Executive Compensation and Product Market Competition in the
Netherlands
Thesis
Master of Science Business Administration
Specialization: Finance
University of Groningen
Faculty of Economics and Business
Acknowledgements
My special thanks go to Peter Smid for his guidance and supervision during the preparation of this thesis, and for his fruitful comments on earlier versions of the paper. I would also like to thank Lammertjan Dam for his role as second supervisor and for his help on the methodology of the paper.
Abstract
This study empirically examines the relation between executive com-pensation and product market competition in the Netherlands. Theory diverges on this subject; some studies suggest a negative impact of com-petition on compensation, other studies suggest a positive relation. A third line of research argues that compensation contracts have an impact on competition. The results from this study are in line with research performed in the United States. I show that more competitive industries pay more executive compensation, which consists of a higher variable component. There is no empirical evidence that compensation contracts influence the extent of competition.
JEL codes: D4, G32, G34, J33, L1
Keywords: Executive compensation, Product market competition, Corpo-rate governance, Incentives, Leverage ratio
1
Introduction
Agency problems in general and (excessive) executive compensation in partic-ular remain an important topic of public debate. Executive compensation has been studied in numerous papers, both theoretical and empirical. Despite the vast amount of literature there still is no consensus on the effects of executive compensation, both in research and in public debate.
expanding the geographical scope of the previous papers. Executive compen-sation generally has three components: a fixed salary, a cash bonus based on current year’s performance, and long term compensation in shares and stock option incentives. I hypothesize that the level of variable compensation (shares, stock options and cash bonuses) increases with the intensity of competition, whereas the total level of compensation is not affected. This has implications for optimizing executive compensation contracts in different industries. Fur-thermore, it helps policy makers in their objective of constructing industry-specific regulations regarding executive compensation (the current debate on compensation in the financial sector is a relevant example).
Existing literature has pointed out two further peculiarities concerning com-petition, which lead to a number of related problems to be investigated in this paper. Both compensation and competition may be related to leverage or, in other words, leverage is a strategic tool in oligopolies. This leads to the hypothesis that the impact of leverage on compensation depends on the intensity of competition. Furthermore, competition may be endogenous, depending on compensation contracts in that market. The hypothesis is that competition is not exogenously determined and in turn depends on compensation contracts. The problem should be treated as a system of equations, where the interrelation is modeled.
This study uses a compact database of the University of Groningen’s Corporate Governance Insights Centre (CGIC) with remuneration data of the 98 largest firms in the Netherlands from 2003 to 2006. Most studies rely on a measure of industry concentration as proxy for competition. Besides the Herfindahl-Hirschman Index (a concentration index), I use the size of market demand and the degree of product differentiation as competition proxies. I employ linear regression analysis, with random panel effects to test for the hypotheses defined above. Furthermore, a two-stage analysis is performed to overcome endogeneity problems.
in Section 4. Section 5 presents the main results and Section 6 provides some robustness checks. Section 7 concludes.
2
Literature Review
The separation of ownership and control (Berle and Means, 1932) is a cen-tral feature of agency costs between executives and shareholders. Many factors influence the extent of agency problems. A large portion of the literature is concerned with aligning the interests of executives and shareholders or, in other words: providing executives with incentives in such a way that they will act in the shareholders’ best interest. Ultimately, the interests of shareholders are defined in terms of risk and return, which are affected by more factors than just by executive behavior. Factors such as market structure and the stage of the business cycle are (mostly) determined outside the firm, and they have a major impact on a firm’s results. However, these factors do influence the way in which an executive should operate. Executive behavior has to be aligned to the competitive environment (e.g. Hermalin, 1992). The importance of product market competition has been identified by numerous theoretical studies (e.g. Hart, 1983; Nalebuff and Stiglitz, 1983; Raith, 2003). However, empirical evi-dence on this subject is scarce, recent, and ambiguous. After a brief introduction in agency costs (Section 2.1) and product market competition (Section 2.2), an overview of theoretical (Section 2.3) and empirical literature (Section 2.4) will be provided.
2.1
Agency costs
Two of the building blocks of agency theory that are of interest here are mon-itoring and remuneration. In order to mitigate agency costs, a principal will try to monitor his agent as to limit behavior that does not maximize the principal’s welfare. Perfect (costless) monitoring of the agent is not possible, since monitoring requires time and effort. The benefits and costs of monitoring depend on individual firm characteristics as well as on the market structure. For a large group of small shareholders, monitoring is a ‘public good’: it is beneficial for the entire group to have a good monitoring mechanism, but for each shareholder it is better not to individually invest in monitoring (i.e. small shareholders will be free-riding). A large institutional investor does have an incentive to monitor the agent. Thus, having a large institutional owner will decrease the free-rider problem and increase monitoring activity. This argument also holds for a firm with a larger amount of debt holders; creditors on average monitor a firm more closely (Stiglitz, 1985).
The power of monitoring is also affected by market structure. In markets with at least one competitor it is possible to monitor outputs using benchmarking. Benchmarking is the use of information on rival firm performance to evaluate one’s own performance. A firm’s performance relative to its competitors stores information about the performance of the executive. Benchmarking can be used as a tool for monitoring executives in a competitive environment. Furthermore, as demand functions and exogenous shocks between firms become more identical, benchmarking can further reduce informational asymmetries and agency costs (Holmstr¨om, 1982). More competitive industries thus allow for better bench-marking and, in effect, better monitoring.1
Monitoring alleviates the agency problem, but it does not completely solve it. Another often practiced tool is the use of monetary incentives. Remuneration is arguably the most heavily debated and studied part of agency literature. The main argument in the executive compensation literature is that monetary incentives (and in some cases perquisites) can be used to align the interests of the executive and shareholders and thus alleviate the costs of the principal-agent
1Consider the extreme example of a market where all characteristics are equal across
problems created by information asymmetries (e.g. Holmstr¨om, 1979; Tirole, 2006).
Holmstr¨om (1979) also notes that a higher degree of information asymmetry induces a higher level of incentive payments. Basically, an executive is rewarded for good performance and fired in case of bad performance. The choices he makes focus on maximizing his wealth (i.e. the expected value of his income and non-monetary perks) without losing his job (i.e. minimizing risk). More intense competition decreases informational asymmetry (Hart, 1983), which increases the managerial risk of losing his job. The increased risk will decrease the opti-mal level of incentive compensation. The literature on executive incentives and executive compensation is extensive (for an extensive overview of the executive compensation literature, see Hallock and Murphy, 1999; Murphy, 1999).
2.2
Competition
Competition is a complex construct, which can be measured in multiple ways. Starting with some industrial organization basics, a monopolist is the sole firm supplying a certain market (in other words, no other firm produces a product that can be used as a substitute). At the other end of the spectrum – in a market with perfect competition – all firms produce a completely homogeneous product and have the same (cost) characteristics. The monopolist can earn maximum excess profits (by setting marginal revenues equal to marginal costs) due to the lack of available substitute products, whereas firms in a competitive market cannot earn any excess profits due to the availability of perfect substitutes. A decrease of the degree of product differentiation (i.e. a decrease in the possibility of earning excess profits) therefore is a commonly used indicator of increasing competition intensity.
Another factor in competition is the existence of strategic interactions in markets characterized by oligopoly. The literature often distinguishes between industries with strategic substitutes (e.g. a differentiated Cournot competition) and indus-tries with strategic complements (e.g. a differentiated Bertrand competition). The strategic interactions in an industry depend on the type of competition. If a competitor decreases his level of output in order to increase the selling price, the best reaction is to increase your firm’s output (and decrease the selling price) in an industry with strategic substitutes. An industry with strategic complements (where firms set prices instead of quantities), the reaction to a rival firm price increase should be to raise your own firm’s price as well. These interactions do not influence the intensity of competition, but they might change the nature of compensation contracts (Aggarwal and Samwick, 1999, a more detailed explanation follows in Section 2.3).
