The performance of first-year CEOs

(1)

1

The performance of first-year CEOs

Name Hidde Roeloffs Valk

Number 10623795

MSc in Business Economics Specialization Finance

Supervisor dhr. dr. I.J. Naaborg Completion 8th_{of December, 2014}

(2)

2 Abstract: CEOs are considered the head of the firm and are accountable and responsible for the firm, its strategy and its performance. With increasing pressure on CEO by boards of directors and shareholders, they are expected to deliver quick results after they have taken their seat into the office. Deteriorating firm performance is one of the most important reasons for a CEO to be fired. This thesis shows that a first-year CEO is able to make a positive impact on firm performance. CEOs that were promoted from within the firm seem to be more capable of affecting return on assets and Tobin’s Q, while outside CEOs do not affect those measures of performance. The exact mechanics behind this performance increase remains unknown. Several measures of investments are utilized to check whether the measures change simultaneously due to the first year in office. None do significantly change in the first year in office of the CEO. Unobserved efficiency increases or managerial skills could be attributed to the increase in firm performance. Future research could investigate what a first-year CEO precisely does in, such as firing or hiring certain managers and how this relates to the increase in firm performance.

Keywords: CEO, turnover, influence, ability, succession, performance, investments, outsider, insider,

RoA, horizon

(3)

3

The chief executive officer (CEO) is seen as the key person in firms, which have taken on an essential role in the modern society and the worldwide economy. A competent CEO is able to revive the firm, if near bankruptcy, and steer towards better weather. However, the lack of management skills and vision could also steer the company towards bankruptcy (Hirshleifer, Low, & Teoh, 2012). The owners of a company, the shareholders, hold a CEO accountable and responsible for the performance of the firm. The typical CEO defines the policies of the firm for seven to ten years (Parrino, 1997). Moreover, average tenure of CEOs is shorter than ever, according to the latest statistics it is less than six years (Kaplan & Minton, 2012). This could increase the pressure on CEOs to perform in a relative short time (Antia, Pantzalis, & Park, 2010).

There is an abundance of variables concerning the decision of hiring a new CEO (Parrino, 1997), however, when a new CEO is in office, he has the incentive to deliver results. His own job is always on the table. The first year in office of a CEO could therefore set the tone for the future of the firm and its performance. To what extent and how a CEO is able to influence firm performance in his first year has not been researched yet. This thesis will bridge the gap between CEO succession and firm performance. The goal of this thesis is to evaluate the impact of CEO succession on firm performance via investments, leverage and acquisitions. The research question of this paper is as follows:

‘Is a CEO able to make a positive impact on the firm’s performance in his first year in office?’

The research question is relevant for all stakeholders, especially shareholders and the boards of firms. The expectation of quick results is embedded in the culture of firms from the US (Fiordelisi & Ricci,

(5)

5 2014). Thus, is it reasonable to expect performance enhancement in the first year after a turnover? This research will try to answer the question via the use of a regression method with panel data. A sample of CEO turnovers from S&P1500 firms is used as the basis for this analysis. The sample contains observations from the period between 1992 and 2012. As an additional robustness check, an event study is implemented to see whether the stock of the firm gives cumulative abnormal returns in a 1-2 year period after the CEO has taken his place into office

The thesis is structured as follows: in chapter two, the existing literature concerning this topic will be reviewed. The relationship between CEO turnover, succession and firm performance will be explained in more detail. Building upon this, hypotheses related to the research question are introduced. In the third chapter, the data collection and sample building behind this research is presented. In chapter four, empirical results will be analysed and discussed, including robustness checks with the event study. Chapter five is the conclusion of this thesis.

2. Literature Review

This chapter gives an overview of the relevant literature on the relation between CEO turnover (succession) and the related firm. It is a broad subject covered in the field corporate finance and governance, as well as the field of behavioural finance. This section is divided into five subsections. The first section describes the literature on CEO turnover and its relationship with firm performance. It will discuss how the probability of a CEO turnover increases when firm performance decreases. Moreover, evidence from the Dutch stock market’s reactions surrounding a CEO turnover will be discussed. This thesis will latter focus on the question if, and to what extent, CEOs are able to influence a firm’s stock performance in the US stock market, namely the S&P1500, specifically in their first year in office. The third section will focus on the decisions made by a CEO surrounding the turnover. This is helpful in building the hypotheses and the methodology. The fourth section it is discussed whether a CEO is able to influence the firm and its policies. Specifically, researchers try to identify a ‘CEO-effect’ via empirical methods. In the last section, hypotheses will be stated and discussed.

(6)

6

2.1. CEO turnover and firm performance

The relationship between CEO turnover and firm performance is twofold. On the one hand, a firm’s (accounting and stock) performance influences the probability of a CEO turnover. On the other hand, a CEO turnover can have an impact on a firm’s stock performance. Both relationships will be discussed in this section.

Existing studies examine various financial measures surrounding CEO turnover. Most early studies have documented that CEO turnover is preceded by declining share-price and earnings performance (Warner, Watts, & Wruck, 1988; Weisbach, 1988). Each of these studies is focused on one financial variably solely, in the period surrounding the turnover of a CEO. One main hypothesis is seen throughout most of the literature that use the probability of a CEO turnover as the dependent variable in their methodology. The main hypothesis, which is discussed in this subsection, states that the probability of a CEO turnover is inversely related with financial variables such as stock price and accounting based performance, investments and R&D expenses (Warnet et al (1988), Weisbach (1988). In other words, if, stock prices are decreasing, the probability of CEO or top management change increases. Evidence shows CEO turnovers are preceded by poor firm performance in the years preceding it. A general model concerning this hypothesis follows from Weisbach (1988) and Warner et al (1988), where f denotes a general function and u is a random error term.

Probability (CE0 turnover (t)) = f (Firm performance,, CEO age,, . . .) + u(t)

Logically, the precision in the calculation and the choice of the performance measure on which the CEO is evaluated upon by the board of directors is crucial in the decision of the firm regarding the turnover. For example, the operating return on assets might be low in absolute terms. Comparing it with the industry mean or median, it might be relatively high. If said CEO is fired or forced to resign based on an absolute measure of performance, it might have not been a good decision. Thus, the measure used in empirical research might not be the measure that is affecting the true probability of a CEO turnover (Parrino, 1997). In a later research, Huson, Parrino & Starks (2001) showed that the probability of a CEO turnover is small after bad firm performance. In their sample, when a given firm sits in the lowest quartile of firm performance (-3.7% operating return on assets), the probability that the CEO is fired is

(7)

7 only 2.8%. This gives rise to the idea that CEOs are not simply evaluated upon one measure of firm performance, but possible other measures and other factors, which determine the success of a CEO.

