Executive pay, tournaments and corporate performance in UK firms

June 2001

Executive pay, tournaments and corporate performance in 1 UK firms Martin J. Conyon and Graham V. Sadler2 This paper explores the relationship between executive pay and corporate performance. First, we focus on the pay-for-performance sensitivity (PPS), review recent evidence (mainly UK) and outline a methodology for calculating the PPS itself. Secondly, we review the evidence on tournament theory. Tournament models predict that career concerns generate incentives for executives and can explain observed variations in pay outcomes in the boardroom. Thirdly, we provide evidence on the distribution of the PPS for 532 executives within 100 large UK stock market companies for 1997. Unlike prior work, we include non-CEO executives in the analysis. We show that the PPS increases through organizational levels. Also the statistic is not constant across firms. Finally, we consider the relationship between corporate performance and incentives. We show, consistent with prior evidence, that there is a positive relationship between firm performance and the effective ownership of stock-based compensation by management.

Martin J. Conyon is from The Wharton School, University of Pennsylvania, Philadelphia, USA. Graham V. Sadler is from Warwick Business School, University of Warwick, Coventry, UK.

ß Blackwell Publishers Ltd 2001, 108 Cowley Road, Oxford OX4 1JF, UK and 350 Main Street, Malden, MA 02148, USA

1 Introduction The determination of executive compensation has emerged as an issue of considerable academic importance. Some of the recent UK research on executive compensation includes Benito and Conyon (1999), Conyon (1998), Conyon and Peck (1998b), Cosh and Hughes (1997), Gregg et al. (1993) and Main et al. (1996). The more voluminous work from the US includes Aggarwal and Samwick (1999), Garen (1994), Hall and Liebman (1998) and Jensen and Murphy (1990a).

International Journal of Management Reviews Volume 3 Issue 2 pp. 141–168

Recent reviews and collected works examining executive pay issues are provided by Hallock and Murphy (1999) and Murphy (1999). A central theme within the UK debate is whether executive pay is adequately related to measures of company performance. This paper explores this relationship and makes a number of contributions to the existing literature. First, we focus on the pay–performance sensitivity (PPS), review recent evidence and outline a methodology for calculating the payfor-performance term itself. Secondly, we

141

Executive pay, tournaments and corporate performance in UK firms

ß Blackwell Publishers Ltd 2001

142

review evidence on tournament theory. Tournament models predict that career concerns generate incentives for executives and can explain observed variations in pay outcomes in the boardroom. Thirdly, we document new empirical evidence on the distribution of the pay-for-performance link within UK stock market companies. Prior UK evidence has only considered the PPS for the CEO. We extend the analysis to non-CEO executive directors. Finally, we consider the relationship between corporate performance and incentives, demonstrating how incentives may affect managerial behaviour by promoting higher levels of effort and subsequently increase firm performance. We outline some issues in conducting this experiment. We also provide empirical evidence on the relationship between measures of corporate performance and the stock-based reward structure for executives in UK firms. The paper is organized as follows. Section 2 considers the relationship between executive compensation and firm performance. Most of the applied executive compensation literature focuses on the strength of the relationship between executive pay and company performance. The motivation for this research agenda is pretty clear. Shareholders cannot directly observe the activities of management and so a potential opportunity arises for managers to pursue their own interests. An incentive contract that relates management pay to performance, though, may ensure that managers do indeed promote shareholder wealth maximization. We present a simple theoretical model that illustrates this. Also in Section 2, we review applied work on the pay-for-performance link. The econometric and empirical evidence, until quite recently, suggested that the linkage between management pay and performance was quantitatively quite small and that the incentives for senior executives may be blunted. Many authors thus concluded that the small changes in executive wealth brought about by changes in firm performance were not consistent with agency models. However, we

also review recent contributions to the literature that have argued that the relationship has become stronger over time. Indeed, an important contributing factor to the growth in the pay-for-performance link is the central role of stock-based compensation such as share options. Section 2 concludes by illustrating how the pay-for-performance link can be calculated (rather than estimated) from information on executives’ holding of stockbased compensation (such as options, equity and long-term incentive plans). Section 3 focuses on an alternative stream of executive pay research, centred on tournament theory, which predicts executives may exert effort in order to be promoted to a better-paid job position. An important implication of this work is that the structure of compensation is central to understanding wage setting in the boardroom and executive incentives (Lazear 1995). The propositions arising from theoretical tournament models, though, have received far less empirical investigation. Indeed, much of the existing research pertains to sporting contexts such as golf, basketball and motor racing. There are relatively few papers that deal with the issue in a business context. We outline some of the predictions of tournament theory that have appeared in the literature and review existing evidence. Section 4 presents new evidence on the PPS within UK firms. Using data on 532 executives at 100 UK stock market companies in the fiscal year 1997/8, we calculate the PPS for each executive separately. Our data show that the PPS is not constant within firms. Importantly, we show that the median PPS increases as executives move up the organizational levels of the firm. One reason for this observation is that at the top of the organization, the incentives provided from career concerns (i.e. the possibility of future promotions) become blunted. To compensate for the lost opportunity of playing in a promotion tournament, greater financial incentives are required. This is consistent with tournament models.

In Section 5, we turn to the relationship between firm performance and incentives. We discuss how the provision of incentives may affect managerial behaviour and promote higher levels of effort and subsequently improve firm performance. We provide some empirical evidence on the relationship between measures of corporate performance and the executive reward structure of UK firms. Specifically, we show that there is a concave relationship between measures of firm performance (namely return on assets) and the PPS. This statistic measures the financial incentives that managers face to promote shareholder interests.

2 Pay for Performance This section deals with the PPS. In Section 2.1 we outline the basic principal–agent model. Section 2.2 considers estimating econometrically the pay-for-performance term and reviews evidence on the magnitude of the statistic. Finally, Section 2.3 deals with a procedure for directly calculating the pay-forperformance sensitivity/statistic.

2.1 A Principal±Agent Model This subsection summarizes the basic results of an incentive scheme within a principal– agent framework. Such theoretical models typically form the foundation of applied work in the executive compensation literature. However, a comprehensive review of this literature is beyond the scope of this paper. Appendix A provides a more formal derivation of a basic principal–agent model, and Kevin Murphy’s (1999) executive compensation review outlines key insights from contract theory for applied executive compensation research. The focus in this section, then, is on a linear reward structure where the main issue becomes the slope of the reward function. That is, how sensitive is pay to performance? The hidden-action (moral hazard) model assumes that the CEO takes action to produce

stochastic shareholder value and receives a payment based on shareholder value. The shareholders’ (principal) pay-off is a function of output, net of incentive pay to the agent. The CEO’s (agent) pay-off is a function of their incentive pay net of the cost-of-effort or action. The production function that relates the CEO’s effort to subsequent firm value is common knowledge, but the CEO’s effort level is private information to him/her. Shareholders know what levels of effort, or actions, they want the CEO to take but cannot directly observe the CEO’s effort. The problem facing the principal is to design a contract, subject to the imposed constraints of the agent’s optimizing behaviour. Two constraints need to be satisfied. The participation constraint (or individual rationality constraint) requires the pay-off to the agent must be at least as great as those presented by outside opportunities (i.e. the CEO’s reservation utility). Secondly, the incentive compatibility constraint requires that the contract offered to the agent lead him/her to choose the best selfinterested course of action. As Murphy (1999) notes, ‘‘The fundamental insight emerging from the traditional principal–agent models is that the optimal contract mimics a statistical inference problem; the payouts depend on the likelihood that the desired actions were in fact taken.’’ This is the so called ‘‘informativeness principle’’ introduced by Holmstrom (1979). It suggests that rewards to CEOs are based on stock-based measures of company performance, because these provide useful information in deciding what actions the CEO actually took in promoting shareholder value. In addition, the informativeness principle rationalizes the role of other performance measures in the CEO contract. Accounting measures of performance, such as return on assets, can be used to the extent that they provide valuable information that the CEO took the desired action. Also, the traditional principal–agent model provides other important insights into the nature of CEO compensation contracts. A

June 2001


143



144

particularly important one is the trade-off between risk and incentives. Appendix A shows that the optimal pay-for-performance term can be specified as follows:

1 1 rc00 a2

2:1

where is the pay-for-performance term, r is the agent absolute risk aversion parameter, c(a) is the agent cost of effort function, so c00 (a) is the slope of the marginal cost of effort, and 2 is the variability of firm value. Equation (2.1) says that the optimal pay-forperformance term is related to three factors (in more complicated models this optimality condition can cater for the productivity of the agent, relative performance evaluation and potential distortions in the performance term). The optimal PPS will be equal to one when output is certain (2 0) or when the agent is risk neutral (r 0). As the uncertainty in the firm value increases, and/or the risk aversion of the agent increases, the resulting optimal PPS declines. The intuition of the result is clear, when output is known with certainty and observed by the principal, then a one-to-one relation exists between managerial actions and rewards. Similarly, if the agent is risk neutral, in effect he willingly assumes the riskiness embodied within the firm’s assets, hence a pay–performance parameter of one. The agent cost-of-effort function enters in a perhaps less intuitive way (see Appendix A). If the slope of the marginal cost-of-effort function is zero, then the optimal sharing rate is one. For quadratic cost-of-effort functions, for example, ceteris paribus, the sharing rate will be less than one, reflecting the disutility of extra effort. The optimal pay-forperformance term is further discussed in Murphy (1999). Note, in this example the optimal depends on agent risk aversion r, the cost-of-effort function c00 (a) and risk (2). Since each company has a different CEO/agent, it is likely that there will be heterogeneity in each of these factors. Accordingly, it is most

unlikely that circumstances will conspire to promote a common across all agents (see Garen 1994). This point is important, since (as will be shown below) most researchers in the executive compensation literature retrieve an average pay–performance term using econometric techniques. This issue is addressed in the recent literature by Murphy (1999) and Conyon and Murphy (2000), where the distribution of the pay–performance term is considered by calculating across all agents individually using non-econometric methods. To summarize the model, the pay-off to a higher level of effort stochastically dominates that to a lower level. This pay-off, whose probability distribution is affected by the unobservable effort, is verifiable, however, and provides an enforceable argument in the optimal (but second best) contract set by the principal (see Hart 1995). Subject to the constraints imposed by ensuring the participation and the individual rationality of the agent, the argument focuses on defining the optimal contract or sharing rule.

