Inferring information frequency and quality

Abstract

We develop a microstructure model that, in contrast to previous models, allows one to estimate the frequency and quality of private information. In addition, the model produces stationary asset price and trading volume series. We find evidence that information arrives frequently within a day and that this information is of high quality. The frequent arrival of information, while in contrast to previous microstructure model estimates, accords with non-model based estimates and the related literature testing the mixture-of-distributions hypothesis. To determine if the estimates are correctly reflecting the arrival of latent information, we examine intervals around earnings announcements when private information flow is likely. We find evidence that flows are more likely prior to earnings announcements, which indicates that the estimates reflect the arrival of latent information.