Government Control, Regulatory Enforcement Actions, and the Cost of Equity

Abstract

Using a comprehensive manually collected dataset of regulatory enforcement actions against fraud in the Chinese capital market and a difference-in-differences (DID) research design, we examine the impact of such actions on the implied cost of equity and the role of the government as the controlling shareholder in moderating this relationship. We find that regulatory enforcement actions increase firms’ cost of equity, and that government controlling shareholders can mitigate the effect of these actions. Our results are robust to various sensitivity tests, including alternative measures of the cost of equity, alternative samples, additional control variables, and an alternative DID design. Additional analysis provides supporting evidence that the effect of enforcement actions on the cost of equity arises from investors’ perception of higher long-run information risk in the case of fraud firms. Further, government controlling shareholders can mitigate the impact of such actions on the cost of equity by lowering investors’ perceived information risk.

Keywords:

• Ownership
• Enforcement
• Cost of equity
• Fraud
• China

JEL codes:

• G32
• G34
• G38
• M41

Acknowledgements

The authors thank Yuan Ding (editor) and two anonymous reviewers for their helpful comments. We also thank Greg Shailer, Sonali Walpola, Mark Wilson and conference participants at the 2016 The International Journal of Accounting Symposium for their helpful comments on earlier versions of this paper. Kun Tracy Wang acknowledges financial support from the Australian National University’s College of Business and Economics Research School Grant (No. R62860-60D4).

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1 Examples are the 1997 Asian financial crisis, a series of high-profile global corporate and accounting scandals at the turn of the century (e.g., Enron and WorldCom in the United States [US], the Parmalat scandal in Europe, and One-Tel and HIH Insurance in Australia), and the 2008 global financial crisis.

2 Private enforcement was virtually absent in China until 2002, when the Supreme People’s Court issued the Notice Regarding Accepting Tort Cases Arising from Stock Market False Disclosure, which for the first time explicitly allowed private securities litigations. In 2005, the Securities Law was revised to recognize the legal basis for civil suits on securities misrepresentation, as established by the Court. However, only a few private securities prosecutions have been successful, given the several obstacles, including the prohibition of class action lawsuits and the local courts’ ineffectiveness in dealing with such litigation, which cause significant difficulties and delays in the pursuit of civil securities litigation (Chen et al., 2006; Huang, 2013). In addition, civil securities litigation is limited to misrepresentation and can only be filed based on sanctions imposed by government regulators, mainly the CSRC, or on criminal judgments of courts (Huang, 2013).

3 See, for example, Smith et al. (1984), Howe and Schlarbaum (1986), Feroz et al. (1991), Nourayi (1994), and Dechow et al. (1996) for the US market, and Chen et al. (2005), Yang and Xie (2008), Aggarwal et al. (2015), and Li (2016) for the Chinese market.

4 Much of the government ownership research rests on the agency and grabbing hand theorization. Studies argue that the government shareholders’ non-financial goals result in principal–principal conflicts of interests between government controlling shareholders and minority shareholders, which create inefficiencies and dysfunctions in firms and lead to their poor financial performance (e.g., Cuervo & Villalonga, 2000; Dharwadkar et al., 2000; Megginson & Netter, 2001; Shleifer & Vishny, 1994; Shleifer, 1998; Wang & Shailer, 2018).

5 According to Article 179 of the 2005 Securities Law, the main duties of the CSRC include formulating policies and regulations for the securities market; supervising the listing and trading activities of all types of securities; regulating the securities market behaviors of the listed companies, shareholders, and other market participants; supervising securities exchanges and other organizations engaged in the securities business; and investigating and penalizing activities in violation of the relevant securities laws and regulations.

6 Consistent with this view, Piotroski et al. (2015) document that Chinese politicians tend to suppress negative news from listed companies especially government-controlled firms, which they attribute to the greater costs to politicians for the release of negative news about government-controlled firms.

7 For example, Firth et al. (2012) document that SOEs have an advantage as defendants in court trials against other parties when they are subject to enforcement actions.

8 For our final sample of 634 fraud firm-years, one-year ahead and two-year ahead analyst forecast data are available in the WIND Financial Terminal (WIND) for only 61 (9.62%).

9 For example, the regression for 1-year-ahead earnings in year t uses data from year t − 10 to t − 1, the regression for 2-year-ahead earnings in year t uses data from year t − 11 to t − 2, and the regression for 3-year-ahead earnings in year t uses data from year t − 12 to t − 3.

10 For example, if we estimate 1-year-ahead earnings in year 2003, that is, if 2003 is year t and the forecast year is 2004 (year t + 1), we use all firms with available observations from 1993 to 2002 to estimate the coefficients using Equation 2. We then multiply the independent variables of firm i in year 2003 with the coefficient estimates to obtain the earnings forecast for firm i in 2004 (year t + 1). If we estimate 2-year-ahead earnings in year 2003, that is, if 2003 is year t and the forecast year is 2005 (year t + 2), we use all firms with available observations from 1992 to 2001 to estimate the coefficients using Equation 2. We then multiply the independent variables of firm i in year 2003 with the coefficient estimates to obtain the earnings forecast for firm i in 2005 (year t + 2).

11 See Hou et al. (2012) and Li and Mohanram (2014) for details of the HVZ model and the EP model, respectively.

12 The full specification of Equation 3 includes POST, TREATMENT, and POST × TREATMENT, where TREATMENT is an indicator variable that equals 1 if a sample firm has experienced enforcement actions during the sample period, and 0 otherwise. When estimating Equation 3, TREATMENT will always equal 1 for firms that have experienced enforcement actions, and thus it is omitted from the regression because of the inclusion of firm fixed effects. POST × TREATMENT is also omitted from the regression because the interaction takes the value of 1 only for firms that have experienced enforcement actions in the post-reform periods, that is, POST and POST × TREATMENT are the same in practice.

