Strengthened Board Monitoring from Parent Company and Stock Price Crash Risk of Subsidiary Firms. September 2018

Transcription

1 Strengthened Board Monitoring from Parent Company and Stock Price Crash Risk of Subsidiary Firms September 2018 Abstract: This article examines whether and how the strengthened board monitoring from the parent company influences stock price crash risk of the affiliated subsidiary firms. Using a quasi-natural experiment in China, we empirically document that the enhanced board monitoring at the parent company helps reduce crash risk of its subsidiary firms. Three channels of this impact are identified: (i) the enhanced corporate governance at the parent company can somehow prevent it from tunneling wealth from the subsidiary firm; (ii) a better corporate governance of the parent company helps it better monitor the subsidiary firm; and (iii) the governance improvement at the parent company can spill over to the subsidiary firm, which helps constrain managers at the subsidiary firm on misconducts of dealing with negative information. Keywords: Stock price crash risk, Board reform, Central state-owned enterprise (CSOE), Business group 1

2 1. Introduction Prior studies have heavily focused on how a firm s board structure influences the firm itself on the stock price crash risk (hereafter crash risk) (e.g., Yuan et al., 2016; Andreou, et al., 2016), whereas very few scholars have examined the effects that controlling shareholders board structures may have on a firm s crash risk. In this research, we investigate whether and how the strengthened board monitoring from the parent company affects crash risk of its subsidiary firms. Many papers have investigated a parent company's impacts on its subsidiary firms' financing flexibility, financial constraint, and cost of equity capital (e.g., Vijh, 2006; Tam, 2014). It is intriguing, however, to see whether and how the board monitoring from the parent company affects its subsidiary firm s crash risk, a higher moment of firm s financial performance capturing firm s sensitivity to some extreme bad events (Kim et al., 2014; Callen and Fang, 2015). Further, regarding the research on firm-specific crash risk, although some scholars have researched how a firm s crash risk is impacted by its institutional shareholders (e.g., Velury and Jenkins, 2006; Andreou et al., 2016) or controlling shareholders (e.g., Boubaker et al., 2014), no one has studied this topic from the perspective of parent companies. Besides, prior studies investigating the influence of institutional or controlling shareholders use only data samples from developed economies (e.g., America and France); whereas very few literature has tapped into the impact that blockholders may have on a firm s crash risk in the setting of emerging markets. In fact, research on this topic using data sample from emerging countries may be more interesting given that blockholders governance attributes can influence more on a firm s activities particularly in the emerging markets with weak corporate governance (Jiang et al., 2010). The present research is also motivated by a quasi-natural experiment occurring in China. In June 2004, the State-owned Assets Supervision and Administration Commission (SASAC) issued an important guidance of board reform namely On 2

3 Establishing and Improving Board of Directors Trial in Central SOEs, which aims at initially establishing a board of directors and improving the independence of the board and its role in monitoring management on behalf of the SASAC. This red-head document has been guiding a number of Chinese central state-owned enterprises (CSOEs) to carry out business-group-based board reform. In 2005, Baosteel Group Corporation became the first Chinese CSOE commencing the board reform. As of 2015, there are 87 CSOEs who have finished the board reform. Since board reforms carried out within those business groups started from different years, and, for each business group, the reform can put impacts on all units belonging to the same group, it is possible to identify an idiosyncratic shock that is, a publicly traded firm affiliated to a CSOE got exogenously shocked after the reform implementation of its parent company. As such, it is feasible to adopt a difference-in-differences (DID) method in examining a causality effect that a parent company s board monitoring may have on crash risks of its subsidiary firms listed in the capital market. In this study, we hand-collected the key information that identifies the parent-subsidiary relationship between a CSOE and a state-owned firm publicly traded in the Chinese stock markets, and the year of board reform for each of such CSOE groups. These key information as well the relevant financial data provided by third-party vendors allows us to conduct an empirical investigation on how a parent company s board structure influences its subsidiary firm s crash risk. We detect that the source of variation stemming from the board reform at the parent company does exist. In specific, the board reform implemented at the parent company helps decrease crash risk of its subsidiary firms. Also, from the result of a reverse causality test, we show that this effect peaks at the second year of the board reform, and then gradually attenuates afterwards. Furthermore, we examine how the degree of board reform impacts the subsidiary firm s crash risk. We use the ratio of outside directors in the board of the parent company to 3

4 proxy the degree of board reform. A board with higher presentation of outside directors is more independent and plays a stronger monitoring role in the parent company. We find that this effect of parent company s board reform on subsidiary firm s crash risk is positively associated with the degree of reform. Based on our empirical results, it is identified that there are three channels making this effect happen. First, the lower crash risk of the focal firm may be attributed to less serious agency problems from the parent company. In a pragmatic way, managers at both the parent and subsidiary companies may collude to tunnel wealth from the subsidiary firm to the parent company. Due to the board reform, the agency problems from the parent company is somehow alleviated for the focal firm. As a consequence, the focal firm s crash risk may be reduced since this type of agency problems are addressed to some extent. As indicated by our empirical results, the effect of crash risk reduction for the focal firm due to the board reform is bigger in the situation when the agency problem from the parent company is more serious. Second, the board reform improves the parent company s corporate governance which helps the parent company to more effectively monitor its subsidiaries. In practice, there would be some board members on behalf of the parent company sitting on the focal firm s board of directors. We detect that the inclusion of board members from the parent company has actually helped reduce crash risk of the focal firm. We also empirically show that this effect is larger when the focal firm's board consists of more members representing the parent company. Similarly, the corporate governance of the listed firm (i.e. the focal firm) has also been improved thanks to the board reform implemented at its parent company. This improvement is an outcome spilled over from the enhancement of the parent company s corporate governance. An enhanced governance mechanism makes it less possible for those of the focal firm s negative events such as not timely releasing bad information, which eventually could lead to a reduction of crash risk for the focal firm. Our empirical result suggests that this 4

