1
How do Industry Shocks affect the relationship between
performance and CEO turnover?
Program: MSc Accountancy and Control – Accountancy track, Faculty of
Economics and Business
Student: Alexandra Drăguși
Student number: 10636544
First supervisor: dr. P. (Peter) Kroos
Date: 21st of June 2015
2
Statement of Originality
This document is written by student Alexandra Drăguși who declares to take full
responsibility for the contents of this document. I declare that the text and the
work presented in this document is original and that no sources other than those
mentioned in the text and its references have been used in creating it. The Faculty
of Economics and Business is responsible solely for the supervision of completion
of the work, not for the contents.
3
How do Industry Shocks affect the relationship between performance and
CEO turnover?
Abstract:
This paper empirically examines how managerial retention decisions are related to various industry shocks. Prior turnover literature emphasizes that CEOs are punished for poor performance, but do not specify the underlying causes of bad performance. I contribute to the literature by finding a strong positive relationship between factors beyond CEO control and turnover. My results also show that turnover is not sensitive to performance that is contemporaneous with industry shocks. When I examine how corporate governance characteristics affect turnover, I find that CEOs that entrench vis-à-vis boards of directors are able to shield themselves from dismissal. These results are consistent with the fact that, both under normal industry conditions, as well as when shocks occur, boards do not act pro-actively in their supervisory duties.
Keywords:
4 Table of contents 1. INTRODUCTION ... 5 1.1. Background ... 5 1.2. Research question ... 6 1.3. Motivation ... 7 1.4. Structure ... 8
2. LITERATURE REVIEW, THEORY AND HYPOTHESES ... 8
2.1. Constructs of organizational design decisions: delegation of decision rights, information asymmetry and performance measures ... 8
2.2. Theoretical perspectives on CEO dismissal ... 10
2.3. Empirical literature about CEO dismissal ... 10
2.3.1. The relationship between performance and dismissal ... 10
2.3.2. Moderators of the relationship performance - turnover ... 11
2.4. Industry shocks as moderator ... 15
2.5. The effect of industry shocks on the relation between performance and turnover ... 16
3. SAMPLES AND VARIABLES ... 17
3.1. Sample selection and data sources ... 17
3.2. Variable measurement ... 18 3.2.1. Industry shocks ... 18 3.2.2. CEO turnover ... 18 3.2.3. CEO performance ... 20 3.2.4. Control variables ... 21 3.3. Empirical models ... 21 4. EMPIRICAL RESULTS ... 22 4.1. Summary results ... 22 4.2. Multivariate analyses ... 26
4.2.1. Industry shocks and CEO turnover ... 26
4.2.2. Effect of industry shocks on the relation between turnover and performance ... 31
4.2.3. Sensitivity analyses ... 32
5. CONCLUSION ... 37
References ... 39
5 1. INTRODUCTION
1.1. Background
Owners of the firm delegate decision authority to executive managers. This, however, creates problems associated with informational asymmetry as executive managers now have more (inside) knowledge about their effort choices and their abilities relative to corporate outsiders, such as shareholders (Jensen and Meckling, 1976). To address this, firms periodically disclose information to capital markets on the basis of performance measure outcomes. As such, the board of directors can infer whether executives have taken these actions in the best interest of the owners, as well as whether the abilities of the executive match the requirements of the firm. Indeed, prior research shows a negative relationship between performance and dismissal (Coughlan and Schmidt, 1985; Warner et al., 1988; Hermalin and Weisbach, 1998; Engel et al., 2003; Farrel and Whidbee, 2003; Bushman et al., 2010; Dikolli et al., 2014). The explanation is that when financial indicators reveal unfavorable information the CEO is replaced because the firm’s owners infer that she is ineffective in formulating and implementing the appropriate strategies and policies that create shareholder value (Dikolli et al., 2008). Kwon (2005) proposes two motivations for dismissal: one relates to sorting whether the employee is well suited for the job, and the second points out to an incentive explanation where the employer uses the turnover to provide incentives for employees’ unobservable effort. This threat of dismissal has also been analyzed by Huson et al. (2004) under the scapegoat hypothesis. Subsequent research looks into several attributes that affect the performance dismissal relationship. Firstly, the quality of performance measures has been analyzed by various researchers. Ghosh (2011) suggests that only accounting losses result in a higher likelihood of a subsequent CEO turnover and variations in profit level do not impact CEO turnover decision. An explanation is the conservative nature of accounting (Watts, 2003). Losses tend to be more informative than profits because it include current and future losses, while for future profits accountants have a tendency to require a higher degree of verification. As a consequence, losses can show whether a CEO invested in negative NPV projects rather than maximizing the profits. In addition, research has showed that accounting returns receive greater weight in turnover decisions when they are relatively more precise and highly correlated with stock returns (Engel et al., 2003). On the other hand, Kaplan and Minton (2012) show that turnover is more sensitive to stock performance. An explanation lies in the prediction of Hermalin and Weisbach (1998) that stock returns, besides being a function of current management, also reflect the market’s expectations of future management changes.
6 Secondly, CEOs are held accountable for the performance measures outcome. Jenter and Lewellen (2010) find that boards fire CEOs for poor performance and that the threat of dismissal is a first-order source of incentive for most CEOs (Höppe and Moers, 2008). Based on the relative performance evaluation theory, executives’ dismissal should be related only to company’s performance. However, Jenter and Kanaan (2014) prove that under some circumstances boards allow exogenous shocks to firm performance to affect their CEO turnover decision. They find that market and industry performance are also relevant in explaining CEO turnover.
Thirdly, tenure moderates the board’s emphasis placed on firm performance in affecting CEO dismissal. Since boards of directors infer the CEO’s ability based on periodically observing firm performance, their beliefs regarding managerial ability become precise over the employment period. That is, the weight that accounting numbers have in updating owners’ perceptions about the CEO’s uncertain ability decreases with tenure as she gains firm specific experience (Dikolli et al., 2014).
Fourthly, corporate governance characteristics have usually been used to explain how board influences the turnover (Jenter and Lewellen, 2010). The overall opinion is that an independent board makes dismissal more sensitive to performance, while a weak board can be subject to CEO entrenchment and may fail to act against their manager when confronted with bad performance.
Little research (Wiersema and Bantel, 1993; Eisfeldt and Kuhnen, 2013; Jenter and Lewellen, 2010; Guay et al, 2014) specifically looked into the question of how changing circumstances affect the relationship between performance and turnover. While standard economic theory suggests that in assessing the quality of its CEO the board of directors should ignore components of firm performance which are caused by factors beyond the CEO’s control, current literature shows that CEO ability (and the quality of the CEO-firm match) changes due to exogenous shocks to firm performance.
1.2. Research question
The principal agent paradigm describes an environment where the principal creates incentives to align its interests with the agent’s. What the research question wants to examine is the matching between the firm’s need and CEO skills. In other words, when an industry shock creates a change in need for leadership, how is the relationship between performance and turnover affected. The research question is stated as follows:
7
What are the effects of industry shocks on the relation between CEO performance and turnover?
1.3. Motivation
This paper contributes to the emerging body of CEO turnover by extending the stream of research built upon the assumption that industry-level poor performance signals bad CEO-firm matches which becomes an indicator of dismissal. While many papers tackle the CEO pay topic, recent changes in CEO turnover trigged by shifts in the business environment receive little attention. I will not look, as the previous papers did, to the idea that executives are dismissed mainly due to poor financial performance. Instead my research topic will focus on another cause of mismatch between the principal and agent: whether CEOs can adapt to industry changes. It is the future performance that is evaluated by the board in assessing the continuing of employment.
