How Does Investor Protection Influence the Usefulness of Cash Flow Information in CEO Turnover Decisions? International Evidence

Transcription

1 How Does Investor Protection Influence the Usefulness of Cash Flow Information in CEO Turnover Decisions? International Evidence Jinshuai Hu Institute for Financial and Accounting Studies Xiamen University, China Jeong-Bon Kim City University of Hong Kong, Hong Kong Lin Wang Institute for Financial and Accounting Studies Xiamen University, China Abstract This study investigates how investor protection affects the usefulness of cash flow information in CEO turnover decisions. We contend that in an environment of weak investor protection, where cash flows are better able to capture the underlying firm performance than accrual-based earnings and stock price, shareholders and corporate boards rely more on cash flow information to assess CEO performance and replace poorly performing CEOs. With a large sample of listed firms from 41 countries, our results reveal that: (1) the sensitivity of CEO turnover to cash flow (accruals and stock returns) performance decreases (increases) with the strength of investor protection; and (2) analysts cash flow forecasts enhance the sensitivity of CEO turnover to both cash flow and market performance, and such an effect of analyst following is more pronounced for firms in countries with weaker investor protection. Our findings are consistent with the view that cash flow information plays a more important role in facilitating efficient contracting when shareholders rights are poorly protected. February 2014 We thank Agnes C.S. Cheng, In-Mu Haw, Bingbing Hu, Siqi Li, Stephen H. Penman, and Xinrong Qiang for their comments and suggestions. We have also received useful comments from seminar participants at City University of Hong Kong, Fudan University, and Xiamen University. The second author acknowledges partial financial support from the GRF grant of the Hong Kong SAR government (Project No.: ). The usual disclaimer applies.

2 How Does Investor Protection Influence the Usefulness of Cash Flow Information in CEO Turnover Decisions? International Evidence Abstract This study investigates how investor protection affects the usefulness of cash flow information in CEO turnover decisions. We contend that in an environment of weak investor protection, where cash flows are better able to capture the underlying firm performance than accrual-based earnings and stock price, shareholders and corporate boards rely more on cash flow information to assess CEO performance and replace poorly performing CEOs. With a large sample of listed firms from 41 countries, our results reveal that: (1) the sensitivity of CEO turnover to cash flow (accruals and stock returns) performance decreases (increases) with the strength of investor protection; and (2) analysts cash flow forecasts enhance the sensitivity of CEO turnover to both cash flow and market performance, and such an effect of analyst following is more pronounced for firms in countries with weaker investor protection. Our findings are consistent with the view that cash flow information plays a more important role in facilitating efficient contracting when shareholders rights are poorly protected. Keywords: cash flow; investor protection; CEO turnover; analyst forecast JEL classification: M41; G14; G15

3 How Does Investor Protection Influence the Usefulness of Cash Flow Information in CEO Turnover Decisions? International Evidence 1. Introduction A strand of recent research shows that over the last two decades, cash flow information in US firms has become increasingly important and valuable to outside investors for both contracting and valuation purposes (e.g., Nwaeze et al. 2006; Wasley and Wu 2006; Banker et al. 2009; Lehavy 2009; Balsam et al. 2011). 1 Another strand of research, using a sample of international firms from around the world, provides evidence suggesting that market participants have a tendency to rely more on cash flows than on accrual earnings in countries with weaker investor protection, where accrual earnings and market price are less likely to capture the underlying firm performance (Bartov et al. 2001; Hung 2001; DeFond and Hung 2007; Miranda-Lopez and Nichols 2012). Given the above background, this study aims to provide systematic, large-sample evidence on two important, but yet unexplored, questions: (i) whether shareholders and/or corporate boards use cash flow information, in addition to earnings and market information, when evaluating chief executive officers (CEOs ) performance and terminating poorly performing 1 For example, Banker et al. (2009) document that while both value relevance and compensation weight on earnings decline from the period to , both value relevance and compensation weight on cash flows increase from the earlier period to the later period. Similarly, Balsam, Yang, and Yin (2011) show that the executive compensation of US-listed firms becomes more (less) sensitive to the cash flow (accruals) component of earnings over the period Another example illustrating the increasing use of cash flow information is the significant increase in cash flow forecasts (both analyst and management forecasts) over the last couple of decades (e.g., DeFond and Hung 2003; Wasley and Wu 2006; Call 2008; Givoly et al. 2009). 1