13 In our sensitivity test, when estimating the cost of equity capital based on Chen, Chen, et al.’s (2011) realized return approach, to obtain realized EPS to proxy for earnings forecast for periods from t + 1 to t + 3, we require firms to have data available for financial years 2012–2014 to remain in the sample.

14 The search engines we used are Baidu (www.baidu.com), Google (www.google.com), and the website of Shenzhen Securities Information Co., Ltd. (www.cninfo.com.cn), which is the official website authorized by the CSRC to disclose Chinese listed companies’ announcements and reports.

15 A typical example of such enforcements is equity investors’ failure to fulfil disclosure obligations and/or comply with trading rules when their trading triggers a disclosure requirement. To avoid confounding results, we exclude those firms related to such announcements from our sample.

16 The announcements are counted as one announcement if they are made in separate documents for different violators but involve the same fraud event.

17 Studies evaluating the implied cost of capital estimates report mixed relationships between the two factors (Guay et al., 2011). Gode and Mohanram (2003) note that if the CAPM holds, then beta should be the only risk factor. Guay et al. (2011) further argue that the fact that CAPM beta is not successful in precisely estimating the cost of capital has motivated studies to estimate the implied cost of capital, and thus the relationship between beta and the implied cost of capital is not helpful in evaluating the cost of equity estimates.

18 In an untabulated test, we also examine whether fraud firms that inflated profits or fabricated assets incur a higher cost of equity than firms committing other information-related frauds do. Data on profits and assets are important information for investors’ evaluation of the return and riskiness of their investment. Moreover, the CSRC uses profit performance (including the amount of profit and profitability ratios) as an important criterion for firms’ initial public offerings, rights offerings, and maintaining stock exchange listing status (Aharony et al., 2000; Chen & Yuan, 2004), which may motivate firms to inflate profitability. We do not find the cost of equity for fraud firms that inflated profits or fabricated assets to be higher than that of firms committing other information-related frauds. This result suggests that investors perceive other information-related frauds as jeopardizing information quality as well.

19 In our subsequent trend analysis in Section 6.4, we find that fraud firms do not incur a high cost of equity in the year before the enforcement actions although their information quality has already deteriorated in that year, as documented in this analysis. Taken together, our findings suggest the importance of the CSRC’s enforcement actions in raising shareholders’ awareness of the information risk of fraud firms.

20 Although the F-tests of equality of coefficients show that the magnitude of the positive coefficient on YEAR 1 is significantly smaller than the coefficient on YEAR 0 and YEAR 2+ (p-values are 0.014 and 0.037 respectively for the F-tests), there is no significant difference between the coefficients on YEAR 0 and YEAR 2+ (p-value is 0.639 for the F-test).

21 The tests of the equity of coefficients among the three interaction terms do not show a significant difference. Further, F-tests show that there is no significant difference in the information quality between non-fraud firms and government-controlled fraud firms during the post-enforcement periods.

22 In addition, compared to non-fraud firms, government-controlled fraud firms do not have lower information quality two years before the enforcement year (β_GOV + β_{YEAR −2} + β_{YEAR −2 × GOV} = −0.013, p-value = 0.12), but there is weak evidence that they have lower information quality in the year prior to the enforcement year (β_GOV + β_{YEAR −1} + β_{YEAR −1 × GOV} = 0.012, p-value = 0.10).

23 Supporting this conjecture, we find that all of our information risk measures are positively correlated with both severe and less severe frauds measured by the severity of punishment and CSRC enforcement, and that the correlations between the information risk measures and the measures of the severity of fraud are higher for less severe frauds than for severe frauds.

24 In untabulated tests, we find similar results when measuring political connection using a series of alternative measures. Following Fan et al. (2008), we define a firm as politically connected if its CEO is politically connected, or if its CEO, deputy CEO, chairperson, or deputy chairperson is politically connected. We also follow Marquis and Qian (2014) and identify a person as having political connection if he/she is currently serving or has formerly served in the government or the military, or is/was a member of the Chinese People’s Political Consultative Conference (CPPCC) or the National People’s Congress (NPC). Marquis and Qian (2014) note that having former or current members of the CPPCC or the NPC may provide greater symbolic benefit to a firm, because the CPPCC is an advisory board for the Chinese government and the NPC is the only legislative body in China.

25 Our results remain similar when we use longer test windows, such as three-year [−1, 1] and five-year [−2, 2] windows.

26 Previous studies suggest that regulatory enforcement actions affect trading volume (e.g., Ferris et al., 1992), firms’ auditor choice (Brocard et al., 2018; Chen et al., 2005), and firm values, and thus distress risk (e.g., Chen et al., 2005), which in turn may affect the cost of equity (Brennan & Subrahmanyam, 1996; Fama & French, 1992, 1995). Therefore, in our main tests we do not control for these factors to avoid the ‘bad control’ problem and to capture the total effect of enforcement actions more accurately. In our robustness tests using Equations 3 and 4, we add trading volume (VOLUME) to control for share liquidity, auditor quality (AUDITOR) to control for auditor choice, and the book-to-market ratio (BMRATIO) to control for distress risk. The information on the top 10 auditors is collected from the official website of the CICPA (http://cicpa.org.cn). The definitions of these variables are provided in Appendix B.