Corporate governance, violations and market reactions

Pacific-Basin Finance Journal 21 (2013) 881–898

Contents lists available at SciVerse ScienceDirect

Pacific-Basin Finance Journal journal homepage: www.elsevier.com/locate/pacfin

Corporate governance, violations and market reactions☆ Roy Kouwenberg a, b,⁎, Visit Phunnarungsi c, a a b c

Mahidol University, College of Management, Bangkok, Thailand Erasmus University Rotterdam, Erasmus School of Economics, Rotterdam, The Netherlands Assumption University of Thailand, Martin de Tours School of Management and Economics, Bangkok, Thailand

a r t i c l e

i n f o

Article history: Received 19 August 2011 Accepted 21 June 2012 Available online 30 June 2012 JEL classifications: G30 G14 Keywords: Corporate governance Violations Event study Market reaction

a b s t r a c t We test the relation between firm-level corporate governance and the market reaction to announcements of violations of rules and regulations by Thai listed firms. We find no significant difference in market reaction when firms with high and low governance scores commit violations. We do find a larger negative abnormal return when firms with low past violation records violate the rules. The market reaction is especially strong, −8.1% on average, when firms with low past violations and low governance scores commit violations. The evidence suggests that investors rely on a combination of observed behavior (violations) and the firm's formal governance policies to learn about the firm's true governance practices. © 2012 Elsevier B.V. All rights reserved.

1. Introduction In this paper we test whether the market reacts differently when firms with good and poor governance commit violations of the listing rules. In theory corporate governance should decrease the firm's discount rate and increase firm value, as governance mechanisms are designed to mitigate agency conflicts and prevent expropriation by managers and controlling shareholders. However, in practice minority shareholders typically face a severe asymmetric information problem when assessing a firm's corporate governance practices. The problem for outside investors is to determine whether the adoption of formal good governance policies is a sign of good governance practices, or window-dressing to improve a firm's external image. Violations of listing rules by firms can provide new information to the market about whether firms implement governance policies substantively or not. If governance policies are fully discounted in stock

☆ Financial support from the Capital Market Research Institute of the Stock Exchange of Thailand is gratefully acknowledged. ⁎ Corresponding author at: Mahidol University, College of Management, 69 Vipawadee Rangsit Road, 10400, Bangkok, Thailand. Tel.: +66 22062017. E-mail address: [email protected] (R. Kouwenberg). 0927-538X/$ – see front matter © 2012 Elsevier B.V. All rights reserved. doi:10.1016/j.pacfin.2012.06.006

882

R. Kouwenberg, V. Phunnarungsi / Pacific-Basin Finance Journal 21 (2013) 881–898

prices, we expect that the market reacts more negatively when firms with high governance scores violate the listing rules, compared to firms with low governance scores. Violations by a firm with a high governance score can reveal that the firm window dresses corporate governance, leading to a negative surprise and a higher discount rate, whereas expectations are lower to begin with for firms with low governance scores. Failure to detect a significant difference between market reactions to violations between firms with low and high governance scores would lead to doubts about whether governance scores provide relevant information to the market about substantive governance practices that is reflected in firm-level discount rates. We use a short-window event study of firms listed on the Thai stock exchange to uncover the relation between governance and market reactions to violations. We sort listed firms into two groups based on a score for the adoption of good governance policies. The governance policies include having a high proportion of independent directors, separation of the positions of chairman and CEO and the establishment of a remuneration committee, among others. We test whether the market reacts differently when firms with high and low governance scores violate the listing rules. As a benchmark for comparison, we also sort firms based on past violations into good firms (with a record of low past violations) and bad firms (with high past violations). If a firm's historical violation record affects the discount rate, announcements of violations by good firms should lead to a stronger marker reaction than new violations by bad firms. We find that violation announcements are bad news for investors: the average abnormal return market reaction is − 2.2% during the three-day event window around the announcement (days − 1, 0, + 1). The market reaction is especially negative when firms commit violations classified as severe: − 4.1%. Surprisingly, we find no significant difference in the market reaction to violations between firms with high and low CG scores: the average abnormal return is − 1.1% for high CG firms and − 2.7% for low CG firms, but the difference is not significant. In contrast, we do find a significant difference in market reaction between firms with low and high past violation records. The average abnormal return is − 4.4% for good firms (low past violations), while for bad firms the market reaction is − 1.3%. Closer inspection reveals that the market reaction is especially negative when firms with a good track record and low governance scores commit violations: the average abnormal is − 8.1% for good firms with low CG. In stark contrast, the average market reaction to violations is only − 0.3% for good firms with high CG. This result suggests that the market scrutinizes and discounts a good firm's governance policies once violations occur, leading to a strong negative reaction for good firms with low CG. On the other hand, for firms with a bad track record governance problems appear already discounted, as the market reaction is not different between bad high CG firms and bad low CG firms (− 1.6% vs. − 1.1%). In sum, the results suggest investors gradually learn about a firm's substantive governance practices by observing both the firm's behavior (violations) and its formal governance policies. Formal governance policies (e.g., having independent directors) may not provide sufficient information when considered in isolation, as bad firms can try to window-dress governance and not all good firms may be willing to bear the costs of improved governance policies. We test our hypothesis in an emerging market with relatively weak legal setting, Thailand. Existing studies on the relation between market reactions to violations and corporate governance are rare, and mainly conducted in the U.S., a market with strong investor protection. Carcello et al. (2011) find that the market reaction to restatements is less negative when the firm has a fully independent audit committee, but only if the CEO is less involved in the director selection process. Lisic et al. (2011) find that the market reaction to restatements is less negative for firms with at least one financial expert on the audit committee, but this effect is reduced when CEO power is high. We find that the market reaction to violations of good firms in Thailand is less negative when the positions of CEO and Chairman are separated (lower CEO power) and when the audit committee is chaired by an independent director. To the best of our knowledge, the only related study in an emerging market is a recent working paper by Cumming, Hou and Lee (2011) who find that the market reaction to fraud announcements in China is less negative for state-owned enterprises, while firm-level governance (board independence and CEO duality) has no significant impact. In contrast to our paper, Cumming, Hou and Lee (2011) do not provide explicit hypotheses or an explanation for the insignificant effect of governance on fraud announcement returns, as their study mainly focuses on the role of financial analysts in deterring fraud. Further, as China's capital market is unique due to widespread state ownership of firms and special listing requirements,

R. Kouwenberg, V. Phunnarungsi / Pacific-Basin Finance Journal 21 (2013) 881–898

883

studies in other emerging markets are valuable in our opinion. Our motivation for selecting Thailand in particular is that the legal environment and enforcement are weak, similar to other emerging markets. However, because disclosure requirements are relatively strict in Thailand and the stock market regulators are active in detecting and announcing violations, we can construct a large dataset of market reactions. Our paper is indirectly related to several studies that investigate the effect of corporate governance on firm value (see Love, 2011, for a literature review). Most studies find a positive relation, suggesting that the adoption of good governance policies helps to increase firm value. Another strain of research, for example Chen et al. (2009), focuses on the direct effect of corporate governance on the firm's cost of capital (or discount rate). All of these studies suffer from potential validity problems due to endogeneity and reverse causality. Our study compliments the existing literature by investigating the effect of corporate governance on firm value indirectly, through market reactions to violation announcements. This paper is organized as follows. Section 2 provides the theoretical framework and develops the hypotheses. Section 3 explains the methodology and the data used for the study. Section 4 provides empirical results and Section 5 concludes the paper.

