Financial leverage and managerial compensation: Evidence from the UK

Research in Economics 65 (2011) 36–46

Contents lists available at ScienceDirect

Research in Economics journal homepage: www.elsevier.com/locate/rie

Financial leverage and managerial compensation: Evidence from the UK✩ Gianluca Papa a , Biagio Speciale a,b,∗ a

ECARES, Université Libre de Bruxelles, Belgium

b

FNRS, Belgium

article

info

Article history: Received 8 January 2009 Accepted 11 March 2010 Keywords: Principal-agent theory Executive pay Financial leverage Asymmetric information

abstract Using the data on a panel of quoted UK firms over the period 1995–2002, this paper studies the effects of financial leverage on managerial compensation. The change in the investors’ expectations that caused the recent collapse of the stock market tech bubble has been used as a source of plausibly exogenous variation in the firm’s debt. We find that pay-forperformance sensitivity is increasing in financial leverage, with the exception of the 10% most levered firms. © 2010 University of Venice. Published by Elsevier Ltd. All rights reserved.

1. Introduction This paper empirically studies the effects of financial leverage on managerial compensation. Since Jensen and Meckling (1976) the firm has been recognized as a nexus for a multitude of contracting relationships and the literature that followed their seminal paper recognized that managerial compensation depends on the conflict of interests between different contracting parties that are asymmetrically informed. On the one hand, executive compensation is designed to align managerial incentives with the interests of shareholders who are uninformed about the level of effort exerted by their managers and who therefore link part of the remuneration to the firm’s performance.1 This is done to minimize the agency cost coming from the separation between ownership and control. Since the firm’s debt is recognized to reduce shareholder–manager conflicts, several theoretical works predict a negative association between pay-for-performance sensitivity and financial leverage. There are different reasons that explain this negative relationship. First, debt could decrease the firm’s free-cash flow, which should reduce the manager’s ability to use corporate resources for empire-building purposes (Jensen, 1986). Second, higher debt could increase the threat of bankruptcy (Grossman and Hart, 1982), which could imply that managers act in the interest of shareholders even in the presence of low-powered explicit incentive schemes. Third, an increase in the firm’s debt could increase monitoring by lenders, which could substitute for the provision of monetary incentives by shareholders. On the other hand, higher financial leverage could increase the stockholder–bondholder conflicts. Indeed, a compensation that is designed to align managerial incentives with shareholder interests could induce risk-shifting incentives for managers. ✩ We are indebted to Bruno M. Parigi and two anonymous referees for helpful comments and suggestions that significantly improved the paper. We would also like to thank Marco Becht, Michele Cincera, Mathias Dewatripont, Abdul Noury, seminar participants at ECARES (Université Libre de Bruxelles), the University of Catania, the Vrije Universiteit Amsterdam, the JEI 2005 (Bilbao) and the 2006 SOLE meetings (Cambridge, MA), in particular Roberto Cellini, Vicente Cuñat, Maria Guadalupe, Kevin Hallock, Patrick Legros and Joseph Tribo, for useful discussions. We have also received valuable suggestions from Marco Gulisano, Heiko Karle, Alexander Sebald, Saverio Simonelli and Sergey Stepanov. Biagio Speciale is a postdoctoral researcher at the F.R.S-FNRS and gratefully acknowledges their financial support. The usual disclaimer applies. ∗ Corresponding address: ECARES, Université Libre de Bruxelles, 50 Avenue F.D. Roosevelt, 1050 Brussels, Belgium. E-mail address: [email protected] (B. Speciale). 1 Incentive compensation may also induce correlation between the portfolio of managers and the cash flow of the firms they manage. Bisin et al. (2008)

studied the agency problem between a manager and shareholders when the former can hedge her compensation against the poor performance of her firm using financial markets and shareholders can monitor the manager’s portfolio in order to keep her from hedging, but monitoring is costly. 1090-9443/$ – see front matter © 2010 University of Venice. Published by Elsevier Ltd. All rights reserved. doi:10.1016/j.rie.2010.03.003

