Executive Summary: We develop a novel measure of financial statement similarity using vector-based cosine similarity.
Abstract: Firms with similar businesses are expected to display similar relations between their financial statement items. We examine whether a firm is more likely to draw SEC scrutiny when its financial statement relations deviate from its industry peers or from its past. We propose measures for the similarity (or dissimilarity) of a firm’s financial statement relations to its peers in the same industry-year or its own previous year and further decompose peer-to-peer similarity into the predicted component, which is justifiable by a firm’s business model, economic performance, and accounting policies, and the residual component. We find that firms with greater peer-to-peer dissimilarity and year-over-year dissimilarity are more likely to receive SEC comment letters, especially on financial statement issues. The findings about peer-to-peer similarity are largely driven by its residual and therefore unjustifiable component. Further, we separate the sample period before the SEC launched a data analytics peer-comparison model (referred to as “RoboCop” in the media) from the period after the launch. Our findings about peer-to-peer similarity are driven by the post-RoboCop period, whereas we observe no difference in the role of year-over-year similarity before vs. after RoboCop.
Citation: Brown, S. V., G. Ma, and J. W. Tucker. 2022. Financial Statement Dissimilarity and SEC Scrutiny. Working Paper.