5 impact is indeed greater when the managerial agency costs of a firm is bigger. This paper potentially makes contribution in several aspects. First, this article enriches the lacking literature on whether and how a subsidiary firm is affected by the governance improvement in its parent company. To the best of our knowledge, this is the first article investigating the impact of a strengthened board monitoring from the parent company on the crash risk of its subsidiary firm. We empirically show that the board reform at the parent company can lower its subsidiary firm s crash risk. Our research further distinguishes the reform effect that channels through lowering agency costs from the parent company and that through lowing agency costs existing at the subsidiary firm itself; moreover, we identify that the reform effect is also partly arisen from less negative interaction activities (i.e., collusions) between parent and subsidiary firms. Second, our research offers an empirical evidence from the perspective of parent companies for the bright side standpoint that the problem of crash risk faced by a firm can somehow be eased by its controlling shareholders. The conclusions on whether a blockholder positively or negatively impacts a firm s crash risk is seemingly ambiguous. Some have shown that the monitoring from institutional shareholders constrains the focal firm s top management team in opportunistically managing abnormal accruals, thereby improving the quality of earnings data disclosed by the focal firm (e.g., Velury and Jenkins, 2006; Andreou et al., 2016) and reducing the focal firm s crash risk (Habib et al., 2017). Whereas some argue that controlling shareholders with excessive control tend to withhold negative information in order to cover up their expropriation of minority shareholders interest, thereby resulting in higher crash risk (Boubaker et al., 2014). Previous investigations (e.g., Boubaker et al., 2014) are based on the research setting of developed economies, while ours is based on an emerging market. The misconducts of controlling shareholders in the emerging economies are even more prevalent due to their relatively weak legal environment. Indeed, our investigation suggests that although the 5

6 expropriation by the parent company induces the crash risk in listed subsidiary firms, the strengthening of board monitoring at the parent company can reduce this kind of negative effect. Finally, our study offers some empirical evidences for the positive effect of the ongoing board reform in China. As a significant driving force of China s economy transition, state-owned enterprise groups have to implement a series of marketization reforms undoubtedly. For any reform, it is critical to pretest and evaluate its impacts. Our research suggests that the economic impact of the ongoing board reform at Chinese CSOEs are overall positive, and a deeper reform on CSOEs corporate governance may be encouraged. The rest of this paper proceeds as follows. Section 2 briefly introduces the context of our focus as well as our hypothesis. Section 3 presents our empirical methodology, where our data set, measurement method, and the empirical models are discussed. In Section 4, we discuss the empirical results. Specifically, we summarize our major finding with several robustness tests, and provide a mechanism analysis. Section 5 concludes the article. 2. Institutional background and hypothesis In this section, we briefly introduce the board reform at CSOE groups, and discuss how a group-based board reform may affect crash risk of the member firms within the group. 2.1 Board reform at the SOE groups In China, there are two types of SOEs: CSOEs and local SOEs, which are supervised and administered by central government and local government respectively. Before 2003, most CSOEs did not establish the mechanism of board of directors. This is due to the fact that the CSOEs established before 1993 are not required to have board of directors. For 6

7 those established between 1993 and 2003, it is required to set up board of directors; however, the management team members and board directors within those business groups are mostly overlapping -- meaning the rights of decision-making and execution are not separated, and so the mechanism of board of directors actually is not effective. To deal with this issue, SASAC sent people namely inspection commissioners to the CSOEs in order to better monitor them. Later on those inspection commissioners are renamed as chairmen of supervisory committee. This, however, is not a real mechanism of board of directors. From 2003, the Chinese government issued some red-head documents to guide CSOEs to implement board reform. The purpose of the reform is to establish a real mechanism of board of directors in the CSOEs. In general, a reforming CSOE hires both inside and outside people to sit on its board of directors; also, in order to address somehow the problem of overlap between executive and board members, the number of outside directors is usually set to be bigger than that of inside directors. Thus, board of directors is able to play a real role in the business group s important strategic decisions, risk managements and the C-Suite level executions. Take Baosteel as an example. Baosteel Group Corporation is the first CSOE carrying out the board reform in October Baosteel set up a board of directors which was comprised of nine directors, among which five were outside members. Baosteel is the first CSOE whose board of directors has over-half outside directors. We demonstrate the changes in the governance structure of Baosteel arisen from the board reform in Appendix Ⅱ. As of 2015, 87 CSOEs (business group based) have established their boards of directors over the past decade. 2.2 Board reform at the parent company and listed subsidiaries crash risk Although some scholars argue that the new board mechanism has no effect and many 7

