Pay at the top: a study of the sensitivity of top director remuneration to company specific shocks.

Conyon, Martin J. ; Gregg, Paul

1. Introduction

Public attention has recently focused on pay awards of senior executives in UK companies. This focus involves both the absolute sums involved and an implicit question mark over whether such executive pay increases are justified by the performance of their companies. In spite of media and public interest it is als the case that determination of executive compensation has, unlike in the United States, attracted relatively little attention from academic economists in the UK.(1) This article aims to partially rectify this situation by examining the determination of top directors' pay in a sample of UK quoted companies between 1985 and 1990.

The empirical modelling of top directors' pay is usually couched in terms of th principal agent paradigm.(2) The shareholders, who are collectively the principals, attempt to control the chief executives, who are their agents, by the use of explicit incentives, including bonuses or share options, which relat their remuneration to the performance of the firm. In this way they overcome th problem which can arise if the interests of the executive, who actually runs th firm, are not the same as the shareholders. However, there are other considerations which may also be important in shaping the pay of senior executives (see Tirole, 1988). The behaviour of chief executives may be constrained to some degree by competition in the product market, and the risk o take-over or bankruptcy. The performance of the chief executive may be judged i terms of yardstick competition, comparing the growth of the firm with others in the same industry for example, and this will provide information to the shareholders of the firm concerned and also other potential employers (see for example Gibbons and Murphy, 1992).

This article assesses the importance of changes in the company specific operating environment, which we call company shocks, in shaping the compensatio packages of chief executives. Using a unique panel of industrial companies(3) w address two related issues:

(1) To what extent, if any, are the compensation packages of chief executives justified by the performance of their companies?

(2) What is the importance of company specific information available to shareholders (company specific shocks) in conditioning executive pay?

The article is organised as follows. In section 2 we briefly discuss the role o principal agent theories and the limits to managerial discretion. The discussio motivates our subsequent empirical analysis. In section 3 we describe the data together with our modelling strategy. In section 4 we consider the results and, finally, in section 5 we offer some concluding remarks.

Our results indicate that growth of directors' pay has averaged just under 10 per cent per annum in real terms between 1985 and 1990 in our sample of around 170 firms. Over the period as a whole, therefore, pay of top executives has risen by 77 per cent in real terms. This contrasts with real average earnings growth of 2.6 per cent per annum over the same period for the same firms or 17 per cent over the period as a whole. In our econometric results it turns out that whilst top director pay growth is positively related to sales growth it is only weakly related to a long-term performance measure (total shareholder returns) and not at all to current accounting profits. There is evidence to suggest that implicit constraints on managerial discretion are important. For example, relative measures of performance, yardstick competition, appear to be applied with sales growth but not shareholder returns. The ownership status of the firm run by the top manager (whether it is a parent or subsidiary), the extent to which the firm is cash constrained and whether the firm derecognised trade union are all important factors in shaping top pay. More surprisingly, th growth in executive pay is systematically higher amongst enterprises that expanded by take-over activity rather than through internal growth and the returns to sales growth by this method are equally or more highly rewarded than internal growth, whilst performance as measured by shareholder returns appears to be unrelated to pay in these circumstances. This is important because sales growth is substantially higher after a take-over but performance is often below normal. This appears to be at odds with the principal agent framework and so raises questions as to whether effective control of managerial pay setting by shareholders is in place.

2. Compensation, company performance and operating environment

The principal agent framework has become a widely used theoretical model to explain the remuneration of high level management and chief executives (see Main, 1992). Such models typically predict that a positive relationship between compensation and company performance can emerge. In the typical agency framewor shareholders (who collectively act as a risk neutral principal) delegate decision making authority to managers (the risk averse agent) whose interests potentially diverge from those of the shareholders. This hierarchical structure has an important source of market failure namely that the effort levels of the manager are not directly observable by the principal and so cannot be fully contracted upon. Moreover, the shareholders and the managers' interests potentially diverge since managerial effort positively affects the output variable which the shareholder is interested in and hence adds to the shareholders' payoff, but is costly to the agent and so detracts from the managers interests.

The problem for the principal is to design a contract such that the expected monitoring costs for the shareholders are minimised bur still induce the executive manager to act in the best interests of the shareholders although now at the executive's own volition (see Tirole, 1988). The contract offered will depend on the relative risk attitudes of the parties involved and will also be subject to a participation and incentive constraint. The participation constraint requires that the manager receive at least his fall back outside option. The incentive constraint requires that it is in the agent's interest to undertake the costly action. The contract offered typically ties the reward received by the manager to a variable that the principal is interested in such as company performance or shareholder returns (see Gibbons and Murphy, 1990).