There are various calls that ask for empirical research to explore the causes of CEO turnover.
Firstly, Brickley (2003) observes that “firm performance continue to explain very little of the
variation in CEO turnover.” He calls for other detailed exploration of factors that influence turnover, besides performance variables. Secondly, Kaplan and Minton (2012) suggest that CEO’s job has become riskier over time. Their study explores the annual CEO turnover for a sample of large US companies from 1992-2007 and discovers that after the year 2000 the turnover increased from 15.8% to 16.8% implying a decreased average tenure as CEO from 7 to 6 years. This perspective requires some explanations. My current study wants to look deeper into the CEO “bad performance” causes. Thirdly, Wiersema and Bantel (1993) are the first to argue that top management turnover is influenced by the environment and that CEOs might find it difficult to adapt to new conditions. Therefore, this thesis intends to contribute by examining to what extent different exogenous factors influence boards dismissal decision.
My main motivations stands behind the Guay et al. (2014) paper. The researchers direct the attention to the idea that CEOs need to adapt to industry shocks concluding that this “skill” plays a significant role in manager turnover decision. Also, PwC’s 18th annual global CEO survey released in January 2015 points out that tomorrow’s CEO must have capabilities such as strategic thinking and adaptability.
Overall, a better understanding of the causes behind executive turnover can offer a better understanding of the shrinking managers’ tenure. Also, by having an insight into the variables that add a greater risk for executive dismissal, shareholders and stakeholders can discover the challenges that the company has to be aware off and infer the direction of the firm new strategy.
8 1.4. Structure
The remainder of the paper is structured as follows. Section two describes the theoretical framework, then various key papers related to the research question are presented and at the end the hypotheses are formulated. Section three presents the research methodology starting by defining the sample, variable measurement and ending with building the empirical model. The results and sensitivity analyses are discussed in section four. Finally, section five concludes the thesis and points to the limitations encountered.
2. LITERATURE REVIEW, THEORY AND HYPOTHESES
2.1. Constructs of organizational design decisions: delegation of decision rights, information asymmetry and performance measures
A dominant theme that led to an abundance of economic research is the principal – agent relationship. The agency theory is underlined in Jensen and Meckling (1976) and is built upon the division of ownership and control due to the fact that owners who are defined as principals do not usually manage the company by themselves, but they hire executives, referred as agents, who have to act on behalf of them. However, as both parties are interested in maximizing their own utility, it is expected that the agent will be more inclined to pursue its own interest by gaining prestige, power and job security against the owners who seek profit maximization. Along these lines, two characteristics emerge: goal incongruence and information asymmetry. The assumption behind goal incongruence is that the agent (i.e. CEO) will try to maximize her utility function at the expense of principal (i.e. shareholders) own utility. The second characteristic, information asymmetry, refers to the fact that the principal has less information about the agent’s characteristics and actions. As a consequence it has to use performance measurement and monitoring.
The previously mentioned unequal distribution of information between the owners and CEO creates two agency problems: adverse selection and moral hazard. The first occurs before the CEO is hired and refers to the principal’s inability to observe the agent’s private information that may lead the candidate to misrepresent his/her abilities. In this setting, the board can select a poorer CEO. However, the selected manager will be dismissed as over time the principal learns about the agents’ ability and updates its view. Another proposed solution to mitigate this issue is to hire an insider CEO.
9 The second agency problem, moral hazard, takes place after the CEO accepts the position. It is based on the assumption that individual actions cannot be observed and hence contracted upon. So, CEOs can shirk and invest less than the agreed amount of effort. The mechanism to respond to moral hazard is the use of monitoring. As owners of the company have little interest in collecting information on their agent, it is the board of directors’ responsibility to monitor the CEO by using performance measurement and dismiss her in case they encounter poor performance.
As such, performance measures are used to address the adverse selection by assessing the manager’s abilities and mitigate moral hazard by making sure that top management acts in the best interest of shareholders and stakeholders. Two perspectives emerge with regards to them. The first one relates to incentives as board of directors takes the decision to reward employees (e.g. promote) based on the observation of performance measures. The second links performance measure to dismissal decision.
However, concerning the last perspective, there is not always a clear understanding of whether the dismissal is motivated by the fact that the employee is not well suited for the job or because the principal wants to provide incentives for effort. Current literature provides two reasons for the threat of dismissal. Firstly, the sorting explanations is developed on the idea of a skill matching between the employee and the job so it is expected that a good performance to increase the principal’s belief in the agent’s “type”, such as skills or matching quality, and reduce the dismissal probability (Kwon, 2005). Secondly, the incentive explanation argues that the principal uses dismissal as a mechanism to provide “incentives” for agents’ unobservable efforts because turnover can lead to a loss of guaranteed future income, decreased reputation in the labor market, and a corrosion of the value of equity held (Höppe and Moers, 2008). In the incentive model, good performance indicates a higher likelihood of the agent’s high effort and thereby reduces dismissal probability (Kwon, 2005). Although current literature has not tried to distinguish between the two types, there are evidences that over the agent’s tenure the sensitivity between performance and dismissal reacts differently under each model. Kwon (2005) disentangles these effects and concludes that the slope (sensitivity) of dismissal probability with respect to performance increases over time when using the incentive
explanation. However, with regards to the sorting explanation, the slope is expected to decrease
because over the agent’s tenure the principal receives more observations on the agent’s ability. As a consequence dismissal becomes less sensitive to performance due to the fact that new observation has a smaller impact on the principal’s belief about the agent’s type.
10 2.2. Theoretical perspectives on CEO dismissal
The two aforementioned perspectives also have implications for the question whether ability of the manager and firm performance should increase following CEO turnover. This is explained by the ability and scapegoat hypotheses, but only the former view finds an increase in both managerial ability and firm performance. The underlying assumption in the ability hypothesis is that managers have different qualities. However, the board cannot observe its CEO quality and hence must rely on realized performance. Poor performance signals a low quality type of manager and board reacts by dismissing her because it is supposed that the benefits of replacement exceeds the costs. As such, the new CEO is expected to have superior quality and by that to provide an increase in the managerial ability. Further, because bad luck can trigger poor performance, by dismissing the incumbent CEO the board expects that managerial luck reverts to normal thus increasing the company performance.
The scapegoat hypothesis proposed by Huson et al. (2004) is built on the proposition that managers dislike effort and therefore they must be threatened with dismissal if their performance is low. In contrast to the ability hypothesis, under the scapegoat approach it is assumed that managers have the same ability. Following this perspective dismissing the incumbent CEO will not improve the managerial ability because the new CEO will have the same quality as the previous. The only explanation of turnover is that board fires only unlucky managers to maintain the credible threat of dismissal and they expect that firm performance will improve due to the reversal of bad luck.
2.3. Empirical literature about CEO dismissal
2.3.1. The relationship between performance and dismissal
The prior literature is clear that poor organizational performance tends to precede executive departure. The hypotheses in this stream of research are predicted on financial indicators revealing public information about the CEO ability in increasing the company’s value. Therefore, owners use periodic accounting reports or firm stock performance to learn about their CEO ability. Further, it is expected that their decision to terminate an executive employment contract to be triggered by the historical performance revealing inferior managerial ability.