4 CEOs; and (ii) how the strength of a country s investor protection institutions affects the role that cash flow information plays in CEO replacement decisions. 2 The legal protection of investor rights plays two different, albeit related, roles in determining the usefulness of information about firm performance for assessing CEO performance and making CEO replacement decisions. On the one hand, well-functioning legal institutions foster effective governance systems (e.g., La Porta et al. 1997, 2000; Bushman and Smith 2001; DeFond and Hung 2004; Nagar 2004), and consequently enhance the ability of shareholders and/or the board to evaluate CEO performance and terminate poorly performing CEOs (DeFond and Hung 2004; Lel and Miller 2008). 3 Under this view, information about firm performance is more useful for CEO turnover decisions in countries with strong investor protection than in countries with weak protection, because strong investor protection institutions constrain managerial opportunism and facilitate the efficacy of corporate governance. On the other hand, investor protection is also a crucial factor determining the ability of a 2 As pointed out by Engel, Hayes, and Wang (2003), the setting of CEO turnover may yield greater insights than that of executive compensation into how information is used in corporate boardrooms. First, executive compensation contracts are often explicitly set, ex ante. Once the contracts are fixed, the boards may lose the authority to choose information to evaluate executive performance, ex post. On the contrary, in considering retention/termination decisions, directors may of course make use of accrual earnings, cash flows and/or market performance measures, but the directors themselves must make the decision, ex post, about retaining/replacing the CEO. Second, stock prices reflect the market s expectations regarding CEO turnover decisions. This effect can make the disaggregated accounting information (including accruals and cash flows) more important in inferring a CEO s specific actions for the termination decision (Bond et al. 2010; Bushman and Smith 2001; Armstrong et al. 2010). 3 For example, DeFond and Hung (2004) document that CEO turnover is more closely related to poor performance in countries with strong investor protection. Lel and Miller (2008) find that foreign firms that are cross-listed on the US exchanges from countries with weak investor protection are more likely to terminate poorly performing CEOs than non-cross-listed firms, suggesting that stringent investor protection in the US improves the corporate governance of cross-listed firms. 2

5 firm s accounting system and stock price to capture underlying economic events. Previous research suggests that in countries with stronger investor protection, accounting and price information is more useful for outside stakeholders to evaluate firm fundamentals. This is because, in countries with strong institutional infrastructures, such as stringent legal enforcement and high penalties (including reputation losses) for financial misreporting, accounting reports are better able to reflect the economic consequences of managerial actions in general and CEO performance in particular (e.g., Ball et al. 2000; Bushman and Smith 2001; Leuz et al. 2003; Bushman and Piotroski 2006). Similarly, strong investor protection promotes outside investors to acquire information and trade on the information, and thus enhances the ability of stock price to capture management value-creating/impairing activities (Morck et al., 2000; Fernandes and Ferreira, 2009). Though cash flows are of contracting relevance, 4 the aforementioned research has paid little attention to whether and how the relative importance of cash flows versus other performance information (e.g., accrual earnings and market price) is influenced by the strength of investor protection or related institutional infrastructures from the contracting perspective. 5 Consequently, little is known about the conditions under which a corporate board relies more on 4 A growing body of research indicates that cash flow information has stewardship value (e.g., Natarajan 1996; Nwaeze et al. 2006, 2010; Banker et al. 2009). See Nwaeze et al. (2006) for a brief review of the literature on the contracting relevance of cash flows. Nwaeze et al. (2006) also provide real instances of firms using cash flows as a performance measure in incentive plans. 5 One notable exception is DeFond and Hung (2004), who document that market-based performance is more useful to explain CEO turnover events in countries where investors rights are better protected and/or stock price is more informative. However, their results suggest that the effect of investor protection on the CEO turnover-earnings relation is basically insignificant. Therefore, our study focuses mainly on how investor protection influences the relative usefulness of earnings components, say, cash flows versus accruals, although market-based information is also incorporated in our analyses. 3

6 cash flows when evaluating CEO performance and terminating poorly performing CEOs. To fill this gap, our study investigates whether and how the likelihood of CEO turnover is more sensitive to cash flows than to accounting accruals and market performance, and how this sensitivity is influenced by a country s investor protection environment. Compared with accounting accruals, cash flows are relatively less discretionary and less susceptible to manipulation. Cash flows provide investors with supplementary information that helps them interpret earnings information (Hung 2001; DeFond and Hung 2003, 2007). For example, a growing body of recent research suggests that cash flow information (e.g., analysts and management s cash flow forecasts) enhances outside investors ability to uncover a firm s true underlying performance and to constrain managerial opportunism (DeFond and Hung 2003; Wasley and Wu 2006; McInnis and Collins 2011; Call 2008). To the extent that accrual earnings are of poor quality and stock prices are of low informativeness in a weak investor protection environment, stakeholders (e.g., outside shareholders and corporate board members) are likely to demand cash flow information in addition to accrual-based earnings and market information. The poor quality of accrual-based earnings and low informativeness of stock price information in weak investor protection environments make shareholders and corporate boards rely more on cash flows than on accrual earnings and market information when assessing CEO performance and terminating poorly performing CEOs. We therefore propose and test our first prediction that the likelihood of CEO turnover becomes more sensitive to cash flow performance than to accrual and market performance as the strength of investor protection decreases. We next examine whether and how analysts cash flow forecasting activities affect the 4