2. Hypothesis development and related literature 2.1. Hypothesis development Effective corporate governance may lower the risk premium required by investors, and thereby decrease the cost of capital and increase firm value. Several studies, including Klapper and Love (2004) and Durnev and Kim (2005), investigate the direct effect of corporate governance on firm value and they find a positive relation, suggesting that the adoption of good governance policies helps to increase shareholder value. However, potential endogeneity can create a problem in these studies as relevant variables affecting both corporate governance and firm value may have been omitted. Reverse causality is another problem that may arise because firm value can also affect the adoption of corporate governance policies. Our study investigates the effect of corporate governance on firm valuation indirectly. We test whether the market reaction to violations of listing rules by firms depends on the governance policies adopted by the firm. We focus on Thai firms that violate the rules and regulations set by the Thai SEC and Stock Exchange of Thailand. Our first hypothesis is that the announcement of violations by firms is bad news for investors, a sign of potential negligence or expropriation, and hence the market reacts negatively. Hypothesis 1. The market reacts negatively to the announcements of violations of rules and regulations. In theory corporate governance policies can mitigate agency conflicts and help prevent expropriation, thereby lowering the required risk premium demanded by investors and increasing firm value. However, in practice investors face an asymmetric information problem when assessing whether a firm implements governance policies substantively or symbolically. Announcements of violations provide new information to the market about whether firms implement corporate governance policies substantively or window‐ dress governance. Especially for firms with high governance scores the occurrence of violations or fraud should create a negative surprise and lead to doubts about whether the governance policies have been truly implemented. On the other hand, for firms with low governance scores the market should anticipate a higher frequency of violations and discount the market value in advance. Hence, we expect that the market reacts more negatively when firms with high governance scores commit violations, compared to firms with low governance scores. Hypothesis 2. The market reacts more negatively to the announcement of violations of rules and regulations by firms with high corporate governance scores, compared to firms with low corporate governance scores. Another source of information that investors can exploit to assess governance practices, and the likelihood of future violations, is the firm's past violation track record. For ease of exposition, we refer to firms with more past violations as ‘bad firms’, whereas firms with less past violations are referred to as ‘good firms’. We expect that the announcement of violations committed by good firms is more surprising for investors and leads to a

884

R. Kouwenberg, V. Phunnarungsi / Pacific-Basin Finance Journal 21 (2013) 881–898

stronger market reaction, as the firm's good track record deteriorates. In contrast, for bad firms violations are in line with prior expectations. Thus, our third hypothesis is as follows: Hypothesis 3. The market reacts more negatively to the announcement of violations of rules and regulations by good firms, compared to bad firms. If investors use a firm's track record as an additional source of information about governance practices, then we expect that the market reacts especially negative for violations committed by firms with good governance and a good track record. For firms with either a bad track record or poor governance, the market may already expect future violations. However, for well-governed firms with good track records, violations are unexpected. In other words, we expect an interaction effect between good governance and having a good track record. Hypothesis 4. The market reacts more negatively to the announcement of violations of rules and regulations by good firms with high corporate governance scores, compared to other firms. Apart from testing Hypotheses 1, 2, 3 and 4 separately, we also estimate a multiple regression model with the abnormal return around the violation date as dependent variable, and using violation severity and other controls as explanatory variables.

2.2. Related literature We expect the market to react negatively to announcements of violations, as these events can damage the reputation of the firm and may also lead to direct loss of wealth through penalties, legal costs and settlements. The existing empirical literature confirms that the market reacts negatively to fraud cases (Karpoff and Lott, 1993) and that the market responds negatively to earning restatements by firms (Palmrose et al., 2004; Agrawal and Chadha, 2005). Karpoff et al. (2008) find that most of the long-term loss in firm value results from damaged firm reputation, including a higher cost of capital, and not from direct penalties. In theory corporate governance mechanisms mitigate agency problems and can reduce expropriation by managers and controlling shareholders (see, e.g., Durnev and Kim, 2005; Doidge et al., 2007). The existing empirical evidence in the literature indeed confirms that several firm‐level governance mechanisms, such as having independent directors and having financial experts in the audit committee, are associated with a lower frequency of fraud and earnings restatements (see, e.g., Beasley, 1996; Uzun et al., 2004; Agrawal and Chadha, 2005; Farber, 2005; Chen et al., 2006). Only few studies have investigated whether market reactions to fraud cases and earnings restatements depend on the firm's internal governance quality. In a study about CEO involvement in selecting board members, Carcello et al. (2011) find that the market reaction to earning restatements is less negative when the firm has a fully independent audit committee, but only if the CEO is less involved in the director selection process. In a similar study, Lisic et al. (2011) find that the market reaction to restatements is less negative for firms with at least one financial expert on the audit committee, but this effect is reduced when CEO power increases. Thesev two studies have been conducted in the U.S., a developed market with strong investor protection laws. 1 Evidence for emerging markets with a weaker institutional setting is almost non-existent, except for a recent working paper by Cumming, Hou and Lee (2011). Cumming, Hou and Lee (2011) find that the market reaction to fraud announcements is less negative for state-owned enterprises in China, while governance characteristics like board independence, CEO/Chairman duality and ownership concentration have no significant impact. As China's capital market is unique due to state ownership and special listing requirements, studies in other emerging markets of market reactions to violations are still necessary in our opinion. Further, Cumming, Hou and Lee (2011) do not provide hypotheses or explanations for the insignificant effect of governance on fraud announcement returns, as their paper focuses mainly on the role of financial analysts in deterring fraud. 1 Chhaochharia and Grinstein (2007) study the effect of the Sarbanes–Oxley Act and find positive abnormal returns for firms with lower SOX compliance around the announcement of SOX, suggesting that investors expect these firms to benefit more from mandatory imposition of stricter governance.

R. Kouwenberg, V. Phunnarungsi / Pacific-Basin Finance Journal 21 (2013) 881–898

885

As we conduct our empirical study in a developing market with a relatively inefficient legal system and poor investor protection, namely Thailand, we discuss the institutional setting and corporate governance in Thailand in the next sections, as well as the potential impact on the expected outcomes of our research. 2.3. Institutional setting Thai listed companies fall under the supervision of Stock Exchange of Thailand (SET) and the market regulator, the Thai SEC. SET supervises the dissemination of information by listed companies; it can impose fines and suspend trading when listed firms fail to submit financial reports or any information that is likely to have a significant effect on a firm's stock price. Further, SET monitors the trading of stocks for signs of insider trading and market manipulation, using computer surveillance systems.2 Apart from SET, the supervision of listed companies is the responsibility of the Thai Securities and Exchange Committee (SEC). Using a large dataset of regulatory actions by SET and SEC regarding fraud and failure of firms to disclose relevant information, covering 333 firms during the period 1990–2010, our data reveal that Thai regulators are active, with an average of 0.17 violations of rules and regulations announced per listed firm per year. Although SET and SEC are active in detecting and announcing violations, ultimate penalties in the form of jail sentences, large monetary fines and delisting are rare. In the first ten years since its establishment in May 1992, the SEC referred more than 25 criminal cases to the police, but as of May 2002 (the starting point of our sample) none of these cases had led to a conviction. The first conviction for a securities law violation eventually occurred in 2004 when a former CEO and vice-chairman of a listed firm were sentenced to five years in jail for expropriation. Two other successful cases followed in 2005 and 2007. The SEC also has the authority to directly impose fines for certain less serious offences, such as insider trading. However, the fines imposed tend to be rather small, limiting their value as a deterrent. Djankov et al. (2008) develop an index for laws regulating self-dealing transactions and the expropriation of minority shareholders by corporate insiders. Thailand has a very high score for disclosure requirements (1 out of 1, far above the sample average of 0.38), but below-average scores on liability standards. Cumming, Johan and Li (2011) construct indices for exchange trading rules in 42 countries. The Thai stock market ranks above average on rules limiting market manipulation and below average for insider trading rules. Overall, the institutional setting for the stock market in Thailand can be summarized as follows: disclosure requirements are strict and the two stock market regulators are active and regularly announce violations by listed firms, but large penalties for offenders are rare due to inefficiencies in the legal system. 2.4. Corporate governance in Thailand The Stock Exchange of Thailand introduced a corporate governance code for Thai listed firms in the year 2002. Similar to existing codes in developed countries, the code addresses the protection of rights of minority shareholders and other stakeholders, the importance of independent directors and the disclosure of potential conflicts, among other issues. The Thai corporate governance code was introduced on a ‘comply or explain’ basis. Only a small number of the recommended policies are either compulsory by law or part of the listing requirements: establishing a proxy, having an audit committee, and having at least three independent directors. In a survey of all listed firms in 2003, SET found that 98% of firms indeed followed these required policies. The voluntary adoption of the remaining non-compulsory policies in the code was not widespread in 2003. For example, only 47% of firms had a board consisting of at least one-third independent directors, as recommended by the code. The Thai corporate governance code focuses on protecting the rights of minority shareholders, as high ownership concentration is prevalent. A shareholder with a block of 25% or more of the voting rights of a Thai listed firm has the legal right to nullify any corporate decision and to inspect the operations and financial condition of the company, among others rights. Using 25% as a minimum cutoff point for control, 78.7% of the 333 listed firms in our sample in 2002 have at least one controlling shareholder. Control by families and family business groups is most common (54.0% of all listed firms). As a result of high 2 We refer to Cumming and Johan (2008) for more information on global market surveillance practices. Cumming and Johan (2008) show that cross-market surveillance is crucial for market development. Thailand has only one stock exchange organization (SET) and dual listings abroad are virtually absent.