G. Papa, B. Speciale / Research in Economics 65 (2011) 36–46

37

In a context of asymmetric information concerning risk choices taken by managers in their investment decisions, bondholders would anticipate that managers have risk shifting incentives by simply observing the management compensation structure. This would affect their decision about the price of the bond. Low pay-for-performance sensitivity can thus be seen as a precommitment device to minimize the agency costs of debt related to the risk-shifting problem (John and John, 1993). Either by reducing shareholder–manager conflicts or by nourishing stockholder–bondholder ones, the mentioned theoretical works predict that an increase in financial leverage lowers pay-for-performance sensitivity. There has been little empirical work to test these theoretical predictions. Studying management compensation policies in 77 publicly traded firms that filed for bankruptcy or privately restructured their debt from 1981 to 1987, Gilson and Vetsuypens (1993) find that during periods of financial distress pay-for-performance sensitivity is extremely low (even insignificantly different from zero), while it increases after firms restructure their debt. Using a sample of 1652 CEOs in the largest publicly traded US companies for the period 1993–1999, Ortiz-Molina (2007) find that the elasticity of managerial remuneration to total shareholder returns is decreasing in total financial leverage, except for firms with convertible debt (the latter finding is in line with John and John, 1993 and with Green, 1984). In contrast to the findings of the mentioned theoretical and empirical literature, in our paper we show how pay-forperformance sensitivity may be increasing in financial leverage, with the exception of the most levered firms. We estimate a panel of quoted firms from the UK, ranging from 1995 to 2002. In the United Kingdom, remuneration is set by an independent committee. There is also nearly full disclosure and intense scrutiny of remuneration practices by institutional investors. It is probably fair to say that it is unlikely that managers set their own pay. We employ robust regression techniques to deal with the presence of outliers and firm dummy variables to take account of unobserved heterogeneity. A methodological contribution of this paper is that we exploit the tech bubble burst as a source of variation in firm’s debt. The collapse of the tech bubble after the year 2000 implied different changes in financial leverage for sectors with different technological levels. The sharp change in investor’s expectations was very likely to be exogenous to the incentive schemes of the firms. In contrast to results in previous empirical works, we find evidence of an increasing relationship between financial leverage and pay-for-performance sensitivity for almost all the firms in our sample. This relationship turns into negative for the 10% most levered firms. Our estimates indicate that, on average, the negative effect of financial leverage concerned UK companies with long term debt larger than 34% of total assets. Moreover, our estimates show an increasing relationship between financial leverage and total short-term compensation (basic salary plus bonus) as well, with the exception of firms with a level of financial leverage that is larger than the 99th percentile of the financial leverage distribution. In auxiliary regressions, we also show that for our sample both total shareholder returns and their variance are positively associated with financial leverage. These additional findings are helpful to shed some light on possible mechanisms driving the relationship of interest. We discuss these potential channels in the context of the standard agency theory with moral hazard. For firms with high levels of financial leverage, it was less easy to align managerial and shareholder interests through an incentive scheme because total shareholder returns may have become a noisier signal for managerial effort. For all other firms, the estimates show a positive effect of financial leverage on incentives. In this case, financial leverage may have raised the marginal value of the agent’s effort for the principal. Additionally, it may have implied a positive effect on pay-forperformance sensitivity through its effect on risk. This mechanism may be along the lines of a recent literature showing a positive relationship between risk and incentives through effects on delegation of responsibilities, endogenous market structure and/or endogenous matching between managers and firms (see Prendergast, 2002; Raith, 2003; Wright, 2004; Serfes, 2005). The rest of the paper proceeds as follows. In Section 2, we present an empirical analysis of the relationship between financial leverage and pay-for-performance sensitivity. Section 3 proposes some possible explanations for our estimation results. We conclude in Section 4. 2. Empirical analysis of the effects of financial leverage on pay-for-performance sensitivity 2.1. Data description We have observations on an unbalanced panel of UK quoted firms belonging to the FTSE350TM and FTSE Small CapTM indices. Data come from two sources: companies accounting data are taken from the Amadeus database, while data on manager compensation and firm stock returns come from the Datastream database. Our initial dataset is made up of 10 yearly observations (ranging from 1993 to 2002) for a sample of 754 quoted companies for which we have availability of data on financials and manager compensation from the 1135 ones that make up the FTSE350TM and FTSE Small CapTM indices. The dependent variable (ManComp) is the compensation of the firm’s highest paid director (usually the CEO), taken by the Datastream ‘‘Equities’’ dataset. It is in constant ’97 £. Following Jensen and Murphy (1990), Core et al. (1999) and Cuñat and Guadalupe (2005) – among others – the dependent variable is in levels. This allows to estimate the pound change in managerial compensation associated with an increase in the measure of firm’s performance. ManComp includes the basic salary plus a bonus usually linked to short term performance. It does not include long term elements of compensation, such as stock options, which are not readily available for UK firms. This study thus follows the definition of remuneration variable that has been used in most empirical studies on managerial compensation in the UK (among others, see Conyon, 1997; Girma et al., 2006). We therefore focus on the effects of firm’s debt on short term compensation. A remark is worth mentioning. We rely on the work of Conyon and Murphy (2000), who collect UK stock option data for a cross section of firms, for the

38

G. Papa, B. Speciale / Research in Economics 65 (2011) 36–46

Table 1 Summary statistics. Variable

Mean

Median

St. dev.

Min

Max

Remuneration TSR Financial leverage Concentration Total assets (millions of £)

455 779 16 13.4 5.8 1037.9

309 361 5.8 9.1 4.2 137.3

518 783 73.4 14.6 5.4 4948.9

4820 −96.9 0 −7.3 0.7

9 583 385 1332.3 112 41.3 165 900

Remuneration (ManComp) is the compensation of the firm’s highest paid director (in constant 1997 £). TSR is yearly total shareholder returns. Financial leverage (Lev) is long term debt (debt with maturity higher than one year) as a percentage of total assets. Concentration (Conc) is the median value of the ratio of profits on sales, for each Sic UK92 3-digit sector. Totalassets is in millions of constant 1997 £.

year 1997, directly from the remuneration reports. They compare the average composition of total pay in the UK and in the US: in 1997 the CEOs of the largest British companies received on average 59% of their total remuneration in the form of base salaries, 18% in bonuses and 10% in share options, while US executives received 29% of their total pay as base salary, 17% as bonuses and 42% as share option grants. Conyon and Murphy (2000) thus report that share option grants and other components of long term compensation represent a fairly small percentage of total remuneration for the British CEOs, while they are much more important for the American CEOs.2 Moreover, in the average composition of total pay that is presented above, the share options are valued at grant-date using the Black and Scholes (1973) formula. As reported by Conyon and Murphy (2000) and Hall and Murphy (2000), the Black and Scholes (1973) formula is at best a measure of the company’s opportunity cost of granting the option and typically overstates its value to the manager–recipient.3 It is therefore plausible to assume that for our sample of UK firms more than 80% of the total pay was made up of base salary and bonus. For this reason, if the option grants were available and included in our compensation measure, results might be qualitatively similar to the ones presented in this section. As a measure of firm’s performance we consider a market based one: yearly total shareholder returns (TSR). This variable is defined as the percentage increase in value of an investment in shares including reinvested dividends minus UK cpi inflation to obtain real returns. In order to control for risk, the variance of TSR is calculated by using monthly TSR for the three years preceding the one under consideration. Following Aggarwal and Samwick (1999), the variance of TSR is included in the form of cumulative distribution function (VarTSR). As Core and Guay (2002) document, this variable transformation is obtained by ranking the observations from 1 to the sample size, subtracting 1, and then dividing by the sample size minus 1. These transformed ranks are uniformly distributed between zero and one, which correspond respectively to the minimum and maximum values of the variance of TSR observed in the sample. The normalization of the variance of TSR variable to the unit interval lowers the impact of outliers in the regression specification and enables ease of interpretation of the estimated coefficient. In addition, we rely on the estimation results of Conyon (1997), who finds that relative performance evaluation (i.e., using the performance of other firms operating within a similar market environment) has not been a statistically significant determinant of the highest paid director compensation in the UK. As a measure of financial leverage, we use long term debt (debt with maturity higher than one year) as a percentage of total assets (Lev).4 Following Rajan and Zingales (1995), we use a stock measure rather than a flow one (e.g. interest coverage). It is worth noting that the choice of a book measure instead of a market measure does not qualitatively influence our results because all firms follow the same UK accounting standard and our analysis does not deal with a cross-country comparison. Our competition measure is based on the Lerner Index or price–cost margin, that is (price − marginal cost)/price: in fact, in less competitive industries the mark-up is expected to be higher. Since we lack data on marginal cost, we follow the literature and build a proxy for the Lerner Index by taking the ratio of profits on sales. This ratio is computed for every firm on each Sic UK92 3-digit sector: for each industry the median value (Conc) represents our inverse measure of competition. Furthermore, we control for the firm’s size by introducing totalassets (in millions of constant ’97 £) as a regressor. Table 1 presents summary statistics for the variables described above. The mean annual remuneration for the highest paid director is about £456 k, while the median is 32% lower (£309 k). The high skewness of the remuneration variable is also confirmed by its minimum (£4820, The Sanctuary Group, 1995) and maximum (£9.583 mil, Vodafone, 2000) values. The average long term debt is about 13.4% of total assets. Although some firms display a null amount of debt, there is also a bunch of highly leveraged firms. 2.2. Estimation strategy and empirical results To answer the research question of interest, we estimate the following econometric model: ManCompit = α1 TSRit + α2 TSR · Levit + α3 TSR · Lev2it + α4 Levit + α5 Lev2it + α6 Xit + ci + dt + εit