Factors that do influence (and measure) the intensity of competition are the degree of product substitutability (which is explained at the beginning of this section), the concentration of market supply, and the size of market demand (see e.g. Raith, 2003; Karuna, 2007, who also study the relation between competition and compensation). In the industrial organization literature more sophisticated competition estimates are being used, like the Panzar-Rosse and the Hall ap-proaches (for an overview, see Hyde and Perloff, 1995). Although they can estimate the level of competition more accurately, these techniques have the disadvantage that they do not allow for cross-industry comparison.
Concentration is often employed as a measure for product market competition. In more concentrated industries, market supply is concentrated around less firms. The intuition behind concentration is that if less firms supply most of the market, they have more market power and the industry is less competitive. The concentration ratio is a crude measure of seller concentration; it sums the market shares of the largest m (often three to eight) firms of the industry.2The
disadvantage of the concentration ratio is that it focuses on the largest firms; it completely ignores the dynamics in the lower tail of the market. Another commonly used concentration measure is the Herfindahl-Hirschman Index (HHI henceforth), which equals the sum of all squared market shares and therefore
2The i th firm market share is defined as S
does not merely focus on the largest firms (see Curry and George, 1983; Geroski, 1983; Tirole, 1988, for further reading, and also for an overview of more exotic alternative concentration measures).
An important note by Sutton (1991) is that the relation between competition and concentration does not need to be linear if the number of firms is endogenous (i.e. entry and exit is possible). Intense price competition between incumbents can serve as an entry barrier, leading to a low level of new entrants, hence leading to a concentrated market with fierce competition. ‘The intuition underlying this result is simply that the anticipation of a tougher competitive regime makes entry less attractive, thus raising equilibrium concentration levels.’ (Sutton, 1991, pp. 9).
The size of market demand (measured, for instance, by net sales) also provides information on the intensity of competition. Competition is more intense in large markets, since larger market demand will attract more entrants. Furthermore, incumbent firms will try harder to increase their market share, since this will lead to a larger reward (i.e. sales) than in a small market (Raith, 2003).
In order to analyze competition, it is necessary to define ‘industry’, since compe-tition is always defined within a certain industry. Most empirical studies use an industry classification such as the Standard Industrial Classification (SIC) or Statistical Classification of Economic Activities in the European Community (NACE) as an industry framework. There is no one-to-one correspondence between the frameworks and the actual product market, but it is the best quantitative approximation there is (Aggarwal and Samwick, 1999). Industry classifications in this area of research are usually defined at the two, three, or four-digit level (e.g. Aggarwal and Samwick, 1999; DeFond and Park, 1999; Falato and Kadyrzhanova, 2007; Graboyes, 2010; Karuna, 2007; Giroud and Mueller, 2010).
and the US. The Netherlands arguably is too small to be considered an isolated geographical market (this argument follows from Aghion et al., 2005, who find the same problem for their sample of UK firms).
Competition is measured in endless ways. Based on the pros and cons of the literature discussed above, relying on a single measure of competition may be flawed (Karuna, 2007). In cross-industry examinations, only the competition proxies that can be cross-sectionally compared can be used, like the size of market demand, the degree of product differentiation, and an index of concen-tration. Furthermore the literature made it clear that the definition of a market is problematic in size and scope. Industry classifications alleviate the problem, and it is important to have a geographical market scope that is large enough.
2.3
Compensation and product market structure
Product market competition influences the extent of agency problems, mainly through its effect on information asymmetry and on firm-level profits. However, there may also be a causal link the other way around: incentive compensation might influence the degree of competition. Interestingly, theories about the relation between compensation and competition diverge. I will give a concise overview of conflicting theories in this area.
which gives him the incentive to work hard. More intense competition provides better benchmarks and would thus decrease (the need for) monetary incentives.
The latter argument is also supported by the income effect argument of Herma-lin (1992): since competition lowers the firm’s expected profits, it also lowers an executive’s expected profit (like the ’profit destruction’ effect of Tirole, 2006). An executive will not be able to extract as much agency rents when the industry is more competitive. This may be due to a decrease in room-for-error when competition is fierce (similar to the benchmarking effect). To put it in other words: an increase in competition decreases the expected value of all actions for the executive, and increases the risk (i.e. the chance of getting fired). Hermalin (1992) decomposes the effect of product market competition on executive compensation in four perspectives, and the income effect is the only perspective with an unambiguously negative effect.
The three other perspectives have an ambiguous sign with respect to compe-tition. Assume an executive has to choose an action; ‘better’ actions consume less agency goods. A change in competition can change the relative riskiness of implementing different actions (a risk-adjustment effect ), and the difference in profitability between the good and the bad action (a change-in-the-relative-value-of-actions effect ), therefore changing the executive’s choice of action. The sign of these effects is ambiguous. Consider union negotiations, where the execu-tive can choose to meet the demands (which is more pleasant to the execuexecu-tive) or not to meet them (which is the ‘better’ action). We have two firms: a monopolist and a firm in a market characterized by oligopoly. Weak union negotiations may be noticed sooner in oligopoly because of benchmarking (making it more risky than in monopoly), but the executive might convince shareholders that the wage rise happens in the entire industry and therefore is inevitable (making it less risky than in monopoly). Furthermore, an oligopolist may lose more than the monopolist if higher wage costs lead to a lower market share, or lose less if the rival firms’ unions demand the same wage rise (the relative value of the action changes).
action’ (Hermalin, 1992, pp. 381).3 An executive must convince its principals that he takes good actions. Because of adverse selection, if the information asymmetry is high, an executive has to put more effort (and hence, lose more agency rent) in convincing that his actions are good. If the commitment of the executive reduces uniformly for all actions when information asymmetry declines, executive income (i.e. the consumption of agency goods) will increase. However, if the reduction of commitment is largest for better actions, he may switch to the better action; then the sign is ambiguous.
Returning to benchmarking, this concept can also be used directly in incentive contracts using relative performance evaluation (RPE), where executive pay-ments depend both on own firm performance and the performance of rival firms. RPE can be used to either increase or decrease competition. In product markets with strategic complements, executives are compensated positively on the basis of their own profits and their rival firm’s profits. This softens competition, since all executives benefit from higher rival-firm profits. Conversely, in product markets with strategic substitutes, executives are compensated positively on the basis of their own profits and negatively on the basis of their rival firm’s profits. Managers will then try to increase their own profits at the expense of their competitors, leading to more intense competition and thus lower profits for all firms. Relative performance evaluation implies that competition and compensation are interrelated; competition may not be exogenously determined (Aggarwal and Samwick, 1999).
The complex relation between compensation and competition is also proven by Spagnolo (2000). Focusing on own-firm performance, Spagnolo argues that long-term incentives (i.e. shares and stock options) will decrease competition in product markets with strategic substitutes, like the RPE discussed by Aggarwal and Samwick (1999). Long-term stock-based incentives induce tacit collusion, since present compensation is related to the expectations of future profitability. This provides rivals with a credible signal that the executive wants to maximize future profits as opposed to short-term profits. The tacit agreement will hold in equilibrium as long as the expected value of the long-term incentive payments
3Note that this is an ex ante effect of reduced information asymmetry, a reduction in
is higher than the short term bonus in case of breaching the tacit collusion.