886

R. Kouwenberg, V. Phunnarungsi / Pacific-Basin Finance Journal 21 (2013) 881–898

ownership concentration, the average freefloat of Thai-listed firms is only 37.5% and there is no active market for corporate control. Thai firms are not allowed to issue dual-class non-voting shares and the separation between cash flow rights and voting rights is small (Khanthavit et al., 2003). 3 The Thai good governance initiative seems suited for decoupling of policy and practice, as adoption of most recommended policies is voluntary, while substantive implementation of the policies cannot easily be verified by investors and is not enforced. Ananchotikul et al. (2010) find that there is a small group of Thai listed firms who appear to use governance policies as a form of “cheap talk” (43 firms in a sample of 333, or 12%). These firms have written policy declarations on good governance or business ethics, but relatively low scores for shareholder rights and board independence, and more frequent future violations of listing requirements. The potential for decoupling is one important reason why we expect the market to react more negatively to the announcement of violations by firms with high governance scores in Hypothesis 2, as violations should lead to doubts about the substantive implementation of governance policies by these firms. One potential exception to this argument is when the information asymmetry and decoupling problems are so severe that the market stops discounting information about firm-level governance policies altogether. In that case we do not expect any relation between the market reaction to violation announcements and governance scores. We leave this issue open for our empirical examination. 3. Methodology and data 3.1. Methodology We use a short-window event study to investigate the market reaction to announcements of violations of listing rules by SET and SEC. We focus on abnormal returns for days −1, 0, and + 1, separately, and the 3-day event window (days − 1, 0, + 1 combined), where day 0 is the announcement date of the violation. The abnormal return (AR) for a violation of firm j on day t is defined as
 ARjt ¼ rjt −E rjt

ð1Þ

where rjt is a continuously compounded total return for the stock of firm j and E(rjt) is the expected return for the stock of firm j based on the market model. We follow Scholes and Williams (1977) to estimate beta for the market model to account for illiquidity and thin trading in the Thai stock market. For each day in the event period, the average abnormal return (AAR) is averaged across the number of violations, j¼N

AARt ¼

∑j¼1 t ARjt

ð2Þ

Nt

where Nt is the number of violations over which AR is averaged on day t. The cumulative abnormal return for a violation of firm j over days (t1, t2) is measured as

j

CARt 1 ;t 2 ¼

t2 X

ð3Þ

ARjt

t¼t 1

The cumulative average abnormal return over days (t1, t2) is measured as j¼N

CAARt 1 ;t 2 ¼

j

∑j¼1 CARt 1 ;t 2 N

ð4Þ

where N is the number of cumulative abnormal returns. 3 Wiwattanakantang (2001) finds that the separation of voting and cash flow rights has no significant effect on the value of Thai firms.

R. Kouwenberg, V. Phunnarungsi / Pacific-Basin Finance Journal 21 (2013) 881–898

887

Table 1 Violations of SET and SEC rules, 2003–2010. This table summarizes the data on violations of rules and regulations from the Stock Exchange of Thailand (SET) and the Securities Exchange Commission (SEC). The column “Total violations” shows the number of violations committed by firms during 2003–2010, for each type of violation separately. The column “Number” shows the number of violations during 2003–2010 that can be included in the study (excluding violations of firms in the Rehabilitation sector). The column “Percentage” shows the percentage of each type of violation included (total 238 violations). The column “Source” shows the data source (SET or SEC). The classification for violation level of severity is: 1 = severe violation of corporate governance principles; 2 = medium; and 3 = minor violation. Violation

Source Level of Total Total violations Description of violation severity violations included in the study Number Percentage

Fraud Expropriation

SEC

1

2

1

0.4%

Insider trading

SEC

1

2

1

0.4%

Market manipulation

SEC

1

1

1

0.4%

Failure to disclose information Connected party SET transaction

1

2

1

0.4%

Material information

SET

1

37

31

13.0%

Incomplete information

SET

2

6

6

2.5%

3

3

1.3%

4

3

1.3%

Violation of rules related to tender offers Tender offer SET/ 1 SEC Reporting of share SEC 1 holdings Takeover information

SET

2

8

8

3.4%

Financial statements not correct Accounting Violation SET

1

1

1

0.4%

Financial Statement Amendment Adverse Opinion

SET

1

15

11

4.6%

SET

1

48

21

8.8%

Disclaimer of Opinion

SET

2

92

48

20.2%

Qualified Opinion

SET

2

2

2

0.8%

Failure to submit financial statements according to the procedure Information Deadline SET 2 80 33 Information Procedure, #1

SET

3

52

52

13.9% 21.8%

Expropriation of assets by managers of the firm. Managers or owners of the firm used inside information to trade the firm's shares for their own benefit. Managers or owners of the firm manipulated the trading of company stocks to mislead or lure others to buy or sell.

Company failed to, and hence was forced to, disclose a connected party transaction by the SET. Company failed to, and hence was forced to, disclose material information to the public by the SET. Company submitted incomplete and/or unclear information, and the SET summoned the company to submit complete information.

Company failed to conduct a tender offer when required to so. Company failed to report when the number of stocks held of another company reached a multiple of 5% of the total number. Company failed to submit or submitted incomplete and/or unclear information about a tender offer to the SET or general investors.

Financial statements failed to comply with Generally Accepted Accounting Standards. Company was required to amend their financial statements. Auditors issued an adverse opinion on the firm's financial statement. Auditors issued a disclaimer of opinion on the firm's financial statement. Auditors issued a qualified opinion on the firm's financial statement.

Failed to submit financial statements or other documents by the deadline. Failed to submit financial statements or other documents by the procedures as specified by the SET. (continued on next page)

888

R. Kouwenberg, V. Phunnarungsi / Pacific-Basin Finance Journal 21 (2013) 881–898

Table 1 (continued) Violation

Source Level of Total Total violations Description of violation severity violations included in the study

Information Procedure, #2

SET

3

14

14

5.9%

Information Procedure, #3

SET

3

1

1

0.4%

370

238

Number Percentage Company submitted financial statements or other documents to SET but such information was not completely released to the public. Submitted financial statements or documents during trading hours.