8 governance problems have not yet been resolved 1, it is deemed that no matter what, at least theoretically, the board reform should result in some improvement in terms of corporate governance for the CSOE s parent company. For a subsidiary firm, the implication could be three-fold. First, the new mechanism of board of directors may constrain the parent company s managers in colluding with those at its subsidiary firm in order to misappropriate the subsidiary s interest such as assets. Some prior studies have empirically shown that the state-owned parent companies tend to extract wealth from their subsidiary firms listed in the capital market through ways such as earnings management (Liu and Lu 2007; Peng et al., 2011), related party transactions (Cheung et al., 2009; Jian and Wong, 2011), intercorporate loans (Jiang et al., 2010), and related party loan guarantees (Berkman et al., 2009). Jiang and Kim (2015) state that the expropriation in the SOEs occurs partly because managers in the parent company pursue private benefits at the costs of subsidiary firms. Such misappropriations are the sources that a subsidiary firm s negative information emanates from (Habib, et al., 2017). An enhanced governance of the parent company thanks to the board reform can to some extent constrain managers opportunistic behaviors, thereby resulting in lower crash risk of the subsidiary firm. Second, a subsidiary firm will receive more effective monitoring from its parent company after the board reform. According to the agency theoretical framework proposed by Jin and Myers (2006), the information asymmetries existing within a CSOE group allows managers at the subsidiary firm to withhold bad news for an extended period for the consideration of things such as maximizing compensation, protecting employment, 1 In 2014, as disclosed by the National Audit Office of China, eleven Chinese CSOEs have big issues such as misjudging on investment and getting huge losses, distributing company benefit to employees, misdeeds in using funds, etc.; five of the eleven groups have introduced the mechanism of outside board members. 8

9 minimizing litigation concerns that are stemmed from negative information release (Kothari et al., 2009). When the hoarding negative information is revealed at once to the public, the firm s stock price might drop sharply and lead to the edge of crash (Habib et al., 2017). In our research setting, the information asymmetries can also exist between the parent company and its subsidiary firm within a CSOE group. Hence, a better corporate governance of the parent company might be able to influence the subsidiary firm s crash risk through more effective monitoring. Finally, the subsidiary firm s governance may improve as well after the implementation of board reform at its parent company. This is highly possible because there exists a spillover effect in terms of governance improvement from the parent company to the subsidiary firm. In this circumstance, managers at the subsidiary firm would be less likely to do things that can harm shareholders interest while hiding the negative information, which could reduce crash risk of the firm. Collectively, the board reform taking place at the parent company of a CSOE group helps improve the corporate governance of both the parent company and its subsidiaries, which eventually leads to lower crash risk of the subsidiaries. We thus postulate the following hypothesis. H1. A firm whose parent company enacted board reform achieves lower crash risk compared to a matched company. 3. Empirical design In this section, we first introduce the approaches for measuring firm-specific crash risk, and then we describe the data sample and variables used in our study. In the final part we present the empirical strategy in testing our hypothesis. 3.1 Measuring firm-specific crash risk Existing literature provides four methods for measuring firm-specific crash risk 9

10 (Habib et al., 2017). To ensure that the measure reflects firm-specific factors of crash risk rather than market-level movement, each of the four methods requires firm-specific weekly return estimated as the residuals from the market model (Chen et al., 2001), which can be mathematically expressed with the following form: R i, t i 1Rmt, 2 2Rmt, 1 3Rmt, 4Rmt, 1 5Rmt, 2 i, t (1), where Ri,t is the return that takes into account the yields on cash dividend reinvestment of firm i in week t and Rm,t is the value-weighted market return in week t. This estimation model includes both lead and lag terms of the market return so that non-synchronous transactions are not an issue (Dimson, 1979). The firm-specific weekly return for firm i in week t is calculated as the natural logarithm of one plus the residual return (i.e., W i, t ( i, t Ln 1 ), where is the residual in Eq. (1)). Based on the firm-specific weekly i, t return calculated above, the first method for measuring crash risk is to take a binary value -- i.e., coding one if a firm s weekly returns falls at least 3.09 standard deviation below its mean value in a given year, and zero otherwise (Hutton et al., 2009). It is worth to note that the threshold of 3.09 standard deviation is set in order to cut off the 0.1% of the normal distribution in terms of the firm s weekly returns. Any weekly return of a firm dropping to the interval of the 0.1% tail is deemed to be a substantial drop, a so-called stock price crash. The second method initially proposed by Chen et al. (2001) is based upon skewness of firm-specific weekly returns distribution. Specifically, it considers the left-side skewness by taking the third moment of firm-specific weekly returns for a given year and normalizing it by the standard deviation of the returns raised to the third power. This measure is frequently adopted by scholars (Habib, et al., 2017). Denote NCSKEW as the measure of a firm s crash risk, then Eq. (2) below offers a mathematical form of this approach. 10