A research conducted by Warner et al (1988), published in the same issue as Weisbach (1988), specifically investigates the stock price movements of firms and the following changes in top management. Here, top management is defined as the CEO, the president or the chairman of the board. Firms in the sample are randomly selected firms from the New York Stock Exchange (NYSE) in the period of 1963 to 1978. The authors argue that their research provides new evidence on the internal mechanisms behind maximization of shareholder value by removing inefficient managers. In contrast to earlier studies like Weisbach (1988), this study took three mechanisms of corporate control into consideration. These three mechanisms are: board monitoring, mutual monitoring between managers and monitoring by large block holders (Shleifer & Vishny, 1986). In addition to a regression method, Warner et al (1988) also implemented an event study. Results showed that on the day the CEO change is announced, the mean excess return is not distinguishable from zero, but that the variance of the excess return increases on the event day. This finding implies that an announcement of a CEO turnover contains valuable information but that the information is good news for some firms and bad news for others. This research shows support for the main hypothesis: the probability of top management changes is inversely related to the stock performance. Moreover, it shows that information about management is reflected in the stock price and that such information is used in evaluation of top management (Warner et al, 1988). Thus, a CEO is significantly able to influence the firm’s stock performance and is evaluated upon that. However, there seems to be a lag of up to two calendar years before a firm’s reaction to decreasing share performance with a turnover of its CEO. In contrast to Warner et al (1988), Weisbach (1988) shows evidence, via an event study, of all CEO turnover announcements resulting in positive excess returns, even in the case of anticipatable retirements. An explanation for this could be sample selection bias, according to the author. Weisbach (1988) also used a random sample from NYSE firms but for a different time period, namely 1974 to 1983, in the 1-3 days surrounding the announcement of the turnover. This creates the probability that different samples lead to different outcomes of a similar

(8)

8 research. The empirical method of the event study used in the robustness check is similar to the methods of Weisbach (1988) and Warnet et al. (1988) but with an increased time period: one to two years.

Murphy & Zimmerman (1993) brought novelty to this area of research by examining a variety of financial variables in a time horizon of ten years, surrounding CEO turnovers. This is in contrast to the earlier studies, such as Weisbach (1988) and Warner et al (1988) who focused on one single financial variable and thus offer only explanation concerning one specific variable. Murphy & Zimmerman (1993) add variables such as R&D, advertising expenses and capital expenditures, because those are more under the influence of the firm and thus more directly controllable for the CEO. The authors extract a random sample of 1000 CEO turnovers from Forbes 500 firms in the period of 1971 to 1989. The authors argue that firm’s stock performance is more likely to reflect macro-economic conditions and thus the CEO should not be evaluated upon that by the board. Furthermore, Murphy & Zimmerman (1993) test the existence of three non-mutually exclusive problems concerning CEO turnover. First, a departing CEO that expects his department (such as a retirement), could make accounting or investment decisions aimed at increasing accounting based earnings. This is also known as the decision horizon problem. Second, active CEOs in poor performing firms could also make such decisions to cover up the firm’s health, known as the cover-up problem. Third, incoming CEOs could boost future earnings at the expense of transition year earnings by writing off investments with a negative return, known as a ‘big bath’. This leads to the conclusion that CEOs have the power and incentives to manipulate a firm, and its performance, to their own good. This is consistent with the classic principal/agent theory. These three problems are taken into account in the building of the empirical methodology in chapter three. To conclude, Murphy & Zimmerman (1993) show a mechanical and structural link between certain expenditures and CEO turnover because of the horizon problem. In contrast to earlier studies, the authors show evidence that the decline in certain expenditures before a turnover is more explained by the decline in firm performance.

In a more recent paper by Antia, Pantzalis, & Park (2010), the horizon problem described by Murphy & Zimmerman (1993) is shown via the tenure of a CEO. The authors use a sample of S&P 1500 firms in the period of 1992 to 2003. Via a multi-factor regression analysis, the authors provide

(9)

9 evidence that longer tenure should lead to higher market valuations. In addition to the increase of accounting performance, as evidenced by Murphy & Zimmerman (1993), this problem creates agency costs at the expense of long-term performance. Thus, a shorter CEO tenure is related to higher agency costs and could lead to a decline of the firm’s value. This should concern shareholders because average CEO tenure is decreased in the last decades to an average of less than six years. One of the reasons for this fact might be that boards respond more frequently and aggressively to deteriorating firm performance. Boards act in response due to a difference in their decision horizons. CEOs are limited in their horizon because of their relative tenure, while stakeholders are affected by the firm for its lifespan. Even stock options compensation, with the goal to lengthen the decision horizon, seem to shorten the CEO’s decision horizon and lead to a negative effect on firm performance. (Antia et al, 2010).

Contrary to the literature above, which is focused on CEO turnover in relationship with the firm’s own performance, such as return on assets or stock price. A large body of literature is focused on the relative performance of firms to peers, such as the firm’s industry in which it operates. Kaplan & Minton (2012) focus on this aspect and show empirical evidence via a sample of all Fortune 500 firms in the 1992 to 2005 period. The negative relationship of firm performance with CEO turnover also holds for relative firm performance, for all types of turnovers except for external turnovers. In an external turnover, a new CEO is hired from outside the firm. These almost only occur in the case of a hostile takeover, in which a reason of the takeover is a growing firm performance. Moreover, if an industry performs poorly, but the firm does not, it could be efficient for a firm to bring on a new CEO, in order to respond to the changing industry. To conclude, CEO turnover seems to be more sensitive to stock performance rather than accounting based performance measures, such as return on assets. The methodology in this paper will use both measures of firm performance.

Jenter and Kanaan (2014) have recently proposed a novel approach that splits firm performance into two components: systematic and firm specific. The authors build their sample on firm year observations of S&P1500 firms from 1993 to 2001 and test whether industry performance explains the probability of a CEO turnover. Via a two-stage least squares regression method, with the industry mean performance as the instrumental variable, the authors show the link between the industry’s performance

(10)

10 with the firm’s stock performance and the probability that a CEO turnover occurs. Their research shows evidence of CEOs being significantly more likely to be fired from their jobs after poor industry or poor stock market performance. The authors argue that there are factors beyond the control of CEOs on which they are evaluated on. Jenter & Kanaan (2014) suggest that the current standard CEO hypothesis does not capture the full relationship between firm performance and CEO turnover. The methodology of this thesis will include industry-adjusted measures of firm performance because of the seeming relevance for the CEO. Moreover, there seems to be a high degree of luck involved in some industries, such as the oil industry according to Bertrand & Mullainathan (2001). Oil prices are determined by the market and a CEO benefits from high oil prices via higher profits, but that is more a matter of luck than an effect of the CEO.