Coughlan and Schmidt (1985) are one of the earliest documenting the relation of stock price performance and management turnover. The researchers use a timeframe of three years (i.e. 1977-1980) and by ranking the firms on their stock return they report that for CEOs whose age is less than 64 years the probability of a management change for the lowest 1% is seven times
11 that for CEOs in the top 1% (i.e. 21.3% chances of leaving the company in the next year to 3.1%). Warner et al. (1988) confirm that stock price reflects information about management performance and that an inverse relation exists between share performance and turnover, although not highly sensitive (i.e. there is a time lag of two years). The study analyzes 269 random firms over an extended period of years 1962-1980. As the previous research, it ranks firms by performance and places them in deciles. The results hold only for extreme performance and indicate that the probability of top management change in the bottom decile is 1.5 times larger than in the top of the firms (i.e. 12.8% to 8.6%).
Rather than performance alone, Farrel and Whidbee (2003) suggest that board of directors focuses on deviation from earning performance expectations. As such, they argue that one year analyst forecast errors provide incremental information about managerial performance and that there is a negative relation between earnings growth forecasts and CEO dismissal. However, the results show that the economic effect is quite small because if the industry-adjusted analysts forecasts decline by one standard deviation then the probability of a managerial turnover increases by only 0.88%.
2.3.2. Moderators of the relationship performance - turnover
2.3.2.1. Information
A general relationship exists between poor performance and executive departure, however the literature discovered that the relation is not uniform, but moderated by other variables. A primary moderator is the information provided by various performance measures and how board uses it to motivate their turnover decisions. Prior studies develop metrics for poor performance, yet there is no definite answer into what measure is an appropriate indicator of bad performance that dominates in determining CEO turnover.
Ghosh (2011) provides an intuition on the use of the accounting measures and makes the assumption that poor performance can be captured only into accounting losses as they are more informative in judging managerial performance due to their conservative nature. A key implication of accounting conservatism principle is that losses are more timely and reliable signals of deteriorating managerial performance than small or declining profit (Ghosh, 2011). Further, a loss indicates that the CEO failed in being a good steward leading boards of director to make a critical evaluation about the executive abilities. This may increase the boards focus on collecting more adverse information (i.e. private information) which might lead to frequent CEO turnover (Hermalin and Weisbach, 1998). Two reasons sustain the previously mentioned
12 assumption. Firstly, in-depth evaluation can expose detrimental information about the CEO’s ability, secondly boards want to see when confronted with serious difficulty, whether the current leadership will be able to address them. Finally, Watts (2003) explains that the CEO dismissal is caused by a failure to meet shareholders expectations as reporting accounting losses will lead to erosion in their equity value.
Other researchers emphasize the use of stock prices. Kaplan and Minton (2012) find that since the year 2000 board of directors became sensitive to firm’s stock performance. This is due the fact that boars are more focused on shareholder value than previously thought (Jenter and Lewellen, 2010). Another explanation is provided by Engel et al. (2003) who predict that directors will put a greater weight on information in stock returns when earnings timeliness1 is low or when stock returns are less variable.
Overall, the pattern of CEO turnover due to poor performance is robust, however firm performance does not prove to be a powerful explanatory variable. Although it is statistically associated with executive dismissal, the variance explained by performance was in most of the studies below 50%, particularly in the range of 10-20%.
2.3.2.2. Noise
The contracting theory proposes that the relative importance (i.e. weight) of performance metrics should be a decreasing function of their noise and an increasing function of their sensitivity to employee effort or decision. Engel et al. (2003) use both types of measure and conclude that accounting performance is more predictive of a CEO dismissal than stock performance. The underlying intuition provided by Höppe and Moers (2008) is that stock prices include market expectation and therefore provide a noisier signal, and also markets are not always sufficient informed about the firm’s strategy which reduces the ability to envision the future company plans.
Although boards may use different measures for evaluation, the turnover decision should be triggered only by analyzing the firm performance that is under the CEO’s control. By filtering out luck, executives will be held accountable only for their decision that influence the firm value and not for things that are beyond their control. Following this view, Bushman et al. (2010) look upon the sensitivity between CEOs performance and turnover caused by two types
1 Timeliness refers to that fact that earning in the current period should contain the consequences of actions undertaken in this period.
13 of risks, one that emerges from the factors that are under the executive control, the second relates to uncontrollable factors. They conclude that when there is idiosyncratic volatility (i.e. related to CEO’s abilities) the turnover/performance sensitivity is high. On the other hand, factors out of the CEO control (i.e. noise, industry-wide effects) pose a systematic risk that decreases the sensitivity. This is due to the fact that boards find it more difficult to clearly asses their executive talent.
A stream of research looks into the use of RPE (relative performance measure) in CEO turnover. The underlying idea is that besides public and private information, board of directors evaluates their CEO based on a group of peer firms that are subject to the same industry and market shocks. In this way it will offer an insurance to their CEO that she will not be punished for bad luck. However, new empirical results show that boards fail to filter all exogenous observable shocks. Eisfeldt and Kuhnen (2013) find that in the context of a competitive assignment CEO turnover is affected not just by the firm and CEO characteristics, but also by the industry conditions which will change the appropriate industry skills required by a company. Jenter and Kanaan (2014) prove that, in contrast to RPE, low industry stock returns and (to a lesser extent) low market returns increase frequency of forced CEO turnovers. Particularly, if the industry component of a company performance decreases from 90th to its 10th percentile then this doubles the probability that the CEO will be forced to leave. The results are explained by the timeframe used (i.e. recent period 1993 – 2009) which includes the recession that signals more information about the CEO skills. In contrast to a boom period, recessions reveal more about the CEO qualities as whether she anticipated the downturn and prepared for it. However, dismissals only occur on CEOs who underperform their peer group.
2.3.2.3. Tenure
A more recent paper confirms that boards behavior has not changed over time and bad performance still trigger turnover. Dikolli et al. (2014) show that subsequent negative quarterly performance encountered by a company increases the likelihood of CEO dismissal. The researchers enforce once again the assumption that performance measures reveal information about the CEO uncertain ability and that boards act when they see bad performance. However, they find that CEO’s tenure can attenuate the use of performance measure in turnover decisions. The chances to be dismissed following negative quarterly performance report depend on whether the CEO was newly appointed or she had a longer career. Indeed, new CEOs have a turnover likelihood of 34-43%, in contrast with the 4-11% for experienced executives. The explanation for this result points out to two hypotheses. The first one, learning hypothesis,
14 highlights that CEO ability becomes precise over time, and such the uncertainty of managerial ability is resolved over tenure. The second one, managerial hypothesis, sees tenure as a proxy for executive power. This implies that board will lessen the emphasis put on performance measures not because they choose to monitor less, but mainly due to weak corporate governance. Over a long tenure the CEO can became entrenched and acquire greater negotiating skills which may lead to less independent boards.
The current prediction is derived from Hermalin and Weisbach (1998) who find evidence that board of directors are prone to increase their monitoring only when faced with new CEOs because their abilities are unknown. The study approaches both learning and managerial hypotheses mentioned earlier and emphasizes that entrenchment has a positive relation to CEO tenure and that the impact of (poor) performance on the likelihood of forced turnover will decrease with CEO tenure. The assumption behind this is that over the executive employment horizon the board will become less independent. To demonstrate this, the researchers create a model of a Nash bargaining game between the CEO and board. As the CEO proves herself valuable for the firm her bargaining power increases. Intuitively, new CEOs lack bargaining power as they have the same expected value. In this context, the board has to monitor more a newly appointed CEO than the incumbent, which comes with additional costs. Hence, a less independent board will be more willing to retain a mediocre CEO, than dismiss her. The results are consistent with the belief that entrenched CEOs are less monitored. However, the study also shows that when faced with a decrease in the corporate performance the board members are replaced by independent members.
