1.1 – Metrics for large shareholdings

20The literature on large shareholdings has so far provided little guidance for choosing metrics capable of assessing effective corporate control and gauging the intensity of management monitoring. Table 1 supports this statement by providing a survey of the metrics found in the literature dealing with the effects of ownership. Survey A in the table specifically provides a review of large shareholding measurements found in some seminal and reference studies. At first glance, there is a broad and heterogeneous range of alternative metrics. The first major option is to focus either on ownership concentration and the existence of blockholders (Survey A1 – blockholders defined as shareholders owning more than five percent of equity), or on the controlling interest held by the largest or ultimate largest shareholder (Surveys A2 and A3). [2]

21An alternative lies in the measure of the degree of control as displayed in the second column of the table. Among the set of studies focused on ownership concentration and blockholders (Survey A1), two early studies rely on the weight and concentration of the top five shareholders (Demsetz and Lehn, 1985; Dyl, 1988). However, most authors measuring ownership concentration first identify the companies that have at least one blockholder, then they choose either a discontinuous metric (an indicator variable takes value one when at least one blockholder exists, and zero otherwise) or a continuous one (the portion of equity shares held by all blockholders). Other authors have introduced a discontinuous and dichotomous viewpoint: between “owner-controlled” companies with at least one blockholder, and “management-controlled” companies with no blockholders and hence supposedly controlled by their managers. Among the set of studies focused on the largest or ultimate largest shareholder (Surveys A2 and A3), the authors adopt similar approaches by choosing either a categorical discontinuous variable (Holderness and Sheehan, 1988) or a continuous one. Among the studies listed in Survey A, only Dyl (1988) and Hambrick and Finkelstein (1995) discuss the choice between continuous and discontinuous specifications. Dyl contends that a continuous metric is more appropriate in an “agency” context, but provides no rationale or empirical evidence to support this assertion. Conversely, Hambrick and Finkelstein assert that a discontinuous metric is more appropriate for the assessment of monitoring. They argue that as soon as a major shareholder holds a significant position an increase by a few percentage points in equity shares should not substantially increase its vigilance. The authors run a test for the respective impact of a discontinuous vs. a continuous metric (see metrics on Table 1) on CEO pay and conclude that a threshold effect in shareholdings is more relevant than a continuous relationship.

22Large shareholders are in most cases defined as blockholders – i.e. shareholders owning at least 5% of the common stock – or as shareholders crossing a threshold of 10% or 20% ownership (see column 3). However, these thresholds have no theoretical or empirical rationale, and they mostly rely on the regulatory thresholds that trigger the mandatory public reporting of ownership positions. This threshold is 5% in the regulations of the American SEC and other stock exchange authorities around the world; although in some other countries the threshold is 10%, and this explains the minimum level of ownership found in the international survey of La Porta et al. (1999). Some other studies use a threshold of 20% or 25%. For instance, Holderness and Sheehan (1988) put forward a maximum threshold of 20% to define their category of diffusely-held firms – they suggest that below this threshold shareholders would not have sufficient voting power to affect firm policies. La Porta et al. (1999) also use a 20% cutoff to define the chain of control of the ultimate largest shareholder based on a similar assumption that “this is usually enough to have an effective control of a firm”. The reported studies do not provide any theoretical or empirical support for this assumption. Hence the thresholds used in the literature regarding large shareholders are based essentially on mandatory disclosures or anecdotal evidence. To the best of my knowledge there is no study that proposes defining the appropriate level of ownership required to have effective control over a firm.

23A potential explanation for this absence of investigation could lie in the ownership structure of large American listed firms, which are perceived to be mostly diffusely-held. Because blockholders would be rare and typically holding small blocks of equity shares, a simple indicator for their existence (as in Beatty and Zajac (1994), Tosi and Gomez-Mejia (1989), and others listed in Table 1) or a simple linear relationship (as in Mehran (1995) or Konijn et al. (2011)) would be considered sufficient. However, Holderness (2009) demonstrates that the conception that the ownership pattern of U.S. companies is diffuse is essentially a “myth”. La Porta et al. (1999) and Faccio and Lang (2002) also show in their seminal studies that large shareholdings are a widespread ownership pattern all over the world. As pointed out by a number of authors (Shleifer and Vishny, 1997; Bhagat et al., 2004; Holderness, 2009), and in spite of a growing literature on large shareholders, very little is known about the forms taken by their influence, and how this can be measured.

1.2 – Continuous metrics for managerial ownership

24By contrast, the literature on managerial ownership is more insightful in its investigation of the impact of different levels of shareholdings (Survey B in Table 1) and provides relevant references for comparison with large shareholders’ levels of control. The issue of managerial ownership arise from Berle and Means’ (1932) vision of a modern corporation with no large controlling owners, leaving the effective control of the firm in the hands of the management. Later, Jensen and Meckling (1976) developed a theory according to which top managers are the agents of the equity holders and should manage the firm in the interests of the latter: this would be achieved in particular by providing the agents with equity shares. On this basis, a large number of studies investigated the amount of shareholdings a manager should acquire to have a positive impact on the market value of the equity. Morck et al. (1988) show that the firm value as measured by Tobin’s Q has a non-monotonic relationship with insider ownership: [3] they specify a piecewise (or spline) regression and find a positive impact up to 5%, then negative up to 25%, and slightly positive above 25%. [4] However, their two break-points are predetermined and are not provided endogenously in their model. The 5% and 25% points are respectively explained by the mandatory disclosure level, and by the suggestion in Weston (1979) that hostile bids cannot succeed when a shareholder holds more than 20-30% of equity. Cho (1998) uses the same piecewise linear regression with two break-points, but defines the values of the points using an “iterated search technique”: he finds a first point ranging between 7% and 10%, and a second between 34% and 38%.

25Some subsequent studies use polynomial specifications for identifying the values of change points. First, McConnel and Servaes (1990) hypothesize a curvilinear relationship between managerial ownership and firm value. Their specification takes the form of a quadratic function allowing for one turning point (i.e. the extreme value of the function) and two regimes. They find an inverted U-shaped relationship with a maximum value at around 40% to 50%. Then Short and Keasey (1999) refer to the spline regression of Morck et al. to specify a cubic function allowing for two turning points and three regimes. Finally, Davies et al. (2005) assert that a cubic specification is too restrictive because it cannot represent the more complex evolution of managerial behavior. [5] They find significant coefficients for their quintic equation, and define four turning points and five regimes. These polynomial specifications provide a better estimate of the turning points, compared to the piecewise regression of Morck et al. (1988), because the point values are defined endogenously in the model. However, they are still limited, since the number of regimes is predetermined by the number of degrees initially hypothesized in the polynomial function. Furthermore, these investigations do not discuss the possibility of threshold effects in managerial shareholdings instead of a continuous relationship with turning points.

26These studies show that the impact of managerial shareholdings on firm value is positive at low and high levels of equity shares, and negative at intermediate levels (Morck et al., 1988; Short and Keasey, 1999; Davies et al., 2005). The authors interpret these results as evidence of alignment of interests at low and high levels of ownership, and of managerial entrenchment at intermediate levels. In the context of the present study this non-monotonic effect of managerial shareholdings on firm value suggests that the monitoring behavior of controlling shareholders, and the way they design compensation packages, may vary in a similar way, depending on their degree of control.

1.3 – Controlling shareholders and CEO pay

27In the absence of a major shareholder on the board of directors CEOs may have sufficient power to influence the board and maximize their own pay. The level of pay is then solely limited by an “outrage” constraint (Bebchuk and Fried, 2003), which involves not drawing the attention of public opinion and outside minority shareholders. In this managerial power perspective risk-averse CEOs bargain for high levels of salary rather than over significant elements of variable remuneration (Harris and Raviv, 1979; Beatty and Zajac, 1994).

