The effect of uncertainty on pay dispersion’s impact on firm performance

‘The effect of Uncertainty on Pay Dispersion’s

impact on firm performance”

Master Thesis

Student:

Athanasia- Grammatiki Antoniou

Student Number:

11084324

Date:

01/02/2017

Supervisor:

Dr. J. Van de Ven

TABLE OF CONTENTS

ABSTRACT

This research elaborates the relationship between intra-firm pay dispersed compensation schemes and firm’s performance, in 105 North American companies that are enlisted in Compustat and Execucompt Databases of Standard & Poor’s. Contributing to the literature of agency and tournament theory, I address this matter with respect to the situational factor of economic uncertainty. I model uncertainty following Dess & Beard (1984) and approaching it as a system of three different aspects: availability/ lack of resources (munificence), environment’s sensitivity to change (dynamism) and market share of the firm (complexity). Specifically, I try to find the effect of these three different aspects of uncertainty in the impact of pay dispersion on firm’s performance. Results indicate that there is not a significant effect of uncertainty on the relationship between firm performance and pay dispersion.

INTRODUCTION

In the business world, we often witness situations where the CEO and the chief executives are distinguished by the average employees by a pay difference so large, that almost equals the absolute amount of their total wage. This phenomenon is known as pay dispersion, also referred to as spread, range, variation, and inequality, and is defined as the differences in pay levels between individuals within and across jobs or organizational levels (Shaw et al. 2002). Two theories that address this issue, each of them approaching it from different aspects, are agency and tournament theory. Agency theory (as developed by Ross 1973-4, Mitnick 1975, Jensen & Meckling 1976) tries to shed some light, by highlighting the fact that principals and agents that stand in different positions within a firm do not share the same interests and goals. Therefore it is most likely that they won’t act in favor of each other’s benefit. For that reason, principals introduced pay-for-performance measures to obtain control over the agents’ actions, by compensating them based on their measurable output. These payment schemes create pay differences between employees. In a relative spirit, tournament theory (developed by Lazear & Rosen 1981, Rosen 1986, Nalebuff & Stiglitz 1983) suggest that wage differences serve as a signaling of the potential and future career that employees can achieve. That way, being paid by your rank in a company acts as a motivation to raise in class and employees increase their effort and boost firm’s performance.

Many empirical studies have been conducted on the subject of pay dispersion, giving confusing evidence, sometimes in favor and sometimes against the use of wage differentials. For instance, Erikson (1999), Main et al. (1993), Bloom & Mitchel (2002) found proof in favor of tournament theory while on the contrary Bloom (1999), Pfeffer and Langton (1992/3) found the opposite. I will elaborate more on this on the literature review.

My research contributes to this literature and focuses on the fact that there are situational factors that might affect the relationship of pay structure and firm performance. Specifically, I examine the effect of uncertainty on this relationship. When employees’ productivity is affected by exogenous factors or noise, risk- averse employees will reduce their efforts, because such efforts are less likely to affect their productivity and their output (Lazear& Rosen 1981, Lee et al. 2008). A larger pay dispersion will counteract (Lazear 1995) or, according to others (Hart 1995), will intensify the impact of uncertainty on employees’ efforts and productivity.

Therefore, I decided to condition the examined relation of pay differentials and firm's performance on the interaction between uncertainty and pay dispersion, following the model of Lee et al. (2008) who examined the same thing but conditioning it on the interaction of payment distribution with agency costs and corporate governance. I will discuss this in detail in the methodology section. My research question is formed as: “Does uncertainty enforces or damages the impact of pay dispersion on firm’s performance?” As far as I know, there has not been a research yet that focus on this particular matter. Another contribution is that this research focuses not on pay dispersion generally but on intra-firm compensation dispersion, the one that happens not between different companies but inside the same firm. The stream of literature that concerns intra-firm differences has not been developed as much as the one that analyses between corporations and industries wage structures.

My research suggests that uncertainty does not have an influence on pay dispersion’s impact on firm performance. When I use the whole sample of firms but also when I focus specifically on circumstances where uncertainty is high, this conclusion is repeated. Moreover, some variables of uncertainty have consistent effects (always positive or negative but never significant) regardless of the sample I am using.

This research is structured as follows: In the next section there will be a literature review and analysis, and then a section that refers to the aspects of this particular research. The paper will proceed with a chapter about methodology, and then the results will follow. Finally, there will be a discussion of the main findings and a conclusion.

LITERATURE REVIEW

The tremendous difference between a CEO’s compensation compared to that of an average employee is rational to challenge our perception of fairness. One might wonder about the morality of someone earning a life's fortune annually, in the expense of someone earning much less. But what might seem peculiar to public opinion, economist have achieved to justify with their models. The topic of intra-firm pay dispersion on firm productivity embodies the analysis and perspectives of different theories and views. In this section, my aim is to give an overview of these theories and focus on their influence on the investigated topic and to clarify the foundations upon which the modern payment system is based.

Agency theory

Agency theory is relevant to comprehend the intuition on which compensation is based. It addresses two problems (Eisenhardt, 1989). According to Jensen & Meckling, (1976) an agency relationship is defined as a contract under which a Principal delegates some of his authority to another person (the Agent) to perform a task on his behalf. The first problem is that if both parties are utility maximizers, then it is very probable that the Agent will not always act in the best interest of the Principal, but in his own best interest. The second problem is a risk-sharing issue that occurs when principal and agent hold different attitudes on risk. For instance, Principal might be risk neutral, but Agent might be risk averse and dislikes it when the outcome of his efforts is uncertain. Then the principal and the agent may prefer different actions because of their different risk preferences. (Wilson, R. 1968). For example, the Agent might avoid doing his best in a task if he is not sure that he will succeed.

To make this more concrete, I will refer to Hart (1995). The Agent’s performance depends on the effort e he will exert and on some randomness ε, that is beyond his control (I will elaborate more on ε later on the literature review), so 𝑞 = 𝑔(𝑒, 𝜀). The output q is observable by the Principal, but neither e nor ε is. Assume also that the Agent dislikes expressing effort, so he bears a cost 𝐶(𝑒). Finally, Principal gains from Agent’s performance by 𝑟(𝑞) and he is risk neutral while Agent is risk-averse. In that case, Principal’s utility would be 𝑈_𝑃(𝑟(𝑞) − 𝑃) = 𝑟(𝑞) − 𝑃 and agent’s utility would be 𝑈𝐴(𝑃, 𝑒) = 𝑉(𝑃) − 𝐶(𝑒) where P is Agent’s payment,

and V is a concave function. If the principal could observe e, then he would offer the Agent a fixed amount 𝑃∗ _{if Agent would choose the optimal effort 𝑒}∗_{. This amount 𝑃}∗_{ensures optimal}

risk sharing: risk-neutral principal bears all risk of the realization of 𝑒, and risk averse Agent none. But since Agent dislikes to exert effort and effort is not observable, he would only exert 𝑒 < 𝑒∗. To get the Agent to work for him, Principal must compensate him according to his produced output 𝑞, with 𝑃 = 𝑃(𝑞) and 𝑃(𝑞) = 𝑤 + 𝑏 ∗ 𝑞. That way, the Agent’s final compensation will depend on his performance. With this incentive scheme, both parties face the trade-off between optimal incentives and optimal risk sharing. The bigger is the variable part of the compensation (𝑏 ∗ 𝑞) that depends on 𝑞, the more incentivized is the Agent to work. On the other hand, he is exposed to greater risk.

To conclude, the Principal has to establish appropriate incentives for the Agents so that they will not deviate from the Principal’s interests, and to choose between a fixed oriented contract and an outcome-based one (Eisenhardt, 1989). In other words, he has to consider pay-dispersed compensation schemes, where employees are paid differently according to their produced output.

