The relationship between CEO risk-taking incentives in unionized firms and exploratory innovation

The relationship between CEO risk-taking

incentives in unionized firms and exploratory

innovation

Daniël Hendriks 11419628 July 1, 2018

Master Thesis

MSc Finance: Corporate Finance University of Amsterdam

ABSTRACT

Extensive research has been conducted about the effect of executive

compensation on executive risk-taking incentives and the impact of unions on executive

compensation. However, the direct relationship between unions and risky innovation, i.e.

exploratory innovation, remains unexplored. This research studies the direct effect of labor

unions on firm-level exploratory innovation. The CEO equity pay ratio and the board

independence ratio are tested as moderating variables of the main relationship.

OLS and differences-in-differences regressions are conducted using panel data on

North American firms from multiple databases for the period of 1992 through 2009. The

differences-in-differences regression makes use of an exogenous shock to unions’ bargaining

powers from legislation changes. The differences-in-differences results provide strong

evidence of a negative direct relationship between union bargaining power and exploratory

innovation. Support was also found for a negative moderation of the equity pay ratio on the

effect of union bargaining power on exploratory innovation. Weak support was found for a

negative moderation of the effect of board independence on the negative direct relationship.

The academic, managerial and public policy implications of this study are discussed at the

end of the thesis as well as the suggestions for future research.

Keywords: Labor unions · Risk-taking incentives · CEO compensation · Exploratory

innovation · Equity pay · Board independence · Right-to-work legislation

Statement of originality

This document is written by Daniël Hendriks 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.

TABLE OF CONTENT

1. Introduction ... 5

2. Literature Review ... 8

2.1 The relationship between CEO compensation and risk-taking incentives ... 8

2.2 The relationship between unions and CEO compensation ... 10

2.3 The relationship between unions and innovation ... 12

2.4 Research gap and research question ... 13

3. Theoretical framework ... 15

3.1 The effect of unions on firms’ exploratory innovation ... 15

3.2 The moderating effect of the share of equity pay in CEO compensation ... 17

3.3 The moderating effect of board independence ... 18

4. Data and variables ... 21

4.1 Sample selection and data collection ... 21

4.2 Variables ... 26 4.2.1 Dependent variables ... 26 4.2.2 Independent variables ... 29 4.2.3 Moderator variables ... 29 4.2.4 Control variables ... 30 4.3 Summary statistics ... 32

5. Statistical models and results ... 37

5.1 OLS models ... 37 5.2 OLS results ... 40 5.3 Differences-in-differences models ... 43 5.4 Differences-in-differences results ... 46 5.5 Robustness tests ... 49 6. Discussion ... 52 6.1 Academic relevance ... 52 6.2 Managerial implications ... 53

6.3 Public policy implications ... 53

6.4 Limitations and suggestions for future research ... 54

7. Conclusion ... 58

REFERENCES ... 61

APPENDIX ... 68

LIST OF TABLES

Table 2 Variable descriptions and operationalization ... 68

Table 3 Cross-correlations ... 69

Table 4 Exploratory innovation proxies and the industry unionization rate ... 70

Table 5 Exploratory innovation proxies and the union dummy ... 71

Table 6 Exploratory innovation proxies and the right-to-work transition dummy ... 72

Table 7 Robustness checks ... 73

LIST OF FIGURES

Figure 2 The development of the share of equity pay in total CEO compensation ... 36

1. Introduction

A firm’s innovation process can involve both exploration and exploitation (March, 1991). Exploratory innovation is the radical design of new products, services or distribution channels to enter unfamiliar or distant market segments (Benner & Tushman, 2003; Danneels, 2002; Jansen, Van Den Bosch & Volberda, 2006). It requires new knowledge or the departure of existing knowledge in the firm (Jansen et al., 2006; Levinthal & March, 1993). Exploitative innovation on the other hand is the incremental design of existing products, services or distribution channels (Benner & Tushman, 2003; Jansen et al., 2006.). Exploitative innovation only broadens existing knowledge and competences (Abernathy & Clark, 1985; Jansen et al., 2006). The returns to exploratory innovation are more uncertain than the returns to exploitative innovation. Therefore, exploratory innovation bears more risk to the firm and its executives than exploitative innovation.

Executive compensation is an essential mechanism to attract, retain and incentivize top executives. Much of the literature has assigned crucial importance especially to the incentivizing mechanism of compensation for firm innovation. Many studies find that efficient executive

compensation designs can reduce risk-related agency problems (Agrawal & Mandelker, 1987; Beladi & Quijano, 2013; Brockman, Martin & Unlu, 2010; Chava & Purnanandam, 2010; Coles, Daniel & Naveen, 2006; Core & Guay, 1999; Gormley, Matsa, & Milbourn, 2013; Heron & Lie, 2013; Jensen & Meckling, 1976; Knopf, Nam & Thornton, 2002; Mao & Zhang, 2018; Rajgopal & Shevlin, 2002). The reduction of risk-related agency problems can stimulate executives to pursue more risky

innovation. Effective executive compensation designs are thus essential for innovation. The executive compensation in a firm is determined by a company’s compensation committee which is part of the company’s board of directors. The board of directors and the compensation committee are therefore crucial in influencing firm innovation.

Executive compensation is often criticized for being excessive or creating perverse incentives (Huang, Jiang, Lie & Que, 2017). Consequently, external or internal stakeholders often put pressure on executive pay. Several studies find that labor unions are one of the stakeholders who often target executive compensation (Huang et al., 2017; Singh & Agarwal, 2002). Unions target executive compensation because they are concerned that executives receive an excessive and unfair portion of

firm cash flow or that wage inequality harms employee morale (Huang et al, 2017). Unions have powerful tools at their disposal to exert pressure on companies. The most important tools are the right to strike, which is protected under the National Labor Relations Act (NLRB, 2017), and the use of shareholder proposals (Ertimur, Ferri & Muslu, 2011). Many studies indeed find evidence that unions are effective in reducing average chief executive officer (CEO) compensation (Banning & Chiles, 2007; Ertimur et al., 2010; Gomez & Tzioumis, 2006; Huang et al., 2017). More specifically, boards of directors very often use such CEO pay reductions as a signal to unions to improve their bargaining positions with unions.

The effects that termination risk, i.e. the risk for the executive of losing his or her job, can have on risk-taking behavior has been studied extensively. Much of the literature has found evidence that a higher termination risk is associated with less risk-taking and therefore also with less risky types of innovation (Balsmeier et al., 2017; Chakraborty, Sheikh & Subramanian, 2007; Frydman & Jenter, 2010; Manso, 2011). However, little literature has specifically focused on the effect of the risk of compensation reductions, i.e. a manager’s compensation risk, on executive risk-taking and

exploratory innovation.

The extant literature has thus studied two important relationships. On the one hand, the literature has described the relationship between executive risk-taking incentives from executive compensation and exploratory innovation. On the other hand, the literature has studied the relationship between unionization and chief executive officer compensation. However, no studies have focused on the direct effect of unions on risk-taking and exploratory innovation. To fill this gap in the literature, this thesis studies the direct relationship between labor unions and exploratory innovation.

