Is the pay-performance relationship industry-specific? Evidence from the U.S. pharmaceutical sector.

(1)

University of Amsterdam Faculty of Economics and Business

Bachelor Business Administration with Concentration: Finance

Is the pay-performance relationship industry-specific?

Evidence from the U.S. pharmaceutical sector.

Author: Leon Mofardin 11623586 Thesis supervisor: Mr. Adam Fehér Final Submission: 30.06.2020

(2)

Abstract

In the past thirty years the importance of compensation incentives has been debated at length. Theory suggests that aligning CEO pay with shareholder wealth will stimulate firm

performance. However, there is little agreement amongst scholars on the efficacy of such practices. From an examination of the extant literature, it surfaced that industry

characteristics were often omitted from their analyses. Using panel data on 27 U.S. companies from 2012 to 2017, the determinants of executive compensation in the pharmaceutical industry were compared to a wider sample of corporations from various backgrounds. It emerged that the traditional determinants of pay such as volatility and firm size do not significantly influence compensation for pharmaceutical CEOs. Nonetheless, no significant difference was found for the pay-performance sensitivity. These results imply that the pay-performance relationship might be improved by adjusting it to industry-specific factors. However, further research needs to corroborate this intuition.

Statement of Originality

This document is written by student Leon Mofardin who declares to take full responsibility for the contents of this document.

I declare that the text and the work presented in this document are original and that no sources other than those mentioned in the text and its references have been used in creating it.

The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

(3)

1. Introduction

Over the past two decades, debates surrounding issues of executive opportunism have become increasingly prominent. Headlines such as the 2001 Enron scandal and the 2002 Worldcom debacle have left the public incredulous. Despite the economic cost, a great deal has been learnt about the issues which arise from providing CEOs with excessive freedom in their decision-making. Consequently, today even the most unsophisticated of investors

should be able to recognize the signs of a strongly compromised company. Nonetheless, more recent events such as the 2016 Bayer-Monsanto acquisition show that some CEOs continue to neglect the interests of their shareholders in their actions. Disincentivising this kind of

conduct is crucial for a well-functioning company (Jensen & Meckling, 1976). A widely accepted method to motivate CEOs to operate with the firm’s best interest at heart, is strengthening the link between CEO compensation and shareholder wealth (Hill & Jones, 1992). Scholars around the world have devoted much effort to studying the efficacy of various compensation schemes. To measure the strength of the pay-performance relationship, academics have controlled for a wide variety of factors, for example size, risk, executive ownership and countless other metrics. Nonetheless, the empirical evidence for a direct relationship between pay and performance is strongly conflicting. Consistently, it seems that many studies have forgone an analysis of the effect that the industry environment can have on the pay-performance sensitivity. Thus, this paper attempts to measure the link between

compensation and performance by isolating a single industry and comparing it to a wider sample of firms. Most notably, the aim is to investigate how the industry affects the pay-performance relationship. Therefore, the central question of this paper is: To what extent does an analysis of the pay-performance relationship in the pharmaceutical industry differ from cross industry analyses? To answer this question, a sample of twenty-seven U.S. firms has been collected from the NASDAQ. Sixteen of these firms have been randomly selected from the pharmaceutical sector and provide the central dataset for this research. The remaining eleven firms were selected with the only restriction being that they do not belong to the pharmaceutical industry. These serve as the comparison group in this study. After running various regressions, it surfaced that the pay-performance relationship is just as weak in the pharmaceutical sector as when measured across industries. However, this study found that the traditionally recognized determinants of pay did not appear to have a significant effect on pay in the pharmaceutical industry. This suggests that different industries might require different compensation schemes to incentivise adequate executive conduct. Moreover, a small-scale

(7)

attempt at measuring opportunistic tendencies amongst pharmaceutical CEOs was carried out by means of a two stage least squares regressions. This analysis is contingent on the accuracy of the efficient market hypothesis, which will be contextualised below. Notwithstanding, the results suggest that incentives are not strong enough to prevent CEOs from engaging in value-destroying mergers and acquisitions. Thus, further research into incentive management is required.

The first part of this paper lies out the theory which frames this study. The second section illustrates the debate on pay-performance techniques. Here, the reader is provided with an introduction of the underlying logic. The subsequent methodology segment clearly defines the variables and the econometric procedure applied in the technical part of this study. Following on from this, the results section reports the main findings, whilst the discussion offers an empirical deliberation on the status quo in the analysed sample. From these findings, an attempt will be made to draw inferences to inform further research. Lastly, some limitations to the validity of the results will be presented. To conclude, a brief summary of this paper is offered to consolidate the key ideas and findings.

2. Literature Review

Agency Theory

To set a framework for this paper, let us start from the assumption that the modern corporation is characterized by the separation of ownership and control. As Jensen and Meckling (1976) explain at length in their “Theory of the Firm: Managerial behaviour,

agency costs and ownership structure”, shareholders delegate the management of the firm’s

resources to an independent agent, who is expected to act in the shareholders’ best interest. This entails increasing the value of the firm by taking on positive NPV projects. Essentially the foundation of the firm is a separation between management and risk bearing. However, as Jensen and Meckling (1976) posit, this principal agent contract comes at the cost of

information asymmetry. Shareholders are unable to monitor the entirety of the agent’s decision-making process. They will often not be aware of the various investment alternatives a CEO faces, nor do they know whether the decisions made by the CEO will maximise firm

(8)

value (Jensen & Meckling, 1976). In accordance with neo liberalist ideology, financial theory is based on the belief that individuals are opportunistic and self-serving (Pratt & Lavett, 2001). Thus, agents will act in a way that most benefits themselves, potentially to the detriment of their shareholders’ interests. Examples of such costs include expropriation of firm resources, excessive use of corporate perks and neglecting managerial responsibilities (Ang, Cole & Lin, 2000). Therefore, it is imperative that such opportunistic behaviour be prevented with appropriate incentive schemes.

Agency theory is the pillar around which most research on incentives revolves. The theory postulates that aligning the agent’s wealth to that of the principal is conducive to resolving this principal-agent conflict (Hill & Jones, 1992). In the corporate world this is often achieved by linking executive pay to performance. Accordingly, the prospects of increasing one own’s wealth should motivate CEOs to act in line with what’s best for their shareholders (Jensen & Meckling, 1976). However, evidence seems to present mixed results on the efficacy of such methods (Jensen & Murphy, 1990; Hall & Liebman, 1998; Aggarwal & Samwick, 1999; Bertrand & Mullainathan, 2001). To provide a background for the

analysis in this paper, the following paragraphs introduce some of the most influential literature on the topic.

2.1 Empirical Evidence

Stock ownership

In a frequently cited study, Jensen and Murphy (1990) compared the effect of various forms of compensation on performance. Their findings suggest that the relationship between pay and performance is weak, and the strongest incentives are offered by executive stock ownership. This makes sense intuitively, as the worth of the CEO’s share holdings is directly proportional to the value of the firm. While Hall and Liebman (1998) confirm the efficacy of CEO stock ownership and option awards, they find stronger evidence for the

pay-performance relationship than Jensen and Murphy’s study (1990). Their main argument is that during the sample period analysed by Jensen and Murphy, the use of stock options was less prominent than it became following the rise of the 1990’s bull market. Secondly, Jensen

(9)

and Murphy (1990) used only one measure to estimate the sensitivity of CEO wealth to firm value, which Hall and Liebman (1998) argue is too limited. Mainly, Hall and Liebman (1998) draw attention to the sensitivity of performance to stock and stock option pay. However, they conclude that there is not sufficient evidence to corroborate the usefulness of this practice.