(1)

where X is a vector of control variables. The other variables follow the notation presented in the previous subsection. 2 Different corporate tax rules, institutional and cultural factors may explain these differences (see Conyon and Murphy, 2000 for further details). 3 See also Girma et al. (2006) for further discussion on the evaluation of stock options using the Black and Scholes (1973) formula. 4 We obtain similar results using alternative definitions of financial leverage, such as the ratio of long term debt on capital (long term debt + total shareholder funds).

G. Papa, B. Speciale / Research in Economics 65 (2011) 36–46

39

Table 2 The effects of financial leverage on pay-for-performance sensitivity. ManComp dependent variable

(1)

(2)

TSR

117.7***

54.26**

99.08***

TSR · Lev

1.52

9.61

9.47***

−0.17***

−0.17***

TSR · Lev2

(3.85)

(1.08)

TSR · Conc

−8.95**

Lev

1295.87***

Lev2

(−2.05)

(4.17)

(3)

(2.26)

***

(2.79)

(−2.68)

(3.10)

(2.73)

(−2.63)

−9.45** (−2.17)

1098.21***

1114.58***

−7.97

−8.26

(3.54)

(−1.57)

(3.57)

(−1.61)

VarTSR

9099.09

9089.15

9390.02

Conc

3713.78***

3584.96***

3699.02***

Totalassets Type of estimation Year dummies Turning point R2 # of observations

(1.08)

(4.86)

42.65 ***

(1.10)

(4.83)

42.67 ***

(1.12)

(4.86)

42.68 ***

(113.74)

(115.24)

(114.41)

Firm’s fixed eff. Yes

Firm’s fixed eff. Yes 28.26 (85th pc.) 0.65 3960

Firm’s fixed eff. Yes 27.85 (85th pc.) 0.65 3960

0.64 3960

ManComp is the compensation of the firm’s highest paid director (in constant 1997 £). TSR is yearly total shareholder returns. Lev is long term debt (debt with maturity higher than one year) as a percentage of total assets. VarTSR is the variance of TSR included in the form of cumulative distribution function. Totalassets is in millions of constant 1997 £. Coefficients for year and firm dummies are not reported. T -statistics are in parentheses. * denotes significance at 10%. ** Denotes significance at 5%. *** Denotes significance at 1%.

Pay-for-performance sensitivity is estimated by the coefficients on TSR and its interaction terms. We are mainly interested ∂ 2 ManComp

in the effects of financial leverage on pay-for-performance sensitivity (i.e., ∂ TSR∂ Lev ), which is captured by the coefficients on the interaction terms TSR · Lev and TSR · Lev2 , i.e.,  α2 and  α3 . The effects of financial leverage on the fixed part of managerial ∂ ManComp compensation (i.e., ∂ Lev conditional to TSR being equal to 0) can instead be studied from the coefficients of the variables ∂ ManComp

Lev and Lev2 , i.e.,  α4 and  α5 . Finally,  α2 TSR + 2 α3 TSRLev + α4 + 2 α5 Lev (i.e., ∂ Lev ) allows to compute the effects of financial leverage on total short-term compensation at specific values of total shareholder returns and financial leverage. We include firm fixed effects (ci ) to control for all the unobserved variables that do not vary over time. Since we have observations concerning the presence or not of large shareholders only for one year, we need to assume that the effect of ownership structure is captured by the firm dummies. Finally, through the time dummies (dt ) we control for all the unobserved variables that change over time but that are common to all the cross-section units (for instance, the implementation of a set of recommendations related to managerial compensation following the Greenbury code of conduct in 1995). As in Hermalin and Wallace (2001), Core et al. (1999), among others, we lag by one year the explanatory variables, because a large part of the bonus is paid in the year following the one in which performance is measured. Moreover, the fixed part of the compensation is highly influenced by past performances. Through the analysis of the Cook’s (1977) distance and the (1992) criterion,5 we detect the presence of several outliers. Thus, we use both OLS and a robust regression technique that gives less weight to influential observations.6 Table 2 shows the results of our robust estimates.7 The estimation period covers the years 1995–2002, since the initial years are used in building the variance of TSR. In all columns, we control for time and firm dummies (not reported in the table). In the first column, we report estimates without including the term Lev2 and its interaction with TSR. As can be easily seen by comparing the first column with the other ones, the omission of these two terms significantly biases our estimation results. Through the coefficients of TSR · Lev and TSR · Lev2 , columns (2) and (3) show that pay-for-performance sensitivity increases with financial leverage for most of the firms in our sample. The relationship of interest is non-monotonic. The turning point of the inverted-U is given by long term debt equal to about 28.3% of total assets, which is larger than the median value of Lev, and corresponds to the 85th percentile of the financial leverage distribution. Firms that have a level of financial leverage smaller than the turning point have higher payfor-performance sensitivity than they would have in the absence of debt. The result that the estimated turning point is larger than the median value of financial leverage may also be consistent with the findings of Graham (2000) and Morellec (2004) that firms use debt too conservatively.

