How do different CEO types affect cash holdings?

U

NIVERSITY OF

A

MSTERDAM

A

MSTERDAM

B

USINESS

S

CHOOL

MS

F

–

C

F

M

T

H

OW DO DIFFERENT CEO TYPES AFFECT CASH

HOLDINGS

?

Author:

Supervisor:

Dávid Pataki

Dr. Vladimir Vladimirov

Statement of Originality

This document is written by Student David Pataki 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.

Acknowledgment

I would like to take this opportunity to thank my thesis supervisor Dr. Vladimir Vladimirov for the continuous support and availability throughout the Master Thesis and Thesis Seminar. The timely suggestions and flexibility are greatly appreciated. Also, I would like to thank the authors of the paper Custódio et al. (2013) making their research data available to the public.

Abstract

I examine the relation between certain CEO characteristics and firms’ cash holdings among publicly traded US firms from 1993 until 2007. Using panel regressions, I find evidence that if a generalist CEO replaces a specialist manager, the firm experience an increase in cash holdings. Also, overconfident managers tend to hold higher cash balances compared to more rational ones. Further investigations suggest that firms with financial constraints display a positive, statistically significant relation between managerial overconfidence and cash holdings. Additionally, financially constrained firms exhibit a higher sensitivity of overconfidence to cash holdings than financially unconstrained ones.

Table of Contents 1. Introduction ... 2 2. Literature review ... 5 2.1. Cash holdings ... 6 2.2. CEO types... 11 2.3. Hypotheses development ... 15 3. Data ... 16 3.1. Variables ... 17 3.1.1. Cash holdings ... 17

3.1.2. CEO type measures ... 17

3.1.3. Financial constraints ... 19

3.1.4. Control variables... 20

3.2. Descriptive statistics ... 22

4. Methodology ... 23

5. Empirical Results ... 25

5.1. General ability index and cash holdings ... 25

5.2. Overconfidence and cash holdings... 28

5.3. Financial constraints ... 30

5.3.1. General ability index and constraints ... 31

5.3.2. Overconfidence and constraints ... 31

5.3.3. Interaction analysis ... 33

6. Robustness Analysis ... 34

7. Conclusion ... 38

References ... 39

2 1. Introduction

Cash holdings have become a widely discussed topic in the past decade and the academic literature has provided us with many explanations, theories and applications of the matter. Cash holding, accumulating cash and cash equivalents, in other words cash hoarding has been rapidly and steadily increasing from the beginning of the 1990s until just recently. Between 2009-2010 a US non-regulated firm which has $5 million market capitalization on average held 21.28% of its asset in cash (Pinkowitz, Stulz, and Williamson, 2013). The media is also attracted by this phenomenon. One particular article reports that the US non-financial companies total cash holdings account to $1.7tn by the end of 20151. One third of this amount, $505bn of cash was held by only five tech giants. The largest company, namely Apple, held $216bn in cash which is a tenth of all the US cash holdings. In the second quarter of 2017 this number reached $256.8bn which is larger than the market capitalization of General Electric2. Also, the majority of these reserves are abroad due to the unfavorable repatriation taxes. This has also reached the attention of the world leaders, Donald Trump plans to offer a one-time taxation on overseas cash which would allow the companies to repatriate their cash reserves with a lower cost, however, details are still unknown3_.

Despite the fact that it is a relatively new topic, there is an extensive academic literature on the cash holdings phenomenon. However, there is still no evidence on how specific manager characteristics affect these cash reserves. There are numerous papers discussing why firms hold extensive amount of cash, what the value of cash is and whether cash holding is increasing or destroying shareholder value. Although, most of the articles try to tackle the question by focusing on firm characteristics, there is one concept that focuses on CEOs influence on cash holdings, the agency cost motive. However, it only questions whether the management as a whole is aligned with the shareholders’ interest or not. If they do not consider the welfare of the shareholders, they might use the accumulated cash for their own benefits and get tempted to invest into negative NVP projects like value destroying acquisitions (Moeller, Schlingemann and Stulz, 2005). Nevertheless, there is still no explicit research about how CEOs characteristics, traits are linked to cash holdings,

1 _{“Apple's cash hoard swells to record $256.8 billion”, CNBC May 2nd 2017.}

http://www.cnbc.com/2017/05/02/apples-cash-hoard-swells-to-record-256-8-billion.html. Accessed June 24th ₂₀₁₇ 2 _{“Apple's cash hoard swells to record $256.8 billion”, CNBC May 2nd 2017.}

http://www.cnbc.com/2017/05/02/apples-cash-hoard-swells-to-record-256-8-billion.html. Accessed June 24th ₂₀₁₇ 3 _{“Apple's cash hoard swells to record $256.8 billion”, CNBC May 2nd 2017.}

3 even though it is clear that firm cash policies and operational decisions affecting cash reserves are determined or at some extent influenced by the management. As the issue became more prominent, the firms and managers are put under pressure to answer stakeholders’ doubts. However, it is still not apparent whether having large cash reserves hinders performance. On the one hand, Harford (1999) finds that firms with excess cash use it to make value destroying acquisitions. On the other hand, Mikkelson and Partch (2003) show that firm that hold high cash levels persistently have similar or greater performance than matched firms. Ultimately, it is still important to understand if there is relation between CEO characteristics and cash holdings. By looking at how certain types of CEOs determine cash holding I can showcase an unexplored area of the cash holding puzzle and further improve the explanation of this new trend.

The effect of different CEO types on cash holdings is examined by focusing on two CEO characteristic. The first trait I consider is the ability of the manager by implementing the publicly available general ability index (GAI) provided by Custódio et al. (2013). They measure the managers’ ability by five category of CEO experience and divide the CEOs into specialists and generalists. Specialist managers have a more firm-specific knowledge thus have little experience in other industries and held less positions overall. Generalists have accumulated a more universal and complex background, like multiple CEO titles or many different workplaces. Generalists are considered to have a broader knowledge and hence might be better in producing future revenues and as Custódio et al. (2013) show they are most valued in merger-intense companies. Therefore, I predict that generalist CEOs hold lower cash balances compared to specialists. The second characteristic is the level of managerial overconfidence. Malmendier and Tate (2005) argue that overconfident managers overestimate their ability to produce results and thus the profitability of available investments. They show that overconfident managers positively affect firms’ sensitivity of investment to cash flow and that these CEOs use the firms’ incoming cash flow to make more investments. According to the static tradeoff theory by Opler et al. (1999) and in line with the findings of Bates et al. (2009), there is a positive relation between investments and cash holdings, anticipating that overconfident CEOs have higher cash balances. Additional to these measures, I also consider financial constraints and their effect on either characteristics’ influence on cash holdings. I mainly rely on Almeida et al. (2004) study on this matter, examining five measures to proxy for financial constraints. Financial constraints play a large role in cash holdings, constrained firms on average hold more cash and that cash is more valuable for them than for unconstrained

4 firms (Faulkender and Wang, 2006; and Denis and Sibilkov, 2010). On the one hand, I suspect that constraints are a stronger force determining cash holdings than the CEO ability thus leaving the GAI measure insignificant. On the other hand, overconfidence can further influence the firms’ cash balances alongside with the financial constraints.