2.3.2.4. Board quality
While Hermalin and Weisbach (1998) control for board characteristics, Jenter and Lewellen (2010) investigate board characteristics as a moderating effect for the relation between turnover and performance. Their results indicate that strong boards2 are more likely to replace their CEOs after bad performance. For example, a CEO in the year four of her tenure has 52% chances to leave the company if the performance is in the bottom quintile, in comparison with 8% for the top performance. Therefore, there is a 44% spread between turnover of top and bottom that is driven by performance. In contrast to Hermalin and Weisbach (1998) who predict that board has a preference for the incumbent CEO, Jenter and Lewellen (2010) show a much more
2 Strong boards are defined as small boards with a majority of independent directors and higher director ownership. These variables have been used as prior literature show that they affect the boards’ effectiveness.
15 aggressive response of boards to bad performance and the fact that long tenured CEOs are as prone to turnover as their younger peers. This is surprising since mature CEOs should be more entrenched (as they have enough time to prove themselves and establish their power) as well as more valuable to the company (because a longer tenure should reflect higher firm-specific ability). Indeed, the researchers enforce the assumption made by early literature that board quality affects the turnover decision since in case of a bad corporate performance over the first five years of tenure the CEO turnover probability is 83% for companies that have a strong board, and only 49% for weak boards. However, this governance spread is significant mostly at lower performance level. Overall, the results prove that boards are more focused on shareholder value than previously believed and that the threat of turnover continues to be a source of incentive for most managers.
2.4. Industry shocks as moderator
While most of the literature focuses on firm characteristics in order to explain the matching between the CEO and company needs, there are few studies that go in depth into analyzing the effects of external environment. As a business can be affected by market changes (i.e. customer preferences or suppliers’ and competitors’ actions) and technological changes (i.e. advent of products/ services/ knowledge), the organization has to adapt to dynamics of environment and CEOs are expected to influence this fit by means of strategic thinking and actions. Thus, replacement of top executives becomes a response to overcome organizational inertia and to adapt strategically to a changing context.
An early study of Wiersema and Bantel (1993) argues that top management is influenced by the instability encountered into the external environment which can destabilize the match between the company and its manager, thus leading to greater turnover. The researchers propose a model of adaptation to environmental context, however, the study is concentrated not solely on the CEO, but on the whole management team3. The framework developed defines three dimensions for environment: munificence (i.e. organization growth), instability (i.e. unpredictable change pertinent to strategic decision-making) and complexity (i.e. range of environmental factors that need to be considered in strategic-decision making). Each of the previous environmental conditions is expected to link with firm performance. Such that lack of munificence and increase in environmental instability lead to high competition between
3 The management team is composed of a first tier (Chairman, CEO, President and COO) and second tier (senior
16 companies and thus unstable market conditions that will lower industry and firm performance due to the fact that some organizations are trying to gain market share at the expense of others. With regards to complexity in business operations, an increase in its level means that management will need more information for taking decisions that may affect the sharing of information among team members and slow down the decision making process which in turn points out to lower performance outcomes. Researchers conclude that CEOs can find it difficult to adapt to these conditions which lead to increased turnover.
The role industry conditions play in CEO turnover has not been researched by prior literature until recently. Mainly because past CEO turnover studies emphasize the role of the board of directors as an effective monitoring mechanism that learns over time about their CEO abilities and acts in accordance with deteriorating corporate performance. However, newly research by Eisfeldt and Kuhnen (2013) points out to the relevance of industry conditions in triggering CEO turnover. They construct a competitive framework to analyze the firm trade-off between preserving the match of CEO skills to company needs and choosing an outside option (i.e. replace the executive). According to their model, an industry shock is defined as a shock to firm’s skill demands that deteriorates the quality of the match between firm and CEO. The optimal incumbent manager will lack the skills necessary for the new industry conditions, which motivates the board to find a replacement that possess the proper skills to adapt the new industry changes. This empirical implication is beyond the principal agent paradigm where the CEO turnover was analyzed by means of RPE. The current research yields novel prediction, as such in addition to managerial and firm characteristics, industry conditions have been found to drive dismissal as well.
Particularly, Guay et al. (2014) analyze a number of industry shocks and suggest that shocks to asset growth, market-to-book, investment, research and development expenses, sales, competition and globalization have positive and significant effects on CEO dismissal. The evidence points out that industry shocks may signal the importance of other abilities than those that have been successful in the past. In the light of the recent literature that discusses determinants of turnover beyond firm performance my first hypothesis is formulated as follows:
Hypothesis 1: Industry shocks increase the likelihood of CEO turnover. 2.5. The effect of industry shocks on the relation between performance and turnover The learning hypothesis indicates that over time managers learn about their CEO abilities, however recent research find evidence that is inconsistent with this hypothesis. Indeed, Jenter
17 and Lewellen (2010) examine CEO turnover and discover that boards find the recent performance being more useful in evaluating their managers, than distant past from three or four years ago. Guay et al. (2014) explores the turnover – performance relation and the effect shocks have on board’s assessment of its CEO. The results suggest that when there is a change in the business environment the board of directors will consider more useful to evaluate the current rather than past CEO performance. The researchers share the idea that earlier periods are less informative, and therefore it is more useful to weight the present manager’s performance. This is because past performance is not any more an informative measure as it captures only a backward looking perspective not correlated to the new business changes and manager’s abilities to react.
The second hypothesis should show the effects of industry shocks on the use of information sources to consider when deciding upon the CEO turnover. Hence, the second hypothesis can be formulated in the following way:
Hypothesis 2: Industry shocks decrease the use of (a) past performance information, and increase the use of (b) current information for considering CEO turnover.
3. SAMPLES AND VARIABLES 3.1. Sample selection and data sources
The empirical analysis obtains data on CEO turnover from Compustat’s ExecuComp database which contains roughly 1,500 companies each year that are in the S&P 500, S&P mid-cap 400 and S&P small-cap 600 indices. The sample spans 2006 – 2014, a timeframe that is interesting to observe as it includes both the economic expansion (until the late 2007) and recession (starting with December 2007 until June 2009 leading to crisis to various industries such as automotive industry). This period is marked with increased volatility that pushed companies to change in order to adapt to new competition and market demand. The data collected from ExecuComp yields an initial sample of 12,979 observations. Further, for each firm, data on accounting and market returns and other firm characteristics are obtained from CRSP/Compustat Merged Fundamentals Annual. Following prior research, financial firms (SIC codes 6000 to 6999) and utilities (SIC codes 4900 to 4942) are excluded since they are subject to financial regulation, supervision that makes the firms to adopt certain capital requirements thus influencing their financial indicators. Further data is lost because the first year of the sample serves only for building lagged variables, also missing items within the
18 calculations (such as total assets) and winsorizing (at 1% and 99% levels) reduce the sample to 8,835 firm-year observations with the requisite information. The final sample contains 1,462 unique firms.
