Abstract
Ball and Shivakumar [(2006), The role of accruals in asymmetrically timely gain and loss recognition. Journal of Accounting Research, 44, 207–242] show that the observed smoothness of earnings (i.e. negative contemporaneous correlation between accruals and cash flows) is the joint product of the role of accruals in smoothing out transitory fluctuations in operating cash flows (noise reduction role) and the role of accruals in providing timely recognition of economic gains and losses (contracting role). These two roles of accruals have opposite effects on earnings smoothness properties. Using a regression framework that allows us to simultaneously consider both roles, we show that failing to control for changes in timely gain and loss recognition as firms shift to IFRS can lead to erroneous inferences regarding the effects of IFRS adoption on earnings smoothness, and consequently on researcher’ conclusions about how IFRS adoption has affected accounting quality. Our results are consistent with mandatory (2005) IFRS adoption resulting in a change in the contracting role rather than the noise reduction role (or smoothness role) of accruals. A decrease in timely loss recognition, an increase in timely gain recognition, and a net decrease in asymmetric timely loss recognition are what drives the change in observed smoothness properties of earnings around mandatory IFRS adoption.
Acknowledgements
We would like to thank Daniel Beneish (editor) and two anonymous referees for the help in the revision process. We are especially grateful for the help received from Thomas Jeanjean. We would like to thank the participants at seminars at HEC Paris and University of Iowa for their feedback. Vedran Capkun is a member of the CNRS unit GREGHEC, UMR 2959.
Notes
1 Some studies view earnings smoothness (i.e., higher negative correlation between accruals and cash flows) as improving the quality of earnings (Dechow, Citation1994; Tucker & Zarowin, Citation2006).
2 IFRS standards that rely in some way on fair value measurements include: IAS 16 (Property, Plant and Equipment), IAS 19 (Employee Benefits), IAS 28 (Investments in Associates), IAS 32 (Financial Instruments: Presentation), IAS 34 (Interim Financial Reporting), IAS 36 (Impairment of Assets), IAS 37 (Provisions, Contingent Liabilities and Contingent Assets), IAS 38 (Intangible Assets), IAS 39 (Financial Instruments), IAS 40 (Investment Property), IAS 41 (Agriculture), IFRS 1 (First-time Adoption of International Financial Reporting Standards), IFRS 2 (share-based Payment), IFRS 3 (Business Combinations), IFRS 4 (Insurance Contracts), IFRS 5 (Non-current Assets Held for Sale and Discontinued Operations), IFRS 7 (Financial Instruments: Disclosure), IFRS 9 (Financial Instruments: Classification and Measurement), and IFRS 13 (Fair Value Measurement). We acknowledge that several of the IAS and IFRS standards listed here will report fair value gains and losses as part of other comprehensive income rather than conventional bottom-line earnings. However, for a number the standards listed, the fair values measurement effects will end up in bottom-line earnings. Therefore, in the broad cross section of firms that comprise our sample, we maintain that changes in fair value are more likely to impact earnings after mandatory IFRS adoption than was the case under domestic GAAP standards.
3 One might argue that greater use of fair value accounting may also increase timely loss recognition. But this is less clear to us given that domestic GAAP rules gave rise to asymmetric timely loss recognition under historical cost measurements. The results we report below in section V demonstrate that timely loss recognition, in fact, decreased rather than increased following IFRS adoption.
4 Using bottom-line net income instead does not change our results qualitatively.
5 We repeat all the tests of the Basu specification using total accruals at the dependent variable (see Collins, Hribar, and Tian Citation2014) and our results are qualitatively similar.
6 All the regression models standard errors are Huber–White–Sandwich robust standard errors. Given the number of years, industries, and countries in our sample, we do not cluster our standard errors by two categories (for a discussion of pros and cons of double clustering with small number of clusters see Petersen (Citation2009), Thompson (Citation2011), Cameron, Gelbach, and Miller (Citation2011), and Christensen, Hail, and Leuz (Citation2013)). However, clustering standard errors by year and industry, year and country, or country and industry yields qualitatively unchanged results. Similarly, introducing country and industry fixed effects, or year fixed effects, in our regression models does not qualitatively change our results.
7 Including all observations starting with 1990 does not change our results qualitatively, nor does including observations beyond 2008. However, as IFRS standards experienced major changes in 2001 and 2009, we do not include pre-2002 and post-2008 observations when we conduct our main tests.
8 Adoption years beyond 2005 are included in our sample for two reasons. First, in countries outside the EU/EEA (outside the European Commission jurisdiction, e.g., China or South Africa), firms were allowed to transition to IAS/IFRS after 2005. Second, in EU/EEA countries firms had to adopt IAS/IFRS in the fiscal year beginning in calendar 2005 which allowed for the possibility of firms’ first IAS/IFRS fiscal year ending in 2006.
9 All variables are winsorized at the 2.5% and 97.5% level. Results do not change qualitatively if we winsorize at the top and bottom 5% (1%).
10 We use the Compustat Global database for the data for this test as this component of accruals is not available in Datastream. The number of observations for this test (10,710) is substantially lower than the number of observations in Table (19,750). We compute ‘Other Accruals’ as (Net Income + Depreciation and Amortization – changes in WC accounts) – Cash flow from operations.