Empirical results in this topic provide us with mixed signals about the validity of the theory and do not always prove a positive relationship between pay-for- performance incentives and realized performance. According to Barkema and Gomez-Mejia (1998), many researchers have found weak or none statistical significance between output-based compensations and performance. One study that they refer is conducted by Jensen& Murphy (1990). They investigated the magnitude of incentives provided by different payment systems, in a sample of publicly owned firms. Their results failed to state an explicit support of the agency model, as the relationship between CEOs’ compensation and performance of the firm was rather weak compared to the theoretical suggestions. To interpret this phenomenon, the authors point at the significance of external factors such as political forces, public opinion or labor market institutions, factors that would be included in the randomness 𝜀 of Hart’s model. On the other hand, Lazear (2000) confirmed agency’s theory hypothesis about output-based payment. In his research in Safelife Glass Corporation, when the company changed the compensation scheme from a fixed wage to piece rates, productivity increased around 44% in output per worker. The principal – agent relationship help us understand the theoretical framework on which compensation on the modern business world was built. In order, though to obtain a more rounded view on the matter of pay dispersion, we also need to address tournament theory.

Tournament theory

While Agency theory studies the effect of absolute payment on employees’ provision of effort, tournament theory comes to contribute by focusing on the relative payment within a firm. Formulated by Lazear and Rosen this theory acted as a counterweight to the standard notion of compensation being based on the value of the produced marginal product. They supported that another model is possible and more efficient, a model of competitive lotteries. According to that model, prizes and penalties are attributed to winners and losers employees in a setting of a market contest (Lazear and Rosen 1981). The innovative element is that these prizes are not assigned relative to the produced output, but rather by a comparison between hierarchical levels. A tournament competition is created between employees, and they are getting paid based on their rank within the firm. That way, workers effort increases to rise in ranks and salary. Precisely, Lazear and Rosen (1981) in their model considered a two-player (𝑖, 𝑗) tournament and a set of two fixed prizes: 𝑊1 for the winner and 𝑊2 for the loser. A worker’s produced

outcome is 𝑞 = 𝜇 + 𝜀, where 𝜇 is the worker’s effort, and 𝜀 is a random component drawn from a known distribution with 𝐸(𝜀) = 0 and 𝑉𝑎𝑟(𝜀) = 𝜎2_{. There is also a cost of effort 𝐶(𝜇), and}

it is assumed that both 𝐶′ and 𝐶′′ are positive. The probability that worker j wins depends on his effort 𝜇𝑗 and negatively on his opponent’s effort 𝜇𝑖 and it is also affected by the distribution

of 𝜀. The probability that j wins is:

𝑃 = 𝑝𝑟𝑜𝑏(𝑞_𝑗 > 𝑞_𝑖) = 𝑝𝑟𝑜𝑏(𝜇_𝑗− 𝜇_𝑖 > 𝜀_𝑖 − 𝜀_𝑗) = 𝑝𝑟𝑜𝑏 (𝜇_𝑗− 𝜇_𝑖 > 𝜉) = 𝐺(𝜇_𝑗− 𝜇_𝑖) (1) , where 𝜉 = 𝜀_𝑖− 𝜀_𝑗, ξ~g[𝜉], G is the cumulative distribution function of 𝜉, 𝛦(𝜉) = 0 and 𝐸(𝜉2_{) = 2𝜎}2 _{(because 𝜀}

𝑖 and 𝜀𝑗 are i.i.d.)

Α worker’s expected utility is (𝑃)[𝑊₁− 𝐶(𝜇)] + (1 − 𝑃)[𝑊₂− 𝐶(𝜇)] = 𝑃𝑊₁+ (1 −

𝑃)𝑊2− 𝐶(𝜇) (2). Each worker chooses 𝜇 tο maximize his expected utility, so

(𝑊₁− 𝑊₂)𝜕𝑃 𝜕𝜇− 𝜕𝐶(𝜇) 𝜕𝜇 = 0 (3) and (𝑊1− 𝑤2) 𝜕2𝑃 𝜕𝜇2− 𝜕2𝐶

𝜕𝜇2 < 0. Both players maximize (3) by

taking as given the other player’s effort choice. So from (1) it follows that for worker j:

𝜕𝑃 𝜕𝜇𝑗= 𝜕𝐺

(𝜇𝑗−𝜇𝑖)

𝜕𝜇𝑗 = 𝑔(𝜇𝑗− 𝜇𝑖) which by substitution at (3) gives the reaction function of

worker j: (𝑊₁− 𝑊₂)𝑔(𝜇_𝑗− 𝜇_𝑖) − 𝐶′(𝜇_𝑗) = 0 (4). The other player’s reaction function is symmetrical. This symmetry implies that when the Nash equilibrium exists, 𝜇𝑗 = 𝜇𝑖 and

𝑃 = 𝐺(0) =1

2. So finally, by substituting 𝜇𝑗 = 𝜇𝑖 at (4) we get 𝐶 ′_(𝜇

𝑗) = (𝑊1− 𝑊2)𝑔(0).

This final equation proves that that the worker’s effort level is increasing as the spread between the prizes increases.

Tournament theory has raised quite a debate among economists. The main criticism lies with the fact that this kind of competition challenges perceptions of morality and fairness between employees, resulting in a hostile environment for cooperation and damaging firm’s productivity. As an answer to this, Lazear (1989) says that with an equal compensation, the morale of the employees in the upper tail of the value creation distribution might as well be negatively affected. To expand the theory, Lazear elaborates about the importance of the reference group. To determine the reference group based on which employees will be compared, there must be caution as to intra-firm relationships- who works with who (Lazear 1989). When cooperation is necessary, compensation contests should be avoided in order not to cause rivalry. For instance, managers of the same position should be compensated similarly, while payment can vary between hierarchical levels. Empirical evidence exactly on this matter is provided by Erikson (1999). In his research, he proves that hierarchical levels significantly determine payment schemes. The more levels in the hierarchy, the more pay dispersion we encounter. Moreover, the wage dispersion phenomenon is more intense in the upper hierarchical levels, while payments remain equal within the same rank. The more someone rises in the corporate structure, the more payment diversifies.

Furthermore, an important determinant of tournament theory are the characteristics of the human capital itself. Lazear (1989) notes that personality should be considered as a hiring criterion and a company should carefully decide whether to mix or match between personality types. His point is that taking into account that average productivity is lower in firms with more competitive employees and also that wage compressed mechanisms are more appropriate in competitive firms; then the average wage would be lower in compressed systems (Lazaer 1989). That way, he correlates lower wages with egalitarian payment and personality characteristics. Lazear's argument is also confirmed empirically by Erikson (1999). Theory and empirics also suggest that between the higher levels of hierarchy, if the setting is such that encourages comparisons, then the firm should be oriented in less dispersed compensations systems. But if the setting is such that does not leave space for personality and social characteristics to cause conflicts between the higher executives, then tournament models are more appropriate (Frederikson et al. 2010).

Uncertainty of Environment

In the previous sub-sections, we saw that agency theory suggests the use of performance-based payments, which leads to payment differences among employees, as an incentivizing mechanism and as a method of interest alignment between principal and agent. A more extreme belief is expressed by tournament theory which treats wages as competitive lotteries and uses their distribution as signaling and motivational scheme.

It is important though to keep in mind that there are many determinants in the equation between payment structures, provision of effort and, consequentially, firm’s performance. A change in these factors might intensify or on the contrary reverse this relationship. One of the factors that influence this balance is risk and uncertainty of the environment (Eisenhardt, 1988). The more external factors beyond management's control can affect the firm’s outcome, the more risk is created, the more uncertain becomes the environment on which the company operates.