Falato and Kadyrzhanova (2007) provide another interesting assumption: indus-try leaders have lower incentive payments than smaller firms within the indusindus-try, because industry leaders tend to have less valuable investment opportunities than smaller competitors.4For shareholders it is more important to provide the
smaller firm executive with incentives to utilize these opportunities.
Brander and Lewis (1986) develop a model which shows sequential linkages between product market output and financial structure. Like ordinary oligopoly games, the executive selects an output level given his knowledge about market demand and his own and his competitors’ cost functions. As in most of the literature, Brander and Lewis assume limited liability for the executive, as op-posed to the investors who can lose their initial investment.5The authors argue
that higher financial leverage affects the output level selected by the executive. Abstracting from market imperfections, the debt-equity (D/E ) ratio should not affect the value of the firm and is therefore not important to the investors (the capital structure irrelevance principle of Modigliani and Miller, 1958). The executive however, protected by limited liability, only benefits from operating profits when they exceed the level of debt that has to be repaid. Therefore the executive does not maximize the expected level of profit, but rather the level of residual profit (i.e. profit after interest payments). A firm with higher leverage therefore selects higher output levels to increase the expected value of residual profit. This argument is closely related to that of asset substitution (see Jensen and Meckling, 1976), and also occurs in monopolistic and perfectly competitive environments.
In oligopolies, debt serves another purpose. As Brander and Lewis show, a higher level of debt signals the willingness to increase the firm’s output. Competitors will react by decreasing their output level if the signal is credible. Debt thus also serves a strategic purpose (in fact, this signal is similar to that of a Stackelberg leader in industrial organization literature, see e.g. Tirole, 2006).
4For example: a small Dutch-based company can increase shareholder’s value by expanding
to another country, whereas the global industry leader has no such opportunities anymore.
5In fact, in the original analysis by Brander and Lewis, the agency conflict is modeled
Theoretical literature thus learns that competition and compensation are ex-pected to be related. The various papers do not unambiguously tell that the relation is positive or negative. Furthermore, it is not clear from these papers, which of the variables is the cause of the other. Some empirical studies have been conducted on this subject, an overview is presented in the next section.
2.4
Overview of empirical studies
The theoretical literature on the relation between product market competition and executive compensation has not yet been able to provide a clear-cut model. Abovementioned papers show that the relation is ex ante ambiguous and that the variables have a two-way relation. Moreover, other factors (e.g. competitive position and leverage), may have an additional or interacting influence on the relation. In this section the current empirical literature in the area will be discussed, although this area remains under-explored by empirical research. Furthermore, I will discuss some related papers that have added value with respect to control variables.
The importance of industry factors in remuneration studies is known in empirical studies. Some studies incorporate industry dummies as control variables (e.g., see Ely, 1991; Oxelheim and Randøy, 2005). Papers that explicitly study the relation between compensation and competition are scarce (e.g. Aggarwal and Samwick, 1999; Karuna, 2007; Graboyes, 2010).
stan-dard approach in the incentive literature.6The measure of industry competition is the HHI.
Karuna (2007) relates equity-based executive incentives with competition. The main finding is that executives receive stronger incentives when product market competition is more intense. Another interesting addition to the literature is his measurement of competition. Besides the HHI, Karuna uses three measures of competition (based on the theoretical work of Raith, 2003). Competition depends on the level of product substitutability; an industry with more ho-mogeneous goods has more intense competition. Furthermore, competition is positively related with market size, since a larger market attracts more firms, ceteris paribus. High entry costs decrease the number of firms in an industry, and is therefore negatively related to competition. Product substitutability and market size are positively related to incentives, whereas a negative relation exists between incentives and entry costs.7 More intense competition thus leads to higher compensation. Executive incentives are defined as ‘the sum of changes in the value of stock and option holdings for a manager for a 1% change in the stock price’ (Karuna, 2007, pp. 280). He ignores cash bonuses, which may be relevant. Furthermore, he calculates the present value for options using the Black & Scholes method, which is not theoretically valid for employee stock options (see e.g. Hull, 2009). The recent work by Graboyes (2010) finds a non-linear relation between competition and compensation, where compensation is highest in very competitive and monopolistic markets. The author also uses the various proxies of Raith (2003) and Karuna (2007). Incentives in this paper are defined more in line with the study of Aggarwal and Samwick (1999).
Cu˜nat and Guadalupe (2009) used two deregulation events in the financial sector as natural experiments for increases in product market competition to assess the impact on executive compensation. Total pay only marginally increased with the two deregulations, whereas the proportion of variable compensation increased at the cost of fixed salary. Pay-performance sensitivity thus increased
6Note that current wealth (e.g. current company stock holdings) is not used. In
pay-performance literature this data is of interest, since executive behavior is also influenced by his current holdings. However, the relation under investigation is between the monetary incentive shareholders currently offer an executive to align their interests and the product market structure.
with competition in the financial sector. The conclusions do not necessarily hold for other industries.
Industry leaders tend to have lower pay-performance sensitivity than smaller rival firms (Falato and Kadyrzhanova, 2007, find the hypothesized effect). More-over, Albuquerque (2010) argues that large firms are more efficient than small firms, leading to differences in product market competition strategies. Therefore, compensation may vary with firm size.
Variable compensation thus seems to be correlated with competition (Karuna, 2007; Cu˜nat and Guadalupe, 2009), although the relation may not be linear (Graboyes, 2010). The empirical studies provide similar results in this respect. Cu˜nat and Guadalupe have found evidence that unlike variable compensation, the level of total compensation is not related to competition. This finding is not discussed by Karuna (2007) and Graboyes (2010). Furthermore, the studies by Karuna and Graboyes do not mention possible problems with endogeneity, although the theoretical literature suggests that an examination of this problem is necessary.8 I will try to improve the empirical knowledge in the field by
building on the methodology and the results of abovementioned studies. Next section contains a detailed explanation of the methodology of this study.
3
Methodology
The relation between the level of executive compensation and the intensity of product market competition is investigated using regression analysis. As pointed out in the previous section, this analysis is problematic because of the possibility of endogeneity. There is no clear-cut causal link between the variables, they affect each other. Endogenous relations are problematic when conducting a regression analysis. Using an exogenous shock in competition to measure its impact on compensation (or the other way around) provides more valid results (e.g. Cu˜nat and Guadalupe, 2009). Using exogenous shocks, however, poses problems with
8Cu˜nat and Guadalupe (2009) do not have this problem since they investigate an
inter-industry comparison, as I mentioned in the literature overview section.
In order to alleviate the endogeneity problem, the analysis is done in two steps. First, I will investigate the impact of competition on compensation as if competition is exogenously determined. In an additional analysis I use two-stage least squares regression.
The basic specification of the model that is being tested is
Compensationit= α0+ α1 Competitionit+ α2 Competitionit∗ D/Eit+
α3 POSITION + α4 PERFit+ α5 MVALit+
α6 MTBit+ α7 D/Eit+ α8 VOLATit+ α9AGEit+
α10 TENUREit+ α11 DUTCHit+ it (1)
where Compensation is the level of executive compensation for a given year; Competition is the level of competition in the product market; D/E is the debt-equity ratio of the firm; POSITION is the firm’s competitive position (i.e. the firm’s market share); PERF is the firm’s annual stock return; MVAL is the firm’s market value of equity; MTB is the firm’s market value of equity divided by its book value of equity; VOLAT is the volatility in stock price; AGE is the executive’s age; TENURE is the number of years the executive currently works at the firm in his current position; and DUTCH is a dummy variable that determines whether the executive has the Dutch nationality. See Section 4 and Appendix A for a more exact definition of the variables. The interaction variable Competition * D/E addresses the suggestion by Brander and Lewis (1986) that the impact of leverage on compensation depends on the intensity of competition.