28Firms with poor governance and little effective monitoring have to deal with agency costs, including rent extraction, and with retention issues. Hence they need to design optimal compensation contracts that include performance-based plans. In particular, equity-based pay provides CEOs with an incentive to maximize firm value and shareholders’ wealth (Jensen and Murphy, 1990). Also, in the absence of a large shareholder engaged in monitoring and able to evaluate their CEO’s ability to reach performance targets, firms have to rely on compensation packages that are a posteriori sensitive to performance and informative of CEO’s actions (De Angelis and Grinstein, 2015). These higher shares of contingent-based pay in firms with poor governance result in higher levels of total pay, because risk-averse CEOs seek pay premiums to compensate for their bearing more risk (Beatty and Zajac, 1994; Haubrich, 1994; Hall and Murphy, 2002).

29Firms with no large shareholder and with poor governance may also be more sensitive to conditions in the labor market for CEO talent. Chaigneau and Sahuguet (2018) predict that widely held firms are more subject to retention issues, since they have less ability to select and monitor specific CEOs for their firm. An absence of engaged monitors with insight into the specific managerial skills needed for the firm leads to the selection and hiring of generalist CEOs. These are much sought after in the labor market, and they have high reservation wages (Murphy and Zábojník, 2007). Once they are appointed poorly-governed firms have to deal with retention issues, since their generalist CEOs are more likely to receive outside offers. As stated by Oyer (2004), the need to retain CEOs leads to their being rewarded not only for their impact on firm-level performance, but also for industry-level performance unrelated to their actions. Compensation packages that include a large variable component are therefore more effective in keeping levels of pay competitive because they provide more flexibility in times of good business conditions, when CEOs may receive better outside offers (Campbell and Thompson, 2015).

30In summary, firms with no large shareholders are expected to design compensation packages with higher salaries and higher performance-based components. Higher salaries can be explained by the fact that in these firms CEOs can gain more power and are able to bargain for higher fixed pay, and because they tend to have more generalist skills with higher reservation wages. Higher shares of performance-based components are a substitute for monitoring in these firms, so that the interests of a CEO can be aligned with those of outside shareholders; they also circumvent the lack of a priori knowledge about a CEO’s ability to reach performance targets; and they help retain generalist CEOs when they receive improved outside offers.

31In firms with concentrated ownership, large shareholders are expected to monitor the management directly and improve governance in the firm. Empirical studies find that large shareholders do play a monitoring role. They are able to curb the influence of powerful CEOs on the design of their own pay, and they reduce agency costs in the form of lower levels of CEO pay (Dyl, 1988; Lambert et al., 1993; Hambrick and Finkelstein, 1995; Cyert et al., 2002; Chhaochharia and Grinstein, 2009). They are also less likely to have recourse to alternative mechanisms of monitoring, such as performance-based pay (Beatty and Zajac, 1994; Lippert and Moore, 1995; Mehran, 1995; Core et al., 1999).

32The aforementioned literature implies that performance-based remuneration in firms with poor governance is a substitute for direct monitoring by large shareholders. It also implies that the interests of large shareholders are aligned with those of outside shareholders, and that they monitor their CEOs with respect to the maximization of firm value. But large shareholders may also have their own agenda and try to use their dominant position to extract private benefits (Claessens et al., 2002; Morck et al., 2005), leading to a new type of agency conflict between entrenched large shareholders and outside minority shareholders (Shleifer and Vishny, 1997). Several empirical studies show that, in the context of large shareholders that use excess voting rights to reinforce their dominant position, entrenchment translates into higher levels of CEO pay (Masulis et al., 2009) and a lower sensitivity of pay to performance (Haid and Yurtoglu, 2006; Cao et al., 2011). In this perspective CEOs are rewarded for pursuing the private benefits of entrenched shareholders, at the expense of maximizing outside shareholders’ wealth.

33In the context of the present study and the search for various potential regimes of control it is to be expected that some of these regimes are associated with the alignment and optimal-contracting hypothesis, although other regimes may be associated with the entrenchment hypothesis. For instance, large shareholders may be in a contestable position at some levels of ownership – where they fear being subject to hostile takeovers – and it is in their interest to align with other shareholders; while at other levels they would have a sufficiently protected position to enable the extraction of private benefits.

34Another way the present paper seeks to identify the manner in which large shareholders engage in monitoring is associated with the seniority of control. The greater seniority of large shareholders controlling a firm is suggestive of more experience, and has implications for monitoring. Drawing on the view that better quality monitoring allows for the better evaluation of a CEO’s inherent ability (Hermalin, 2005) and that monitors learn more about a CEO’s ability in the year following his/her hiring (Chaigneau and Sahuguet, 2018), it can be expected that large shareholders improve their knowledge of the incumbent CEO in the years after they themselves take control of the firm. If they do not have enough firm-specific experience to efficiently scrutinize a CEO’s actions they would need to rely on performance-based plans that are a posteriori informative of a CEO’s ability (De Angelis and Grinstein, 2015). Also, in the case where new controlling shareholders want to hire a new CEO, they would lack sufficient experience to select CEOs with firm-specific skills; and they would tend to hire, as in firms with diffuse shareholding, generalist CEOs with higher reservation wages and greater retention issues. In either case, incumbent or hired CEOs, it is anticipated that new controlling shareholders are associated with higher levels of pay.

2.1 – The controlling shareholder

35A preliminary step in this study concerns the appropriate identification of controlling shareholders; and so different metrics for controlling shareholding are tested. First, the broadest measure of controlling shareholdings is the percentage of equity held by all blockholders. A distinction can then be made between blockholders that are directly or indirectly represented on the board of directors, and those that are not. Second, the largest shareholder may have a specific influence, either as the largest shareholder represented on the board of directors, or as the largest external shareholder if no more significant shareholder is represented on the board. Third, the largest shareholder can increase its control with devices such as shareholder agreements (having other shareholders acting in concert) and deviations from the “one share-one vote” principle. Fourth, the largest shareholder may be a company or any type of organization that may itself ultimately be controlled or diffusely-held, with differing possible effects on its monitoring role. Most of these alternative metrics can be found in articles listed in Table 1. The literature provides no empirical or theoretical support for differentiating between these various alternatives, nor for proposing a hypothesis regarding the most relevant measure of controlling shareholdings. The following specification includes a set of measures aimed at empirically identifying and isolating large shareholding patterns that are associated with control:

36

37The subscript j is equal to {1, 2} and stands for, alternatively, cash compensation or total compensation, as defined in Table 2. The subscripts i and t respectively stand for firms (i = {1, ……, 123}) and year (t = {1, ……, 10}). The Holdings variable is subdivided into eight categories as presented in Table 2. These variables are aimed at identifying the large shareholders that effectively play a management monitoring role and providing a relevant definition of controlling shareholdings.

38Equation (1) is specified for pooled panel data, with robust errors clustered at the firm level in order to account for correlations within the firms. Panel data with fixed individual effects would not be a relevant alternative in the present study, notably because CEO compensation and ownership patterns mostly vary from one firm to another rather than within each firm. The above-specified equation aims to estimate the impact of differentiated ownership patterns on CEO compensation from one firm-year to another, and not the impact of ownership variations within each firm over the sample period, as a fixed effect model would do.