100.0%

3.2. Violation data Announcements of violations used for our study include when the Thai SEC announces offences of securities law by listed firms regarding expropriation of firms' assets by insiders, falsification of financial statements, insider trading and market manipulation. We also include violations of listing rules detected by the SET. Examples include (i) violations of rules regarding the disclosure of related party transactions and other material information, (ii) violations of rules and procedures regarding financial statements (financial statements submitted by the firms contains errors, or does not comply with generally accepted accounting principles, or financial statements were not submitted by the deadline, or not submitted following the relevant procedures), (iii) violations of rules regarding tender offers, (iv) when the firm's auditor issues a qualified opinion, and adverse opinion, or a disclaimer of opinion about the firm's financial statement. SET announces violations of listing rules to the market by posting trading signs (Notice Pending, Halt, and Suspension), accompanied by a brief written statement explaining the violation. Table 1 provides an overview of all the SEC and SET information used to measure violations, as well as the number of violations involved. The total number of violations in the sample period 2003–2010 is 370. However, we can only include 238 violations (64%) in our study because the remaining violations were committed by firms in rehabilitation, with the shares suspended from trading. Shares of firms in rehabilitation, usually involving reorganization and debt restructuring, do not trade until the reorganization has been completed successfully. We classify violations based on severity, into minor, medium and severe violations (see column 3, Table 1). Out of 238 violations in our sample, 28%, 41%, and 31% are minor, medium and severe, respectively. We use the severity-weighted average number of violations as a proxy to classify firms into bad firms (with high past violations) vs. good firms (low past violations) to test Hypothesis 3. 4 In particular, we use the variables Violtot02S, Violtot04S, Violtot06S and Violtot08S to classify firms into good and bad groups in the periods 2003–2004, 2005–2006, 2007–2008 and 2009–2010, respectively, updating the classification every two years. Table 3 shows descriptive statistics of the violation variables. 3.3. Corporate governance data Since 2002 Thai listed firms are required to disclose information about their governance policies in their annual report and information disclosure forms. We collect firm-level corporate governance data in the years 2002, 2004, 2006, and 2008, and use it to construct a corporate governance index CGScore. The index consists of 17 elements listed in Table 2, including CEO/chairman duality, the structure and independence of the board, the presence of board committees for remuneration and governance, whether the firm has its own corporate governance policy, and disclosure of remuneration of directors and executives. Seminal studies on U.S. firms use governance indices based on takeover defenses and voting rights, such as the G-index of Gompers et al. (2003) and the entrenchment index of Bebchuk et al. (2009). As high ownership concentration is the norm in the Thai stock market and hostile takeovers are extremely rare, takeover defenses are not included in our governance score. Our governance index unfortunately also does not include shareholder voting rights, as our source of data (annual reports and disclosure forms) does not provide this information. However, given the prevalence of majority shareholders in Thailand (average 4

Severe violations have a weight of 3, medium violations a weight of 2, and minor violations 1.

R. Kouwenberg, V. Phunnarungsi / Pacific-Basin Finance Journal 21 (2013) 881–898

889

Table 2 Definition of the corporate governance summary score. This table presents a list of elements of the corporate governance policy of Thai listed firms that are collected in the years 2002, 2004, 2006, and 2008. The weights used to create the corporate governance summary score (CGScore) are shown in the column “Score”. The minimum total score is 0 and the maximum is 10 for a firm following all policies listed. Category Sub-category

Score

A. Board independence and CEO duality Separation of CEO and Chairman positions Independent director ratio is 33% or higher Independent director ratio is 40% or higher Independent director ratio is 50% or higher Number of independent directors is 4 or higher Number of independent directors is 5 or higher B. Board committees Board has a remuneration committee Remuneration committee has an independent chairman Board has a nominating committee Nominating committee has an independent chairman Board has a corporate governance committee Corporate governance committee has independent chairman C. Written CG and ethics policies The firm has its own written corporate governance policy The corporate governance policy has been approved by the board The firm has a business ethics code for employees D. Disclosure of remuneration details Director remuneration disclosed (for each director separately) Executive remuneration disclosed (for each executive separately) Total

4.0 1.0 0.5 0.5 1.0 0.5 0.5 3.0 0.5 0.5 0.5 0.5 0.5 0.5 1.5 0.5 0.5 0.5 1.5 0.5 1.0 10.0

free-float is 37.5%), we expect shareholder voting rights to matter less for governance than in developed markets. As a robustness check, we also use an alternative governance index provided by the Stock of Exchange Thailand based on information in 2002 that covers voting rights. We use our governance index CGScore to classify firms into high CG vs. low CG, to test Hypothesis 2. Similar to the past violations' track record, we update the governance index every two years. In particular, we use CGScore02, CGScore04, CGScore06, and CGScore08 to classify firms in the periods 2003–2004, 2005–2006, 2007–2008 and 2009–2010, respectively. Table 3 shows descriptive statistics of the governance scores.

Table 3 Descriptive statistics of CG scores and past violations. This table shows descriptive statistics of the corporate governance summary score (CGScore) and the past violations variables (VioltotS) during the period 2002–2008. Each variable is measured four times: in 2002, 2004, 2006, and 2008. See Table 2 for the definition of the corporate governance score. The past violations variable measures the average number of violations per year listed in the period since 1990, weighted by severity (×1 for minor, ×2 for medium and ×3 for severe). Variable

Period

Number of firms

Mean

Median

A. Corporate governance summary score Score on a scale from 0 (min) to 10 (max), using CG information and weights CGScore02 CG score in 2002 320 2.02 1.50 CGScore04 CG score in 2004 312 2.98 2.50 CGScore06 CG score in 2006 303 3.37 3.00 CGScore08 CG score in 2008 292 3.96 3.50 B. Past violations Average number of violations per year listed in the period, weighted by severity. Violtot02S Period: 1990–2002 318 0.341 0.23 Violtot04S Period: 1990–2004 332 0.340 0.20 Violtot06S Period: 1990–2006 321 0.343 0.19 Violtot08S Period: 1990–2008 310 0.339 0.17

Std. dev.

Min

shown in Table 2. 1.24 0 1.70 0 1.88 0 1.99 0

0.53 0.52 0.53 0.52

0 0 0 0

Max

Skew

Kurtosis

7.5 9.0 9.0 9.0

1.44 1.02 0.76 0.41

3.32 0.98 0.13 −0.48

7.20 4.80 4.89 4.31

7.25 4.08 3.83 3.32

85.8 26.5 24.6 18.0

890

R. Kouwenberg, V. Phunnarungsi / Pacific-Basin Finance Journal 21 (2013) 881–898

Table 4 Average and cumulative average daily abnormal return of all violations. This table summarizes the average and cumulative average daily abnormal return during day −30 to day +30 around the violation announcement (day 0). The daily return is calculated using continuously compounded total returns. Abnormal returns are market model adjusted, with the SET Total Return Index as the market index. The average abnormal return (AAR) is calculated based on Eq. (2) and cumulative average abnormal return is calculated based on Eq. (4). The number of violations used varies by day as some stocks are illiquid and have no closing price on that particular day. The number of violations used for calculating AAR on day 0 is relatively low because some stocks are temporarily suspended from trading by SET. Values significantly different from zero at a significance level of 10%, 5%, and 1% are marked ⁎, ⁎⁎, and ⁎⁎⁎ respectively. Event day

No. of violations

AAR

Std. dev.

t-statistics

No. of violations

CAAR

Std. dev.

t-statistics

−30 −25 −20 −15 −10 −9 −8 −7 −6 −5 −4 −3 −2 −1 0 +1 +2 +3 +4 +5 +6 +7 +8 +9 +10 +15 +20 +25 +30

179 180 189 191 191 191 195 190 192 189 190 185 197 188 111 167 165 168 169 172 173 174 174 172 172 175 188 191 189

0.53% −0.04% 0.42% −0.08% 0.24% −0.25% −0.51% 0.32% −0.79% 0.29% 0.17% −0.35% −0.42% −0.80% 0.41% −1.71% 0.03% −0.56% −0.31% 0.58% 0.33% −0.06% −0.12% −0.33% 0.31% −0.09% 0.21% 1.06% −0.50%

5.70% 4.94% 4.26% 5.10% 4.77% 5.12% 4.85% 5.40% 5.05% 4.30% 4.98% 4.07% 4.67% 8.19% 5.57% 6.51% 6.06% 5.10% 7.99% 4.62% 6.58% 5.11% 5.29% 6.18% 6.24% 5.82% 5.82% 12.00% 3.83%

1.247 −0.122 1.362 −0.228 0.686 −0.662 −1.474 0.828 −2.167** 0.916 0.457 −1.176 −1.254 −1.339 0.766 −3.387*** 0.065 −1.417 −0.501 1.652 0.655 −0.146 −0.295 −0.706 0.652 −0.211 0.498 1.226 −1.800*