Since data are collected both cross-sectional and over time, they have panel data properties. Following Karuna (2007) the model will be tested as a panel data model with random effects, clustering firms at the industry level. The companies are not drawn randomly from a population of Dutch companies, which violates one of the assumptions of the random effects model. However, in order to be able to compare the results with prior literature, the main specification will be using random effects.9 The standard errors are corrected using White’s period
standard errors and covariance, to control for potential autocorrelation and heteroskedasticity.
Endogeneity is inherent in agency theory studies, which poses a problem when interpreting the results from the ordinary least squares (OLS) regressions. To overcome potential endogeneity problems, a system of equations will be esti-mated using two-stage least squares regression (TSLS). The studies on compen-sation and competition mentioned in Section 2 do not use TSLS, but it will provide us with additional information.
Compensationit= α0+ α1 Competitionit+ α2 Competitionit∗ D/Eit+
α3 POSITION + α4 PERFit+ α5 MVALit+
α6 MTBit+ α7 D/Eit+ α8 VOLATit+ α9AGEit+
α10 TENUREit+ α11 DUTCHit+ it (2a)
Competitionit= β0+ β1Compensationit+ FIRMSit+ νit (2b)
4
Data
The sample consists of the 98 largest Dutch companies in the Netherlands. Data regarding CEO compensation over 2003 – 2006 are obtained from the University of Groningen’s Corporate Governance Insights Centre. The original dataset is composed of 107 companies, but the financial companies are excluded from the final sample because their characteristics differ too much from the other companies.10 Compensation data is subdivided into fixed salary, cash bonuses,
and long-term compensation (consistent with Aggarwal and Samwick, 1999). The variables are explained in detail in Section 4.1. Data on competition is ob-tained from Eurostat, and the HHI is calculated using firm level sales data from Orbis (Section 4.2). The control variables are defined in Section 4.3. Descriptive statistics of the variables are presented in Table 2; for the exact definitions, see Appendix A. Table 3 provides descriptive statistics on the original values of the variables that have been transformed.
All compensation variables have a minimum value of zero (for STOCK, negative values are truncated at zero). Seven observations show zero total income for the CEO, four firm-year observations have no fixed income, and 198 observations for STOCK have a value of zero. Since this number of zero-values is substantial this possibly changes the interpretation of STOCK.
The number of firms in a specific market has a wide range, from 356 to over 300,000 firms. FIRMS has a strong correlation with the competition variables, and a weak correlation with the compensation variables (see Table 1). This makes FIRMS a candidate for being instrumental variable. MTB and D/E show a wide range as well, signifying some outliers in both variables.11Below, I discuss the groups of variables in detail.
10Furthermore, AFC Ajax NV is excluded from the sample since no reliable competition
data was available.
11In nontabulated tests, controlling for outliers did not materially change the outcome of
4.1
Compensation
Data on compensation is provided by the CGIC. The dataset includes detailed compensation data of CEOs of 107 publicly traded companies with their main listing in the Netherlands, from 2002 to 2006. Compensation can be decomposed into short-term and long-term compensation. Short-term compensation consists of annual fixed salary, cash bonus, and other annual payments.12 Long-term
compensation consists of the fair value of granted shares and stock options over a given year. The fair value of stock options is calculated using an adequate version of the binomial model correcting for the characteristics of employee stock options. The empirical studies mentioned in Section 2.4 use the Black & Scholes method, which is not valid in theory (Hull, 2009), but in the CGIC sample the Black & Scholes and the binomial method are highly correlated (Van der Laan, 2008, r2 = 0.99). This allows for comparison with the literature while using
the more valid approach of valuation. Changes in CEO wealth (e.g. from stock options exercised in the past) are not used in the analysis. See also Van der Laan (2008) for definitions of all compensation variables.
Three specifications of compensation will be tested. Total compensation (TO-TAL) is regressed to obtain information on the effect of competition on the level of total compensation. In order to test for the effect of competition on incentives, long-term compensation (STOCK) and cash bonus (BONUS) are used in the model. I take the natural logarithm of the compensation variables, because of the non-normal distribution of the variables.13
4.2
Competition
It is mentioned above that competition is a complex construct, which can be approximated in multiple ways. I use market size, product differentiation, and industry concentration, and follow the measurement approach of Karuna (2007) for these variables. Market size and product differentiation are obtained from the Eurostat database, industry concentration is extracted from company
12The level of fixed salary is annualized if a CEO did not work at the company from 1
January to 31 December.
information from Orbis.
Market size (MSIZE) is measured by the natural logarithm of total industry sales. A larger market size implies more intense competition. Product differ-entiation (DIFF) is estimated as the price-cost margin, which is defined as industry level sales divided by operating costs. Operating costs are defined as sales minus gross profit. Sales (‘production value’) and gross profit (‘gross operating surplus’)are obtained from the Eurostat database.
The definitions of the variables are closely related – but not equal – to the accounting definitions. The definitions come from the European Union (1998). Production value is defined as ‘turnover, plus or minus the changes in stocks of finished products, work in progress and goods and services purchased for resale, minus the purchases of goods and services for resale, plus capitalised production, plus other operating income (excluding subsidies)’ (European Union, 1998, pp. 6); extra-ordinary and financial income is excluded. This definition of industry sales is used instead of turnover, since it is less subject to short-term demand shocks that do not represent the long-term level of competition in a market. Moreover, the incorporation of inventory levels allows this measure to show decreasing market size (and thus competition) before the actual turnover levels decline. Turnover and production value are highly correlated (r2> 0.75 for all
years).
‘Gross operating surplus is the surplus generated by operating activities after the labour factor input has been recompensed’ (European Union, 1998, pp. 11). It is composed of turnover plus the adjustments for inventory level changes minus material costs, personnel costs, and nondeductible taxes on intermediate products. Personnel costs are defined as ‘the total remuneration, in cash or in kind, payable by an enterprise to an employee in return for work done by the latter during the accounting period’ (European Union, 1998, pp. 17). This should equal the accounting definition of cost of employees for companies that use accrual accounting.
therefore I calculate a raw HHI of the EU25 using the same methodology as the EU KLEMS project (see O‘Mahony et al., 2008), using the Orbis database. The resulting raw HHI is adjusted in the same way as has been done in the EU KLEMS database
HHI =HHIraw− 1/N
1 − 1/N
to alleviate the reporting bias (i.e. not all companies are included in the database, and therefore the actual level of the HHI will be lower than the reported level).
For all companies in the sample, the NACE rev.2 code is reported at a two-digit level in the Orbis database. The HHI is calculated on the basis of the two-digit NACE rev. 2 classification.14 The Eurostat data is reported in industries based
on NACE rev. 1 classification at a two digit level. All NACE rev. 2 codes (as reported in Orbis) are converted to NACE rev. 1 codes in order to combine the Eurostat database with the CGIC company accounts.
4.3
Controls
Executive compensation is designed to link an executive’s reward to the per-formance of the firm. Therefore a positive relation between compensation and performance is expected. Firm performance (PERF) is used as a control variable to capture this variation in compensation. It is defined as the logarithmic stock return for a given year, and calculated as ln(Pt) − ln(Pt−1), where P is the
firm’s share price at 31 December. Twenty-five firm-year observations have an annual stock return of more than 100%. These outliers are deleted from the main sample; in a robustness analysis the tests are performed using the raw performance data.