7 usefulness of performance information in determining the CEO turnover performance relation. Recent studies document that a growing and economically significant proportion of firms in both the US and around the world are followed by analysts who issue cash flow forecasts (e.g., DeFond and Hung 2003, 2007; Lee 2011; McInnis and Collins 2011). Information intermediaries such as analysts have incentives to make cash flow forecasts when the demand for cash flow information is high (DeFond and Hung 2003, 2007). In line with this notion, DeFond and Hung (2007) provide evidence suggesting that analysts cash flow forecasts play an important role in attenuating the adverse effects of weak investor protection on earnings usefulness. To date, however, little is known about whether analysts cash flow forecasts satisfy investors demand for cash flow information regarding real corporate decisions such as CEO dismissals. To address this unexplored question, we then examine the role of cash flow forecasts in shaping the CEO turnover performance relation. Specifically, we propose and test our second prediction that analysts cash flow forecasting activities enhance the sensitivity of CEO turnover to firm performance, and that such an effect of cash flow forecasts is more pronounced for firms in countries with weaker investor protection institutions. To test our first prediction, we construct a large sample of 114,742 firm-years (26,853 unique firms) from 41 countries over the period of Using a logit model with the likelihood of CEO turnover as the dependent variable, we find that the impact of investor protection on the sensitivity of CEO turnover to cash flows versus accounting accruals varies in an opposite direction as the strength of investor protection increases. Specifically, our results show that in countries with weaker shareholder rights protection, CEO turnover is more sensitive 5

8 to lagged cash flows, while it is less sensitive to lagged accruals and stock returns. The findings are robust to a variety of alternative proxies for investor protection and to controlling for stock price synchronicity and noise in performance measures at the firm level. These findings remain qualitatively unaltered when a country-level measure of earnings quality is additionally controlled for. To test our second prediction, we construct a reduced sample of 72,328 firm-years with data on analyst forecasts to examine the role of analysts cash flow forecasts in shaping the sensitivity of CEO turnovers to firm performance. Briefly, we find that analysts cash flow forecasting activities enhance the sensitivity of CEO turnover to cash flow performance, but this effect is significant only in countries with weak investor protection. We also find that analysts cash flow forecasts improve the usefulness of market-based performance for evaluating CEO performance and terminating poorly performing CEOs. We find, once again, that this effect is restricted to firms in countries with weak investor protection. The above findings are in line with the view that analysts cash flow forecasts supplement potential deficiencies of accrual earnings as a performance measure in countries with weak investor protection. In short, our results, taken together, suggest that analysts cash flow forecasts play an important governance role in facilitating CEO replacement decisions in countries with weak investor protection. This study contributes to the literature in several ways. First, we document an important role of investor protection in determining the usefulness of cash flows and accounting accruals for the contracting purpose. Such a role provides a possible explanation for the mixed effect of investor protection on the usefulness of earnings for CEO turnover decisions documented by the existing 6

9 literature (e.g., DeFond and Hung, 2004). 6 Previous research also shows that cash flows have an incremental value over and beyond accrual earnings under certain conditions for both valuation and stewardship purposes (e.g., Dechow 1994; Natarajan 1996; Xie 2001; Nwaeze et al. 2006; Banker et al. 2009). Our study contributes to these two strands of literature by demonstrating that the legal protection of shareholder rights is a critical factor in determining the contracting usefulness of both cash flows and accruals. Second, to our knowledge, our study is the first to investigate the relative importance of cash flow versus accrual earnings information in the context of an important real decision, that is, a decision to terminate poorly performing CEOs. Third, we find that in an environment of weak investor protection, where earnings and market prices are less likely to reflect the underlying performance, cash flow information plays a more important role in helping the board identify and terminate poorly performing CEOs. Our finding helps us explain why market participants are more likely to demand cash flow information and analysts are more likely to supply cash flows forecasts when shareholder rights are poorly protected (e.g., DeFond and Hung 2003, 2007; Lehavy 2009). 7 6 Using a law enforcement index as their main investor protection proxy, DeFond and Hung (2004) show that investor protection s effect on the sensitivity of CEO turnover to the earnings performance is insignificant (see, for example, Model 3 of Table 6, p. 292; Models 2 through 4 of Table 5, p. 287). As discussed in more detail in Section 4.1, our study suggests that such insignificant results on earnings may be due to the fact that the usefulness of accruals and cash flows, two main components of earnings, varies in an opposite direction with the strength of investor protection. 7 Givoly, Hayn and Lehavy (2009) document that analysts cash flow forecasts are of low quality and appear as a naive extension of their earnings forecasts, which, as noted by Lehavy (2009), raises questions about the validity of the demand hypothesis regarding cash flow forecasts. Our findings, however, suggest that analysts are more likely to provide cash flow forecasts in countries with poor investor protection where accrual earnings are less revealing of the true CEO performance (DeFond and Hung 2007), because outside stakeholders demand and rely on cash flow information to make CEO retention decisions. 7

10 The remainder of this paper is organized as follows. Section 2 reviews related literature and describes empirical predictions. Section 3 explains the sample, data, and research design. Section 4 presents the empirical results. Section 5 explains the results of various robustness checks. The final section concludes. 2. Related Literature and Empirical Predictions 2.1. Investor protection and the sensitivity of CEO turnover to cash flows Institutional infrastructures are an important factor in determining the ability of reported earnings to reflect a firm s true underlying performance. For firms in countries with strong investor protection, accounting earnings can better recognize the economic consequences of managerial actions over time (Ball et al. 2000), because accrual-based earnings provide a better matching of revenues and expenses than cash flows. As such, accounting information is more revealing of underlying managerial performance (Ball and Brown 1968; Dechow 1994; Cheng et al. 1996). However, the discretionary portion of accounting accruals is subject to manipulation by executives for their private gains. Weak investor protection exacerbates managerial opportunism in financial reporting by providing managers with incentives and opportunities to mask a firm s underlying performance via earnings management (Hung 2001; Leuz et al. 2003). For example, Hung (2001) finds evidence suggesting that managers are more likely to manipulate earnings in countries with weaker investor protection institutions, thereby causing accrual-based earnings to be less value-relevant in these countries. Leuz et al. (2003) provide 8

