Real Earnings Management and Information Asymmetry in the Equity Market

Abstract

The literature suggests that real earnings management (REM) activities can increase adverse selection risk in capital markets. Due to their opacity and the difficulties in understanding their implications, REM strategies may increase the level of information asymmetry among investors. This paper examines the association between earnings management through real activities manipulation and information asymmetry in the equity market. To estimate the level of adverse selection risk we use a comprehensive index of information asymmetry measures proposed by the market microstructure literature. For a sample of Spanish listed firms, we find that firms’ strategies of increasing earnings through REM are associated with higher information asymmetry in those firms that meet last year’s earnings. Our findings are consistent with the hypothesis that earnings management through real activities manipulation garbles the market, enhances private information production, and exacerbates information asymmetry in the stock market.

Acknowledgements

We thank an anonymous reviewer and the editor, Professor Guochang Zhang, for their useful comments and suggestions, which have improved the paper greatly. We also appreciate the comments made by Thomas Gilliam, Francisco Villanueva, José A. Gonzalo Angulo, Begoña Giner, J. Samuel Baixauli, and the participants at the European Accounting Association 38th Annual Congress, the XVIII AECA Congress, the XI Workshop on Empirical Research in Financial Accounting, and the XXIV Finance Forum.

ORCID

José Yagüe http://orcid.org/0000-0002-3589-5022

Notes

1 In their review of REM literature, Xu et al. (2007) consider a wider definition of REM strategies by including financing transactions. Financing activities include stock repurchases, use of stock options in compensation packages, use of financial instruments, and structuring financing transactions.

2 Bushee (1998) hypothesizes that the monitor role of institutional investors could affect managerial incentives to manipulate R&D to meet earnings targets. In this study, we do not analyze this aspect, as we only seek to highlight that sophisticated or informed investors, unlike individual investors, are concerned about real activities manipulation and its firm's value implications.

3 Zhang (2001) theoretically examines incentives behind public disclosure by the firm and trading by informed investors, the interaction between both two forms of information dissemination, and their consequences on the extent of information asymmetry among traders. Assuming that the amount of private information production by informed traders (public disclosure by the firm) increases (reduces) information asymmetry, Zhang's model derives an equilibrium in which the amount of private information production, the level of disclosure, and information asymmetry are all linked to specific characteristics of the firm.

4 Income increasing real earnings management does not always affect cash flows and earnings in the same direction (Roychowdhury, 2006) because, whereas price discount and overproduction have a negative effect on cash flows, cutting discretionary expenses has a positive effect. Although this has led some studies to disregard abnormal cash flows in REM measures, and thus focus only on abnormal production costs and abnormal discretionary expenses, other authors include abnormal cash flows in order to take into account the possibility of sales manipulation.

5 Note that ACFO and ADISPEXP are the residuals of models (4) and (6) multiplied by (−1), so these are the values we add to APROD in REM1, REM2 and REM3.

6 A potential concern about the use of ASY as proxy for information asymmetry for our sample is that the PCA is sensitive to sample size. To check the robustness of the index, we evaluate the performance of the PCA by applying computer-based resampling (bootstrap) techniques. Thus, we draw a large number of samples (1000, 5000, and 10,000) of different sizes –smaller than (234 observations), equal to (468), and larger than (1000) our sample size. We perform PCA analysis to all the samples and compute confidence intervals (basic percentile) at the 1% level. We observe that our full-sample estimations for all relevant parameters (the eigenvalues and the component weights for the first factor) are always included in the bootstrap intervals.

7 The results do not change (the coefficient on discretionary accruals is not significant) if we include the absolute value of discretionary accruals, as in Kim et al. (2012).