Market capitalization, measured as the natural logarithm of market value of equity at 31 December of a given year, is added to the regression to control for size effects (MVAL). If executive compensation depends on firm size, it may either depend on total firm size, or on the market capitalization of the firm (i.e. the size of the equity value of the firm). I choose to control for the latter, since
14For robustness, all competition variables are also calculated at the three and four-digit
this is in line with Karuna (2007). Market capitalization is a volatile proxy, and does not reflect the total size of a firm, therefore the book value of total assets, the number of employees and net sales will be used for robustness. A related but distinct factor is a firm’s competitive position (POSITION), measured as firm sales divided by total industry sales.
The market-to-book ratio (MTB) is used as a proxy of growth opportunities. This variable is defined as the market value of equity divided by the book value of equity, calculated at 31 December of the year under investigation. To control for financial constraints, the D/E ratio at 31 December of the given year is added. Debt is expressed in book value, and equity in market value.15
Volatility of stock returns controls for the riskiness of a particular firm (VOLAT). The annualized standard deviation of daily stock returns over the current year is used.
Three controls of CEO characteristics are used. AGE; TENURE, the number of years working for the current firm as CEO; and DUTCH, a dummy that takes a value of 1 if the CEO has the Dutch nationality, and 0 if he is not. The data on CEO characteristics are obtained from the CGIC database.
5
Results
The empirical specifications are tested in this section. Presentation of the results follows the hypothesis structure. Section 5.1 provides the results on the main hypothesis, whether (variable) compensation increases with the intensity of com-petition. The results of the effect of leverage are provided in Section 5.2. Lastly, the results on the possibility of endogeneity of the intensity of competition are provided (Section 5.3).
15Book and market values of debt are approximately the same in most cases. Book values
5.1
Main results
This section covers the results of the least squares regressions, assuming that competition is exogenously determined. The results are based on Equation (1) without the interaction term Competition*D/E, and they are presented in Ta-ble 4. Since I use three proxies for competition, this leads to three testaTa-ble equations using MSIZE, DIFF, or HHI as the sole proxy for the Competition variable. The results are presented in Columns I, II, and III, respectively. The testable equation presented in Column IV includes all proxies for the Competi-tion variable in the equaCompeti-tion. Furthermore, three specificaCompeti-tions of the dependent variable are tested. Panel A provides the test results for total compensation (TOTAL), which shows that of the main exogenous variables, only DIFF signifi-cantly influences compensation. MSIZE and HHI (Column I and III) are negative but insignificant for all reported intervals. In Column II, DIFF is negative (-1.234) and significant at the 5% level. When all variables are included in the equation (Column IV), DIFF becomes -1.304 and more strongly significant (1% level). More competition (a lower level of product differentiation) thus leads to higher compensation. For a 1%-point decrease in product differentiation, total
compensation increases with EUR 7,000.16 Furthermore, PERF, MVAL, and
D/E are all positively related to compensation (at the 5%, 1%, and 1% level, respectively).
Panel B and C provide results similar to Panel A. MSIZE and HHI are still not significant at the 10% level. DIFF is still negative and significant (now at the 10% level), but the coefficient increases: -7.462 for STOCK and -5.448 for BONUS. This implies that the median executive receive EUR 6,000 more in cash bonus for a 1%-point decrease in product differentiation. A decrease in product differentiation also increases STOCK; because of the large amount of zero-values we can conclude that firms in industries with less product differentiation more often pay their executive long term incentives. When all measures of competition are included the significant negative relation between STOCK or BONUS and DIFF remains (at the 5% and 10% level), the levels of the coefficients do not
16All compensation variables in the tests are logarithmic transformations of the original
change materially.
In conclusion, results in Table 4 provide (weak) evidence that competition has a positive influence on compensation, and that the influence is larger on the level of variable compensation than on the total level of compensation. The higher coefficients for BONUS and STOCK suggest that in more competitive markets, a larger part of executive compensation is variable. Interestingly, the effect of competition is only significant for DIFF. MSIZE and HHI do not provide significant results on their own, nor do they become significant after regressing all competition measures together.
5.2
Interaction between competition and D/E
The results in Table 4 show that a companies with a higher D/E ratio pay the executive more compensation, both the level of compensation and the cash bonus are higher (for long-term variable compensation D/E is not significantly positive). A possible explanation is that on average CEOs of highly leveraged companies make better decisions. However, according to Brander and Lewis (1986), the relation between D/E ratio and decision making depends on the competitive environment (see Section 2.3). Therefore I construct another test, which includes the interaction between competition and the D/E ratio, using the complete Equation (1). The results are presented in Table 5, which has the same structure as Table 4. Column IV with the results of the equation including all proxies for competition is omitted for brevity, since it could be tested with either one of the interaction terms, and all interaction terms combined, leading to four testable equations in each panel.17
17Tests with the three competition proxies and one interaction term, or all three interaction
Table 4: Main results
This table estimates the effect of competition on executive compensation using 304 firm-year observations from 98 Dutch firms. Compensation data and AGE, TENURE, and DUTCH are obtained for the period 2003–2006 from the University of Groningen’s Corporate Governance Insights Centre, data on competition (MSIZE and DIFF) are obtained from Eurostat, HHI is derived from data from Orbis, and the other data are obtained from Datastream. Results come from least squares regression analysis with cross-sectional random panel effects. MSIZE = natural log of total industry output level. DIFF = industry output level, divided by industry level output less gross operating surplus. HHI = sum of squared market shares of all firms in the industry. (Industries are defined at the two digit level.) POSITION = firm level sales divided by total industry level output. PERF = annual stock return. MVAL = natural log of the market value of equity. MTB = market-to-book ratio of equity. D/E = debt-equity ratio. VOLAT = annualized standard deviation of daily stock returns. AGE = age of the executive in years. TENURE = number of years that the executive is currently employed in his current position. DUTCH = dummy variable, 1 if the executive is Dutch, 0 otherwise.
Independent variables
Competition measure
I II III IV
MSIZE DIFF HHI All
Panel A: Dependent variable = Total compensation
Intercept 10.911∗∗∗ 11.122∗∗∗ 10.019∗∗∗ 12.471∗∗∗ MSIZE −0.077 −0.108 DIFF −1.234∗∗ −1.304∗∗ HHI −0.468 −0.185 POSITION −7.035 −10.432 −6.079 −11.609 PERF 1.012∗∗ 0.964∗∗ 1.007∗∗ 0.968∗∗ MVAL 0.281∗∗∗ 0.309∗∗∗ 0.273∗∗∗ 0.323∗∗∗ MTB 0.055 0.050 0.054 0.050 D/E 0.112∗∗∗ 0.111∗∗∗ 0.109∗∗∗ 0.115∗∗∗ VOLAT 0.253 0.242 0.328 0.147 AGE −0.012 −0.012 −0.012 −0.013 TENURE 0.001 −0.002 0.000 0.000 DUTCH −0.142 −0.119 −0.133 −0.131 Adjusted r2 0.119 0.125 0.118 0.120