I conduct my research on US firms, due to high shareholder protection and greater alignment between managers and shareholders’ interest (Dittmar et al., 2003). However, under this country-level protection, entrenched managers in a firm-country-level setting in fact hold less cash to avoid proxy fights and media attention which could trigger shareholder activism (Harford et al., 2008). Only publicly traded firms were included from 1993 until 2007 and all financial companies and utilities were removed from the sample. First, I examine whether there is a difference between specialist and generalist cash policy controlling for other motives for holding cash using several regression types, including ordinary least squares, the Fama-MacBeth method, and fixed effects. Next, I investigate overconfidence employing two proxies while using similar empirical methods as mentioned before. The first overconfident variable is measuring the managers’ willingness to sell his stock option when it would objectively be the rational course of action. If they fail to do so (two times), they are considered overconfident (Malmendier and Tate, 2005). The second proxy is implemented by following the method of Campbell et al. (2011). The authors utilize Malmendier and Tate’s (2005) finding of linking investment levels and overconfidence together. Lastly, several financially constrained measures were used to split the sample into financially constrained and unconstrained subgroups in order to re-investigate the original regressions. Additionally, as a separate analysis I employ the constraints proxies as dummy variables into the regressions and also interact them with the GAI or overconfident variables to produce interaction terms.

Considering the GAI, I find that after a turnover, if the new manager is a generalist then the firm hold more cash. There is also limited evidence that as the CEO acquire more general ability they hold more cash. However, I find no evidence that the general ability index has a meaningful impact on subgroups when considering financially constrained and unconstrained firms. Overconfidence on the other hand, proves to be an important CEO characteristic that determines cash levels. The base regression models display a positive relation between overconfidence and cash holdings. Furthermore, considering the financial constraint settings, I find evidence suggesting, that a change in the CEO behavior to act as overconfident will result in a higher cash balance for constrained firms which is also larger than what an unconstrained firm would experience. Overall, the GAI

5 shows limited influence affecting cash balances, however, there is a strong relation between managerial overconfidence and cash holdings.

Several problems arise from keeping high cash reserves. First, when corporations do not spend their accumulated cash reserves the potential to further enhance the economy is unexploited. The companies investing their excess cash could benefit all the stakeholders by creating for instance new projects, new workplaces, innovations and better environments. Secondly, there are some concerns regarding not the existence but the use of these high cash levels. Considering that there are agency conflicts between CEOs and the shareholders, doubt will inevitably arise. One can assume that the management would be more carless or that they could use the excess cash to realize their own agenda. Also, Harford (1999) provides evidence that CEOs in cash rich firms often use cash to make value destroying acquisitions, which certainly hurt investors. Thirdly, keeping a high amount of liquid assets (e.g. cash and short term investments) enables firms to potentially pay back all of their debt at once and hence have a net debt value around zero. Many institutions look at debt when considering firm leverage, but examining the net debt values, we find a completely different situation for firms’ leverage levels in the US (Bates et al., 2009). This can aggravate policymakers in regulating the debt levels of firms and provide accurate indicators about the companies’ credibility. My research can help shareholders and boards to better understand the impact that CEOs have on cash holding, and in a broader concept allow stakeholders to assess how certain managerial characteristics affect the firm itself.

The remained of the thesis is organized as follows. The next section summarizes the academic literature of cash holdings and CEO characteristics then develops the hypotheses. In Section III, the sample data and variable construction are presented. Section IV describes the methodology of this thesis. In Section V, I demonstrate and examine the empirical results of the models. Section 6 investigates the robustness of these findings. Section 7 concludes.

2. Literature review

The aim of this thesis is to examine the relation between CEO characteristics and cash holdings. I focus on two managerial traits: the ability and the overconfidence levels of the CEO. Apart from the agency cost motive, there is little research on how managers influence cash holdings. Also, the agency cost motive only considers the managers as a link between shareholders and their firms, examining whether they are maximizing shareholder value or not. Implementing proxies from the

6 existing academic literature on managerial ability and overconfidence I can examine these measures relation to cash holdings. Furthermore, I include financial constraints as well, because many researches present it as a dominant factor explaining hoarding cash.

In order to complete my research, I first need to review the rich academic literature on cash holdings to know what and to what extent is influencing firms’ cash policies. Reviewing previous studies, I can examine their methods determining motives to hold cash and employ them as control variables in my regression to see if the ability or the overconfidence of the CEOs alone has an effect on the firms’ cash balances. Investigating previous studies helped to improve the methodology of the thesis like the inclusion of the Fama-MacBeth model extending the validity of my research. In the second part, I describe the relevant papers that focus on CEO types and characteristics to have a base understanding of their influence on firms in general and if there is a theoretical connection to cash holdings as well. Lastly, I showcase the studies that touch on both aspects of my research to be able to compare results if available.

2.1. Cash holdings

One of the first theories on corporate cash holdings are originated form Keynes (1936). The author presents two motives for holding cash and cash equivalents. First is the transaction cost, which states that firms hold cash to avoid the costs of liquidating their assets in time and without loss in value. The second theory from Keynes (1936) is the precautionary motive. If other sources of funding are too expensive or not available firms can use their accumulated liquid assets to finance their investments. The paper from Opler et al. (1999) is one of the most cited studies of this topic, also investigates these motives. They develop the static tradeoff theory and the financing hierarchy theory. The former suggests that there is a tradeoff between the costs and benefits of holding cash and that firms strive to find the optimal level of cash if managers are maximizing shareholder value. The latter implies that there is no optimal cash level to be held, the firms’ debt and cash change according to their profitability. The transaction motive and the precautionary motive are two benefits of holding cash in the static tradeoff model. Moreover, the paper adds that if information asymmetries are higher between investors and the management then external financing becomes more expensive, thus those firms that invest into projects where information asymmetry is excessive are expected to hold more cash. Lastly, they introduce the agency cost motive of holding cash which is aligned with the financing hierarchy theoretical framework. One

7 possible reason for the management to accumulate cash is if they are risk averse but in doing so they are not focusing to maximize shareholder value. The second reason for entrenched management to hold cash is to pursue their own interest and realize personal visions. It is hard to distinguish between the two main theories (static trade off and financial hierarchy) because most of the proxies are ought to show the same signal for both view. However, there is one significant difference. The tradeoff theory suggests a positive relationship between investment and cash holdings where the financing hierarchy theory predicts a negative effect.

They also present the concept of excess cash, measured by the residuals of the Fama-MacBeth model, meaning that a firm hold excess cash if the predicted cash amount by the regression is lower than the firm actually holds. They divide the sample into high and low excess cash firms and find that firms with positive excess cash spend considerably more on investment or acquisitions but the tendency to use excess cash itself is rather limited. Developing and using the excess cash method is rather influential to later papers. Many studies follow suit and use this methodology to conduct their research (e.g.: Pinkowitz et al., 2011 and 2012). Furthermore, they also argue that cash is not equivalent to negative debt, because there are cases where the management can use the accumulated cash to invest which, however, would not had been financed by the capital market. This is proven by Acharya et al (2007) where they show that financially constrained firm with high hedging needs will keep and hoard cash from their free cash flow, however, constrained firms without the need of high hedging will pay back their debt with the accumulated excess cash.