11 evidence that managers in countries with poor investor protection tend to manipulate reported earnings to conceal their rent-seeking activities, suggesting that obfuscation in financial reporting allows CEOs to avoid disciplinary actions against them. In addition, Haw et al. (2004) show that strong statutory protection of minority rights constrains opportunistic earnings management induced by the divergence between the control and ownership rights of controlling shareholders. The legal protection of investors rights has a similar effect on the ability of stock price to capture a firm s value-creating/imparting activities. In line with this view, Morck, Yeung, and Yu (2000) show that strong property investor rights promote informed arbitrage, which capitalizes detailed specific information and enhances stock price informativeness. Fernandes and Ferreira (2009) find that stock price becomes more informative after a country initially enforces its insider trading laws. In short, the above evidence, taken together, indicates that reported earnings are of poor quality and stock prices are of low informativeness in countries with poor investor protection. Cash flows provide information that complements the performance information contained in accrual-based earnings. Cash flows are less subject to manipulation than accounting accruals (Penman 2001; Wild et al. 2001; Levitt 2002). 8 Market participants can use cash flows to interpret earnings information, for example, by comparing cash flows to operating incomes (DeFond and Hung 2003). A large body of prior studies show that cash flows are incrementally 8 We note that cash flows can still be influenced by management s discretion (Comiskey and Mulford 2002; Lee 2011). For example, Lee (2011) finds that managers manipulate cash flows by timing certain transactions such as delaying payments to suppliers or accelerating collections from customers. However, because manipulating cash flows involves altering real business activities (such as deferring or accelerating expenditures), it is likely to be more costly than manipulating accounting accruals. 9

12 useful over and beyond accrual earnings, particularly for firms whose accounting accruals are subject to manipulation (e.g., for firms with long operating cycles) (Bowen et al. 1987; Ali 1994; Dechow 1994). Similarly, DeFond and Hung (2003) provide evidence suggesting that cash flow information is useful for helping investors interpret reported earnings, particularly when earnings information is complex and difficult for outside stakeholders to understand. 9 In line with the view that cash flows are potentially useful for uncovering a firm s true underlying performance, several recent studies indicate that cash flow forecasts by both management and analysts play an important role in constraining earnings management, and thus enhancing earnings quality (Wasley and Wu 2006; Call 2008; McInnis and Collins 2011). In environments with poor investor protection, earnings and stock price are less likely to capture a firm s true underlying performance. In such an environment, shareholders and corporate boards are more likely to demand cash flow information to supplement deficiencies in reported earnings and stock price as measures of CEO performance, and thus tend to rely more on cash flows than on accrual earnings and market information when evaluating CEO performance and terminating poorly performing CEOs. As argued by DeFond and Hung (2007), the supplementarity between cash flows and accrual earnings causes cash flow information to play an important role in attenuating the adverse effects of weak investor protection on earnings usefulness. Their study finds that information intermediaries, such as analysts, are more likely to issue cash flow forecasts to satisfy the high information demand in countries with weak investor 9 For example, for firms with large accruals, relatively more heterogeneous accounting choice, high earnings volatility and high capital intensity, etc. 10

13 protection. Consistently, Miranda-Lopez and Nichols (2012) demonstrate that outside investors in a country with weak investor protection, such as Mexico, are more likely than investors in the U.S. to use cash flows (rather than accrual earnings) as their main source of information when making investment decisions. Contracting theories with multiple performance measures suggest that optimal incentive-compatible contracts should place more weight on performance measures that are more precise and more sensitive to the agent s effort (e.g., Banker and Datar 1989; Holmstrom and Milgrom 1991; Feltham and Xie 1994; Natarajan 1996; Engel, Hayes, and Wang 2003). Cash flows are contracting-relevant in the sense that they are incrementally informative about underlying managerial actions, and thus the use of cash flow information for the contracting purpose facilitates efficient risk sharing between contracting parties (Holmstrom 1979; Gjesdal 1981; Lambert and Larcker 1987; Banker and Datar 1989). Consistent with these analytical predictions, Natarajan (1996) and Kumar et al. (1993) document that the incorporation of cash flows into managerial performance evaluation models enhances the model s ability to explain executive compensation. Nwaeze, Yang, and Yin (2006) find that the relative quality of cash flows compared with that of earnings has a positive (negative) impact on the usefulness of cash flows (earnings) for executive compensation. In the context of CEO turnover, Engel, Hayes, and Wang (2003) provide evidence that the relation between various performance measures and CEO turnover is affected by the properties of performance information (e.g., noise in performance measures and information timeliness). To the extent that accrual earnings are of poor quality and stock prices are of low 11

