The Effects of Conservative Reporting on Investor Disagreement

Abstract

We examine whether the level of a firm's conditional conservatism affects investor disagreement around earnings announcement dates. Investor disagreement is relevant for its repercussions on stock market efficiency. However, the literature related to the effect of firms’ reporting policies on disagreement is scant. Prior research suggests that conservatism, by requiring higher verifiability of profits, constrains earnings overstatements and encourages more complete revelations of losses, thus improving the information environment. In this paper, we further hypothesize that these effects of conservatism enhance news credibility and decrease information asymmetry, particularly for bad news announcements. This results in a lower disagreement and improved interpretation of earnings news. We consistently find that conservatism measures are negatively associated with proxies of announcement-time investor disagreement and that this effect is stronger when the firm is reporting bad news. Additional analyses indicate that the impact of conservatism is stronger when market surprise to the announcement is greater, while it is weaker in the presence of frequent and precise voluntary disclosure that preempts the earnings announcement. Finally, we show that a higher percentage of institutional investors’ ownership and a higher level of commitment to conservatism reinforce the impact of the latter.

Acknowledgements

We are grateful to Beatriz Garcia Osma (Associate Editor) and two anonymous referees for their constructive comments and suggestions. Special thanks to Stephen Ryan and Barath Sarath for their advice. We wish also to thank Ari Yezegel, William Rees, Peter Pope, Marina Niessner and seminar participants at Georgia State University, The Hebrew University of Jerusalem, Bocconi University, IE Business School, the University of Utah, Bentley University, the University of California Los Angeles, New York University, the 2012 European Accounting Association Doctoral Colloquium and Annual Congress, the 2012 LBS Transatlantic Doctoral Conference and the 2012 Rookie Recruiting and Research Camp at Miami for their helpful comments.

Notes

1 For theoretical developments on the link between trading activity and disagreement, see, for instance, Harris and Raviv (1993), Holthausen and Verrecchia (1990), Karpoff (1986), Kim and Verrecchia (1991, 1994, 1997), and Varian (1985). Empirical support for the relation between volume and disagreement can be found, among others, in the studies of Bamber and Cheon (1995), Bamber et al. (1997), Garfinkel (2009), and Kandel and Pearson (1995). In particular, Garfinkel (2009) suggests that volume-based measures are the best proxies available to empirical researchers and the only reliable proxy around earnings announcements. Garfinkel (2009) also claims that other measures adopted in the past, especially forecast dispersion, suffer from various biases and shortcomings, which impair their effectiveness in capturing investor disagreement, particularly around earnings events.

2 To further test this association in a multivariate setting, we follow other studies on the consequences of conservatism (Francis & Martin, 2010; Kim & Zhang, 2015; LaFond & Watts, 2008; Ramalingegowda & Yu, 2012). We create a variable that is equal to the quintile rank of the disagreement proxies (qDIS) and add it, together with the control variables, into the Basu model. The results are consistent with those reported above.

3 Additionally, we test whether the magnitude of this association depends on the news’ sign by interacting CONS with a good news dummy in Equation (3). Good news released by conservative firms can be argued to be more informative, since the degree of verification required for its recognition is very high. In this case, the price adjustment when good news is reported may be larger. However, the lower disagreement about bad news reported in the main tests may speed up the price incorporation of information for negative signals relative to positive ones. The results (not tabulated) show that the coefficient of CONS*GOOD is not significantly different from zero (t-statistic −0.24). It could be that the two effects balance each other out or that they are too weak to be detected in our sample. Since a more thorough analysis of the price effect of conservatism is beyond the scope of our paper, we leave this venue for future research.