2.2 – Firm and CEO characteristics

39Firm and CEO characteristics are control variables listed and described in Table 2. According to the literature, one of the most influential determinants of CEO pay is firm size (Murphy, 1999; Cyert et al., 2002). According to the standard labor market perspective, the marginal contribution of CEO talent to firm value is greater in larger firms and justifies higher levels of pay (Terviö, 2008; Gabaix and Landier, 2008). Also as firm size grows, the complexity of the organization and the number of hierarchical levels increases and pushes top compensation upwards (Lazear and Rosen, 1981; Lambert et al., 1993). Similarly, the age of the company denotes the complexity and maturity of a firm’s organization. Leverage refers to the weight of debt in the capital structure of the firm. A higher degree of leverage is expected to reduce reliance upon elements of incentive-based pay because of the disciplinary role and agency costs of debt (John and John, 1993). Capital intensity is a proxy for measuring the asymmetry of information between CEOs and shareholders regarding growth opportunities: a high proportion of tangible assets reduces uncertainty about growth opportunities, resulting in a lower need for flexibility in compensation contracts and lower CEO incentive-based pay (Kole, 1997; Margaritis and Psillaki, 2010). Average ROA stands for profitability of the firm and is expected to have an inverse relationship with CEO pay: low profitability is associated with higher risk for the firm, which in turn increases the contingent portion of pay and the level of total compensation for risk-averse CEOs (Beatty and Zajac, 1994).

40CEO tenure is expected to have a nonlinear impact on compensation: tenure may have a positive impact on pay increases in the first years because of experience and increased CEO bargaining power, but in subsequent years long-tenured CEOs develop firm-specific human capital and lose attractiveness in the managerial labor market, which in turn reduces their bargaining power in the pay-setting process (Hambrick and Finkelstein, 1995; Cyert et al., 2002). In addition, CEOs may accumulate stock ownership over the years that reduces the need for additional equity-based pay for long-tenured CEOs (Chourou et al., 2008). This nonlinear impact is taken into account by adding the square value of CEO tenure in Equation (1). CEO age may also influence the setting of pay, but it is not included here because it results in information correlated to CEO tenure and is redundant (Croci et al., 2012).

41Hambrick and Finkelstein (1995) show that newly-nominated CEOs are paid less than their predecessors if they are promoted internally (new insider CEO); conversely new outsider CEOs are paid a premium so as to attract them, and are expected to have at least the same level of compensation as their predecessors. Lastly, other control variables include indicator variables that control for the positive impact expected from CEOs who also chair the board (Core et al., 1999; Cyert et al., 2002), for the premium provided to CEOs who are firm founders but do not hold a large share of equity, and for industry-specific effects (Gomez-Mejia et al., 2003; Cyert et al., 2002).

42Beyond the degree of control, the type of control as defined by the identity of the largest shareholder can also influence CEO compensation schemes. Two specific types of control, characterized by their diversified or passive involvement in firms (investment companies and passive families), [6] were found to be significantly associated with higher levels of compensation in an earlier study (Almeida, 2014). The effects of these two categories are then controlled for with an indicator variable in order to avoid biased analyses of the degree of control.

2.3 – Degree of control and the PTR model

43To define the different degrees of control, I use the panel threshold regression (PTR) model developed by Hansen (1999). The PTR model provides testing techniques to measure the relevant number m of thresholds that allows for m + 1 regimes to be significantly differentiated, and to determine a confidence region or interval for each point estimate (for γ₁ to γ_m). Here, the threshold variable, denoted CS, is the percentage of equity held by the controlling shareholder. The control variables included in Equation (1), namely firm and CEO characteristics, industry-specific effects and type of control, are unchanged and are denoted as X in the following equations. The Holdings variable that represented a percentage of shares in (1) is herein replaced by a set of M + 1 indicator variables that represents the distinct regimes in the degree of control. The equation takes the following form:

44

45where I{.} denotes the indicator function that takes value one for γ_m–1 ≤ CS_i,t < γ_m and zero otherwise, M is the number of thresholds tested, and the smallest and highest thresholds are set to zero and one (γ₀ = 0, γ_M+1 = 1). This discontinuous measure assumes the degree of control is homogeneous in a given regime. In other words, controlling shareholders have a fixed effect from one threshold point up to the next one, irrespective of the relative percentage of holdings within each regime.

46The first regime, denoted I{γ₀ ≤ CS_i,t < γ₁}, includes firms with no large shareholder (i.e., CS_i,t = 0%), plus firms with large shareholders below the first threshold value (i.e., 1% ≤ CS_i,t < γ₁). [7] Below this γ₁ threshold, large shareholders design compensation schemes similar to the ones offered by firms with no large shareholders. Hence firms falling within this first regime can be considered to be diffusely-held or non-controlled. The PTR specification seeks to identify this first threshold above which large shareholders are deemed to effectively monitor the management. This regime of diffusely-held or non-controlled firms also serves as a benchmark for the specific effects of the other control regimes. For this reason it will be omitted in the estimations, which is equivalent to imposing a constraint making θ_j,0 equal to zero.

47The procedure for estimating the threshold values starts with a single-threshold model (M = 1); the first step involves one iteration aimed at determining the value of γ₁ that minimizes the sum of squared errors of the following equation (where γ₂ is set to a share of 100%) :

48

49The second step involves testing the significance of the single-threshold model as compared to the null hypothesis of a zero-threshold model, based on an F-statistic. Under the null hypothesis, the sum of squared errors ignores the presence of a threshold point γ and Hansen (1996) shows that this creates a nuisance in the F-statistic; he then suggests a bootstrap procedure to approximate p-values.

50If the single-threshold model is validated, the second stage consists in taking the estimated

51These stages are repeated as long as the alternative hypothesis of M against the null of M – 1 threshold points is not rejected. Ultimately, the procedure defines the relevant number of thresholds and regimes for the sample, and the estimated values of the thresholds.

52In the final stage confidence intervals are constructed around the true values of the refined estimated points. The construction of the confidence interval for each

2.4 – Seniority of control and the PTR model

53A controlling shareholder effectively monitors the firm’s top management when it holds more than γ₁% of the equity, and the results derived from the methodology described above may define different degrees of control beyond this point. Among these control regimes new monitors may behave differently because they have not yet acquired enough firm-specific knowledge and experience in monitoring the firm. There may be a number of years of control, denoted ω, below which the controlling shareholder still relies on alternative mechanisms of control in the form of contingent remuneration, so as to mitigate their asymmetry of information with the CEO. In this case, the use of higher ratios of bonuses and other incentive-based pay would also push the levels of cash and total compensation upwards.

54The following specification assumes that there is one threshold ω of seniority below which the controlling shareholder behaves differently, all other things being equal, including a given regime for the degree of control. Other control variables, denoted X, are the same as in Equation (2).

55

56Seniority of control only concerns controlling shareholders holding more than γ₁% of the firm’s equity and coefficient τ is expected to be positive for the aforementioned reasons. The addition of coefficients τ and θ_m provides the effect of a “new” controlling shareholder within the m^th regime of control as compared to the first regime of non-controlled firms (omitted, and associated with θ₀ = 0). The indicator variable for controlling shareholders with a seniority of more than ω years – “long-term” controlling shareholders – is not included in (3), and their impact is as measured by θ_m alone.

57The threshold value ω for the seniority of control is estimated based on the PTR model (Hansen, 1996). The minimum value of the sum of squared errors is searched for by iterating on integer values of seniority from the lowest to highest deciles of seniority on the panel. The test of no thresholds against one threshold is based on the same F-statistic and bootstrap p-values as described in the previous section. Similarly, LR-statistics are eventually calculated to define confidence intervals for ω.

3.1 – Sample data

58The sample firms are selected from 180 French firms listed on Euronext Paris and members of the SBF120 Index for at least one year between 2003 and 2012. The SBF 120 index consists of the 120 companies with the highest traded stock activity at Euronext Paris. The sample is restricted to public limited companies headquartered in France [8] and to firms that have fully available data for at least four years. This produces a sample of 1,119 firm-year observations for 123 firms from 2003 to 2012. CEO characteristics and compensation data are manually collected from annual reports, and firm characteristics are extracted from Datastream. The percentages of equity held by shareholders are extracted from Thomson One Banker – Ownership database.