Evidence from an event study on the Dutch stock market showed no significant effect on stock prices after a CEO turnover announcement (Cools & van Praag, 2007). The event study investigates if there are abnormal returns surrounding the announcement dates of the CEO turnover of firms listed at the Amsterdam Stock Exchange from the period of 1991 – 2000. The event window that is used is ten days before the announcement and one day after. The most important event window is the day of the announcement, since market efficiency suggests that all information should be priced immediately by the market into the stock price (Cools & van Praag, 2007). A possible explanation is given by Warner et al (1988): the effect on stock returns at the announcement of a CEO change is the sum of two separate effects. One positive effect is called the real monitoring effect, resulting from an unanticipated change of the CEO, in the interest of shareholders. The other effect is negative, called the information effect and follows from a signal that management is currently poor at managing the firm. With these two effects taking place, a clear abnormal return on the event date might not always be proven. In contradiction to this event study, Weisbach (1988) reported significantly positive stock returns at a three day window surrounding the CEO announcement in a random sample of NYSE firms. Event studies seem to be a powerful and widely used method in evaluating the impact of a CEO turnover and will thus be used, with a longer horizon, as a robustness check in chapter four.

(11)

11 To conclude, it is a widely proven hypothesis that the probability of a CEO turnover is inversely related to firm performance. However, the measurement of performance matters. Either it could be absolute performance (return on assets) or relative performance to either the market or the industry. Overall, this hypothesis holds with all measures of performance but differs in sensitivity in relation with the probability of a CEO turnover. Certain measures of firm performance are more under the influence of the CEO, such as accounting-based performance, as illustrated by the horizon problem. Boards might fire a CEO more easily of the stock price of the firm goes down, relative to a decrease in return on assets. The sensitivity of the probability of a CEO turnover therefore differs between measures of firm performance. Moreover, boards are more eager to replace a CEO than three decades ago, therefore increasing the pressure on the shoulders of a CEO. This thesis wants to extend on that and investigate how CEOs deal with the high pressure in their first year in office, by investigating if CEOs are able to affect firm performance in their first year in office and how investments and leverage are affected.

2.2. CEO origin

One of the important debates within this area of corporate governance research, is whether it matters if a CEO is promoted from within the firm or hired from outside the firm, e.g. does the origin of the CEO matter? Zhang & Rajagopalan (2010) described the origin of the CEO as follows:

’’The origin of the CEOs is hence expected to influence their ability to formulate and implement strategic change, and therefore to also affect the relationship between the level of strategic change and firm performance’’ (Zhang & Rajagopalan, 2010, page 2)

Parrino (1997) argues that the cost of outside succession matters in the probability of a CEO turnover. The author bases its analysis from a sample of directors that are reported in the Forbes annual compensation surveys between 1971 and 1989. Firm data is extracted from the S&P1500, via the Compustat database. Information about the CEO’s origin is collected by hand research from Wall Street Journal. Regression analysis is used to investigate the research question. There is no straight-forward answer for the debate, the boards should consider its strategy when choosing a new CEO. Insiders are more attractive when the firm needs to continue its current strategy. Outsiders are the most desirable option when the firm wants to change it strategic direction. The availability of an outside candidate is

(12)

12 an important factor in the determination of a board to replace a poor CEO. Moreover, there is a strong negative relationship between relative firm performance (the author uses ROA as the performance measure) and the probability of an outsider getting hired as the new CEO. In other words, if the firm performs badly in relation to the industry, an outside CEO is more likely to be appointed. The relationship holds for both forced and unforced turnover. However, it seems to matter if the industry is homogeneous or heterogeneous. Homogeneous industries are industries in which products are similar for all firms, such as cigarettes. Heterogeneous industries have products that differ more, in the eyes of the consumer, such as smartphones. It appears to be less costly to replace CEOs in homogeneous industries (Parrino, 1997). Logically, outside directors from the same industry are familiar with the production technology of the firm and the product markets. In Hornstein’s (2013) paper about investment decisions concerning CEO turnover, the author found no performance difference between outsider and insider CEOs. The author concludes that comparing the efficient capital budgeting decisions made by a CEO and its firms, via its marginal q, the origin of the CEO does not matter. This paper will be discussed in more detail in the next subsection. Kaplan & Minton (2012) argue that most outsider successions are the result of an acquisition. The authors base their result on a large sample of Fortune 500 firms in the 1992 to 2005 period. In their paper, it is shown that the hypothesized negative relationship between stock performance and outsider turnover does not hold. A reason might be that when a firm has good stock performance, it could be a target to an acquiring firm which replaces management with a golden parachute.

To conclude this subsection, there is mixed evidence concerning the outsider or insider CEO discussion. The question remains largely unsettled and seem to depend on the situation of the firm and the industry, according to Parrino (1997).

2.3. CEO decision making surrounding turnover

In the following section, an analysis about CEO behaviour and decision making surrounding the turnover will be given. The methodology of this thesis will partly be based on CEO decisions that could

(13)

13 have an effect on firm performance. These decisions are related to capital expenditures, leverage and acquisitions.

As described in the previous section, there are some problems concerning CEO turnover, such as the horizon problem, the cover-up problem and the big bath problem (Murphy & Zimmerman, 1993). All problems relate to decisions made by a CEO prior or after turnover. The main idea behind these problems are the CEOs that are acting differently, because of the turnover. The horizon problem seems to be the most important and it is described extensively. Dechow and Sloan (1991) provide empirical evidence for the horizon problem. The evidence shows that a CEO is more likely to cut R&D expenses before leaving the firm, thereby increasing accounting based earnings and possibly their personal wealth, if their compensation structure is dependent of the height of accounting based earnings. This problem is most severe when CEOs approach their retirement. A reason might be the fact that a CEO exits the labour market when he retires. In other words, a CEO becomes more ‘myopic’, in the sense that they tend to place less weight on cash flows occurring after their employment time horizon. Antia et al. (2010) describe the consequences of this problem. One consequence is the fact that compensation structure is negatively related to a CEO’s decision horizon, therefore worsening the horizon problem. Compensation structures with a large degree of equity or stock options, create incentives to boost short-term performance. In turn, this could lead to lower firm performance on the long short-term. Slightly alleviating this problem might be the fact that CEOs are able to take a number of (supervisory) boards, at other firms, after their retirement. A CEO should have a good reputation before asked to do so (Brickley, 2003). This could create an incentive to create long-term value.

Hornstein (2013) examined patterns in the efficiency of corporate capital budgeting surrounding CEO turnover. The author continues on the assumption made by Bertrand & Schoar (2003), in the sense that CEOs differ in their decisions concerning investments and corporate capital budgeting. In the paper, the marginal q is used to determine if a firm is under or overinvesting. The estimation approach of the marginal q, developed by Durnev et al. (2001), can be defined as: ‘the ratio of the

unanticipated incremental change in firm market value divided by the contemporaneous marginal investment. Thus, an optimal capital budgeting process would be one where a firm invests until the last

(14)

14

investment generates a marginal q of 1.0. A positive (negative) deviation of estimated marginal q from 1.0 would thus reflect under- (over-) investment.’ (Hornstein, 2013, page 42) Through this approach,

the author reveals several implications via a six year window surrounding CEO turnover. The sample is constructed from the Compustat database, which holds both CEO and firm year information, concerning the period 1992 – 2005. The most important implication that is to be taken from the research of Hornstein (2013) for this thesis, is the evidence that firms tend to overinvest prior to a forced turnover, due to agency problems such as empire building or shirking. On average in the transition year, many firms cut down their investments. After the turnover, the expenditures increased tremendously. Moreover, the marginal q in the two years after the turnover have a mean value above one, which shows evidence for underinvestment after a turnover. Consequently, we could expect an increase in expenditures in the first year of CEO turnover in the presentation of the empirical results in chapter five. A reason might be that there are less agency costs due to the replacement of the CEO, therefore resulting in more efficient investments (Hornstein, 2013). Another interesting suggestion of the data shown by Hornstein (2013), is that firms that hold down their investments in the first three years after the CEO turnover, are the firms with the most inefficient capital budgeting decisions. In other words, firms seem to quickly learn from their inefficient investment decisions by reducing investments, even if they have replaced their CEO.