11 i, t i, t 3/ / 2 NCSKEW [ n( n 1) W ]/[( n 1)( n 2)( W ) ] (2), where n is the number of transaction weeks of firm i's stock in year t. A higher absolute value of NCSKEW means a higher skewness coefficient, corresponding to a greater stock crash risk. The third method for measuring crash risk takes into account the crash likelihood using a down-to-up volatility measure (Chen et al., 2001). Denote DUVOL as the crash risk, then it can be calculated as shown in Eq. (3): i, t u i, t d down up 2 2 DUVOL log{[ n 1 W ]/[( n 1 W )]} (3), i, t i, t where nu and nd are the number of up and down weeks in year t, respectively. A higher value of DUVOL indicates a greater crash risk. Since the down-to-up volatility measure does not involve the third moment, it is less likely to be overly impacted by extreme weekly returns (Chen et al., 2001). The last measure of firm-specific crash risk is called implied volatility smirk, which is introduced by Kim and Zhang (2014). The implied volatility is the volatility implied by the option s market price based upon an option pricing model such as Black-Scholes model. The implied volatility smirk is defined as the difference between the implied volatility of out-of-money (OTM) puts and that of at-the-money (ATM) calls, which is calculated as follows: IV-SKEW = IV OTMP - IV ATMC (4), where OTM puts are put options with a delta value ranging from to , and ATM calls are call options with a delta value ranging between and The daily IV-SKEWs computed following Eq. (4) are averaged throughout a twelve-month period from the fourth month of each fiscal year. Despite no justification is seen in extant literature of which one of the four methods is the best, both NSKEW and DUVOL are frequently used in prior studies (e.g., Kim et al., 11

12 2011a, b; Xu et al., 2013). Hence, we adopt these two measures for estimating firm-specific crash risk in our investigation. We do this way also based two other considerations. First, there is no real OTC market in Chinese stock markets; therefore it is not feasible to compute data of firm-specific crash risk using the forth measure. Second, given the feature that Chinese government intervenes its stock market frequently, the first measure may not be able to well capture firm-specific crash risk. 3.2 Model We adopt the following model to test our hypothesis: CRASHRISKi,t = 0 + t + i + 1REFORMi,t + 2CRASHRISKi,t-1 + 3SIZEi,t-1 + 4LEVi,t-1 + 5ROAi,t-1 + 6RETi,t-1 + 7SIGMAi,t-1 + 8DTURNOVERi,t-1 + 9BMi,t OPAQUEi,t-1 + i,t (5), where i indexes a listed subsidiary firm and t indexes a year; t and i are year and firm fixed effects, respectively. CRASHRISK measures the crash risk in the listed subsidiary firms, which is introduced in Section 3.1. REFORM equals 1 if the parent company affiliated to the listed firm has implemented the board reform by time t, and 0 otherwise. Notably, model (5) is a generalized DID model since it accounts for the many board reforms staggered in the parent companies over time (Bertrand and Mullainathan, 2003). Our treatment group includes the listed firms whose parent companies have implemented board reform by year t. The control group is not restricted to the parent companies that never conducted the reform, but incorporates any firms whose parent companies have not yet conducted the reform by year t, even if they implement the reform in the future. Following prior literature of crash risk (e.g., Chen et al., 2001; Kim et al., 2011a, b), we control for several firm characteristics, such as firm size (SIZE), leverage (LEV), profitability (ROA) and book-to-market ratio (BM). SIZE is calculated as the natural logarithm of the book value of a firm s total assets. LEV is calculated as a ratio of total 12

13 debt to total assets. ROA is return on total assets. BM is measured as the book value of a firm s total assets divided by the market value of equity plus book value of debt. In addition, we include several factors that may influence the crash risk likelihood, such as stock volatility (SIGMA), stock return (RET), de-trended share turnover (DTURNOVER). SIGMA is defined as the standard deviation of the firm specific weekly return over the fiscal year. Stocks with higher volatility are more likely to experience crashes in the future. RET is the mean value of the firm specific weekly return over the fiscal year. DTURNOVER is calculated as the average monthly share turnover in the current year minus the average monthly share turnover in the prior year. Besides, we also control for the opaqueness of the firm information (OPAQUE), which is measured as the moving average of the three-year absolute values of annual performance-adjusted discretionary accruals. Hutton et al. (2009) show that firms with earnings management are more likely to suffer from a future crash. Finally, we include the lag value of crash risk to control for the serial correlation of crash risks. All of the control variables are lagged by one year. All the variables are described in AppendixⅠ. 3.3 Sample description Chinese listed companies started to disclose detailed information of shareholder structure and ultimate controlling shareholders from So we set the sample period as By combining the information of shareholder structure and ultimate parent controlling companies, as well as a name list of the 112 CSOEs released by SASAC ( we manually match the parent-subsidiary relationship between a CSOE and a state-owned listed company. Through this process, we identified that there are 87 CSOEs who have finished board reform, and some of them have affiliated subsidiary firms that are publicly traded in the Chinese stock markets (either Shanghai or Shenzhen Stock Exchange). By eliminating those with missing data needed 13