5 Hadi’s (1992) criterion for outlier detection recursively computes the distance of an observation from a cluster of observations in the model. It is effective in dealing with masking and swamping problems. 6 Robust regressions are estimated using iteratively reweighted least squares, i.e., the rreg routine in Stata. Among others, this robust estimation technique is used by Croxson et al. (2001) and Hall and Liebman (1998). Results in all the tables presented in the paper are from robust regressions. 7 The presence of several outliers and leverage points makes the OLS estimates less reliable.

40

G. Papa, B. Speciale / Research in Economics 65 (2011) 36–46

Table 3 Financial leverage before and after the tech bubble burst. Variable

Mean

Median

St. dev.

Min

Max

Lev before the tech bubble burst (1995–1999) Lev after the tech bubble burst (2000–2002) H0 :Lev2000–2002 -Lev1995–1999 = 0 (t-stat)

12.7 15.1 4.99

8.6 10.7 4.37

14.1 15.7

0 0

112 99

Lev is long term debt (debt with maturity higher than one year) as a percentage of total assets. Table 4 The effect of the collapse of the stock market tech bubble on financial leverage and total shareholder returns. Dependent variable

Lev

Lev

TSR

TSR

Bubdum

0.63***

0.83***

−7.34***

−4.26***

(4.08)

(4.65)

(−5.18)

(−2.61)

Bubdum ∗ HTMan

−1.89

Bubdum ∗ MHTMan

0.5

−9.07**

−1.89***

−47.03***

(−2.59)

(−2.21)

(−2.78)

Firm’s fixed eff. 0.63 3960

Firm’s fixed eff. 0.63 3960

−11.16* (−1.67)

(1.11)

Bubdum ∗ HTSer Type of estimation R2 # of observations

***

(−7.56)

Firm’s fixed eff. 0.60 3960

Firm’s fixed eff. 0.60 3960

Bubdum is a dummy variable that is equal to 1 for the years 2000 onwards. HTMan and HTSer are dummies equal to 1 for high technology manufacturing and service firms, respectively. MHTMan is a dummy equal to 1 for medium–high technology manufacturing firms. Coefficients for firm dummies are not reported. T -stat are in parentheses. * Significance at 10%. ** Significance at 5%. *** Significance at 1%.

The coefficients of Lev and Lev2 show that financial leverage also has a significant impact on the fixed part of managerial remuneration. The relationship between firm’s debt and total short term compensation (basic salary plus bonus) is increasing, with the exception of the most levered firms. We have computed the turning point of the non-monotonic relationship at the median value of total shareholder returns: it is equal to a level of long term debt of about 63% of total assets (

∂ ManComp ∂ Lev

> 0 ⇐⇒ Lev <

 +  α2 TSR α4 ,  + −2( α3 TSR α5 )

 is the median value of total shareholder returns), which corresponds to where TSR

the 99th percentile of the financial leverage distribution. We now briefly describe the results related to the control variables. Competition has a positive and significant effect on the pay-for-performance sensitivity (see the coefficient of TSR · Conc in columns 1 and 3 of Table 2). This result is interesting since most theoretical papers that study the effect of product market competition on managerial remuneration stress the many countervailing forces that shape the optimal incentive scheme. It is in line with the results of Cuñat and Guadalupe (2005). Moreover, managers of firms in highly concentrated industries receive higher base salaries relative to managers working for firms in more competitive sectors (see the coefficient of the variable Conc). Finally, an increase in total assets of 1 mil £ implies an increase in managerial remuneration of about 43 £. 2.3. The collapse of the stock market tech bubble as a source of exogenous variation in firm’s debt In this subsection, we address the issue of possible endogeneity of the financial leverage variable. Financial leverage is a decision variable for the management. Endogeneity may arise if, for instance, the incentive scheme affects the managerial choices concerning the firm’s level of debt. We estimate the effect of financial leverage on pay-for-performance sensitivity using a natural experiment as an exogenous source of an increase in financial leverage. This natural experiment is the collapse of the stock market tech bubble, which started in 2000. Due to a sharp change in investor’s expectations, the tech bubble burst can be considered as exogenous to the internal organization of the firms and, in particular, to their incentive schemes. Moreover, it is plausible to assume that its magnitude was hardly predictable by the firms. Table 3 presents summary statistics for the financial leverage variable before and after the bubble burst. After the year 2000, there was an increase in the mean value of the financial leverage variable.8 This provides the exogenous variation that helps in the identification of the coefficient of interest. The change in investor’s expectations implied different changes in financial leverage for sectors with different technological levels. This intuition is confirmed by a naive fixed-effect estimation with Lev as dependent variable. Bubdum is a dummy variable that is equal to 1 for the years 2000 onwards (after the collapse of the stock market bubble), 0 for the previous years. The first column of Table 4 shows that after the stock market bust UK firms increased the level of their long 8 The descriptive evidence in Table 3 does not necessarily suggest that – during financial crises – firms tend to replace stock financing with debt financing. For instance, in the 2008–2009 crisis the financial leverage/capital ratio of banks increased not because of higher borrowing, but because of capital destruction.

G. Papa, B. Speciale / Research in Economics 65 (2011) 36–46

41

Table 5 The effect of the collapse of the stock market tech bubble on pay-for-performance sensitivity. ManComp dependent variable

(1)

(2)

TSR

99.08***

63.30

(3.10)

(1.56)

−17.92

TSR · Bubdum

(−0.21)

9.47***

TSR · Lev

9.48**

(2.73)

(2.18)

17.81**

TSR · Lev · Bubdum TSR · Lev

2

TSR · Lev2 · Bubdum TSR · Conc

(2.04)

−0.17

***

(−2.63)

(−1.69)

−0.24* (−1.68)

−9.45

−3.93

**

(−2.17)

(−0.62)

−10.14

TSR · Conc · Bubdum 1114.58

Lev

***

(3.57)

Lev · Bubdum Lev2 2

−0.16*

−8.26 (−1.61)

Lev · Bubdum

(−1.19)

−182.23 (−0.53)

2810.6*** (6.35)

16.35*** (2.92)

−45.92*** (−5.79)

VarTSR

9390.02

10841.35

Conc

3699.02***

3798.75***

Totalassets Type of estimation Year dummies Turning point R2 # of observations

(1.12) (4.86)

42.68 ***

(1.32)

(5.09)

42.48 ***

(114.41)

(116.43)