Regarding their results, on the one hand, Opler et al. (1999) find that firms with larger size, net working capital, leverage, and those with dividend or credit rating likely to hold less cash. On the other hand, cash holdings increase with growth opportunities, industry volatility, cash flow ratio, capital expenditures ratio and R&D ratio. These findings are consistent with the precautionary motive, however, there are not enough evidence to confirm any significant role for the agency cost in determining cash levels. I rely on this paper in more ways. First, the regression is an exemplary method which I utilize in this thesis. Secondly, the theoretical framework helps to establish further connections in developing my hypotheses.

Due to the fact that Opler et al. (1999) is such an all-rounded study, many of the subsequent papers discuss the unsolved agency conflict theory. Harford (1999) research is mainly focusing on this unproven hypothesis. He studies the US market and determine the agency cost as the primary

8 reason for excess cash holdings and examine the cash-rich firms’ investment decisions. As in Opler et al (1999), this paper uses excess cash as their independent variable as well. Herford’s (1999) hypothesis comes from Jensen’s (1968) free cash flow hypothesis, which predicts that on average, investments made by managers that hoard large amounts of cash are value decreasing. Jensen argues that this comes from the agency conflict and it is sternest at firms with great cash flows. Harford (1999) indicates that cash holdings is essentially cash flow, hence predicting that acquisitions – the most visible form of investment – made by cash-rich firms are value destroying. As mentioned, cash-rich firms are determined by similar method as in Opler et al. (1999), but Harford applies a stricter threshold to excess cash levels to determine a company as cash-rich. They find that firms with excess cash are more likely to make acquisitions and these acquisitions are in fact value decreasing as predicted by the free cash flow hypothesis, shown by the negative reaction of the market to the announcement and later by the weak operating performance of the merged entity. Another paper by Dittmar et al. (2003) conduct a worldwide research on how the agency costs theory affect cash holdings on a global scale. They include firms from 45 countries around the world. After controlling for industry effects, capital market developments and other firm characteristics which influence cash holdings, they find that firms in countries of the lowest level of shareholder protection hold twice as much cash than firms where shareholders are best protected. The paper shows that with lower shareholder protection, thus higher agency cost, firms hold significantly greater levels of cash holdings. Furthermore, in countries of poor shareholder protection, other causes explaining cash holdings are seemed to be weaker. Moreover, there is no evidence that managers hoard more cash when access to the capital market is limited in countries with low shareholder protection. On the contrary, the paper finds that companies keep more cash when external funding is more accessible. The authors are canvassing the findings with care, completing large set of robustness tests which yield similar results, but they still note that it is hard to capture the international differences in accounting systems across this broad setting. A study by Harford et al. (2008), however, documents that US corporations with weaker corporate governance, thus firms with higher agency cost are in fact holding less cash. Also, firms with lower corporate governance all else equal, would use up their high cash reserves more quickly than companies with strong corporate governance. These spending mainly focus on acquisitions rather than R&D or capital expenditures. Further, the paper argues that these investment options made by firms with the worst governance are diminishing future profitability and firm value. Harford et al (2008) also

9 investigate the seemingly contradiction that in an international setting, firms in countries of poor shareholder protection hold higher cash reserves (Ditmar et al., 2003). Harford et al (2008) imply that the effects of county-level shareholder protection overrule the effects of the firm-level managerial entrenchment. They argue that the US has high shareholder protection thus managers are not as entrenched as others in less protected countries. The CEOs in the US can be targeted by proxy fights because hoarding too much cash is noticeable and attracts shareholder activism. In order to avoid that, managers spend their cash reserves as mentioned before. According to Brisker et al. (2013) as well, there is a negative relationship between corporate governance and cash holdings. Beside the effects of the precautionary motive (firms’ increased credibility help them raise external funding), they show that firms after being admitted to the S&P index have significantly lower cash holdings due to higher agency cost. After inclusion, corporate governance declines as a result of an increase in managers’ entrenchment. Firms and the CEOs gain more attention after managing to be accepted into the S&P 500, which in turn, aligned with Harford et al. (2008) findings, leads to a cutback in cash holdings. Lastly, Gao et al. (2013) provide a unique study on the relation between cash holdings and agency conflict. They study publicly and privately held firms in order to better understand the public companies’ cash polices. They find that on average, privet firms hold about fifty present less cash than public firms, even though they have more limited access to the capital markets. The authors consider agency cost as the main reason for public firms to hold much more cash. Also, they show that well-governed firms diminish their debt levels or pay out their excess cash, when worse-governed firms use the cash pile to invest or to acquire which hinders its performance.

There are studies dedicated to examine whether high cash firms behave differently than others. Mikkelson and Partch (2003) are examining how persistence high cash ratio firm perform, whether they are different compared to lower cash balanced firms or transitory cash-rich firm. They look at firms that have at least 25 percent of their assets in cash for 5 straight years. They find no evidence that these firms which had adopted a large cash holding policy have different operating performance compared to firms similar in size and industry. These persistence cash-rich firms invest more, mostly into R&D and are faster growing. Lastly, the authors find no evidence that agency conflict is responsible for the differences in performance among cash-rich firms. The paper by Pinkowitz et al. (2011) also adds to the argument that managers do not use cash in cash-rich firms to (unnecessarily) acquire other firms. The authors define a firm cash-rich if its hold 45% of

10 their assets in cash. They show that cash-rich firms are as likely to use cash for acquisitions as other firms. Furthermore, cash-rich firm are more likely to use equity as the method of payment than cash. Moreover, these firms on average use stock to acquire the target company when the bidders are overvalued. Lastly, the paper demonstrates that cash-rich firms are more likely to use cash when both its stock and the target are undervalued.

Other papers focus on the financial constraints aspect of the cash holding phenomenon. Almeida et al (2004) measure whether financial constraints have an effects on firms’ cash-cash flow sensitivity. They show that financially constrained firms display a significant positive cash flow sensitivity of cash, meaning that these firms stockpile their free cash flow into cash holdings rather than pay it out as dividend or use it for investment. However, as predicted by the authors, unconstrained firms do not show any particular pattern. They use 5 different proxies for financial constraints which I implement in this thesis as well. Building on Almeida et al (2004) findings’ Faulkender and Wang (2006) and Denis and Sibilkov (2010) shows with similar proxies that the value of cash is higher for firms that are financially constrained and this effects is even stronger when these firms are in need of capital to invest. This suggests that cash holdings are an effective way to allow firms to capitalize on investment opportunities which otherwise would not be accessible and support the view that firms can mitigate the negative effects of financial constraints by maintaining higher cash levels.