3.2. Variable measurement 3.2.1. Industry shocks
Consistent with recent prior literature, CEO retention decision is analyzed in light of a matching between the executive skills and firm’s need. In this paradigm, any shock that affects the industry business conditions can modify the skills needed in the market and thus make the board reconsider the matching mechanism. A shock is a change in the industry environment that may imply that the best practices in the past do not work anymore in the future. Industries are defined according to the two-digit Standards Industrial Classification (SIC). Each industry shock is computed by aggregating the firm-level variable to the industry-level, then calculating the absolute percentage change in industry averages from one fiscal year to another (between year
t-1 and t). Further, both an extreme increase and decrease can affect the business conditions,
but the primary analyses use unsigned industry shocks. However, subsequent analysis tests the shock sign.
The industry shocks are selected based on reading the papers of Guay et al. (2014) and Wiersema and Bantel (1993), as well as the PwC’s 18th annual global CEO survey. It is been acknowledged that the environment in which CEOs operate is continuously changing. Executives must learn how to deal with the breakthrough innovations, maintain a sustainable business in a competitive landscape that rapidly reshape, and adapt to the new patterns of consumer behaviors. Specifically, six shocks are defined: AssetGrowth is the change in industry assets, MB is the change in industry market-to-book ratio, Investment is the change in industry capital expenditure, Competition is the change in Herfindahl index, Sales is the change in industry sales and Technology is defined as the change in industry research and development (research and development) expenses4.
3.2.2. CEO turnover
CEO experiences turnover if the executive listed in ExecuComp in year t is not listed as CEO in t+1. The executive ID number is used to verify this. Formally, Turnover is an indicator
4 The primary analysis uses continuous variables for the shocks. However, in untabulated analysis, the continuous
variables are replaced with discrete measures. Dummy variables are defined for large shocks, particularly, an indicator variable takes the value one for shocks in the 5th quintile, zero otherwise.
19 variable equal to one if the CEO in year t leaves the office in year t+1, zero otherwise. Most of the academic research distinguishes between “voluntary” and “involuntary” turnover, but drawing this distinction implies the subjective interpretation of press releases. That can be prone to misinterpretation (i.e. probably an underestimation of involuntary turnover). However, the majority of academic papers tackling CEO dismissal separate the two types of turnover. Engel et al. (2003), or Guay et al. (2014) run their regression for the total turnovers as well as for “involuntary”.
Therefore, a simplified model is used to distinguish between the two types of turnover. As such, according to prior literature (Bushman et al., 2010) the current thesis considers forced turnover if the CEO that leaves the company is aged 60 or under, also using the “reason” variable from Compustat, executives that are mentioned to be “retired” or “deceased” are excluded from the “forced” sample. However, if it is mentioned that she “resigned” this will still be interpreted as
forced turnover, since a CEO can be pressured to depart. Forced turnover (defined by the
variable ForcedTurnover) takes binary values, one if the CEO is dismissed in the following fiscal year, zero otherwise. The empirical models will be run for forced turnover as the dependent variable.
For illustrative purposes, an example of a forced CEO turnover within the computer industry that was influenced by two of the examined shocks is discussed. Between 2008 and 2010 when the economy was facing global financial downturn most of the industry grew at a slower rate. However, at the end of 2010 and beginning of 2011, the US GDP grew with 2.5%, respectively 1.6% rate, increasing the confidence in economy and the willingness of consumers to spend. Also, since countries such as India or China increased their need for technology, the PC hardware manufacturing industry had a context to boost its sales. In the sample, the computer industry in 2011 experienced a positive shock to sales (increase of 13.44%) and market-to-book ratio (increase of 10.62%), both above the 75th percentile. Twelve of the ninety-nine computer-industry executives in the sample were forced out, including Hewlett Packard (HP) CEO, Leo Apotheker. Although, HP could benefit from the increased demand since it is a hardware company, the executive had difficulties in adapting to the industry growth.
„ HP's board said the company needed a change at the top. The company cut its financial outlook three times in Apotheker's tenure, and on Thursday, HP said it was not confident it would be able to meet its sales targets for the current quarter.
20
"We are at a critical moment and we need renewed leadership to successfully implement our strategy and take advantage of the market opportunities ahead," said Ray Lane, who on Thursday was named executive chairman of HP's board.”
(Extract from CNN Money, September 22, 2011) 3.2.3. CEO performance
Prior literature has used a diverse set of predictors for executive turnover. Ghosh (2010) suggests that only accounting losses lead to CEO dismissal. Earnings data are more reliable over market return due to the fact that they measure short-term profits. On the other hand, stock price reflects the presented discounted value of the expected future cash flows of the firm. Also, it includes the market’s estimate of the probability that a poor performing CEO will be dismissed. However some researchers use stock return as performance measure, e.g. Kaplan and Minton (2012) since boards nowadays are more focused on shareholder value. While there is still not a unique shared view into the appropriate choice of performance measure, for the empirical tests both accounting (ROA, LagROA) and market performance (RET, LagRET) are used. ROA is preferred because explicitly shows if the firm is able to generate an adequate return on the assets it has. CEO performance measures are constructed over a whole fiscal year. ROA is defined as industry-adjusted earnings before extraordinary items in year t scaled by the beginning of period assets, while LagROA is the industry-adjusted return on assets over year
t-1. RET is the industry-adjusted percentage change in the market capitalization calculated using
the Compustat variable “mkvalt”, particularly (mkvaltt-mkvaltt-1)/ mkvaltt-1. “Mkvalt” is the total fiscal year common shares outstanding multiplied by the month-end price that corresponds to the period end date. In case “mkvalt” is missing, the variable is replaced by multiplying the share price at the fiscal year with the shares outstanding. Due to the relatively high variation in market capitalization, the logarithm of market value is reported.5 LagRET is the industry-adjusted market return over year t-1. The values are industry industry-adjusted to filter out industry trends. For each firm, industry adjusted performance measures were computed by subtracting the industry average from the firm measure. For example, an industry adjusted ROA is obtained by subtracting the industry ROA from the firm ROA. The industry averages were calculated on the basis of all firms with the same aggregate two-digit SIC code as the sample firm during a particular year.
21 3.2.4. Control variables
Control variables measurement is done by using a vector (‘Controls’) that includes the following firm variables: Size and Growth because larger and growing firms have a greater demand for high quality CEOs. Size is defined as the natural logarithm of total assets at the end of year t and Growth is the ratio of market to book value of equity at the end of year t. Proxies for firm risks, such as earnings volatility (VolatilityROA) and stock volatility (VolatilityRET) are used because when firms have increased volatility they are more exposed to shocks that destabilize their practices. Both are calculated as the standard deviation of accounting return, respectively market return over five years. Also, the sample is controlled for industry concentration (Concentration) since executive turnover is more often encountered in highly concentrated industries. Concentration is the Herfindahl-Hirschman Index (HHI) at the end of year t, calculated as the sum of squared market shares of all firms in the industry (2-digit SIC). Managerial characteristics are included such as CEO Age that is equal to the age of the executive at the end of year t. Finally, the second variable used to control for managerial entrenchment is equity ownership (EquityOwn) defined as the percentage of the company’s shares owned by the CEO.
3.3. Empirical models
The regression used to test CEO turnover and industry shocks is defined according to Guay et al. (2014).
The following regression is used to explore the first hypothesis:
Hypothesis 1: Industry shocks increase the likelihood of CEO turnover.
ForcedTurnover = α + β Shockt + γ Controls + ɛ [1]. Where,
ForcedTurnover = Indicator variable equal to one if the CEO is replaced in the year following the year for each observation. This relates only to the forced turnovers, defined in Appendix
Shockt = Vector that includes all six shocks (i.e. Asset growth, MB, Investment,
Competition, Sales and Technology)
Controls = Vector of controlling variables including Growth, Size, Concentration,
VolatilityROA, VolatilityRET, Age, EquityOwn and also containing the
22 With regards to shocks, that are the main variables of interest, it is expected a positive and significant β for all of the shocks. Therefore, a β>0 substantiates the first hypothesis.