Based on the model of Hart& Holmstrom (1987) and Hart (1995) that I referred in the agency theory sub-section, the incentive intensity principal is derived and indicates the factors that determine the optimal pay for performance contract. As I mentioned before, the output 𝑞 in the Principal – Agent model equals 𝑞 = 𝑒 + 𝜀 (𝑒 is agent’s effort and is unobservable, and ε is noise with 𝛦(𝜀) = 0 and 𝑉𝑎𝑟(𝜀) = 𝜎2_{) and cost of effort equals 𝑐(𝑒) =}1

2𝑐𝑒

2 _{. The agent has CARA}

utility function, 𝑟 is the measure of risk aversion, and contract 𝑃(𝑞) is linear: 𝑤 = 𝑠 + 𝑏 ∗ 𝑞 . In the Hart&Holmstrom (1987) and Hart (1995) model, Agent’s expected utility is: 𝐸(𝑃) −

𝑟

2𝑣𝑎𝑟(𝑃) − 1 2𝑒

2_{. Principal is trying to maximize his own utility by max 𝐸(𝑞 − 𝑠 − 𝑏 ∗ 𝑞) =}

(1 − 𝑏)𝑒 − 𝑠 (1) subject to agent’s 𝑚𝑎𝑥𝐸(𝑃(𝑞)) −𝑟 2𝑣𝑎𝑟(𝑃(𝑞)) − 1 2𝑒 2 _{(2) and 𝐸(𝑃(𝑞)) −} 𝑟 2𝑣𝑎𝑟(𝑃(𝑞)) − 1 2𝑒 2 _{> 0 (3). Because 𝐸(𝑃(𝑞)) = 𝑠 + 𝑏 ∗ 𝑒 and 𝑣𝑎𝑟(𝑃(𝑞)) = 𝑏}2_𝜎2_{, (2)}

reduces to e=b. Formula (3), also named Participation Constrain, is binding in optimal so it must be 𝑠 + 𝑏2 −𝑟 2𝑏 2_𝜎2 ₋1 2𝑏 2 _{= 0 → 𝑠 =}𝑟 2𝑏 2_𝜎2₋1 2𝑏

2_{. Then principal’s utility}

maximization in (1) becomes 𝑚𝑎𝑥(1 − 𝑏)𝑒 −𝑟

2𝑏

2_𝜎2₋1 2𝑏

2 _{. By taking the first order}

condition, we get the optimal bonus 𝑏 = 1

1+𝑟𝑐𝜎2 and the incentive intensity principal. The incentive intensity principle tells us that the variable part of compensation (b) decreases when the cost of effort (c) increases, the degree of risk aversion (r) increases and the randomness’s variance (𝜎2_{) increases. As I mentioned before, this randomness includes factors beyond}

measure uncertainty in the following sections is consistent with this definition of randomness 𝜀. We can see, therefore, that according to agency theory, when uncertainty rises, incentives b diminish, and pay dispersed compensations are not preferred. Empirical evidence complies with the theory. Eisenhardt (1988) hypotheses that, in a setting of salesmen, outcome uncertainty will be positively related to the use of salaries and negatively related to the use of commission. Indeed results indicate that in risky occasions, performance-based payments are avoided. Similarly, Bloom & Milkovich (1997) found a negative relationship between the use of contingent pay and firm performance with regard to external factors and specifically with the instability of the environment. They showed that companies in more uncertain environments that tend to use incentive pay present a decline in their performance.

On the contrary, tournament theory suggests that riskier environments demand more intense dispersion in the distribution of payment to overcome the damage made by the risk (Lazaer 1995). More precisely in the model of Lazaer and Rosen (1981) that I analyzed before, they find that 𝐶′_(𝜇

𝑗) = (𝑊1− 𝑊2)𝑔(0) (look in p.16). This indicates that the provision of effort

increases according to the spread between the prizes 𝑊1 and 𝑊2. In this expression,

𝑔(𝜇_𝑗− 𝜇_𝑖) = 𝑔(0) because in the Nash equilibrium 𝜇_𝑗 = 𝜇_𝑖. This expression also implies that

the greater is the importance of the randomness 𝜀, or else uncertainty ¸in the output (i.e., the smaller 𝑔(0) becomes), then the optimum level of effort for a given spread 𝑊₁ and 𝑊₂ decreases. Hence, in environments in which factors beyond the agent’s control have a substantial effect on his output, firms use a larger prize dispersion to offset the effect of uncertainty (Lazear and Rosen 1981, Erikson 1999)

In his research Erikson (1999) interprets uncertainty in terms of demand and cost condition. He finds a positive and significant relationship between the variability of sales, which proxies for the environmental uncertainty, and the gap between payments within a firm. He concludes in a steeper compensation level hierarchy when the environment is characterized by instability. Lazear (1995) elaborates that companies that operate in a noisy environment will demand intense pay dispersion to overcome the effort-damaging effect of the increased risk. Unstable environments make the monitoring and evaluation of performance rather difficult in absolute terms, and because tournament compensations are based on relative performance standards, they are considered much more appropriate. Similarly, Bloom et al. (2002) in their research they find a positive relationship between noisy environments and pay dispersion. Specifically, in an uncertain environment, firms with greater investment opportunity needed more dispersed compensation schemes, and there was also a positive relationship between pay differentials and

firm’s performance.

THIS RESEARCH

In this section, I will describe the purpose and the theoretical framework of my research. As it was made apparent in the literature review, pay dispersion has been the issue of interest in the agency and tournament theories. These theories have been analyzed, expanded and enriched, but empirical results give mixed signals. Some studies provide evidence in confirmation of tournament theory and pay differentials while others fail to do so. But one can support that reality lies somewhere in between these perspectives and that there are various determinants of the appropriate payment structure, dispersed or compressed. The purpose of this research is to suggest which compensation scheme is more suitable, regarding firm's performance, with respect to environmental uncertainty. In simple words, to investigate if environmental uncertainty has a positive or adverse effect on the relationship between pay dispersion and firm’s performance and if this effect gives evidence in favor or against the use of pay differentials. Thus, the research question is formed as follows:

“Does the interaction effect of pay dispersion and uncertainty strengthens or attenuates the effect of pay dispersion’s on firm performance?”

There are already studies on the determinants of firm performance. The contribution of this study to the existing literature is that it intends to focus only on the case of uncertainty and, moreover, on its combined effect with pay dispersion. To my knowledge, there has not been a research yet that examines the interaction term of the compensation structure with the elements of environmental uncertainty, as I will attempt to do so. The exact model that I will use will be explained in detail in the methodology section.

In the literature, uncertainty is approached as a coherent system of three different components, such as in Child (1972), Khandwalla (1972) Dess& Beard (1984) Tan& Litschert (1994). In my research, I will follow Dess& Beards model of munificence, dynamism, and complexity (Dess& Beard 1984, Tan& Litschert 1994, Anderson & Tushman 1986). Munificence in Dess& Beard, or else hostility in Tan& Litschert, captures the organization's capacity which is based on the availability or scarcity of resources. Dynamism in Tan& Litschert, named uncertainty in Anderson, describes the volatility and intensity of change that characterizes an environment and that makes foreseeing of events and decision-making difficult. Complexity finally represents

the heterogeneity or homogeneity of firm’s activities and market’s segments (Dess& Beard 1984, Khandwalla 1972). The exact measurement of these components will be described in the methodology.

Based on the existing literature, I will form three hypothesis, each one referring to one of the three components of environmental uncertainty. In addition, I will develop my hypothesis so that they will be expressing an argument in favor of tournament theory.