***, **, * Represent significance at the 1%, 5%, and 10% levels respectively.
Table 4 – Continued
I II III IV
MSIZE DIFF HHI All
Panel B: Dependent variable = Long-term variable compensation
Intercept 13.400 8.456 1.328 22.533∗∗ MSIZE −1.008 −1.181 DIFF −7.462∗ −8.540∗∗ HHI 9.115 10.001 POSITION 62.392 57.913 69.033 37.766 PERF 5.418∗∗∗ 5.219∗∗∗ 5.440∗∗∗ 5.164∗∗∗ MVAL 0.805∗∗ 0.871∗∗∗ 0.676∗∗ 1.046∗∗∗ MTB 0.030 0.021 0.023 0.028 D/E 0.032 0.043 0.007 0.067 VOLAT −7.178∗ −6.973∗ −6.955∗ −7.818∗∗ AGE −0.086 −0.090 −0.082 −0.082 TENURE −0.037 −0.034 −0.045 −0.030 DUTCH −0.639 −0.388 −0.503 −0.513 Adjusted r2 0.158 0.159 0.153 0.163
Panel C: Dependent variable = Cash bonus
Intercept 1.563 9.589∗ 5.181 8.885 MSIZE 0.259 0.066 DIFF −5.448∗ −5.139∗ HHI −13.093 −8.442 POSITION −23.945 −40.743 −21.071 −36.337 PERF 1.102 1.142 1.090 1.123 MVAL 0.917∗∗∗ 1.093∗∗∗ 0.959∗∗∗ 1.087∗∗∗ MTB 0.201∗∗∗ 0.186∗∗∗ 0.206∗∗∗ 0.192∗∗∗ D/E 0.501∗∗ 0.548∗∗∗ 0.515∗∗ 0.549∗∗∗ VOLAT −15.668∗∗∗ −16.505∗∗∗ −15.490∗∗∗ −16.172∗∗∗ AGE −0.134∗∗ −0.134∗∗ −0.144∗∗ −0.141∗∗ TENURE 0.074 0.072 0.081 0.075 DUTCH 0.428 0.259 0.40 0.278 Adjusted r2 0.161 0.181 0.163 0.176 Observations 304 298 304 298
The results go in the same direction as in the previous test; DIFF is negatively related to the three specifications of compensation, whereas MSIZE and HHI do not show any significant relation. The coefficient of DIFF increases in all panels with respect to the results in Table 4. Furthermore, the results become more significant for DIFF (5%, 1% and 5% level, respectively). It is worth noting that the interaction variable is not significantly related to compensation for specification I and III (the variables MSIZE and HHI were not significant in the first place). In Panel A, HHI*D/E does show a weak relation (p=0.094), but this result is not robust for Panel B and C. In Column II (with DIFF as main exogenous variable) the newly added variable is significantly positively related to compensation. The effect of leverage resulting in higher compensation is thus more pronounced in industries with less competition (i.e. more product differentiation). The control variable D/E became negative and significant for TOTAL and STOCK, making the total effect unknown.
5.3
Two stage least squares results
Table 5: Results with D/E interaction
This table estimates the effect of competition on executive compensation using 304 firm-year observations from 98 Dutch firms. Compensation data and AGE, TENURE, and DUTCH are obtained for the period 2003–2006 from the University of Groningen’s Corporate Governance Insights Centre, data on competition (MSIZE and DIFF) are obtained from Eurostat, HHI is derived from data from Orbis, and the other data are obtained from Datastream. Results come from least squares regression analysis with cross-sectional random panel effects. MSIZE = natural log of total industry output level. DIFF = industry output level, divided by industry level output less gross operating surplus. HHI = sum of squared market shares of all firms in the industry. (Industries are defined at the two digit level.) Comp*D/E = respective competition measure times D/E ratio. POSITION = firm level sales divided by total industry level output. PERF = annual stock return. MVAL = natural log of the market value of equity. MTB = market-to-book ratio of equity. D/E = debt-equity ratio. VOLAT = annualized standard deviation of daily stock returns. AGE = age of the executive in years. TENURE = number of years that the executive is currently employed in his current position. DUTCH = dummy variable, 1 if the executive is Dutch, 0 otherwise.
Independent variables
Competition measure
I II III
MSIZE DIFF HHI
Panel A: Dependent variable = Total compensation
Intercept 11.358∗∗∗ 11.641∗∗∗ 10.085∗∗∗ MSIZE −0.108 DIFF −1.702∗∗ HHI 1.662 Comp*D/E 0.023 0.356∗∗∗ −2.021∗ POSITION −6.790 −11.612 −5.568 PERF 1.021∗∗ 0.952∗∗ 1.037∗∗ MVAL 0.280∗∗∗ 0.315∗∗∗ 0.268∗∗∗ MTB 0.056 0.049 0.057 D/E −0.177 −0.303∗∗ 0.191∗∗∗ VOLAT 0.168 0.293 0.217 AGE −0.012 −0.013 −0.014 TENURE 0.001 −0.002 −0.001 DUTCH −0.137 −0.123 −0.112 Adjusted r2 0.116 0.124 0.117
***, **, * Represent significance at the 1%, 5%, and 10% levels respectively.
Table 5 – Continued
I II III
MSIZE DIFF HHI
Panel B: Dependent variable = Long-term variable compensation
Intercept 15.423 14.454∗∗∗ 1.158 MSIZE −1.162 DIFF −13.372∗∗∗ HHI 17.703 Comp*D/E 0.116 3.757∗∗∗ −7.470 POSITION 62.497 40.429 68.151 PERF 5.477∗∗∗ 5.268∗∗∗ 5.620∗∗∗ MVAL 0.810∗∗ 0.969∗∗∗ 0.682∗∗ MTB 0.031 0.025 0.025 D/E −1.463 −4.269∗∗∗ 0.306 VOLAT −7.367∗ −6.497∗ −7.231∗ AGE −0.088 −0.095 −0.089 TENURE −0.037 −0.036 −0.046 DUTCH −0.619 −0.488 −0.420 Adjusted r2 0.156 0.170 0.153
Panel C: Dependent variable = Cash bonus
Intercept 6.161 12.034∗∗ 5.224 MSIZE −0.074 DIFF −7.731∗∗ HHI −8.023 Comp*D/E 0.251 1.512∗∗ −4.473 POSITION −22.453 −46.658 −20.413 PERF 1.222 1.149 1.200 MVAL 0.915∗∗∗ 1.125∗∗∗ 0.953∗∗∗ MTB 0.206∗∗∗ 0.182∗∗∗ 0.207∗∗∗ D/E −2.720 −1.196 0.694∗∗ VOLAT −16.273∗∗∗ −16.357∗∗∗ −15.691∗∗∗
***, **, * Represent significance at the 1%, 5%, and 10% levels respectively.
Table 5 – Continued
I II III
MSIZE DIFF HHI
AGE −0.140∗∗ −0.136∗∗ −0.148∗∗
TENURE 0.073 0.070 0.079
DUTCH 0.474 0.224 0.439
Adjusted r2 0.162 0.183 0.162
Observations 304 298 304
***, **, * Represent significance at the 1%, 5%, and 10% levels respectively.
The small correlation between the competition variables and the residuals suggests that competition may not be closely interrelated with compensation. This is an indication that the theoretical argument – which is persistent – does not hold up in practice. The next step to check for endogeneity is to perform a two-stage regression, the results of which are presented in Table B.1 in the Appendix. Again, DIFF is negatively related to compensation, whereas DIFF*D/E is positively related to compensation (although the results are only significant for STOCK as dependent variable). Comparing the results from Table 5 with those from Table B.1, the signs of most variables remain same. The independent variables no not appear to be significant, which adds to the initial finding that TSLS is not appropriate and therefore there is no two-way relation between compensation and competition.
6
Robustness checks
186.1
Panel effects
When fixed effects are introduced in the original analysis most of the significant results disappear. Nontabulated regressions reveal that in most specifications the competition variables lose their significance, as do the control variables. This may be because the data is not suitable for a fixed effects model. The level of competition variables varies between companies (the cross-sections), but most are nearly time-invariant (e.g. product differentiation does not double in value in one year). In the process 18Results of the robustness analyses are not tabulated for brevity; tables are available upon
of demeaning the competition variables per cross-section most of the cross-sectional difference will be lost.
