Discussion—The Effect of Risk on Price Responses to Unexpected Earnings

Abstract

Chambers, Freeman and Koch test to see whether earnings response coefficients (ERCs), or stock returns per unit of unexpected earnings, are increasing in total risk. They generate this prediction from a model in which the sensitivity of dividend expectations to earnings news increases in total risk. They document a strong positive relation between ERCs and total risk, after controlling for variation in systematic risk and persistence. They do not find the negative relation between ERCs and systematic risk that prior ERC research had predicted and documented.I examine in turn the two central questions that Chambers, Freeman and Koch pose: (a) Which kind of risk do ERCs reflect: systematic, idiosyncratic or both (total)? and (b) Is the relation between ERCs and risk positive, zero or negative? I argue that accounting researchers’ inability to find a relation between ERCs and systematic risk is no surprise, given that modern finance research cannot document a relation between systematic risk and stock returns. I point out recent empirical work suggesting that earnings news has both numerator and denominator effects. I argue that the many research design differences between the current paper and previous ERC papers make it difficult to evaluate how much measurement error or model misspecification contributes to the different findings. I review recent empirical and theoretical finance research on idiosyncratic risk to place the paper’s findings in a broader context. Finally, I suggest avenues for further research.