Another important strategic decision for a CEO to create value is a decision about mergers and acquisitions (M&A). M&A are investments that have the potential to create large value impact and are easily observable by the market (Zhao, 2013). Morck, Shleifer, & Vishny (1990) found evidence that acquisitions are systematically overpaid by CEOs. One reason might be overconfidence, as extensively described by Morck et al (1990) and Malmendier & Tate (2005). Specifically, CEOs overestimate the returns and overpay for acquisitions. Moreover, CEOs would rather finance internally than externally, because internal funds is more easily available. Roll (1986) was one of the first to describe the phenomenon of overconfidence, relative to M&A, in his famous Hubris Hypothesis. The hypothesis states that the valuation of a target is the key element in a takeover. The valuation is a subjective process and it reflects individual decisions. Thus, valuations of acquisitions are subject to errors and

(15)

15 subsequently, overpaying is the result (Roll, 1986). Another view of overpayment is that the managers of a bidding firm prioritize personal goals rather than the shareholder’s goals (Morck et al, 1990). This is more in line with the principal-agent theory, where managers might not always act in the interest of shareholders. It has not been researched yet if incoming CEOs undertake acquisitions in their first year and if this pays off. This thesis wants to bridge that gap.

To conclude, recent evidence shows that activity are changing around and are in response to CEO turnover. Moreover, CEOs could be overconfident concerning their ability to acquire firms efficiently, and thus overpay which decreases firm value and performance. Also, all decisions are subject to agency problems and could therefore be costly for the shareholders of a firm.

2.4. The ‘CEO effect’

Whether the CEO even matters, is an important question for this thesis. First, it is important to know if the CEOs is able to influence the corporate strategy and investment decisions. In turn, these decisions could lead to an effect on firm performance. Thus, this section will look at the empirical research surrounding the question whether CEOs have an effect on strategic decisions (such as investments and leverage) and firm performance.

There is evidence of a relationship between a CEO’s and strategic changes made by a firm (Jensen and Zajac, 2004). Mackey (2008) shows in his study that there is a significant ‘CEO effect’ in firm performance, suggesting that CEO’s do matter. Beatty and Zajac (1987) study shows CEO succession effect and evidenced a significant influence of a CEO in the firm’s investment decisions. Thus, both theoretical rationale and empirical evidence lend support to the idea that the CEO is a firm's executive leader, and thus is able to have a significant effect on the firm’s strategy.

According to Wasserman, Nohria, & Anand (2001) a CEO is able to significantly influence the firm’s performance directly. To a certain, yet high, degree the CEO matters in delivering the firms performance, both market and accounting based measures. Roll (1986) was one of the first to recognise that managers are not completely rational and differ in their valuations of target firms. In a similar vein, Bertrand & Schoar (2003) argue in their paper that CEOs do matter and are able to distinguish

(16)

16 themselves from each other via their own ‘style’. Most importantly, CEOs have a clear effect on firm policies such as leverage and investments, which is important for our hypotheses. Bertrand & Schoar show this empirical relation by looking at CEOs and how they influence firms. With a sample from Forbes 800 firms and Execucomp data from 1992 to 1992, the authors track the decisions of several CEOs across multiple firms. The average length of tenure is just higher than five years in their data.

The authors argue that early empirical studies surrounding CEOs contain a quite narrow perspective. The neoclassical view, which the authors refer to, assumes implicitly that CEOs are homogeneous and unselfish inputs for the firm’s productivity. Managers and CEOs, in this view, are perfect substitutes. Thus, it doesn’t matter for corporate decisions which CEO is steering the firm. Two firms with exactly the same characteristics will make the same decisions. This perspective seems out-dated due to the contributions from the behavioural finance field, which assume that CEOs are not perfect substitutes (Bertrand & Schoar, 2003).

Beatty and Zajac (1987) argue that the succession/performance relationship is a function of two distinct, complementary concepts: manager effects and succession effects. The authors suggest that announcements of CEO changes are typically associated with a reduction in the value of the firm, as reflected in the perceptions of the stock market, and that CEO successors tend to significantly influence the production and investment decisions of their firms. These results hold for both insider and outsider CEO turnover.

A recent study by Bennedsen et al (2006) shows evidence of family firms that appoint family members as new CEOs, have a negative impact on firm performance. The authors use a unique dataset from Denmark, because Denmark has high percentage of family firms. The authors utilize a regression method that looks for the difference in performance after and before the turnover. This thesis will use a similar method. The negative effect that is reported by the authors is quite large: at least four percentage points on operating profitability. The authors conclude that non-family CEOs have a larger, positive effect on firm performance than family CEOs. Moreover, CEO age does not affect firm performance in the research of Bennedsen et al (2006) but as evidenced by (Fiordelisi & Ricci, 2014), CEO age does matter in corporate culture.

(17)

17 However, there could also be a factor affecting firm performance that involves luck and should not be attributed to the skills and decisions of the CEO. Bertrand (2001) showed evidence that CEOs of oil industry firms are benefitting from the oil price, which they are not able to control. If the oil price is high, performance is high and the opposite also holds. Unfortunately, controlling for such implications of luck is difficult for other industries and might not even be possible to do accurately, according to Bertrand (2001).

To conclude, it is safe to assume that a CEO is able to influence firm performance. However, the research in this thesis wants to explore if, and to what extent, a CEO is able to influence the firm in his first year in office.

2.5. Hypotheses

In the last subsections, the driving forces behind CEO turnover and succession have been described. Based on the literature, this research will focus on the first year in office of a CEO and if he/she is able to influence firm performance. This subsection contains the hypotheses central to the analysis. CEOs are expected to influence firm performance and have an effect on the firm. This relationship will be tested for its first year in office. The distinction between forced and unforced turnovers is not incorporated in the scope of these hypotheses and methodology, as data on this distinction must be hand-collected. However, outsider CEOs are more likely to be appointed in the case of a forced turnover, according to Parrino (1997).