Firm’s fixed eff. Yes 27.85 (85th pc.) 0.65 3960

Firm’s fixed eff. Yes 34.11 (90th pc.) 0.66 3960

See variable definition in the notes of Tables 2 and 4. ManComp is in constant 1997 £, totalassets is in millions of constant 1997 £. Coefficients for year and firm dummies are not reported. T -statistics are in parentheses. * Significance at 10%. ** Significance at 5%. *** Significance at 1%.

term debt as a percentage of total assets by 0.6%, on average. In the second column of Table 4, we have added interactions of Bubdum with dummies denoting the technological level of the firm.9 After the stock market bust, there was a significant reduction in the financial leverage variable for high technology manufacturing (HTMan) and service (HTSer) firms. This result should not be considered very surprising since the evaluation of these firms by potential investors heavily depends on future prospects of their R&D projects. Following the collapse of the tech bubble, the sharp decline in investors’ confidence may have implied an increase in financing costs for high-tech firms, due to their quasi absence of collateral and to the threat of bankruptcy. On the contrary, low-tech firms could rely on collateral. After the change in investor’s expectations low-tech firms increased the level of their financial leverage. Columns 3 and 4 of Table 4 show results from regressions similar to the previous ones, but with TSR as dependent variable. On average, the collapse of the stock market bubble implied a decrease in total shareholder returns of about 7%. The decline of shareholder returns concerned UK companies regardless of their technological level. High technology service firms experienced a decline in total shareholder returns of more than 50% (4.26 + 47.03). Table 5 reports results of a specification in which we interact the experiment (the dummy variable Bubdum that takes value one after the collapse of the stock market bubble) with our measure of financial leverage.10 For comparison purposes, the first column of Table 5 reproduces results already presented in the last column of Table 2. The estimation strategy that is used in this subsection is a simple ‘‘one group before and after design’’, i.e., a before and after design without an untreated comparison group (see Meyer, 1995). The key identifying assumption is that, absent the stock market bust and conditional on the time dummies, there would have been no change in average compensation: E [ManComp1 ] − E [ManComp0 ] = 0, where E [ManComp1(0) ] is the average compensation after (before) the year 2000. 9 For manufacturing firms we have used the OECD classification (NACE classification in brackets). High technology manufacturing (HTMan): Space and aviation (353), Computers and office machinery (30), Electronics-communications (321–322), Pharmaceuticals. Medium–high technology manufacturing (MHTMan): Scientific instruments (33), Electrical machines and equipment (2971, 31, 323), Motor vehicles (34, 352), Chemicals (24, excl. 244), Other transport equipment (354, 355), Non-electrical machines (29 excl. 2971). High-tech services (HTSer) include the following sectors: Telecommunications (642), Computer and related activities (72), Research and development (73). 10 Cuñat and Guadalupe (2005) exploit the appreciation of the British pound in 1996 as a natural experiment associated to an increase in competition to study how product market competition shapes incentive contracts.

42

G. Papa, B. Speciale / Research in Economics 65 (2011) 36–46

The interaction term TSR∗Bubdum accounts for the fact that firms may have had different pay-for-performance sensitivity after the collapse of the stock market bubble regardless of their level of debt. Lev ∗ TSR is instead included to take into account that firms with different financial leverage may have had different pay-for-performance sensitivity regardless of the stock market bust. The main coefficients of interest (i.e., the coefficients of the variables TSR ∗ Lev ∗ Bubdum and TSR ∗ Lev2 ∗ Bubdum) are statistically different from zero. They capture the differential effect of financial leverage on pay-for-performance sensitivity after the year 2000. In this regression, the sharp change in investor’s expectations is used as a source of variation in debt that is exogenous to the internal organization of the firms and, in particular, to their incentive schemes. The estimates show an inverted-U relationship between financial leverage and pay-for-performance sensitivity. The turning point is given by a level of long term debt equal to about 34% of total assets, which corresponds to the 90th percentile of the financial leverage distribution. Similar qualitative results are obtained for the relationship between financial leverage and total short-term compensation (basic salary plus bonus). When computed at the median value of TSR, the turning point of the latter relationship is 43.69 (the 95th percentile of the financial leverage distribution). For firms with levels of long ∂ ManComp term debt smaller (larger) than 43.69% of total assets, the derivative conditional on Bubdum being equal to 1 is ∂ Lev positive (negative). Finally, we briefly discuss the magnitude of some coefficients of interest. At the median value of financial leverage (Lev = 9.1), after the collapse of the stock market bubble an increase in TSR by one percentage point raised ManComp by about £ 215. This result is in line with previous empirical literature showing that managerial pay is marginally influenced by firm performance. Moreover, for a hypothetical firm with median TSR (5.8) and financial leverage (9.1), the second column of Table 5 suggests that after the collapse of the stock market bubble a one percent increase in financial leverage over total assets raised the short-term compensation of the highest paid director by £ 2388. 2.4. Additional results on TSR and variance of TSR We present auxiliary regressions that allow to explore the effects of financial leverage on total shareholder returns and the variance of TSR. This is done to better understand the mechanisms driving the relationship between financial leverage and incentive pay. In particular, we estimate the following two equations: TSRit = a1 Levit + a2 Lev2it + ci + dt + εit

(2)

VarTSRit = b1 Levit + b2 Lev2it + ci + dt + εit

(3)