I hypothesize that there is a negative direct relationship between the degree of unionization in a firm and exploratory innovation. Five proxies for exploratory innovation are used to estimate this relationship. I expect this negative relationship to be stronger for a higher degree of board

independence. In addition, I suggest that a larger share of equity pay in total CEO compensation will positively moderate the negative effect of unionization on exploratory innovation.

To test these hypotheses, I will conduct panel data regression analysis at the firm-level, based on a panel dataset from multiple databases. The dataset is composed of data from ISS, Compustat, NBER, USPTO, Kogan, Papanikolaou, Seru, and Stoffman (2017), the Union Membership and Coverage Database from the Current Population Survey (CSP), MSCI and NCSL. The final dataset contains data on firms from the United States during the period 1992-2009.

This thesis is organized as follows. In the literature review I will provide a detailed overview of the literature on the relationship between CEO compensation and risk-taking incentives, the relationship between unions and CEO compensation and the relationship between unions and innovation. The literature review is followed by the theoretical framework in which the hypotheses are developed. The subsequent chapter will address the data, sample and variables. The statistical models and results are reported in the following chapter. The results will be discussed in the discussion section. Finally, I conclude my thesis with limitations of the research and will I provide directions for further research.

2. Literature Review

The central subject of this thesis is exploratory innovation in unionized firms. Three strands of literature are particularly relevant for this subject. Firstly, the literature describing the relationship between executive compensation and executive risk-taking incentives can be used to make

expectations and theorize about the effects of executive compensation on exploratory innovation. Secondly, because of the first strand of literature it is also relevant to this study how unions influence chief executive officer compensation. Thirdly, a large subset of literature has focused on the effect of unions on R&D expenses and patent output. However, there are no studies studying the direct effect of unions on exploratory innovation, i.e. the type of innovation that is riskier and not measured by total patent counts. This section will address the consensus in the literature on the effect of unions on innovation in general. The next section will first describe the literature on the effects of CEO

compensation on risk-taking incentives.

2.1 The relationship between CEO compensation and risk-taking incentives

The distinction and benefits of the two types of innovation have long gained much attention in the scholarly literature (Holland 1975; Kuran 1988, March, 1991; Schumpeter 1934). The returns to exploratory innovation are uncertain, distant and often negative whereas the returns to exploitative innovation are more predictable, proximate and often positive (Geiger & Makri, 2006; March, 1991). In sum, exploratory innovation bears more risk to the firm given its unfamiliarity, distant returns and low predictability to the firm than exploitative innovation. Therefore, it is of interest how CEO risk-taking incentives are involved which are very important for exploratory innovation. Next, the agency theory concerned with resolving problems that can exist in agency relationships due to unaligned levels of risk preferences will be described.

There exists consensus in the literature that there are many possibilities for potential conflicts of interests between shareholders and executives. The central notion in the agency conflict is that there is separation of power between the executives who control the firm’s assets and the

shareholders, who own the firm. Consequently, a conflict can arise when the interests of the principal and the agent are misaligned. One important conflict that can arise is managers undertaking too little

risky innovation, when more risky innovation is preferred by the shareholders. The reasons for managers undertaking too little risky innovation are, among others: general risk-aversion; having undiversified human and financial wealth tied to the firms’ performance (Fama, 1980; Sheikh, 2012; Stulz & Smith, 1985); having a preference for simply enjoying ‘the quiet life’ (Bertrand &

Mullainathan, 2001); or career or reputation concerns (Bertrand & Mullainathan, 2003; Gormley & Matsa, 2016; Grossman & Hart, 1983; Holmström, 1979). Finally, heterogeneous internal and external firm pressures might increase manager myopia (Amihud & Lev, 1981; Belghitar & Clark, 2015; Hirhsleifer & Thakor, 1992). In sum, there are many possibilities for a misalignment between firm risk preferences and executives’ personal risk preferences.

Much scholarly literature finds that linking executive compensation to firm performance helps in overcoming the executive’s different risk preferences (Fama & Jensen, 1983; Gibbons & Murphy 1992; Mao & Zhang, 2018; Smith & Stulz, 1985). Much of the literature has focused on the alignment of a CEO’s compensation with firm risk as a stimulating factor for innovation. A central tool to align firm risk with executive risk is using equity pay as a compensation measure (Hall & Murphy, 2003). Equity pay adds convexity to executives’ payoffs and increases the sensitivity of executive wealth to firm risk (Core & Guay, 1999; Core & Guay, 2002; Milidonis & Stathopoulos, 2014; Smith & Stulz, 1985). Multiple previous studies have found evidence that executive

compensation sensitivity to stock return volatility does encourage risk taking behavior by aligning the risk tolerance of the shareholders and the chief executive officer (Agrawal & Mandelker, 1987; Beladi & Quijano, 2013; Brockman, Martin & Unlu, 2010; Chava & Purnanandam, 2010; Coles, Daniel & Naveen, 2006; Core & Guay, 1999; Gormley, Matsa, & Milbourn, 2013; Heron & Lie, 2013; Jensen & Meckling, 1976; Knopf, Nam & Thornton, 2002; Mao & Zhang, 2018; Rajgopal & Shevlin, 2002).

However, not all literature agrees on the effects of such equity compensation as an incentive for executives to take more risk. Several studies argue that executive investment decisions depend not only on how an executive’s compensation changes with firm risk but also on how the executive’s job is affected if the project fails. Heterogeneous combinations of corporate governance characteristics and equity pay can actually decrease risk-taking. One of such corporate governance characteristics is the termination risk for executives imposed by the board of directors.

Many studies consider how termination risk influences executives’ risk taking. Chakraborty et al. (2007) find that executive innovation decisions not only depend on how an executive’s

compensation changes with the risk of the project but also on the likeliness of being fired after bad performance. Jenter and Lewellen (2014) introduced the term performance-induced CEO turnover to define forced turnover that would not have happened had the results been good. In other words, termination risk equals the risk of performance-induced turnover. Core & Guay (1999) and Pratt (1964) argue that a trade-off exists between the wealth effect, which is affected by the risk-taking incentives, and the risk-aversion effect, which is affected by the termination risk. The wealth effect is the change in a manager’s expected wealth from a change in firm risk and risk-aversion is the impact of risk aversion on executive utility (Core & Guay, 1999; Pratt, 1964). Termination risk will increase the negative impact of risk-aversion on executives’ utility. Therefore, executives will take the termination risk into account when determining the risk levels of innovation decisions they will accept. In other words, the negative impact of increased termination risk can partially offset the positive impact of increased risk-taking incentives from equity compensation. Similarly, many studies argue that the interaction between higher innovation risk and termination risk depends on managers’ individual utility functions (Carpenter, 2000; Core & Guay, 1999; Ju, Leland & Senbet, 2003; Lewellen, 2006; Milidonis & Stathopoulos, 2014; Pratt, 1964; Ross, 2004; Smith & Stulz, 1985).

The findings of Chakraborty et al. (2007) corroborate the conjecture that an increase in the termination risk is associated with a decline in the corresponding firm’s stock returns volatility. Balsmeier et al. (2017) and Manso (2011) find that in situations with high termination risk managers will tend to pursue less exploratory innovation projects.