14 informativeness in a weak investor protection environment, shareholders and corporate boards rely more on cash flows than on accrual earnings and market information when assessing CEO performance and making CEO replacement decisions. Given the trend that cash flow information has been increasingly used in a variety of contexts over the past twenty years (e.g., Nwaeze, Yang, and Yin 2006; Banker et al. 2009; Lee 2011), our analyses focus on the impact of investor protection on the sensitivity of CEO turnover to cash flows. Specifically, we test our first prediction that the likelihood of CEO turnover is more sensitive to cash flows than to accounting accruals and stock performance as the strength of investor protection decreases Analysts cash flow forecasts and the sensitivity of CEO turnover to performance Previous research provides evidence suggesting that analysts issue cash flow forecasts in response to investors demand for cash flow information, particularly when earnings are of poor quality (e.g., DeFond and Hung 2003, 2007; Wasley and Wu 2006; Call 2008; McInnis and Collins 2011). For example, DeFond and Hung (2007) find that financial analysts are more likely to provide cash flow forecasts in countries with weak investor protection. In such countries, as discussed earlier, managers have fewer incentives for truthful reporting, and are more likely to engage in opportunistic earnings management (e.g., Hung 2001; Leuz et al. 2003; Haw et al. 2004), thereby making earnings information less revealing of the true underlying performance. Stated in another way, DeFond and Hung s (2007) finding suggests that cash flow forecasts supplied by information intermediaries, such as analysts, are valuable to outside investors, because the information on cash flow helps to attenuate the adverse effects of poor institutional 12

15 infrastructures on earnings usefulness. Consistently, McInnis and Collins (2011) find that analysts cash flow forecasts have a significant effect on constraining earnings manipulation and improving earnings quality. An important implication from the aforementioned studies, as Lehavy (2009) points out, is that analysts cash flow forecasts are informative for outside stakeholders to evaluate firm performance. In line with this view, prior studies document that earnings quality improves and analyst forecast errors decline following the provision of cash flow forecasts in addition to earnings forecasts (Call et al. 2009; McInnis and Collins 2011). These studies, however, have paid little attention to the role of analysts cash flow forecasts in influencing a firm s real decisions. As a result, little is known about how analysts cash flow forecasts satisfy investors demand for cash flow information in the context of a firm s real decisions such as a decision to terminate poorly performing CEOs. Therefore, it is interesting to examine whether cash flow forecasts affect the usefulness of performance information (e.g., earnings and their components, cash flows and accruals) for CEO turnover decisions, and how such effects vary with the strength of investor protection institutions. Specifically, we test our second prediction that analysts cash flow forecasting activities enhance the sensitivity of CEO turnover to performance, and that such an effect of cash flow forecasts is more pronounced for firms in countries with weaker investor protection. 3. Data and Research Design 3.1. The sample 13

16 We collect CEO data for the period of from the OSIRIS database compiled by Bureau van Dijk Electronic Publishing (BvD). Following DeFond and Hung (2004) and Lel and Miller (2008), in the first step, we use the titles of CEO, chief executive officer, and chief executive to identify the CEO for each firm. For firms without executives bearing any of the aforementioned titles, alternative CEO titles are used in the second step. The title used for each country is listed in Appendix B. To mitigate the problem of CEO misidentification, we use the OSIRIS disks released by BvD at each quarter-end of our sample years to identify the names of the CEOs. For identifying CEO turnover, we delete duplicate CEOs so that only one record is kept for each CEO identified in each year. Firm-year observations are deleted if more than two different individuals are identified as the CEO in the same year. In the case of two CEOs identified in a given year, we retain the year-end CEO if one of the two identified CEOs appears as CEO in the following year, and the other identified CEO is recognized as the prior year-end CEO if that person appears as CEO in the prior year. We assume that such cases represent CEO turnover during the year. All CEO turnover events are identified by comparing the CEOs names from two successive years. The financial data are extracted from Worldscope, and the stock market data are from Datastream. After merging the top executive data with the financial and stock market data and excluding country-years with fewer than 30 observations, we obtain the final sample of 114,742 firm-year observations (26,853 firms) from 41 countries over the period of to test our 14

17 first prediction. 10 We collect analyst forecast data from the I/B/E/S database. Following prior studies (e.g., DeFond and Hung 2007; Lehavy 2009), we use one-year-ahead annual analyst forecast data to calculate the number of analysts that forecast cash flow and/or earnings. To mitigate the potential effect of I/B/E/S coverage bias on our analysis, we restrict our sample to firms with at least one analyst forecast (either earnings or cash flow forecasts). This restriction reduces our observations from 114,742 to 72,328. We use this reduced sample to test our prediction regarding analyst forecasts Model specifications To test our first prediction that CEO turnover is more sensitive to cash flow performance in countries with weaker investor protection, we estimate the following logit model using the full sample: CEOTO t = a 0 + a 1 OCF t-1 + a 2 AC t-1 + a 3 RET t-1 + a 4 OCF t-1 *IP + a 5 AC t-1 *IP + a 6 RET t-1 *IP + a 7 IP + a 8 SIZE t-1 + a 9 LEV t-1 + Fixed Effects+ µ (1) To test the effect of analysts cash flow forecasts on shaping the CEO turnover performance sensitivity, we estimate the following logit model using the reduced sample of firms with analyst 10 Our analysis does not distinguish routine CEO turnovers from forced CEO turnovers and is unable to control for CEO-specific characteristics such as CEO age and tenure because of limited data availability. As pointed out by Hermalin and Weisbach (2003), DeFond and Hung (2004), and Lel and Miller (2008), the non-distinction of routine versus forced CEO turnovers may increase measurement errors in the dependent variable, namely, the CEO turnover indicator. However, the error in the dependent variable is unlikely to distort our inference on the test variables unless it is systematically correlated with the measurement errors of the test variables. 15