Hypothesis 1: New CEOs have a positive effect on firm performance in their first year in office It is clear from multiple existing literature, such as Bertrand & Schoar (2003) and Wasserman, Nohria, & Anand (2001) that a CEO matters for firm performance. Moreover, there is increasing pressure on CEOs due to shorter average tenure and increasing pressure from supervisory boards that are sensitive to performance (Kaplan & Minton, 2012). Section 2.5 described this ‘CEO effect’ in detail. However, these studies document different periods of time surrounding the turnover, e.g. several years via a regression analysis or only days via an event study. This thesis will use both, with a focus on the first full year in office. The hypothesis stated above is aimed at proving a ‘CEO effect’ in the first year of

(18)

18 tenure. Hypothesised here is that CEOs have an effect, either negative or positive on first year firm performance.

Hypothesis 2: CEOs that are promoted from within the firm have a larger positive effect on firm performance in their first year in office

One of the largest debates in the literature concerning CEO turnovers is the question which is better: a CEO that has experience within the company or a CEO from outside of the company that has more diverse experience and brings in a fresh perspective or strategy. As described in section 2.2., there is no straight-forward answer. According to Zhang & Rajagopalan (2010), the origin of the CEO (outsider or insider), matters in developing and implementing strategic change and therefore affects firm performance. As Parrino (1997) points it, the industry in which the firm operates seems to determine whether a new CEO is an insider or outsider. It is hypothesised here that inside CEOs are better able to influence firm performance in their first year in office. Reason for this might be the firm-specific knowledge and certain efficiency or managerial problems which the new CEOs has learnt from earlier positions within the firm. Certainly in the first year in office there could be changes implemented by the new CEO that have a substantial and relatively fast effect on the firm’s performance. Moreover, outsider CEOs are more likely to be hired when a firm is in a period of bad performance, thus it is harder to turn around for such a CEO(Fee & Hadlock, 2003). This creates the expectation that insider CEOs are more able to positively influence firm performance in their first year in office relative to outsider CEOs.

Hypothesis 3: There is a relationship between CEO age and firm performance

According to Bennedsen et al (2006), CEO age does not affect firm performance. This was based on a dataset from Denmark, where this relationship might be different. As shown by Fiordelisi & Ricci (2013), the age of the CEO matters in defining the corporate culture. This thesis will use data on firms from the US with international CEOs. Higher age CEOs might have more experience concerning management, the industry and education and a larger influence on corporate culture.

(19)

19 As described earlier in the literature review, Hornstein (2013) showed evidence that investments decreases prior to CEO turnover and halt during the turnover and after that, are reverting to their normal levels. Agency costs might be causing this variance in investments. Bertrand & Schoar (2003) argue that capital expenditures can be seen as internal investments to increase the internal growth of the firm. Thus, a CEO willing to increase investments should have a long decision horizon and this could lead to increasing firm performance due to lower agency costs of this longer horizon (Murphy & Zimmerman, 1993). If both hypothesis one and two are proven, it might be evidence of the big bath problem, in which a new CEO takes on temporary expenses in its transition year to boost earnings in the subsequent years. Hypothesised here is the expectation of first year CEOs increasing investments.

Hypothesis 5: CEOs are less likely to acquire firms in their first year in office

Acquisitions

Another important driving force for the successor of a CEO might be the decisions made about acquisitions of other firms. As Bertrand & Schoar (2003) argue, acquisitions can be seen as a way to develop external growth, in contrast to capital expenditures which is seen as internal investments. Acquisitions are aimed at increasing the revenues of a firm. However, at most they have a slightly positive return for the acquiring firm and mostly negative returns, according to Morck (1990). Malmendier & Tate (2005) agree on this and add that overconfident CEOs are more likely to make acquisitions that lead to negative returns. Thus, it is hypothesised that CEOs are will engage in less acquisitions in their first year in office.

3. Methodology & Data

This section will discuss the methodology and sample data involved in answering the hypotheses and the research question. First, the methodology is explained in detail, including dependent and independent variables and why those specific variables should be included, according to the literature. Second, sample construction and descriptive statistics are discussed. The data is extracted from similar databases as the papers mentioned in the literature review: Compustat and Execucomp.

(20)

20

3.3. Methodology

The methodology used in this thesis, is a panel data regression with fixed effects at CEO and year levels, aimed at proving the research question. This method is most appropriate since time-series data is used and fixed effects help to deal with omitted variable bias, by controlling for unobserved variables between CEOs. It is also known as manager fixed effects, used by Bertrand & Schoar (2003). It is therefore assumed that CEO style of management, education or other variables that we cannot observe, are constant through time. This fixed effects thus absorbs effects that are constant throughout time and already controls for the type of industry the CEO-firm observations operate in.

The dependent variable is the firm’s performance difference between the first year in office and the transition year. In econometrics terminology this is known as the first difference dependent variable. This is similar to Bennedsen et al (2006), who used a three year difference, surrounding the turnover. We subtract the performance of the transition year to remove the effect of the previous CEO. Thus, the methodology removes the effect of the transition year firm performance and the effect of the previous CEO, on firm performance.

An important question is to either use accounting-based performance measures or stock market returns as the dependent variable in the regression method. Indicators for accounting profits can be EBITDA, ROE and ROA and could be used as a proxy for past or current firm performance. Market value is a proxy for future firm performance. In literature about CEO turnover and succession, ROA is mostly used (Gordon et al., 2002). This research will use three measures: Tobin’s Q, ROA and Operating ROA as the dependent variable. The exact construction of each variable can be found in the appendix.

A possible endogeneity concern is the reverse causality of the (change in) firm performance and the first year dummy. Previous CEOs might have steered the firm in negative or bad performance and this results in his turnover, which in turn gives a new CEO his first year in office. As pointed out by Huson, Parrino & Starks (2001), the probability that a CEO is fired because of firm performance, is relatively small. For the worst performing firms, this is only 2.8%. This leaves room for many other factors contributing to the event of a CEO turnover. Moreover, the methodology controls for the previous CEO’s and firm’s performance by taking the change in firm performance, relative to the previous year.

(21)

21 The general regression formula is as follows:

𝐶ℎ𝑎𝑛𝑔𝑒 𝑖𝑛 𝑓𝑖𝑟𝑚 𝑝𝑒𝑟𝑓𝑜𝑟𝑚𝑎𝑛𝑐𝑒_𝑖,𝑡

= 𝛼𝑖+ 𝛽1 ∗𝐹𝑖𝑟𝑠𝑡 𝑦𝑒𝑎𝑟 𝑑𝑢𝑚𝑚𝑦_𝑖,𝑡 + 𝛽2 ∗𝐶𝐸𝑂 𝑎𝑔𝑒_𝑖,𝑡+ 𝛽3 ∗𝐿𝑒𝑣𝑒𝑟𝑎𝑔𝑒_𝑖,𝑡+ 𝛽4 ∗ 𝐿𝑛 𝑡𝑜𝑡𝑎𝑙 𝑎𝑠𝑠𝑒𝑡𝑠𝑖,𝑡+ 𝛽5 ∗ 𝐿𝑛 𝑜𝑓 𝑠𝑎𝑙𝑒𝑠𝑖,𝑡+ 𝛽6 ∗𝑌𝑒𝑎𝑟 𝐹𝑖𝑥𝑒𝑑 𝐸𝑓𝑓𝑒𝑐𝑡𝑡 + 𝑒𝑖𝑡

Where the subscript i denotes the individual firm and t the year in which the observation is taken. 𝛼𝑖 denotes the CEO fixed effect, also known as the intercept for an individual CEO, 𝑒𝑖,𝑡 is the error term. The year fixed effect variables controls for unobserved differences between years.