To extend the scope of the above analysis, Aggarwal and Samwick (1999) included risk measures to test whether greater risk reduces the responsiveness of pay to performance. From their research, it appears that firms with more consistent levels of stock returns

displayed the highest degree of pay-performance sensitivity. However, Bertrand and Mullainathan (2001) found no evidence that equity incentives had a persistent beneficial effect on firm performance. In fact, their research showed that often CEOs may be rewarded for luck rather than performance. They found that factors such as recurring price fluctuations within an industry, exchange rate changes and year to year effects of industry-wide

conditions are often captured in pay. These risk-loaded factors are essentially out of a CEO’s control. Nonetheless, their study offers proof that CEOs are consistently being compensated for such events. Moreover, Garvey & Milbourn’s (2006) research suggests that not only are managers rewarded for luck but also, through their influence on boards, they can isolate their pay from bad luck via asymmetric benchmarking. This entails purposively choosing different benchmarks to evaluate performance in both good and bad years. More recent studies, such as the one conducted by Malik and Makhdoom in 2016, have drawn attention to the influence of board composition and board practices on firm performance. They find that higher paid CEOs tend to be less engaged with the firm’s interests, and also demonstrate the moderating effect of board independence on this negative relationship. Consistently, the evidence they find points against an excessive focus on pay in directing managerial decision-making. Lastly, some specialists even go as far as suggesting that stock option plans are designed for accounting purposes rather than to incentivise firm performance altogether (Elson, 2003). Thus, there is little agreement amongst scholars as to the efficacy of pay for performance.

Industry effects

Notwithstanding the mixed evidence, the importance of incentive-based compensation in the process of motivating CEOs remains uncontested and it has inspired much research. Accordingly, scholars have continued to investigate the pay-performance relationship in a wide variety of settings. Frydman and Jenter (2010) explain that analysing the

(10)

pay-performance relationship is rendered particularly difficult by the influence of a plethora of confounding observable and unobservable variables. However, in most of the literature on executive compensation, broad samples of publicly traded firms have been used (Sloan, 1993; Hartzell & Starks, 2003; Conyon, 2006). This may cause the estimates to be biased by the inclusion of various sources of random noise. In response, some academics have

narrowed their research to the national level (Zhou, 2000; Elston & Goldberg, 2003; Conyon & He, 2011). However, this implies that measurements will still be tainted by the effect of industry-specific events. Therefore, other studies have limited their scope to individual

industries. Surprisingly, industries such as banking and the financial sector seem to have been preferred for this purpose (Chen, Steiner & Whyte, 2006; Cuñat & Guadalupe, 2009). This is counterintuitive for the study of an already complex phenomenon. The high volatility

inherent to these industries is likely to bias estimations due to the lack of consistent performance measurements. Consistently, in an analysis of the pharmaceutical industry, Roach and Goedde (2003) succeeded to find much stronger evidence. In fact, they showed that there is a strong relationship between firm market value and executive compensation in the pharmaceutical industry. It should be noted that the pharmaceutical industry is

characterized by a set of unique traits. Firstly, the pharmaceutical sector is research intensive and has an abundance of M&A activity (Pavlou & Belsey, 2005). Further, due to the lack of substitutes for medical treatments, there is consistent demand for medications, which

significantly reduces the volatility in this sector (Ganuza, Llobet & Domínguez, 2009). Moreover, competition is protected by patents, which further reduces the risk of sudden changes. Roach and Goedde’s (2003) findings suggest that analysing the performance relationship in specific industry settings could help shed more clarity on the incentive

compensation debate. Therefore, this paper will compare the pay-performance relationship in the pharmaceutical industry to a broader sample of firms from various backgrounds. If a significant difference in the determinants of pay is revealed, there is evidence for the need to tailor incentive schemes to particular industry characteristics.

However, Roach and Goedde (2003) used data for U.S. pharmaceutical companies from 2001. Economically this was a rather unique year for the U.S. as a recession initiated by the dot com bubble was further strengthened by the 9/11 terrorist attack on the World Trade Centre (Schifferes, 2001). This warrants a renewed analysis of the pharmaceutical industry. Thus, the contribution of this study lies in that it will offer a more representative and

(11)

which has been largely devoid of disruptive economic events. By localising the effort, the aim is to isolate the relationship between pay and performance to reveal the extent to which the industry setting affects this relationship. This will be achieved by comparing the results from the pharmaceutical industry to a wider sample of firms from various sectors. As in line with Roach and Goedde’s (2003) findings and the analysis carried out by Jensen and Murphy (1990), this study will use the lagged firm market value (i.e. T-1) to quantify the main

regressor: performance. In accordance with agency theory, this appears to be the most straight-forward way to capture shareholder wealth. However, to avoid incurring the same bias as Roach and Goedde (2003), the data for this paper has been purposively selected from the 2012-2017 period (please note: to obtain the lagged value for firm market value, data from 2011 was added). The chosen time interval should minimize the impact of the 2008 crisis on the global economy, allowing for a more accurate measurement of performance. Additionally, as Frydman and Jenter (2010) outlined, the pay-performance relationship can suffer from strong influences of other factors. Therefore, the inclusion of adequate control variables is crucial to devise a reliable system of measurement. The main control factors for this study are outlined below, a more elaborate list of the utilized variables will be offered in the methodology section.

Firstly, the influence of executive ownership on performance was noted early on, as in Jensen and Murphy’s 1990 study. Additionally, Anderson, Banker, and Ravindran (2000) note that this is a two-way relationship. On the one hand, executive ownership can lead to entrenchment, thus enabling executives to tunnel corporate resources by increasing their own compensation. On the other hand, greater executive compensation might lead to a decrease in pay as it ties an executive to the firm, in turn decreasing his bargaining power. However, this is not a linear relationship. According to Griffith (1999), both CEOs with low and high stakes in a company seem to exert the highest levels of effort, whereas moderate levels of ownership are most closely linked with lower performance. Therefore, to obtain accurate results from a linear regression model it is important to mitigate the effect associated with varying levels of executive ownership through the inclusion of a control variable.

Next, managers in various industries have been found to decrease research spending to

positively affect short-term performance evaluations. A move which is designed to serve their own interest (Bushee, 1998). Accordingly, Cheng (2004), showed that firms are aware of this phenomenon and include pre-emptive measures via stock options in compensation design. Moreover, the business cycle in the pharmaceutical industry is characterized by great upfront

(12)

investments in R&D, long testing phases during which new medications await FDA approval, and eventually profits arise at a much later stage. This causes a lag between the registration of expenses and the recording of profits. Therefore, a control variable for R&D expenditures will be used in this study.

Lastly, Pavlou and Belsey, 2005 illustrate the prominence of acquisitions in the

pharmaceutical industry to gain access to patents which enable firms to produce specific medications. However, M&As are a widely adopted strategy by self-serving executives to entrench themselves (Berger, Ofek & Yermack 1997). This provides executives with more freedom to tunnel corporate resources to their private accounts, increase their compensation and act in other opportunistic ways (Jensen, 1986; Grinstein & Haribar, 2004; Kumar, Kuo & Ramchand 2012). Thus, pharma CEOs might be exploiting the wide acceptance of

acquisitions as a way to disguise their own opportunistic tendencies. This hypothesis will be addressed by means of a two stage least square regression. The underlying intuition will be more precisely outlined in the next section.

