteristics of the observations data, because of their different requirements for additional data.In terms of observations lost and contribution to explanatory power, the use of last year's valuation error is far superior to the use of the one-year-ahead valuation error. Further, our results suggest that the two measures impose sample selection and the selection biases are different between the two measures. Nonetheless, the measures do appear to contain complementary information such that they can both contribute to the explanation of market value in tandem. To do so, however, requires losing a majority of the available observations. Whether the loss of observations is compensated for by the additional statistical power provided by incorporating more significant explanatory variables, combined with any attendant sample selection bias, can only be decided in a particular context