Dependent variables

Return on assets and Tobin’s Q will be used as the main methods for estimating firm performance. Return on assets is defined as net income divided by total assets. Tobin’s Q is a proxy for the stock market value of the firm. Both measures are used widely in literature concerning CEO turnover and succession.

As discussed earlier, and in congruence with Kaplan & Minton (2012) and Bennedsen et al (2006), firm performance can also be adjusted for industry. Moreover, CEOs are evaluated on relative performance with the industry (Jenter & Kanaan, 2014). This controls for industry shocks that affect firm performance but are not under the control of the CEO. The performance of a certain industry can affect the performance of the firm and gives a certain degree of luck via firm performance, to an individual CEO and his compensation. (Bertrand & Mullainathan, 2001). This is controlled for in the third measure of firm performance, where the mean industry performance is subtracted from the firm’s performance, resulting in the industry-adjusted firm performance, similar to Bennedsen et al (2006).

Explanatory variables

The main variable of interest is the first year dummy. This variables takes on a value of 1 if it is the first complete year in office for a given CEO, and 0 otherwise. Via this variable, we try to measure the CEO effect for his first year in office. Also, similar to Bennedsen et al (2006), CEO age is included as an

(22)

22 explanatory, to account for the effect any age difference a CEO could have on managing the firm. This results in the second regression equation where the independent variables are added.

Control variables

As pointed out by Mackey (2008), firm size and industry also determine the firm’s performance and should be controlled for when looking at CEO’s influence on firm performance. The firm’s industry is constant through time and controlled for via 𝛼𝑖, the CEO fixed effect. Firm size is measured by the value of the natural logarithm of sales. Firm performance is expected to be positively related to firm size, because larger firms normally have more market power. Bennedsen et al (2006) use an extra measure for controlling for the size of the firm, namely the natural logarithm of assets. This will also be included in the regression. As firm performance is affected by leverage, this should be included as a control variable to prevent omitted variable bias (Antia et al., 2010).

Secondary regressions

Aimed at proving hypotheses four and five, additional regressions will be run, to check for the effect of a first year in office on investments and acquisitions. These regressions might explain the mechanics behind a relationship of the CEO’s first year in office and firm performance. Investments are an important determinant in the firm’s accounting and stock performance. To remove the effect of the previous CEO, the change in investments will be used, similar to the regression formula with firm performance as the dependent variable. As described in the literature review, we could expect an increase in investments due to the CEO’s first year in office.

Two measures of internal investments will be used as the dependent variable, based on Bertrand & Schoar (2003). The first is defined as capital expenditures divided by the value of net property plant and equipment. The second is defined as net capital expenditures (capital expenditures minus depreciation), divided by net property plant and equipment. Also, acquisitions can be seen as a form of external investment for growth (Bertrand & Schoar, 2003). The acquisitions dummy variable takes on a value of 1 if there have been one or more acquisitions in a given firm year observation and will also be used as a dependent variable.

(23)

23 The control variables for these regressions are based on the control variables used by Aivazian, Ge, & Qiu (2005). These control variables include the lagged Tobin’s Q ratio, leverage and the lagged ratio of sales, divided by fixed assets. Tobin’s Q is included because a higher market valuation gives a firm more possible investment opportunities and easier financing. Leverage determines the cost of financing and might constrain the level of investments, if it is too high. This is also known as the debt overhang problem. the ratio of sales and fixed assets accounts for the size of the firm. Larger firms have more investment opportunities due to, but not limited to, market power. Standard firm control variables such as industry and CEO and year fixed effects are included, similar to the earlier regression formula. This results in the following formula:

𝐶ℎ𝑎𝑛𝑔𝑒 𝑜𝑓 𝑖𝑛𝑣𝑒𝑠𝑡𝑚𝑒𝑛𝑡𝑠𝑖,𝑡

= 𝛼𝑖+ 𝛽1 ∗𝐹𝑖𝑟𝑠𝑡 𝑦𝑒𝑎𝑟 𝑑𝑢𝑚𝑚𝑦_𝑖,𝑡 + 𝛽2 ∗𝐶𝐸𝑂 𝑎𝑔𝑒_𝑖,𝑡+ 𝛽5 ∗𝑇𝑜𝑏𝑖𝑛′𝑠 𝑄𝑖,𝑡−1+ 𝛽5

∗ 𝐿𝑒𝑣𝑒𝑟𝑎𝑔𝑒_𝑖,𝑡+ 𝛽5 ∗ ( 𝑆𝑎𝑙𝑒𝑠

𝐹𝑖𝑥𝑒𝑑 𝐴𝑠𝑠𝑒𝑡𝑠)𝑖,𝑡−1+ 𝛽7 ∗𝑌𝑒𝑎𝑟 𝐹𝑖𝑥𝑒𝑑 𝑒𝑓𝑓𝑒𝑐𝑡𝑡 + 𝑒𝑖𝑡

3.4. Data

In this section, the sample, data sources and dataset construction are discussed. Also, a table of summary statistics will be analysed. US firms will be used for this research, as data is most accessible and expansive. Sample period will be 1992-2012, as information about CEO turnover dates is only available since the year 1992. Financial fundamentals time series data concerning firms is retrieved from the Compustat North America Fundamental Annual dataset. This concerns information regarding capital expenditures, R&D expenses, stock price, assets etc. The data is based on the annual report files of firms. This will be the initial dataset on which several sample filters will be applied.

CEO turnover time series data is retrieved from Execucomp. This dataset starts in 1992, thus limiting our time window to 20 years. This database is based on the full Standard & Poors 1500 firms. CEO turnover dates from this dataset will be used to create a variable that contains a value of 1 if in the calendar year has a CEO turnover has taken place. Otherwise, the value will be 0. Another variable will count the length of tenure of a CEO, starting from the moment he has entered the Execucomp database.

(24)

24 This allows us to have data on a CEO, and his firm, before he went into office. This is needed to make conclusions concerning hypotheses two and three. Information about mergers & acquisitions is extracted from the Securities Data Company platinum database. These three datasets will be merged and results in a panel data set with CEOs matched with firms in yearly observations, ranging from 1992-2012.