2.2 Estimating (Econometrically) the PPS The principal–agent model described above predicts an optimal solution for the pay-forperformance term dependent on the risk aversion of the agent, the uncertainty of firm value and the function describing the cost-ofeffort to the agent. Empirical studies have thus attempted to estimate how sensitive executive compensation actually is to measures of company performance. The usual way in which such empirical models proceed is to estimate a simple reduced form equation rather than the parameters of a specific principal–agent model (see Conyon et al. 1995). A standard or typical regression equation would model the compensation of an individual director i at time t as log (Compensation)it

Performanceit it

2:2

where the term is the reaction coefficient reflecting the sensitivity of director compensation to corporate performance. The magnitude of the coefficient is interpreted as reflecting the operation of principal–agent-type mechanisms, with higher values of suggesting closer alignment of owner and management interests. The value of is thus an estimate of the term in the principal–agent model described in the previous section. An important feature of this modelling procedure is that by estimating in first differences (the term is a difference operator such that for any director i in period t then Xit Xit ÿ Xi;tÿ1 ) the estimate is free from company fixed effects bias (see Murphy 1985). The fixed effects bias may be thought of as arising from an omitted variable problem. Suppose that managerial ability/talent is thought to affect managerial pay and firm performance, but is difficult to observe or measure so is excluded from the original estimated pay equation. OLS estimates of are biased and difficult to interpret. Since ability and firm performance are positively correlated, the term can be the effect of ability on pay rather than how pay varies with changes in firm performance. If ability is presumed (relatively) constant over time, then an indicator variable for each CEO-manager effectively controls for their differences in talent. The resulting estimate of , then, reflects the performance effect on pay, rather than the fixed effect, which in this case is managerial ability (see Johnston and DiNardo 1997, Chapter 12, for extensions and limitations along with a discussion of panel data issues generally). There has been a certain amount of UK research estimating such models, but this contrasts with the much more voluminous US literature (see Bruce and Buck. 1997). So, what estimates of have been reported in the literature? The earlier US literature has frequently found the link between directors’ compensation and company performance to be small. In the widely cited analysis of US executives, Jensen and Murphy (1990b)

estimate that pay–performance relation (including pay, options, stockholdings and dismissal) to be $3.25 for every $1000 change in shareholder wealth. They concluded that such a value-for-pay–performance term was too low to be consistent with principal–agent theory. ‘‘We believe that our results are inconsistent with the implications of formal agency models of optimal contracting’’ (p. 227). Such a small PPS might be a matter of concern for shareholders and policy-makers, since the implied small private returns to CEOs for significant changes in shareholder worth implicitly questions the incentives for top management to pursue shareholder interests. The early UK evidence, too, suggests that directors’ compensation was only weakly related to company performance, i.e. early estimates of were small or insignificant (see Conyon et al. 1995). Before looking at the evidence in detail, however, it is important to stress some general features of the early UK data. First, the measure of compensation typically used in UK studies is a time series on the cash compensation of the highest-paid director. This contrasts with the relevant unit of analysis, which is the individual executive. This can cause problems for the estimated relationship between pay and performance, since the compensation time series may actually represent rewards to several different individuals. For instance, a large annual increase in the salary and bonus of the highest-paid director may reflect a recruitment payment (golden handshake) for a new CEO and not a pay rise to the particular individual who was the highest-paid director in the previous year. Secondly, there is the (potentially) controversial area of how exactly to measure director/executive pay. Until comparatively recently, most UK studies have used only the direct emoluments of the highest-paid director available from company accounts, or secondary sources such as Datastream and Hemmington Scott (the latter are electroni-

June 2001


145



146

cally stored, making the construction of firmlevel panels, i.e. where large numbers of companies are followed through time, relatively straightforward). This measures only current cash compensation, namely salary and bonus, but excludes long-term compensation such as the estimated value of share options and other forms of deferred compensation. Bruce and Buck (1997) argue that, by excluding these extra components of the directors’ overall compensation, the estimated relationship between compensation and performance may be biased (see also Conyon et al. 1995). The use of cash received as a measure of compensation does, however, have an empirical grounding. Lewellen and Huntsman (1970) reported that cash compensation represented an excellent proxy for total compensation (estimated as the sum of cash and the equivalent value of all deferred and contingent compensation) and even provided supe rior re sults whe n re gresse d on independent variables such as company size and profits. The structure of compensation, however, has dramatically changed since the Lewellen and Huntsman study. In the US, non-cash elements of pay are now often larger than the cash elements (Murphy 1999), and executive compensation packages in the UK are following a similar trend (Conyon and Murphy 2000). A second and more fundamental reason as to why the wider compensation measures have typically not been used in the UK has been a lack of available and consistent data. Finally, there is the question of how to measure company performance. Some empirical models use market-based measures of corporate performance such as shareholder returns or shareholder wealth, whereas others use accounting-based measures such as earnings per share or return on capital employed. It is not immediately apparent which is the correct performance measure to use, however, since principal–agent mechanisms stress returns to shareholders, a market-based measure reflecting share price

appreciation and dividend yield (i.e. total shareholder return) does seem more intuitive. Table 1 extends that provided by Conyon and Peck (1998a) and reports some recent UK evidence on the relationship between directors’ pay and company performance along with significant US studies. Some important general themes emerge. First, estimates of the pay-for-performance relationship in the UK are small, suggesting that incentives may not be very strong. Secondly, what statistical link between directors’ pay and corporate performance had been observed in UK companies appeared to have been decoupled in the period since 1989 (Gregg et al. 1993). By the early 1990s, one could not detect any significant relationship between the basic pay of UK executives and the stock market performance of their companies. Even allowing for the changing nature of compensation packages (i.e. towards more long-term performance pay in the form of stock options and other deferred mechanisms), Gregg et al. (1993) found little change in the estimate of . This contrasts with Benito and Conyon (1999). They found in a much larger sample of companies (in excess of 1000) between 1985 and 1994 that the link between executive pay and shareholder returns had become quantitatively larger. Main et al. (1996) were the first to produce a UK study that used a pay measure which incorporated the value of option grants. Their paper identified the change in the Black– Scholes value of options over a given year and added this to the cash compensation to yield total pay figure. Their sample was only 60 large UK firms over the period 1983–1989, but using this data they did find a much stronger relationship between pay and performance than had previously been reported. However, some immediate observations with their approach are noteworthy. First is the measurement of the compensation variable. Two distinct measures were used, one being the standard cash compensation measure of base pay plus bonus. The second measure is termed ‘‘total

Table 1. Some recent evidence on the pay±performance relationship Study

Data

Compensation measure

Performance measure

Estimated (standard error)

Remarks

Jensen and Murphy (1990b)

US data on 2213 CEOs, 1974±86

(1) Change in salary and bonus of CEO (2) Change in wealth (=salary+bonus+value of restricted stock+other benefits+present value of salary increment+change in value of options) of CEO

Change in shareholder return dated at (a) period t and (b) period tÿ1

1(a) 0.0000139 (0.0000017) 1(b) 0.0000080 (0.0000015) 2(a) 0.000176 (0.000034) 2(b) 0.000131 (0.000034)

Performance effects regarded as small

Main (1992)

512 UK companies 1969±89

Change in salary and bonus of highest-paid director

Stock market return

0.038 (0.012)

Gregg et al. (1993)

288 UK companies, 1983±91


Change in shareholder returns

1983±88: 0.027 (0.013) 1989±91: ÿ0.024 (0.022)

Effect of performance on compensation displays time heterogeneity. Disappears after 1988

Main and Johnston (1993) Conyon and Leech (1994) Conyon and Gregg (1994)

220 UK companies, 1990 294 UK companies, 1983±86 169 UK companies, 1985±90

Salary and bonus of highest-paid director Change in salary and bonus of highest-paid director Change in salary and bonus of highest-paid director

Risk adjusted market return Change in shareholder wealth Shareholder return

0.100 (0.135)

Cross section evidence

0.052 (0.020)

Effects of governance discussed Role of unions, mergers and financial structure on director compensation evaluated

Conyon (1995)

28 UK privatized companies, 1990±94


Return on shareholders' equity

0.0039 (0.0042)

Levels modelled, rather than first differences; fixed effects

Cosh and Hughes (1997)

44 UK companies in electrical engineering sector, 1989±94

Level and change in CEO pay

(1) Return on capital employed (2) shareholder return

(1) ÿ0.02 (0.5) (2) 0.11 (0.047)

Effects of shareholdings evaluated; relative performance considered

1985±87: 0.076 (0.032) 1988±90: 0.020 (0.036)

June 2001


147


148

Data

51 quoted UK companies, 1981±91

213 large UK companies, 1988±93


94 FT-SE 100 companies, 1991±94

478 US companies, 1980±94

1500 US companies, 1993±96

Study

Smith and Szymanski (1995)

Conyon (1997)

Main et al. (1996)

Conyon and Peck (1998b)

Hall and Liebman (1998)

Aggarwal and Samwick (1999)

Table 1 (continued)

(1) Change in salary, bonus and value of current option grants of CEO (2) The above plus change in value of option and equity holdings of CEO

(1) Change in salary and bonus of CEO (2) Change in salary, bonus and value of current option grants of CEO

Change in salary and bonus of highest-paid director in companies where (1) Proportion of outside directors on remuneration committee is above the median (2) Same proportion is below the median

Board and top directors' remuneration. (1) Salary and bonus (2) Total remuneration (including stock options)


Level of directors' remuneration including performance related pay, benefits and basic salary (for all directors)


Change in shareholder wealth

Shareholder return dated at (a) period t and (b) period tÿ1

Shareholder return

Share performance

Shareholder return

(1) Sales (2) Earnings per share

Performance measure

Model also considers other compensation measures and extends analysis to other executives

Option data derived by tracking option holdings through progressive proxy statements

(1a) 0.163 (0.012) (1b) 0.0596 (0.011) (2a) 0.280 (0.022) (2b) ÿ0.016 (0.024) (1) 0.432 (0.053) (2) 1.036 (0.313)

Data derived directly from annual reports. Board structure effects on pay evaluated. Outcome ambiguous.