59The shares held by blockholders (i.e., shareholders holding more than a five-percent share) are cross-checked with ownership data provided in annual reports, and corrected when necessary. [9] Based on the information about directors disclosed in annual reports, blockholders that are directly or indirectly represented on the board of directors are identified. If no blockholder is represented on the board, I identify the largest shareholder represented on the board and owning more than one percent of equity. [10] Shareholders that are part of a shareholder agreement and act in concert with the largest shareholder are also identified. Lastly, voting rights are hand-collected when company bylaws provide for double voting rights. [11]

60As for the seniority of control, ownership databases do not provide this information. I therefore collected information about the largest shareholder represented on the board of directors from firm’s annual reports. For each largest shareholder represented on the board, seniority of control starts in the first year that a representative was nominated to the board. The seniority takes into account the various directors who may have succeeded each other as representatives of a given shareholder. In cases where a shareholder was already nominated to the board prior to becoming the largest shareholder, seniority as a non-first shareholder on the board is also taken into account, since the shareholder is considered to have acquired experience and firm-specific knowledge in those previous years. The available data from annual reports allows this information on the first year of nomination to the board to be collected back to the late 1990s. The seniorities prior to this period are obtained from the history of the company provided in annual reports or on company websites, or from other sources found in the press. In specific cases, the first year of control coincides with the founding of the firm [12] for founding shareholders, with nationalization for governmental shareholders, or with a takeover in other cases.

3.2 – Descriptive statistics

61Table 3 gives some descriptive statistics for the data, based on the variables defined in Table 2. On average, the sample firms have sales amounting to €11,764 million and a market capitalization of €9,477 million, in constant 2007 euros. The average firm has a leverage of 0.80, a capital intensity of 23.39% and a five-year average ROA of 3.01%, was first established 87 years ago, and went public 23 years ago. The average CEO has served as CEO for 10 years. He/she is a new CEO for 21% of the firm-year observations, including 12% as an insider and 9% as an outsider. He/she is also the board chairperson in 54% of firm-year observations, and is the founder of the company but not a blockholder in 4% of instances. His/her mean cash and total compensation are respectively €1.27 million and €2.10 million. The value of stock-options is based on the fair value that firms disclose in their annual reports. This disclosure has been mandatory since 2008. Before that time many firms did not disclose fair value; in such cases it is estimated by the option pricing model of Black and Scholes (1973), using the number of options granted, the exercise price and time until exercise, historical volatility and dividend yield of the company, and the annual risk-free interest rate based on government bond yields. Other stock-based pay is estimated by the value of the stock on the day it is granted.

Table 3

Descriptive statistics

Variable Mean Median Min 1st decile 9th decile Max Firm characteristics: Sales (€000) 11 764 045 2 621 648 2 133 458 232 35 900 500 167 610 992 Market capitalization (€ 000) 9 477 162 2 894 800 6 904 396 368 26 301 308 148 470 400 Leverage 0,80 0,24 -2,31 -0,08 1,52 40,30 Capital intensity 23,39% 16,58% 0,03% 2,30% 53,90% 98,32% Average ROA (over five years) 3,01% 3,29% -229,39% -1,12% 9,11% 28,99% Years since foundation 87,3 78,0 3,00 23,0 161,0 348,0 Years since listed 22,7 19,0 1,00 6,0 38,0 128,0 CEO characteristics: Tenure 9,8 6,0 1,0 1,0 24,0 47,0 New insider CEO 0,12 0,00 0,00 0,00 1,00 1,00 New outsider CEO 0,09 0,00 0,00 0,00 0,00 1,00 CEO is chairman 0,54 1,00 0,00 0,00 1,00 1,00 CEO is founder 0,04 0,00 0,00 0,00 0,00 1,00 CEO compensation: Cash compensation (€ 000) 1 274 1 136 29 362 2 431 7 478 Total compensation (€ 000) 2 100 1 518 29 391 4 063 22 811

Descriptive statistics

Blockholders: % All blockholders 43,09% 46,72% 0,00% 8,24% 72,64% 94,22% % Blockholders on BoD 36,18% 37,03% 0,00% 0,00% 70,29% 89,22% % Largest shareholder 32,41% 29,17% 1,10% 6,75% 66,25% 89,22% % Largest sh. on BoD 31,13% 28,93% 0,00% 0,00% 66,25% 89,22% % Largest sh. on BoD + Concert 33,89% 32,58% 0,00% 0,00% 68,32% 93,40% % (voting rights) Largest sh. on BoD + Concert 38,62% 39,11% 0,00% 0,00% 73,51% 93,40% Num. blockholders 2,1 2,0 0,0 1,0 4,0 7,00 Num. blockholders on BoD 1,3 1,0 0,0 0,0 2,0 7,00 Seniority of largest sh. on BoD 33,4 19,0 0,0 0,0 92,2 309,0 N. = 1119 firm-year obs.

Note: Amounts in Euros are expressed in constant 2007 Euros. Tenure is expressed in years and other CEO characteristics are indicator variables. Seniority is expressed in years.

62In the final decile of total compensation there is significant dispersion and high relative values. This can be explained by the fact that companies do not provide equity-based compensation according to any standard timetable. Some grant it on an annual basis, others from one year to another, or it is triggered by a specific event (newly nominated CEO, IPO, an exceptional operating or financial success…); other firms do not provide any stock-based pay. Hence the regressions of the following sections will be based on two-year averaged equity-based compensation in order to attenuate differing practices in making these grants. This does not however completely eliminate the very high relative values for a number of these grants. Equity-based compensation is then winsorized so that the ratio of equity-based to total compensation is trimmed to the top decile value of the sample (i.e., 56% – not reported in the table).

63In the average sample firm about 2.1 blockholders hold 43.09% of the common stock. Out of these 2.1 blockholders, 1.3 are insiders (i.e., they are represented on the board of directors) and hold an average of 36.18% of equity. The largest shareholder holds on average 32.41% of equity [13] and is in most instances represented on the board of directors, with an average equity holding of 31.13% for the largest insider shareholder. Other shareholders that concluded an agreement with the latter (i.e., those acting in concert) add an average control of 2.76%, reaching a combined interest of 33.89%. [14] Another device for increasing control involves conferring double voting rights on certain categories of shares. This gives nearly 5% of additional interest to the average largest insider shareholder, alone or in concert, who thus reaches a voting control of 38.62%. [15] The largest insider shareholder has on average 33.4 years of seniority.

64The data show a number of alternative ways of identifying which shareholders may exert effective control over the firm, and of measuring that control. Out of the 1,119 firm-year observations, 190 firm-years have no blockholders represented on the board, and 119 of these do not have large insider shareholders either (not reported). Among firms that have a large insider shareholder, some also have other insider and/or outsider blockholders, and some outsider blockholders may hold greater positions than the largest insider. Also, some of the largest insider shareholders act in coalition with other shareholders, and some have additional control with double voting rights. The following section aims to separate the respective influences of each of these various categories of shareholder by estimating their impact on CEO compensation.

4.1 – Alternative measures of controlling shares

65A number of studies provide evidence of a negative relationship between the level of CEO compensation and the existence of monitoring by large shareholders. Table 4 presents the results after running Equation (1), with cash compensation as a dependent variable in columns 1 to 3, and total compensation in columns 4 to 6. The respective first two columns (columns 1 and 2, and columns 4 and 5) look for the relevant metrics for large shareholdings, as presented in Table 2, that are associated with a significant impact on CEO pay, and hence with effective monitoring.