179 180 189 191 191 191 195 190 192 189 190 185 197 188 111 167 165 168 169 172 173 174 174 172 172 175 188 191 189

0.53% 0.11% 1.70% 1.93% 2.55% 2.24% 1.91% 2.70% 1.07% 1.79% 2.08% 1.57% 0.98% 0.20% 2.85% −0.16% −0.50% −1.27% −0.92% −0.02% −0.17% −0.32% −0.79% −0.71% −0.72% −1.28% −1.28% −3.72% −5.24%

5.70% 10.62% 16.67% 16.09% 19.47% 20.06% 20.35% 21.26% 22.77% 23.59% 24.21% 24.25% 23.48% 25.28% 18.81% 25.86% 27.23% 27.05% 28.68% 29.07% 30.58% 30.56% 31.39% 32.19% 32.96% 35.41% 36.25% 40.23% 42.16%

1.247 0.135 1.398 1.655* 1.813* 1.541 1.311 1.752* 0.651 1.042 1.186 0.882 0.588 0.107 1.594 −0.081 −0.234 −0.608 −0.418 −0.008 −0.074 −0.136 −0.333 −0.291 −0.286 −0.478 −0.484 −1.279 −1.707*

4. Empirical results 4.1. Event study and abnormal returns Table 4 shows the average abnormal return (AAR) and cumulative average abnormal return (CAAR) during a period of 30 days before and 30 days after the violation announcement (day 0). On the day immediately after the violation announcement (day +1), the abnormal return is −1.71% and significantly different from zero (t‐stat = −3.387, p-value = 0.001). On the event day (day 0) itself the abnormal return is not significantly different from zero. The number of violations used for calculating AAR on the event day is relatively low because some stocks are temporarily suspended from trading. Before the event day no information leakage seems to take place, as the abnormal return on most days is insignificant and the CAAR in the pre-event period is positive. Table 5 shows the abnormal return on the three days surrounding the announcement date (day 0), namely on days −1, 0, +1 and in the 3-day event window, classified by severity of the violation. For severe violations the market reaction is significantly negative on day +1 (−2.43%) and in the 3-day event period (−4.10%). For medium violations the abnormal return is significantly negative on day −1 (−1.14%), day +1 (−2.58%) and in the 3-day event window (−2.07%). For minor violations the abnormal return is not significant. When all types of violations are combined, we find significant negative returns of −1.71% on day +1 and −2.16% in the 3-day event period. In sum, we find that the market reaction to severe and medium violations is

R. Kouwenberg, V. Phunnarungsi / Pacific-Basin Finance Journal 21 (2013) 881–898

891

Table 5 Abnormal return surrounding violation on day 0 classified by severity of violations. This table shows descriptive statistics of the abnormal returns surrounding a violation on day 0, classified by the severity of the violation. Daily returns are calculated based on continuously compounded total returns. Abnormal returns are market model adjusted, with the SET Total Return Index as the market index. Row “Violations (number)” is the number of violations having abnormal return (trading) data on that day under each category. Row “Negative abnormal return (%)” is the percentage of violations with negative abnormal return over the total number of violations in each event window. Mean abnormal returns (AAR) significantly different from zero at a significance level of 10%, 5%, and 1% are marked *, **, *** respectively. Minor violation Day −1 Day 0 Mean (AAR) Median Std. dev. t-statistics Violations (number) Negative abnormal return (%)

0.12% 0.02% 3.14% 0.28 52 50%

−0.10% 0.06% 4.83% −0.15 54 48%

Medium violation Day +1

Days −1, 0, +1 Day −1

Day 0 Day +1

0.01% 0.02% 3.92% 0.03 53 49%

−0.03% −0.13% 9.02% −0.02 54 54%

−1.14% −0.73% 4.88% −2.08** 79 63%

3.31% 1.53% 7.80% 2.04* 23 22%

−2.43% −2.38% 7.45% −2.46** 57 72%

−4.10% −2.58% 13.76% −2.36** 63 73%

−0.80% −0.19% 8.19% −1.34 188 57%

Severe violation Mean (AAR) Median Std. dev. t-statistics Violations (number) Negative abnormal return (%)

−1.17% −0.20% 13.44% −0.66 57 56%

−0.77% 0.02% 4.20% −1.06 34 41%

Days −1, 0, +1

−2.58% −2.54% 7.19% −2.71*** 57 79%

−2.07% −2.06% 10.31% −1.80* 80 73%

−1.71% −1.43% 6.51% −3.39*** 167 67%

−2.16% −1.73% 11.28% −2.69*** 197 68%

All violations 0.41% 0.20% 5.57% 0.77 111 41%

significantly negative, but not to minor violations.5 Further, there is no sign of information leakage during the period before the announcement day. 4.2. Hypothesis test results about governance and past violations To test Hypothesis 2 we use the median of CGScore for all listed firms to classify firms into high CG firms vs. low CG firms, with the governance measure updated every two years, in 2002, 2004, 2006 and 2008. We compare the AAR during the event window (− 1, 0, + 1) around violation announcements between these two groups. Table 6 shows the results. The mean AAR of high CG firms (− 1.1%) is less negative than that of low CG firms (− 2.7%), in contrast to Hypothesis 2, but the difference is not significant (p‐value = 0.299). In sum, the results do not support Hypothesis 2: we find no significant difference in market reaction when low CG and high CG firms violate the rules. To test Hypothesis 3 we use the median of the severity-weighted violations of all firms (VioltotS) to classify firms into bad firms (high past violations) vs. good firms (low past violations). The violations measure and the firm classification are updated every two years, in 2002, 2004, 2006, and 2008. We then compare the mean abnormal return during the violation announcement event period (− 1, 0, + 1) between these two groups. Table 6 shows that the mean AAR for bad firms is − 1.3%, while when good firms violate the rules the average abnormal return is − 4.4%. We find a significant difference in the mean abnormal return between good firms and bad firms (−3.1%, p‐value = 0.088), supporting Hypothesis 3. With respect to Hypothesis 4, Table 6 shows that the mean AAR for firms with a good track record and high CG is less negative (2.1% difference) than for other firms. Based on Hypothesis 4 we expected the opposite: for good firms with high CG violations should come more as a surprise, leading to a stronger negative reaction. Hypothesis 4 is therefore not supported. We have also double-sorted the firms in four groups, based on CG score (high vs. low) and past violations record (good vs. bad): the mean AAR in each group is shown in Table 6. By far the largest market reaction (− 8.1%) occurs when firms with a good track record and low CG scores commit violations. In 5 We have also tested for differences in the mean (and median) abnormal return between the categories of violations. We find that the AAR of severe and medium violations is significantly more negative than the AAR of minor violations. The AAR of severe violations is more negative than the AAR of medium violations, however, the difference is not significant.

892

R. Kouwenberg, V. Phunnarungsi / Pacific-Basin Finance Journal 21 (2013) 881–898

Table 6 Abnormal event window return classified by governance and past violations. This table shows the mean abnormal return (AAR) in the event window (−1, 0, +1) surrounding the violation on day 0, classified by CG score and past violations (two-way sorting). Low (high) CG firms are firms with governance score lower (higher) than the median of CGScore of all listed firms. Bad (good) firms are firms with past violations higher (lower) than the median of VioltotS for all listed firms. The governance score (CGScore) and past violations (VioltotS) measures are updated every two years: in 2002, 2004, 2006 and 2008. Abnormal returns are market model adjusted, with the SET Total Return Index as the market index. The row “Violations” displays the number of violations with abnormal return data in each category. The lower part of the table shows tests of Hypotheses 2, 3 and 4 (H2, H3 and H4). Panel A shows results for the full sample 2003–2010, Panel B the first half 2003–2006 and Panel C the second half 2007–2010. Mean abnormal returns (AAR) significantly different from zero at a significance level of 10%, 5%, and 1% are marked *, **, *** respectively. Panel A