Following Karuna (2007), the tests are also carried out disregarding the panel proper-ties altogether, using pooled least squares analysis. The results do not materially differ from the random effects model: the significant variables in the original analysis remain significant; the signs of the significant variables do not change; and the explanatory
power improves a little (adjusted r2 increases 0 to 0.056 with respect to the random
effects model). Disregarding panel properties altogether does not change the results presented in the previous section. Including a firm specific intercept and demeaning the variables makes all main explanatory variables insignificant.
6.2
Industry classification
6.3
Other robustness checks
The literature suggests various other robustness checks; they will be discussed in this section. Furthermore, some of the control variables are replaced with other proxies, the results of which are also provided in this section. In a first step, I construct a new variable, which is (the natural logarithm of) the sum of the cash bonus and long-term incentive payments. Using total variable compensation (VARIABLE) as dependent variable provides the same results as those in Section 5 (both with and without interaction term). The coefficient of DIFF is between the values of the tests with STOCK and BONUS separately.
The second step is to construct another new variable (INCENTIVE), representing
the value of variable compensation divided by the value of total compensation.19
This variable replaces Compensation in Equation (1); the results show that variable compensation as a share of total compensation increases as product differentiation decreases (-0.851), and the interacting variable DIFF*D/E has a positive influence on INCENTIVE (0.211, both variables are significant at the 1% level).
Graboyes (2010) found a nonlinear relation between competition and compensation. I test for this relation by adding the squared version of the competition variable to Equation (1). This test provides neither a significant relation to the linear competition variables, nor to the squared transformations.
Firm size can be determined in multiple ways; in the main tests MVAL was used fol-lowing previous studies (e.g. Karuna, 2007). However, market value of equity depends is an incomplete measure of a firm’s size, since this measure incorporates investor’s expectations of the company and does not adequately represent the complexity of a firm. The test results are robust however, for (the natural logarithm of) assets, number of employees, and net sales as proxy for firm size. The main test results are also unchanged when PERF is replaced with the raw version of firm performance.
7
Conclusion
Consistent with the hypothesis and prior empirical results, executive compensation is positively related to the intensity of competition. More specifically, total compensation does increase with competition, but the most pronounced increase is in long-term and short-term variable compensation. Consequently, compensation depends more on the variable component as competition increases. This is consistent with the theoretical argument that executive effort is more valuable in competitive environments. I also find that in a more competitive market less financial leverage is rewarded, whereas in markets with less competition having more leverage is rewarded. The result is not in line with the hypothesis from Brander and Lewis (1986), who suggest that leverage is in fact a strategic tool to gain market share from competitors. The results may be explained by the change-in-information effect (Hermalin, 1992). Whenever financing is needed, an executive has to decide whether to use debt or equity. In more competitive (and more informative) markets, it is easier to convince shareholders to provide the cash: the adverse selection problem is less pronounced. Most good projects (and little or no bad projects) can be financed with equity. In markets with little or no competition obtaining equity for good investments is harder; managers will therefore more often select debt financing, also for good projects. Executives that use debt financing in competitive environments (or equity financing in less competitive markets) may therefore thus select more bad projects, leading to less compensation.
The hypothesis that competition is also driven by compensation is not supported by my data. There is little evidence that competition is not exogenous, and simultaneous tests on the impact of compensation on competition and vice versa do not provide additional information.
The results of this study are in line with previous empirical literature and contribute to the literature in a number of ways. The findings in this study are obtained in the Netherlands, proving that the relation between compensation and competition holds outside the United States. The interaction of leverage and competition on compensa-tion reveals to be an important factor in the compensacompensa-tion and competicompensa-tion literature, which has been previously omitted in empirical studies. Furthermore, this study has empirically disproved the theoretical argument that competition is not exogenously determined and depends on compensation contracts.
degree of product substitutability is the only proxy that provides significant results, the size of market demand and the Herfindahl-Hirschman Index (HHI, a ratio of industry concentration) do not prove to be significantly related to compensation. The lack of robust results may be attributed to the fact that market size and HHI do not serve as good proxies for competition, or to the relative small dataset used in this study. The dataset at the author’s disposal is limited in time, number of firms, and geographical scope. A broader set of data might improve the robustness of the results from this study, leading to a further understanding of compensation contracts in competitive environments.
More specifically, the effect of leverage both on competition and compensation needs further research. This study demonstrates that having less financial leverage is re-warded in more competitive markets. The implications of this finding can be checked by testing whether the leverage ratio depends on the intensity of competition, and whether companies with a lower leverage ratio outperform their peers in competitive markets.
References
Aggarwal, R. and A. Samwick, 1999. Executive Compensation, Strategic Competition,
and Relative Performance Evaluation: Theory and Evidence. The Journal of
Finance, 54(6):1999–2043.
Aghion, P., N. Bloom, R. Blundell, R. Griffith, and P. Howitt, 2005. Competition and Innovation: an Inverted-U Relationship. The Quarterly Journal of Economics, 120(2):701–728.
Albuquerque, A., 2010. Does Firm Heterogeneity lead to Differences in Relative
Executive Compensation? Finance Research Letters, 7(2):80 – 85.
Berle, A. A. and G. C. Means, 1932. The Modern Corporation and Private Property. Harcourt, Brace & World.
Brander, J. A. and T. R. Lewis, 1986. Oligopoly and Financial Structure: The Limited Liability Effect. The American Economic Review, 76(5):pp. 956–970.
Cu˜nat, V. and M. Guadalupe, 2009. Executive Compensation and Competition in the
Banking and Financial Sectors. Journal of Banking & Finance, 33(3):495 – 504. Curry, B. and K. D. George, 1983. Industrial Concentration: A Survey. The Journal
DeFond, M. L. and C. W. Park, 1999. The effect of competition on CEO turnover. Journal of Accounting and Economics, 27(1):35 – 56.
Ely, K. M., 1991. Interindustry Differences in the Relation between Compensation and Firm Performance Variables. Journal of Accounting Research, 29(1):pp. 37–58.
European Union, E. C., 1998. Concerning the definitions of characteristics for
structural business statistics. Commission Regulation (EC) No 2700/98, of 17
December 1998.
Falato, A. and D. Kadyrzhanova, 2007. Optimal CEO Incentives and Industry
Dynamics. Working Paper (Federal Reserve Board and University of Maryland). Geroski, P. A., 1983. Some Reflections on the Theory and Application of Concentration
Indices. International Journal of Industrial Organization, 1(1):79 – 94.
Giroud, X. and H. Mueller, 2010. Does Corporate Governance Matter in Competitive Industries? Journal of Financial Economics, 95(3):312–331.
Graboyes, E., 2010. What is the Effect of Product-Market Competition on Chief
Executive Officer Compensation? Honors Projects. Paper 27 (Macalester College). Hallock, K. F. and K. J. Murphy, 1999. The Economics of Executive Compensation,
volume I, II. Edward Elgar Publishing Limited.
Hart, O. D., 1983. The Market Mechanism as an Incentive Scheme. The Bell Journal of Economics, 14(2):pp. 366–382.
Hermalin, B. E., 1992. The Effects of Competition on Executive Behavior. The RAND Journal of Economics, 23(3):pp. 350–365.
Holmstr¨om, B., 1979. Moral Hazard and Observability. The Bell Journal of Economics,
10(1):pp. 74–91.
Holmstr¨om, B., 1982. Moral Hazard in Teams. The Bell Journal of Economics,
13(2):pp. 324–340.
Hull, J. C., 2009. Options, Futures, and Other Derivatives. Pearson/Prentice Hall, 7 edition.