Table 4

Regression results for CEO compensation on alternative measures of controlling shareholdings

Ln(Cash compensation) Ln(Total compensation) (1) (2) (3) (4) (5) (6) coef p-value coef p-value coef p-value coef p-value coef p-value coef p-value Alternative measures of contr. sh.: % Largest sh. not on BoD -0,19 (0,70) 0,33 (0,56) % Other block. not on BoD 0,12 (0,72) -0,10 (0,80) % Largest sh. on BoD -0,42 (0,00) -0,69 (0,00) % Concert with largest sh. -0,63 (0,08) -0,61 (0,09) -0,93 (0,03) -0,96 (0,03) % Other block. on BoD 0,05 (0,87) 0,02 (0,95) % Largest sh. on BoD: * Ultimately controlled -0,39 (0,00) -0,71 (0,00) * Ultimately diffusely-held -0,41 (0,08) -0,72 (0,04) % Excess in voting rights -0,18 (0,66) -0,03 (0,95) % Largest sh. on B. + Concert -0,43 (0,00) -0,74 (0,00) Firm characteristics: Ln(Sales) 1,35 (0,00) 1,34 (0,00) 1,34 (0,00) 1,49 (0,00) 1,49 (0,00) 1,48 (0,00) Ln(Sales)² -0,04 (0,00) -0,04 (0,00) -0,04 (0,00) -0,05 (0,00) -0,05 (0,00) -0,05 (0,00) Ln(Market capitalization) 0,16 (0,00) 0,15 (0,00) 0,15 (0,00) 0,23 (0,00) 0,23 (0,00) 0,23 (0,00) Leverage -0,01 (0,11) -0,01 (0,10) -0,01 (0,13) -0,02 (0,07) -0,02 (0,10) -0,02 (0,12) Age of company 0,08 (0,00) 0,08 (0,00) 0,08 (0,00) 0,08 (0,01) 0,08 (0,01) 0,08 (0,01)

Regression results for CEO compensation on alternative measures of controlling shareholdings

Capital intensity -0,36 (0,01) -0,36 (0,01) -0,35 (0,01) -0,40 (0,04) -0,42 (0,03) -0,40 (0,04) Av. ROA -1,45 (0,00) -1,44 (0,00) -1,42 (0,00) -1,64 (0,00) -1,64 (0,00) -1,60 (0,00) CEO characteristics: Tenure² -0,004 (0,00) -0,004 (0,00) -0,004 (0,00) -0,001 (0,00) -0,001 (0,00) -0,001 (0,00) New insider CEO -0,09 (0,03) -0,088 (0,03) -0,09 (0,03) -0,10 (0,06) -0,10 (0,05) -0,10 (0,05) New outsider CEO 0,07 (0,21) 0,065 (0,23) 0,07 (0,22) 0,13 (0,04) 0,14 (0,03) 0,14 (0,03) I(CEO is Chairman) 0,11 (0,03) 0,11 (0,02) 0,11 (0,02) 0,09 (0,11) 0,09 (0,10) 0,09 (0,10) I(CEO is founder) 0,71 (0,00) 0,71 (0,00) 0,71 (0,00) 0,91 (0,00) 0,92 (0,00) 0,92 (0,00) Type of control: I(Investment companies) 0,22 (0,00) 0,21 (0,00) 0,22 (0,00) 0,31 (0,00) 0,31 (0,00) 0,30 (0,00) I(Passive families) 0,51 (0,00) 0,50 (0,00) 0,51 (0,00) 0,75 (0,00) 0,74 (0,00) 0,74 (0,00) Intercept -7,13 (0,00) -7,06 (0,00) -7,01 (0,00) -8,71 (0,00) -8,74 (0,00) -8,65 (0,00) Industry fixed effects Yes Yes Yes Yes Yes Yes Firm-year obs. 1119 1119 1119 1119 1119 1119 Adj. R² (%) 78,16 78,15 78,10 78,05 78,02 77,99

Note: Pooled panel regressions with standard errors clustered at the firm level, as specified in Equation (1). The dependent variables are cash and total compensation (expressed in thousands of constant euros and log-transformed) in models 1 to 3 and in models 4 to 6, respectively. Alternative measures of controlling shares are expressed as a percentage of equity shares, except if voting rights is mentioned. BoD stands for Board of directors. Control variables include firm and CEO characteristics, the type of control, and industry specific effects. All variables are as defined in Table 2. I(.) denotes the indicator function. The adjusted R² is expressed in percentages.

66Overall, the results are consistent with the previous literature, as they show a negative relationship between the levels of CEO pay and the percentage of large shareholdings.

67Among the various ways of measuring large shareholdings, only the holdings of the largest shareholder represented on the board of directors, plus the holdings of the shareholders that have concluded an agreement with him/her/it (i.e., those acting in concert with the largest shareholder), present a significant negative relationship with cash and total compensation (Table 4, columns 1 and 4). The holdings of other blockholders, whether represented on the board or not, have no impact. A largest shareholder that is not represented on the board also has no impact on the compensation scheme.

68The results then suggest that, among the various categories of shareholders, only those that can have their interests represented on the board of directors are able to monitor the CEO pay-setting process, and the effects of other large shareholders are not significant. [16]

69The largest shareholders are then subdivided between those that are ultimately controlled (including, by definition, families and the State, and companies or institutions that are themselves controlled) and those that are not. The results (Table 4, columns 2 and 5) show that both have a significant negative impact on cash compensation, although that of the former is more significant than that of the latter. This result is consistent with previous studies showing that controlling shareholders have similar effects, whether they are themselves controlled or diffusely-held (Claessens et al., 2002). The largest insider shareholders that are ultimately diffusely-held are then kept as a measure of effective control, together with those that are ultimately controlled.

70The regression in columns 2 and 5 also includes the percentage of voting rights in excess of equity rights held by the largest insider shareholder, plus those acting in concert with it. The results show no significant impact from these excess voting rights. This contradicts previous studies showing that shareholders who enhance their control through excess voting rights tend to entrench and extract private benefits (Claessens et al., 2002), notably in the form of higher managerial pay (Masulis et al., 2009). This contradictory result can however be explained by the fact that these previous studies involved dual-class shares, which yield much larger excess voting rights compared with the device of double voting rights used in the French context.

71Columns 3 and 6 present the regression with the relevant measure of the degree of control, i.e. the portion of equity held by the largest shareholder represented on the board of directors and by the shareholders acting in concert with him/her/it. The holdings of these shareholders are hereafter referred to as “controlling shareholdings”, and will serve as the measure of the degree of control in the following section.

4.2 – Firm and CEO characteristics

72As for firm characteristics, the size effect is positive and significant as measured both by sales and market capitalization (see Table 4). Unreported results show that both sales and market capitalization contribute independently to the levels of CEO compensation: sales explain the base salary component better, and market capitalization the contingent components of pay; hence both measures are retained in the base specification. The age of the firm, in relation to the size and the maturity of the firm’s business, is positively and significantly associated with the levels of CEO pay. The coefficients associated with leverage are negative, consistent with the disciplinary role of debt and the negative impact on CEO incentive-based pay found in previous studies (Masulis et al., 2009; Conyon et al., 2011); but it is not significant at the usual thresholds, except for one specification for total pay. Capital intensity has a significant negative coefficient, consistent with the prediction that in firms with a high ratio of tangible assets the top manager has less discretion and requires less monitoring through contingent pay (Kole, 1997). Average ROA is also negatively and significantly related to CEO compensation. An explanation for this relationship is that average ROA is highly negatively correlated with the standard deviation of ROA (not reported), so that low performance is associated with increased risk for the firm, and CEOs in these firms may receive a higher level of compensation to compensate for uncertain contingent pay (Cyert et al., 2002).