where we control for all the time or firm invariant variables through the inclusion of fixed effects ci and dt . The estimated coefficients  a1 and  a2 in Eq. (2) are respectively 0.379 (t = 1.56) and −0.0036 (t = −2.44). While the linear term is less precisely estimated (p-value = 0.118), the two coefficients are jointly significant (p-value = 0.03). Total shareholder returns are increasing in financial leverage both at the mean and median value of Lev. For the 2% most levered firms (firms with a level of long term debt of more than 53% of total assets), the variable TSR is instead decreasing in financial leverage. These highly levered firms also include firms with (accounting) negative net worth. In Eq. (3),  b1 and  b2 are 0.003 (t-stat: 4.76) and 0.00001 (t-stat: 1.26), respectively. VarTSR is therefore increasing in financial leverage. As we describe below, this positive association between risk and financial leverage may affect the provision of incentives. 3. Possible explanations for a non-monotonic relationship between financial leverage and pay-for-performance sensitivity The previous section shows evidence of a non-monotonic relationship between financial leverage and pay-forperformance sensitivity. The relationship of interest is expected to be decreasing only for the 10% most levered firms, while for all other firms there seems to be a positive effect of financial leverage on managerial pay. This implies that at least two theoretical effects are at work, each prevailing in different ranges of financial leverage. Using data on US firms, the related literature surveyed in the introduction has instead shown a negative effect of financial leverage on pay-for-performance sensitivity. While plausible reasons for this negative effect have already been presented above, in this section we discuss some possible explanations for the increasing part of the inverted-U curve (up to the 90th percentile of the financial leverage distribution). As results from the estimation of Eq. (3) show, financial leverage is positively associated with risk. Consequently the increasing part of the inverted-U curve may in principle depend on a positive relationship between risk and incentives. In the standard agency model, incentive pay is often considered lower in more uncertain environments (see Holmström and Milgrom, 1987; Aggarwal and Samwick, 1999). There is however a branch of the literature that has shown how under alternative assumptions a positive relationship between incentive pay and risk may arise. For instance, Prendergast (2002) finds that uncertain environments result in the delegation of responsibilities, which may in turn generate stronger incentives. Wright (2004) and Serfes (2005) show that similar results can be obtained in the presence of endogenous matching of managers to firms if managers differ in their degree of risk aversion. In particular, if more risky principals match

G. Papa, B. Speciale / Research in Economics 65 (2011) 36–46

43

with less risk averse agents, then this may result in a positive association between risk and incentives. Raith (2003) studies a model of an oligopolistic industry in which firms provide incentives to managers and market structure is endogenous. His results confirm that risk and incentives may be positively related. Our dataset does not allow us to check whether an increase in financial leverage has raised the delegation of responsibility within firms or has affected the matching between managers and firms. The data we employ are not suitable for the study of market structure dynamics either. However, estimation of (2) has shown that higher financial leverage is positively associated with total shareholder returns, with the exception of the 2% most levered firms. This may have implied an increase in the marginal value of the agent’s effort for the principal. In the following subsection we present a simple theoretical illustration where this effect may explain the increasing part of the inverted-U shape, while the decreasing part of the curve comes from the higher uncertainty due to higher financial leverage (as already found by Aggarwal and Samwick, 1999). 3.1. Financial leverage and managerial compensation: a simple theoretical illustration A risk-neutral principal hires a risk-averse manager who has to exert effort in order to reduce production costs. For instance, this effort could concern the development of new products or of new production technologies, the reorganization of the company, etc. The marginal cost is written as follows: c =K −e−ε

(4)

where K is a positive constant, e is the effort exerted by the agent and ε is a random variable. As usual in moral hazard problems, the effort is not observable to the principal and, since it is costly for the agent, the optimal contract relates the agent’s compensation to the realized cost that we assume to be contractible. The additional hypothesis we make in the present setting is that the uncertainty on the realization of c depends on firm’s financial leverage L and in particular we assume the following distribution for the random term:

ε ∼ N (µ(L), σ 2 (L)).

(5)

We assume that neither shareholders nor the manager use debt as a tool to affect the agency relationship, that is we consider an exogenous level of financial leverage. This assumption implies that this theoretical illustration is simply aimed to go with our empirical analysis, where we address the possible endogeneity of financial leverage and consequently analyze the effects of an exogenous shock in firm’s debt on managerial pay.11 This theoretical exercise is however of little help in the analysis of the properties of the financial leverage equilibrium level. Let debt be raised for a random cost-reducing investment (for instance, a R&D one) prior to the exertion of effort by the manager. µ(L) is the expected benefit (loss) of the cost-reducing investment, net of repayment. For instance, following Myers’ (1984) pecking-order theory, tax advantages and lower agency costs imply that debt may have a comparative advantage with respect to equity as a financing source for cost-reducing investments. Positive levels of financial leverage may imply that the firm is exploiting potentially profitable (cost-reducing) investments or also that is enjoying tax advantages of debt. However, for high levels of financial leverage, the probability of bankruptcy worsens credit ratings and raises both financial and operating costs (for instance, because personnel requires higher salaries to accept the higher risk of unemployment in case of firm’s liquidation), reducing the equity investor’s return. σ 2 (L) denotes the variance of the return from the cost-reducing investment. In Section 2, the data show that both µ(L) (proxied by the total shareholder returns) and σ 2 (L) (proxied by the variance of the TSR) may well be increasing in financial leverage. The property of the function σ 2 (L) implies that shareholders of a firm that is highly indebted because of its investment activity cannot easily distinguish whether the cost reduction is due to managerial effort or to the realization of the random term. The manager has a CARA utility function of the following form: u(w, e) = − exp(−r (w − ψ(e)))

(6)

where w is manager’s compensation, r is the positive (Arrow–Pratt) coefficient of absolute risk aversion (−u /u ) and ψ(e) is the disutility from exerting effort, that for simplicity we consider quadratic: ψ(e) = e2 /2. The wage contract w is assumed to be linear. In addition to a fixed part, the manager’s remuneration has a variable part that is linked to the expected cost reduction. In particular, the expected value and the variance of w are respectively given by E (w) = α + β(e + µ(L)) and Var (w) = β 2 σ 2 (L). Given the normality of ε , maximizing expected utility is equivalent to maximize the certainty equivalent: ′′

′

e2

1 − r β 2 σ 2 (L). (7) 2 2 Thus, the agent’s certainty equivalent consists of the expected wage minus the disutility from exerting effort and minus a risk premium. CE = α + β(e + µ(L)) −

11 An exogenous shock in financial leverage may be due to reasons that are external to the firm. For instance, it may be related to changes in bankruptcy laws, tax codes, laws affecting investor protection, Merger & Acquisition activities, composition of the board of directors, etc. In Section 2, the source of exogenous variation in firm’s debt, which allows the identification of the coefficient of interest, is the collapse of the stock market tech bubble in 2000.

44

G. Papa, B. Speciale / Research in Economics 65 (2011) 36–46

The principal ensures participation of the agent by offering a contract with an expected utility of at least her reservation utility which, without loss of generality, we normalize to zero. The principal’s problem is then to solve: max E (Π (c )) − E (w)

(8)

α,β

subject to: CE ≥ 0

(9)

e∗ ∈ arg max α + β(e + µ(L)) −

e

2

2

1

− r β 2 σ 2 (L)

(10)

2

where Π is the profit function and e is the effort exerted by the agent. (9) and (10) are respectively the individual participation constraint and the incentive compatibility condition. From the IC condition we get e∗ = β ∗ , where β ∗ solves: max Π (K − β − µ(L)) − β

β2 2

1

− r β 2 σ 2 (L).