Models include sector performance term and lagged dependent variable

Effects of board-room controls evaluated: outcomes ambiguous

Argue for the need to include effect of average executive pay as an `outside option'

Remarks

(1) 0.088 (0.047) (2) 0.033 (0.087)

For CEO: (1) 0.146 (0.113) (2) 0.729 (0.282)

0.061 (0.020)

Cross section (1) 0.43 (0.06) (2) 0.03 (0.10) Time series (1) 0.41 (0.20) (2) 0.03 (0.24)



Salary and bonus of highest-paid director (Hemmington Scott measure)

1093 quoted UK companies, 1985±94


40 small to medium sized companies 1985±92

510 UK firms and 1666 US firms in fiscal year 1997

Benito and Conyon (1999)

Conyon (1998)

Conyon and Nicolitsas (1998)

Conyon and Murphy (2000)

Implied change in CEO firm wealth. Includes stock options, equity, and LTIPs

Change in remuneration of the highest-paid director

Change in the salary and bonus of highest-paid director


Data

Study

Table 1 (continued)

(1) Time series heterogeneity in considered. It increases over time (2) Examine board governance effects for a sub-sample of 211 companies and find little direct effects (1) Dynamic panel data estimation (2) Effects of boardroom controls evaluated: outcomes ambiguous (3) CEO turnover models also estimated (1) Dynamic panel data estimation (2) Controls for CEO turnover (1) Models and compares levels of pay in the US and the UK (2) Models CEO incentives (3) CEO turnover models also estimated

Fixed effects: (1) 0.067 (0.020) Random effects: (1) 0.076 (0.019)

0.067 (0.025)

(1) Profits per employee 0.0026 (t=0.87) (2) Sales growth 0.245 (1.17) (1) UK mean 2.33%; median 0.25%. (2) US mean 4.18%; median 1.48%

(1) Shareholder return (2) Relative stock price performance

Shareholder return

Shareholder wealth

(1) Profits per employee (2) Sales growth

Remarks


Performance measure

June 2001


149



150

remuneration’’, which they define as ‘‘the sum of the emoluments and any change in value that year in the Black and Scholes ‘cashequivalent’ value of option holdings’’ these two measures were applied to three categories of employee, the highest-paid director, the CEO and the total board. The total remuneration measure more accurately reflects the change in firm-specific wealth owned by the CEO. For instance, other authors (e.g. Conyon and Murphy 2000; Murphy 1999; Yermack 1995) distinguish between annual pay which includes the value of current grants of options and wealth effects which refer to the change in value of the whole portfolio of options held. In essence, the econometric method used by Main et al. estimates a panel data econometric model equivalent in form to Equation (2.2) above. The compensation variable can be alternatively salary and bonus or total remuneration, yet they still estimate a single common, average pay-for-performance term ( ). However, as discussed earlier, the payfor-performance term is likely to display cross-sectional, i.e. firm, heterogeneity. This is due to the fact that the optimal (in a simple principal–agent model) depends on the (second derivative) agent cost-of-effort function, variability of firm wealth and agent risk aversion, all of which are likely to vary over different agents and companies. So, in common with much of the US and UK literature Main et al. provide an average estimate of the pay–performance term. However, recent US and international comparisons have calculated directly the pay–performance term for each CEO separately. Examples are Conyon and Murphy (2000), Jensen and Murphy (1990b), Murphy (1999) and Yermack (1995). This observation may be important if the distribution of pay-forperformance sensitivities is non-normal. Indeed, our new results below shows marked differences between the calculated mean and median pay-for-performance sensitivities. The Main et al. (1996) results are easily summarized. The mean (median) salary and bonus compensation of the highest-paid

director in 1989 was £223,000 (£165,000). In contrast, the mean (median) total remuneration of the highest-paid director was £317,000 (£199,000). The econometric results based on the dynamic panel of companies revealed a statistically significant relationship between cash compensation and current dated share price. Similarly, there is a significantly positive relationship between total remuneration and current dated share price. Specifically, their estimated models forecast that a 10% increase in stock returns yields a 2.25% increase in salary and bonus. This translates into an £8018 increase on the median highest-paid director salary and bonus in 1989 of £357,000. The specification that uses total remuneration as the dependent variable forecasts that a 10% increase in stock returns yields an 8.94% increase in total remuneration, translating into a £50,600 increase on the median highest-paid director total remuneration of £566,000. These results, the authors suggest, demonstrate ‘‘a more robust connection between executive pay and performance in British firms than has hitherto been reported’’ (Main et al. 1996, 1641). To recap, much of the evidence from early empirical work in both the US and the UK concluded that there was no link between direct executive compensation and the stock market performance of their companies. Even where a link had been identified, its magnitude seemed to be extremely small (Gregg et al. 1993; Jensen and Murphy 1990b) and thus seemed to offer little support to principal– agent model. Haubrich (1994), however, argued that, despite being small, given the risk aversion of CEOs, such estimates of the pay-forperformance term could still indeed be consistent with the predictions of agency theory. Garen (1994) empirically tested the principal–agent model by examining whether CEOs’ stock-related compensation is decreasing in the standard deviation of firm returns and whether CEOs’ salary-based compensation is increasing in the standard deviation of the firm returns. Using the Jensen and

Murphy (1990b) sample of 430 US firms in 1988, Garen does find weak evidence in support of these propositions, although none of his regressions provides a statistically significant coefficient on the standard deviation of firm returns variable. More recent studies, however, have reported results that lend even greater support to the theoretical model. Results presented by Aggarwal and Samwick (1999), for example, strongly support the principal–agent model. They use the variation in stock return volatility across firms to test whether executives at riskier firms have lower pay–performance sensitivities, as is predicted by most principal– agent models. They find that the PPS of a manager’s compensation is decreasing in the variance of the firm’s returns and that the PPS for executives at firms with low stock price volatility is an order of magnitude greater than it is for executives at firms with highly volatile stock returns. Other more recent evidence (typically from the US) also suggests the pay–performance link may be becoming stronger. Murphy (1999) reports that pay–performance sensitivities in the US have nearly doubled between 1988 and 1996. This has been driven primarily by executive share options and direct equity ownership, with the author stating that 95% of the estimated 1996 PPS for CEOs in manufacturing companies comes from options (64%) and equity (31%). Furthermore, Murphy reports an inverse relation between company size and PPS. This can be intuitively explained, since the CEOs of large companies tend to own a smaller proportion of their company through shares and options. While increases in the PPS would suggest a lessening of agency problems, Murphy (1999) also offers evidence which suggests that agency problems might be increasing. This stems from the observation that, although the value of shares held by S&P-500 CEOs has increased substantially over the past decade, the percentage of outstanding equity held by such CEOs has

been declining and it is the percentage ownership, rather than the absolute value of shareholdings that indicates the severity of the agency problem. Several other authors have also documented recent increases in the PPS. Hall and Liebman (1998) used data from 478 companies over a period of 15 years from 1980 to 1994 and further concluded that there was a strong relationship between firm performance and CEO pay. Their method differed from most previous studies because, in a similar vein to Main et al. (1996), they constructed a pay measure that included changes in the value of the stock of equity and options held by the CEO. Excluding these elements, they find similar although slightly larger pay–performance sensitivities to that of Jensen and Murphy. But when these elements are included, their elasticity of pay estimates are some 30 times larger than previously reported elasticities. In terms of dollar returns, Hall and Liebman estimate median and mean values of the PPS at $5.29 and $25.11 per $1000 in 1994, compared with Jensen and Murphy’s median estimate of $3.25. Aggarwal and Samwick (1999) find even larger pay–performance sensitivities. Using their sample of 1500 companies over the period 1993–1996, they construct a pay measure that, like Hall and Liebman, includes the change in the market value of the executive’s holdings of shares and options as well as salary, bonus and the value of any grants of options or other long-term incentives. Based on this measure of pay, they find median and mean dollar pay–performance sensitivities of $14.52 and $69.41 per $1000; even excluding the revaluation of option holdings the respective figures are $6.59 and $58.61. Note that in addition to rewarding managers based on their own company performance, the owners of a company may wish to make pay dependent on performance relative to that of other companies operating in the same industry or sector (see Holmstrom 1982; Nickell 1995; Tirole 1988). The idea is

June 2001


151


simple. In the model outlined above, owners want to reward effort, but can only observe the output. Some shocks, though, are common to the industry or sector as a whole (e.g. industry profits may fall independent of the actions of the manager). To control for such shocks, the owner of the company merely looks at the profit outcome of the company relative to other firms in the same industry. One would expect then to see managerial pay not only directly related to company performance but also to the performance relative to other companies (see Gibbons and Murphy 1992).