2.2 Hypotheses

As previously explained, most research on the pay-performance relationship has overlooked the impact that differences across industries may have on the relationship between pay and performance. Consistent with Roach and Goedde’s (2003) paper, it is assumed that analysing this relationship in a specific industry will add more clarity to the pay-performance debate. Accordingly, the first hypothesis of this study is:

H1: The interaction between pay and performance is stronger for pharmaceutical firms than for other companies.

Executive compensation Firm Performance

(13)

Once the pay-performance relationship has been established, it is appropriate to measure how well it functions in relation to entrenchment theory. In fact, the pharmaceutical industry has strong M&A activity (Pavlou & Belsey, 2005). These acquisitions are widely accepted and justified as a means to obtain access to relevant patents and licenses. Multiple studies show that M&As are the preferred strategy used by CEOs to establish their

dominance in a corporation, with detrimental consequences to firm value (Berger, Ofek & Yermack, 1997; Jensen, 1986; Grinstein and Haribar, 2004; Kumar, Kuo & Ramchand, 2012). According to the strong version of the efficient market hypothesis (Malkiel, 1989), all private and public information on a company is entirely reflected in market prices. Thus, if M&A’s are undergone with opportunistic intentions (i.e. value destroying), they should cause a reduction in firm market value. Therefore, regressing changes in firm market value on M&A expenditures should provide a measure of net value creating (or destroying) M&A activity. Next, based on the evidence that CEOs entrench themselves to freely increase their own pay, it should be possible to detect executive opportunism by regressing the net M&A factor on compensation. If pharmaceutical CEOs initiate mergers with the sole goal of self-dealing, there should be a positive relationship between value destroying M&As and

executive compensation. This will be tested by means of a 2sls regression. In the first stage, the relation between year on year changes in firm market value and yearly M&A spending is analysed. Next, the second stage of this model will regress the newly devised joint effect on CEO compensation. Accordingly, the second hypothesis is:

H2: Pharmaceutical CEOs opportunistically exploit M&As to increase their own compensation.

(14)

3. Methodology

3.1 Sample Design and Data

To investigate the above described hypotheses, a sample of 27 publicly traded U.S. companies has been selected from the NASDAQ. Data on their performance and CEO compensation has been downloaded from the CompuStat and ExecuComp database of the Wharton school of the University of Pennsylvania. To ensure the accuracy and validity of measurements, the downloaded values have been cross references with the companies’ official 10k annual reports and DEF 14a proxy statements. Discrepancies have been adjusted based on the official values in the statements. These statements were retrieved from the U.S. Securities and Exchange Commission archives. As such, all data is in line with the GAAP and, thus, comparison across entities is possible. The sampling period spans from 2012 to 2017, as mentioned above this has been done with the goal of minimising the impact of the 2008 financial crisis on company performance.

The following paragraphs will outline the various regressors used for the analysis.

Short term compensation

Salary and bonuses constitute the short-term fraction of compensation. Due to the fact that salary is fixed by contract, it is considered not to be part of incentive compensation. In fact, Jensen and Murphy (1990) note that even prospects of future increases in base salary have only a minimal effect on total performance. Thus, salary merely functions as a bottom line that provides the executive with a guaranteed and independent level of income. This warrants the exclusion of base salary level from the pay variable.

(15)

on the achievement of a threshold. These thresholds are typically expressed as a percentage of ratios such as return on assets (ROA), return on investment (ROI) or return on equity (ROE). Though they don’t directly narrow the gap between shareholder wealth and executive wealth, long term sustainable profitability lies at the base of growing a successful firm. Thus, bonuses will be included in the measure of pay.

Long term compensation

Stock Awards

Stock awards typically come in restricted form. This implies that the receiver stays with the firm over a specific amount of time (commonly five years) before acquiring the right to dispose of the stock as they wish. On the one hand, this ensures executive retention as one would not want to lose the accumulated stock. On the other hand, it strengthens the link between the performance of the company and the executive’s pay. In this way, executive pay increases when shareholder value is grown. This is expected to motivate executives to

maximise their own value in ways that are also accommodating to the shareholders (Jensen & Murphy, 1990). Consequently, the inclusion of stock awards is crucial to obtaining a

representative measure of the pay-performance sensitivity.

Stock Options

Stock options provide the executive with the right to call for a specified amount of stock up until maturity of the option. Amounts and dates are specified a priori in the option contract. This form of pay provides the executive with a disincentive to waste corporate resources. In fact, a CEO’s future wealth is directly proportional to the change in shareholder wealth (Jensen & Murphy, 1990). Thus, options are awarded to executives to increase their sensitivity to firm performance and, therefore, need be included in the pay variable.

Non-Equity Incentives

Just like bonuses, non-equity incentive plans are also linked to accounting metrics. However, instead of rewarding yearly performance, they are tied to long term performance measures. Thus, their impact may be spread over time. This makes it more difficult to assess

(16)

their influence on the pay-performance relationship. As the selected interval encompasses just a few years, this category of remuneration won’t be included in the regression.

Pension and Other Income

As Fama (1980) explains, pension and other income arise from optimal contracting. Thus, they are effectively part of incentive pay. Other income includes contributions to severance payments, perquisites and discounts on stock purchases. These function as an allowance to the executive which should incentivize performance while limiting rent

extraction. Consequently, it is imperative that this fraction of pay be included in the analysis. Based on the nature of the various categories of pay, the dependent variable, namely executive compensation, will be defined by an aggregate measure. This measure includes bonuses, stock awards, option awards, pension and other income. As mentioned above, salary and non-equity incentive plans are not directly linked to shareholder wealth growth and, thus, will not be included. The newly generated variables will be referred to as sensitive pay for the remaining part of this paper. It must be noted that incentive compensation is typically paid out after a performance appraisal has been carried out. This can only be executed ex-post and it is commonly done in the following fiscal year. Therefore, sensitive pay in one year will be matched with the performance measure for the preceding year.

Performance Measures

From the above described studies, it arises that a great deal of firm performance measures have been used in research. Amongst these, the most common were net income, return on assets (ROA), market value of equity, earnings per share (EPS) and market value added (MVA) (Mehran, 1995; Bhagat & Bolton 2008; Jensen & Murphy, 1990; Bertrand and Mullainathan, 2001; Aggarwal and Samwick, 1999; Hall and Liebman, 1998,

Sheikholeslami, 2001). Market value of equity measures the aggregate market price of common shares outstanding, and as such is a direct measure of shareholder wealth. Therefore, based on Roach and Goedde’s (2003) study, market value of equity is the preferred measure for this research. However, in line with Jensen and Murphy’s (1990) method, the lagged value of a firm’s market price will be used to better describe the link with compensation.

(17)

Control Variables

This section sets out to briefly explain the control variables which will be used within the study. It must be noted that no values for Acquisition and R&D expenditures have been collected for the mixed group. This is because this sample merely serves as a comparison group, and thus it should not include variables for industry-specific traits. In fact, the aim of this study is to isolate the pharmaceutical industry and its characteristics in order to measure whether the inclusion of industry-specific traits can lead to a more conclusive analysis of the pay-performance relationship.