73Regarding CEO characteristics, the square of CEO tenure has a negative impact on the levels of pay, which is consistent with the nonlinear relationship found by Hambrick and Finkelstein (1995). During their first two years of tenure, CEOs coming from outside the firm (New outsider CEO) are paid the same as their predecessors in terms of cash compensation, but they receive higher equity-based compensation, resulting in higher total compensation; CEOs recruited from inside the firm (New insider CEO) are paid less both in terms of cash and total compensation. This is also consistent with the predictions of Hambrick and Finkelstein (1995). CEOs who also chair the boards of directors receive higher cash compensation but do not receive higher long-term incentives. CEOs who are also firm founders but do not hold large blocks of shares are paid significantly more.

74Lastly, type of control, in the form of investment companies or passive families, has a significant positive impact on CEO pay, as expected for diversified or passive largest shareholders. Industry fixed effects are also accounted for: consistent with the analysis of Demsetz and Lehn (1985), the media are among those who pay the highest amount of cash and total compensation, while regulated industries (utilities, banks and insurance companies) are among those who pay the lowest amount of cash and total compensation (not reported).

4.3 – Estimation of thresholds in the degree of control

75This section follows the PTR methodology presented earlier so that one or more thresholds in the controlling shareholdings can be identified (denoted CS). Following the results of the previous section, controlling shareholdings are defined as the equity shares held by the largest shareholder, alone or in concert, represented on the board of directors. In the first stage I look for the value of γ that best discriminates between a first class of non-controlled firms (including firms with no large shareholders on the board) and a second class of controlled firms, based on the differing effects of the two regimes on CEO pay monitoring.

76In order to have a sufficient number of observations in each regime and to avoid regimes driven by outliers, the iteration is applied with values of γ starting with the lowest decile of controlling shareholdings in the sample, incremented by 1% up to the top decile. γ can thus take all integer values from the lowest to the highest deciles of the controlling shareholding distribution, i.e. from 1% (actually above the first decile, as 119 out of 1,119 firm-years have no large shareholders on the board) to 68%. If the threshold effect is significant, the next step consists in looking for the second, third, etc. threshold, while taking the thresholds previously found as given, and as long as the m^th threshold is significant.

77Table 5 displays the results of this sequential iterative procedure. The first threshold found for cash and total compensation has an F-statistic that is much higher than the bootstrap 1% critical value (84.71 vs. 11.40 for cash compensation, and 118.95 vs. 12.12 for total compensation); the same applies for the second threshold found. Thus, the zero- and the single-threshold models are successively rejected. The existence of a third threshold point is also validated for cash and total compensation at the 1% and 5% levels of confidence, respectively (bootstrap p-values are 0.8% and 2%, respectively). Finally, the tests for a quadruple-threshold model against a triple-threshold model are rejected for both dependent variables, with bootstrap confidence levels of 97% and 21%, respectively.

Table 5

Tests for an m^th threshold given m – 1 threshold(s)

First Second Third Fourth Cash compensation Threshold estimate 46% 11% 34% 8% F-stat. for m vs. m – 1 thresholds 84,71 46,15 12,87 2,30 p-value (bootstrap) 0,00 0,00 0,008 0,97 (10%, 5%, 1% critical values) (7.13, 9.03, 11.40) (8.08, 9.29, 11.34) (8.40, 9.63, 12.75) (9.48, 10.71, 12.80) Total compensation Threshold estimate 46% 11% 34% 5% F-stat. for m vs. m – 1 thresholds 118,95 52,03 12,83 7,58 p-value (bootstrap) 0,00 0,00 0,02 21,00 (10%, 5%, 1% critical values) (6.89, 8.39, 12.12) (8.35, 10.05, 12.05) (8.48, 9.95, 14.42) (8.95, 9.84, 12.33)

Tests for an m^th threshold given m – 1 threshold(s)

Note: Looking for the relevant number of thresholds in controlling shareholdings, based on a panel threshold regression (PTR) model as developed by Hansen (1999), and based on its impact on CEO cash and total compensation. The first column shows the first threshold estimate based on a single-threshold model (see Equation (2’)) and using all possible integer values of γ between 1% and 68% (the top decile of controlling shares on the panel). The null hypothesis of a zero-threshold model against the alternative of a single-threshold model is tested, based on an F-stat. The asymptotical distribution of the F-stat requires a bootstrap procedure (Hansen, 1999), which is run with 400 iterations and leads to p-values and critical values presented in the table. The null of zero threshold is rejected. The null of a single-threshold (column 2) and of a double-threshold (column 3) are similarly rejected, both for cash and total compensation. The existence of a fourth threshold (column 4) is not accepted with p-values of 0.97 and 0.21 for cash and total compensation, respectively.

78In the single- and double-threshold models, the estimate values are respectively 46% and 11% both for cash and total compensation. These two values are used in the triple-threshold model to find the third threshold, which is 34% for both dependent variables. The first two values were estimated while ignoring the existence of a second or third threshold point. The refined estimation consists in first taking the 34% and 46% thresholds as given and looking again for the first threshold, and then taking 34% and 11% as given and looking again for the second threshold. The refined estimates confirm the 11% and 46% threshold points found in the single and double threshold models, and are displayed in Table 6.

Table 6

Threshold estimates in a triple threshold model

Cash compensation Refined threshold estimates 11% 34% 46% 95% confidence interval [10%, 16%] [31%, 42%] [45%, 46%] 99% confidence interval [9%, 17%] [27%, 42%] [45%, 48%] Total compensation Refined threshold estimates 11% 34% 46% 95% confidence interval [10%, 14%] [31%, 42%] [46%, 48%] 99% confidence interval [9%, 15%] [28%, 45%] [45%, 48%]

Threshold estimates in a triple threshold model

Note: Threshold values found in Table 5 are re-estimated (refined estimates) in the framework of a triple-threshold model: is estimated taking and as given, and so on. Refined estimates confirm the initial estimates of . Following Hansen (1999), the confidence interval of is constructed as follows: is calculated for each possible integer value of γ (i.e. [1%; 68%]), S₃ being the sum of squared errors of the triple threshold model (with and taken as given); by construction LR₁ = 0 when . The confidence interval is determined as a “no-rejection” region around the null point of LR₁, where LR-stats are below the critical values defined by Hansen (i.e. 7.35 and 10.59 for the 5% and 1% critical values, respectively). The same procedure is applied to and .

79As a test of robustness, I also run a simultaneous search for γ values in the triple threshold model, i.e. looking for the three threshold values that simultaneously minimize the sum of squared errors in Equation (2), instead of the described sequential procedure. The results provide the same three estimate points as the refinement estimates, i.e. 11%, 34%, and 46% both for cash and total compensation.

80Table 6 also displays the confidence interval for the three refined threshold estimates, based on the LR-test as specified by Hansen (1999). The confidence intervals for the 11% and 46% threshold values are quite tight, and are reassuring about the estimated value of the break points separating one regime of the degree of control from another. The confidence interval for the 34% threshold value is wider: [31%, 42%] at the 5% level of significance. The specific pattern of this regime, which presents a CEO pay design similar to the omitted regime – further described below – may account for such a relaxed confidence interval.

4.4 – Estimation of seniority thresholds

81The previous section narrows the definition of controlling shareholders to the largest shareholders on the board holding, alone or in concert, at least 11% of the equity share. The present section aims to specify effective control more precisely, using the criterion of the number of years of control. The assumption is that the largest shareholder on the board needs time to acquire firm-specific knowledge and improve the quality of the monitoring of CEO compensation.

82Among the controlling shareholders above the 11% threshold in degree of control, the lowest and highest deciles of seniority are three years and 107 years respectively. The same iteration as in the previous section is applied for each integer value in this range. For each tested value, an indicator variable takes value one when the seniority is below the tested threshold, and zero if not. It thus accounts for the specific effect of “new” controlling shareholders among the three effective-control regimes identified in the previous section.