14 for our model, in our sample we include 62 groups controlling 354 listed firms. The information about board reform launched for each CSOE is released on SACSAC s official website or the official website of each CSOE. We hand-collect the data from these websites about the reform year and the number of outside directors that were selected into the newly-formed boards. The other related accounting and financial data used in this research is downloaded from the China Security Markets and Accounting Research Database (CSMAR). There are 3128 firm-year observations fulfilling our criteria. By deleting the observations with missing values, we have 2050 firm-year observations in total at the end. Table 1 summarizes some descriptive statistics for our sample. The means of the two dependent variables of our interest, NCSKEWt and DUVOLt, are and respectively, and their standard deviations are and 0.477, which are quite large relatively. This implies that the sample firms crash risk coefficients are dispersing around the mean value in a wide range. The average of REFORMt is 0.313, indicating 31.3% of the sample firms have been influenced by their parent companies board reform. It is worth mentioning that overall both the mean and median values of the control variables included in our model are close, appearing to us that there is not a serious outlier issue. For instance, the average and median values of firm size in natural logarithm is (e = 5,678,767,178 Yuan in RMB originally) and (4,291,919,905 Yuan in RMB originally) respectively, while the standard deviation is 1.529, suggesting the sizes of the sample companies are close after the logarithm transformation. In short, the value ranges of the variables lie within reasonable intervals for using DID models. For any listed firm belonging to a CSOE, we treat the year in which the CSOE launches its board reform (reform launch year hereafter) as the time point of exogenous shock. Note that different focal firms have different reform launch years. 14

15 [Insert Table 1 Here] 4. Empirical results In this section, we present our major finding that answers our research question. We then conduct several tests to examine the robustness of our major finding. In the final part of this section, we provide mechanism analyses where the channels are discussed. 4.1 Major finding Table 2 reports the results of whether the board reform in the parent company reduces the crash risk in listed subsidiary firms. As shown in the results of the basic model, Columns (1) and (4), the coefficient of REFORM when using NCSKEW to measure firm-specific crash risk is statistically significant at the 10% level (coefficient = , t-statistics = -1.73), while the coefficient of REFORM when using DUVOL is statistically significant at the 1% level (coefficient = , t-statistics = -2.91). Economically, the results in both Columns (1) and (4) suggests that the reduced crash risk of the listed firm can be attributed to the board reform undertaken at its ultimate parent company. [Insert Table 2 Here] In the models of Columns (2) and (5), we control for the related characteristics of the listed firm as well as a lead term of the dependent variable when regressing either NCSKEW or DUVOL against REFORM. As presented, while the coefficient of REFORM when using DUVOL as dependent variable (Column (5)) is statistically significant at the 1% level (coefficient = , t-statistics = -3.20), which is the same as the one in the basic model; the significance level of the coefficient of REFORM when using NCSKEW as response variable (Column (2)) has increased to 5% (coefficient = , t-statistics = 15

16 -2.10); moreover, the economic magnitudes of both coefficients in Columns (2) and (5) are greater than those in Columns (1) and (4). We further consider two variables of external corporate governance, the audit quality (BIG4) and analyst coverage (ANALYST), as many papers have documented that these two factors may affect firm s crash risk. For example, Xu et al. (2013) find that an increase in a firm's analyst coverage leads to an increase in stock price crash risk due to the analyst optimism. Callen et al. (2017) find that firms with higher audit quality have lower future crash risk. BIG4 equals 1 if the listed firm is audited by a BIG 4 auditor, and 0 otherwise. ANALYST is measured as the natural logarithm of 1 plus the number of analyst coverage. The coefficients of REFORM demonstrated in Columns (3) and (6) are similar to those in Columns (2) and (5). One thing noteworthy is that the values of F statistics for the models throughout Columns (1) (6) are quite large, suggesting that the null hypothesis of our main hypothesis is jointly statistically significant (Wooldridge, 2013, pp ). To summarize, the above analyses suggest that overall the treatment effect of the parent company s board reform (REFORM) on the subsidiary firm s crash risk is statistically significant and negative, which supports our main hypothesis. In other words, the board reform taking place within the ultimate parent company helps reduce the stock price crash risk of its subsidiary firms publicly traded in the capital market. 4.2 Robustness tests In this section, we perform several tests to check the robustness of our findings Reverse causality check Although we find that the board reform taking place at the ultimate parent company reduces its subsidiary firm s crash risk, it is possible that the implementation of the board reform is driven by a serious issue of crash risk observed prior to the board reform. To 16

17 discern the effect of reverse causation, we therefore conduct a test by following Bertrand and Mullainathan (2003) and Giroud and Muller (2010). In specific, we regress either NCSKEW or DUVOL against four REFORM-related dummy variables and control for the necessary variables that are the same as those included in models (2) and (5) of Table 3. The four dummies are: REFORM(-1) which equals 1 when a listed firm s ultimate parent company will implement board reform one year from now, or 0 otherwise; REFORM(0) which equals 1 when a firm s ultimate parent company implements board reform this year, or 0 otherwise; REFORM(+1) which is equivalent to 1 if a firm s ultimate parent company implemented board reform one year ago, or 0 otherwise; and REFORM(2+) which equals 1 if the ultimate parent company implemented board reform two or two more years ago, or 0 otherwise. Theoretically, if there exists a reverse causation effect between the ultimate parent company s board reform and its subsidiary s crash risk, then the coefficient of the lead term REFORM(-1) tends to be statistically negative. However, if it is the board reform undertaken at the parent company that reduces the crash risk in listed subsidiary firms, then the coefficients of lag terms (i.e., REFORM(+1) and REFORM(2+)) tend to be statistically negative whereas the coefficient of REFORM (-1) should be insignificant negative. Table 3 presents the results of the reverse causality test. In Column (1) where we use NCSKEW as a proxy of crash risk, the coefficient of REFORM(-1) is significant but positive, which is contrary to the prediction of reverse causality effect. In addition, the coefficients of REFORM(+1) and REFORM(2+) become negative although insignificant after the board reform in the parent company. Column (2) illustrates the results when using DUVOL as a proxy of crash risk. The coefficient of REFORM(-1) is positive and insignificant; however, the coefficient of REFORM(+1) is negative and significant (coefficient= , t-statistic= -2.15), and the coefficient of REFORM(2+) is also negative but it has become statistically insignificant. Therefore, the results in Table 3 do 17

