Using Market Reaction to Infer Persistence of Earnings Surprises

Abstract

We measure the market's assessment of the information in a particular earnings surprise by calculating a firm- and time-specific earnings response coefficient (FTERC). We use the FTERC to infer the market's expectation of the persistence of unexpected earnings and also develop an interpretive framework. Examining the market's response to a particular earnings surprise - rather than whether it, on average, over- or underreacts - allows researchers to use the FTERC as a dependent variable (e.g. in a study of disclosure quality) or as a control when each response is unique (e.g. a firm before, during and after fraud). Our model implies seven classifications of expected persistence: growing, permanent, decaying, transitory, partially offsetting, offsetting, and subsuming. We find that approximately 1 in 4 earnings announcements results in an FTERC within the 'normal' permanent-to-transitory range; over 70% of expectation revisions are growing or subsuming.

Keywords: earnings response coefficient, persistence

JEL Classification: M41, G12