The coming regression, developed by Guay et al. (2014) defines the second hypothesis. Hypothesis 2: Industry shocks decrease the use of (a) past performance information, and increase the use of (b) current information for considering CEO turnover.
ForcedTurnover = α + β Shockt + φ1 ROAt* Shockt + φ2 LagROA * Shockt + φ3 RETt * Shockt + φ4 LagRET * Shockt + φ5 ROAt + φ6 LagROA + φ7 RETt+ φ8 LagRET + γ Controls + ɛ [2]. Where (in addition to the variables defined in the above regression),
ROA = Industry-adjusted return on assets over year t
LagROA = Industry-adjusted return on assets over year t-1
RET = Industry-adjusted market return over year t
LagRET = Industry-adjusted market return over year t-1
Controls = Vector that includes firm specific characteristics (Growth, Size
Concentration, the proxies for risk: VolatilityROA and VolatilityRET)
and managerial characteristics (Age and EquityOwn).
As business conditions change, it is expected that board of directors assign a greater (small) weight to current (lagged) performance in assessing CEO adaptability to the shocks. To test this assumption, both accounting return (year t) and lagged accounting return (year t-1) are interacted with industry shocks. The same applies for stock return. Therefore it is predicted that φ1, φ3 < 0 (more negative weight on current period when a shock occurs) and φ2, φ4 > 0 (less negative weight on prior period in the presence of an industry shock).
4. EMPIRICAL RESULTS 4.1. Summary results
Table 1 provides descriptive statistics for my sample. Panel A presents managerial and firm characteristics. Regarding executive information, the panel reports a mean CEO turnover rate of 25%, while the average forced turnover is equal to 17%. Noting that turnover measure is a
23 one-year construct, these turnover rates are in line with prior academic literature6. The average age is 56 years with the majority of CEOs ranging between 47 and 65 years old. This is similar with Engel et al. (2003) and Ghosh (2011) that reports an average of 56 years. The mean percentage of total shares owned by a manager is 2.32%, close to the value of 4.44% reported by Dikolli et al. (2014). The percentage is not high, so it does not point to an entrenched CEO. The statistics for firm characteristics are similar to Guay et al. (2014). The performance measures have low mean, for example the industry adjusted ROA is 0.1%. The reason is that the majority of the companies are asset intensive (the sample is dominated by manufacturing industry) so items such as property, plant and equipment frequently represents an important percentage of company’s total assets. This capital-intensive business carry a high large fixed asset base, thus leading to a lower rate of return. The sample seems to include more mid-cap firms having a mean (median) market capitalization around 9 billion (2 billion). These companies feature considerable growth options, since their market value of equity is about three times bigger than the book value of equity. The size (ln assets) is 7.565, which corresponds with an average company size (i.e. average assets) of 8.8 billion. The mean value for Herfindahl-Hirschman Index is close to zero (0.115) in line with Ghosh (2011) that reports a median of 0.041, indicating that there is a large number of firms competing. Finally, the market return volatility is higher than the accounting volatility (mean of 0.523, respectively 0.064). This can be explained by the conservative practices of accounting that reflects stability of asset prices (making them to be more accurate over the long run). In contrast, the market prices are affected by their noisy nature.
Table 1: Summary statistics
Table 1 reports descriptive statistics. The sample includes firms from merging two WRDS datasets: ExecuComp and CRSP/Compustat merged, with data from 2006 through 2014. Financial firms (SIC codes 6000 to 6999) and utilities (SIC codes 4900 and 4942) were dropped. The final sample yields 8,835 firm-year observations with non-missing data for all variables. Panel A reports descriptive statistics for firm and managerial characteristics, and Panel B reports for industry shocks. Panel C reports the composition of the sample over industries and years. Panel D reports Pearson correlation. Bold correlation coefficients are significant at the 10% level. All variables are defined in Appendix A.
6 Engel et al. (2003) report a rate of forced turnover of 13%, Huson et al. (2004) report 16% and Guay et al. (2014) find 10%. Looked at it another way, Kaplan and Minton (2012) report about the board-driven turnover that only 16.35% of CEOs in the place in 2000 remained CEOs in 2007.
24 Panel A: Managerial and firm characteristics
Variable Mean Std.Dev. 10% 25% 50% 75% 90%
Turnover 0.247 0.431 0.000 0.000 0.000 0.000 1.000 ForcedTurnover 0.171 0.377 0.000 0.000 0.000 0.000 1.000 ROA 0.001 0.127 -0.087 -0.027 0.011 0.052 0.099 RET 0.001 0.066 -0.062 -0.029 -0.002 0.026 0.061 LagROA 0.002 0.127 -0.085 -0.025 0.014 0.056 0.103 LagRET 0.000 0.067 -0.063 -0.029 -0.002 0.028 0.064 VolatilityROA 0.064 0.100 0.009 0.016 0.032 0.072 0.145 VolatilityRET 0.523 0.690 0.166 0.246 0.375 0.584 0.918 Size 7.565 1.631 5.608 6.458 7.462 8.593 9.738 Growth 2.944 3.460 0.893 1.353 2.145 3.439 5.526 Concentration 0.115 0.087 0.046 0.061 0.077 0.135 0.216 Age 55.843 7.154 47.000 51.000 56.000 60.000 65.000 EquityOwn 2.321 5.888 0.000 0.060 0.624 1.958 5.100
Panel B provides the sample statistics to the six industry shocks defined. For many of the shocks, the mean value approaches the 75th percentiles, which indicates that the distribution of these variables is right skewed. Shocks have an average that is below 18 %. This shows that they are not a disruptive change. Although, the sample spans the years affected by the financial crisis, probably it is expected to see larger shocks. However, companies from the financial sector that was most affected by the crisis were dropped from the sample.
Panel B: Industry shocks
Variable Mean Std.Dev. 10% 25% 50% 75% 90%
AssetGrowth 0.11 0.11 0.02 0.04 0.08 0.14 0.19 MB 0.10 0.13 0.01 0.04 0.07 0.12 0.21 Investment 0.17 0.16 0.02 0.05 0.11 0.25 0.35 Competition 0.12 0.21 0.00 0.02 0.05 0.10 0.22 Sales 0.11 0.12 0.02 0.04 0.08 0.13 0.25 Technology 0.16 0.19 0.03 0.05 0.10 0.19 0.32
Panel C provides the sample composition over industries and years. The sample is concentrated in the manufacturing (SIC 20-39: 60.61%). Regarding the years, the sample seems equally distributed (only the year 2014 has a minimum number of 431 sample observations to be analyzed).