As I showed in the Uncertainty sub-section, tournament theory states that in the case of uncertainty, pay dispersion should be intensified. Higher levels of munificence indicate the availability of resources that increases the firm’s capacity to deal with risk and uncertainty. Therefore, higher levels of munificence signal more certainty on the environment, while lower ones indicate uncertainty. In cases of stability, it is reported in the literature, even by tournament theorists, that compressed compensation structures are more appropriate, while dispersion might backfire by damaging cooperation among employees. In that case:

H1: Munificence will have a negative effect on the relationship between Pay dispersion and firm’s performance/ Scarcity will have a positive effect on the relationship between Pay dispersion and firm's performance

Dynamism indicates the variability of the environment and the degree to which it is exposed to change. Therefore it shows high levels of uncertainty. That way, hypothesis two would be: H2: Dynamism will have a positive effect on the relationship between pay dispersion and firm’s performance.

Similarly, complexity also indicates a sensitivity to uncertainty, as a company involved in many different activities is faced with greater uncertainty.

H3: Complexity/heterogeneity will have a positive effect on the relationship between pay dispersion and firm’s performance.

METHODOLOGY

In this section, I will describe the methodology that I will use. It will include information on the research design, the data sample, the description and measurement of the chosen variables.

Data

All data used in this research are acquired from Standard and Poor’s Compustat and Execucompt Databases, reached after subscription at Wharton Research Data Services. Compustat includes the necessary financial information, and in Execucompt are available all data regarding executives’ compensation. These databases include only companies that operate in North America and, consequently, this research will refer to North’s America corporate environment.

The data is a panel data sample consisted by 105 firms in a period from 2004 to 2014, resulting in 1155 observations. The initial size of the Execucompt dataset was 1268 active firms. This number has been so considerably reduced due to the scarcity of Compustat’s data that refer to the total employees’ expenses and the number of employees, which are necessary for the construction of the chosen variables. As a result, the majority of this 1268 firms was dismissed. There was a slight reduction in the sample size due to the availability of performance and past performance data and, finally, the sample was stabilized at 105 firms. Although more options of the dependent and independent variables could have been used, I avoided it so the sample would not be further diminished. Furthermore, it is essential for the demands of this research that these companies belong to different industry sectors, and it is a criterion that they fulfill. Industries are distinguished by their four-digit Standard Industrial Classification (SIC) code. The names of the included firms and the SIC codes of industries are available in the Appendix.

Τhe Model

This research’s objective is to find if the environment’s uncertainty strengthens or attenuates the relationship between pay dispersion and firm performance. The model I will use draws its inspiration from Lee et al. 2008. The effect of the compensation spread can be found by regressing our measure of pay dispersion on firm’s performance controlling for the important factors, in this case, uncertainty, firm's size and past performance as we will discuss next. To find the effect of uncertainty on the relationship between pay dispersion and firm's performance, I included the interaction terms between the measures of uncertainty and pay dispersion. Precisely, the model will be:

𝐹𝑖𝑟𝑚′𝑠 𝑃𝑒𝑟𝑓𝑜𝑟𝑚𝑎𝑛𝑐𝑒_𝑖𝑡

= 𝛽0+ 𝛽1𝑃𝑎𝑦𝐷𝑖𝑠𝑝𝑒𝑟𝑠𝑖𝑜𝑛𝑖𝑡+ 𝛽2𝑃𝑎𝑦 𝐷𝑖𝑠𝑝𝑒𝑟𝑠𝑖𝑜𝑛𝑖𝑡∗ 𝑀𝑢𝑛𝑖𝑓𝑖𝑐𝑒𝑛𝑐𝑒𝑖𝑡

+ 𝛽₃𝑃𝑎𝑦𝐷𝑖𝑠𝑝𝑒𝑟𝑠𝑖𝑜𝑚_𝑖𝑡∗ 𝐷𝑦𝑛𝑎𝑚𝑖𝑠𝑚_𝑖𝑡 + 𝛽₄𝑃𝑎𝑦𝐷𝑖𝑠𝑝𝑒𝑟𝑠𝑖𝑜𝑛_𝑖𝑡

∗ 𝐶𝑜𝑚𝑝𝑙𝑒𝑥𝑖𝑡𝑦_𝑖𝑡+ 𝛽₅𝑀𝑢𝑛𝑖𝑓𝑖𝑐𝑒𝑛𝑐𝑒_𝑖𝑡+ 𝛽₆𝐷𝑦𝑛𝑎𝑚𝑖𝑠𝑚_𝑖𝑡+ 𝛽₇𝐶𝑜𝑚𝑝𝑙𝑒𝑥𝑖𝑡𝑦_𝑖𝑡 + 𝛽8𝑆𝑖𝑧𝑒𝑖𝑡+ 𝛽9𝑃𝑎𝑠𝑡𝑃𝑒𝑟𝑓𝑜𝑟𝑚𝑎𝑛𝑐𝑒𝑖𝑡+ 𝜀𝑖𝑡

, where i indicates the specific company and t the specific year. Including the interaction effects, allows the effect of a change in Pay Dispersion on firm's performance to depend on the values of munificence, dynamism, and complexity. Τhe interactions mean that the effect of the compensation spread on the company's performance is different for the different values of munificence, dynamism, and complexity. That way the effect of Pay Dispersion on Returns On Assets is not just 𝛽₁, but 𝛽₁+ 𝛽₂+ 𝛽₃+ 𝛽₄. The coefficients of interest are those of the interaction terms (𝛽₂, 𝛽₃ 𝑎𝑛𝑑 𝛽₄), since they represent the effect of uncertainty on the relationship between firm performance and pay dispersion and they can either enforce it or diminish it. Also, it is important to include the individual covariates of munificence, dynamism, and complexity to avoid an omitted variable bias. This, of course, doesn’t mean that they have some other profound or causal meaning.

To take this model one step further, we have to notice the following. In the way it is formed now, the constant 𝛽₀ is firm’s performance when there is no pay dispersion within a company and when there is also no uncertainty in the environment. But it is rather unrealistic to assume that there is a case where all employees have exactly the same compensation or the environment in which the firm operates is completely neutral. Besides that, because both measures of pay dispersion and uncertainty will be continuous variables, the model requires more specification to be interpretable. For this reason, I will center the variables first, by subtracting the means from each case, and then compute the interaction terms and the regression. That way the means of the centered variables will be zero. Positive (negative) values in dynamism and complexity (munificence) will mean uncertainty (as they will be above the average) - while negative (positive) values in dynamism and complexity (munificence) will mean the opposite. Hence, centering doesn’t change what the model predicts, but it changes the interpretation of the variables (Williams 2015).

Variables

Dependent Variable

The dependent variable in this research will be firm’s performance. There is a wide variety of measures of firm's performance in empirical studies such as Sales, Net Income, Tobin’s Q (Carpenter &Sanders 2002) ,Return on Equity (O’ Reilly, Brian G. Main & Graef S. 1988), stock performance (Tosi, Jr. and Luis R. Gomez-Meji 1994), Return on Assets (Carpenter &Sanders 2002), wage levels (Winter-Ebmer & Zweimuller 1999). In my study I will use the Return on Assets, to better investigate and confirm or reject the relationship between dependent and independent variables.

Return on Assets is the ratio between firm’s Net Income and the total book value of firm’s assets.

𝑅𝑂𝐴 =

𝑁𝑒𝑡 𝐼𝑛𝑐𝑜𝑚𝑒

𝑇𝑜𝑡𝑎𝑙 𝐵𝑜𝑜𝑘 𝑉𝑎𝑙𝑢𝑒 𝑜𝑓 𝐴𝑠𝑠𝑒𝑡𝑠

Net Income can be directly downloaded from Compustat, but Total Book Value of Assets has to be calculated. In this case, I used the Total Assets - Debt in Current Liabilities (Total) – Long-term Debt (Total). (In Compustat codes: AT- DLC- DLTT)

Independent variables

I have used two kinds of independent variables: Pay Dispersion and the interaction of Pay Dispersion with Environmental Uncertainty. The primary coefficients of interest are the coefficients of the interaction terms.