In sum, many scholars argue that the increasingly widespread use of equity-based executive compensation and the termination of CEO’s for poor corporate performance will exacerbate, rather than alleviate risk-related agency conflicts (Frydman & Jenter, 2010; Jenter & Lewellen, 2014).

2.2 The relationship between unions and CEO compensation

The way that unions influence CEO compensation has received some attention in the literature. It is not fully in agreement on the effect of unions on compensation. However, recent

literature has added new insights into the effects of unions on executive compensation and the components of it.

Huang et al. (2017) and Singh and Agarwal (2002) state that unions target executive compensation because they are concerned that executives receive an excessive and unfair portion of firm cash flow or that wage inequality harms employee morale. In other words, without necessarily bargaining for higher salaries of employees, unions can target excessive CEO compensation when it deems it necessary. The effect of chief executive officers’ compensation reductions through pressures from unions is called a dampening effect of unions (Singh & Agarwal, 2002).

The following relationships are found between the height of executive compensation and unions. Singh and Agarwal (2002) find that CEO’s in unionized firms received more pay than CEO’s in non-unionized firms. They find evidence of a ratcheting effect, which means that higher

compensation at lower levels in the firm will also lead to higher compensation at executives’ levels (Singh & Agarwal, 2002; Westphal & Zajac, 1997). However, numerous studies find evidence for a dampening effect of unions on CEO pay. Banning and Chiles (2007) find that on average CEO’s in unionized firms receive compensation that is significantly lower than in non-unionized firms. Ertimur, Ferri and Muslu (2010) find that unions that use shareholder voting to target excess CEO

compensation experience a significant reduction in CEO compensation. Huang et al. (2017), who instead focused one labor negotiations, find that looming union negotiations compel compensation committees to reduce chief executive officer compensation. Firms benefit from sending such strategic signals to increase the chance of moderate lower level wage increases, which will simultaneously maintain workers’ motivation levels (Huang et al., 2017). More specifically, reducing executive compensation might serve as a signal that executives and the board view the prospects to be negative and thereby instigating everyone, especially labor, to show moderation in contract negotiations.

Several studies corroborate the findings of Huang et al. (2017) that boards of directors send signals to labor unions. Bova (2013) argues that directors project a negative outlook to their unions to decrease the union’s bargaining position. Klasa, Maxwell and Ortiz-Molina (2009) and Matsa (2010) find that executives and boards strategically adjust their cash holding policies to improve their bargaining power over its unionized workers.

Of central importance to this study is the effect that unions have on equity compensation, which is found to be crucial in increasing risk-taking incentives and increasing risky innovation. There has however been little attention to the specific effect that unions have on equity compensation. Some studies find that unions do effectively reduce chief CEO equity compensation (Ertimur et al., 2011; Gomez & Tzioumis, 2006). Most recently, Huang et al. (2017) found that, especially around negotiations with labor unions, board of directors make use of CEO compensation reductions to send signals to the labor unions. Such signals are used to make union negotiations more easy. In addition, they find that this pattern is stronger for equity grants than for other components of CEO pay. This pattern probably emerges for equity pay especially because of the ease with which firms can determine both how many options to grant and when to grant them.

2.3 The relationship between unions and innovation

The subject of this study is exploratory innovation in unionized firms. However, there are, to my knowledge, no studies studying the direct effect of unions on exploratory innovation. A large literature has however focused on the effect of unions on R&D expenses and patent output. This section will briefly address the consensus in the literature on the effect of unions on innovation in general.

The scholarly literature has mainly studied the effects of union rent-seeking on R&D investment and innovation. Union rent-seeking refers to the desire by labor unions to retain rents of the returns generated in the firm. Consistent with this conjecture, Connoly, Hirsch and Hirschey (1986) find that R&D investments add relatively less to the market value of unionized. Therefore, unionized firms respond by investing less intensively in R&D. Findings by Betts, Odgers and Wilson (2001) and Hirsch (1991) corroborate these findings. In other words, union rent-seeking reduces the returns to research and development and produces a corresponding limiting influence on firms’ research and development intensity.

Most scholarly literature finds evidence that unions are related with a decline in R&D

expenditures or patent output. Bradley, Kim and Tian (2016) find that passing a union election results in a significant decline in patent quantity. Hirsch and Link (1987) find that innovative activity is

significantly less important for unionized firms than for non-unionized firms. Many additional studies find a negative relationship between unions and R&D expenditures (Acs & Audretsch 1987; Acs & Audretsch, 1988; Addison & Wagner, 1994; Audretsch & Schulenburg, 1990; Bronars, Deere & Tracy, 1994; Mansfield, 1980; Terleckyj, 1980). While these studies reveal that unions have a negative effect on firm R&D outputs, a few studies find that labor unions have a positive effect on innovative output. For example, Menezes-Filho, Ulph, Van Reenen (1998) find that in the U.K. R&D intensity appears to have been higher in unionized firms than in non-unionized firms.

In sum, these results indicate rather strong negative relations between unionization and innovation in general. However, the literature has paid no attention the relationship between unions and exploratory innovation.

2.4 Research gap and research question

Firstly, this chapter addressed how exploratory innovation bears more risk to executives and the firm than exploitative innovation. Therefore, the effects of CEO compensation on risk-taking incentives is of interest to this study. Many studies find that that equity pay that links the executive’s wealth to the firm’s risk indeed increases executives’ risk-taking (Agrawal & Mandelker, 1987; Beladi & Quijano, 2013; Brockman, Martin & Unlu, 2010; Chava & Purnanandam, 2010; Coles, Daniel & Naveen, 2006; Core & Guay, 1999; Gormley, Matsa, & Milbourn, 2013; Heron & Lie, 2013; Jensen & Meckling, 1976; Knopf, Nam & Thornton, 2002; Mao & Zhang, 2018; Rajgopal & Shevlin, 2002). On the contrary, several studies argue that managerial investment decisions depend not only on how a manager's wealth changes with firm risk but also how termination risk influences a manager’s risk-aversion. Consequently, a manager’s risk-aversion associated with the extra

termination risk can offset the manager’s wealth increase associated with the extra risk-taking. In fact, a line of studies indicates that the increasingly widespread use of equity-based executive

compensation and the termination risk of chief executive officers for poor corporate performance exacerbate risk-related agency conflicts (Balsmeier et al., 2017; Chakraborty et al., 2007; Frydman & Jenter, 2010; Manso, 2011).

The way that unions influence CEO compensation has also received scholarly attention. It is not fully in agreement on the effect of unions on compensation. Several studies find a ratcheting effect of unions on CEO pay, thereby increase executive compensation in unionized firms. However, most extant literature finds that, to enforce compensation equality in the firm and less excessive CEO compensation, unions are effective in decreasing CEO compensation (Banning & Chiles, 2007; Ertimur et al., 2010; Gomez & Tzioumis, 2006; Huang et al., 2017). What’s more, boards of directors often use such compensation reductions as a signal to unions that can improve the board’s

negotiations with labor unions. Finally, some studies find that especially equity pay is reduced by boards as a signal to unions, presumably because of its discretionary nature.

