Why More Forward-Looking Accounting Standards Can Reduce Financial Reporting Quality

Abstract

A premise of standard setters and of much empirical research is that improving the quality of accounting standards and their implementation increases information in capital markets. This paper challenges this premise and shows that there are situations in which ‘better’, that is, more forward-looking, accounting standards reduce the information content of financial reports. The reason is that a forward-looking accounting standard affects the smoothness of reported earnings, which can conflict with the manager's smoothing incentive and her willingness to incorporate private information in the financial report. Although the manager could eliminate the effect by earnings management, it is too costly to do so. As a consequence, the capital market's ability to infer the financial and nonfinancial information in reported earnings declines. This finding should increase the awareness that an ‘improvement’ in accounting standards, without considering incentives and other information residing in firms, can adversely affect the quality of financial reporting.

Acknowledgements

We thank Wolfgang Ballwieser (discussant), Anna Boisits, Judson Caskey (discussant), Paul Fischer, Frank Gigler, Mirko Heinle, Rick Lambert, Stefan Schantl, Jeroen Suijs (associate editor), Marco Trombetta (discussant), anonymous reviewers, and participants at the meeting of the Ausschuss Unternehmensrechnung in the Verein für Socialpolitik 2012, EAA Congress 2013, Accounting Research Workshop at the University of Basel, Tel Aviv International Conference in Accounting, and workshop participants at the University of Minnesota, University of Pennsylvania, and University of Technology, Sydney, for helpful comments. A prior version of this paper was entitled ‘Accounting standards, earnings management, and earnings quality’.

Notes

1 Tucker and Zarowin (2006) document that smoothing can increase the information in market prices.

2 The expected utility functions of the manager and the investor in Stocken and Verrecchia (2004) are convex in the information, which leads to effects that differ from the present paper. Moreover, many of their analyses implicitly allow for negative investment by the investor.

3 For a discussion, see also Ewert and Wagenhofer (2011).

4 It would be possible to assume different projects, for example, projects with a growth rate that is less than the discount rate. However, the additional insights gained would be low.

5 In line with most accounting standards, we assume that the expected profit μ is recognized in the second period. Since μ carries no information to investors, this assumption is not restrictive.

6 It should be noted that varying the degree of inclusion of also affects the overall precision of the manager's private information because it is ‘real' variance in contrast to the constant accounting noise.

7 An alternative assumption is , so higher k increases both the information about and the noise in , for example, because more forward-looking information is prone to higher measurement error. We do not employ this assumption in order to separate the effects of varying information content and precision.

8 Possible reasons for the optimality of incentive systems combining market price and accounting numbers can be found in Dutta and Reichelstein (2005) and Dutta (2007).

9 Dichev, Graham, Harvey, and Rajgopal (2013) report that hitting earnings benchmarks, influencing executive compensation, and smoothing earnings are among the most frequent motives for earnings management.

10 De Jong, Mertens, van der Poel, and van Dijk (2012) report that also analysts recognize a benefit of smooth earnings, although they dislike earnings smoothing if it reduces transparency.

11 Huson, Tian, Wiedman, and Wier (2012) provide empirical evidence that compensation committees adjust the weights in the variable compensation in the final year of a CEO to counter the CEO's greater incentives for earnings-increasing manipulation.

12 The reason that the horizon effect only indirectly affects the first-period earnings management is that the bias in the second period is more extreme and unrelated to the manager's nonfinancial information, which is anticipated in the first-period earnings management decision. Increasing the tenure of managers would, therefore, not qualitatively affect our results derived for the first period.

13 In contrast to many other papers, we do not require r strictly greater 0 because the smoothing incentive creates a cost of increasing the bias too much. In other words, the cost of earnings management is not a necessary ingredient for the existence of a non-trivial equilibrium in our model. However, as we show later, our main result requires r > 0.

14 Many papers model investment decisions, which affect the cash flows directly (for a survey see Kanodia, 2006). In our model, assuming that the manager is protected by a minimum reservation utility constraint, the cost of earnings management would reduce the firm's cash flows. For simplicity, we assume that the manager incurs the cost, leaving (gross) cash flows to the firm constant.

15 Despite we exclude asymmetric information about the weights, the resulting equilibrium is not degenerate because the private nonfinancial information of the manager introduces noise into the earnings report that market participants cannot fully back out.

16 To facilitate readability, we do not index the random cash flow components if it is not confusing.

17 See, e.g. Fischer and Verrecchia (2000). Guttman, Kadan, and Kandel (2006) prove the existence of equilibria with partial pooling. Einhorn and Ziv (2012) show that such equilibria do not survive the D1 criterion of Cho and Kreps (1987).

18 In this equation, we show the market discount factor ρ separately, so that β is only driven by the market's revision of expectations.

19 This definition is similar to that, e.g. in Francis, Olsson, and Schipper (2006).

20 We exploit the property of normal distributions that the information content of a signal is independent of the realization of the signals, which avoids a signal-specific definition of FQ.

21 While Lemma 1 is a result of our model, a similar result obtains in Sankar and Subramanyam (2001). They assume a risk averse manager who maximizes the expected utility based on time-additive CARA utility functions and focus on the effects of the magnitude of reversal of earnings management over time. They find that without any requirement of bias reversal, the manager boosts current earnings in each period without incorporating any private information. But if the bias reverses more than at a certain portion, the manager smoothes earnings over periods and includes private information into the bias. Consequently, without a smoothing incentive and private information, reported earnings would not be incrementally informative over the accounting signal.

22 This expression follows from (17) by substituting .

23 This effect is economically similar, but converse, to an increase in the smoothing parameter s.