18 not support the reverse causality hypothesis. These results also suggest that although the board reform may reduce the crash risk of the listed subsidiary firms, the effect exists only in the next year of the reform, and then disappears as time goes by. [Insert Table 3 Here] The degree of board reform and crash risk To be further, we look into how varying degrees of board reform occurring at an ultimate parent would affect its subsidiaries crash risks. In specific, we replace the dummy explanatory variable REFORM with a ratio-typed variable namely REFORM_DEGREE based on the models in Columns (2) and (5) of Table 4, where REFORM_DEGREE is defined as the ratio of outside directors sitting on the board of the parent company where a listed subsidiary firm belongs at the time of board reform. According to the empirical result corresponding to our main hypothesis, we surmise that, ceteris paribus, a higher degree of board reform being implemented at the parent company would lead to a lower level of crash risk for the subsidiary firm. We find that, indeed, the regression results in Table 4 are consistent with our surmise. Specifically, both coefficients of REFORM_DEGREE corresponding to NCSKEW and DUVOL are negative and statistically significant at the levels of 5% and 1%, respectively. [Insert Table 4 Here] 4.3 Mechanism analyses In this section, we analyze several channels through which a parent company s board reform impacts the crash risk of its subsidiary firms The effect of less agency problem from the parent company 18

19 First of all, due to the higher presentation of outside directors in the board therefore the strengthened board monitoring in the parent company, the board reform might somehow alleviate the agency problem from the parent company that can harm the subsidiary firm s interest. In practice, the managers at both the parent and subsidiary companies may collude to do something that can negatively impact the subsidiary firm s performance. If this is true, it is predicted that for those listed firms with more serious agency problems arisen from parent companies, the board reform should help reduce more on their crash risk. To examine this effect, we use related party transactions as a proxy to measure this type of agency problem following Cheung et al. (2009) and Jian and Wong (2011). We rank the sample companies according to the amount of related party transactions between parent and subsidiary companies. A dummy variable namely H_TUNNEL is defined to label the sample companies that are above the median (H_TUNNEL = 1) or below the median (H_TUNNEL = 0) in terms of the amount of related party transactions. Thus, we are able to separate the sample into two groups: one (the other) with more (less) serious agency problem arisen from parent companies. We rerun the model of Eq. (5) separately using the two subsamples. As we can see from the results exhibited in Table 5, for the former group, the coefficients of REFORM in both regressions using either NCSKEW or DUVOL are statistically significant, whereas for the latter, both coefficients are statistically insignificant. The results are exactly consistent with our prediction. Moreover, for the former group, the economic magnitudes of the both coefficients (i.e., and ) are larger than those using the full sample (i.e., and in Table 2). That is also correct by intuition. [Insert Table 5 Here] 19

20 4.3.2 Monitoring effect from parent companies It is noted an ultimate parent company may influence its subsidiaries crash risk through the way of monitoring. More effective monitoring from ultimate parent company can reduce the information asymmetry level between the ultimate parent company and its subsidiary firm; as such, it becomes more difficult for managers of the subsidiary firm to hide bad news (Jin and Myers, 2006), which may ultimately decrease the crash risk of the subsidiary firm. The channel of this type of monitoring can be stronger for those listed companies whose board of directors has more members coming from their ultimate parent companies. To identify this monitoring effect, we use the median of the ratio of the listed subsidiary firm s board directors coming from parent companies 2 to divide the sample into two groups: one is those who have more board directors coming from parent companies (H_MONITOR = 1), and the other is those who have less board directors coming from parent companies (H_MONITOR = 0). By running the same models using the two groups of data sample separately, we obtain the results summarized in Table 6. As reported, for the group with more board directors coming from parent companies, the coefficients of REFORM in Columns (1) and (3), which correspond to both the measures of crash risk as dependent variables, are statistically significant; whereas the coefficients of REFORM (i.e., in Columns (2) and (4)) in the group of less board members coming from parent companies are not statistically significant. Intuitively, the results appear reasonable. It is because the group with more board members from parent companies intrinsically has a bigger channel that passes parent companies monitoring effect to their subsidiary firms. Consequently, the subsidiary firms crash risk are reduced to a more extent. In fact, the effect size for the group with more board members from parent 2 As an alternative, we use a dummy of whether the parent company's president or CEO sits on the board of the focal firm. The result is similar. 20