Concerning the subject of this study, exploratory innovation in unionized firms, there are no studies studying the direct effect of unions on exploratory innovation specifically. However, a large amount of literature has focused on the effect of unions on R&D expenses and patent output. These results indicate rather strong negative relations between unionization and innovation in general.

In sum, the literature has studied two important relationships. On the one hand, the literature has described the relationship between executive compensation and executive risk-taking. On the other hand, the literature has studied the relationship between unionization and CEO compensation. A third important strand of literature to this study has also described the direct effect of unions on innovation in general. However, no studies have focused on the direct effect of unions on risk-taking and exploratory innovation. To fill this gap in the literature, this thesis studies the direct relationship between unions and exploratory innovation.

3. Theoretical framework 3.1 The effect of unions on firms’ exploratory innovation

Exploratory innovation is associated with significant risks. Without the appropriate risk-taking incentives, managers are likely to underinvest in risky innovation. Equity pay is an important tool for the compensation committee and the board of directors to align the shareholders’ and executives risk-taking preferences.

The literature also argues that the interaction between firms’ corporate governance

characteristics and executives’ risk preferences influence the attractiveness of exploratory innovation. One of such corporate governance characteristics is the board of directors’ likeliness of firing the executive. This likeliness determines the termination risk associated with risky activities of the executive. Core and Guay (1999) and Pratt (1964) argue that a trade-off exists between the wealth effect, which is affected by the risk-taking incentives, and the risk-aversion effect, which is affected by the termination risk. Termination risk will increase the negative impact of risk-aversion on executives’ utility and can partly offset the wealth effect from incentives provided by equity pay. What’s more, much literature argues that equity compensation can exacerbate the agency problems when it is combined with high termination risk, and that thus managers will pursue less exploratory innovation.

For unions, an additional corporate governance factor will have more influence than termination risk on an executive’s utility function. For unions, the most important corporate

governance characteristic arguably is the board of directors’ likeliness of reducing CEO compensation as a signal to labor unions. In this study, the term compensation risk is introduced which embodies the executive’s risk of receiving compensation reductions. The compensation risk consists of two parts. First, the compensation risk embodies the risk of the chief executive officer of receiving

compensation reductions after weak performance. This is similar to performance-induced termination risk from the study by Jenter and Lewellen (2014). This part of compensation risk thus embodies the likeliness of receiving performance-induced compensation cuts. Secondly, compensation risk also embodies the likeliness of receiving compensation reductions that have the purpose of easing imminent labor negotiations. This compensation risk is therefore so important for managers in

unionized firm because boards often use compensation cuts as signals to unions to signal future negative outlooks to the labor unions or simply to engender the labor unions’ willingness of only moderate salary increases. Bova (2013) and Huang et al. (2017) found that boards indeed use

compensation reductions ahead of looming labor negotiations. An important component of executive compensation that is used as signal to unions are equity pay reductions.

Expectedly, this type of signaling is more or only necessary in case of weak results or misunderstood decision-taking. I argue that managers deem the possibility of the outcomes to be weak, or misunderstood to be higher for exploratory innovation. There is simply more risk associated with exploratory innovation as the returns are highly unpredictable and often negative (Geiger & Makri, 2006; March, 1991). In addition, exploratory innovation is often misunderstood by the board of directors and unions because of their lack of precise technical knowledge that executives do exhibit.

Compensation risk will increase the offsetting effect of risk-aversion over the wealth effect. This will decrease the optimal risk-taking levels in executives’ innovation decisions. What’s more, to decrease the necessity for boards of using CEO compensation reductions as signals, a CEO will reduce its risk-taking. Consequently, executives will reduce exploratory innovation if compensation risk is increased. The compensation risk from signaling to unions is only present in unionized firms. Termination risk will remain present similarly in unionized and non-unionized firms. Therefore, ceteris paribus, it is likely that managers in unionized firms pursue less exploratory innovation than managers in non-unionized firms, because the possible CEO compensation reductions reduce the manager’s incentive to take on excessive risk.

This conjecture is not necessarily in contradiction with the evidence that equity pay stimulates risk-taking. It simply means that some factors are likely to partly offset the stimulating effects,

corroborating with previous studies (Balsmeier et al., 2017; Chakraborty et al., 2007; Frydman & Jenter, 2010; Manso, 2011). Moreover, the equity pay could still be associated with riskier exploitative innovation. Exploratory innovation is however more easily misunderstood than

exploitative innovation by the board of directors and unions because of knowledge gaps with the chief executive officer. To avoid innovative actions to be misunderstood, CEO’s will focus on quantifiable

results, such as a greater number of patents. They will foster an increase in patent count to satisfy demands for innovation and performance (Balsmeier et al., 2017)

In sum, equity pay will generally be effective in stimulating managers to pursue innovation. This stimulating mechanism will be however partially offset by the widespread use of CEO payment reductions as a signal to unions. The increased compensation risk associated with signaling decreases managers’ utility of exploratory innovation through higher risk-aversion levels. Therefore, chief executive officers in unionized firms exhibit more incentives for a reduction of risk-taking than managers in non-unionized firms in which there are no signaling needs for the board of directors. Consequently, because of higher incentives for risk reduction for managers in unionized firms than in non-unionized firms it is likely that managers in unionized firms will pursue less exploratory

innovation. These managers will resort to more quantifiable innovative results, such as a greater number of patents or innovation in technologies types that the board of directors and unions will sooner approve of. Consequently, I formally hypothesize:

Hypothesis 1: There is a negative relationship between unionization and exploratory innovation

3.2 The moderating effect of the share of equity pay in CEO compensation

The previous section hypothesized that a chief executive officer in a unionized firm would pursue less exploratory innovation to reduce the risk of compensation reductions than a CEO in a firm with no unions. This compensation risk could offset the incentivizing effect of equity pay, for which so much of the literature has argued it to be an incentive for risk-taking. Heterogeneous combinations of corporate governance, executive preferences and compensation designs influence the relationship. More specifically, it seems likely that certain compensation compositions are more sensitive to compensation risk than others.

As found by Huang et al. (2017), unions especially use equity pay reductions to send signals to labor unions. Presumably, this is because of the discretionary nature of equity pay. Board of directors can easily adjust how many options it grants and when it grants them. Therefore, executives

who receive relatively large amounts of equity pay are influenced the most by boards who make use of compensation reductions as signal to unions. Managers who receive relatively small amounts of equity pay are less influenced by boards who make use of compensation reductions as a signal. Consequently, it seems likely, when comparing unionized firms, that executives who receive larger amounts of equity pay would be more prepared to reduce the compensation risk. Consequently, managers who receive larger amounts of equity pay are more likely to reduce the risky type of innovation, i.e. exploratory innovation, than managers who receive smaller amounts of equity pay.