2.3 Calculating the PPS Much of the evidence contained in Section 2.2 is based on estimating the pay-for-performance parameter from a reduced form panel data equation. In contrast, Murphy (1999) and Conyon and Murphy (2000) note that the payfor-performance parameter can be calculated directly. This is because the main elements that go to make up the PPS are holdings of stock-based compensation by executives rather than cash emoluments. The PPS can be written simply as: PPS

Shares Held as % of Firm Shares


152

Options Held as % of Firm Shares

LTIP Shares as % of Firm Shares

LTIP Delta

Option Delta

(2.3)

Equation (2.3) decomposes the aggregate payfor-performance term into three elements (see Conyon and Murphy 2000). Data on each of these elements can typically be collected for each UK executive under new reporting requirements. However, by far the most demanding in terms of data collection are the necessary inputs to calculate the value of stock options and the associated option deltas (see below). The intuition and explanation of Equation (2.3) is straightforward. The first element recorded on the right-hand side of the equation is the incentive from holding ordinary equity. An executive holding 1% of common equity receives £1 of a £100 increase in shareholder

wealth. Conyon and Murphy (2000) refer to this as the ‘‘sharing rate’’ or ‘‘effective ownership’’ arising from a change in shareholder wealth that is translated into executive equity wealth. The second element on the right-hand side of Equation (2.3) is the stock option pay– performance term. Unlike equity, the option sharing rate is not simply the percentage of outstanding options on common equity, since the change in value of share options is not one-for-one with a change in the share price (see Appendix B). The actual increase in the option value for a unit increase in the share price is termed the delta () of the option, and is determined by the derivative of the Black– Scholes option value with respect to the price of the underlying asset [for a European call on a share paying continuous dividends, the option delta is given as: eÿqT N d1 ]. The option delta varies between zero and one (see Cox and Rubenstein 1985). Deep in the money options (that is where the share price is considerably in excess of the exercise price) have deltas that are close to unity whereas deep out of the money have deltas close to zero. CEOs who only hold deep out of the money options will, independent of the fraction of options on outstanding equity held, have low pay–performance sensitivities. The final element recorded in Equation (2.3) is the incentive from long-term incentive plan (LTIP) grants. UK LTIP shares are potentially subject to forfeiture if certain employment and/or performance objectives are not achieved. Thus, as with options, the LTIP sharing rate is weighted by an equivalent ‘‘LTIP Delta’’, which represents the change in value of each LTIP share from an incremental change in the share price. Calculating analytical UK LTIP deltas for performancecontingent grants is potentially complicated and, as a simplification, we assume here an LTIP delta of one independent of performance-vesting contingencies. The evidence contained in Murphy (1999) has indicated that the PPS in the USA has increased over time and that this is largely due

to the direct relationship between pay and performance arising from stock-based compensation. Conyon and Murphy (2000) provide a comparison of differences in CEO pay and incentives in the US and the UK for the fiscal year 1997. They show that, after controlling for size, sector and other firm and executive characteristics, CEOs in the US earn 46% higher cash compensation and 190% higher total compensation (including share options, etc.). Stock-based incentives, too, are higher in the US. The calculated PPS for the US sample of about 1500 firms implies that the median CEO receives 1.48% of any increase in shareholder wealth. This compares with a value of 0.25% in the sample of 510 UK firms. They attribute the differences, which are interesting given the similarity of the economies and corporate governance structures, to greater share option awards in the US arising from institutional and cultural differences between the two countries.

3 Tournament Theory and Incentives The focus of Section 2 has been the principal– agent model and the associated literature investigating the link between company performance and executive compensation. Although more recent research in the US, has suggested a strengthening of the pay– performance link (Aggarwal and Samwick 1999; Hall and Liebman 1998; Murphy 1999), as detailed above, most early studies suggested this link was neither strong nor consistent (Conyon et al. 1995; Jensen and Murphy 1990b). Thus in an attempt to align economic theory and empirical reality further, economists proposed an alternative theory of executive pay known as tournament theory. This theory, initially developed by Lazear and Rosen (1981), also tried to explain the large disparity between CEO pay and the pay of executives located one level down the organizational hierarchy. Lazear and Rosen comment that ‘‘On the day that a given individual is promoted from vice-president to president,

his salary may triple. It is difficult to argue that his skills have tripled in that one-day period, presenting difficulties for standard theory where supply factors should keep wages in those two occupations approximately equal. It is not a puzzle, however, when interpreted in the context of the prize’’ (p. 847). Lazear and Rosen suggest that, even though the salary of the top executive may well exceed all measures of their marginal product, it can still be economically efficient. The justification is that the high salary of the CEO acts as an incentive to those on lower management levels to accept wages at less than their own expected marginal product. The underlying theme of tournament theory, then, is that agents will exert effort in order to get promoted to a higher position in the management hierarchy associated with which is a higher level of compensation. Individual agents thus compete with each other, increasing their effort, in an attempt to increase the likelihood of winning the prize of promotion. In this framework, it is of course only relative performance that is of importance. As in a competitive sports game, agents need only be concerned that they outperform their rivals and not with their absolute level of performance.

June 2001

3.1 A Tournament Model For simplicity, consider a simplified version of the Lazear–Rosen tournament model. A firm has two employees and two job slots (supervisor and subordinate). The two individuals compete with each other, the loser gets the employee job while the winner becomes the supervisor. The overall prizes are fixed in advance, with the winner receiving W1 and the loser W2. The probability of winning the contest depends on the level of effort that each contestant exerts, together with a random shock component. Denoting the individuals as j and k then: qj j aj

qk k ak ß Blackwell Publishers Ltd 2001

153


where qj and qk are individual outputs, j and k are the effort levels of the respective individuals, and aj and ak are the white noise chance factors. Dealing with the individual labour supply decision first, each individual wants to maximize their expected pay-off. Looking first at employee j, the problem then is

Max j : PW1 1 ÿ PW2 ÿ Cj where P is the probability of j winning and C(j) is a convex cost-of-effort function, i.e. the monetary value associated with a particular level of effort. The first-order condition for j is thus:

W1 ÿ W2 @P=@j ÿ @C=@j 0

3:1

There is of course a corresponding problem for agent k. Employee j wins the contest if he/ she produces more output than employee k, that is, j wins if qj > qk. The probability that j wins is therefore given by:

P probqj > qk prob ak ÿ aj < j ÿ k probj ÿ k > ak ÿj Gj ÿ k where G is the distribution function on the random variable akÿaj. Also, note that @P= @j @Gj ÿ k =@j gj ÿ k . H o w ever, since individuals j and k are ex ante identical, there is a symmetric Nash equilibrium where j and k choose the same effort level, thus jÿk 0, and so Equation (3.1) can be written:

W1 ÿ W2 g0 @C=@j

3:2

Now consider the optimal wage chosen by the firm, given the labour supply decision characterized by Equation (3.2). Lazear (1995, 30–31) demonstrates that the average wage necessary to attract employees to the firm and the optimal wage spread are given by

W1 ÿ W2 =2 C W1 ÿ W2 1=g0 ß Blackwell Publishers Ltd 2001

154

Tournament models, together with Equation (3.2), have some testable implications. First,

an increase in the wage spread, W1ÿW2, implies a higher equilibrium effort, since C() is convex. So a bigger rise in the pay gap will induce workers to compete harder for promotion. Furthermore, the absolute level of the prize does not affect effort, if both prizes rise by the same amount, so the prize differential remains unchanged, then the effort level will also remain unchanged. The intuition is that the value of winning is not only the prize at that level, but also the possibility to compete for larger prizes at higher levels. However, the higher up the organizational hierarchy the individual moves, the smaller the opportunity for promotion becomes, since there are fewer positions to move into. One substitute for the loss of the chance to compete further is higher current compensation. In consequence, tournament models predict that compensation will be an increasing function of organizational level (Lambert et al. 1993; Main et al. 1993). Indeed, Rosen (1986, 701) comments: ‘‘The extra weight of rewards at the top is due to the no tomorrow aspects of the final stage of the game.’’ Thus tournament models predict that there is a convex relationship between executive compensation and organizational level. Tournament theory has not gained universal acceptance though. Dye (1984), for example, provides a comprehensive critique of tournament theory raising doubts about several features, including the feasibility of constructing appropriate handicaps, the difficulty of judging multidimensional performance in an ordinal sense and the problems of collusion and sabotage among contestants under such arrangements. Baker et al. (1988) also question the wisdom of using promotions as an incentive device, pointing to the costs of promoting an individual with skills inappropriate to the promoted post. Tournament models also predict that the tournament prize is increasing in the number of competitors (see Eriksson 1999; Lambert et al. 1993; Main et al. 1993; O’Reilly et al. 1988; Prendergast 1999). Each tournament participant implicitly gives up some of the

expected salary associated with his marginal product or performance. This excess then becomes part of the overall tournament prize. As O’Reilly et al. (1988, 261) remark: Given this fact, then it should follow that, in general, the more players in the tournament, the larger the prize should be. In an organizational context, this should mean that, after controlling for other possible economic determinants of CEO compensation, the more vice presidents, the larger should be the observed gap between the CEO’s salary and bonus and those of the vice presidents.

This provides a second testable proposition, namely: Tournament models predict that the tournament prize (gap) and the number of contestants are positively correlated.

3.2 Tournaments: Some Empirical Evidence On an empirical level, there have been few tests of tournament theory in general and only a handful in the context of executive compensation. One such study, however, is that carried out by O’Reilly et al. (1988). In their study, O’Reilly et al. tested the hypothesis that the larger the number of candidates competing for a CEO position, represented by the number of vice-presidents, the greater would be the disparity in pay between the CEO and other executive levels. However, although they did report a statistically significant result, it was in the opposite direction to that predicted by tournament theory. In a further test, the authors re-defined the boundaries for inclusion within the tournament. Rather than incorporating all executives, only those with significant responsibilities were included. It was proposed this was a fairer representation of those individuals who would be most likely to be involved in succeeding the CEO. However, with this refined sample, no statistically significant results were reported. In contrast, Main et al. (1993) do isolate a positive relationship between the number of

tournament participants and pay differentials. However, although finding results that were consistent with the operation of tournaments, they concluded that there was little ‘‘support for the empirical importance of consideration of pay equity at the top of corporations’’ (p. 606). Further support for tournament theory is provided by Lambert et al. (1993). Using internal firm data, they show that differences in compensation between hierarchical levels are consistent with tournament theory. In a more recent study using data on 2600 managers from 210 Danish firms during the four-year period 1992–1995, Eriksson (1999) concluded that ‘‘almost all of my findings are consistent with tournament models’’ (p. 241). He reports finding a positive relationship between the number of participants and the prize of the tournament and a stable convex relation between pay and job level. Conyon et al. (2001), too, test tournament predictions using data on 100 large stock market companies in the fiscal year 1997. They test the tournament prediction that the ratio of pay between adjacent levels increases as one moves up the hierarchical level (i.e. the relationship is convex). Their empirical results provide some confirmatory evidence of this. Moving from the level just below the CEO to the Group CEO job position generates approximately a 60% increase in pay. This compared with Main et al. (1993), who find that CEOs enjoy a level of pay that is about 140% greater than those at the next reporting level down. Also, they show that the premium for winning the tournament (i.e. the prize) is increasing in the number of executive directors, again consistent with tournament models. Although empirical research on tournament theory in a business context has been limited, strong support for it has been found in a sporting setting. Ehrenberg and Bognanno (1990), looking at professional golfers, and Becker and Huselid (1992), looking at professional NASCAR drivers, both report results in favour of tournament theory. A further example is Fernie and Metcalf (1996),