Pay Dispersion

Pay Dispersion, like company's performance, has also been measured in several ways. Erikson (1999), measured it as the difference between the logarithms of CEO payment minus the logarithm of the average managerial pay. Other instruments are the Gini coefficient ( Bloom& Mitchel,2002), wage differentials conditional on worker’s observable characteristics (Winter- Ebmer and Josef Zweimuller 1999, Fredrik Heyman 2005), the coefficient of variation of wages and the 90–10th percentile wage ratio (Fredrik Heyman 2005). In this study as a measure of

intra-firm pay dispersion I will consider Pay Gap and Pay Gap Ratio (Lallemand, Plasman & Rycx, 2004).

Pay Gap is the difference between the average employees’ compensation and the average top 5 executives’ compensation, inside a firm. Pay Gap Ratio is the above result but divided with the average workers’ compensation, indicating how many times the average workers’ wage fits to the wages spread. Precisely:

PayGap= Top 5’s average compensation- average employees’ compensation

and

𝑃𝑎𝑦 𝐺𝑎𝑝 𝑅𝑎𝑡𝑖𝑜 =Top 5

′_{s average compenssation − average employees′ compesation}

Average employees′compenation

Compensation of top executives can be found directly through Execucompt Database and can easily be averaged. It includes salary, bonus, other annuals, restricted stock grants, LTIP payouts, values of option grants. To calculate the worker’s average compensation, Total Staff Expenses (XLR) must be divided by the number of employees (EMP). Both can be downloaded from Compustat. Then, the top 5’s executives’ averaged compensation must be extracted from it. To get the Pay Gap Ratio, that difference must be divided by the number of employees.

Uncertainty of environment

As I mentioned before, in the literature, uncertainty is approached as a coherent system of three different components, and I will use Dess& Beards three dimension model. To measure these dimensions, I will refer to Boyed’s (1990) research on boards of directors confronting environmental uncertainty, as well as to Bloom& Michel (2002) who investigated the effect of pay dispersion on managerial turnover and tenure.

According to Boyed, munificence represents the availability of resources which is measured by the growth of industry sales. Precisely, I will regress the logarithm of the annual four-digit SIC industry sales of the previous five years into a measure of time. Specifically, following Keats& Hitt (1988) the regression I will use is 𝑦 = 𝑏0+ 𝑏1∗ 𝑡+ ε, where y is the five years industry

sales and t is a time index. Then the regression slope coefficient divided by the mean value of sales measures the munificence. Dynamism is measured as the variability of the industry sales

growth rate over the same period (Boyed 1990) and is calculated as the standard error of the slope coefficient of the above regression, divided as well with the mean value of sales.

Environmental complexity refers to the heterogeneity of activities, so consequently has an impact on the inputs and outputs produced but the firm as well as those of its competitors. The Herfindahl ‐ Hirschmann index is a suitable measure for it. It is calculated as the sum of the squared market shares of all firms in each industry (Boyed 1990). Precisely:

Herfindahl‐ Hirschmann index = ∑( 𝑆𝑎𝑙𝑒𝑠 𝑜𝑓 𝑓𝑖𝑟𝑚 𝑆𝑎𝑙𝑒𝑠 𝑜𝑓 𝐼𝑛𝑑𝑢𝑠𝑡𝑟𝑦)

2 𝑁

𝑡

, where the ratio 𝑠𝑎𝑙𝑒𝑠 𝑜𝑓 𝑓𝑖𝑟𝑚𝑠

𝑠𝑎𝑙𝑒𝑠 𝑜𝑓 𝑖𝑛𝑑𝑢𝑠𝑡𝑟𝑦 is the relevant market share. According to Boyed, the smaller

the Herfindahl Index would be, the more equal is the market share among firms, the more equal their size and power. Consequently, the more complex the environment would be. To conclude, as Herfindahl Index drops, uncertainty rises.

Control Variables

In the context of this research, it is possible to use various control variables. The ones that I chose are the organizational size (Erikson 1999) as well as firm’s past performance (Bloom 1999). In order to define past performance, the performance indicator of Return on Assets will be used in a five years depth. In addition, instead of including five different indicators for each of the past years, I will use the average of past performance’s return on assets. As far as firm size is concerned, there are plenty of options. It can be measured directly by sales or assets (O'Reilly III, Brian G. Main& Graef S. 1988) as well as by the number of total employees (Tosi, Jr., and Gomez-Mejia). Erikson (1999) used the logarithm of sales. In this research, I will use the logarithm number of total assets as it has been used by Vijayakumar et al. (2010), Papadogonas (2007) and Maja et al. (2012). These studies are conducted in a similar setting like this one, addressing effects on firm’s performance as well ,and distinguished the logarithm of assets as the most appropriate measure of organizational size

RESULTS

In this section, I will discuss the results of the econometric analysis. It includes descriptive statistics of the variables, analysis of correlations between variables and regressions’ results.

Summary statistics are illustrated in Table 1 below:

We notice that numbers differ due to the distinct nature of each variable. Pay Gap’s mean, for instance, is around 3.24 million, a number quite large since it represents pay dispersion in absolute terms. On the other hand, Pay Gap Ratio’s mean represents how many times does the average employee’s wage fit to the spread with the top executives’ average pay, so it is a number far smaller than Pay Gap’s mean. All variables count 1155 observations each, so our panel data is balanced. It is also worth noticing that the variables that measure uncertainty have rather low mean values suggesting that probably the economic sectors where the firms of the sample operate, indicated by the four- digit SIC number, have low dynamism and munificence but higher complexity.

Table 2 summarizes the correlations between the variables. It helps us to evaluate the relationships between pay dispersion and firm performance and to decide which combination of them is more appropriate. It is also interesting to look the correlations between the other variables as well. The correlations I used are Pearson pairwise correlations. Between the measures of pay dispersion and Return on Assets, the combination of Returns on Assets and PayGapRatio is, significantly correlated (ρ=0,235*).This, of course, does not imply causation, but it indicates that PayGapRatio has some predictive content on firm’s performance. For this reason, I will use the pair of PayGapRatio with ROA to proceed in the analysis, and the combination of PayGap with

ROA will be excluded.

Return on Assets and Average Past Return on Assets are also positively and significantly correlated (0.379*), which is expected since they both measure firm performance. Similarly, PayGap and PayGapRatio are also positively and significantly correlated (0.276*), which is also expected since PayGapRatio is calculated based in PayGap. Also, Munificence and Dynamism are correlated (-0.404*), which implies a connection between the availability of resources and the intensity of change. Finally, it also worth noticing that from the three measures of uncertainty only Herfindahl Index has a significant correlation with PayGapRatio (0.134*).

The hypothesis are tested using linear regressions. In total, 7 Fixed Effect regressions with clustered standard errors are run. Firstly, the hypothesis are tested using the whole sample of the 105 selected firms, from 2004 to 2014. Following the model I described before, depended variable in these models is firm’s performance, measured by ReturnOnAssets (ROA), independent variables are the interaction terms of PayGapRatio with uncertainty, and control variables are also included. Results are illustrated in Table 3.