In sum, it is likely that the effects of potential compensation reductions as a signal are different for varying compositions of CEO compensation. Because Huang et al. (2017) found that boards especially use equity pay reductions for signaling, that executives with relatively large equity pay arguably have more incentives to reduce risk-taking. Consequently, when comparing unionized firms, chief executive officers with higher equity compensation decrease exploratory innovation more than chief executive officers with lower equity compensation. Therefore, I hypothesize as follows:

Hypothesis 2: The relationship between unionization and exploratory innovation is positively moderated by the ratio of equity pay in total CEO compensation

3.3 The moderating effect of board independence

The compensation committee, which is part of the boards of directors, formally sets CEO pay (Huang et al., 2017). Executives provide input on their pay. However, executives and board of directors have divergent goals. Although both parties arguably like to keep union relations healthy, executives are more likely to resist a pay cut to reach this goal. Therefore, it seems likely that boards with more power over executives are better able to curtail CEO compensation as a signal to unions. In other words, the board’s ability to use CEO compensation reductions as a tool to keep union relations healthy is influenced by its power over the executives. Reduced power of the board of directors over executives diminishes the opportunities for sending executive compensation cuts as a signal to unions. An important indicator of a board’s strength over executives is board independence. Huang et al

(2017) indeed find evidence that firms with strong and independent boards are more likely to curtail total CEO compensation to deal with powerful unions.

As stated previously, a CEO in a unionized firm would pursue less exploratory innovation to reduce the risk of compensation reductions than a CEO in a non-unionized firm, thereby possible offsetting the risk-taking incentivizing effect of equity compensation. Heterogeneous combinations of corporate governance, managerial preferences and compensation designs influence the net

incentivizing relationship between compensation and risk-taking. Board independence is one of such corporate governance characteristics that will arguably influence this relationship.

Compensation risk will be higher in unionized firms with more independent boards than in unionized firms with less independent board, using the evidence of Huang et al. (2017).

Compensation risk will be lower in firms with less independent boards because these boards simply are less equipped to determine the chief executive officer’s compensation design. Thus, if indeed independent boards are more likely to curtail total CEO compensation to deal with powerful unions, managers in firms with independent boards will have more incentive to reduce risk-taking. Therefore, managers in unionized firms with more independent boards are likely to pursue less exploratory innovation, which bears more compensation risk.

In addition, there often exists a significant knowledge gap between executives and the board of directors. Because board members are less familiar with the opportunities of exploratory innovation than with exploitative innovation, these board members might more easily disapprove of exploratory innovation than exploitative innovation. The higher likeliness of disapproval could lead to higher termination and compensation risk for exploratory innovation than for exploitative innovation. This will lead to decreased incentives for exploratory innovation for CEO’s of firms in which the board has more power of the executive, i.e. is more independent.

In sum, the board of directors and the executives have divergent goals concerning executive compensation. Executives are likely to resist pay cuts, which the compensation committee of the board of directors might use for sending signals to unions. Therefore, higher strength of executives in resisting pay cuts would increase their ability to ensure stable compensation and no (significant) pay cuts. However, a higher ability of the board to reduce CEO pay would decrease the CEO’s willingness

to pursue exploratory innovation. Because board independence determines a board’s power relative to the executives, this board independence influences the incentives of managers to reduce risk-taking. A higher degree of board independence would decrease the CEO’s willingness to pursue exploratory innovation for managers in unionized firms. In addition, exploratory innovation is in general less attractive than exploitative innovation if the new technologies from exploratory innovation are more likely to be misunderstood by the board. The negative impact of board independence on managers risk-taking preferences will be stronger the more independent the board is. Summing up, managers in unionized firms are more likely to take innovation decisions which are less likely to lead to

termination and compensation risk, which are preferred by the board and the least likely to be misunderstood. Hence, I hypothesize:

Hypothesis 3: The relationship between unionization and exploratory innovation is positively moderated by board independence

The previously discussed hypotheses result in the conceptual framework as illustrated in Figure 1.

Figure 1

4. Data and variables 4.1 Sample selection and data collection

This study uses a panel data research design to examine the direct effect of a firm’s union bargaining power on the firm’s exploratory innovation. Industry unionization levels are used as a proxy for a firm’s union density as well as a firm-level dummy indicating high union density. Two moderators are included in the regression models. These moderator variables are the firm’s degree of board independence and the share of equity pay in total CEO compensation.

I will conduct a multiple regression analysis at the firm-level, based on a panel dataset from multiple databases. This dataset is determined by the joint availability of data on industry level unionization rates from the Union Membership and Coverage Database; firm-level indicators of high union density from the MSCI database; patent data from the National Bureau of Economic Research (NBER), the United States Patent and Trademark Office (USPTO) and Kogan et al. 2017; board of directors data from ISS; basic firm characteristics data and CEO compensation data from Compustat; and state-level right-to-work legislation data from the National Conference of State Legislatures (NCSL).

This study uses a similar time-period as recent studies who employ similar variables and similar statistical approaches to effectively compare the outcomes of this study with the outcomes of these studies. For example, the study of Balsmeier et al. (2017) on the relationship between board independence and exploratory innovation uses a time-period of 1996 through 2006. The study of Huang et al. (2017) on the effect of unions on CEO compensation uses a time-period of 1993 through 2011. The use of the patenting data from NBER for studying firm innovation is also in line with the study by Bradley et al. (2017) that has, like this study, investigated the relationship between unions and innovation.

I require the sample of firms to have available information on key variables used in the following statistical analyses such as industry unionization rates, firm level unionization indicators, CEO compensation, corporate governance indicators and financial information. These data

Next, the data collection and construction of this dataset is described in more detail. First, the collection of disaggregate patenting data for constructing yearly firm-level innovation indicators is described. In addition, the decisions made regarding cleaning and transformation of this raw data are described in the next section

A.

Several studies have used patenting data on North American firms for studies on firm-level innovation (Balsmeier et al., 2017; Bradley et al., 2018; Gu, Mao & Tian, 2018; Jaffe, 1989; Mao and Zhang, 2018). In line with these studies, I obtain patenting and citation data from the latest version of the NBER Patent Citation database for 1963–2006. The NBER Patent Citation Dataset was initially created by Hall, Jaffe and Trajtenberg (2001). This dataset is based on patenting data on U.S. firms provided by the United States Patent and Trademark Office (USPTO). The dataset contains patents’ application and grant dates, technology classifications and citing and citation information. For almost all U.S. firms that received at least one patent between 1975 and 2006, the NBER provides a unique time-invariant assignee code as well as a CUSIP match. Using the CUSIP match, the reach of the NBER patent database could be extended to 2010 with raw data from the USPTO and with data from Kogan et al. (2017). Next, the decisions made regarding cleaning and transformation of this raw data are described.