There are factors that could play a role in determining cash holdings other than the previously mentioned transaction, precautionary and agency motives. One of them is the tax-based explanation introduced by Foley et al. (2007). The study argues that US multinationals hold higher cash balances because it is costly to repatriate their foreign earnings, thus leaving the income abroad. The authors report an economically and statistically significant positive relation between the proxy for the taxes firms would pay if they repatriated their foreign earnings and their cash holdings. They also show that this effect is more pronounced for affiliates where the repatriation costs is higher. Additionally, the theory of R&D smoothing by Brown and Peterson (2010) serves as an additional explanation to the cash holdings phenomenon. It is excessively costly to adjust R&D expenses due to temporary financial frictions, hence the authors predict that firms use their cash balances to smooth R&D investments. They find that this hypothesis is true for financially constrained companies, but find no evidence that unconstrained companies use cash holdings to smooth R&D expenses.

11 Lastly, regarding cash holdings, the study by Bates et al. (2009) examine the three main hypotheses of holding cash as well: the transaction cost motive, the precautionary motive and the agency conflict motive. The authors acknowledge the taxation motive of Foley et al. (2007) but do not investigate the matter. Rather, they confirm and shed new lights on cash holding theories. First, they show that the dramatic increase in the average cash holdings in the US is concentrated among non-dividend payer firms, firms in more recent IPO listing cohorts, and firms in industries with the highest idiosyncratic volatility. They argue that the increase in cash ratio mostly can be explained by the evolution of firm characteristics. The changes in firm characteristics that mainly affect cash holdings are the lower inventories, the increased cash flow volatility, the declined capital expenditures, and the rise of R&D investments. While their findings affirm the precautionary motive as well, they do not find evidence for the agency cost motive pronouncing the discussed puzzle of the effects of the agency conflict on cash holdings. Furthermore, they note a large reduction of net debt from 1980 through 2006 which is due to the rise of cash holdings and not to the decrease of debt. At the end of the sample period on average a firm could pay back all of its debt only using cash holdings. This study clearly summarizes the most relevant theories and findings regarding cash holdings. Some of their empirical methods will be incorporated to this thesis as well.

2.2. CEO types

In this next paragraph, I examine the literature for different CEO traits that could influence cash holdings in a given company. First, I consider the managers’ ability. It is hard to measure the abilities of a human being, but Custódio et al. (2013) manage to formulate a rather accurate proxy. They look at CEOs work experiences and determine whether the accumulated human knowledge is general in a sense that it is beneficial to any company and valuable in any given situation. The other side of the coin is when the managerial capital is concentrated and limited to one specific company or role. They use five characteristics to construct this general ability index. They consider the number of (1) positions a CEO held, (2) firms worked in, (3) different industry workplaces, and the existence of (4) previous CEO position and (5) conglomerate work experience. Then they use a factor model to compute one index which helps to enhance the explanatory power of the regression by reduce multicollinearity. They divide the CEOs into two types: generalist and specialist (an index value below the median). They find an average of annual pay premium for

12 generalist CEOs of 19% relative to specialist. This compensation package premium is even larger when managers are hired to implement complicated and universal tasks like acquisitions and restructuring. Also, this generalist pay premium is higher in industries that experienced shocks in the last decade. Furthermore, the authors also show that CEOs’ pay increases the most when firms externally hire a new CEO and switch from a specialist to a generalist manager. They raise alternative explanations for their findings but they confirm that the conclusion is robust. In this thesis, I will use the publicly available data of this research to examine how generalists and specialists affect cash holdings.

Secondly, I apply overconfidence as a type of CEO classification. Dividing CEOs whether they are overconfident or not, we can examine if there is difference in cash holdings between the firms that employ overconfident CEOs and non-overconfident ones. Malmendier and Tate (2005) measure the relation between overconfident managers and investment policies. They argue that overconfident investors believe that they better than average and that their investments can yield higher returns. Overconfident CEOs overestimate the profitability of available projects. If the firms’ internal funds allow and there is no restriction from other sources, they overinvest. However, overconfident CEOs are reluctant to raise external funds, issue equity because they are convinced that their firms’ stocks are undervalued. When short on cash, these companies cut back on investment. To test these hypotheses, they develop proxies for overconfidence. These measures are all based on the same concept. A rational investor diversifies its portfolio by not holding one kind if shares. When CEOs’ option compensation packages are vested, in order to diversify, they should sell those stocks if they are in the money. The absence of this behavior is argued to be caused by the overconfidence of the CEO. The CEO thinks that the stocks are undervalued and waits to sell it later, when it is better priced. A detailed description on one proxy that I implement can be find in section III. The results are confirming the authors’ hypothesis. Overconfidence has a significant positive effects on the firm’s sensitivity of investment to cash flow, overconfident CEOs on average use cash flow to make more investments. This result is most prominent in equity-dependent firms. Further, they show that CEOs with engineering backgrounds have higher investment-cash flow sensitivity, while managers with finance education display lower investment- cash flow sensitivity. The paper from the same authors (Malmendier and Tate, 2008) examine overconfident CEOs’ acquisition activities. They utilize the previous proxies for overconfidence and implement another where they look at how CEOs are portrayed in the press. They find that overconfident CEOs are

13 an average make more acquisitions than non-overconfident counterparts. This happens the most when firms only use internal funds or if they partake in diversifying mergers. Moreover, the returns after the merger announcements for firms with overconfident managers are reflecting a significantly larger negative market reaction as well. Besides the agency cost, this paper provides an alternative cause of firms engaging in value destroying activities. The study from Brown and Sarma (2007) is also investigating the effects on acquisition activity related to both CEO overconfidence and dominance. They examine the Australian market and confirm what the previous study found, overconfident managers make more acquisition. However, they also find that CEO dominance is at least as significant of a factor than overconfidence. Manager traits can influence other aspect of firm policies. For example, Galasso and Simcoe (2011) show that overconfident CEOs are also more innovative on average. The authors measure innovation activity by citation-weighted patent numbers and find a positive relation with managerial overconfidence. This effect is stronger in more competitive industries. Furthermore, Deshmukh et al. (2013) investigates overconfidence and dividend policy. The theory argues that overconfident CEOs believe that their stock is undervalued which means that they are reluctant to fund investments with issued equity. In order to be able invest into positive net present value projects they do not pay dividend to accumulate the necessary internal funds. Consistent with the prediction, they find that overconfident managers pay out lower levels of dividends. Also, the difference between high- and low growth firms’ dividend policy is lower when firms employ overconfident CEOs. Further, they find that market reacts more positively after a dividend increase announcement if they had higher uncertainty about the CEO overconfidence level, because increased dividend implies that the CEO is more rational (less overconfident). In addition, Otto (2014) studies the relation between CEO overconfidence and compensation packages. Using stock options (as implemented by previous papers) and earnings forecasts to proxy overconfidence, they find that managers that are considered overconfident payed less in total, have lower stock grants and decreased bonuses compared to non-overconfident CEOs. If non-overconfident CEOs’ actions lead to value destroying mergers, then these managers should be fired more often. Campbell et al. (2011) seeks to reveal whether this statement is true and to what degree. They first determine that overconfidence leads to overinvestment but it is true for the other spectrum of confidence, low-confident managers tend to underinvest. They show that managers with both of these extreme confident-levels are significantly more likely to be fired than moderate confident managers. Next, they demonstrate that forced turnover can only be

14 found if the board is aligned with maximizing shareholder value. Lastly, they prove that overconfidence is considered only for forced turnovers, they do not affect other types of turnovers. The findings are consistent with the hypothesis that there is an optimum confidence level for the CEO. One of the measures the authors use for overconfidence is a modified proxy from the work of Malmendier and Tate (2005). It is an important contribution, because following these steps others can test overconfidence on a bigger scale. Campbell et al. (2011) also implement a second proxy based on Malmendier and Tate (2005) where they link investment and overconfidence. These measures are implemented in my research as well and described in details later in the section III.