Pay Gap Pay Gap

Ratio ROA

Average Past ROA

Logarith

of Assets Munificence Dynamism

Herfidahl Index Pay Gap 1 Pay Gap Ratio 0.2769* 1 ROA 0.0207 0.2351* 1 Average Past ROA -0.0344 0.2625* 0.3792* 1 Logarithm of Assets 0.6193* 0.0750* -0.2109* -0.2963* 1 Munificence -0.0072 -0.0294 -0.0142 -0.0714* 0.0801* 1 Dynamism -0.0462 -0.0287 0.0388 0.0169 -0.1221* -0.4044* 1 Herfidahl Index 0.0744* 0.1344* 0.1148* 0.2380* -0.1746* -0.0529 -0.007 1

Table 2: Pearson Correlations

Table 2 illustrates pairwise Pearson correlations between measures of firm performance, pay dispersion, uncertainty and control variables. Significance indicated with a * at 5% level.

The coefficients of interest are not significant, so we fail to confirm our Hypothesis, and we can say that there is no effect of uncertainty on the relationship between firm performance and pay dispersion. Nevertheless, it is also worth to take a closer look at the signs of the coefficients of interest and see what additional conclusions we can draw from them.

The coefficient of PayGapRatio in positive in all cases, although not substantial in magnitude. That comes in favor of tournament theory, as it seems that Pay Dispersion boosts firm performance.

Model 3 includes the interaction terms, which are of interest in this research. Munificence, as I said before, indicates the availability of resources. Higher levels of munificence indicate plenty of resources in the industry where the firm operates and signal certainty. Lower levels indicate a scarcity of resources and uncertainty. The variable of munificence is centered therefore values

Variables Model 1 Model 2 Model 3

Independent

Pay Gap Ratio 0.0001*** 0.000088* 0.000084

Interaction: Munificence* PayGapRatio 0.0005

Interaction: Dynamism* PayGapRatio 0.0011

Interaction: Herfidahl I.* PayGapRatio 0.000019

Control

Average Past ROA -0.328*** -0.33*** -0.33***

Logarithm of Assets 0.00024 0.012 0.011 Munificence 0.085 0.0911 Dynamism 0.012 0.028 Herfidahl Index -0.016 -0.0212 Constant 0.054 -0.069 -0.058 R^2 Within 0.036 0.044 0.047 R^2 Between 0.709 0.644 0.702 R^2 Overall 0.076 0.091 0.097 N 1155 1155 1155

TABLE 3: Results of regression analysis between ROA, Pay Gap Ratio and Interaction terms with Uncertainty

Model 1-3 are Fixed Effects models with clustered standart errors. Model 3 is the one of focus, as it includes the variables of interest. The interaction terms of the uncertainty measures with PayGapRatio are all positive, but small in magnitude and lack significance. Their sign is in contrast with Hypothesis 1& 3 and in line with Hypothesis 2

below zero signal munificence below average and uncertainty. The coefficient of the interaction is (0.0005) which shows that for an increase in PayGaRatio or an increase in Munificence there will be an increase of (0.0005) on the effect of pay dispersion on firm performance. Then the overall effect of PayGaRatio on ROA will be 0.000084+0.0005=0.000584. This is against our first hypothesis, that munificence (certainty of the environment) has a negative effect on the relationship between dispersed compensation schemes and firm’s performance.

The interaction term between Dynamism and PayGapRatio is (0.0011). We have hypothesized that dynamism represents the degree of which the company’s environment is exposed to change. Therefore it indicates high levels of uncertainty and will have a positive effect on the relationship between pay dispersion and firm’s performance, according to tournament theory. Therefore, the result is in line with Hypothesis 2.

In Hypothesis 3, it is stated that a complex environment will have a positive effect on the relationship between the compensation range and firm performance. Our measure of complexity, Herfindahl Index, the smaller it is, the more complex the environment. Therefore, as HI drops, uncertainty rises. The coefficient of this interaction indicates that if HI rises by 1 (therefore uncertainty drops), this will have a 0.000019 increase in the impact that PayGapRatio has on ROA. This is against Hypothesis 3 that predicts the opposite: that an increase in complexity (decrease of HI) will have a positive effect.

To summarize, the lack of significance in results indicates that there is not a particular effect of uncertainty on the relationship between firm performance and pay dispersion. By interpreting only the signs of the coefficients of interest, we see that they are in line with Hypothesis 2, but in contrast with Hypothesis 1&3.

I repeated the above analysis but using binary variables for environment’s uncertainty. I calculated the average of Munificence, Dynamism and Herfindahl Index and created three binary variables: Scarcity, B.Dynamism, and Complexity. Scarcity equals one for each firm for which munificence is below average, indicating high levels of uncertainty, and zero otherwise. Similarly, Complexity equals 1 when Herfindahl Index is below average and zero otherwise. Finally, B.Dynamism equals one if Dynamism is above its average. Results are illustrated in Table 4.

Again, there is not a significant effect of uncertainty on the impact of pay dispersion on firm’s performance, so we do not confirm our Hypothesis. By taking a closer look at the signs of coefficients, we see that B.Dynamism’s coefficient is positive, in line with Hypothesis 2 and tournament theory. On the other hand, Scarcity and Complexity are negative, suggesting that

higher uncertainty decreases the impact of pay dispersion, and this is in contrast with Hypothesis 1&3. We can notice that there is consistency in the behavior of the interaction coefficients between Model 3 and Model 4.

I decided to implement the same analysis under circumstances where uncertainty is more explicit. I divided the sample into subsamples where the presence of uncertainty is clearer, based on the average value of each of uncertainty’s measures. The first group includes all firms that are suffering from scarcity of recourses (Scarcity=1). The second includes all enterprises which face an above average dynamic environment (B.Dynamism=1), and the third includes those that operate in a high complex environment (Complexirty=1). Note that these sub-samples, are

Variables Model 4

Independent

Pay Gap Ratio 0.0001***

Interaction: Scarcity* PayGapRatio -0.0000829 Interaction: B.Dynamism* PayGapRatio 0.0000542 Interaction: Complexity* PayGapRatio -0.0000124

Control

Average Past ROA -0.336***

Logarithm of Assets 0.001 Scarcity -0.0109 Dynamism -0.0045 Complexity 0.0136 Constant 0.037 R^2 Within 0.039 R^2 Between 0.679 R^2 Overall 0.081 N 1155

Model 4 is a Fixed Effects model with clustered standart errors. Scarcity, B.Dynamism & Complexity are binary variables: when 1 they indicate high levels of Uncertainty and 0 otherwise. The signs of the interaction terms reject Hypothesis 1& 3, and confirm Hypothesis 2.

TABLE 4: Results of regression analysis between ROA, Pay Gap Ratio and Interaction terms with Uncertainty, where uncertainty variables are

unbalanced panel data. They are illustrated in Graph 1.

Results are illustrated in Table 5. In this case of explicit uncertainty as well, there is no effect of uncertainty on the relationship between pay dispersion and firm's performance, as the lack of significance indicates. In models 5, the analysis is run on the sample where uncertainty exists due to the scarcity of resources, so we will focus only on the coefficient of interaction between Scarcity and PayGapRatio. The coefficient is small in magnitude (0.00095) and positive, and it is in line with Hypothesis 1 that lack of resources enforces the effect of pay dispersion on firm performance. Respectively, in model 6 the coefficient that concerns us is that of the interaction of PayGapRatio with Dynamism (0.0015). Its positive sign is also in line with Hypothesis 2. On the other hand, Hypothesis 3 is in contrast with the results in model 7. The interaction of PayGapRatio with Complexity is negative (-0.00074) and indicates a decrease on the relationship between firm performance and pay dispersion.

To conclude, the scheme with the separate sub- samples that indicate an environment with clearly uncertain conditions, did not result in different results than using the whole sample. In this case as well, uncertainty did not have a significant effect on pay dispersion and firm performance since results were not significant. But we did saw a change in the sign of the scarcity interaction with PayGapRatio from negative to positive.