• As outlined above the pharmaceutical industry is strongly characterized by its investments in research and development. However, the relationship between R&D and pay is not straight forward. A control variable taking on the value of yearly R&D expenditures is put in place to isolate the pay-performance

relationship from this effect.

• As it has already been illustrated in the theoretical framework, Mergers and acquisitions can be used in two ways within the pharmaceutical industry. On the one hand they ensure access to patents. On the other hand, they can establish a CEOs authority in the firm. Thus, a variable for M&As as

measured by the dollar denominated yearly expenditure for acquisitions will be included in the regression. In order to test the second hypothesis of this study a two stage least square model is used. The procedure is described in more detail in the next section of this paper.

• As noted above, Anderson, Banker, and Ravindran (2000) explained that executive ownership can affect pay in two opposite ways, either entrenchment or greater commitment. Moreover, it emerges from Griffith’s (1999) study that the amount of ownership claims that CEOs hold in a company, strongly

influences their commitment to the performance of the firm. Therefore, a control variable measuring the percentage of stock held by the CEO over the total equity of the firm is indispensable to capture the pay-performance relationship in its entirety.

(18)

• As illustrated by Kostiuk (1990), executive compensation has notably been an increasing function of firm size. Consistently, a control variable for size will be included by means of the book value of the firms’ total assets.

• Due to the aforementioned influence of risk on the pay-performance relationship, a measure of stock volatility based on the Black-Scholes valuation method has been retrieved from CompuStat. As CompuStat only reports monthly or quarterly data, average annual values have been generated.

The following segment provides a description of the specific econometric techniques applied in carrying out this study.

3.2 Analytical plan

Fixed Effects Regression

To test the first hypothesis the following ordinary least squares regression (OLS) is applied to each of the two sub-samples of this study. As explained in the previous segment, the M&A and R&D variable will be included only for the pharmaceutical sample.

Ln Sensitive pay = α + β1 ln market valuet-1 + β2 M&A + β3 R&D + β4 executive stake + B5 Volatility + µ

This OLS analysis will offer a preliminary image of how the analysis of the pay-performance relationship differs if concentrated on a single industry rather than a broader sample. To avoid issues arising from heteroskedasticity, these two regressions will be carried out with robust standard errors.

Following, a fixed effects model is employed to control for the state of the general economy and the underlying differences across entities. Entity clustered standard errors will be used in this case to limit bias from autocorrelation when using panel data. The specification of the fixed effect model is:

Ln Sensitive payit = β0 + β1 ln market value it-1 + β2 M&Ait + β3 R&D it + β4 Executive stake it + β5 Volatility it + β6 Sizeit + αi + µit

(19)

However, the above described evidence (Grinstein & Haribar, 2004; Kumar, Kuo & Ramchand 2012) suggests that M&As can often be undertaken by CEOs to grow their own wealth. Consequently, it is worth testing the strength of the pay-performance relationship in light of the combined effect of M&As and changes in firm market value. This procedure is outlined below.

Two-Stages Least Squares

To more accurately capture the degree to which remuneration limits managerial opportunism, a two stage least square model has been devised. For this purpose, managerial opportunism is proxied by value destroying acquisitions. Based on the efficient market hypothesis, changes in firm market value will be instrumentalized on acquisition expenditures in the first stage regression. This process should yield the degree to which executives engage in value destroying M&As. In the second stage, this value is regressed on the change in yearly compensation (to generate the variable for change in pay and in market value, data on 2011 compensation has been added). If this regression yields a positive coefficient, one can deduce supporting evidence for the entrenchment theory that CEOs increase their wealth through value destroying M&As.

The following section will present the findings obtained from the application of the above described models.

4. Results and Analysis

4.1 Summary Statistics

It is important to mention that to simplify the interpretation of the regression output, some variables have been transformed into their natural log form. The ln(Y)=βln(X)

regression form allows us to interpret the pay-performance sensitivity in terms of elasticities, i.e. a 1% change in X is associated with a β% change in Y. For the non-transformed

(20)

X causes a 100β% change in (Y). Summary statistics for the two subsamples (i.e. pharma vs non pharma) can be found under Appendix 1. These values are based on the amounts

retrieved for the 2012-2017 period. To ensure comparability, market value and compensation for 2011 have not been included in this calculation as none of the other variables have been collected for that year.

Firstly, the average sensitive pay of a pharma CEO in the sample is $10,800,000.00. The average Market value of the pharmaceutical firms in the sample is $81,800,000,000.00. ROA averages at 0.0418, while average earnings per share are 1.95 after rounding. The average sensitive pay of non-pharma CEOs is $ 11,500,000.00. The average Market value in the mixed sample is $ 78,500,000,000.00. Average ROA is approximately 0.0452, while average earnings per share amount to 3.68 after rounding. At first glance, it can be noted that on average pharmaceutical CEOs are receiving lower incentives to manage more valuable firms. However, their ability to generate profits as shown by ROA does not differ

significantly from their counterparts in the non-pharmaceutical sectors. Next, there is a clear difference between earnings per share. Pharmaceutical companies seem to generate lower earnings for their shareholders. This could in part be due to the business cycle of the

pharmaceutical industry which has been outlined in the theoretical framework. Lastly, to test whether pay is in any way related to the lagged market value the correlation between these two variables has been tested. The Pearson coefficient is 0.706 (Appendix 1c), hence it is worth further analysing this relationship.

The following sections are dedicated to the econometric analysis. Thereafter, practical implications will be derived from the findings.

4.2 Ordinary Least Squares Regression

To introduce the relationships between the variables, two separate ordinary least squares regressions have been conducted. The second column in the table below displays the main results for the sixteen pharmaceutical companies in the sample. The third column is based on the data collected for the eleven publicly traded U.S. non pharmaceutical firms.

(21)

Robust standard errors in parentheses, *** p<0.01, ** p<0.05, * p<0.1

The above output displays the coefficients and in parentheses the respective standard errors for the two precursory regressions. To control for company specific characteristics, size (measured by the accounting value of a firm’s total assets), stock volatility (compounded

VARIABLES Pharma ln sensitive pay Non-Pharma ln sensitive pay Ln market value t-1 0.417*** 0.414*** (0.080) (0.057) M&A 0.000 NA (0.000) NA R&D 0.000 NA (0.000) NA Volatility 2.823 51.401*** (14.42) (10.709) Executive stake -7.504 -5.076*** (7.217) (1.912) Size 0.000 0.000 (0.000) (0.000) Constant 6.060*** 5.121 *** (2.134) (1.487) R-squared 0.555 0.582

(22)

based on the Black-Scholes procedure and retrieved from CompuStat) and executive stake have been included in the regression. Whilst the above is a very simple OLS regression, this does not limit us from making some broad inferences based on the findings.

When analysing the pharmaceutical sample, the only significant variable is the market value of the firm. The relationship between shareholder wealth (i.e. ln market value t-1) and compensation (i.e. ln sensitive pay) is significant at the 1% level. Thus, the null hypothesis that there is no effect of firm market value on the variable fraction of compensation can be rejected. However, it appears that the pay-performance sensitivity in the Pharmaceutical industry (0.417) does not differ significantly from that in the mixed sample (0.414). This disproves the first hypothesis, namely “The interaction between pay and performance is stronger for pharmaceutical firms than for other companies”. The 0.417 coefficient can be interpreted as follows: every 1% change in market value changes the average CEO’s salary by 0.417%. Considering an average market value of $81,800,000,000.00 for pharmaceutical firms, this implies that a $818,000,000.00 (i.e. 1%) increase in firm market value causes the average executive’s sensitive pay to rise from $10,800,000.00 to $10,845,036.00 (i.e. 1.00417 times).