The first decision made regarding the raw data is the type of classification system that is used for classifying patents. This study categorizes patents according to the classification system that was created by Hall et al. (2001). The original USPTO database uses 400 USPTO technological classes to assign patents to. Hall et al. (2001) developed a new classification system that reduced the total number of technological classes from 400 to 36. The second decision concerns the date that is assigned to the patents. This study will assign the patent to the year in which the firm applied the patent to the USPTO. This is in line with recent scholarly literature (He & Tian, 2013). Thirdly, only utility patents are considered for the firms. Fourthly, only new patents are included, meaning that reassigned patents are not included in the analyses. The resulting set of data is aggregated to the firm level which results in variables on firms’ total patenting portfolios, yearly additions to patent

portfolios and the number of citations made and to which firm these citations were made. Firms’ total patenting portfolios are again disaggregated into the 36 technological categories, leading to variables on firms’ total patenting portfolios per technological category and the number of additions to these category portfolios. The final decision made regarding the raw data is the specification of a patent’s life span to be 20 years. From 1861 until 1994, the USPTO determined the term of the patent to be 17 years, which was however often extended by 3 years. From 1994 through 2012 the USPTO

determined the term of the patent to be 20 years (USPTO, 2017). Therefore, this study drops firms’ patents from the aggregate patent portfolio and the individual patent portfolios for the different technological subcategories after 20 years. To merge this dataset with data from other databases, the 8-digit CUSIPs accompanying the NBER data are transformed into 6-digit CUSIP numbers. The next section will describe the collection of industry level unionization rates and the decisions made regarding the cleaning and transformation of this raw data.

B.

Previous empirical evidence supports the use of industry unionization rates as a proxy for the unionization rates of firms within an industry and for union bargaining power (Bronars and Deere, 1994; Chen, Kacperczyk & Ortiz-Molina, 2011; Connolly et al., 1986; Huang et al., 2017; Karier, 1985; Klasa et al., 2009; Rosen, 1969). Following this literature, this study uses industry unionization rates as a proxy for individual firms’ union density and union bargaining power. The industry

unionization rates are collected from the Union Membership and Coverage Database of CPS. The Union Membership and Coverage Database annually reports the fraction of total workers in a 3-digit Census Industry Classification (CIC) industry who are represented by unions in collective bargaining agreements. For the time-period of 1992 through 2009, which is used in this research, two different versions of CIC codes are used by the Union Membership and Coverage Database of CPS. Therefore, for panel data from 1992 through 2002, the CIC code classification system from 1990 is used to match industries with the industry identifier codes from the initial dataset with patenting data. The initial dataset identifies industries with SIC codes in 1987 format. For panel data from 2003 through 2009, the CIC code classification system from 2000 is used to match industries with the industry

identifier codes, which are 1987 format SIC codes, from the initial dataset.

An important decision made regarding this raw data is the inclusion of all industries. Many studies exclude one or several industries from their analyses. This study will however include all industries, which is in line with the study by Balsmeier et al. (2017) on the effect of board

independence on exploratory innovation. The next section will describe the collection of firm-level indicators of high union density and the decisions made regarding the cleaning and transformation of this raw data.

C.

In addition to industry unionization rates, firm-level indicators of high union density are collected. These indicators of high union density are provided by the MSCI database (Bhandari & Javakhadze, 2015) and will serve as a proxy for a union’s bargaining power. The MSCI uses yearly indicators of high union density inside firms. The time-period of being indicated as having high union density for a certain year is extended to two years before and two years after. Several similar studies have extended firm-level unionization rates to two years before and two years after a rate was measured (Huang et al., 2017; Marciukaityte, 2015).

Even though this measure is expected to provide a strong proxy for a firm’s bargaining power, matching firms and year observations are scarce. This results in a small number of firm-year observations (a total of 111 observations) in which firms are indicated as having high union density. All other firm-year observations are indicated as firm-years in which the firm had no high union density. These indicator outcomes are merged in the existing dataset for the entire panel of years. The next section will describe the collection of state-level union legislation, more specifically right-to-work legislation. This section will also address the decisions that are made regarding the cleaning and transformation of this raw data.

D.

labor union, thus weakening union power (Huang et al., 2017). The adoption of right-to-work

legislation in firms’ state of incorporation is by many scholars considered to be an exogenous shock to unions’ bargaining power (Matsa, 2010). Scholarly literature finds evidence that indicates that right-to-work legislation has a significant impact on union organizing activity. Right-right-to-work legislation leads to less union organizing and therefore significantly reduces union membership in firms (Ellwood & Fine, 1987; Matsa, 2010). Lower average union membership will reduce union

bargaining power. In addition to reducing the threat of new union organizing activity, right-to-work laws arguably directly affect collective bargaining at firms with existing unions. Several studies argue that the passing of right-to-work legislation makes union membership less attractive to employees (Ellwood & Fine 1987; Matsa, 2010). Fewer people joining unions or maintaining membership will affect the strike threats that unions can make during labor negotiations. Strike threats from unions are diminished by reduced financial resources and by less participation. The reduced strike threats during labor negotiations will decrease unions’ bargaining powers.

Data on right-to-work legislation for individual North American states is collected from the National Conference of State Legislatures (NCSL). However, there is a significant lag after which right-to-work laws become fully effective. Collective bargaining agreements entered before the adoption of right-to-work legislation remain effective until they expire (Marciukaityte, 2015). Thus, for a few years, the law is not fully effective (Marciukaityte, 2015). Consequently, for indicating the effect of states passing right-to-work legislation on firms, I exclude the first three postadoption years. Thus, firms in states that have passed right-to-work law legislation are indicated three years after the passing of right-to-work legislation with a dummy as treated firms. All firms in the dataset are assigned to their state of incorporation. Right-to-work law legislation will therefore be matched to the state of incorporation of the firms included in the dataset. The next section will describe the collection of firm-level corporate governance data and the decisions made regarding cleaning and transformation of this raw data.

E.

States from 1996 to 2006. This dataset is currently known as ISS Legacy and was formerly known as the IRRC Takeover Defense database. This dataset from 1996 through 2006 was used by the recent study of Balsmeier et al. (2017). This study complements the data from the ISS Legacy database with additional board data, also from ISS, for the period of 2006 through 2009. Additional corporate governance data on executive compensation is collected from Compustat’s Execucomp. Due to data and time restrictions no individual board independence ratios are collected from 1992 through 1995. To mitigate these restrictions and enable board independence ratios to be included in regression models for the full time-period used in this study, firms’ independence ratios from 1996 are also used in the previous four years. The next section will describe the collection of additional firm and

financial data and the decisions made regarding cleaning and transformation of this raw data

F.

Compustat provides firm and financial data on almost all the firms covered in the existing dataset for the time-period of 1992 through 2009. This data is used to construct control variables in the statistical models. The following control variables constructed from Compustat data that will be mentioned later in this study are: research and development intensity; age of the company; Tobin’s Q; capital expenditures; size; leverage; above average tenure of CEO; an indicator when the CEO is chairman of the board of directors; and an indicator when the CEO is on the compensation committee. All variables and their operationalization will be described in more detail in the following section.

4.2 Variables

4.2.1 Dependent variables

To measure exploratory innovation as dependent variable, the same proxies for exploratory innovation are used as in Balsmeier et al. (2017). These five proxies as dependent variables are described next.