Hyde, C. E. and J. M. Perloff, 1995. Can market power be estimated? Review of Industrial Organization, 10:465–485.
Karuna, C., 2007. Industry Product Market Competition and Managerial Incentives. Journal of Accounting and Economics, 43(2-3):275 – 297.
Modigliani, F. and M. H. Miller, 1958. The Cost of Capital, Corporation Finance and the Theory of Investment. American Economic Review, 48(3):261–297.
Murphy, K. J., 1999. Executive Compensation, volume 3, chapter 38 in: O. Ashenfelter and D. Card (ed.), ”Handbook of Labor Economics”, pages 2485–2563. Elsevier, 1 edition.
Nalebuff, B. J. and J. E. Stiglitz, 1983. Prizes and Incentives: Towards a General
Theory of Compensation and Competition. The Bell Journal of Economics,
14(1):pp. 21–43.
O‘Mahony, M., C. Castaldi, B. Los, E. Bartelsman, Y. Maimaiti, and
F. Peng, 2008. EU KLEMS - Linked Data: Sources and Methods.
(http://www.euklems.net/data/linked/EUKLEMS LinkedData SourcesAndMethods 220708.pdf). Oxelheim, L. and T. Randøy, 2005. The Anglo-American Financial Influence on CEO
Compensation in Non-Anglo-American Firms. Journal of International Business Studies, 36(4):pp. 470–483.
Raith, M., 2003. Competition, Risk, and Managerial Incentives. The American
Economic Review, 93(4):pp. 1425–1436.
Spagnolo, G., 2000. Stock-Related Compensation and Product-Market Competition. The RAND Journal of Economics, 31(1):pp. 22–42.
Stiglitz, J. E., 1985. Credit Markets and the Control of Capital. Journal of Money, Credit and Banking, 17(2):pp. 133–152.
Sutton, J., 1991. Sunk Costs and Market Structure: Price Competition, Advertising, and the Evolution of Concentration. MIT Press Books. The MIT Press.
Tirole, J., 1988. The Theory of Industrial Organization. The MIT Press, Cambridge, MA.
Tirole, J., 2006. The Theory of Corporate Finance. Princeton University Press. Van der Laan, G., 2008. Behavioral Corporate Governance: Four Empirical Studies.
Appendix A
Data
Table A.1: Definition of variables and their source
Definition of all variables used in the models. The column Variable denotes the names of the variables as used in the model specifications and in the output tables. An explanation and exact definition of the variables is given in the middle column. The source of each variable is denoted in the right column. CGIC is the University of Groningen’s Corporate Governance Insights Centre.
Variable Explanation Database
FIXED fixed salary in thousands of euros CGIC
Annualized if executive does not work the entire year. Not used in the analysis
BONUS cash bonuses in thousands of euros CGIC
The logarithmic transformation of the original variable plus 1 is used in the tests.
STOCK long-term compensation in thousands of euros CGIC
The sum of stock option grants and share grants in the current year. The logarithmic transformation of the original variable plus 1 is used in the tests.
TOTAL sum of fixed salary, cash bonus and long-term
variable compensation
CGIC
The logarithmic transformation of the original variable plus 1 is used in the tests.
VARIABLE sum of cash bonus and long-term variable
compensation
CGIC
The logarithmic transformation of the original variable plus 1 is used in the tests.
INCENTIVE part of total compensation that is variable CGIC
Cash bonus plus long-term variable compensation divided by total compensation
AGE age of the CEO CGIC
TENURE number of years the CEO is in his current position
at the firm
CGIC
DUTCH dummy variable CGIC
Variable equals 1 if the CEO has the Dutch nationality, and zero otherwise
Table A.1 – Continued
Variable Explanation Database
MSIZE natural logarithm of the market size Eurostat
Market size is defined as the industry level production value. Production value is defined as year t sales plus changes in inventory
DIFF extent of product differentiation Eurostat
Industry level production value, divided by production value minus gross operating surplus.
HHI adjusted herfindahl index Orbis
Herfindahl index is the sum of squared market shares of all companies in an industry in year t. Market share is defined as a firm’s year t turnover as reported in its annual report. The Herfindahl index is adjusted as follows: HHI = HHIraw−1/N
1−1/N .
POSITION competitive position of the firm Orbis
SALES divided by MSIZE Eurostat
FIRMS number of firms in industry Orbis
Total number of firms in included in the Orbis database within an industry classification
PERF year t stock return Datastream
logarithmic return, obtained by ln(Pt) − ln(Pt−1) ,
where P is the closing share price at 31 December.
MVAL market value of equity Datastream
Natural logarithm of the market value of equity at 31 December
ASSETS total assets Datastream
Natural logarithm the book value of total assets at 31 December
EMPL number of employees Datastream
Natural logarithm of the number of employees at 31 December
SALES net sales Datastream
Natural logarithm of net sales. Net sales is calculated from 31 December of t -1 to 31 December of t
MTB market-to-book ratio Datastream
Market value of equity divided by the book value of equity, both measured at 31 December
Table A.1 – Continued
Variable Explanation Database
D/E debt/equity ratio Datastream
Book value of total liabilities divided by the market value of equity, both measured at 31 December
VOLAT volatility Datastream
Appendix B
Results TSLS
Table B.1: Two-stage least squares results
This table estimates the effect of competition on executive compensation using 304 firm-year observations from 98 Dutch firms. Compensation data and AGE, TENURE, and DUTCH are obtained for the period 2003–2006 from the University of Groningen’s Corporate Governance Insights Centre, data on competition (MSIZE and DIFF) are obtained from Eurostat, HHI is derived from data from Orbis, and the other data are obtained from Datastream. Results come from two-stage least squares regression analysis with cross-sectional random panel effects. FIRMS is used as instrumental variable.
MSIZE = natural log of total industry output level. DIFF = industry output level, divided by industry level output less gross operating surplus. HHI = sum of squared market shares of all firms in the industry. (Industries are defined at the two digit level.) Comp*D/E = respective competition measure times D/E ratio. POSITION = firm level sales divided by total industry level output. PERF = annual stock return. MVAL = natural log of the market value of equity. MTB = market-to-book ratio of equity. D/E = debt-equity ratio. VOLAT = annualized standard deviation of daily stock returns. AGE = age of the executive in years. TENURE = number of years that the executive is currently employed in his current position. DUTCH = dummy variable, 1 if the executive is Dutch, 0 otherwise. FIRMS = number of firms in the industry.
Independent variables
Competition measure
I II III
MSIZE DIFF HHI
Panel A: Dependent variable = Total compensation
Table B.1 – Continued Independent
variables
Competition measure
I II III
MSIZE DIFF HHI
TENURE 0.003 −0.003 −0.002
DUTCH −0.141 −0.128 −0.039
Adjusted r2 0.114 0.121 0.090
Panel B: Dependent variable = Long-term variable compensation
Intercept 46.390 34.758∗ −2.862 MSIZE −3.776∗ DIFF −34.714∗ HHI 100.541∗ Comp*D/E 0.462 8.733∗∗ −20.382 POSITION 38.965 −18.567 31.991 PERF 5.612∗∗∗ 5.074∗∗∗ 5.945∗∗∗ MVAL 1.094∗∗ 1.446∗∗∗ 0.729∗∗ MTB 0.055 0.016 −0.005 D/E −5.864 −9.929∗∗ 0.794 VOLAT −8.884∗∗ −6.104 −10.259∗∗ AGE −0.091 −0.101 −0.059 TENURE −0.025 −0.044 −0.078 DUTCH −0.862 −0.760 −0.034 Adjusted r2 0.128 0.005 0.095
Panel C: Dependent variable = Cash bonus