18 forecast data: CEOTO t = b 0 + c 1 OCF t-1 + c 2 AC t-1 + c 3 RET t-1 + c 4 OCF t-1 * ANA + c 5 AC t-1 * ANA + c 6 RET t-1 * ANA + c 7 ANA + c 8 SIZE t-1 + c 9 LEV t-1 + Fixed Effects+ µ (2) In the above models, the dependent variable CEOTO t is an indicator of CEO turnover that equals one if a CEO turnover occurs during the event year t, and zero otherwise. OCF is measured as the cash flows from operations divided by lagged total assets. AC is measured as operating income (E) minus cash flow from operation activities (OCF) divided by lagged total assets. Our market performance measure, RET, is calculated by the market-adjusted stock returns over the fiscal year. ANA is one of the two analyst following variables, ANA_CF and ANA_E: The variable, ANA_CF, is the number of analysts who follow the firm and provide one-year-ahead annual cash flow forecasts, and the variable, ANA_E, is the number of analysts who provide one-year-ahead annual earnings (non-cash-flow) forecasts. Firm size, denoted by SIZE, is measured by the natural logarithm of total assets in millions of U.S. dollars at the end of the fiscal year. Financial leverage, LEV, is measured by the ratio of total liabilities to total assets. We also control for year and industry (based on the Fama-French 48 industries classification) fixed effects. All of the explanatory firm-level variables used in our analyses are lagged by one year to avoid the possibility that the performance of a departing CEO overlaps with that of his/her successor. IP is one of our investor protection proxies. Following Mclean, Zhang, and Zhao (2012), among others, our primary proxy for investor 16

19 protection (IP) is a comprehensive investor protection index (PROTECT) developed by La Porta, Lopez-de-Silanes, and Shleifer (2006). For robustness checks, we also employ a variety of alternative investor protection proxies, including the origin of laws (LAWORG), the aggregate anti-self-dealing index (ASDI), the ex post anti-self-dealing index (XPSD), the liability standards index (LITSTD), and the public enforcement index (PUBENF). Appendix A provides the detailed definitions of these proxies. Our first prediction is supported if we observe a 4 > 0 and a 5 < 0, along with a 1 < 0 and a 2 < 0, in Eq. (1). This suggests that as the strength of investor protection increases, the board places less weight on cash flows (a 4 > 0),while it place more weight (a 5 < 0) on accruals, when identifying and terminating poorly performing CEOs. To test our second prediction regarding the effect of analyst followings and the associated role of investor protection, we partition the reduced sample of firms with analyst following data into the below two subsamples: (i) the sample of firms from countries with weak investor protection; and (ii) the sample of firms from countries with strong protection. We then estimate Eq. (3) for each of these two subsamples Descriptive statistics Panel A of Table 1 presents the sample size and CEO turnover rate for each of the 41 countries in our sample. The numbers of firms and firm-years vary widely across countries: the U.S. (7,196 and 33,537, respectively) and Japan (4,268 and 24,260) have the two largest numbers of firms and firm-years, while Sri Lanka (17 and 36, respectively) has the smallest. The CEO turnover rate ranges from 3.27% in Taiwan to 17.08% in Finland, with an average of 17

20 11.56%. Panel B of Table 1 reports the number of firm-years with CEO turnovers and the turnover rate by year. The turnover rate varies from the lowest of 7.95% in 2009 to the highest of 17.41% in 2001, but does not reveal any clear trend over the 10-year test period. [ Insert Table 1 here ] Table 2 shows the descriptive statistics of our research variables. Panel A is for country-level variables and Panel B is for firm-level variables. As shown in Panel A, our main proxy for the strength of a country s investor protection (PROTECT), its alternative proxies (LAWORG, ASDI, XPSD, LITSTD, and PUBENF), and our proxy for country-level earnings quality (EQ) vary widely across countries, with a general trend that investor protection is relatively strong in common law countries (U.S. and U.K.) and relatively weak in code law countries (e.g., France and Germany). We measure firm performance using operating incomes (E) and market-adjusted stock returns (RET). With respect to the two earnings components, the operating cash flow variable (OCF) has a mean of 0.039, while the mean of the accruals variable (AC) is Overall, the mean and median statistics of our firm-level independent variables suggest that the data are reasonably distributed, with a sufficiently large standard deviation. As shown in Panel B, our sample firm has, on average, about 2.5 cash flow forecasts and about 7 earnings forecasts issued by analysts. [ Insert Table 2 here ] 4. Empirical Results 4.1. Investor protection and the sensitivity of CEO turnover to cash flow versus other 18