Full sample: 2003–2010

Mean (AAR) Violations

High CG

Mean (AAR) Violations

Low CG

Mean (AAR) Violations

Bad −1.6% 36 Bad −1.1% 100 All bad −1.3% 136

Good −0.3% 26 Good −8.1%⁎⁎⁎

All high CG −1.1% 62 All low CG −2.7%⁎⁎

28 All good −4.4%⁎⁎

128 All violations −2.1%⁎⁎

54

190

Tests of hypotheses

High vs. low CG

Good vs. bad record

Good & high CG vs. others

Hypothesis: expected sign Mean (AAR) t-stat

H2: − 1.6% 1.04

H3: − −3.1%⁎ −1.71

H4: − 2.1% 1.43

Panel B

First half of sample: 2003–2006

Mean (AAR) Violations

High CG

Mean (AAR) Violations

Low CG

Mean (AAR) Violations

Bad −0.6% 21 Bad −0.4% 68 All bad −0.4% 89

Good 2.7% 11 Good −8.6%⁎⁎ 25 All good −5.2%⁎⁎

All high CG 0.5% 32 All low CG −2.6%⁎ 93 All violations −1.8% 125

36

Tests of hypotheses

High vs. low CG

Good vs. bad record

Good & high CG vs. others

Hypothesis: expected sign Mean (AAR) t-stat

H2: − 3.1% 1.64

H3: − −4.8%⁎ −1.96

H4: − 4.9%⁎⁎ 2.21

Panel C

Second half of sample: 2007–2010

Mean (AAR) Violations

High CG

Mean (AAR) Violations

Low CG

Mean (AAR) Violations

Bad −3.0% 15 Bad −2.8%⁎ 32 All bad −2.8% 47

Good −2.5% 15 Good −3.9% 3 All good −2.8%⁎⁎

All high CG −2.8% 30 All low CG −2.9%⁎

18

35 All violations −2.8%⁎⁎ 65

Tests of hypotheses

High vs. low CG

Good vs. bad record

Good & high CG vs. others

Hypothesis: expected sign Mean (AAR) t-stat

H2: − 0.1% 0.03

H3: − 0.1% 0.05

H4: − 0.4% 0.21

R. Kouwenberg, V. Phunnarungsi / Pacific-Basin Finance Journal 21 (2013) 881–898

893

stark contrast, for firms with good track record and high CG scores the mean AAR is only −0.3%. This result suggests that the market scrutinizes and discounts a good firm's governance policies once violations occur, leading to a strong negative reaction for good firms with low CG. On the other hand, for firms with a bad track record governance problems appear already discounted, as the market reaction is not much different between bad high CG firms and bad low CG firms (− 1.6% vs. − 1.1%). As the Thai corporate governance code was introduced in March 2002, investors may have needed time to fully learn and appreciate the difference in governance scores between firms. Bebchuk et al. (2010) argue that learning effects are important and can explain why abnormal governance returns have disappeared over time in the U.S. stock market. Panel B and Panel C of Table 6 show the market reactions to violations in the two sub-periods 2003–2006 and 2007–2010. We find that the results in the first period (2003–2006) are similar to the full sample. There is a strong negative reaction of − 8.6% when good firms with low CG commit violations and Hypothesis 3 is supported. In the second half of the sample, mean abnormal returns are negative, but of similar magnitude in all four categories. The mean AAR is not significantly different across the four groups (low/high CG, good/bad firm) in 2007–2010. Taken together, these results suggest that learning effects took place in the first half of the sample (2003–2006). Initially, the market may have not fully discounted low governance scores, but only the firm's more tangible track record of past violations. However, once a firm with a good track record commits violations, the market scrutinizes and discounts weaknesses in governance policies: the abnormal return is drastically more negative for good firms with low CG, compared to good firms with high CG (mean AAR difference: − 11.3% in 2003–2006). In the second half of the sample, 2007–2010, there are no noticeable differences in market reactions to violations, suggesting the market has now fully discounted both firms' formal corporate governance policies and past violations record. We have investigated which components of our governance index contribute most to the negative market reaction when good firms with low CG commit violations. We find that only two governance aspects create significant differences in market reactions for good firms: CEO/chairman duality and whether the audit committee is chaired by an independent director. For good firms who do not separate the positions of Chairman and CEO, or whose audit committee is not chaired by an independent director, the mean AAR is − 10.4%. The market seems to recognize and discount these governance weaknesses once violations occur. Among firms with a bad past track record, the market reaction is − 2.9% for firms without CEO/Chairman duality or audit committee without independent chairman, suggesting that for bad firms weaknesses in governance were already discounted. As a robustness check, we have further classified the violations based on severity (minor, medium and severe) and re-examined the hypotheses. A drawback is that test power is poor due to the low number of observations in each category. For example, when good firms with low CG commit a severe violation the mean AAR is − 16.7%, however there are only 7 violations in this category and the mean is not significantly different from zero. Ignoring tests of significance, the findings are qualitatively similar across the violation severity categories. Hypothesis 3 is supported: the market reacts more negatively when firms with a good track record commit violations, regardless of violation severity. The strongest market reaction always occurs when firms with good track record and low CG commit violations. 6 As a further robustness check, we have repeated the tests with an alternative governance score based on an index constructed by the Stock Exchange of Thailand for all listed firms in 2002 (see Ananchotikul et al., 2010). This CG measure includes information about shareholder rights, but it is only available for the year 2002. We use this measure to sort firms into low and high governance groups in the first half of the sample, 2003–2006. We find that the results are the same as in Panel B of Table 6, with Hypothesis 2 not supported (no difference between low and high CG firms), and the strongest market reaction occurring for good firms with low CG (− 5.8%). As a final robustness check, we have also calculated abnormal event returns by subtracting the total market return from the firm's stock return in the event period, i.e. without correcting for risk (beta) with the market model. All results are similar when using market-adjusted abnormal returns (results available upon request).

6 The mean AAR is -16.7% for severe violations (N=7, p-value = 0.121), -3.6% for medium violations (N=11, p-value = 0.310) and -7.1% for minor violations (N=10, p-value = 0.044). We cannot reject that the mean AAR for good firms with low CG is equal across the three violation severity categories.

894

R. Kouwenberg, V. Phunnarungsi / Pacific-Basin Finance Journal 21 (2013) 881–898

4.3. The determinants of market reactions to violation announcements The results of the event study in the previous section do not control for other variables that may explain the market reaction to violation announcements. In this section we estimate a regression model to examine the factors explaining the stock market reaction to the announcement of violations. The model is described below, whereas the independent variables used in the model are explained in detail in Appendix A.   ARi ¼ β0 þ β1 DHighCGi þ β2 DGoodFirmi þ β3 DHighCGi xDGoodFirmi ′ þβ4 DSeverei þ β5 DMediumi þ β6 Xi þ εi;