83The threshold that yields the minimum sum of squared errors is eight years, both for cash and total compensation, as reported in Table 7. The F-statistic is considerably higher than the bootstrap-estimated 1% critical value, leading to the rejection of the null of a zero-threshold model against the alternative of a single-threshold model, both for cash and total compensation. The values of the LR-statistic lying below the critical values of 5% and 1% define a confidence interval with a minimum value of six years for both compensation measurements, and a maximum value of 14 or 16 years.

Table 7

Estimation of seniority thresholds

Cash comp. Total comp. Threshold 8 years 8 years F-stat. for 1 vs. 0 threshold 31,19 30,64 p-value (bootstrap) 0,00 0,00 (10%, 5%, 1% critical values) (7.27, 8.85, 11.26) (76.78, 8.46, 11.61) Threshold estimates 8 years 8 years 95% confidence interval [6, 14] [6, 9] 99% confidence interval [6, 14] [6, 16]

Estimation of seniority thresholds

Note: The threshold denoted ω is estimated by minimizing the sum of squared errors in a single-threshold model as specified in Equation (3); the iteration is run with a range of integer values from 3 to 107 years (the lowest and highest deciles of seniority among controlled firms). The F-stats, and their bootstrap estimated p- and critical values, lead to reject the null of a zero-threshold against the alternative of a single-threshold model, both for cash and total compensation. Then the values of the LR-stats are calculated for each integer value of ω in [3; 107]; the confidence interval for is defined as the range of ω values for which LR-stats lie below the 1% and 5% critical values defined by Hansen (1999).

84This leads to the identification of a threshold of about eight years, six at the minimum, in the seniority of a controlling shareholder, after which it significantly changes the way it monitors the CEO pay-setting process.

85The specification tests for a unique threshold for seniority of control, as there is no rationale for the existence of two or more break points in the acquisition of monitoring insight by the controlling shareholders. As a robustness check, a second point estimate and a test for a double-threshold model against a single-threshold model are also processed (not reported). This check rejects the existence of a second threshold for total compensation. For cash compensation, the alternative hypothesis of a double-threshold model is not rejected with a p-value of 3%. The point estimate is 47 years of control but is not quite precise, since the 95% confidence interval is large (from 36 to 87 years), and the 99% confidence interval cannot be defined: all seniority values have LR-statistics below the 1% critical value. This leads to a rejection of the double threshold model for seniority in the sample. However, a second threshold may be further investigated in future research, notably within a sub-sample of family firms and with respect to generational issues in family control.

4.5 – Thresholds for the degree of control

86The dummy variables for the degree of control regimes are integrated into the base regression specified in Equation (2) and displayed in columns 1 and 2 of Table 8. As the dependent variable is log-transformed, the coefficients of the dummy variables can be interpreted as approximated deviations in percentage from the omitted regime of non-controlled firms. Firms falling into the second regime of control with controlling shareholdings ranging from 11% to 34% pay about 17% less cash compensation to their CEOs. After adding the long-term incentives, the discrepancy between the first and second regimes increases with an approximately 32% lower level of total compensation in the latter.

Table 8

Regression of CEO compensation on control regimes

Ln(Cash comp.) Ln(Total comp.) Ln(Cash comp.) Ln(Total comp.) (1) (2) (3) (4) coef p-value coef p-value coef p-value coef p-value Controlling shareholders: I(Seniority of control < 8 years) 0,22 0,00 0,26 0,00 Degree of control: 1st regime: I(0% to 11%) – omitted 2nd regime: I(11% to 34%) -0,17 (0,00) -0,32 (0,00) -0,25 (0,00) -0,42 (0,00) 3rd regime: I(34% to 46%) -0,03 (0,64) -0,15 (0,05) -0,07 (0,30) -0,19 (0,02) 4th regime: I(≥ 46%) -0,31 (0,00) -0,50 (0,00) -0,32 (0,00) -0,52 (0,00) Intercept -7,16 (0,00) -8,50 (0,00) -6,81 (0,00) -8,07 (0,00) Control variables Yes Yes Yes Yes Firm-year obs. 1119 1119 1119 1119 Adj. R² (%) 78,40 78,82 79,01 79,44

Regression of CEO compensation on control regimes

Note: Pooled panel regressions as specified in Equation (2), with standard errors clustered at the firm level. The dependent variables are either cash (columns 1 and 3) or total (columns 2 and 4) compensation, expressed in thousands of constant euros and log-transformed. The regime of non-controlled firms (degree of control below 11% of shareholdings) is omitted and serves as a benchmark for the other regimes. Columns 1 and 2 include indicator variables (denoted I(.)) for the three regimes of controlled firms. In columns 3 and 4, an indicator for the regime of new control (below eight years of seniority) is added, as specified in Equation (3). Control variables (firm and CEO characteristics, type of control, industry specific effects) are the same as in Table 4. The adjusted R² is expressed in percentages.

87Controlling shareholders in the fourth regime, holding 46% or more of the common stock, have a substantially stronger impact on CEO pay: cash and total compensation are respectively about 31% and 50% lower as compared with the first regime. Unreported Wald tests for the equality of the coefficients show that the negative impact from the fourth control regime is significantly stronger than the negative impact from the second regime (confidence levels of 98% and 99.7% for cash and total compensation, respectively).

88The third regime of control has a specific pattern. Controlling shareholders who hold between 34% and 46% of equity behave no differently from the first regime of non-controlled firms as concerns cash compensation, and only have a 15% downward effect on total compensation (with significance at the 5% level). Higher levels of pay in this intermediate class of controlled firms, as compared with the second and fourth regimes, seem to be mostly in the form of non-equity components of pay. To check for robustness I run a regression specifically for equity-based compensation (not reported) – based on a Tobit model, since this is a limited dependent variable. The results show that the level of this component is actually lower than in the first regime of non-controlled firms (with a 97% confidence level) and similar, ceteris paribus, to that found in the second regime of control. Consequently the higher level of total compensation in the third regime, by comparison with that in the second and fourth regimes, mainly takes the form of higher cash compensation.

4.6 – Thresholds for the seniority of control

89In columns 3 and 4 of Table 8 an indicator for controlling shareholders with fewer than eight years of seniority is added to the specification (see Equation (3)). The indicator variable takes value one if the controlling shareholder holds, alone or in concert, at least 11% of the equity and has been represented on the board of directors for less than eight years, and zero if not. The coefficients are significantly positive both for cash and total compensation. This means that “new” controlling shareholders provide their CEOs with higher pay than “long-term” controlling shareholders.

90These coefficients should be added to those associated with the regimes of control so that the effect of these new controlling shareholders, as compared with CEO pay practices found in non-controlled firms, can be evaluated. Unreported regressions that separate out new and long-term controlling shareholders are also run in order to test for statistical significance. They show that new controlling shareholders provide similar cash compensation but lower total compensation compared with non-controlled firms.

91Unreported regressions of the two components of cash compensation, namely base salary and bonuses, reveal that higher cash pay found for new control compared to long-term control is mostly in the form of bonuses rather than base salary. Another unreported regression restricted to long-term incentives as a dependent variable confirms that new controlling shareholders grant significantly more equity-based compensation than long-term ones, though it remains lower than in non-controlled firms.

92Newly controlled firms, compared with long-term ones, mostly provide their CEOs with higher levels of pay in the form of variable components of pay. Compensation packages in firms with low seniority of control are thus quite similar to those found in non-controlled firms, although they rely slightly less on equity-based pay.

93These results provide evidence that new controlling shareholders have to deal with agency and retention issues, and so reveal some similarity with the poor governance found in non-controlled firms.

