Differential Interpretation of Public Information: Estimation and Inference

We propose a new measure of differential interpretation in the context of a Bayesian learning model, which allows us to abstract from other sources of disagreement, such as differences in priors. We then develop a likelihood ratio statistic for testing the null hypothesis that agents interpret public information identically. Using financial analysts’ earnings forecasts, we find evidence that there is significant heterogeneity in the interpretation of public information among investors. In addition, we validate our new measure of differential interpretation and demonstrate its superiority over other proxies, such as Kandel and Pearson’s (1995) and Garfinkel’s (2009) metrics. Finally, we find that differential interpretation increases firm cost of capital, which has important implications to regulators, managers and academics.