Revisiting Accounting Entropy: Measuring Information of Accounting Classification

July 16, 2024

Nan Li, University of Minnesota
Pierre Jinghong Liang, Carnegie Mellon University
Gaoqing Zhang, Carnegie Mellon University and University of Minnesota

In this paper, we derive theoretically, and empirically validate and apply, a new measure of accounting classification based on the entropy concept from information theory of Shannon in the 1940s and pioneered by Theil in the 1960s in the use of financial statement analysis. We show this measure has several desirable conceptual attributes. First, it is constructed internally from accounting numbers on financial statements. Second, it summarizes information across different financial statement line-items. Lastly, it is built on the premise of accounting classification. In our empirical exercise, we calculate entropy-based summary statistics using balance sheet information for a large sample of US publicly traded firms. We next validate our new measure by demonstrating its association with existing empirical proxies that indirectly measure information content about the traded firms. As an empirical application, we also use the setting of financial analyst attention allocation to demonstrate that our measure predicts analysts’ behaviors.