ð5Þ

where ARi is the abnormal return during the event window (− 1, 0, + 1), D_HighCGi, and D_GoodFirmi are dummy variables used to classify firms in high/low CG groups and good/bad track record groups, D_Severei and D_Mediumi are dummies for severe and medium violations, Xi is a (k x 1)-vector of control variables, β6 is a (k x 1)-vector with regression coefficients, and εi is the error term. Detailed explanations of the controls variable and expected signs are shown in Table A1 in Appendix A, whereas descriptive statistics of all variables used in the regression model are in Table A2. The full model with interaction effects in Eq. (5) is not suited for testing Hypotheses 2, 3 and 4. For example, if we would like to test Hypothesis 2 (market reacts more negatively to violations by firms with high CG scores compared to low CG firms) then we should only include the dummy D_GoodFirmi in the model to make the appropriate comparison.7 The first three columns of Table 7 (labeled H2, H3, and H4) show estimation results for models that allow us to precisely test the hypotheses. The last two columns of Table 7 (labeled 5a and 5b) show results for the full model in Eq. (5), to test for interaction effects. In the full model the base category for comparisons, represented by the constant, is a firm with bad track record and low CG. In the full model specifications in Table 7 (last two columns, 5a and 5b), the interaction effect between the governance dummy and the past track record dummy is strong and significant. This confirms our earlier finding that market reacts especially strong to violations of firms with a good track record and low governance scores. 8 When we include all nine control variables in the model, we find that model fit is poor (column 5b in Table 7): adjusted R 2 is negative, the F-test is insignificant, and all controls have insignificant beta. As including many redundant variables inflates standard errors, we remove highly insignificant control variables with p-value greater than 30% one by one. Column 5a in Table 7 shows the model after eliminating redundant variables: it includes return on assets (ROA, p-value b 0.001) and a dummy for having at least one analyst covering the firm (Popularity, p-value = 0.261). The model is now significant (F-stat = 5.085), with adjusted R 2 of 13.1%. Columns H2, H3 and H4 of Table 7 show the models suited for testing the hypotheses. Only Hypothesis 3 is supported: the market reacts stronger when good firms commit violations, compared to firms with a bad track record. The market does not react stronger when high CG firms commit violations, compared to low CG firms (Hypothesis 2). Further, Hypothesis 4 is not supported: the market does not react more negative to violations of firms with good governance and high CG. Among the control variables, only return on assets (ROA), a proxy for firm profitability, is significant. When firms with poor profitability (low ROA) violate the rules, it creates a stronger negative market reaction. Managers or controlling shareholders of a firm with low profitability have a higher incentive to expropriate (Durnev and Kim, 2005), and the market seems to take this into account when firms commit violations. As a robustness check we have also estimated the regression models using severe violations only. The results are similar as reported in Table 7, with a significant interaction effect due to the strong negative market reaction in the group of good firms with low CG. When we split the sample in two periods, 2003– 2006 and 2007–2010, we find that all results reported in Table 7 are present in the first half of the sample (2003–2006). In the second half of the sample (2007–2010) governance and past track record do not have 7 In contrast, in the full model in Eq. (5), the dummy D_GoodFirmi only measures the difference between bad firms with high CG and bad firms with low CG. 8 The exact interpretation of the interaction term in the full model is not straightforward: the coefficient of D_HighCGi x D_GoodFirmi measures the difference between “good high CG and bad low CG” firms compared to “bad high CG and good low CG” firms. But, intuitively, the interaction effect is simply a reflection of the fact that the market reaction is much stronger for good firms with low CG.

R. Kouwenberg, V. Phunnarungsi / Pacific-Basin Finance Journal 21 (2013) 881–898

895

Table 7 Estimation results for the regression model explaining the stock market reaction to violation announcements. This table shows estimation results of the regression model for explaining the stock market reaction to violation announcements (Eq. (5)). The dependent variable is the mean abnormal return (AAR) in the event window (−1, 0, +1) surrounding the violation on day 0. Columns H2, H3 and H4 show model specifications with dummies that test exactly Hypotheses 2, 3 and 4. The expected sign for the hypothesis is shows in the column ‘Hypothesis’. Columns 5a and 5b show results for the full model shown in Eq. (5), including the interaction between D_HighCG (high CG dummy) and D_GoodFirm (low past violations dummy). Model 5b includes a full set of 9 control variables. Other models include only 2 controls with p‐value ≤ 30%, namely return on assets (ROA) and a dummy for at least one analyst covering the firm (Popularity). D_Severe and D_Medium are dummies for severe and medium violations. See Table A1 in Appendix A for a full list of control variables, definitions and expected signs. Standard errors are reported below the estimated coefficients in parentheses. Coefficients significantly different from zero at a significance level of 10%, 5%, and 1% are marked *, **, and *** respectively. Testing hypotheses Variable

Hypothesis

Constant D_Severe D_Medium D_HighCG

H2: −

D_GoodFirm

H3: −

D_HighCG x D_GoodFirm ROA

H4: −

Popularity Other controls (7 variables) Adj. R2 F-statistic Observations

Interaction effect

H2

H3

H4

5a

5b

−1.4% (1.8%) −2.6% (2.2%) −0.1% (2.1%) −0.5% (1.8%)

0.2% (1.8%) −3.8%⁎ (2.1%) −0.5% (2.0%)

−1.6% (1.8%) −2.6% (2.2%) −0.1% (2.1%)

0.3% (2.5%) 3.9%⁎⁎⁎

1.3% (1.9%) −3.6%⁎ (2.1%) −0.7% (2.0%) −3.2% (2.1%) −9.2%⁎⁎⁎ (2.3%) 10.1%⁎⁎⁎ (3.7%) 4.5%⁎⁎⁎

1.0% (3.0%) −2.4% (2.4%) −2.0% (2.3%) −1.8% (2.3%) −6.7%⁎⁎ (2.6%) 8.9%⁎⁎ (4.2%) 2.3% (1.9%) −1.9% (2.4%) Yes −0.010 0.901 148

−5.1%⁎⁎⁎ (1.8%)

4.0%⁎⁎⁎ (1.1%) −1.3% (1.8%) No 0.066 3.680⁎⁎⁎

4.3%⁎⁎⁎ (1.0%) −0.8% (1.8%) No 0.105 5.448⁎⁎⁎

(1.0%) −1.4% (1.9%) No 0.066 3.664⁎⁎⁎

(1.0%) −2.0% (1.8%) No 0.131 5.085⁎⁎⁎

190

190

190

190

a significant impact on abnormal returns, suggesting that investors have learnt to assess the substance of governance practices. 5. Conclusion In this paper we study the market reaction to announcements of violations of rules and regulations by Thai listed firms. First, we find that the market reacts negatively to violation announcements: we document a significant average abnormal return of − 2.2% during the event window (day − 1, 0, + 1). The market reaction is especially strong when firms commit violations classified as severe: − 4.1%. Second, we find no significant difference between the abnormal returns of firms with high and low corporate governance scores. However, we do find a significant difference in market reaction when we classify firms as “good” or “bad” based on their past violation track record (below or above the median). We also find a strong interaction effect between governance and a firm's track record: the abnormal return is most negative, − 8.1% on average, when good firms with low CG score commit violations. In stark contrast, the average market reaction to violations is only − 0.3% for good firms with high CG. When bad firms commit violations, the market reaction does not depend on governance: − 1.1% for bad firms with low CG and − 1.6% for bad firms with high CG. The pattern in abnormal returns suggests that the market initially did not fully discount governance policies, but relied mainly on firms' more tangible track record of past violations. However, once a firm with a good track record commits a violation, the firm's governance is scrutinized and the market then discounts weaknesses in governance policies to account for higher expropriation risk and future violations. This may explain why there is such a large difference in average abnormal return for good firms with low CG (−8.1%)

896

R. Kouwenberg, V. Phunnarungsi / Pacific-Basin Finance Journal 21 (2013) 881–898

and good firms with high CG (−0.3%). Further, we find that this difference is mainly present in the first half of the sample, 2003–2006, when good governance policies were still relatively new in the Thai stock market (after introduction of a good governance code in 2002). In the second half of the sample, 2007–2010, there are no significant differences in abnormal returns around violation announcements between firms in the four categories (high/low CG and good/bad track record), suggesting that the market by then fully discounted both past violations and the firm's governance policies. The presence of a strong interaction effect between governance scores and the track record of past violations shows that investors assesses a firm's substantive governance practices by relying on a combination of observed behavior (violations) and the firm's formal governance policies (e.g., having independent directors). Formal governance policies considered in isolation may not provide sufficient information about the firm's true governance practices, as bad firms can try to window-dress governance policies and not all good firms may want to bear the costs of improving their governance policies. Violations can help reveal the firm's true underlying governance practices through time. Among all governance policies considered, we find that the market reaction is most negative to violations of good firms who do not separate the positions of CEO and chairman and to violations of good firms where the audit committee is not chaired by an independent director. Other governance policies, like having a higher ratio of independent directors or a remuneration committee, do not have an effect on abnormal violation returns. Results from a regression model, with the abnormal event window return as the dependent variable and several control variables, confirm the main findings from the event study. Appendix A. Regression model variables