It is also interesting to note that whilst volatility seems to be a significant determinant of CEO pay in the control sample, the same cannot be said for the pharmaceutical industry. This contradicts Aggarwal and Samwick’s (1999) findings of the moderating effect of risk on the elasticity of pay to performance. Nonetheless, it is in line with the low volatility nature of the pharmaceutical industry. Lastly, contrary to what is taught by managerial entrenchment theory, it appears that firm size does not influences the remuneration of the CEO in either sample. Lastly, the R squared shows that the regressors applied to ln sensitive pay explains 43.5% of the total variance in average sensitive pay in the pharma industry. For the mixed sample this value is even higher, namely 58.2%. Therefore, this model appears to fit the data rather well. Notwithstanding, increasing the number of variables inevitably adds explanatory power to a regression as the additional regressors will capture some random variance. Thus, this R squared value might be inflated by the presence of the control variables.

To conclude, evidence was found for a pay-performance relationship in both samples.

However, in the pharma sample there is not enough proof for a significant effect of volatility, executive stake and size on sensitive pay. An explanation might be that entity and time fixed effects have not been kept Cerberus Paribus. This can be achieved by means of a fixed effects model. This will keep the underlying differences across firms constant and it will reduce the impact of the general economic state on performance measures.

(23)

4.3 Fixed Effects Model

To improve the robustness of this study, the data has been analysed by means of a fixed effects model. However, a surprisingly low correlation between u_i and Xb was detected. Therefore, the fixed effects model was tested against a random effects model by means of the Hausman test. The resulting P-value is 0.45 indicating that a random effects model is more appropriate for this study. The output for both models as well as the Hausman test, can be found under appendix 3. The random effects function can be summarized as follows (note that the λ correction applied by the RE model is omitted here):

Ln Sensitive payit = β0 + β1 ln market value it-1 + β2 M&Ait + β3 RD it + β4 Executive stake it + αi + µit

This format helps in controlling for undetected heterogeneity, assuming that this

heterogeneity will be constant through time and uncorrelated to the independent variables. The below table presents the main coefficients obtained from this analysis.

(Refer to appendix 3 for complete output)

VARIABLES Pharma ln Sensitive Pay Non-pharma ln Sensitive Pay lnmarketvalue 0.365*** 0.377*** (0.113) (0.090) Volatility -9.406 47.207*** (22.339) (15.326) Executive stake -3.093 -4.533* (6.801) (2.415)

(24)

M&A 0.000 NA (0.000) NA R&D -0.000** NA (0.000) NA Size 0.000 0.000 (0.000) (0.000) Constant 7.187** 6.082*** (2.998) (2.262) Rho 0.082 0.275

Robust standard errors in parentheses, *** p<0.01, ** p<0.05, * p<0.1

Firstly, in this more accurate regression form, the coefficient for market value is still significant (at the 1% level) in both subsamples. Its accuracy is further confirmed by the relatively low standard error. Thus, the null hypothesis that market value has no effect on pay can again be rejected. A marginal decrease in the coefficient has been registered, from 0.417 to 0.365 in the pharmaceutical sample and from 0.414 to 0.377 in the comparison sample. This could be due to omitted variable bias which has been largely solved by the use of the random effects regression. Additionally, when looking at the between cluster R squared value, it appears that this model captures a large fraction of the variance between firms (over 80%). However, it is important to note that the comparison group displays a significantly higher rho value than the pharmaceutical sample. This indicates that for the mixed sample

(25)

about 27.5% of the variance in the dependent variable is purely due to differences across firms. This could indicate the presence of an industry effect on pay. However, there is not enough evidence to draw such a strong conclusion. Therefore, further research on this interaction is recommended.

Next, executive stake appears to be significant in relation to pay in the mixed sample only. Similarly, volatility also seems to be a significant factor of compensation only in the comparison sample. Lastly, it is worth noting that R&D expenses for the pharmaceutical sample now appear to have a significant effect on pay (p<0.05). Though the coefficient is of negligible size, it is surprising that it has a negative sign. This contrasts with Cheng’s (2004) finding that firms use remuneration mechanisms to encourage R&D spending. Moreover, despite the wide-spread evidence that M&A activity is often conducted in an opportunistic manner, no significant relationship between acquisitions and CEO pay was found. To understand this relationship in more depth, the effect of M&A on pay has been tested by means of a two stage least square regression which will be outlined in more detail in the following sub-section.

4.4 Two Stage Least Squares Regression

As mentioned in the methods section, M&A has been found to be used as a starting point for self-dealing (Jensen, 1986; Grinstein & Haribar, 2004; Kumar, Kuo & Ramchand 2012). Thus, the usefulness of compensation schemes should become visible in their relation to the degree to which they punish value destroying M&A activity. This is analysed via a 2sls regression (output in appendix 4). An important disclaimer is that the F statistic for the first stage regression is 8.29, indicating a weak instrument (F<10). Moreover, the endogeneity test failed to reject the null hypothesis that the instrument can be treated as an exogenous

variable. Thus, any conclusions derived from this regression need to be considered with absolute care. Nonetheless, some broad inferences for further research can be derived from this study. The below table presents the coefficient and in parentheses the standard error for the second stage regression of M&A on Change in sensitive pay.

(26)

VARIABLES Change in sensitive pay M&A -0.000576* (0.000312) Constant 1.901e+06** (804,199) Observations 96 R-squared -0.016

Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1

Firstly, when looking at the first stage regression in output 4 table 4a, a significant negative relationship is found between the change in market value and M&A expenditure (p<0.01 and β = -0.005). If the underlying intuition is correct, this might indicate a prevalence of value destroying M&As in the pharmaceutical industry. Secondly, in the second stage regression presented in the above output, there appears to be a significant (p<0.1) negative relationship between the instrumentalised value of M&A activity and the year to year change in the sensitive fraction of executive compensation. This contrasts with the second hypothesis of this study. In fact, it appears that value destroying M&As are associated with a decrease in executive compensation. Consequently, it seems that compensation schemes in the pharmaceutical industry are correctly designed, thus not allowing CEOs to disguise self-serving acquisitions as a way to gain relevant patents. Nonetheless, it is important to mention that this regression is built on a very strong

assumption, namely that changes in market prices can be used to predict whether M&As will be value increasing or value destroying. However, if further research can corroborate this relationship, this model could be used as a building block for more accurate econometric studies.