There is little evidence in the academic literature on how CEOs can affect cash policies, let alone how manager characteristics cause any change in a firm’s cash level. However, there is the study of Liu and Mauer (2011) in which they examine the relation between risk-taking managerial compensation incentives and cash holdings. They argue that for risk-averse CEOs to maximize shareholder value, they need to be incentivized by equity based compensation like stock options. These options encourage the managers to take riskier investments, thus maximizing shareholder value, however, these riskier activities bring uncertainty for the bondholders. The authors find a positive relation between the sensitivity of managerial compensation to stock price volatility (vega) and cash holdings. Higher equity-based incentives for a CEO result in higher cash holdings in the firm. There are two explanations for this. One is that risk taking can increase the costs of external found, so the firm needs to hoard cash to be able to invest in the future. The second is that stockholder-bondholder conflict is stronger when riskier investments are made, thus bondholders might protect themselves by requiring a buffer against the riskiness of the firm which helps reduce the cost of debt. Next, they show that vega has a negative effect on the value of cash from the shareholders’ perspective. These finding is consistent with the bondholder theory, because if an additional dollar does not benefit shareholders, then it helps high vega firms to mitigate the shareholder-bondholder conflict. Finally, there is a study by Orens and Reheul (2013) that investigates CEO characteristics and cash holdings. After taking into account the three main motives of holding cash, older managers (compared to younger ones) and CEOs working only in one industry (compered to multiple-industry experienced CEOs) have higher level of cash owing to the precautionary motive. For younger and other-industry experienced CEOs the opportunity cost of cash is more dominant and as a result they hold less cash. Other CEO demographics like tenure and education have no effects on cash levels. These findings are from Belgium privately

15 hold firms, and the study is limited by the low number of observations and that they only include industrial SMEs.

2.3. Hypotheses development

According to Custódio (2013) generalists are considered to have better abilities than specialists in today’s economic environment (they receive a premium). The ability index is designed to capture the more universal traits of the CEO, and assign values accordingly. After computing the index, they show that generalists have larger compensation packages, which implies that the firms and the shareholders value generalist CEOs more. This could suggest that bigger and better situated firms are targeting generalist CEOs to hire, however, according to the study, it is not the case. The authors show that generalists are even more valued when they are hired to accomplish complicated tasks, like acquisition. Several studies identify a significantly negative coefficient for acquisition on cash holdings (e.g.: Bates et al. 2009, Pinkowitz et al. 2013). Additionally, generalists are considered to have a broader knowledge and thus might be better in producing future revenues, which would lessen the precautionary motive to hold cash. In light of these assumption, my first hypothesis is the following:

H1: Firms with generalist CEOs hold less cash than specialist.

According to Malmendier and Tate (2008), overconfident CEOs overinvest if they have enough internal fund but reluctant to issue equity when they do not have sufficient capital. However, as the tradeoff theory suggest and shown by Opler at al. (1999) and Bates et al. (2009), there is a positive relation between capital expenditures and cash holdings. Thus, my second hypothesis is the following:

H2: Overconfident CEOs have larger cash holdings than non-overconfident ones.

I further test if financial constraints drive my results, whether there is a difference between financially constrained and unconstrained subsamples. Examining whether generalist hold lower cash reserves in financially constrained and unconstrained subgroups would further improve my research. Unfortunately, considering the general ability index there is not enough data to develop a theoretically genuine hypothesis. As argued in the literature review, financial constraints are a dominant determinant of cash holdings, thus I predict that the ability of the CEO will not influence the cash holdings of the company in a financially constrained setting.

16 H3: Financially constrained firms hold more cash than unconstrained ones regardless the manager ability specification.

I also investigate if financially constrained firms with overconfident CEOs implement similar cash policies to unconstrained ones or due to the firm being distressed the impact of the manager characteristic is altered one way or the other. Financial constraints forces firms to save cash from their cash flow (Almeida et al., 2004) and keep a higher cash level which is in accordance with the precautionary motive. Also, as discussed, overconfident CEOs if cash is available make more investment, but this behavioral is associated with higher cash balances (Opler et al., 1999 and Bates et al., 2009). Financial constraints and overconfidence should both increase the cash levels of the firm, there should be no reason for overconfident managers to change the cash policy when being financially constrained, rather enhance it. Thus, my next hypotheses are the following:

H4: Financially constrained firms with overconfident CEO hold more cash than with a non-overconfident CEO.

H5: Financially unconstrained firms with overconfident CEO hold more cash than with a non-overconfident CEO.

H6: Financially constrained firms hold more cash than unconstrained ones if they employ an overconfident CEO.

3. Data

Examining the effects of CEO types on cash holdings I first use the Compustat Fundamental Annual database to acquire financial information on the US firms from 1993 until 2007. This sample period restriction is due to the availability of the general ability index measure provided by Custódio at al. (2013). It limits the size of the dataset, but it is still adequate enough to compute meaningful analysis. Next, Execucomp database was used to collect information on CEOs characteristics and information on their option compensation. Following the existing literature, I imply the following restrictions to sample. First, due to regulatory reasons, I exclude all financial firms (sic codes 6000-6999) and utilities (sic codes 4900-4949). Secondly, to be included into the dataset it is required to have at least $1 million in asset, that firms have non-missing data on cash and short term investment and sic codes are present as well. Lastly, I drop every firm-year duplicate

17 observation from the sample. The final sample consists of 2000 unique firms and 16898 firm-year observations over the period of 1993-2007.

3.1. Variables

3.1.1. Cash holdings

My research seeks to shed some additional light on the cash holdings phenomenon which means that my dependent variable is going to be the cash, more precisely cash ratio. As a reminder, when I – and most of the literature – mention cash it refers to cash and short term investments (or cash and cash equivalents). There are several definitions for cash ratio in the academic literature. Namely, cash to total assets, cash to net assets (net asset is total assets minus cash), the natural logarithm of cash to net assets, cash to sales, and the log of cash to sales. The most common measure is the cash to assets, but after Opler et al. (1999) introduced logarithm of cash to net assets4

it became widely used as well. In my thesis I will apply the cash to assets definition and as a robustness test I will check the results with the log of cash to net assets measurement.