Table A1 Definition of regression model variables and expected signs. This table provides the definitions of all variables used in the regression model for explaining the abnormal violation announcement returns in Eq. (5). All variables based on annual report information are updated annually, lagged two years if the violation occurs before the end of June of the year, and lagged one year if the violation occurs after the end of June. The expected sign of the beta coefficient of each independent variable and each control is shown in the column ‘Expected sign of beta’. Variable

Definition

Expected sign of beta

Dependent variable AR −1,0,1 Abnormal violation announcement return during the event window (days −1, 0, +1). Independent variables D_Severe A dummy variable that equals 1 for severe violations and 0 otherwise. D_Medium A dummy variable that equals 1 for medium violations and 0 otherwise. D_HighCG A dummy variable equal to 1 if a firm's CGScore is greater than the median CGScore of all firms; 0 otherwise. D_GoodFirm A dummy variable equal to 1 if a firm's severity weighted violations (VioltotS) is less than the median of VioltotS of all firms; 0 otherwise. Controls for company characteristics PastAAR Avg. abnormal return in the year before the violation announcement. Popularity A dummy variable that equals 1 when the firm has at least one analyst coverage (in the year 2002); 0 otherwise. LnAsset Natural logarithm of total assets. D_Big4 A dummy variable that equals 1 when the firm's auditor is a Big 4 firm in the year of violation; 0 otherwise. Controls for agency conflicts Leverage Total debt to equity ratio, winsorized at the 99% percentile (right tail). Div. payout Dividend payout ratio: dividends divided by earnings, winsorized at the value 200%, and negative payout ratios replaced with 200%. Tangibility Ratio of plant, property and equipment to total assets. ROA Return on assets (%), winzorized at the 1% and 99% percentile. Control Dummy variable for firms with at least one controlling shareholder with 25% or larger block in 2000; 0 otherwise. Updated once in 2004.

− − − −

− − − −

− − − − +/−

R. Kouwenberg, V. Phunnarungsi / Pacific-Basin Finance Journal 21 (2013) 881–898

897

Table A2 Descriptive statistics of regression model variables. This table presents descriptive statistics of the variables in the regression model for explaining the stock market reaction to violation announcements (Eq. (5)). See Table A1 for variable definitions. The sample consists of 190 violation announcements. Firm-level data summarized in the table are for the firms committing the violations, with each violation included separately (one firm can commit more than one violation). As a result, the sample is not representative for all Thai firms. Some variables have missing values: the column ‘Obs.’ shows the number of observations available. Variable AR −1,0,+1 D_Severe D_Medium D_HighCG D_GoodFirm PastAAR Popularity LnAsset D_Big4 Leverage Div. payout Tangibility ROA (%) Control

Mean

Median

Stdev.

Min

Max

Skew

Kurt.

Obs.

−0.021 0.321 0.421 0.326 0.284 −0.001 0.274 7.844 0.479 2.876 0.187 0.375 −3.044 0.661

−0.018 0.000 0.000 0.000 0.000 −0.001 0.000 7.500 0.000 1.410 0.000 0.365 0.395 1.000

0.113 0.468 0.495 0.470 0.452 0.005 0.447 1.358 0.501 9.231 0.365 0.254 18.990 0.475

−0.689 0.000 0.000 0.000 0.000 −0.022 0.000 4.930 0.000 0.050 0.000 0.004 −145.71 0.000

0.396 1.000 1.000 1.000 1.000 0.028 1.000 13.810 1.000 117.95 2.000 0.945 42.050 1.000

−0.6 – – – – 1.3 – 0.9 – 11.3 2.3 0.4 −2.9 –

11.0 – – – – 15.7 – 4.0 – 140.0 9.2 2.1 20.1 –

190 190 190 190 190 188 190 190 190 175 183 189 190 171

References Agrawal, A., Chadha, S., 2005. Corporate governance and accounting scandals. Journal of Law and Economics 48, 371–406. Ananchotikul, N., Kouwenberg, R., Phunnarungsi, V., 2010. Do firms decouple corporate governance policy and practice? European Financial Management 16, 712–737. Beasley, M., 1996. An empirical analysis of the relation between the board of director composition and financial statement fraud. Accounting Review 71, 443–465. Bebchuk, L.A., Cohen, A., Ferrell, A., 2009. What matters in corporate governance? Review of Financial Studies 22, 783–827. Bebchuk, L.A., Cohen, A., Wang, C.C.Y., 2010. Learning and the disappearing association between governance and returns. Working Paper 15912, National Bureau of Economic Research. Carcello, J.V., Neal, T.L., Palmrose, Z.-V., Scholz, S., 2011. CEO involvement in selecting board members, audit committee effectiveness, and restatements. Contemporary Accounting Research 28, 385–395. Chen, G., Firth, M., Gao, D., Rui, O., 2006. Ownership structure, corporate governance, and fraud: evidence from China. Journal of Corporate Finance 12, 424–448. Chen, K., Chen, Z., Wei, J., 2009. Legal protection of investors, corporate governance, and the cost of equity capital. Journal of Corporate Finance 15, 273–289. Chhaochharia, V., Grinstein, Y., 2007. Corporate governance and firm value: the impact of the 2002 governance rules. Journal of Finance 62, 1789–1825. Cumming, D.J., Johan, S., 2008. Global market surveillance. American Law and Economics Review 10, 454–506. Cumming, D.J., Hou, W., Lee, E., 2011. The role of financial analysts in deterring corporate fraud in China. Working paper, SSRN. Cumming, D.J., Johan, S., Li, D., 2011. Exchange trading rules and stock market liquidity. Journal of Financial Economics 99, 651–671. Djankov, S., La Porta, R., Lopez-de-Silanes, F., Shleifer, A., 2008. The law and economics of self-dealing. Journal of Financial Economics 88, 430–465. Doidge, C., Karolyi, A., Stulz, R.M., 2007. Why do countries matter so much for corporate governance? Journal of Financial Economics 86, 1–39. Durnev, A., Kim, E.H., 2005. To steal or not to steal: firm attributes, legal environment, and valuation. Journal of Finance 60, 1461–1493. Farber, D., 2005. Restoring trust after fraud: does corporate governance matter? Accounting Review 80, 539–561. Gompers, P., Ishii, J., Metrick, A., 2003. Corporate governance and equity prices. Quarterly Journal of Economics 118, 107–155. Karpoff, J., Lott, J., 1993. The reputational penalty firms bear from committing criminal fraud. Journal of Law and Economics 36, 757–802. Karpoff, J., Lee, D.S., Martin, G.S., 2008. The cost to firms of cooking the books. Journal of Financial and Quantitative Analysis 43, 581–612. Khanthavit, A., Polsiri, P., Wiwattanakantang, Y., 2003. Did families lose or gain control? Thai firms after the East Asian financial crisis. In: Fan, J., Masaharu, H., Juro, T. (Eds.), Designing Financial Systems in East Asia and Japan: Toward a Twenty-First Century Paradigm. Routledge, New York. Klapper, L., Love, I., 2004. Corporate governance, investor protection, and performance in emerging markets. Journal of Corporate Finance 10, 703–728. Lisic, L.L., Neal, T., Zhang, Y., 2011. Audit committee financial expertise and restatements: the moderating effect of CEO power. Working paper, SUNY at Binghamton. Love, I., 2011. Corporate governance and performance around the world: what we know and what we don't. The World Bank Research Observer 26, 42–70.

898

R. Kouwenberg, V. Phunnarungsi / Pacific-Basin Finance Journal 21 (2013) 881–898

Palmrose, Z.-V., Richardson, V., Scholz, S., 2004. Determinants of market reactions to restatement announcements. Journal of Accounting and Economics 37, 59–89. Scholes, M., Williams, J., 1977. Estimating betas from nonsynchronous data. Journal of Financial Economics 5, 309–327. Uzun, H., Szewczyk, S., Varma, R., 2004. Board composition and corporate fraud. Financial Analysts Journal 60, 33–43. Wiwattanakantang, Y., 2001. Controlling shareholders and corporate value: evidence from Thailand. Pacific Basin Finance Journal 9, 323–362.