(27)

5. Discussion

5.1 Findings

The positive relationship between performance and sensitive pay, which was found in the above regressions, provides evidence for the existence of a pay-performance relationship. However, this study did not reveal any significant difference of pay sensitivity to

performance between the two subsamples. Thus, the first hypothesis “The interaction

between pay and performance is stronger for pharmaceutical firms than for other companies” is not supported by the findings. Accordingly, in both subsamples of this study the

relationship appears to be rather weak. This is in line with the above presented literature. Nonetheless, some important differences surfaced from this analysis. In fact, the expected relationship between executive ownership and compensation could be detected in the pharma sample. This is in contrast with Anderson, Banker, and Ravindran’s (2000) description of the effects of executive ownership. Thus, further research in the underlying dynamics which are intrinsic to the pharmaceutical sector are required to devise an optimal level of executive ownership. It is also interesting to note that while volatility seems to be a significant determinant of CEO pay in the control sample, the same cannot be said for the

pharmaceutical industry. This contradicts Aggarwal and Samwick’s (1999) findings of the moderating effect of risk on the elasticity of pay to performance. Nonetheless, a possible explanation for this finding is that the pharmaceutical industry is characterized by low volatility of demand and thus executives might not request compensation for risk (Ganuza, Llobet & Domínguez, 2009). Next, based on Cheng’s (2004) evidence, a positive relationship between R&D expenditures and pay was expected. However, this study found a negative relationship. This implies that increases in R&D expenditures lead to a reduction in pay. It must be noted that the coefficient is extremely small. Nonetheless, this confirms Bushee’s (1998) evidence that CEOs can opportunistically decrease R&D spending to boost their pay. The revealed relationship is counterproductive to optimal contracting and requires addressing in compensation design. Lastly, a negative effect of value destroying M&As on executive compensation was found. This contradicts the second hypothesis of this study, namely

“Pharmaceutical CEOs opportunistically exploit M&As to increase their own compensation”. However, compensation is not the only way in which CEOs can exploit M&As to their own benefit. Thus, more accurate variables for self-dealing could be generated. Moreover, it is

(28)

important to mention that this second regression was not carried out under accurate circumstances. Therefore, no conclusive evidence can be inferred from this regression. However, the results provide some support for the pay-performance elasticity in the pharma sector. In fact, in contrast with the above described theory of entrenchment, executive

compensation appears to be a decreasing function of value destroying merger and acquisition activity.

5.2 Practical Implications

Consistent with agency theory, this study corroborates the existence of a link between executive compensation and firm performance. Nonetheless, this relationship is weak and does not warrant relying exclusively on compensation schemes to induce appropriate CEO conduct. Accordingly, it be noted that studies such as the one conducted by Malik and Makhdoom (2016) even found a negative effect from tying pay to performance metrics. Further, the negative relationship that was detected between R&D expenditures and compensation highlights an important loophole in the design of optimal contracting in the pharmaceutical industry. Moreover, the analysis of executive opportunism in relation to M&A activity revealed a significant, yet very weak relationship between value destroying acquisitions and compensation. This indicates that while there are incentives put in place to limit value destroying acquisitions, these incentives might not be strong enough to limit managerial opportunism altogether. Lastly, factors such as volatility, firm size and executive stake which are traditionally accepted to have a strong influence on pay, appear to have no significant effect on compensation in the pharmaceutical industry. Thus, it can be concluded that analysing compensation and performance in isolation in a single industry can lead to results significantly different from a cross industry analysis. Notwithstanding, the pay-performance elasticity does not appear to be significantly different in the pharmaceutical industry than in other sectors. In line with the above outlined literature, a combination of optimal contracting and other board control mechanisms is, thus, recommended.

(29)

5.3 Robustness

Firstly, the data used in this study has been cross referenced with official accounting reports and mistakes in the downloaded dataset have been corrected accordingly. In addition, the use of robust standard errors in the OLS regression and entity clustered standard errors in the random effects model further strengthen the accuracy of the measurements. Moreover, by applying a fixed effects model to the study, firm specific influences have been controlled for. To additionally improve the validity of this study, random effects have been adopted for this process. This was recommended by the Hausman test after comparing the fixed effects and random effects regressions. Consistently, the random effects model offered relatively low values for sigma_u (time invariant term) and sigma_e (error term). This implies that there is a low effect of unobserved heterogeneity on the outcome variable, which further supports the strength of this study. Lastly, the significance attached to the coefficients and the R squared measure offer additional support for the robustness of the undertaken regressions. Thus, it can be concluded that the model used in this study strongly captures the relationship between pay and performance. Notwithstanding, some limitation are outlined below.

5.4 Limitations

Omitted Variable Bias

Firstly, due to the limited scope of this study it was not feasible to account for a wide variety of factors. For example, board composition has been proven to be a relevant factor in determining CEO compensation, but the inclusion of such a measure goes far beyond the scope of this study. Therefore, the results are likely to suffer to some degree of omitted variable bias.

Direct to Consumer Advertising

Secondly, the U.S. still allows direct to consumer advertising (DTCA) of medications and treatments. According to the FDA (The Impact of Direct-to-Consumer Advertising, 2015), DTCA has a strong impact on consumer awareness and consumption. Therefore, to isolate the effect of pay on performance, it is crucial that the effect of direct to consumer advertising be kept constant across observations. Due to a lack of data on this specific

category of advertising expenses this could not be done in this research. Therefore, the values presented in the regression output might be overestimated and inaccurate.

(30)

Option Manipulation

Next, Lie (2005) and Heron and Lie (2007) presented evidence of option backdating. Their studies suggest that executives have been found to retroactively choose the option grant date when the price of a firm’s stock was at its lowest for the year. Consistently, Aboody and Kasznik (2000) found evidence that firms time the disclosure of stock option awards

according to the release of good and bad news. This enables executives to capture the

difference between the current price and the lowest price of the firm’s stock without exerting any effort. This can lead to bias in the value of executive compensation and will, thus, influence the validity of the above results.

Uses of Perks

Research has revealed how pension and other compensation can be exploited as a means of rent extraction. Yermack (2006), for example, analysed the use of corporate jets as an indicator of opportunistic attitudes among CEOs. His study revealed that the stocks of firms offering this kind of perquisites consistently underperformed market benchmarks. It is difficult to quantify when use of perquisites turns into opportunistic abuse. Thus, inclusion of the “pension and other income” variable might not be entirely warranted. A better metric could be devised for further research.

Two-Stage Least Squares Assumption

As mentioned above, the two stage least squares regression is based on a very strong assumption. This assumption is grounded in theory. However, empirical evidence on the efficient market hypothesis is mixed. Moreover, from the validity tests it surfaced that the devised model is not entirely accurate for the attempted measurement. It should be noted that the purpose of this study is explorative in nature and it does not attempt to make suggestions to be applied in the real world. Therefore, any conclusions derived from this regression must be considered with utmost care and can at most be used to inspire further research.

Methodology

Lastly, it is important to note that the methods applied in this study are basic

economic regressions. Consistently, more advanced methods might provide different results. Moreover, the limited scope of this study did not allow to carry out a longitudinal analysis of

(31)

the pharmaceutical industry. In fact, due to the research-intensive nature of this industry, projects can take many years to pay out. Further research should attempt to measure the pay-performance relationship with pay-performance measures lagged over several years (e.g.

matching compensation with projects that were undertaken five to ten years ago). Therefore, to repeat, it is not suggested to draw practical implications from the above presented findings. Nonetheless, this study can be used as a starting point for further research.