21 companies in either Column (1) (i.e., ) or (3) (i.e., ) is greater than that in Column (2) (i.e., ) or (5) (i.e., ) of Table 2; whereas for the group with less board directors from parent companies, the comparison result goes to the opposite. [Insert Table 6 Here] For some listed companies, if they have already received strict monitoring from other shareholders, then the monitoring effect from parent companies in reducing the listed company s crash risk should attenuate. We conduct a test to examine this scenario. Specifically, we design a dummy variable namely H_IO which is equivalent to 1 if the institutional shareholding of the listed subsidiary company at the end of fiscal year t-1 is larger than the sample median, or 0 otherwise. The sample is divided into two groups: one is those with H_IO = 1, and the other is those with H_IO = 0. Separately using the two subsamples, we rerun the two models same as models (2) and (5) in Table 2. The results are presented in Table 7. As we can see, the coefficients of REFORM for the group with low institutional shareholding (H_IO = 0) are statistically significant; moreover, the effect sizes (i.e., and ) are quite larger than those (i.e., and ) in Table 2. In contrast, the coefficients of REFORM for the group with high institutional shareholding (H_IO = 1) is not statistically significant, and the effect sizes are quite smaller than those in Table 2. Overall, the result is exactly consistent with our premise. [Insert Table 7 Here] Less managerial agency costs of firm itself and crash risk As noted previously, due to a spillover effect, the corporate governance of a listed firm may be improved along with the implementation of board reform at its ultimate 21

22 parent company. As a consequence, the listed firm s crash risk decreases in part because of the enhanced governance of the firm itself. To test this channel, we use managerial agency costs (H_M&A) that has been well established by existing literature (e.g., Ang et al., 2000) as a proxy of the listed firm s corporate governance. According to our definition, the binary variable H_M&A is coded as 1 if the managerial administrative cost 3 of the listed subsidiary company at the end of fiscal year t-1 is higher than the sample median, and 0 otherwise. We thus divide the sample into two groups: a group of firms with high H_M&A and the other group with low H_M&A. We surmise that the group with high H_M&A is impacted more significantly by the board reforms undertaken at parent companies. The results are presented in Table 8. As we can see, for the group of high H_M&A, regardless of using either measure of crash risk as dependent variable, the coefficient of REFORM is negative and statistically significant. The significance levels for NCSKEW and DUVOL are 5% and 1% respectively. Yet for the group of low H_M&A, the coefficients of REFORM (in Columns (2) and (4)) is statistically insignificant, and both coefficient magnitudes are lower compared to those for the group of high H_M&A. [Insert Table 8 Here] From our empirical results, it appears to us there exists a channel through which the board reform implemented at an ultimate parent company reduces the crash risk of its subsidiaries. That is, the board reform improves the subsidiaries corporate governance (proxied by managerial agency costs), which puts more strict constraints on the managers opportunistic behaviors that can lead to extreme bad outcomes for the firm. The above mechanism eventually results in reduced crash risks of the subsidiaries. 3 As an alternative, we use administration & selling costs for performing this test. The result is similar. 22

23 5. Conclusion Stock price crash risk has been a hot topic in the finance literature. It is well documented that managers misbehaviors in dealing with negative information leads to the future crash risk of their company. However, current literature mainly focuses on how the corporate governance in the listed firm itself rather than the governance in the parent company influences the listed firm s crash risk. In this article we investigate the impact that a strengthened board monitoring from the parent company has on its subsidiary firm s (i.e., focal firm) stock price crash risk using a quasi-natural experiment in China. We show that the focal firm s crash risk decreases thanks to the strengthening of board monitoring from the focal firm s parent company. In specific, the magnitude of this effect reaches the peak about one year after the reform, and then attenuates afterwards; also, the magnitude of this effect is bigger for those firms who have more board members on behalf of the parent company (i.e., degree of board reform). We further find that the positive impact of the board reform may be channeled in several passages. First, the enhanced corporate governance of the parent company can somehow prevent it from tunneling wealth from the subsidiary firm. Moreover, a better corporate governance of the parent company helps it better monitor the subsidiary firm. Last, the governance improvement of the board reform occurring at the parent company can spill over to the subsidiary firm, which helps constrain managers at the subsidiary firm on misconducts of dealing with negative information. Our research makes the first attempt to investigate whether and how a parent company s corporate governance affects the stock price crash risks of its listed subsidiary firms. We document that a firm s stock price crash risk can be reduced by the improvement in corporate governance of the firm s parent company or by the monitoring strengthening of the parent company. 23

24 AppendixⅠ Variable definition Variables Definition NCSKEW t = the negative skewness of firm-specific weekly return over fiscal year t. DUVOL t = the natural logarithm of the ratio of the standard deviations of down-week to up-week firm-specific weekly returns over the fiscal year t. REFORM t = 1 if the ultimate parent company has implemented the board reform at the end of fiscal year t, or 0 otherwise. REFORM_DEGREE t = the ratio of outside directors on the board of the parent company that has implemented the board reform at the end of fiscal year t. REFORM(-1) = 1 when a listed firm s ultimate parent company will implement board reform one year from now, or 0 otherwise. REFORM(0) = 1 when a firm s ultimate parent company implements board reform this year, or 0 otherwise. REFORM(+1) = 1 if a firm s ultimate parent company implemented the board reform one year ago, or 0 otherwise. REFORM(2+) = 1 if the ultimate parent company implemented board reform two or two more years ago, or 0 otherwise. NCSKEW t-1 = the value of NCSKEW at year t-1. DUVOL t-1 = the value of DUVOL at year t-1. SIZE t-1 = the natural logarithm of total assets at the end of fiscal year t-1. LEV t-1 ROA t-1 RET t-1 SIGMA t-1 DTURNOVER t-1 BM t-1 = the total debt divided by total assets at the end of fiscal year t-1. = the net income divided by lagged total assets at the end of fiscal year t-1. = the mean value of the firm-specific weekly return over the fiscal year t-1. = the standard deviation of firm-specific weekly returns over the fiscal year t-1. = the average monthly share turnover during fiscal year t-1 minus the average monthly share turnover during fiscal year t-2, where monthly share turnover is calculated as the monthly trading volume divided by total number of shares outstanding during the month. = the book value of total assets divided by the sum of market value of equity and book value of debt at the end of fiscal year t-1. 24