Sample filters

The sample is restricted by excluding firms that are in finance, insurance and real estate. These are firms that with a SIC code between 6000 and 6999. Firms with assets and sales of a value lower than 1 million US dollars are removed from the sample. Extreme values are excluded, which are values exceed three times the difference between the 25th_{and 75}th_{percentile. If a CEO has left office in the} same year that he got into office, he will be dropped from the dataset, because this will most likely by an interim CEO (Hornstein, 2013).

3.4.1. Descriptive statistics

In this section, the descriptive statistics concerning the panel data sample will be discussed. First, CEO turnovers and the length of their tenure will be discussed. After that, we will look at the financial variables concerning the firms that are led by those CEOs in our sample.

Figure I shows the number of turnover events per year in the sample. There seems to be a peak of turnovers in the years 1999 – 2000. There is a significant increase in CEO turnover in 2011, indicating at responses from firms following the financial crisis of 2008-2009. This could be explained by the delay evidenced by Kaplan & Minton (2012), who showed CEO turnover can be delayed for up two years, after bad firm performance. This increase in turnovers might be the result of firms reacting on bad firm performance after the 2008 – 2009 financial crisis. Figure II shows the average tenure of CEOs in our sample. Average tenure seems to be increasing in the last few years. The figure seems to show the fact that average tenure was at its all-time low, until 2010, as documented by Antia et al (2010) and Kaplan & Minton (2012).

(25)

25 Figure I: CEO turnover (1992 - 2012)

Source: Execucomp database

Figure II: Average tenure of CEOs (1992 - 2012)

Source: Execucomp Database

108 147 164 199 189 209 ₂₁₄ 251 ₂₅₇ 219 186 198 189 216 201 212 208 155 149 188 174 0 50 100 150 200 250 300 N u m b er o f t u rn o ve rs Year

(26)

26 Table I represents descriptive statistics, based on the full sample. Table 2 shows these statistics for first year in office of CEOs. The sample consists of 2485 unique firm-CEO matches with 2580 observations of first-year CEOs. Total firm-year observation sums up to 26043.

[Table I approximately here]

As shown table I, average return on assets for the full length of tenure or all CEOs, reported in row 1, is 4.97%, with a standard deviation of around 6.58%. The average and median age was 55 years, while the oldest CEO in the sample is 96 years. Also, the first measure of investments is always larger than the second measure of investments. The computation of this measure is the reason for this, as investments (2) measures capital expenditures minus depreciation, otherwise known as net capex, investment (1) measures only capital expenditures.

Table II shows descriptive statistics for subsample, where the statistics are taken from the first full year in office of the CEOs in the sample. For instance, operating return on assets is 9% on average for CEOs in their first year in office. Also, age has a lower mean, which is to be expected. The oldest ‘new’ CEO, was 81 years, as can be seen in column 4 of table II. All measures of firm performance (columns 1-3), show lower average values than the grades presented in table I.

[Table II approximately here]

Table III shows the means of the dependent variables included in the regression, otherwise known as the firm performance measures and the level of investments, surrounding the two years of the turnover. As shown in the table, at 2 years before the turnover, average return on assets is approximately 5.22%. As the turnover approaches (t=0), this performance measure drops. The other two measures of firm performance also show a decrease near the turnover event and a slight increase after the first year in office. The decreases in the transition year (year 0) of performance, could also be caused by the problem known as the ‘big bath’ (Murphy & Zimmerman, 1993), which was described in section 2.1. Row 4 show the second measure of investments, where we a large decrease in its average value is shown. In t=0, its average value is 1.08% and in t=1, the average value is only 1.08%, while the investments (1)

(27)

27 measure remains relatively unchanged. This could be an effect of large depreciation sums taken on by the new CEO.

[Table III approximately here]

In appendix I, the correlation coefficient for the variables which are used in the regressions are reported in a matrix. Only measures of a similar financial variable, such as the two measures of investments, show a high correlation coefficient. Investments (1) and investments (2) have a correlation coefficient of 0.93. Also, the various measures of return on assets show a high degree of correlation, such as the industry adjusted return on assets and the return on assets, which have a correlation coefficient of 0.80. Most importantly, no variables that are used simultaneously in the regressions, show a high correlation coefficient.

4. Empirical results

In the following chapter, the empirical results from the regressions will be discussed and related to existing literature. The earlier stated hypotheses will be reviewed in accordance with the results from the regression analysis. Included in the robustness is an event study aimed at showing an increase in a firm’s stock returns after the CEO takes his place into office. All regression are run with robust standard errors to account for heteroskedasicity.

4.1. Firm performance

Table IV presents the results of the regression estimating the change in firm performance. Column (1) – (3) shows the effect of a first year in office of a CEO through the ‘First Year dummy’ variable on three measures of firm performance. Control variables are included for firm size, sector, year and CEO fixed effect. Column (1) and (2) show the effect of only the change in firm’s performance. Column (3) shows the effect of firm performance with a subtraction of the mean industry performance. This results in the change in performance of the firm in a given year, which exceeds the average performance of the industry. This controls for a measure of luck, which only a sector control variable does not capture, according to Bennedsen et al (2006).

(28)

28 The variables of main interest are first year dummy variables, shown at the top of the table. All three coefficients of these variables show a positive, statistically significant relationship with the various measures of firm performance. Only the coefficient for the effect on Tobin’s Q becomes significant at the 5% level. In a given first year in office, the expected change in firm performance is on average 1.2% in return on assets, according to the regression model. Adjusted for the mean industry performance, this is an expected 0.9%, as shown in table IV. Tobin’s Q is also affected, an increase of 0.02 of Tobin’s Q ratio, also known as the market value of the firm, is expected when a CEO completes his first year in office. Tobin’s Q is a measure of market value performance, while the other measures are accounting-based measures of performance. However, this is the only measure that is not significant at the 1% level. According to this table, there is evidence of correlation between a CEO’s first year in office with the firm’s performance, even after adjusting it for the industry’s performance. In the robustness check, an event study will be presented that examines the stock price changes after a CEO turnover, which is also related to the market value of the firm. The CEO’s age does not show a significant relationship with firm performance. This variable only captures the age of the individual CEO. There might be other variables, such as CEO working experience or education that are not incorporated into the age and might affect firm performance. Data on this kind of variables is difficult to obtain and can only be hand collected.

[Table IV approximately here]

The control variables log of sales and assets show significant relationships with firm performance. This was expected, as these are widely used as a control for the size of firms. Higher sales lead to better firm performance through the firm’s net income, which is a clear relationship. Assets show a negative relationship with firm performance, which might be surprising. Reason for this might be that larger firms are more difficult to manage for CEOs. This is similar to the results of Bennedsen et al (2006). Evidence of Bennedsen et al (2006) suggest that assets are reduced to increase short-term profitability. Incoming CEOs might reduce its asset, by writing off bad investments or divisions. The coefficient on log of assets does not incorporate the change of assets in the first year in office. It only measures the

(29)

29 effect on the change in firm performance in all years of tenure, thus preventing us from drawing a clear conclusion.