Graph1 : Subsample's Density

Graph 1 illustrates the number of observations of each sub sample 497 452 765 0 100 200 300 400 500 600 700 800 900

Scarcity Dynamism Complexity

N u m b e r o f O b se rv at io n s Sub Samples

SU B SAMPL E S D E N SITY

Variables Model 5 Model 6 Model 7 Independent

Pay Gap Ratio 0.000058 -0.000041 0.000076

Interaction: Scarcity* PayGapRatio 0.000952 -0.00011 0.00044 Interaction: Dynamism* PayGapRatio -0.00011 0.0015 0.00071 Interaction: Complexity* PayGapRatio -0.0011 0.00001 -0.00074

Control

Average Past ROA -2.21* -2.12* -.30*

Logarithm of Assets 0.019 0.031 0.04 Munificence 0.069 0.065 0.15 Dynamism 0.01 0.014 0.15 Herfidahl Index 0.086 0.377* 0.66 Constant -0.076 -0.184 -0.35 R^2 Within 0.373 0.347 0.039 R^2 Between 0.423 0.435 0.31 R^2 Overall 0.0018 0.0009 0.072 N 497 452 765

TABLE 5: Results of regression analysis between ROA, Pay Gap Ratio and Uncertainty Interactions, on sub samples.

This tables shows results of fixed effects regression with clustered standart erros, on sub samples. Dependent variable is firm performance measured by ROA. Independent variables are Pay Gap Ratio and Uncertainty Interactionss. In Model (5) we are interested only in the interaction of Munificence with PayGapRatio, in Model (6) in the PayGapRatio-

DISCUSSION

As I concluded in the previous section, this research finds no effect of uncertainty on the relationship between pay dispersion and firm’s performance. Therefore, it confirms neither tournament’s nor agency’s theory suggestions. This conclusion is repeated all three times that I ran the analysis: when using continuous variables, when using binary variables, and when using subsamples.

Just by interpreting only the signs the coefficients of interest in the whole sample, we see that munificence and complexity do not confirm Hypothesis 1 and 3, but Dynamism is in line with Hypothesis 2. The issue here is that since uncertainty is consisted of these three components, in order to conclude that uncertainty enforces or reduces the impact that payment differentials have on the performance of the firm, there must be a confirmation/ rejection on the effect of all these three aspects of this relationship. Hence, to come to a conclusion, we must confirm or reject all of the hypothesis together, as a system. From our results, what we can say is that munificence and complexity do not enforce the impact of dispersed compensation structures, but on the other hand the volatility of change, represented by Dynamism, does.

To double-check the result that there is not an effect of uncertainty on pay’s dispersion impact on performance, I decided to repeat the analysis under circumstances where the uncertainty of environment was more explicit and visible. To do so, I divided the initial sample into smaller ones regarding the average of the uncertainty’s measures. The conclusion of no significant impact is confirmed here as well. Again, just by looking at the signs of coefficients in models 5-7, not all them point at their hypothesized directions.

There are some limitations in this study. If one considers the size of the sample, the non-significant results are partially explained. The initial sample consists of 105 companies and 1155 observations overall, while each subsample consists of 452 observations the smallest to 765 the largest. As I mention in the methodology before, due to the requirements and aspects of this research, the (sub)samples’ size could not be increased. It would be very interesting, though, to see if the conclusions of this research would remain the same if we repeat the same regressions in a larger sample. In any case, the larger the sample, the more reliable the results. Another issue is the variables used. Most researchers in the literature control for the effect of past performance and organizational size. I tried different measures of these in my research before concluding in the use of average past returns on assets and logarithm of assets, without a significant change in the final result. But perhaps other parameters should be included in the

regression analysis as well, and the regression model suffers from omitted variable bias. Also, another issue concerns the measurement of pay dispersion. The data used from Compustat to calculate employees’ and executives’ compensations resulted in a pay gap between the firm’s ranks that is very large and almost equals the executives’ total wage. That affects the Pay Gap Ratio as well. It would be interesting to see the regressions’ results if compensation was more narrowly defined and the pay gap was not so enlarged.

Further research should also take into consideration the following. The data in the subsamples are panel data, but unbalanced. In the regression analysis, I proceeded as if the missingness of the data was completely random. Perhaps there should be a further investigation to assess if there is a pattern in the missing data, and different models should be used in the subsamples’ regressions.

CONCLUSION

The goal of this research was to investigate the effect of environment’s uncertainty on the relationship between pay dispersion and firm performance, in order to provide us with proofs in favor or against the use of pay differentials, with regard to the situational factor of uncertainty. It aimed to be a contribution to the literature of tournament and agency theory. To analyze this relationship, Ι used the method of hypothesis testing in a sample of 105 North American firms, from 2004 to 2014 and ran seven regressions. Results did not show an impact of environment’s uncertainty on pay dispersion and firm’s performance. When I divided the sample into smaller ones, in order to have precise circumstances of uncertainty, this conclusion remained. We could also observe that in most cases, dynamism tended to show a small positive effect but, on the contrary, scarcity of resources and complexity tended to show the opposite.

REFERENCES

Aghion and Jean Tirole. (1997) Formal and Real Authority in Organizations. Journal of

Political Economy. 1-29

Anderson, P., & Tushman, M. L. (2001). Organizational environments and industry exit: The effects of uncertainty, munificence and complexity. Industrial and Corporate

Change, 10: 675-711.

Barkema &Gomez-Mejia.(1989) Managerial Compensation and Firm Performance: A General Research Framework. The Academy of Management JournaL. Vol. 41. 135-145 Bloom & Milkovich. (1997). The relationship between Risk, Incentive Pay and Organizational Performance. CAHRS Working paper

Bloom& Michel. (2002).The relationships among organizational context, pay dispersion, and managerial turnover. Academy of Management Journal. Vol(45).33‐ 42

Bloom.(1999). The performance effects of pay dispersion on individuals and organizations. Academy of Management Journal. Vol(42). 25‐ 40

Bourgeois. (1980), Strategy and Environment: A Conceptual Integration. The Academy

of Management Review. Vol. 5, 25-39

Boyed. (1990). Corporate Linkages and Organizational environment: a test of the resources dependence model. Strategic Management Journal. Vol(11). 419-430

Carpender & Sanders (2002). Top Management Team Compensation: The missing link between CEO pay and firm performance. Strategic Management Journal. Vol(23). 367- 375

Child. (1972) .Organization structure, environment and performance: The role of strategic choice. Sociology, 6: 1 -22

Cowherd& Levine.(1992) . Product Quality and Pay Equity Between Lower-Level Employees and Top Management: An Investigation of Distributive Justice Theory.

Administrative Science Quarterly. Vol(37).302-320

Davis et al. (1991).Perceived environmental turbulence and its effect on selected

entrepreneurship, marketing, and organizational characteristics in industrial firms.

Journal of the Academy of Marketing Science. Vol(19).43-51

Dess & Beard. (1984). Dimensions of Organizational Task Environments. Administrative

Eisenhardt .(1989).Agency Theory: An Assessment and Review. Academy of

Management Review, Vol. 14,57-74.

Eisenhardt. (1988),.Agency- and Institutional-Theory Explanations: The Case of Retail Sales Compensation. The Academy of Management Journal.Vol. 31. 488-511

Eriksson. (1999). Executive Compensation and Tournament Theory: Empirical Tests on Danish Data. Journal of Labor Economics. Vol. 17. 262-280

Fredrik Heyman. (2005).Pay inequality and firm performance: evidence from matched employer–employee data .Applied Economics.Vol (37).1313–1327 Fredrikson et al. Sharing the wealth: social comparisons and pay dispersion in CEO’s top team. Strategic Management Journal. Vol(31).1031 – 1053

Jensen & Meckling.(1976). Theory of the firm: managerial behaviour, agency cost and ownership structure. Journal of Financial Economics 3. 305-360

Jensen& Murphy. (1990). Performance Pay and Top-Management Incentives.

