Discussion of “the evolution of 10-K textual disclosure: Evidence from Latent Dirichlet Allocation”

Dyer et al. (2017) examines trends in 10-K textual disclosure from 1996-2013. Using a set of variables from prior textual research, they show a decrease in characteristics generally considered desirable and an increase in attributes considered undesirable. The trend is driven by disclosures to comply with new regulatory reporting standards. The study takes a high-level approach allowing an overview perspective. However, that approach also means readers should take caution in reaching strong conclusions on any one finding. In this discussion I point out a few areas where I believe readers should be particularly aware of limitations this approach imposes.