25 Panel C: Sample composition over industries and years
Frequency of sample observations over industries
SIC Industry N (%) 00-09 Agriculture 19 (0.22%) 10-14 Mining 498 (5.65%) 15-17 Construction 179 (2.03%) 20-39 Manufacturing 5,339 (60.61%) 40-48 Transportation, communications 283 (3.21%) 50-51 Wholesale trade 381 (4.33%) 52-59 Retail trade 454 (5.15%) 70-89 Service 1,682 (18.8%) Total 8,835 (100%)
Frequency of sample observations over years
Year N (%) 2006 932 (10.55%) 2007 979 (11.08%) 2008 975 (11.04%) 2009 1,036 (11.73%) 2010 1,158 (13.11%) 2011 1,131 (12.80%) 2012 1,114 (12.61%) 2013 1,079 (12.21%) 2014 431 (4.88%) Total 8,835 (100%)
Panel D presents Pearson correlation among the main variables. From this univariate analysis, the following points deserve attention. First, all six proxies for industry shocks are correlated, the correlation is highest between the Sales and Asset Growth (Pearson correlation coefficient = 0.67). Performance measures (industry adjusted return on asset and industry adjusted market return) are as well positively correlated (0.25). Second, this univariate analysis indicates that the industry shocks are indeed positively associated with CEO turnover implying that dismissal does not relate solely to financial performance of the company. Third, consistent with prior literature, performance (both past and current) is inversely correlated with turnover. Finally, industry concentration (defined by HHI) is positively correlated with all of the shocks and also with CEO turnover. This is explainable by the fact that for companies operating in competitive industries board of directors can easily identify unfit CEOs. Board learns more quickly about abilities of their CEO in these industries, and thus can remove a poor CEO sooner (Engel et al.,
26 2003). Except for these, most of the other correlations are small in magnitude (the absolute correlation coefficients are less than 0.3), indicating that multicollinearity is not an issue for the multivariate analyses7.
4.2. Multivariate analyses
4.2.1. Industry shocks and CEO turnover
The findings from estimating the empirical model described by equation (1) are reported in Table 2. The regression model is estimated by means of a Probit model. The interest is in the sign of the coefficients on all the six shocks (β). Consistent with my hypothesis, the results suggests that all shocks have positive and significant effects on CEO forced turnover (p < 0.01).8 In terms of economic magnitude, after controlling for performance and other firm characteristic, three shocks (i.e. asset growth, competition and technology) have the largest incremental effect on forced turnover. In untabulated analysis, moving from lowest decile to the highest decile for each shock the probability for a CEO to be dismissed doubles. This signal that when an industry experiences a substantial shock, then the chances for a company to change its leadership are twice higher, than during a small shock.
7 However, a multicollinearity test diagnose is performed to mitigate the concern that the regression model estimates of the coefficients are unstable. The variance inflation factor (vif) command is used after the regressions. As a rule of thumb, the variables must have values smaller than 10, otherwise a further investigation is required. In the model, the variables used have a vif of 1.60. The mean vif is 1.19.
8 In untabulated analysis, the results are similar using Turnover as the dependent variable. Inferences are also robust to logit regression. Coughlan and Schmidt (1985) and Warner et al. (1988) use logit instead of probit regressions.
27 Panel D: Pearson correlation matrix
Variable 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 1. Turnover 1.00 2. ForcedTurnover 0.77 1.00 3. AssetGrowth 0.21 0.21 1.00 4. MB 0.09 0.09 0.51 1.00 5. Investment 0.04 0.04 0.47 0.27 1.00 6. Competition 0.29 0.26 0.54 0.28 0.36 1.00 7. Sales 0.12 0.12 0.67 0.32 0.61 0.52 1.00 8. Technology 0.18 0.21 0.46 0.27 0.33 0.47 0.44 1.00 9. ROA -0.09 -0.08 0.00 0.00 0.01 0.00 0.01 0.00 1.00 10. RET -0.07 -0.07 0.01 0.00 0.00 0.00 0.00 0.00 0.25 1.00 11. LagROA -0.06 -0.06 0.01 0.00 0.01 0.01 0.01 0.00 0.48 -0.07 1.00 12. LagRET -0.05 -0.04 -0.02 0.00 0.00 -0.01 -0.01 -0.01 0.17 -0.12 0.23 1.00 13. VolatilityROA 0.01 0.03 -0.01 -0.03 0.03 -0.01 0.01 -0.01 -0.19 0.06 -0.38 -0.01 1.00 14. VolatilityRET 0.02 0.03 0.01 -0.01 0.06 -0.03 0.03 0.00 -0.05 0.02 -0.05 0.25 0.23 1.00 15. Growth 0.03 0.01 0.01 0.00 0.00 0.01 0.01 0.03 0.04 0.00 0.02 0.01 0.02 0.00 1.00 16. Size 0.00 -0.02 0.02 0.04 0.02 0.05 0.05 0.05 0.20 -0.01 0.21 0.01 -0.26 -0.13 0.01 1.00 17. Concentration 0.10 0.09 0.25 0.28 0.25 0.24 0.27 0.33 0.00 0.00 0.00 0.00 -0.05 0.04 0.02 0.10 1.00 18. Age 0.11 -0.16 0.01 0.00 0.02 0.02 0.02 0.00 0.01 0.00 0.02 0.00 -0.05 -0.02 0.01 0.07 0.05 1.00 19. EquityOwn -0.04 -0.05 -0.01 -0.01 -0.01 -0.03 -0.03 0.01 -0.01 0.01 0.01 -0.01 0.00 0.01 -0.01 -0.18 -0.02 0.16 1.00
28 Further, prior studies show that dismissal is sensitive to performance, specifically that it has a negative relationship. The sign of the regression coefficients comport with previous literature indicating that a poor performance leads to increased chances that the executive leaves the company (Warner et al., 1988; Weisbach, 1988). This is true for both way of measuring performance, either by market or accounting return. With respect to other control variables, there is a positive relation between industry concentration and turnover (p < 0.01), pointing to a greater forced turnover in more competitive industries (Engel et al., 2003) and a positive relation between stock return volatility and turnover (p < 0.01), indicating increased CEO dismissal when companies face greater uncertainty (Bushman et al., 2010). In Panel B the analysis is repeated while also controlling for industry fixed effects. When analyzing dichotomous variable, the natural approach is to apply a nonlinear regression, such as probit or logit, but because these models cannot accommodate a large number of fixed effects, consistent with prior literature (Guay et al., 2014), I use a linear probability model. The model controls for industry fixed effects, since a shock that is common for an industry, can be less normal for another. Panel B reports findings similar to those in Panel A, with positive and significant effects for five out of the six proxies for industry shocks.
Table 2: The effect of industry shocks on CEO turnover This table reports the regression estimates of the following model: ForcedTurnover = α + β Shockt + γ Controls + ɛ
Shock is represented by the six industry-level shock proxies. Controls is a vector of control variables including ROA, VolatilityROA, LagROA, RET, VolatilityRET, LagRET, Size, Growth, Concentration, Age and EquityOwn. Variables are defined in Appendix A. The results of the regression are reported: non-linear model (Probit regression) in Panel A and linear model with industry fixed effects in Panel B. z-statistics, respectively t-statistics appear in parentheses. ***, **, * denotes 1%, 5% and 10% significance levels (two-tail).