June 2001


155


who undertook an empirical test based on the pay and performance of an unbalanced panel of 50 jockeys over a period of eight years. Jockeys are usually paid a percentage of any winnings, and their opportunity to win, that is the number of rides they are offered, depends on their reputation and standing. This transparency of not only the pay but more importantly the performance of the jockeys made the pay–performance link much easier to observe. Fernie and Metcalf conclude that the existence of this, almost ‘‘ideal’’ payment system does improve the level of effort and hence the performance of the riders when compared with other non-performance-related compensation packages. Outside a sports setting, Bull et al. (1987) produced a laboratory study which used paid undergraduate student volunteers as subjects to test whether tournaments produced the desired effort responses, concluding that tournament theory might have some predictive validity. Furthermore, in a study on the broiler chicken industry, Knoeber and Thurman (1994) reported results that were in support of tournament theory. They also reported that farmers who were unlikely to win the tournament engage in riskier actions in an attempt to improve their chances.

4 The PPS within UK Firms


156

This section focuses on the financial incentives faced by executive directors – not simply the CEO. Building upon the discussion of the PPS and the role of tournaments, we illustrate how the PPS varies in a sample of 532 executives in 100 UK listed companies in the fiscal year 1997/8. Previous UK research has only examined the PPS for the CEO (i.e. the most senior job position). We, however, extend the analysis to include non-CEO executives. This allows us to document evidence on how the pay-for-performance varies within firms. As indicated, there has been no UK evidence on how the PPS varies within the firm at board level. There may be important reasons to suppose that the pay-for-

performance may vary within the company. In Section 3, we considered the predictions of tournament models. Prior empirical tournament research has typically demonstrated that (i) pay/compensation varies within the executive team, (ii) the relation between executive pay and organizational level is convex, and (iii) the prize (i.e. the premium for winning the tournament) may be increasing in the number of tournament competitors. However, just as tournament theory predicts a within-firm variation in compensation, such variation is also possible within the overall PPS of company directors. One reason for this is that at the top of the organization the incentives provided from career concerns (namely the possibility of future promotions) may become blunted (there are no future job slots to compete for). To compensate for the lost possibility of participating in the tournament, greater financial incentives may be required to promote effort. This suggests the empirical hypothesis that there is a positive relationship between the PPS and organizational level. A stronger hypothesis is that the relationship is convex. In this sub-section, we present empirical support for this idea. To test this hypothesis, we use data on 100 representative companies drawn from the 150 largest UK stock market companies in 1997/8. The data set used here is explained in Conyon et al. (2001). The 100 firms account for 63% of the market value of all companies on the London Stock Exchange and are allocated to six industry groups. The sample is representative of the top 150 in terms of sales, employment, market value and industry distributions. We needed data on stock options, LTIPs and equity holdings, along with the inputs necessary to compute the stock option delta as defined in the Appendix. This information is now available since the publication of the Greenbury (1995) report (however, the collection of the data is done by hand and is very time consuming!). The 100 companies have 532 executive directors for which we can compute the aggregate PPS for

each. The dependent variable in the analysis is the aggregate PPS computed for each of the 532 executives separately. The pay-forperformance variable is then regressed on organizational job classifications and various control variables. We allocated each of the 532 executives at the 100 companies to three distinct organizational levels using information from the annual report and accounts. For each executive director, we first identified his or her job title/role and then classified them into one of the following three separate levels. Level 1 Level 2

Level 3

Group CEO: the executive with highest authority in the firm. Divisional CEO: an executive with highest authority within a division. This includes CEOs, Presidents, Chairmen and Managing Directors of major divisions or business units. Finance Directors are also included in this level. Other executives: this includes executive directors with mainly functional duties. Their job titles/ roles include Company Secretary, Human Resource Director, Legal Director, Personnel Director, Corporate Affairs/Development Director, or Marketing Director.

The classification scheme is similar to Lambert et al. (1993). In the pay-forperformance regressions that follow, the excluded category is Level 3. The control variables that we include are size, executive age and industry dummy variables. Company size is measured as the log of market value. Murphy (1999) documents that CEOs at larger firms hold smaller percentage stakes of equity in their companies which need to be controlled for. We also control for industry differences in the executive PPS by including industry indicator variables. Finally, we control for executive age. The rationale for including this control is that older CEOs, closer to retirement, have fewer career concern incentives. Accordingly,

higher financial incentives may be required to compensate for loss of career incentives. The descriptive statistics (in terms of individual executive and firm-level variables) are contained in Table 2. The first three rows d o c u m e n t t h e PP S f or e a c h o f t h e organizational levels. The statistic provides a measure of how executive wealth varies for given changes in shareholder wealth (see Jensen and Murphy 1990b). The results indicate that the median PPS for CEOs is about 0.038 for group CEOs (Level 1), about 0.021 for divisional CEOs (Level 2) and approximately 0.018 for other executives (Level 3). This implies that the median group CEO receives under 0.5% of any increase in shareholder wealth. The average numbers reveal a different pattern though. They are 0.12 for Level 1, 0.26 for Level 2 and 0.24 for Level 3 executives. The distribution of the PPS variable, though, is skewed, with the mean PPS figure heavily influenced by outliers. In this data set there are non-group CEO executives with large shareholdings that inflate the mean PPS figures for Level 2 and Level 3 directors. The determinants of the PPS are shown in Table 3. The models are estimated using quantile regression methods. For instance, column (2) in Table 3 is the median (50th percentile) regression. The object in this case is to estimate the median of the dependent variable (i.e. the PPS) conditional on the values of the right-hand side independent variables. The procedure is analogous to the OLS estimator where the object is to estimate the mean of the dependent variable. However, given the skewed distribution of the PPS, we preferred to use quantile methods. The generalized quantile regression estimates a quantile other than the median. We report the 25th and 75th percentiles in columns (1) and (3) respectively of Table 3. The results indicate that, after controlling for firm size, executive age and industry effects, the coefficient of the group CEO indicator variable is positive and significant in each of the quantile regressions. This suggests that group CEOs (Level 1) have higher PPSs

June 2001


157


Table 2. Pay-for-performance sensitivities and firm characteristics No. of obs.

PPS Level 1: Group CEO Level 2: Divisional CEOs Level 3: Other executives Firm level variables Market Value (»m) Shareholder return Return on capital (%) Volatility Board size Total Board PPS Director level variable Age

Mean [SD]

Percentiles 10th

25th

50th

75th

90th

100 337 95

0.1199 [0.2941] 0.2646 [1.5375] 0.2351 [1.2918]

0.0108 0.0042 0.0056

0.0238 0.0098 0.0085

0.0378 0.0208 0.0178

0.0804 0.0422 0.0291

0.2090 0.1099 0.0712

100 100 100 100 100 100

8598 [11047] 0.2579 [0.2372] 26.19 [31.44] 0.2265 [0.0300] 11.2 [2.69] 1.2349 [4.4173]

1972 -0.0547 3.26 0.2000 8.5 0.0433

2593 0.0956 11.00 0.2000 9.0 0.0784

4737 0.2637 17.75 0.2182 11.0 0.1334

8881 0.4116 31.49 0.2482 13.0 0.3365

20595 0.5413 53.42 0.2742 14.0 1.0337

532

52.4 [6.68]

44.0

49.0

53.0

56.0

60.0

Notes: 1 The sample consists of 532 executives at 100 large UK stock market companies in 1997±1998. Market Value is Datastream MV variable recorded on 17/7/97. Shareholder return is log of the increase in the Datastream RI variable over the company's fiscal year. Return on Capital Employed is Datastream variable 707. Board size represents the number of directors serving on the board at the company's fiscal year end as recorded in the annual report and accounts. Volatility is estimated from stock returns over previous four years (but trimmed to lie in the range 0.2±6) 2 The pay±performance sensitivity (PPS) is defined as: Shares Held as Options Held as Option LTIP Shares as LTIP % of Firm Shares % of Firm Shares Delta % of Firm Shares Delta where 0 < Option Delta < 1 is the share-weighted-average slope of the Black±Scholes function at the year-end stock price, for options outstanding at the fiscal year end, and (LTIP Delta) = 1. The total board PPS is the sum of executive PPS for each company separately.

Table 3. Pay-for-performance sensitivity and organizational level at 100 large UK companies in 1997 LHS = PPS

Level 1: Group CEO Level 2: Divisional CEOs Log (total capital employed) Executive age Observations Industry dummies Pseudo R2


158

Quantile regression at 25th percentile



0.0127*** (0.0021) 0.0008 (0.0017) ÿ0.0076*** (0.0008) 0.0002** (0.0001) 532 Yes 0.0169

0.0200*** (0.0029) 0.0040** (0.0023) ÿ0.0110*** (0.0011) 0.0003** (0.0001) 532 Yes 0.0193

0.0352*** (0.0080) 0.0104 (0.0065) ÿ0.0221*** (0.0031) 0.0009** (0.0004) 532 Yes 0.0207

Notes: 1 The table contains results from quantile regressions performed at the 25th, 50th (median) and 75th percentiles. Analogous to an OLS estimator the median regression, for example, predicts the 50th percentile of the dependent variable. 2 The significance is **p < 0.05; ***p < 0.01. Standard errors reported in parentheses. 3 All regressions contain an unreported constant. 4 Levels 1 and 2 are dummy variables indicating that an executive is either the group CEO or the next level down, namely the divisional CEO. 5 The PPS for each executive is defined as in Table 2.