21 performance information Primary results Table 3 presents a series of logit regression in Eq. (1) with PROTECT as our investor protection proxy. PROTECT, created by La Porta, Lopez-de-Silanes, and Shleifer (2006), is a comprehensive investor protection index and measured as the initial value of the index adjusted for the median of our sample counties. Section A of Table 3 reports the estimated results for Eq. (1) when earnings (E) are not decomposed, while Section B of the table reports the same when earnings are decomposed into cash flows (OCF) and accruals (AC). Market-based performance is measured as the market-adjusted stock returns (RET). Our objective here is to examine the impacts of investor protection on the sensitivity of CEO turnover to performance, particularly earnings and earnings components (cash flows and accruals). In column 1, we first report the results of our baseline regression without including our investor protection proxy (PROTECT). Consistent with prior studies, we find that the coefficient on E is negative and highly significant at p = 0.000, suggesting that the corporate board is more likely to replace CEOs with poorer earnings performance. In column 2, PROTECT and its interaction with E (i.e., E*PTOTECT) are included, and in column 3, the market-based performance measure (RET) and its interaction with investor protection (RET*PROTECT) are additionally included in the regression. The results show that the coefficient on E is still significantly negative, but at the same time, the coefficient on E*PROTECT is positive and highly significant (at p = and for columns 2 and 3, respectively). The positive 19

22 coefficient on E*PROTECT suggests that in countries with weaker investor protection, CEOs with poor earnings performance are more likely to be removed from their positions. This finding is inconsistent with DeFond and Hung (2004): with a similar but smaller sample, DeFond and Hung (2004) show that the effect of investor protection on earnings usefulness in CEO turnover decisions is generally insignificant. 11,12 As in DeFond and Hung (2004), we find that the coefficients on RET and RET*PROTECT are both negative and significant (at p = and p = 0.022, respectively), suggesting that CEOs with poor market-based performance are more likely to be terminated in countries with stronger investor protection. [ Insert Table 3 here ] DeFond and Hung (2004) hypothesize and find that strong investor protection institutions foster effective corporate governance and hence facilitate the termination of poorly performing CEOs. However, their postulation cannot provide an explanation for our finding that earnings become less useful for CEO turnover decisions as investor protection becomes weakened. To address this issue, we next decompose earnings into cash flows and accruals and then investigate the effects of investor protection on the two components of earnings. The results are reported in Section B of Table 3. As in Section A, we first report the results of our baseline regression without including PROTECT, and then, those with PROTECT included. 11 See, for example, Model 3 of Table 6 in DeFond and Hung (2004, p. 292), which shows that the coefficients on the interaction between their main investor protection proxy and the earnings measure are statistically insignificant but positive. Similar results are reported in their Table 5 (p. 287) in which two different investor protection proxies are used. 12 We will explain these inconsistent results in the following analyses regarding the usefulness of cash flows and accruals. 20

23 As shown in column 4, the coefficients on OCF and AC are both negative and significant at p = 0.000, suggesting that shareholders and corporate boards rely significantly on two components of earnings, i.e. cash flows and accruals, when assessing CEO performance. The negative coefficients on OCF and AC indicate that CEOs with poorer cash-flow and/or accruals performance are more likely to be removed from their positions. In columns 5 and 6, we present the regression results when operating cash flows (OCF) and accruals (AC) are respectively used as a single performance measure. In column 7, we report the regression results when OCF and AC are jointly included. In column 8, we additionally include RET and its interaction with PROTECT. Across columns 5 through 8, we find that: (i) the coefficients on the three performance measures (OCF, AC, and RET) are all negative and most are highly significant; (ii) the coefficients on OCF*PROTECT are all positively significant at p = 0.000; and (iii) the coefficients on AC*PROTECT and RET*PROTECT are all negatively significant at the level of 5% or better. The negative coefficients on AC*PROTECT and RET*PROTECT are consistent with DeFond and Hung s (2004) position that strong investor protection facilitates the termination of CEOs with poor performance. It is also consistent with the view that accruals and price information are more useful for making CEO turnover decisions in countries with stronger investor protection, where accruals and stock price are more able to capture the true firm performance. The positive coefficient on OCF*PROTECT suggests that cash flow performance is more useful to shareholders and corporate boards for assessing CEOs actions in an environment of weaker investor protection. This finding is in line with those of DeFond and Hung (2003, 21