6. Conclusion

This study compared the pay-performance relationship for firms in the pharmaceutical industry to a mixed sample of firms from different settings. The goal was to reveal how controlling for the underlying industry can affect the measurement of the pay-performance sensitivity. From an initial comparison of means it seemed that pharmaceutical CEOs are being compensated less to handle firms of greater value. However, more accurate regression forms revealed that the pay-performance relationship in the pharmaceutical industry is no more or less intense than in the mixed sample. In fact, the coefficients of performance on pay were almost identical for both samples. Nonetheless, it surfaced that factors which are

traditionally considered crucial in optimal contracting design, such as volatility, size and executive ownership, do not affect pay in the pharmaceutical sector. This indicates that different industries might require the prioritisation of different aspects when setting incentives. Accordingly, a substantial amount of variance in the mixed sample appears to arise simply from differences across firms. (rho=27.5%). As these corporations were not divided per industry yet random effects were controlled for, one could assume that this variance was brought forth by industry factors. This warrants further investigation into the influence that industry characteristics might have on the pay-performance relationship. Moreover, in the pharmaceutical industry there appears to be a negative effect of R&D investment on pay, meaning that R&D expenditures depress pay. This is detrimental in a research driven industry such as pharmaceuticals, and it requires further insight for the sake of optimal contracting. Lastly, an attempt was made to measure CEO opportunism in terms of value destroying M&A activity. Whilst a prevalence of value destroying acquisitions was detected, it surfaced that on average CEOs are being punished for such behaviour via a reduction in compensation.

(32)

confirmed the existence of a weak pay-performance relationship. Thus, an excessive focus on compensation techniques to direct CEO conduct is not recommended. Additionally, no significant difference was found when analysing this relationship in a single industry as compared to a broader sample. Nonetheless, some factors indicate that the nature of the industry in which this relationship is being analysed might cause different elements to be more or less relevant in optimal contracting. To improve the usefulness of incentive-based compensation, further research into these intricacies is warranted.

(33)

References

Aboody, D., & Kasznik, R. (2000). CEO stock option awards and the timing of corporate voluntary disclosures. Journal of accounting and economics, 29(1), 73-100.

Aggarwal, R. K., & Samwick, A. A. (1999). The other side of the trade-off: The impact of risk on executive compensation. Journal of political economy, 107(1), 65-105.

Anderson, M. C., Banker, R. D., & Ravindran, S. (2000). Executive compensation in the information technology industry. Management Science, 46(4), 530-547.

Ang, J. S., Cole, R. A., & Lin, J. W. (2000). Agency costs and ownership structure. the Journal of Finance, 55(1), 81-106.

Berger, P. G., Ofek, E., & Yermack, D. L. (1997). Managerial entrenchment and capital structure decisions. The journal of finance, 52(4), 1411-1438.

Bertrand, M., & Mullainathan, S. (2001). Are CEOs rewarded for luck? The ones without principals are. The Quarterly Journal of Economics, 116(3), 901-932.

Bhagat, S., & Bolton, B. (2008). Corporate governance and firm performance. Journal of corporate finance, 14(3), 257-273.

Bushee, B. (1998). The influence of institutional investors on myopic R&D investment behavior. The Accounting Review, 73, 305-333

Chen, C. R., Steiner, T. L., & Whyte, A. M. (2006). Does stock option-based

executive compensation induce risk-taking? An analysis of the banking industry. Journal of

Banking & Finance, 30(3), 915-945.

Cheng, S. (2004). R&D expenditures and CEO compensation. The Accounting

Review, 79(2), 305-328.

Conyon, M. J. (2006). Executive compensation and incentives. Academy of

Management Perspectives, 20(1), 25-44.

Conyon, M. J., & He, L. (2011). Executive compensation and corporate governance in China. Journal of Corporate Finance, 17(4), 1158-1175.

(34)

Cuñat, V., & Guadalupe, M. (2009). Executive compensation and competition in the banking and financial sectors. Journal of Banking & Finance, 33(3), 495-504.

Elson, C. (2003). What's wrong with executive compensation?. Harvard Business

Review, 81(1), 68-77.

Elston, J. A., & Goldberg, L. G. (2003). Executive compensation and agency costs in Germany. Journal of Banking & Finance, 27(7), 1391-1410.

Fama, E. F. (1980). Agency problems and the theory of the firm. Journal of political

economy, 88(2), 288-307.

Frydman, C., & Jenter, D. (2010). CEO compensation. Annual Review of Financial

Economics, 2(1), 75-102.

Ganuza, J. J., Llobet, G., & Domínguez, B. (2009). R&D in the pharmaceutical industry: a world of small innovations. Management Science, 55(4), 539-551.

Garvey, G. T., & Milbourn, T. T. (2006). Asymmetric benchmarking in

compensation: Executives are rewarded for good luck but not penalized for bad. Journal of

Financial Economics, 82(1), 197-225.

Hall, B. J., & Liebman, J. B. (1998). Are CEOs really paid like bureaucrats?. The

Quarterly Journal of Economics, 113(3), 653-691.

Hartzell, J. C., & Starks, L. T. (2003). Institutional investors and executive compensation. The Journal of Finance, 58(6), 2351-2374.

Heron, R. A., & Lie, E. (2007). Does backdating explain the stock price pattern around executive stock option grants?. Journal of Financial Economics, 83(2), 271-295.

Hill, C. W., & Jones, T. M. (1992). Stakeholder‐agency theory. Journal of

management studies, 29(2), 2.

Jensen, M. C., & Meckling, W. H. (1979). Theory of the firm: Managerial behavior, agency costs and ownership structure. Journal of financial economics, 3(4), 305-360.

Jensen, M. C., & Murphy, K. J. (1990). Performance pay and top-management incentives. Journal of political economy, 98(2), 225-264.

(35)

Kostiuk, P. F. (1990). Firm size and executive compensation. Journal of human

Resources, 25(1), 90-105.

Kumar, P., Kuo, L. W., & Ramchand, L. (2012). Stock Performance or

Entrenchment? The Effects of Mergers and Acquisitions on CEO Compensation. SSRN (July

8, 2012).

Lavalette, M., & Pratt, A. (Eds.). (2001). Social policy: a conceptual and theoretical

introduction. SAGE Publications Limited, Ch 3, pp.32-51

Lie, E. (2005). On the timing of CEO stock option awards. Management

Science, 51(5), 802-812.

Malik, M. S., & Makhdoom, D. D. (2016). Does corporate governance beget firm

performance in fortune global 500 companies?. Corporate Governance 16(4). Emerald, West

Yorkshire.

Malkiel, B. G. (1989). Efficient market hypothesis. In Finance (pp. 127-134). Palgrave Macmillan, London.

Mehran, H. (1995). Executive compensation structure, ownership, and firm performance. Journal of financial economics, 38(2), 163-184.

Paris, J. (2010). The use and misuse of psychiatric drugs. London (GB): John Wiley &

Sons, Ltd. Chapter 3.

Pavlou, A. K., & Belsey, M. J. (2005). BioPharma licensing and M&A trends. Nature

Review Drugs Discovery, 4 (4), 273-274.

Roach, G. P., & Goedde, A. G. (2003). CEO Compensation in the Pharmaceutical Industry. Compensation & Benefits Review, 35(5), 66-81.

Schifferes, S. (2001, December 31). ‘Will the world fall into recession?’ BBC News. http://news.bbc.co.uk/2/hi/in_depth/world/2001/review_of_2001/1701893.stm

Sheikholeslami, M. (2001). EVA, MVA, and CEO compensation. American Business

(36)

Sloan, R. G. (1993). Accounting earnings and top executive compensation. Journal of

accounting and Economics, 16(1-3), 55-100.