(i.e. greater financial incentives) compared with divisional (Level 2) CEOs. To test further the relationship between the pay-for performance statistic and organizational level, we performed further diagnostic tests. First, we estimated the specification in Table 3 using robust regression methods. This procedure begins by estimating the regression, calculating Cook’s D and excluding observations where D > 1. It then works iteratively, performing a regression, calculating weights based on absolute residuals and then regressing again using those weights until the changes in weights drop below the desired tolerance. A significant coefficient (standard error) on the Group CEO indicator variable of 0.158 (0.0025) was recorded. The Level 2 variable was not significant (standard OLS regression yielded insignificant estimates on both variables). Moreover, the relationship between the executive PPS is convex. We tested this using the procedure outlined in Lambert et al. (1993, 450). For the relation between the PPS and organization level to be convex, the coefficient estimate on Level 2 (Divisional CEO) must be positive. In addition, the difference between the coefficients on Level 1 (Group CEO) and Level 2 (Divisional CEO) must be greater than the coefficient on Level 2 (Divisional CEO). For example, in column (2) the test confirms the convexity proposition [F 11.14, p < 0.01] (since the coefficient on Level 2 job position is not significant in the 25th and 75th percentiles, we do not report the convexity test). Dealing with the control variables, we first note that the coefficient on firm size (measured as the log of market value) is significantly negative in each of the regressions, indicating that larger firms have lower pay– performance sensitivities. This is consistent with Conyon and Murphy (2000) and is a consequence of CEOs in larger firms owning a smaller fraction of the total outstanding equity of their firm than CEOs in smaller firms. We draw a number of inferences from Tables 2 and 3. First, the PPS is not constant

for executives within firms. In Section 2, we outlined a simple principal–agent model where the PPS was related to factors such as the agent’s risk aversion, variability of firm wealth and the slope of the agent’s marginal cost-of-effort. We conjectured that the PPS was unlikely to be the same for all agents. Indeed, the evidence here shows that there is significant variation in the PPSs across firms for a given job position and also within firms across job positions. Secondly, our results suggest that the median PPS is an increasing function of organizational level. Group CEOs appear to have greater financial incentives than other executives. This seems consistent with the tournament literature that argues that pay should increase as executives ascend the corporate hierarchy. The new results indicate that financial incentives also increase though the executive ranks. The increase in financial incentives (PPS) at the top of the organization may be due to the blunted options for further promotion possibilities.

June 2001

5 Corporate Performance and Incentives In this section, we turn to the relationship between corporate performance and incentives. We have seen that the issue of the magnitude of the PPS has been central in executive compensation research. An important underlying reason for this is the supposition that stronger links between pay and performance will result in higher managerial effort and ultimately translate into higher levels of corporate performance. However, as Murphy (1999) remarks ‘‘there is little direct evidence that higher pay–performance sensitivities lead to higher stock-price performance’’. In Section 5.1, therefore, we outline some of the difficulties in testing for a relationship between incentives and corporate performance. In Section 5.2, we highlight some of the existing evidence. Finally, in Section 5.3, we present some new evidence on the relationship between corporate performance and financial incentives (as measured by the PPS).


159



160

5.1 The Importance of Incentives Murphy (1999) argues that one of the main reasons for the lack of any direct evidence on the relationship between corporate performance (measured by firm returns) and incentives is due to the efficient capital market hypothesis. Since information about management’s PPS is public, all such information is already contained in the stock price. In companies that have superior managerial incentives, the information that this signals is instantly captured in the new (presumably higher) share price. Similarly, in companies with inferior managerial incentives, such information will have been instantly captured in the new (presumably lower) share price. A research experiment that measures the payfor-performance sensitivities at a given point in time and then examines the subsequent shareholder returns over subsequent years will thus fail to find a relationship between average returns based on the initial holdings of stock (see Murphy 1999). This experimental design problem, arising from the efficient capital market hypothesis, may be obviated by event study analysis. The efficient market hypothesis predicts that the effect of increased incentives, for example by the increased ownership of stock or the innovation of a new stock-based compensation plan, will be instantly reflected in the stock price upon announcement. A test, then, of the importance of incentives is the stock-price reaction to the announcement of these events. The general problem here is the precise timing of the events. When the adoption of a managerial compensation plan becomes public knowledge is often very ambiguous, and this blunts the effectiveness of the statistical tests. For example, the appropriate date for the increase in managerial incentives from increased stockholding might be (i) the date the transaction was anticipated, (ii) the date the transaction was made or (iii) the date that the transaction was filed with the regulatory authorities. Similarly, there are associated problems with the announcement

of the adoption of new compensation plans. The appropriate announcement date may occur (i) when the board ratifies the adoption of the new plan, (ii) when the plan is recorded in the annual report and accounts or (iii) the date that the plan is delivered to shareholders and the media (see Murphy 1999). Murphy (1999) highlights two event studies that have documented links between returns and the adoption of compensation plans. Brickley et al. (1985) document a 2.4% abnormal return for companies adopting compensation plans. The authors’ tests are based on a number of potential announcement dates. Yermack (1997), too, finds that stock prices increase after non-publicly announced grants of executive stock option packages. This is consistent with a reduction in agency costs.

5.2 Corporate Performance and Managerial Incentives Next we turn to the relationship between measures of corporate performance and managerial incentives. A potential avenue to explore in the light of the implications of the efficient market hypothesis is the relationship between management incentives and other measures of corporate performance. Morck et al. (1988) examine the relationship between Tobin’s Q, defined as the market value of the firm divided by the replacement cost of the firm’s assets, and the concentration of managerial stock-equity holdings. They also consider a return-on-assets measure. They argue that there is potentially a non-linear relationship between firm performance and managerial ownership concentration. At relatively low levels of ownership concentration by management, firm performance increases as agency costs are mitigated. The effects of increased incentives translate into higher effort and performance. Indeed, they find that firm performance (as measured by Tobin’s Q) increases with managerial holdings between 0% and 5% for a sample of 371 firms. However, for values of managerial stock

holdings between 5% and 25%, they find that there is a negative effect on Tobin’s Q, although the relationship is less in magnitude and weaker in significance. They attribute this to an ‘‘entrenchment’’ effect and potential conflicts between different shareholder interests. For board concentration in excess of 25%, they find a positive effect. Their other performance measure is the profit rate. In a sample of 215 companies, they find a positive relationship between the profit rate and board ownership concentration between 0% and 5%. The non-linearities are not significant. Mehran (1995) analyses the relationship between firm performance and management ownership structure in a random sample of 153 manufacturing companies between 1979 and 1980. The performance measures under consideration are Tobin’s Q and return on assets. Mehran (1995) measures the importance of equity-based compensation structure as (i) the percentage of total compensation in grants of new stock options, with options valued via Black–Scholes and (ii) the percentage of total compensation that is equity based. The results indicate a positive and significant relationship between both performance measures and equity-based managerial incentives.

5.3 Corporate Performance and the PPS Following Mehran (1995), Morck et al. (1988) and others, we simply model corporate performance as a function of managerial incentives (and other control variables). The sample is the 100 companies in the fiscal year 1997/8 analysed in Section 4. The descriptive statistics for the company-level data are contained in the lower half of Table 2. The main performance measure that we model is the firm’s return on assets. We would also have liked to model Tobin’s Q. However, a near approximation to average Q available from Datastream is the market to book value. When we used this performance measure, we were unable to identify robust managerial incentive effects. The appropriate calculation of Tobin’s Q using UK data (for example, in

providing an economic estimate of the replacement cost of the firms assets) and its relationship to managerial incentives, therefore, is left for future research. We measure managerial incentives as the total board PPS. Recall that for any given executive director, the PPS tells us how managerial wealth changes for a given change in shareholder wealth. The metric tells us what percentage of any increase in firm wealth the executive effectively receives in terms of his holding of equity, stock options and long-term incentive plans. To examine the measure for the board as a whole, as distinct from an individual executive such as the CEO, we simply sum the PPS across executives for each firm separately. The descriptive statistics in Table 2 indicate that the mean (median) total board PPS is 1.23% (0.133%). The results of the regression analysis are contained in Table 4. The model controls for the effects of firm size (measured as log market value), business risk (measured as the standard deviation of firm returns) and industry effects (measured by a set of industry dummy variables). Following, McConnell and Servaes (1990), Mehran (1995) and Morck et al. (1988), the managerial incentive variable, captured here as total board PPS, is included as a quadratic to identify non-linearities. We report the results from our quantile regression analysis. Focusing on column (2), which is the median regression, we identify a concave relationship between return on assets and managerial incentives. The coefficient on both the linear and quadratic terms are significant, but the quadratic effect is smaller in magnitude than the linear effect. This result is replicated for the 75th percentile regression but not for the 25th percentile. The results indicate that companies in which executive compensation is relatively sensitive to firm performance tend to have higher returns on assets than companies where the relationship is relatively weak at low-to-medium levels of effective ownership rates. Similar effects were found with alternative estimation techniques of the same model in