24 2007) and Hung (2001) that cash flows supplement information that interprets the accrual-based earnings and attenuates the adverse effect of weak investor protection on the usefulness of earnings. The results reported in Section B also imply that the effect of investor protection on reducing the sensitivity of CEO turnover to earnings, as reflected by the positive coefficients on E*PROTECT in Section A, is driven by the cash flow component of earnings. 13 Consistent with DeFond and Hung (2004), we find that the strength of investor protection itself (as reflected in the positive coefficient on PROTECT) is associated with a significantly lower likelihood of CEO turnover, suggesting that CEOs in countries with stronger investor protection have a lower risk of being replaced. With regard to the control variables, we find that firm size (SIZE) and leverage (LEV) are positively associated with the likelihood of CEO turnover. These results suggest that CEOs of large firms and highly leveraged firms are more likely to experience dismissals in general, which is consistent with the findings of previous research (e.g., Fich and Shivdasani 2006). Overall, the results in Table 3 are consistent with our prediction that investor protection has 13 Note that, using the data of the period , we find that the earnings information is more useful in CEO turnovers in weaker investor protection countries, whereas DeFond and Hung (2004), with a sample for the period , show that investor protection s effect on earnings usefulness is generally insignificant. The differing results between our study and DeFond and Hung (2004) might be driven by two factors. One factor is that in recent decades firms have an increasing propensity to use market-based measures to provide incentives or to discipline their executives, especially in countries with strong investor protection institutions where capital markets are well developed such as in the U.S. (See, e.g., Core et al. (2003) and Bushman and Smith (2001) for more detailed discussions on this issue; and refer to Armstrong et al. (2010, p ) for a comprehensive review of the related literature). The second, as noted earlier, is the increasing use of cash flow information all over the world, especially after the series of corporate scandals in These two trends may make accruals information relatively less important in contracting with management in countries with strong investor protection over time (Armstrong et al., 2010). Consistent with our conjecture, we find that investor protection s effect on enhancing the usefulness of accruals is not robust to different investor protection measures, as indicated in Table 4. 22

25 opposing effects on the usefulness of cash flows and accruals information and that cash flow information plays a more important role in assessing CEO performance and making CEO turnover decisions when shareholder rights are poorly protected Using alternative investor protection proxies In Table 3, we measure investor protection by PROTECT. Although La Porta et al. (2006) assert that this index measures both the letter of the law and the strength of law enforcement in terms of disclosure requirements, liability standards, and anti-director rights of investors, this index may possibly capture only part of the strength of a country s investor protection institutions. Therefore, we also check whether our results reported in Table 3 are robust to the use of alternative proxies for investor protection institutions. For this purpose, we reestimate Eq. (1) using several alternative IP proxies, and report the results in Table 4. [ Insert Table 4 here ] In table 4, the proxies used in columns 1 through 5 are the origin of laws (LAWORG), the aggregate index of anti-self-dealing rights (ASDI), the ex post index of anti-self-dealing rights (XPSD), the liability standards index (LITSTD), and the public enforcement index (PUBENF), respectively. Similar to our previous findings, the results in Table 5 show that, when cash flows are interacted with various proxies for investor protection, the coefficients on the interaction term, OCF*IP, are all positive and significant at less than the 5% level, irrespective of which proxy for investor protection is used. The results reconfirm our prediction that the cash flow information plays a more important role in assessing CEO performance and making CEO turnover decisions 23

26 in countries with weaker investor protection. At the same time, we find that the coefficients on AC*IP are all negative, albeit insignificant in four of five cases. 14 Consistent with DeFond and Hung (2004), the results in Table 4 also reveal that the coefficients on the interaction term, RET*IP, are all negative and statistically significant (at less than 5% for three out of the five specifications) across all columns except in column 2. Furthermore, the coefficients on the three performance measures, i.e., OCF, AC and RET, are all negative and statistically significant across all columns. The results reported in Table 4, taken together, suggest that our main findings are robust to the use of alternative proxies for investor protection Do analysts cash flow forecasts improve the usefulness of performance information? In this subsection, using the reduced sample of firms with analyst forecast data, we examine whether and how analysts cash flow forecast activities help shareholders and corporate boards to identify and terminate poorly performing CEOs. In so doing, we first analyze the effect of analyst forecast activities on the usefulness of earnings and then that of cash flow information Analyst forecasts effect on the usefulness of earnings information We first estimate Eq. (2) to test the effect of analyst following on the relative usefulness of different performance information when earnings (E) are not decomposed. For brevity, we only tabulate estimated results using PROTECT as a proxy for investor protection. Though not tabulated, our findings using the alternative proxies for investor protection are qualitatively 14 See footnote 14 in Section for a related discussion about the less insignificant result on AC*IP. 24

27 similar to those using PROTECT. The regression results are reported in Table 5. As presented in Section A of Table 5, where all the firm-years with data on analyst forecasts are used, we find that the interactions of the two analyst coverage variables with earnings (i.e., E*ANA_CF and E*ANA_E) are both negative, although statistically insignificant at the conventional levels. Regarding the market-based performance, we find that the coefficients on the two interaction terms, RET*ANA_CF and RET*ANA_E, are both negative and significant at the conventional levels. This indicates that the likelihood of CEO turnovers is more sensitive to performance measures, whether earnings or stock returns, as analyst forecasting activities become more intense. The above results are consistent with the view that shareholders and corporate boards rely more on public information when identifying and terminating poorly performing CEOs for firms followed by more analysts. [ Insert Table 5 here ] To further evaluate the impact of investor protection on the CEO turnover-performance relation, we now partition the sample into two subsamples of observations: (i) the low PROTECT sample of firms from countries with relatively weak investor protection (with PROTECT lower than or equal to its sample median); and (ii) the high PROTECT sample of firms from countries with relatively strong investor protection (with PROTECT higher than its sample median). Sections B and C of Table 5 report the estimated results for Eq. (2) using the low-protect sample and the high-protect sample, respectively. As shown in column 3, the coefficient on E*ANA_CF is negative and significant for the low-protect sample (with p = 0.067). However, as shown in column 5, it is insignificant at 25