Measuring the Information Content of Disclosures: The Role of Return Noise

Abstract

ABSTRACT Disclosure is of fundamental interest to accounting research. When the sign/magnitude of disclosed news is unclear, the information in disclosure events is inferred using the ratio of return volatilities during event and non-event windows (Beaver 1968). We show that return noise due to microstructure frictions and mispricing affects this ratio, and that effect is comparable to or exceeds that of information content. We use the SEC's Tick Size Pilot program to confirm the causal effect of return noise on the ratio, and to evaluate alternative ways to control for it. The most promising approach is to use the difference between, rather than the ratio of, return volatilities during event and non-event windows. We illustrate its benefits by showing how it alters prior inferences regarding time-series and cross-sectional variation in information content as well as changes in the information content of earnings announcements around the 2004 amendments to Form 8-K filings.