The Probability and Magnitude of Information Events

Models of adverse selection risk generally assume that market makers offset expected losses to informed traders with expected gains from the uninformed. We recognize that the expected loss captures a combination of two effects: 1) the probability that some traders have private information, and 2) the likely magnitude of that information. We use a maximum-likelihood approach to separately estimate the probability and the magnitude of private information and test our procedure on a simulated data set. We then estimate the parameters for NYSE-listed stocks from 1993 through 2003, and show that our estimates can be used to predict future extreme returns. Finally, we examine the time-series and cross-sectional properties of the probability and magnitude of information and find a decline in the frequency of private information events following the introduction of Regulation FD. Our results shed light on the price discovery process and have implications for many areas of finance