Conditional Conservatism in Accounting: New Measure and Tests of Determinants

Following Basu’s (1995, 1997) seminal work, accounting literature adopted the Basu coefficient to measure conditional conservatism (among others, Ball et al. 2003; Ball et al. 2000; Ball et al. 2005; Ball and Shivakumar 2005; Lobo and Zhou 2006; Chandra et al. 2004). However, Basu’s choice of proxy for measuring the arrival of good/bad news, stock returns, introduces inaccuracy in the measure of conditional conservatism (Dietrich et al. 2007; Roychowdhury and Watts 2007; Givoly et al. 2007). To address the problem, I introduce a new measure of conditional conservatism, which results from a Least Absolute Deviation (LAD) piecewise regression and adopts the number of changes in financial analysts’ EPS forecasts as a proxy for good/bad news about future earnings and extends the analysis to two-year and three-year time horizons. I use this new measure to test three determinants that prior literature suggested to explain the presence of accounting conservatism. Results show that companies with (1) high debt-to-assets ratio – closer to default on their debt covenants, with large portion of executives’ compensation tied to the firm’s performance, and in the year prior to a going concern opinion from their auditors report aggressively, recognizing future good news in annual earnings more quickly than bad news