29 Panel A: Pooled regression (1) (2) (3) (4) (5) (6)
Variable Pred. AssetGrowth MB Investment Competition Sales Technology
Shock + 2.34*** 0.69*** 0.21** 1.48*** 1.06*** 1.27*** (15.64) (5.99) (2.07) (19.95) (8.41) (16.01) ROA -0.42*** -0.41*** -0.42*** -0.38*** -0.42*** -0.40*** (-2.97) (-2.92) (-2.96) (-2.69) (-2.95) (-2.78) VolatilityROA -0.07 -0.05 -0.08 -0.14 -0.11 -0.06 (-0.41) (-0.27) (-0.43) (-0.74) (-0.61) (-0.37) LagROA -0.42*** -0.37** -0.39** -0.46*** -0.41*** -0.41*** (-2.79) (-2.46) (-2.57) (-3.05) (-2.70) (-2.74) RET -1.53*** -1.41*** -1.41*** -1.55*** -1.42*** -1.50*** (-5.77) (-5.43) (-5.44) (-5.83) (-5.47) (-5.71) VolatilityRET 0.06*** 0.06*** 0.06*** 0.08*** 0.06*** 0.07*** (2.97) (2.92) (2.69) (3.77) (2.74) (3.16) LagRET -0.92*** -0.98*** -0.97*** -0.95*** -0.92*** -0.95*** (-3.50) (-3.79) (-3.73) (-3.62) (-3.58) (-3.67) Size -0.01 -0.01 -0.01 -0.01 0.00 0.00 (-0.46) (-0.48) (-0.366) (-0.96) (-0.63) (-0.72) Growth 0.00 0.00* 0.00* 0.00 0.00 0.00 (1.56) (1.67) (1.66) (1.55) (1.62) (1.10) Concentration 0.70*** 1.18*** 1.41*** 0.55*** 1.06*** 0.43** (3.71) (6.34) (7.76) (2.90) (5.75) (2.24) Age -0.04*** -0.04*** -0.03*** -0.04*** -0.03*** -0.03*** (-14.90) (-14.63) (-14.71) (-15.16) (-14.77) (-14.57) EquityOwn -0.01*** -0.01*** -0.01*** -0.01*** -0.01*** -0.01*** (-3.14) (-3.22) (-3.23) (-2.75) (-3.05) (-3.68) Constant 0.85*** 0.93*** 0.94*** 1.03*** 0.93*** 0.91*** (5.37) (5.95) (6.04) (6.45) (5.95) (5.74)
Industry fixed effects No No No No No No
Observations 8,835 8,835 8,835 8,835 8,835 8,835
Pseudo R2 0.09 0.06 0.05 0.11 0.06 0.09
30 Panel B: Industry fixed effects
(1) (2) (3) (4) (5) (6)
VARIABLES Pred. AssetGrowth MB Investment Competition Sales Technology
Shock + 0.70*** 0.21* 0.05 0.46*** 0.32** 0.39*** (6.54) (1.95) (0.60) (6.13) (2.52) (5.90) ROA -0.11*** -0.11*** -0.11*** -0.10*** -0.11*** -0.10*** (-3.45) (-3.42) (-3.43) (-3.47) (-3.41) (-3.36) VolatilityROA -0.02 -0.01 -0.02 -0.04 -0.03 -0.02 (-0.58) (-0.39) (-0.57) (-0.96) (-0.84) (-0.74) LagROA -0.11*** -0.10*** -0.10*** -0.11*** -0.11*** -0.11*** (-3.66) (-3.39) (-3.64) (-3.62) (-3.59) (-3.90) RET -0.38*** -0.37*** -0.37*** -0.38*** -0.37*** -0.38*** (-5.66) (-5.51) (-5.50) (-5.70) (-5.58) (-5.73) VolatilityRET 0.018*** 0.019*** 0.018*** 0.023*** 0.017*** 0.020*** (3.48) (3.47) (3.51) (4.51) (3.17) (3.55) LagRET -0.23*** -0.26*** -0.25*** -0.24*** -0.24*** -0.25*** (-4.25) (-4.77) (-4.73) (-4.42) (-4.30) (-4.51) Size -0.001 -0.002 -0.002 -0.003 -0.002 -0.002 (-0.80) (-1.00) (-0.93) (-1.24) (-1.15) (-0.95) Growth 0.0002 0.0002 0.0002 0.0002 0.0002 0.0001 (1.45) (1.53) (1.54) (1.50) (1.53) (1.11) Concentration 0.19* 0.31** 0.38** 0.14 0.28** 0.11 (1.73) (2.21) (2.39) (0.90) (2.27) (0.85) Age -0.008*** -0.008*** -0.008*** -0.008*** -0.008*** -0.007*** (-19.41) (-16.23) (-16.65) (-18.57) (-17.30) (-17.02) EquityOwn -0.001*** -0.001*** -0.001*** -0.001*** -0.001*** -0.001*** (-3.94) (-3.88) (-3.94) (-3.02) (-3.41) (-3.73) Constant 0.53*** 0.57*** 0.58*** 0.57*** 0.57*** 0.55*** (15.03) (14.22) (15.57) (11.82) (12.73) (14.07)
Industry fixed effects Yes Yes Yes Yes Yes Yes
Observations 8,835 8,835 8,835 8,835 8,835 8,835
R-squared 0.08 0.05 0.05 0.11 0.06 0.08
31 4.2.2. Effect of industry shocks on the relation between turnover and performance
Table 3 shows the results for the second hypothesis. The model follows Table 2, except that two interactions are included (between the performance measures – both lagged and current -and industry shocks). By allowing the effects of current -and past period performance to vary with the shocks it can be inferred how shocks alter the board’s assessment for their CEO performance.
The shocks have positive main coefficients, indicating that they drive turnover. However, the coefficients for the interaction variables are insignificant. More precisely, findings show that shocks are unrelated to the sensitivity of turnover to current and lagged period accounting performance. So, the results do not suggest that boards assign a greater (smaller) weight to current (lagged) performance for the assessment of the CEO ability. Overall, the findings show that boards condition their dismissal decision just on poor performance, and do not assign more or less weight to the performance measure outcome in determining CEO turnover when it occurs contemporaneously with an industry shock.
Table 3: The effect of industry shocks on turnover-performance sensitivity This table reports the regression estimates of the following model:
ForcedTurnover = α + β Shockt + φ1 ROAt* Shockt + φ2 LagROA * Shockt + φ3 RETt * Shockt + φ4 LagRET * Shockt + φ5 ROAt + φ6 LagROA + φ7 RETt+ φ8 LagRET + γ Controls + ɛ
Results are reported using a probit regression. Model specification follows Table 2, except that both past and current performance measures interact with the shocks. Variables are defined in Appendix A. For parsimony, I report only coefficients on shock and both performance measure and their interactions with the shocks. z-statistics appear in parentheses. ***, **, * denotes 1%, 5% and 10% significance levels (two-tail).
(1) (2) (3) (4) (5) (6)
Variable Pred. AssetGrowth MB Investment Competition Sales Technology
Shock 2.32*** 0.68*** 0.20** 1.47*** 1.05*** 1.26*** (15.43) (5.83) (1.97) (19.62) (8.28) (15.72) ROA*Shock - 0.86 0.62 1.07 0.68 1.53 0.14 (0.54) (0.55) (1.04) (0.92) (1.20) (0.16) LagROA*Shock + 1.58 0.71 0.77 0.89 0.46 -0.06 (1.03) (0.53) (0.75) (1.00) (0.36) (-0.06) RET*Shock - 3.63 -1.45 -0.39 2.41 -0.95 1.36 (1.19) (-0.82) (-0.22) (1.45) (-0.40) (0.93) LagRET*Shock + -2.02 -2.83 -3.53* -1.36 -1.48 -3.17** (-0.78) (-1.30) (-1.95) (-0.92) (-0.71) (-2.29) ROA -0.49** -0.47*** -0.55*** -0.49*** -0.59*** -0.41** (-2.48) (-2.77) (-2.81) (-2.89) (-2.99) (-2.28) LagROA -0.60*** -0.45** -0.54** -0.55*** -0.46** -0.44** (-2.82) (-2.32) (-2.40) (-3.13) (-2.30) (-2.29)