3.1.2. CEO type measures

My research is centered around how CEO characteristics affect cash holdings. I consider two major traits: ability and overconfidence. My first measure of CEO type is borrowed from Custódio at al. (2013) in which they constructed the General Ability Index (GAI) As mentioned in the literature review, they are investigating whether a generalist CEO is payed more compared to a specialist. They consider five aspect of a manager which could measure how general the CEO ability is, to what degree the experience is universal and thus applicable across all firms. The index is computed by using principal component analysis and includes the number of positions, number of firms, number of industries, whether the CEO had multiple CEO positions and conglomerate experience. The data consist of publicly traded US and international firms from 1993 until 2007 and it is publicly available for usage. 5

Secondly, I measure how CEOs’ overconfidence could play a role in a firm cash policy. Managerial overconfidence has a considerable literature but as it is a human feature it is impossible

4 _{At the beginning of the study he does not mention the logarithm, but in the headers of the regression tables the} logarithm is added

18 to measure directly. There are some promising proxies to asses this attribute. One way to measure overconfidence is introduced by Malmendier and Tate (2005) which examines the option based compensation and the CEO decision to exercise them. Assuming that the CEO is rational and risk averse the authors argue that after the stock option is exercisable and it is profitable one should exercise it and diversify (given that they have to hold shares of their own company) to avoid holding all of its stock based capital in one firm. Malmendier and Tate (2005) consider a CEO overconfident if they hold the stock options that are exercisable and 67% in the money. They use a dataset that has detailed information on option packages which is not publicly available so I follow Campbell et al. (2011)’s methodology constructing the modified variable. For a CEO to be considered overconfident, one has to exhibit this option-holding behavior twice during the sample. First, computing Holde67, they calculate an average exercise price from all stock options the CEO has, using Core and Guay’s (2002) approximation method. Particularly, dividing the unexercised exercisable stock option value (ExecuComp variable OPT_UNEX_EXER_EST_VAL) with the number of

these option (OPT_UNEX_EXER_NUM) we get the average realizable value of one option. Next, they

calculate an estimation of the average exercise price by subtracting the average realizable value per option from the stock price (PRCC_F). Finally, to get to the average percent of moneyness they divide

the average realizable value by the average exercise price. If the CEO had their options in the money at least 67% (moneyness >= 67%) and this happened twice during the sample period, then the CEO is considered overconfident from the first time of having displayed this behavior and onwards. Campbell et al. (2011) used a 100% moneyness threshold for high overconfidence which I will include in the robustness test. They acknowledge that this measure holds some potential concerns, but they do run a quite extensive robustness check which confirms that the method should be accurately reflecting the overconfidence levels. Specifically, my first proxy for overconfident CEOs is Holder67, a dummy variable which takes the value 0 if it’s a non-overconfident manager and 1 if it is an overconfident CEO.

My second measure of overconfident CEOs is also from the work of Campbell et al. (2011) which exploit the findings of Malmendier and Tate (2005). The former study employs the latter’s findings that overconfidence affects investments so they categorize CEOs with high (low) optimism if their firms’ industry-adjusted investment rates were in the top (bottom) quintile in two consecutive years. I relax the consecutive two-year criteria to any two years during one’s tenure. Following the Malmendier and Tate (2005) we measure the investment rates as capital expenditure

19 divided by the previous year total net property, plant and equipment (PPENT). My second variable

for measuring overconfidence is Investment level, which is a dummy variable being 0 if the CEO is non-overconfident and 2 if the CEO is overconfident provided the above discussed criteria.

3.1.3. Financial constraints

As Almeida et al. (2004) point out there is no best practice available to measure financial constraints, therefore, following their study I construct five variables. Fist, computing the payout ratio for each year, I take the sum of total dividends payed and stock repurchase and divided by the operating income before depreciation ((DVT + PRSTKC) / OIBDO). Then, I assign every firm each year

to low (high) payout ratio if they have lower (higher) values than the bottom (top) three deciles threshold of the annual payout ratio. A firm is considered financially constrained if it is in the bottom 30%, moderately constrained if between 30-70%, and unconstrained if in the top 30% of firms. The dummy variable Payout ratio takes the values of 0, 1 and 2 for financially constrained, in between and unconstrained firms, respectively. The reason behind using the payout ratio is that financially constrained firms should have significantly lower payout ratios (Fazzari et al., 1988).

My next variable is size, but note, that it’s calculated and used a bit differently as the Firm Size control variable. I rank the firms each year based on their total assets and allocate them into three groups with the same method as the payout ratio. The dummy variable Size takes 0, 1 and 2 for financially constrained, in between and unconstrained firms, respectively The intuition behind the size of the firms is that smaller firms have difficulties to receive external financing due to being less likely to be known therefore less likely to have connections to the capital market as Almeida et al. (2004) explains.

The third variable that I construct is Bond rating. If a firm had no bond ratings and had zero debt or it had missing data for the bond rating, then I determined those firms financially constrained. Having one bond rating during the whole sample period a firm would be considered unconstrained. Firms that had no rating nor debt (or missing) also regarded as unconstrained. I use the Compustat item SPSDRM to obtain the necessary information on the firms’ bond ratings, however, note that the data is only available until 1998. So firms that do not have any rating before 1999 are considered constrained, even though it could easily be the case that there are many companies that had issued bonds with ratings in later years and thus should be viewed as unconstrained.

20 The fourth proxy is based on Bates et al. (2009) study where they argue that negative net income is considered as a good measure for financial constraints. The variable Negative income is a dummy that takes the value 0 if net income was negative at a given year which means the firm is financially unconstrained otherwise 1 meaning the firm is unconstrained.

Lastly, the widely used Kaplan-Zingales index (KZ index) is computed. Several studies are using the original article of Kaplan and Zingales (1997) to construct the KZ index (e.g.: Almeida et al., 2004 and Malmendier and Tate, 2005) and I follow suit. The index is defined as:

𝐾𝑍_𝑖𝑡 = −1.002 ∗ 𝐶𝐹𝑖𝑡 𝐴𝑇𝑖𝑡−1− 39.368 ∗ 𝐷𝑖𝑣𝑖𝑡 𝐴𝑇𝑖𝑡−1− 1.315 ∗ 𝐶𝑖𝑡 𝐴𝑇𝑖𝑡−1+ 3.139 ∗ 𝐿𝑒𝑣𝑖𝑡 + 0.283 ∗ 𝑄𝑖𝑡

Where cash flow (CF) equals income before extraordinary items plus depreciation (IB + DP),

dividend (Div) is the total dividend (DVT), cash (C) is as always the cash and short term investments

(CHE), leverage (Lev) is the long term debt plus the debt in current liabilities divided by the previous

two variables and stockholders’ equity ((DLTT+DLC)/(DLTT+DLC+SEQ)), and Tobin’s Q (Q) which is

the ratio of total assets, minus the book value of equity and deferred taxes, plus the market value of equity to total assets ((AT-CEQ -TXDB +(CSHO*PRCC_F))/AT). The first three variables are normalized

by the one-year lag of total assets and all variables are winsorized except cash. Div and Lev at 1% level, CF at 1% level of the bottom tail and Q at 1% level of the top tail. Then the according to the index values on a yearly bases the firms are sorted to three categories whether they are in the bottom or top 30 percentile or in between. The KZ index is a dummy variable that takes the values of 0, 1 and 2 for financially constrained, in between and unconstrained, respectively.