June 2001


161


Table 4. However, the significance of the results was not as strong as using the quantile regression method. Using OLS, and controlling for arbitrary heteroscedasticity using White (1980) standard errors, the coefficient (standard error) on the total board PPS variable was 14.53 (9.63). The coefficient (standard error) on the square term was ÿ0.548 (0.395). Using Robust Regression techniques the coefficients (standard errors) on the linear and quadratic terms were respectively 3.17 (2.00) and ÿ0.124 (0.083). These results are suggestive of a relationship between firm performance and managerial incentives in UK firms (where incentives are represented by the aggregate PPS for the total board). However, a number of important caveats suggest caution in interpreting these results and a need for further research. First, the results established in this paper report statistical associations but do not control for the endogeneity of managerial incentives and firm performance. As we have seen in Section 2, compensation and indeed compensation structure may well be determined by corporate performance (i.e. companies with higher corporate performance may cause stronger managerial incentives). The fact that managerial incentives and firm performance form part of a simultaneous

system implies that the empirical relationships identified here are not necessarily causal ones. Instead, the results suggest an interesting empirical observation on which to base further research. Secondly, we have only considered one measure of company performance. We experimented with others, including the market-to-book value, the price cost margin (i.e. return on firm sales), and shareholder return. We had varying degrees of success with these measures in identifying a firm performance and managerial incentive relationship. In Table 5, we report (for interest) the relationship between stock returns over the period 1996–1997 and managerial incentives. As before, we can identify a quadratic relationship. However, this result should be treated with some caution given the discussion of efficient markets in Section 5.1. Different methods (i.e. both OLS and Robust Regression procedures) produced similar statistical results to those in Table 5. However, regressions of the return on sales and market to book value on managerial incentives did not conclusively identify any relationship between the performance term and the PPS. Thirdly, the measure of incentives that we have considered here are the aggregate

Table 4. Corporate performance (ROCE) and incentives at 100 large UK companies in 1997 LHS=return on capital employed

Log market value Volatility Total Board PPS Total board PPS squared Observations Industry dummies Pseudo R2


162




2.4544*** (0.9023) 8.7975 (28.5637) 1.3615 (1.2788) ÿ0.0514 (0.0535) 100 Yes 0.1665

2.3960** (1.1747) 46.5129 (36.5700) 11.7661*** (1.2785) ÿ0.4578*** (0.0538) 100 Yes 0.1739

3.2332 (4.8900) 81.5891 (143.2022) 27.8784*** (2.9562) ÿ1.0549*** (0.1288) 100 Yes 0.2422

Notes: 1 See Table 3, notes 1±3. 2 Log market value is measured at July 1998. Volatility is the variability of stock returns over the company's last fiscal year. 3 The total board PPS is calculated from the individual executive PPS. The PPS for each executive is defined as in Table 2.

Table 5. Corporate performance (shareholder return) and incentives at 100 large UK companies in 1997 LHS=shareholder return

Log market value Volatility Total board PPS Total board PPS squared Observations Industry dummies Pseudo R2




0.0910*** (0.0225) ÿ1.3147* (0.6945) 0.0510*** (0.0168) ÿ0.0014* (0.0007) 100 Yes 0.3573

0.1047** (0.0402) ÿ0.3140 (1.1885) 0.0868** (0.03643) ÿ0.0030* (0.0016) 100 Yes 0.2980

0.0821*** (0.0218) 1.1182 (0.6862) 0.0450** (0.0181) ÿ0.0012 (0.0008) 100 Yes 0.3088

June 2001

Notes: 1 See Table 3, notes 1 and 3. 2 The significance is *p < 0:1, **p < 0:05, ***p < 0:01. 3 See Table 4, note 2. 4 The total board PPS is calculated from the individual executive PPS. The PPS for each executive is defined as in Table 2.

financial incentives (i.e. those arising from equity, options and LTIPs) for the whole board. It is perfectly feasible that these effects could be disaggregated. For instance, the effects of option incentives and equity incentives on firm performance could be analysed separately. Also, the effects of CEO incentives, as distinct from the total board or other executive directors, could also be considered separately.

6 Conclusions In this paper, we have explored aspects of the relationship between executive pay and corporate performance. In the first part of the paper, we focused on two broad issues. The first was the pay-for-performance link, and the other was the role of tournament theory in promoting incentives. This formed a review of issues and evidence pertinent to both areas. In the second half of the paper, we provided new UK empirical evidence on the PPS for all executives within firms. We also provided evidence on the relationship between firm performance and managerial incentives. Our analysis of the PPS concluded that, until fairly recently, the econometric evidence suggested that the putative link between management pay and performance was small

(and possibly weak in the UK context) and so, in consequence, the incentives facing senior executives may be blunted. Various authors had thus concluded that the small changes in executive wealth brought about by changes in firm performance were not consistent with agency models. However, we also showed that recent US research had demonstrated that the pay-for-performance sensitivity/statistic had been increasing over time. An important contributing factor to this growth has been the increased importance of stock-based compensation such as share options. In the UK, knowledge of the time-series behaviour of the pay-for-performance term is limited due to financial reporting constraints in place prior to 1995. This effectively made it difficult to isolate the incentives arising from an executive’s portfolio of options. However, since the publication of the Greenbury (1995) report, the necessary information to compute the PPS has been available. An interesting future project, then, would explore how the PPS has changed for UK firms. We also focused on tournament theory, which predicts that executives may exert effort in order to be promoted to a better-paid job position. We illustrated that there is a growing body of evidence to suggest that tournament-like mechanisms may explain the


163



164

pattern of incentives within firms. For example, we showed that there was US, European and UK evidence to suggest that executive pay was increasing in the number of organizational levels and the CEO pay premium was positively related to the number of tournament players. Tournament theory shows that the internal pattern or structure of incentives is important to secure optimal motivation. The second half of the paper then built upon the evidence surrounding the pay-forperformance link and the role of tournament theory. We provided evidence on the PPS within UK firms. Using data on 532 executives at 100 UK stock market companies in the fiscal year 1997/8, we calculated the PPS for each executive separately. Our data show that the PPS is not constant within firms. Neither is the PPS constant across companies. Importantly, we demonstrated that the median PPS increases as executives move up the organizational levels of the firm. One reason for this observation is that at the top of the organization, the incentives provided from career concerns (i.e. the possibility of future promotions) may become blunted. To compensate for the lost opportunity of playing in a promotion tournament, greater financial incentives are required. This is consistent with tournament models. Finally, we considered the relationship between firm performance and incentives. We were interested in how the provision of incentives can affect managerial effort levels and subsequently improve firm performance. Using data on 100 large UK firms, we established a relationship between firm performance, measured as return on company assets, and management financial incentives. However, we are keen to emphasize the exploratory nature of our results and our desire to stimulate further research in this area. The effect of incentives on firm performance is perhaps the acid test of whether the corporate governance system is working well or not.

Notes 1 This paper is based on research for the Corporate Governance and Economic Performance programme at Warwick University and has been financially supported by the Economic and Social Research Council (Award number R000237246) and PricewaterhouseCoopers. We should like to thank Simon Peck, Laura Read, Martin Walker and seminar participants and colleagues at Warwick and Wharton for useful discussions and comments during the preparation of this paper. 2 Corresponding author: e-mail [email protected]; Tel. +44 (0)24 7652 3849; Fax +44 (0)24 7652 3779.

References Aggarwal, R. and Samwick, A. (1999). The other side of the tradeoff: the impact of risk on executive compensation. Journal of Political Economy, 107, 65–105. Baker, G., Jensen, M. and Murphy, K.J. (1988). Compensation and incentives: practice vs. theory. Journal of Finance, 43, 593–616. Becker, B. and Huselid, M. (1992). The incentive effects of tournament compensation systems. Administrative Science Quarterly, 37, 336–350. Benito, A. and Conyon, M.J. (1999). The governance of directors’ pay: evidence from UK companies. Journal of Management and Governance, 3, 117– 136. Black, F. and Scholes, M. (1973). The pricing of options and corporate liabilities. Journal of Political Economy, 81, 637–659. Brickley, J., Bhagat, S. and Lease, R. (1985). The impact of long-range managerial compensation plans on shareholder wealth. Journal of Accounting and Economics, 7(13), 115–129. Bruce, A. and Buck, T. (1997). Executive reward and corporate governance. In Keasey, K., Thompson, S. and Wright, M. (eds), Corporate Governance: Economic, Management and Financial Issues. Oxford: Oxford University Press. Bull, C., Schotter, A. and Weigelt, K. (1987). Tournaments and piece rates: an experimental study. Journal of Political Economy, 95, 1–33. Conyon, M.J. (1995). Directors’ pay in the privatized utilities. British Journal of Industrial Relations, 33, 159–171. Conyon, M.J. (1997). Corporate governance and executive compensation. International Journal of

Industrial Organization, 15, 493–509. Conyon, M.J. (1998). Directors’ pay and turnover: an application to a sample of large UK firms. Oxford Bulletin of Economics and Statistics, 60, 485–507. Conyon, M.J. and Gregg, P. (1994). Pay at the top: a study of the sensitivity of top director remuneration to company specific shocks. National Institute Economic Review, 3, 83–92. Conyon, M.J. and Leech, D. (1994). Top pay, company performance and corporate governance. Oxford Bulletin of Economics and Statistics, 56, 229–247. Conyon, M.J. and Murphy, K.J. (2000). The prince and the pauper? CEO pay in the US and UK. Economic Journal, 110, 640–671. Conyon, M.J. and Nicolitsas, D. (1998). Does the market for top executives work? CEO pay and turnover in small UK companies Small Business Economics, 11, 145–154. Conyon, M.J. and Peck, S.I. (1998a). Recent developments in UK corporate governance. In Buxton, T., Chapman, P. and Temple, P. (eds), Britain’s Economic Performance. London: Routledge. Conyon, M.J. and Peck, S.I. (1998b). Board control, remuneration committees and top management compensation. Academy Management Journal, 41, 146–157. Conyon, M.J., Gregg, P. and Machin, S. (1995). Taking care of business: executive compensation in the UK. Economic Journal, 105, 704–715. Conyon, M., Peck, S. and Sadler, G. (2001). Corporate tournaments and executive compensation. Strategic Management Journal, forthcoming. Cosh, A. and Hughes, A. (1997). Executive r e m u n e r a t i o n, e x e c u t i ve d i s m i s s a l s a n d institutional shareholdings. International Journal of Industrial Organization, 15, 469–492. Cox, J.C. and Rubenstein, M. (1985). Option Markets. Englewood Cliffs, NJ: Prentice Hall. Dye, R.A. (1984). The trouble with tournaments. Economic Inquiry, 22, 147–149. Ehrenberg, R.G. and Bognanno, M.L. (1990). The incentive effects of tournaments revisited: evidence from the European PGA Tour. Industrial and Labor Relations Review, 43, 74–88. Eriksson, T. (1999). Executive compensation and tournament theory: empirical tests on Danish data. Journal of Labor Economics, 17, 224–242. Fernie, S. and Metcalf, D. (1996). It’s not what you pay it’s the way that you pay it and that’s what gets