Technological proximity: Exploratory innovation is the radical design of new products, services or distribution channels to enter unfamiliar or distant market segments. An efficient proxy for exploratory innovation would therefore be a measure that takes into account how the additions to a

firm’s patent portfolio in year t held by the same firm are different from the firm’s patent portfolio help up to year t−1. In other words, it is a continuous measure of whether firms deviate from previous compositions of the patent portfolio. The same formula of Jaffe (1989), that was also used in the study by Balsmeier et al. (2017), is used to calculate a quantifiable measure for technological proximity:

𝑃"# = 𝑓"&#∗ 𝑓"&#()/ 𝑓+_"&# , &-) ∗ 𝑓+ "&#() ) + , &-)

In this formula, fikt is the fraction of firm i’s patents that belong to patent class k at time t, and fikt−1 is the fraction of firm i’s patent portfolio up to t−1 that belongs to patent class k. Pit ranges between zero and one. The highest possible value indicates that the portfolio distribution of patents in technological classes in a year is exactly the same as in the previous year. A value of zero would indicate no overlap in the portfolio distribution in a certain year compared to the previous year (Jaffe, 1989). Positive coefficients in a regression would thus indicate a narrower innovation trajectory within known areas.

Self-citations: A different efficient proxy for exploratory innovation would be a measure that takes into account to what extent a firm’s patents belong to either a new or familiar technological area. Patents that make citations to a firm’s own patents indicate that the firm’s new patents and innovation are focused on technological areas which it is already familiar with because it has already patents in this technological area. Therefore, the logarithm of the number of times a patent cites other patents owned by the same company is used as a proxy for exploratory innovation (Balsmeier et al., 2017; Faleye, Hoitash & Hoitash, 2011; Sorensen & Stuart, 2000). Fewer self-citations indicate firms’ efforts to explore technological areas that are new to the firm, which can be seen as an exploratory innovation effort.

Backward citations: A third proxy for exploratory innovation to be included as a dependent variable would be a measure that takes into account to what extent a firm’s innovative search is in mature technological areas. Patents in mature technological areas arguably make more citations to previous patents in the corresponding technological area than patents in less mature technological

areas. The reason for this simply is that fewer patents have been developed in a less mature

technological area and fewer citations can thus be made to previous patents in this technological area. Therefore, the logarithm of the number of citations that each patent makes to prior patents is used as a proxy for exploratory innovation (Balsmeier et al., 2004; Lanjouw & Schankerman, 2004). Fewer backward citations indicate firms’ efforts to innovate in less mature technological areas, which can be seen as an exploratory innovation effort.

Patents in unknown classes: A fourth proxy for exploratory innovation to be included as a dependent variable would be a measure that takes into account how many patents are filed by a firm in technology classes in which it had no active or previously filed patents. Patenting in technological classes in which the firm is not yet active indicates that the firm’s new patents and innovation is focused on technological areas that are new to the firm. Therefore, the logarithm of the number of patents that are filed in technology classes previously unknown to the firm is included as a proxy for exploratory innovation. Balsmeier et al. (2017) define unknown patent classes as those in which a given firm has not applied for any patents beforehand. Alternatively, this study defines unknown patent classes as those classes in which a given firm has not applied for any patent in the previous twenty years. Therefore, it is possible that firms file patents in classes in which it has previously filed patents but are assessed in this study as patents in unknown classes since it has been more than twenty years that the firm has filed new patents in this technology class. This is in accordance with the legal term of patents. It also just seems reasonable that twenty years is sufficient for a firm to lose much expertise of a technological class so that filing again in such a class is an exploratory effort. More patents in unknown technological areas would indicate firms’ innovative efforts in technological areas unknown to the firm, which can therefore be seen as a proxy for a firm’s exploratory efforts.

Patents in known classes: A final proxy for exploratory innovation to be included as a dependent variable would be a measure that takes into account how many patents are filed by a firm in technology classes in which it has already filed patents. Patenting in technological classes in which the firm is already active indicates that the firm’s new patents and innovation is focused on

technological areas which it is already familiar with because it has already patents in this

classes previously known to the firm is included as a proxy for exploratory innovation. This study defines known patent classes as those in which a given firm has previously applied for any patent in the previous twenty years. Fewer patents in known technological areas would indicate firm efforts in areas of technological areas new to the firm, which can therefore be seen as a proxy for a firm’s exploratory efforts.

4.2.2 Independent variables

Industry-level unionization rate: This study will use industry unionization rates as a proxy for union density indicators of firms within an industry and for their bargaining power. Industry-level unionization rates are defined as the fraction of total workers in an industry who are represented by unions in a collective bargaining agreement.

Firm-level unionization dummy: MSCI provides an indicator that identifies companies with high union density (Bhandari & Javakhadze, 2017) and will serve in this study as a proxy for a firm’s bargaining power. The firm-level unionization dummy is a dummy variable indicating firms with high union density (1= Yes; 0 = No).

Right-to-work transition dummy: The adoption of right-to-work legislation in firms’ state of incorporation is by many scholars considered as an exogenous shock to unions’ bargaining power (Ellwood & Fine, 1987; Marciukaityte, 2015; Matsa, 2010). The right-to-work transition dummy is a dummy indicating firms located in the state of Texas after 1996, three years after the introduction of right-to-work legislation in 1993.

4.2.3 Moderator variables

Equity pay ratio: Equity pay is defined as the CEO’s total compensation less total cash pay. This definition of equity pay is in line with a previous study of Huang et al. (2017) on the effects of labor unions on CEO compensation. This study will include the importance of equity pay for CEO’s in their compensation as the share of equity compensation in their total compensation. Using a fraction of total CEO compensation will remove comparative differences in total CEO compensation levels. The equity pay ratio is therefore defined as the logarithm of the ratio of equity pay in total

CEO compensation.

Board independence ratio: Board independence will be used as a proxy for the strength of the board of directors over its executives. This study will define the independence ratio as the ratio of independent board members on the board of directors.

4.2.4 Control variables

This study controls for several firm characteristics that could distort the relation found in statistical regressions between unions’ bargaining power a firm’s exploratory innovation. Some control variables are specifically related with exploratory innovation and others with innovative activities in general. These control variables are also used in the study by Balsmeier et al. (2017) that makes use of the same proxies for exploratory innovation in their study.

Research and development intensity: Naturally, firms’ research and development activities are positively related to their innovation activities, be it exploitative or exploratory. Because R&D intensity differs vastly for firms in this study’s sample, research and development intensity is included as a control variable. Moreover, it is especially important to control for R&D intensity as it was found in the literature review that most studies find evidence that R&D is less attractive for unionized firms. Neglecting to control for such differences in the attractiveness of R&D for unionized firms and non-unionized firms will lead to an overestimation of the negative effects of unions on exploratory innovation. Therefore, research and development intensity is included as a control variable and is defined as the research and development expenditures over total assets.

Age of company: Older firms may maintain innovation in technological areas in which it has been active since inception. Naturally, the technological areas in which these older firms have been active since inception have had more time to mature and become crowded. Therefore, especially for the number of backward citations, it is necessary to control for firm age. Neglecting to control for firm age in this case will lead to an overestimation of the negative effects of unions on exploratory innovation. Therefore, the logarithm of age (the number of years since the company’s inception) is included as a control variable.

differences in growth opportunities. Growth opportunities, measured by Tobin’s Q determine the possible innovative opportunities, be it exploitative or exploratory, for the different firms.