3.1.4. Control variables

The explanatory variables used in the regressions are inspired by the existing literature but mainly rely on Opler at all. (1999) and Pinkowitz et al. (2013). The following list provides an overview of which control variables (the Compustat data items) are included in the regressions:

1) Market-to-book ratio. It measures the growth and investment opportunities of a firm. Firms with higher market-to-book ratio have higher cash reserves in order to be able to invest and not pass out on positive NPV projects. To calculate the market-to-book ratio I take the book value of assets and subtract the book value of equity then add back the market value of equity and divide this with the book value of assets ((AT-CEQ+(CSHO*PRCC_F))/AT).

21 2) Firm Size. In theory, larger firms should be able to better excess financing through external sources which means they are less like to be financially constrained which in turn suggest that they do not suffer the costs of having lower cash reserves. It is expected to have a negative relationship with cash holdings. Firm Size is the logarithm of the book value of total assets (AT).

3) Cash Flow (CF). Ceteris paribus, firms with higher cash flow hoard more cash, therefore there should be a positive correlation as well. Cash Flow is calculated as the operating earnings after interest, taxes and dividends but before depreciation, then we scale it with the total assets ((OIBDP-XINT-TXT-DVC)/AT).

4) Industry Volatility. Considering the precautionary motive, firms with higher industry risk should hold larger cash reserves. Industry volatility is computed as the mean of the standard deviation of cash flow over assets for 10 years by industry which is determined by 2 digit SIC codes. A minimum of 3-year requirement was set to calculate the variable.

5) Net working capital (NWC). NWC is considered as a substitute to cash and cash equivalents. Thus, I expect a negative relation between cash holding and NWC. The variable is calculated by deducting the cash from NWC and then divided by total asset to get a net of cash measure ((WCAP-CHE)/AT).

6) Capital expenditures (Capex). There are different theories in the literature that are contradicting in terms of the relationship between cash holding and capital expenditures. It can be the case that firms due to high investment costs have lower cash reserves or because they invest more and/or financially distressed keep more cash. Capex is simply calculated as the capital expenditure to total assets (CAPX/AT).

7) Leverage. Leverage is also a dependent variable that is controversial regarding the relationship with cash. One can argue that considering the pecking order theory a firm should accumulate internal funds when issuing debt is costlier which also aligns with the precautionary motive that if a firm cannot access the external resources then it should hold cash. Moreover, if a firm has too much debt then it can be constraining so it would spend cash to reduce the amount owned. However, as Acharya et al. (2007) show financially constrained firms with high hedging needs have a strong propensity to save cash out of cash flows while leaving their debt positions unchanged. Leverage is debt in current liabilities and long term debt divided by total assets ((DLTT+DLC)/AT).

22 8) R&D. R&D also measure growth opportunity. R&D has a high adjustment cost and firms with financial constraints could hold more cash to smooth R&D expenses as Brown and Petersen (2010) argue. However, because R&D heavy firms invest more, ceteris paribus, their cash levels should be lower. I calculate R&D as R&D over sales (XRD/SALE). 9) Acquisitions. The expenditure regarding mergers and acquisitions are cash consuming (if

it’s not payed entirely with stocks), thus it is expected to have a negative relation to cash holdings. Acquisitions are calculated as acquisitions to total assets (AQC/AT).

10) Dividend. Paying dividend is a proxy for financially unconstrained firms. Firms that pay dividends are less likely to suffer from the precautionary motives because they are generally more established and considered less risky. This would suggest a negative relation to cash holdings. Dividend is a dummy variable that takes 1 if the firm payed common dividend in that year and 0 otherwise (using DVC).

11) Debt issuance. Debt issuance is net debt issuance to total assets, where net debt is calculated from the long term debt issuance minus the long term debt reduction ((DLTIS-DLTR)/AT). Pinkowitz et al. (2013) find that cash holdings increase with debt issuance.

12) Equity issuance. Equity issuance is measured by the net equity issuance, the sale of common and preferred stock minus purchase of common and preferred stock, to total assets ((SSTK-PRSTKC)/AT). The same study finds similar result for equity issuance as for debt issuance. Moreover, Bates et al. (2009) argue that these last two measures are important to control for because cash is immediately affected after raising capital.

3.2. Descriptive statistics

Table 1 summarizes the descriptive statistics for the key variables of my methodology. It consists of 16898 firm-year observation of publicly traded US firms from 1993 until 2007. The interest of my research is the cash to asset variable, in other words the cash holdings. On average, firms hold 14.5% of their total assets in cash and cash equivalents. However, the median is only 7% which means that there are some large outliers in the sample tilting it to the right. This skewedness is even more prominent in the to-book ration and NWC to asset. The market-to-book ratio has a mean of 2.168 and a 1.5 standard deviation, where NWC to assets in average is 0.078 with a standard deviation of 0.147. My first proxy for managerial characteristics is the general ability index. It ranges from – 1.5 to almost 7 with a mean (median) of 0.023 (-0.147). The

23 3rd _{percentile of the index is still under 1 suggesting that this variable is also skewed to the right. I}

implement two overconfident measures. Holder67 is a dummy variable that takes a value 1 if the manager is overconfident, 0 otherwise. The mean is close to half suggesting that both spectrum of overconfidence level is adequately represented. The second proxy for overconfidence is the Investment level. It takes the value 0 if the CEO considered non-overconfident, 1 if moderately confident and 2 if the manager is overconfident. As the mean is around 1, both non- and overconfident managers are evenly presented.

Acquisitions, R&D, Capex and Industry volatility are winsorized at 1% level; the NWC and Cash flow are winsorized by their bottom tail at 1% level; Market-to-book ratio is winsorized by the top tail at 1% level and Leverage is winsorized to be between 0 and 1.

4. Methodology

In the previous section, I demonstrated the sample data, the key variables and their construction. In this next part, I will explain the methodology to test my hypotheses. First, I show the process of measuring the relationship between the CEO characteristics and cash holding. Secondly, I display the financial constrained methodology.

I examine two CEO characteristics: one is the ability of the manager, second is the level of overconfidence. In order to test my hypotheses for both traits’ effects on cash holdings, I will implement several regression models. All of the empirical models have the same base equation which is the following:

𝐶𝑎𝑠ℎ 𝐴𝑠𝑠𝑒𝑡⁄ _𝑖,𝑡 = 𝛼 + 𝛽₁∗ 𝐶𝐸𝑂_𝑖,𝑡𝑇𝑦𝑝𝑒+ ∑𝐾_𝑘=1 𝛽_𝑘∗ 𝑋_{𝑘,𝑖,𝑡}+ 𝜀_𝑖,𝑡 (1)

The dependent variable is the cash scaled by total assets. The coefficient of interest is going to be 𝜷𝟏 which determines whether my hypotheses were correct. 𝑪𝑬𝑶𝒊,𝒕

𝑻𝒚𝒑𝒆

is the characteristic of the manager: ability or overconfidence. The next part in the equation ( 𝜷𝒌∗ 𝑿𝒌,𝒊,𝒕 ) connotes the twelve

control variables described in the previous section. It consists of the Market-to-book ratio, Firm Size, Cash Flow, NWC, Capex, Leverage, Industry Volatility, R&D, Acquisitions, Dividend dummy, Debt issuance and Equity issuance. The last term refers to the residuals.