(11)

2

The FOC of (11) can be written as:

− Π ′ (K − β ∗ − µ(L)) − β ∗ (1 + r σ 2 (L)) = 0.

(12)

The SOC for a maximum is satisfied with the following additional assumption:

Π ′′ (K − β ∗ − µ(L)) < 1 + r σ 2 (L).

(13)

The above condition for instance holds if the function Π is concave in its argument c, which implies decreasing returns to cost reduction.12 More generally, it may hold for relatively high values of r and/or σ 2 (L). By implicitly differentiating the first order condition, we can study the effect of an increase in financial leverage on the optimal piece-rate β ∗ :

µ′ (L)Π ′′ (K − β ∗ − µ(L)) − (β ∗ r σ 2′ (L)) ∂β ∗ =− . ∂L Π ′′ (K − β ∗ − µ(L)) − 1 − r σ 2 (L) ∂β ∗

Note that the SOC ensures that the sign of ∂ L is given by the numerator of the RHS term:

 sign

∂β ∂L



  = sign µ′ (L)Π ′′ (K − β ∗ − µ(L)) − β ∗ r σ 2′ (L) .

As discussed above, the regression analysis in Section 2 shows that financial leverage affects positively both TSR (with the exception of the 2% most levered firms) and the variance of TSR. This implies that we can assume a positive sign for both µ′ (L) and σ 2′ (L). It follows that concavity of the function Π would imply a negative effect of financial leverage on incentive pay. Instead, if Π ′′ (c ) > 0 (i.e., strictly increasing returns to cost reduction) then the sign of the relationship of interest is ambiguous and can be written as follows:

 sign

∂β ∂L



 = sign

 µ′ (L) β ∗r − . σ 2′ (L) Π ′′ (K − β ∗ − µ(L))

µ′ (L)

We define σ 2′ (L) as the ‘‘worthiness ratio’’ of financial leverage. A positive and high value of this ratio implies that the expected benefits (in terms of cost reduction) of higher financial leverage compensates adequately for the increased risk. ∂β ∗ The following result summarizes the effect of a change in L on the variable part of managerial compensation (i.e., ∂ L ):



µ′ (L)



Result 1. If the ‘‘worthiness ratio’’ of financial leverage σ 2′ (L) is higher (lower) than leverage raises (lowers) the optimal piece-rate.

β∗r , then an increase in financial Π ′′ (K −β ∗ −µ(L))

There are two effects of a change in financial leverage on β ∗ . On the one hand, an increase in L changes the marginal value of the agent’s effort for the principal. This first effect on executive remuneration is in principle ambiguous and depends on µ′ (L).13 In our context, the data show a positive sign of µ′ (L), i.e., a positive association between financial leverage and total shareholder returns, for almost all the firms in our sample. This effect may therefore explain the increasing part of

12 As stated above, the cost-reducing investment in this theoretical illustration may be related to R&D activities. 13 This effect is analogous to the value of a cost reduction effect analyzed in the theoretical literature on the impact of product market competition on managerial incentives (Hermalin, 1992; Schmidt, 1997; Graziano and Parigi, 1998; Raith, 2003).

G. Papa, B. Speciale / Research in Economics 65 (2011) 36–46

45

the relationship between financial leverage and managerial pay found in Section 2. On the other hand, higher financial leverage implies more uncertainty. An increase in L makes the observed cost reduction a noisier signal of effort. This has been confirmed by the estimation of (3), which suggests that the variance of total shareholder returns is increasing in financial leverage. This effect should tend to lower the optimal piece rate. It is similar to the one proposed by Holmström and Milgrom (1987) and Aggarwal and Samwick (1999), except that in our simple theoretical exercise the increase in uncertainty comes from higher financial leverage. Combining the two effects, the relationship between incentive pay and financial leverage may be non-monotonic. It depends on the condition given in Result 1. More precisely, if the second order condition (13) holds, then this theoretical   µ′ (L)

exercise shows three mutually exclusive cases that depend on how large the ‘‘worthiness ratio’’ of financial leverage σ 2′ (L) is: ∂β β∗r µ′ (L) (1) ∂ L > 0 for ∀L iff σ 2′ (L) > Π ′′ (K −β ∗ −µ(L)) for ∀L. ∂β

µ′ (L)