94By contrast, the results show that controlling shareholders are able to improve the quality of their monitoring after about eight years of presence on the board, and then exert a downward pressure on the levels of CEO pay.

4.7 – Estimated impacts of the control regimes

95The results for the degree and seniority of control are also significant in terms of their economic impact on CEO compensation. The first three rows of Panel A in Table 9 display mean values of firm size and of cash compensation within each degree of control regime. The differences in the levels of pay observed from one regime to another are mainly explained by the size effect, which is controlled for in the base specification. The two following rows provide an estimate for the impact of the degree of control. The regime effects are the coefficients of the indicator variables found in Table 8. These fixed effects are then used to estimate the gap, in thousands of euros, between the actual level of CEO pay in a given control regime and the level he/she would receive in a non-controlled firm (the omitted first regime), all other things being equal. The last three rows provide the same estimate for total compensation.

Table 9

Mean size and compensation for each regime and estimated excess compensation (in €000, except for regime effects)

Panel A: Degree of control (entire sample)

1st regime (0% to 11%) 2nd regime (11% to 34%) 3rd regime (34% to 46%) 4th regime (≥46%) Sales 18 620 497 12 112 069 7 643 961 8 075 550 Market capitalization 16 182 464 7 282 453 6 344 203 7 443 256 Cash compensation 1 718 1 198 1 310 998 Regime effect Benchmark -0,17 n.s. -0,31 Estimated gap 0 225 n.s. 357 Total compensation 2 785 1 764 1 860 1 484 Regime effect Benchmark -0,32 -0,15 -0,50 Estimated gap 0 664 301 965 N (firm-year obs.) 284 296 144 395

Panel A: Degree of control (entire sample)

Note: Mean cash and total compensation for each degree of control regime. The average values of sales and market capitalization hint at the non-linear size effect that partly explains the differences in the levels of pay from one regime to another. The “regime effect” corresponds to the specific impact of controlling shareholders in a given regime, as estimated in Table 8 – “n.s.” stands for non-significant coefficients. The first regime serves as a benchmark and the “estimated gap” in other regimes measures the variation of pay the CEO would have if he/she served in a firm from the first regime, ceteris paribus. It is measured as: exp[ln(Cash or Total compensation) – Regime effect] – Cash or Total compensation. For instance, CEOs receive an average of €1,198,000 cash compensation in firms falling in the second regime; they would receive €225,000 more if they served in a firm falling in the first regime, all other things equal. The last row presents the number of firm-year observations in each regime.

Panel B: Seniority of control (sub-sample)

Sales Market cap. Cash comp. Regime effect Estimated gap Total comp. Regime effect Estimated gap 9 605 012 5 246 163 1 273 0,22 -246 1 803 0,26 -418

Panel B: Seniority of control (sub-sample)

Note: Sub-sample of controlled firms (i.e. 2nd to 4th regimes) with new controlling shareholders (i.e. seniority <8 years). N=175 firm-year observations. The “regime effect” corresponds to the coefficient found in Table 8. The “estimated gap” is calculated as in Panel A. For instance, CEOs in newly-controlled firms receive an average of €1,273,000 cash compensation, they would receive €246,000 less if they served in a long-term controlled firm (i.e. seniority of control ≥8 years), all other things equal.

Mean size and compensation for each regime and estimated excess compensation (in €000, except for regime effects)

96The average CEO in the second class of firms receives a cash compensation of €1.20 million and a total compensation of €1.76 million, which is about 17% and 32% lower respectively than the cash and total compensation the CEO would receive if he/she served in a non-controlled firm, all other things equal. This would represent about €225,000 [17] more for cash compensation and €664,000 more for total compensation. Similarly, the average CEO in the fourth class of firms receives a cash compensation of €998,000 and a total compensation of €1.48 million, and would receive an additional €357,000 in the form of cash and €965,000 in the form of cash plus equity-based pay if he/she served as a CEO in the first class of firms. The cash compensation gap is not significantly different from zero between the first and third regimes (see column 1 in Table 8), but is about €301,000 for total compensation.

97Panel B of Table 9 displays the mean values of size and compensation for the sub-sample of firm-years that are controlled (degree of control greater than 11% of ownership) with less than eight years of seniority. The mean cash and total compensation are close to those found in the second control regime (see panel A), although the size of the firms in the former is smaller than in the latter; this already illustrates the higher relative pay received by CEOs when controlling shareholders have a low seniority. Specifically, CEOs would be paid €246,000 less cash compensation, and €418,000 less in total compensation, if they served in a similar firm with a more experienced monitor.

4.8 – Interpretation of threshold effects

98The second and fourth regimes in the degree of control – controlling shareholders holding, respectively, between 11% and 34% and above 46% of equity – grant significantly lower levels of cash and total pay, with a significantly stronger effect for the fourth regime. This provides evidence of an “optimal governance” effect as these large shareholders engage in monitoring. Specifically, controlling shareholders in the fourth regime hold a majority share of equity, and they are the ones who can better counterbalance managerial power, and deal with agency and retention issues.

99Comparatively, controlling shareholders in the third regime, holding between 34% and 46% of equity, are associated with higher levels of pay, mostly in the form of non-equity components. This latter pattern is consistent with the “entrenchment” hypothesis: on the one hand, entrenched controlling shareholders tend to provide higher salaries to their CEO in order to buy his/her loyalty and, on the other hand, they are not willing to provide equity-based incentives that might align the CEO’s interests with those of outsider shareholders. This also has some similarity with previous studies of the impact of managerial ownership on firm value, which find evidence of entrenchment on the part of CEOs at intermediate levels of ownership (Morck et al., 1988; Short and Keasey, 1999; Davies et al., 2005).

100Entrenchment on the part of controlling shareholders can here be explained because of the way that, at intermediate levels in the degree of control, they bear the full cost of monitoring activities – external minority shareholders being presumably free-riding (Shleifer and Vishny, 1997) – but they do not receive all of its benefits, since they hold minority cash flow rights. They may then try to use their dominant position to extract private benefits at the expense of external minority shareholders, in order to increase their cash-flow return from monitoring activities.

101Controlling shareholders in the second regime could also try to entrench, since they are the principal monitors but receive less than one third of cash flow rights. The results showing “alignment” for these monitors is however consistent with their more contestable position. Some authors refer to anecdotal evidence suggesting that controlling shareholders would be protected from attempted hostile takeovers only above ownership levels of 25% or 30% (Weston, 1979; Holderness, 2003). Below these levels bad governance is likely to lower the value of the firm and increase the risk of hostile takeovers. Controlling shareholders with a low degree of control are therefore oriented towards aligning with and pursuing the same interest as outside minority shareholders, contrary to those in the third regime who can entrench because they are sufficiently dominant and protected from hostile takeovers. [18]

102Controlling shareholders in the fourth regime hold a majority equity position that provides them with a majority share of the cash flow return from their monitoring activities. This can explain why their interests are more aligned with those of external shareholders, so that they do not try to entrench. In addition, their high degree of control produces a higher intensity of monitoring, with a stronger downward pressure on the level of CEO pay compared with the second regime.

103As far as the seniority of control is concerned, the results show that shareholders who take control of a firm tend to pay higher bonuses and long-term incentives to their CEOs during approximately the first eight years of control, compared with shareholders with higher seniority. Higher levels of pay in the form of higher variable and performance-sensitive components of pay do not provide support for an “entrenchment” hypothesis for these new controlling shareholders. Instead, the results show that these shareholders are subject to higher asymmetries of information with respect to top management, and to stronger agency and retention issues. As a consequence, new controlling shareholders need to have recourse, as in non-controlled firms, to “optimal contracting” in order to enhance their monitoring: they rely on compensation packages that are sensitive to market value, a posteriori informative of CEOs ability and prone to retain CEOs in case of increased outside opportunities.