Table V shows the results of regression run for either insider CEO’s and outsider CEO’s and their first year effect on firm performance. The result include both Tobin’s Q and return on assets as the dependent variable. It is hypothesised that insider CEO’s have more firm-specific knowledge and are therefore better suitable to make a significant positive impact on firm performance. In columns (1) and (2), the regression results for a subsample of insider CEOs is run. Columns (2) and (4) show the results of CEOs who were hired from outside of the firm. In the sample, there are more outsider CEOs than outsider CEOs, as can be seen table VII. However, there is strong evidence that insider CEOs have a larger impact on firm performance, both accounting-based and stock-market based. According to column 1, an insider CEO positive raises the change in return on assets by 1.24% due to his first year in office. Similarly, the change in Tobin’s Q is raised by 0.05. Both are statistically different from zero at the 5% level. Outsider CEOs show only a 10% statistical significance for its first year effect on the change in return on assets and have no significant impact on the change in Tobin’s Q. All other variables show similar results in relation to firm performance as in earlier tables. Most importantly, both the firm performance measures differs significantly between these two types of CEOs. This results in some evidence in favour of hypothesis two, where CEOs who are familiar with the firm and have worked their way up, are significantly more able to influence firm performance in their first in office.

[Table V approximately here]

4.2. Investments

There seems to be a positive relationship between the first year in office of a CEO and firm performance. To check what might be driving this positive relationship, several financial variables are regressed.

First, t-tests are run to check whether the means of the dependent variables changed between the transition year and the first full year in office. Only the second measure of investments, defined as net capital expenditures divided by fixed assets, decreases significantly under the 1% level. The growth

(30)

30 rate, defined as capital expenditures divided by total assets, also decreased significantly under the 1% level. Table reports for these t-tests can be found in the appendix.

Table VI reports the regression results of the relationship between the first year in office and two measures of investments. Investments are measured as the change in investments between the first year in office and the transition year. Control variables such as sector, year and CEO fixed effects are included in all regressions. Firm control variables such as leverage are included in the last two columns. Similar to table IV, we are mostly interested in the first year dummy. This captures the effect of the first year in office on the dependent variable, which is the level of investments. It is hypothesized in chapter two that first year CEOs increase the level of investments.

[Table VI approximately here]

As shown in table VI, a first year in office of a CEO has no significant effect on investments in our sample. Columns 1 and 2 report the results without firm control variables. Column 3 and 4 include the control variables. The control variables have a significant effect on the level of investments, as shown earlier by Aivazian et al. (2005). T-tests of the means show a significant decrease in investments of measure (2). However, when controlling for other variables via this regression, we cannot conclude that a first year CEO has influence on the level of investments. One reason might be suggested by Hornstein (2013), as investments are halted around the turnover and are increased after two or more years.

Table VII reports the effects on the acquisition decision, estimated with a probability regression. Evidence in earlier research shows that acquisitions at best give a slight positive return (Malmendier & Tate, 2005). CEOs might know this and thus it would be a weak strategy for incoming CEOs. Column (1) reports the effects without firm controls and column (2) reports the effect where firm controls are included. Column (1) reports a significant negative effect of the first year in office of a CEO on the acquisition probability. A first year in office leads to an 8.9% decrease in the change of an acquisition. However, when controlling for firm characteristics, this effect is no longer statistically significant. Other control variables show expected coefficient, such as Tobin’s Q. A higher market value gives a firm more options to engage in acquisitions, because a firm can use its equity to finance these acquisitions.

(31)

31 [Table VII approximately here]

To conclude, there seems to be insufficient evidence of first year CEOs significantly affect the level of investments or acquisitions. Regression were also run separately for both outsider and insider CEOs but results did not deviate. CEOs are able to influence firm performance in their first year, but through what channels remains unknown. There could be unobserved efficiency increases or managerial skills implemented in the firm, which can be attributed to the increase in both market and accounting-based performance of the firm.

4.3. Robustness checks

As a first robustness check, alternative measures of the dependent variable will be utilized. First the main dependent variable of interest, the change in firm performance. The change in operating return on assets is used as the dependent variable in the next table. Operating return on assets is measured as the ratio of earnings before interest and taxes (EBIT), divided by the book value of assets. This measure of performance is unaffected by differences in the firm’s capital structure decision and therefore a good indicator of performance changes affected by the CEO (Bennedsen et al, 2006). Similar to table IV, an industry adjusted performance measure is included in the regression.

The coefficient of the first year dummy is less significant than the other measure of firm performance via return on assets. Reported in the table is a 0.3% increase in both measures of operating return on assets, much smaller than 1.2% effect of on return on assets that is reported in table IV. The effect is also less significant, only under the 5% level. The results are in line with the findings of Huson, Parrino, & Starks (2001), in the sense of an increase in operating profits after CEO turnovers. According to these results, CEOs seem to have a larger and more significant effect on return on assets than other measures of firm performance. The positive relationship between the first year in office and the change in firm performance still holds.

[Table VIII approximately here]

Table IX presents the regression results of the same model as table VIII, but separately run for outsider and insider CEOs. Columns (1) and (2) show the results for a CEO promoted from within the firm and

(32)

32 columns (3) and (4) show the results for a CEO who was hired from outside of the firm. Results differ for insider CEOs relative to outsider CEOs. Outsider CEOs seem to have a significant impact due to their first year in office. For example, column (3) shows an increase of 0.4 percentage point on the operating return on assets (ORoA) of a firm. The effect on the industry adjusted ORoA is even larger: 0.7 percentage point. Other variables show similar results are earlier regression. These results partially contradict the results of Table V, where outside CEOs did not seem to significantly influence Tobin’s Q and Return on Assets. However, these are different measures of firm performance. Operating return on assets is defined as the earnings before interest and taxes (EBIT), divided by total assets. These results indicate that outsider CEOs seem to significantly affect EBIT while insider CEOs do not, in their first year in office. While this does not invalidate earlier main results in favour of hypothesis (1), it prevents us from accepting hypothesis (2). There is only partial evidence that insider CEOs are able to influence firm performance in their first year in office.

[Table IX approximately here]

Concerning the regressions which used the change in investments as the dependent variable, alternative definitions of investments were used but showed no different relationship. Using the level of investments instead of the first difference of the investments (the difference between the levels of investments between two subsequent years) showed no significance when correctly controlling for firm characteristics. Thus, there is insufficient evidence to accept hypothesis three and four.

4.3.1. Event study

The empirical regression from the regression methodology shows some evidence of market value being positively affect by the CEO’s first year in office. A popular method in corporate finance and used by many researchers in the corporate governance field is the event study method. Multiple researchers on the field of CEO turnover also utilized this approach, such as Weisbach (1988), Campbell, Gallmeyer, Johnson, Rutherford, & Stanley (2011) and Hornstein (2013). The approach is relatively straight-forward and is based on the Market Effiency hypothesis of Fama (1988) where stock prices directly

The performance of first-year CEOs