β∗r

(2) ∂ L < 0 for ∀L iff σ 2′ (L) < Π ′′ (K −β ∗ −µ(L)) for ∀L. β∗r µ′ (L) µ′ (L) (3) An inverted-U relationship between β and L iff ∃!L such that σ 2′ (L) > Π ′′ (K −β ∗ −µ(L)) for ∀L < L and σ 2′ (L) < β∗r for ∀L > L. Π ′′ (K −β ∗ −µ(L)) Financial leverage may therefore increase (decrease) pay-for-performance sensitivity at any level of L, as case 1 (case 2) suggests. Case 3 shows that financial leverage and pay-for-performance sensitivity may also have a non-monotonic relationship, in line with the estimation results. 4. Concluding remarks In this paper, we have studied the effects of financial leverage on managerial pay. We have estimated a panel of quoted firms from the UK, ranging from 1995 to 2002. The change in investor’s expectations that caused the collapse of the stock market tech bubble after the year 2000 has been used to capture a plausibly exogenous variation in firm’s debt. Our estimates show that pay-for-performance sensitivity is increasing in financial leverage, with the exception of the most levered firms. A similar relationship has been found between financial leverage and total short-term compensation. Our results indicate that in firms with high levels of financial leverage it is less easy to create an incentive scheme that aligns managerial and shareholder interests. In the presence of high levels of firm’s debt, total shareholder returns may become a noisier signal for managerial effort. Our estimates show that on average this negative effect of firm’s financial leverage concerned UK companies with long term debt larger than 34% of total assets (the 10% most levered firms in our sample). Our work also sheds some light on possible reasons why during periods of crisis financial leverage tends to grow together with variable pay.14 During the 2008 financial crisis, the fact that banks which had to be bailed out with tax payer’s money continued to pay high bonuses to their managers was at the center of a harsh debate. In periods of stock market bust, the financial leverage/total assets ratio tends to increase. Since this may raise the volatility of the payoff from the manager’s effort, shareholders have to increase the variable part of the managerial remuneration to provide enough incentives to their manager whose utility is decreased because of the higher volatility. Finally, a caveat of this study is our restricted focus on the effects of financial leverage on short-term compensation (basic salary plus bonus). Due to data availability, our paper has followed most of the other empirical work on managerial remuneration in the UK that use cash compensation alone as dependent variable. Recent corporate reforms in the UK are generating more complete and detailed information on stock options, pension arrangements and long-term incentive plans, which represent a small percentage of total remuneration for the British CEOs, especially if compared to the long-term compensation of the American CEOs (Conyon and Murphy, 2000). The availability of these data will allow future research to examine the relationship between firm’s debt and long term compensation in the UK. References Aggarwal, R.K., Samwick, A.A., 1999. The other side of the trade-off: the impact of risk on executive compensation. Journal of Political Economy 107, 65–105. Bisin, A., Gottardi, P., Rampini, A.A., 2008. Managerial hedging and portfolio monitoring. Journal of the European Economic Association 6, 158–209. Black, F., Scholes, M., 1973. The pricing of options and corporate liabilities. Journal of Political Economy 81, 637–659. Conyon, M.J., 1997. Corporate governance and executive compensation. International Journal of Industrial Organization 15, 493–509. Conyon, M., Murphy, K., 2000. The prince and the pauper? CEO pay in the US and in the UK. Economic Journal 110, F640–F671. Cook, R.D., 1977. Detection of influential observations in linear regressions. Technometrics 19, 15–18. Core, J., Guay, W.R., 2002. The other side of the trade-off: the impact of risk on executive compensation: a revised comment. Mimeo. Core, J.E., Holthausen, R.W., Larcker, D.F., 1999. Corporate governance, chief executive officer compensation, and firm performance. Journal of Financial Economics 51, 371–406. Croxson, B., Propper, C., Perkins, A., 2001. Do doctors respond to financial incentives? UK family doctors and the GP fundholder scheme. Journal of Public Economics 79, 375–398.

14 Looking at the US financial sector, Philippon and Reshef (2008) find that financial jobs were relatively highly paid until the 1930s and after the 1980s, but not in the interim period. They show that financial deregulation and corporate activities linked to IPOs and credit risk tended to increase the demand for skills in financial jobs and consequently the wages in the financial sector.

46

G. Papa, B. Speciale / Research in Economics 65 (2011) 36–46

Cuñat, V., Guadalupe, M., 2005. How does product market competition shape incentive contracts? Journal of European Economic Association 5 (3), 1058–1082. Gilson, S., Vetsuypens, M.R., 1993. CEO compensation in financially distressed firms: an empirical analysis. Journal of Finance 48, 425–458. Girma, S., Thompson, S., Wright, P.W., 2006. The impact of merger activity on executive pay in the United Kingdom. Economica 73, 321–339. Graham, J.R., 2000. How big are the tax benefits of debt? Journal of Finance 55 (5), 1901–1941. Graziano, C., Parigi, B.M., 1998. Do managers work harder in competitive industries? Journal of Economic Behavior & Organization 34, 489–498. Green, R.C., 1984. Investment incentives, debt, and warrants. Journal of Financial Economics 13, 115–136. Grossman, S., Hart, O., 1982. Corporate financial structure and managerial incentives. In: McCall, J. (Ed.), The Economics of Information and Uncertainty. University of Chicago Press, Chicago. Hadi, A.S., 1992. Identifying multiple outliers in multivariate data. Journal of the Royal Statistical Society 54, 761–771. Hall, B.J., Liebman, J.B., 1998. Are CEOs really paid like bureaucrats? Quarterly Journal of Economics 113, 653–691. Hall, B.J., Murphy, K.J., 2000. Optimal exercise prices for executive stock options. American Economic Review 20 (2), 209–214. Hermalin, B.E., 1992. The effects of competition on executive behavior. Rand Journal of Economics 23, 350–365. Hermalin, B.E., Wallace, N.E., 2001. Firm performance and executive compensation in the savings and loan industry. Journal of Financial Economics 61, 139–170. Holmström, B., Milgrom, P., 1987. Aggregation and linearity in the provision of intertemporal incentives. Econometrica 55 (2), 303–328. Jensen, M.C., 1986. Agency cost of free cash flow, corporate finance, and takeovers. American Economic Review 86, 323–329. Jensen, M.C., Meckling, W., 1976. Theory of the firm: managerial behavior, agency costs and ownership structure. Journal of Financial Economics 3, 305–360. Jensen, M.C., Murphy, K.J., 1990. Performance pay and top-management incentives. Journal of Political Economy 98 (2), 225–264. John, T.A., John, K., 1993. Top-management compensation and capital structure. Journal of Finance 48, 949–974. Meyer, B.D., 1995. Natural and quasi-experiments in economics. Journal of Business & Economic Statistics 13 (2), 151–161. Morellec, E., 2004. Can managerial discretion explain observed leverage ratios? Review of Financial Studies 17, 257–294. Myers, S.C., 1984. The capital structure puzzle. Journal of Finance 39, 575–592. Ortiz-Molina, H., 2007. Executive compensation and capital structure: the effects of convertible debt and straight debt on CEO pay. Journal of Accounting and Economics 43 (1), 69–93. Philippon, T., Reshef, A., 2008. Wages and human capital in the US financial industry: 1909–2006. Mimeo. Stern School of Business. New York University. Prendergast, C., 2002. The tenuous trade-off between risk and incentives. Journal of Political Economy 110 (5), 1071–1102. Raith, M., 2003. Competition, risk and managerial incentives. American Economic Review 93, 1425–1436. Rajan, R.R., Zingales, L., 1995. What do we know about capital structure? Some evidence from international data. Journal of Finance 50, 1421–1460. Schmidt, K., 1997. Managerial incentives and product market competition. Review of Economic Studies 64, 191–213. Serfes, K., 2005. Risk sharing vs. incentives: contract design under two-sided heterogeneity. Economics Letters 88 (3), 343–349. Wright, D.J., 2004. The risk and incentives trade-off in the presence of heterogeneous managers. Journal of Economics 83 (3), 209–223.