The Impact of Direct-to-Consumer Advertising. (2015, October 23). U.S. Food & Drug Administration. Retrieved June 12, 2020, from https://www.fda.gov/drugs/drug-information-consumers/impact-direct-consumer-advertising

Yermack, D. (2006). Flights of fancy: Corporate jets, CEO perquisites, and inferior shareholder returns. Journal of financial economics, 80(1), 211-242.

Zhou, X. (2000). CEO pay, firm size, and corporate performance: evidence from Canada. Canadian Journal of Economics/Revue canadienne d'économique, 33(1), 213-251.

(37)

Appendix

Appendix 1

Summary Statistics

(Values in USD except for stock held/firm equity and Return on Assets)

Table 1a. Pharma sample

Variable Observ ations

Mean Std.Dev. Min Max

sensitivepay 96 1.08e+07 1.48e+07 214000 1.38e+08 lnsensitivepay 96 15.668 1.175 12.274 18.744

Marketvalue 96 8.18e+10 1.30e+11 1.60e+07 1.04e+12 lnmarketvalue 96 23.584 2.282 16.588 27.672

ROA 96 .042 .133 -.373 .457

eps 96 1.949 3.097 -6.63 12.37

M&A 96 1.85e+09 4.85e+09 0 3.52e+10 R&D 96 3.03e+09 3.29e+09 7120000 1.10e+10 Executive

stake

96 .008 .017 0 .134

Size 96 41745.02 52357.84 104.26 186000 netincome 96 3.49e+09 5.24e+09 -4.72e+08 2.20e+10

(38)

Table 1b. Non-pharma sample

Variable Observ

ations Mean Std.Dev. Min Max sensitivepay 66 1.15e+07 1.29e+07 1130000 9.36e+07 lnsensitivepay 66 15.831 .937 13.939 18.355 Marketvalue 66 7.85e+10 9.43e+10 6970000 3.66e+11 lnmarketvalue 66 23.984 1.997 15.757 26.627 ROA 66 .045 .028 -.014 .092 eps 66 3.676 3.064 -10.37 13.07 Executive stake 66 .017 .038 0 .142 Size 66 272000 699000 3967.89 2570000 netincome 66 4.80e+09 7.07e+09 -8.03e+07 2.47e+10

Table 1c. Correlations Ln sensitive pay Ln Market Value t-1

Variables (1) (2) (1) lnsensitivepay 1.000 (2) LnMarketValue t-1 0.706 1.000

(39)

Appendix 2

OLS

Table 2a. Ordinary Least Squares Regression Pharmaceutical Companies

lnsensitivepay Coef. Robus St.

Error t-value p-value [95% Confiden ce Interval] Sig LnMarketValu e t-1 0.417 0.080 5.22 0.000 0.258 0.576 *** M&A 0.000 0.000 0.46 0.650 0.000 0.000 R&D 0.000 0.000 -1.06 0.291 0.000 0.000 Volatility 2.823 14.418 0.20 0.845 -25.831 31.476 Executive stake -7.504 7.217 -1.04 0.301 -21.847 6.839 Size 0.000 0.000 -0.47 0.642 0.000 0.000 Constant 6.059 2.134 2.84 0.006 1.819 10.300 ***

Mean dependent var 15.672 SD dependent var 1.181 R-squared 0.555 Number of obs 95.000

F-test . Prob > F .

Akaike crit. (AIC) 233.305 Bayesian crit. (BIC) 246.075

(40)

Table 2b. Ordinary Least Squares Regression Non-Pharmaceutical Companies

lnsensitivepay Coef. Robust

St. Error t-value p-value [95% Confiden ce Interval] Sig LnMarketValu e t-1 0.414 0.057 7.21 0.000 0.299 0.529 *** Volatility 51.401 10.709 4.80 0.000 29.988 72.814 *** Executive stake -5.077 1.192 -4.26 0.000 -7.460 -2.693 *** Size 0.000 0.000 0.45 0.656 0.000 0.000 Constant 5.121 1.487 3.44 0.001 2.146 8.095 ***

F-test 63.703 Prob > F 0.000

(41)

Appendix 3

Time and entity fixed effects

Table 3a. Entity and Time fixed effects regression Pharma sample

lnsensitivepay Coef. Entity Clustered St. Error t-value p-value [95% Confiden ce Interval] Sig LnMarketValu e t-1 0.243 0.154 1.58 0.135 -0.085 0.570 M&A 0.000 0.000 0.19 0.855 0.000 0.000 R&D 0.000 0.000 -2.21 0.043 0.000 0.000 ** Executive stake 1.477 3.838 0.39 0.706 -6.704 9.659 2012b.year 0.000 . . . . . 2013.year 0.161 0.118 1.36 0.193 -0.091 0.412 2014.year 0.903 0.207 4.36 0.001 0.461 1.345 *** 2015.year 0.565 0.429 1.32 0.208 -0.349 1.479 2016.year 0.480 0.224 2.15 0.049 0.003 0.957 ** 2017.year 0.877 0.328 2.67 0.017 0.178 1.577 ** Constant 9.967 3.524 2.83 0.013 2.456 17.478 **

F-test . Prob > F .

Sigma_u 0.84882095 Sigma_e 0.76123802

Rho: 0.55423685 R-sq:

(42)

between = 0.1475

overall = 0.1982 corr(u_i, Xb) =

0.0413

*** p<0.01, ** p<0.05, * p<0.1

Table 3b. Entity and Time fixed effects regression Non-pharma sample

lnsensitivepay Coef. Entity Clustered St. Error

t-value p-value [95% Confiden ce Interval] Sig LnMarketValu e t-1 0.208 0.174 1.19 0.261 -0.180 0.595 Executive stake 2.066 3.749 0.55 0.594 -6.288 10.419 2012b.year 0.000 . . . . . 2013.year 0.047 0.207 0.23 0.827 -0.415 0.508 2014.year 0.111 0.180 0.62 0.551 -0.290 0.513 2015.year 0.214 0.347 0.61 0.553 -0.561 0.988 2016.year -0.044 0.348 -0.13 0.903 -0.819 0.732 2017.year 0.330 0.252 1.31 0.219 -0.231 0.891 Constant 10.723 4.074 2.63 0.025 1.645 19.800 **

F-test 3.096 Prob > F 0.056

Sigma_u 0.63084581 Sigma_e 0.57189156

(43)

R-sq: within = 0.1367 between = 0.4408 overall = 0.3345 corr(u_i, Xb) = 0.2566 *** p<0.01, ** p<0.05, * p<0.1

Table 3c. Hausman test for fixed effects vs random effects

Coef. Chi-square test

value 6.8

P-value .45

Table 3d. Random effects regression Pharma sample

lnsensitivepay Coef. Entity Clustered St. Error

t-value p-value [95% Confiden ce Interval] Sig LnMarketValu e t-1 0.365 0.113 3.24 0.001 0.144 0.585 *** M&A 0.000 0.000 0.34 0.731 0.000 0.000 R&D 0.000 0.000 -1.30 0.194 0.000 0.000 Volatility -9.406 22.339 -0.42 0.674 -53.189 34.378 Executive stake -3.093 6.801 -0.46 0.649 -16.423 10.238 Size 0.000 0.000 0.02 0.985 0.000 0.000 2012b.year 0.000 . . . . .

Is the pay-performance relationship industry-specific? Evidence from the U.S. pharmaceutical sector.