Journal of Political Economy. Vol. (98).225-26

Keats& Hitt (1988) .A Causal Model of Linkages among Environmental

Dimensions, Macro Organizational Characteristics, and Performance. The

Academy of Management JournalVol. 31, No. 3, pp. 570-598

Khandwalla (1972), Environment and its impact on organizations. International

Studies of Management & Organization. Vol. 2. 297-313

Kor, Y.Y., Watson, S., Mahoney (2004). Industry effects on the use of board and institutional investor monitoring and executive incentive compensation.

University of Illinois at Urbana-Champaign Working Paper.

Lazear& Rosen. (1981). Rank-Order Tournaments as Optimum Labour Contracts.

The Journal of Political Economy. Vol (89).841-864

Lazear. (1989). Pay Equality and Industrial Politics. Journal of Political Economy. Vol. 97, 561-580

Lazear.(1995).Personell economics.Cambridge: MA:MIT press

Lazear.(2000). Performance Pay and Productivity. The American Economic

Review. Vol. 90. 1346-1361

Lee, Lev& Yeo. (2008). Executive pay dispersion, corporate governance and firm performance. Review of Quantitative Finance and Accounting. Vol 30, 315-338 Martin. (1982) .The Fairness of Earnings Differentials: An Experimental Study of

33

the Perceptions of Blue-Collar Workers. The Journal of Human Resources. Vol (17).110-122

Maja P, Josipa V (2012). Influence of firm size on business success. Croatian

Operational Research Review (CRORR), 3: 213-224.

Main, B., O'Reilly, C. A., & Wade, J. (1993). Top executive pay: tournament of teamwork? Journal of Labor Economics, 11: 606 ‐ 628

Miller &Friesen. (1983), Strategy-Making and Environment: The Third Link.

Strategic Management Journal. Vol. 4.221-235

Mitnick Barry. (1975). The theory of agency: The policing paradox and regulatory behavior. Public choice.24: 27-42

Nalebuff & Stiglitz (1983). Prizes and Incentives: Towards a General Theory of Compensation and Competition .The Bell Journal of Economics, Vol. 14, No. 1 pp. 21-4

O'Reilly et al. (1988), CEO Compensation as Tournament and Social Comparison: A Tale of Two Theories. Administrative Science Quarterly. Vol. 33, 257-274 Papadogonas, T.A. (2007), “The financial performance of large and small firms: evidence from Greece”, Int. J. Financial Services Management, Vol. 2, No. ½, pp. 14 – 20

Pfeffer. (2007).Human Resources from an Organizational Behavior Perspective: Some Paradoxes Explained. Journal of Economic Perspectives. Vol(21), 115–134 Pfeffer, J., & Langton, N. (1993). The effects of wage dispersion on satisfaction, productivity, and working collaboratively: evidence from college and university faculty. Administrative science quarterly, 38‐ 3: 382‐ 407

Pritchard. (1969).Equity Theory: A Review and Critique. Organizational Behaviour and Human Performance. Vol(4).176-211

Tan & Litschert. (1994). Environment- Strategy Relationship and its Implications: An empirical study of the Chinese electronics Industry. Strategic Management

Journal, Vol(15), 1-20

Tosi & Gomez-Mejia. (1994). CEO Compensation Monitoring and Firm Performance. The Academy of Management Journal. Vol. 37, 1002-101

Shaw et al. (2002). Pay dispersion and workforce performance: moderating effects of incentives and interdependence. Strategic Management Journal. Vol (23). 491-512

Vijayakumar, A. and Tamizhselvan, P. (2010), “Corporate Size and Profitability-An Empirical Profitability-Analysis”, College Sadhana – Journal for Bloomers of Research,

Vol. 3, No. 1, pp. 44 – 53.

Williams.(2015). Interpreting Interaction Effects; Interaction Effects and Centering.University of Notre Dame. Working paper

Wilson, R. (1968) On the theory of syndicates. Econometrica, 36, 119-132.

Winter-Ebmer&Zweimuller.(1999). Intra-firm Wage Dispersion and Firm Performance. Kyklos. Vol (52). 555-572

Zabojnik, Jan (1996). Pay-Performance Sensitivity and Production Uncertainty.

35

APPENDIX

List of firms

ATLAS AIR WORLDWIDE HLDG INC HEARTLAND EXPRESS INC SUNTRUST BANKS INC

BANK MUTUAL CORP HUB GROUP INC -CL A SUSQUEHANNA BANCSHARES INC

BANK OF AMERICA CORP HUDSON CITY BANCORP INC SVB FINANCIAL GROUP BANK OF HAWAII CORP HUNT (JB) TRANSPRT SVCS INC SYNOVUS FINANCIAL CORP BANK OF NEW YORK MELLON CORP HUNTINGTON BANCSHARES TCF FINANCIAL CORP

BB&T CORP INDEPENDENT BANK CORP/MI TD AMERITRADE HOLDING CORP BOB EVANS FARMS INTERPUBLIC GROUP OF COS TRUSTCO BANK CORP/NY BOSTON PRIVATE FINL HOLDINGS IRON MOUNTAIN INC U S BANCORP

BROOKLINE BANCORP INC JETBLUE AIRWAYS CORP UNION PACIFIC CORP

BROWN & BROWN INC JPMORGAN CHASE & CO UNITED BANKSHARES INC/WV

CASCADE BANCORP KANSAS CITY SOUTHERN UNITED PARCEL SERVICE INC

CATERPILLAR INC KEYCORP UNITED STATES STEEL CORP

CATHAY GENERAL BANCORP KINDRED HEALTHCARE INC WEBSTER FINANCIAL CORP CENTRAL PACIFIC FINANCIAL CP KNIGHT TRANSPORTATION INC WELLS FARGO & CO

CHEESECAKE FACTORY INC LEGG MASON INC WERNER ENTERPRISES INC

CITY NATIONAL CORP LEUCADIA NATIONAL CORP WESTAMERICA BANCORPORATION CNO FINANCIAL GROUP INC M & T BANK CORP WILSHIRE BANCORP INC

COMERICA INC MARSH & MCLENNAN COS WINTRUST FINANCIAL CORP

CON-WAY INC MCDONALD'S CORP WORLD FUEL SERVICES CORP

CRACKER BARREL OLD CTRY STOR MORGAN STANLEY YRC WORLDWIDE INC

CSX CORP NEW YORK TIMES CO -CL A YUM BRANDS INC

CULLEN/FROST BANKERS INC NORFOLK SOUTHERN CORP ZIONS BANCORPORATION

DELTA AIR LINES INC NORTHERN TRUST CORP

DENNYS CORP OLD DOMINION FREIGHT

DIME COMMUNITY BANCSHARES PANERA BREAD CO

EAST WEST BANCORP INC PAPA JOHNS INTERNATIONAL INC

FEDEX CORP PAYCHEX INC

FIFTH THIRD BANCORP PIPER JAFFRAY COS INC

FIRST COMMONWLTH FINL CP/PA PNC FINANCIAL SVCS GROUP INC FIRST FINL BANCORP INC/OH PRICE (T. ROWE) GROUP

FIRST HORIZON NATIONAL CORP RED ROBIN GOURMET BURGERS FIRST MIDWEST BANCORP INC REGIONS FINANCIAL CORP FIRST NIAGARA FINANCIAL GRP RUBY TUESDAY INC

36 List of industries’ SIC codes

1389 1600 2011 2711 3312 3531 4011 4210 4213 4512 4513 4522 4731 5172 5812 6020 6035 6211 6282 6321 6324 6361 6411 6798 7311 7389 8011 8051 8721