Capital expenditures: Financial constraints are known to influence firms’ innovative activities and output, be it exploitative or exploratory. Therefore, capital expenditures, which are divided by total assets, are included to account for financial constraints that reduce total innovative output.

Size: A firm’s size is naturally positively related with firms’ innovation activities. Therefore, the logarithm of assets is included to reduce the influence of size on innovative opportunities, in general.

Leverage: Financial constraints are known to influence firms’ innovative activities and output in general. Therefore, leverage, which is measured as long-term debt divided by total assets.

Year dummies: Firms’ innovative opportunities are likely to be influenced by yearly economic conditions and systematic changes in firms’ patenting activity over time. Therefore, this study includes year dummies to correct for year effects and time trends.

Industry dummies: Studies from industrial economics have demonstrated that a firm’s innovative opportunities can be influenced by the sector it is active in (Fernández-Olmos, Gargallo-Castel & Giner-Bagües, 2016). To correct for industry effects, this study includes SIC industry dummies. Table 2 in the appendix summarizes the variables used and their operationalization.

4.3 Summary statistics

Table 1 presents summary statistics for the full sample with unionization rates, statistics for sample 2 with firm-year observations of firms incorporated in states that passed right-to-work legislation and statistics for sample 3 of firms with firm-level indicators of high union density. The

Table 1

Summary statistics

This table presents summary statistics for the full sample with industry unionization rates, the subsample of treated firms with industry unionization rates and the subsample of firms with firm-level indicators of high union density. Unionization rates in the full sample and in sample 2 are measured at the industry level and calculated as the percentage of total workers in an industry who are represented by unions in collective bargaining agreements. The full sample consists of 11,152 firm-years during the period 1992–2009. Sample 2 consists of 740 firm-year observations for firms located in the state of Texas in 1996, three years after the introduction of right-to-work legislation in 1993. Sample 3 consists of 111 firm-year observations for firms that are indicated as having high union density. All continuous variables except for unionization rates are winsorized at the 1st and 99th percentiles.

1 2 3

Full sample with industry unionization

rate Subsample of treated firms with industry unionization rate Subsample with firm unionization indicator

(N = 11,152) (N = 740) (N = 111)

Mean Median Std. Dev. Mean Median Std. Dev. Mean Median Std. Dev. Technological proximity 0.986 0.999 0.043 0.982 0.998 0.048 0.994 0.999 0.028 Self-citations 58.670 1.000 401.497 65.150 0.000 349.381 227.333 0.000 713.547 Backward citations 364.879 28.000 1,825.084 396.697 21.000 1,649.318 475.937 11.000 1,128.800 Patents in known classes 40.468 5.000 170.760 42.522 4.000 125.627 74.162 15.000 121.986 Patents in unknown classes 0.457 0.000 1.161 0.492 0.000 1.028 0.108 0.000 0.340 log (1+ Self-citations) 1.448 0.693 1.846 1.308 0.000 1.949 2.033 0.000 2.590 log (1+ Backward citations) 3.300 3.367 2.388 3.204 3.091 2.381 2.896 2.485 3.033 log (1+Patents in know classes) 2.152 1.792 1.489 2.066 1.498 1.650 2.891 2.773 1.781 log (1+ Patents in unknown classes) 0.359 0.000 0.588 0.331 0.000 0.514 0.229 0.000 0.454 Yearly added patents 40.925 6.000 170.793 43.014 4.000 125.606 74.270 16.000 121.933 Total Pay (in $ millions) 5.442 2.621 14.368 6.136 2.943 8.824 8.762 5.271 8.726 Cash pay (bonus + salary, in $

millions) 1.292 0.934 1.396 1.451 0.964 1.384 1.916 1.381 1.331

Equity pay (Total pay - cash pay, in $

millions) 3.479 1.131 13.703 3.942 1.231 7.483 5.394 2.738 6.823

Equity pay ratio 0.445 0.467 0.291 0.439 0.481 0.306 0.513 0.507 0.234 Board independence ratio 0.676 0.714 0.166 0.694 0.727 0.163 0.709 0.706 0.134 Industry-level unionization rate 11.976 8.600 11.062 12.659 9.700 11.337 22.583 20.550 16.285 Firm-level unionization dummy 0.009 0.000 0.095 0.005 0.000 0.073 1.000 1.000 0.000 Research and development over assets 0.069 0.035 0.096 0.036 0.006 0.064 0.016 0.009 0.016

Tobin's Q 2.524 1.801 3.084 2.146 1.799 1.298 1.843 1.422 1.017 log (Tobin's Q) 0.703 0.588 0.581 0.635 0.587 0.479 0.504 0.352 0.433 Capital Expenditures 0.057 0.045 0.046 0.075 0.059 0.060 0.056 0.048 0.033 Size (log(Assets)) 6.846 6.720 1.989 7.225 6.948 1.992 9.271 9.469 1.491 Leverage 0.159 0.125 0.168 0.145 0.130 0.134 0.270 0.266 0.125 Age of company 23.814 19.000 17.192 23.132 19.000 16.601 41.189 47.000 15.813 log (Age of company)

variables included are all variables described in the previous section or additional variables that are used to construct these variables. The full sample consists of 11,152 firm-year observations. Sample 2 consists of 740 firm-year observations for firms located in the state of Texas after 1996, three years after the introduction of right-to-work legislation in 1993. Sample 3 consists of 111 firm-year

observations for firms that are indicated as having high union density. All continuous variables except for unionization rates are winsorized at the 1st and 99th percentiles. Next, the summary statistics for the full sample are described.

The average technological proximity is 0.986, which is much higher than the 0.52 found for the same operationalization in Balsmeier et al. (2017) but similar to the sample average found by Jaffe (1989). The difference found between this and Jaffe’s (1989) study with the study of Balsmeier et al. (2017) stems from the different classification systems that are used. This study and Jaffe’s (1989) study make use of the classification system from Hall et al. (2001) of 36 classes whereas Balsmeier et al. (2017) made use of the original class system from the USPTO with over 400 different types of classes. The average (median) of the yearly self-citations made is 58.670 (1.000). The average (median) of the yearly backward citations made is 364.879 (28.000). These summary statistics for self-citations and backward citations indicate large differences between the median firm and the firm making the most amount of backward citations and self-citations. On average 0.457 patents are filed yearly in unknown technology classes, while 40.468 are filed in known classes. The summary statistics of the five dependent variables which will be used in the following statistical analyses are shown in the table. The average number of yearly added patents is 40.925. Average (median) total cash CEO compensation is $1.292 ($0.934) million whereas the average (median) amount of equity pay is $3.479 ($1.131) million resulting in average (median) total CEO compensation of $5.442 ($2.261) million. Similar to these statistics, Huang et al. (2017) find an average (median) amount of total cash CEO compensation of $1.380 ($0.970) million and average total CEO compensation of $5.360 ($2.850) million for the period of 1993 through 2011. As the summary statistics show, equity pay makes up, on average, a large part of total CEO compensation. Also, the large standard deviations for total pay and equity pay indicate large differences. The average CEO equity pay ratio is 0.445. In other words, on average, equity makes up 44.50% percent of total CEO compensation. The mean of