results. London School of Economics Discussion Paper. Garen, J. (1994). Executive compensation and principal–agent theory. Journal of Political Economy, 102, 1175–1199. Gibbons, R. and Murphy, K.J. (1992). Optimal incentive contracting in the presence of career concerns: theory and evidence. Journal of Political Economy, 100, 468–505. Greenbury, R. (1995). Directors’ Remuneration: Report of a Study Group Chaired by Sir Richard Greenbury. London: Gee. Gregg, P., Machin, S. and Szymanski, S. (1993). The disappearing relationship between directors’ pay and corporate performance. British Journal of Industrial Relations, 31, 1–10. Hall, B.J. and Leibman, J.B. (1998). Are CEOs really paid like bureaucrats? Quarterly Journal of Economics, 113(3), 653–691. Hallock, K. and Murphy, K. (1999). The Economics of Executive Compensation. Northampton, MA: Edward Elgar. Hart, O. (1995). Corporate governance some theory and implications. Economic Journal, 105, 678– 689. Haubrich, J. (1994). Risk aversion, performance pay and the principal agent problem. Journal of Political Economy, 102, 258–276. Holmstrom, B. (1979). Moral hazard and observability. Bell Journal of Economics, 10, 74– 91. Holmstrom, B. (1982). Moral hazard in teams. Bell Journal of Economics, 13, 324–340. Holmstrom, B. and Milgrom, P. (1987). Aggregation and linearity in the provision of intertemporal incentives. Econometrica, 55, 303–328. Jensen, M. and Murphy, K.J. (1990a). Performance pay and top-management incentives. Journal of Political Economy, 98(2), 225–264. Jensen, M. and Murphy, K.J. (1990b). CEO incentives: it’s not how much you pay, but how. Harvard Business Review, 68(May/June), 138–149. Johnston, J and DiNardo, J. (1997). Econometric Methods, 4th edition. New York: McGraw-Hill. Knoeber, C. and Thurman, W. (1994). Testing the theory of tournaments: an empirical analysis of broiler production. Journal of Labor Economics, 12, 1551–79. Lambert, R., Larker, D. and Weigelt, K. (1993). The structure of organizational incentives. Administrative Science Quarterly, 38, 438–461. Lazear, E.P. (1995). Personnel Economics.

June 2001


165



166

Cambridge, MA: MIT Press. Lazear, E. and Rosen, S. (1981). Rank order tournaments as optimum labour contracts. Journal of Political Economy, 89, 841–864. Lewellen, W. and Huntsman, B. (1970). Managerial pay and corporate performance. American Economic Review, 60, 710–720. Main, B. (1992). Top executive pay and company performance. University of Edinburgh Mimeograph. Main, B.G.M. and Johnston, J. (1993). Remuneration committees and corporate governance. Accounting and Business Research, 23, 351–362. Main, B.G.M., O’Reilly, C.A. and Wade, J. (1993). Top executive pay: tournament or teamwork? Journal of Labor Economics, 11, 606–628. Main, B.G., Bruce, A. and Buck, T. (1996). Total board remuneration and company performance. Economic Journal, 106, 1627–1644. McConnell, J. and Servaes, H. (1990). Additional evidence on equity ownership and corporate value. Journal of Financial Economics, 27, 595–612. Mehran, H (1995). Executive compensation structure, ownership and firm performance. Journal of Financial Economics, 8, 163–184. Morck, R., Shleifer, A. and Vishny, R. (1988). Management ownership and market valuation: an empirical analysis. Journal of Financial Economics, 20, 293–315. Murphy, K.J. (1985). Corporate performance and managerial remuneration: an empirical analysis. Journal of Accounting and Economics, 7, 11–42. Murphy, K.J. (1999). Executive compensation. In Ashenfelter, O. and Card, D. (eds), Handbook of Labor Economics, Vol. 3. Amsterdam: North Holland. Nickell, S. (1995). The Performance of Companies. Oxford: Basil Blackwell. O’Reilly, C.A., Main, B.G.M. and Crystal, G.S. (1988). CEO compensation as tournament and social comparison : a tale of two theories. Administrative Science Quarterly, 33, 257–274. Prendergast, C. (1999). The provision of incentives in firms. Journal of Economic Literature, 37, 7–63. Rosen, S. (1986). Prizes and incentives in elimination tournaments. American Economic Review, 76, 701–715. Smith, R. and Szymanski, S. (1995). Executive pay and performance: the empirical importance of the participation constraint. International Journal of the Economics of Business, 2, 485–495. Tirole, J. (1988). The Theory of Industrial

Organisation. Cambridge, MA: MIT Press. White, H. (1980). A heteroskedasticity-consistent covariance matrix estimator and a direct test for heteroskedasticity. Econometrica, 50, 1–16. Yermack, D. (1995). Do corporations award CEO stock options effectively? Journal of Financial Economics, 39, 237–269. Yermack, D. (1997). Good timing: CEO stock option awards and company news announcements. Journal of Financial Economics, 52, 449–476.

Appendix A: Agency Model This appendix outlines a simple example of an incentive scheme within a principal–agent framework. We derive the optimal pay-forperformance term specified in Section 2.1 of the text. The following basic model is based on Holmstrom and Milgrom (1987). Let a be the effort of the agent and let x a be the output observed by the principal, where the random variable is normally distributed: N 0; 2 . Let the principal’s choice of incentive scheme be linear so that the pay-off to the agent is given by sx x a ; where and are to be determined by the principal. The term x may be thought of as shareholder wealth or profit, while can be interpreted as the fixed salary component of pay. Finally, represents the PPS, and it is this term in particular that researchers attempt to estimate. The principal is assumed to be well diversified and hence risk neutral. Accordingly, their utility can be represented by the pay-off, which is equivalent to output net of incentive pay. The principal’s expected wealth is thus given by

Ex ÿ sx Ea ÿ ÿ a ÿ 1 ÿ a ÿ

A1

Suppose now that the agent, by contrast, is risk averse with a constant absolute risk averse (CARA) utility function given by

Uw ÿeÿrw , where w is wealth and r is the absolute risk aversion level of the agent. By d e f i n i t i oRn , i t i s t h e c a s e t h a t EUw Uwf wdw, where f(w) is the probability density R function of w. Thus, in this case EUw ÿeÿrw f wdw. If wealth is normally distributed, w N w; 2w , then using the properties of the normal function, this simplifies to give

EUw ÿe

r2w 2

ÿ ÿr w

The problem for the agent is determine the level of effort (a) in order to maximize this utility minus the cost of that effort c(a):

2r 2 ÿ ca 2

(a)

(b) (c) (d)

r 2 2 : a ÿ 2

(e)

A3

The first-order condition is simply c0 (a) . The principal’s problem is to maximize his/ her own utility by determining and subject to the above first-order condition and a second constraint that the agent receives a reservation utility u*. Thus the maximization problem becomes:

max ; ; a : 1 ÿ a ÿ subject to

a ÿ

2r 2 ÿ ca u 2

A4

and

c0 a The first-order condition is simply: 1 ÿ rc0 ac00 a2 ÿ c0 a 0. Solving for c(a)0 yields

1 1 rc00 a2

A5

Equation (A5) is given in Section 2.1 of the text. The following are noteworthy.

(A2)

Given the properties of the exponential function, the same ordering will be preserved ÿ r2w =2 as an equivalent utility by using w measure. Furthermore, since xN(a,2) and sx x, then s(x)~N(+ a, 22). The agent’s utility of wealth is thus given by

max a : a ÿ

(f)

The optimal pay-for-performance term, in this particular model, is related to three factors (risk aversion, cost of effort, and variability of firm wealth). Other models can be developed which can include the productivity of the agent, relative performance evaluation and distortion in the performance measures used by firms. The optimal PPS will be equal to one when output is certain (2 0). The optimal PPS will be equal to one when the agent is risk neutral (r 0). As the uncertainty in the firm value increases, and/or the risk aversion of the agent increases, the optimal PPS declines (d /d < 0, d /dr < 0). The optimal PPS will be equal to one when the increase in the marginal costof-effort is zero [c00 (a) 0]. This may be less intuitive than previous results. It implies that, if the marginal cost of effort does not change as the agent undertakes more effort, then the optimal sharing rate is set equal to one. This will occur with a linear cost-of-effort function. The intuition here is how much additional effort the employee exerts as incentives increase. If the CEO is unresponsive to increased incentives, high incentive compensation imposes extra risk on the executive but induces little extra effort, so there is less reason to provide more incentive compensation. As the change in the marginal cost-ofeffort increases, the pay-for-performance term decreases, d /d[c00 (a)] < 0.

June 2001

Appendix B: Black--Scholes Executive stock options may be valued according to the Black–Scholes (1973) pricing


167


formula adjusted for continuously paid dividends. The standard Black–Scholes value (c) of a single European call option is

c SeÿqT N d1 ÿ X eÿrT N d2 where d1 flnS=X r ÿ q 2 =2Tg=fT 1=2 g d2 flnS=X r ÿ q ÿ 2 =2Tg=fT 1=2 g The six inputs to the function are: S, the share price; X, the exercise or strike price; T, the time to maturity; q, the dividend yield; r, the


168

risk free rate of interest; , the standard deviation of returns on the share; N(.) is the cumulative probability distribution function for a standardized normal variable. Note that the terms S, q, r and are all firmspecific variables, but the inputs X and T are option-tranche specific (i.e. vary within the company). The change in the call value of the option with respect to the underlying price of the asset is given as dc/dS eÿqTN(d1) and 0 1 (see Cox and Rubenstein 1985). This is the option delta, .