25 AppendixⅠ(cont.) OPAQUE t-1 = the mean value of the absolute value of discretionary accruals over the prior three years, where discretionary accruals are estimated from the Modified Jones Model adjusted for performance (Kothari et al., 2005). BIG4 t-1 = 1 if the listed firm is audited by the BIG 4 auditors, or 0 otherwise. ANALYST t-1 = the natural logarithm of 1 plus the number of analyst coverage. H_M&A = 1 if the managerial administrative cost of the listed subsidiary company at the end of fiscal year t-1 is larger than the sample median, or 0 otherwise. H_MONITOR = 1 if the ratio of directors assigned by the ultimate parent company for the listed subsidiary company is larger than the sample median, or 0 otherwise. H_TUNNEL = 1 if TUNNEL is larger than the sample median, or 0 otherwise. TUNNEL is measured as the other receivables of the listed subsidiary company due from the ultimate parent company and its affiliated companies divided by total assets at the end of fiscal year t-1. H_IO = 1 if the institutional shareholding of the listed subsidiary company at the end of fiscal year t-1 is larger than the sample median, or 0 otherwise. 25

26 AppendixⅡ. The setting up of board of directors in the China BaoWu Steel Group, 2005 Before 2005 After 2005 Board of directors Top executives Top executives China Baosteel Group (Ultimate Parent) 57.22% 77.86% Baoshan Iron and Steel Co., Ltd 100% Shanghai Baogang Engineering Technology Limited Company and Other 11 non listed subsidiaries Shanghai Baosight Software Co., Ltd Other Subsidiaries Figure 1. The Pyramidal Structure of Baosteel Group and the Board Reform in 2005 Figure 1 describes the pyramidal structure of the Baostell business group, and the board reform in Specifically, the Baosteel group includes three important companies in our study, that is, the China Baogang Steel Group, Baoshan Iron and Stell Co. Ltd and Shanghai Baosight Software Co. Ltd. Among these, China Baosteel Group is the ultimate parent that controls two listed subsidiary firms: Baoshan Iron and Stell Co. Ltd and Shanghai Baosight Software Co. Ltd, where the former stands in the layer 3 in the pyramid and the latter is in layer 4. Before October 2005, the ultimate parent (China Baosteel Group) did not establish the board, and the management as the insider generally controlled the group. In this case, the management in the group might have great incentive to tunnel the firm resources for private benefits. After October 2005, Baosteel Group set up the board and hired five outside directors, including a chairman and an executive of two foreign-based corporations, two former leaders in two central SOEs, and an accounting professor from a Chinese university. At the same time, four committees were established under the board of directors, namely a standing committee, a nomination committee, a compensation and evaluation committee, and internal auditing committee. 26

27 References Ang, J.S., Cole, R.A., and Lin, J.W. (2000) Agency costs and ownership structure. Journal of Finance, 55(1): Andreou, P.C., Antoniou, C., Horton, J., Louca, C. (2016). Corporate governance and firm-specific stock price crashes, European Financial Management, 22, Berkman, H., Cole, R. A., Fu, L. J. (2009). Expropriation through loan guarantees to related parties: Evidence from China. Journal of Banking & Finance, 33(1), Bertrand, M., Mullainathan, S. (2003). Enjoying the quiet life? Corporate governance and managerial preferences. Journal of Political Economy, 111, Boubaker, S., Mansali, H., Rjiba, H. (2014). Large controlling shareholders and stock price synchronicity. Journal of Banking & Finance, 40, Callen, J.L., Fang, X. (2015). Short interest and stock price crash risk. Journal of Financial and Quantitative Analysis. 50, Callen, J.L., Fang, X. (2017). Crash risk and the auditor-client relationship. Contemporary Accounting Research, 34(3), Chen, J., Hong, H., Stein, J.C. (2001). Forecasting crashes, trading volume, past returns, and conditional skewness in stock prices. Journal of Financial Economics, 61, Cheung, Y.L., Jing, L., Lu, T., Rau, P. R., Stouraitis, A. (2009). Tunneling and propping up: An analysis of related party transactions by Chinese listed companies, Pacific-Basin Finance Journal, 17(3), Dimson, E. (1979). Risk measurement when shares are subject to infrequent trading. Journal of Financial Ecomonics, 72, Giroud, X., Mueller, H.M. (2010). Does corporate governance matter in competitive industries?. Journal of Financial Economics, 95(3), Habib, A., Hasan, M.M., Jiang, H. (2017). Stock price crash risk: Review of the empirical literature. Accounting and Finance (forthcoming). 27