This equation is the foundation of my models. I will include a pooled ordinary least square (OLS) regression; a Fama-MacBeth model; firm, industry and CEO fixed effects and also each

24 fixed effects regression with the addition of year fixed effects. These are the most common regression types used in the literature to describe relationships to cash holding. The OLS regression is included to have a basic understanding of the variables’ behavior. In a panel data setting, it is common that observations are not all independent, thus causing the standard errors to be incorrect in this model. The method used by Fama and MacBeth (1973) (referred to as Fama-MacBeth model) is a sectional regression computed annually, evaluating each year as a separate cross-section. This method gets rid of the problem discussed for the OLS regression erasing the serial correlation between the residuals. Furthermore, by using fixed effects regressions, I eliminate unobserved, time-invariant, entity specific omitted factors that can influence the outcome of my regression. Including fixed effects means that we only observe changes within one entity, excluding time-invariant factors . Specifically, when including firm fixed effects and examining within changes in firms, the only time that a change in managers’ traits happens is when the firm has a turnover (or in the overconfidence measure case, when the CEO becomes overconfident). Moreover, including year fixed effects, I control for any omitted variables which is constant across all entities (firm, industry, manager) at each year but vary over time, thus capturing the effects of an aggregate time trend. Heteroskedasticity and autocorrelation consistent standard errors (HAC) are used in the fixed effects regressions by clustering with the respected fixed effects. This estimate assumes no correlation across entities and robust to both arbitrary correlations within clusters and heteroskedasticity.

Implementing these models, I investigate the CEO traits on cash balances independently using two separate analyses. For the first measure of CEO type, I use the General Ability Index (GAI) developed by Custódio at al. (2013). They classify CEO ability ranging from being specific and concentrated to being general and well-rounded. As explained in the literature review and stated as the first hypothesis, I expect that firms employing generalist CEOs have lower cash holdings compared to specialist.I test the hypothesis with equation (1), in this case using the GAI variable, predetermined and provided by the previously mentioned study. It ranges from -1.504 (most specialist) to 6.868 (most generalist) and I expect a negative coefficient as discussed before (H1). The second measure for CEO types is overconfidence. The methodology is replicated here with the exception of the main independent variable, our interest in the regression. Instead of using the general ability index, I apply the overconfidence dummy variables separately, namely Holder67 and Investment level, providing each proxy a regression. The overconfident measures are described

25 in details in the previous section. I would expect that the coefficient of interest is positive as explained in the hypotheses development section (H2).

Lastly, to test my hypotheses regarding financial constraints, I will divide the sample to financially constrained and unconstrained groups and rerun the regressions that had significant results from the first two hypotheses. In the case of the Bond Rating and KZ index overconfidence measures I divide the sample by considering if the values are 0 or 2 which corresponds to financially constrained and unconstrained firms. I exclude firms that are considered in between. For the other dummy proxies, they are binary, reflecting the presence and the absence of financial constraints. The definitions for the financial constraints variables are previously specified in the section III. Regarding the general ability index, I assume that the financial constraints should still be effective enough to balance out the presumed negative effects of generalist on cash holdings. I anticipate a larger coefficient on financially constrained firms than unconstrained ones (H3). For the overconfident measures I developed three hypotheses. The first two (H4 & H5) predict a positive and statistically significant coefficient on the overconfident proxy, meaning with overconfident manager on the board both financially constrained and unconstrained firm hold more cash. Also, I predict that in financially constrained firm overconfident CEOs hold more cash than in unconstrained firms, which suggest a higher coefficient for the financially constrained subgroup (H6). In order to thoroughly examine the effects of financial constraints, I use the base regression equation (1) and add the constraints dummy proxies and their interaction term with the GAI or the overconfident variables. These additional tests allow me to further investigate my last four hypotheses (H3-H6) and have a more complete conclusion.

5. Empirical Results

In this section, I provide the empirical results of my hypotheses. First, I examine the relation between CEOs’ ability and cash holdings in publicly traded US companies. Then, I investigate the effects of CEO overconfidence on firms’ cash levels. Lastly, I present additional tests on financial constraints and its effects on the CEO types’ cash hoardings.

5.1. General ability index and cash holdings

There had been several regression forms implemented by the academic literature so far. I include the following ones: the ordinary least square (OLS), the Fama-MacBeth (F-M), and the fixed

26 effects regressions as discussed in the methodology. Model I and II uses firm fixed effects, but the later includes year fixed effects. Model III and IV has industry fixed effects, and again, the second uses year fixed effects, lastly, Model V and VI includes CEO and the later year fixed effects as well. Each regression is calculated for the entire sample period (1993-2007) with heteroskedasticity-robust standard errors. My dependent variable in all cases is the ratio of cash to total asset. The variable of interest is the GAI index which here is defined as the index associated to each CEO in each year, hence can change over time. The index values range from -1.5 to about 7. The original paper (Custódio et al., 2013) uses above and below median values, which will be examined in the last financial constraints analysis and in the robustness section as well. The control variables are described in the Data section in depth.

Considering each control variable in Table2, most of them are consistent across all models in sign, in significance level and in magnitude as well. Moreover, most of the variables are in line with the findings of the existing literature (e.g.: Opler et al., 1999, Harford et al, 2008, Bates et al., 2009 and Pinkowitz et al., 2013). The market-to-book ratio has positive and significant coefficient at the 1% level, which means that firms that have higher growth opportunities are saving cash in order to avoid passing up on positive net present value investments in the future due to the lack of capital. Firm size and cash holdings display a negative relationship as expected. Larger firms can better access the capital market and do not need to have a cash buffer for unexpected depletion of future cash flows or for unhedged events as the precautionary motive would suggest. My cash flow measure is insignificant which is unexpected, but using different measurements yield similar, insignificant results (missing values treated as zero, or using the definition as in construction of the KZ index).One exception to the similarities between the literature’s and my findings is the industry volatility. Both significance and the magnitude of the coefficients is proportionally lower that can be found in other studies. Net working capital is treated as an equivalent of cash. The significant and negative coefficient is proof of this and is consistent with the literature as well. Capital expenditure and acquisitions is said to have the same relation to cash, because both essentially represent investments. There is no clear academic agreement on the relation between these variables and cash holdings. On the one hand, the firm knows that if the upcoming years are investment-intense then it is well advised to save cash for upcoming opportunities, resulting in a positive effect on cash holdings. On the other hand, investments are cash consuming, leaving the company with lower cash levels, creating a negative relation. In my sample these two variables