The recent empirical literature has paid much attention to the notion of relative performance evaluation (yardstick competition). Here the compensation of the executive, [y.sub.i], in firm i is a function of own firm performance (o profits), [z.sub.i], but also of other firms' output [z.sub.i] in the product market. That is the pay scheme is designed such that [y.sub.i]=[y.sub.i] ([z.sub.i], [z.sub.j]). See Lazear and Rosen (1981) and Gibbons and Murphy (1990) for an empirical application of relative performance evaluation. In essence the output of other firms provides shareholders with important information about the effort levels of own firm managers. If other firms in similar circumstances out-perform the original firm, the manager cannot attribute own poor performance to adverse shocks not under his control.

Strong and Waterson (1987) further entertain the idea that there is a broader range of company specific signals which also reveal information about manageria effort levels. In the case of yardstick competition this was achieved by observing the outturn of other firms. All information available to shareholders which describes firm performance potentially reveals underlying effort level of managers. Hence, the shareholder might offer an incentive payment scheme which can be characterised as y(z, s) where s is an alternative signal indicating other observable information available to shareholders upon which the contract may be conditioned unless of course z itself is a sufficient statistic. This ca include indicators of yardstick competition, but also firm specific characteristics. Indeed this provides an important route by which product marke structures, the risk of bankruptcy and so on can potentially influence executiv effort, and hence performance, by increasing the information base for incentive contracts.

As well as providing information to shareholders signals may also influence managers outside employment opportunities. Fama (1980) argues that explicit incentives contracts may be redundant since managers are disciplined through th managerial labour market. That is, superior performers are suitably rewarded with high wage offers whereas inferior performers receive low offers. Holmstrom (1982) augmented this theoretical notion arguing that whilst the disciplining effect of the managerial labour market is not insubstantial it cannot be regarded as a pure substitute for efficient contracts. In the absence of contracts he shows that executive effort falls as the retirement period approaches. The discipline in the managerial labour market assumes that manager improve their outside options by effort, congruent with maximising shareholder wealth (Gibbons and Murphy, 1992, test this empirically). However, it might be possible for outside options to be related to other factors not necessarily congruent with the interests of shareholders.(4)

3. Econometric modelling strategy and data description

Modelling strategy

To explore the factors which influence top pay we characterise the following very general chief executive compensation equation:

COMPENSATIO[N.sub.it] = g (CORPORATE PERFORMANC[E.sub.it], SIGNAL[S.sub.it]) + [e.sub.it]

where the left hand side variable is defined as the natural logarithm of total salary plus bonus for the highest paid director in company i at time t.(5) We use two direct measures of corporate performance, namely shareholder returns an company operating profits.(6) Shareholder return is defined as the ln[([P.sub.t + [d.sub.t])/[P.sub.t-1]] where P is share price and d is dividend per share. Operating profits are profits before tax plus total interest payments and depreciation. In our estimating model the performance variables pre-date the compensation variable for two reasons. First, since we suspect that current performance is not observable at the time of actual pay determination so that shareholders are likely to use past performance measures as the relevant information variable. Second, using the lagged performance variable can avoid any bias induced by joint determination of pay and performance. In addition, we include firm size as measured by sales. This, of course, reflects past performance but also measures the importance of firm size in managerial organisation and structure. Firm size has been found to be a key determinant of top pay in previous studies (see Main, 1990, Gregg et al, 1993 for the UK and the survey by Rosen, 1990).

Section 2 highlighted firm characteristics as important sources of information to shareholders about the origins of performance. Many such variables are not directly available in our data set. To control for unobserved latent firm characteristics we first difference our data matrix. This procedure implicitly nets out time invariant firm fixed effects. This methodology is proposed by Murphy (1985), Main (1992) and Gregg et al (1993) in relation to the determination of top directors' pay. Expressing the model in first differences provides a very strong test of which factors are important in shaping top directors' pay.(7) Also we include a vector of up to T (where T = time) aggregate time specific variables to filter out aggregate macro economic shocks Thus, we estimate the following first difference equation:

[Delta] COMPENSATIO[N.sub.it] = g([Delta]CORPORATE PERFORMANC[E.sub.it-1], [Delta]SIGNAL[S.sub.it-1], [Delta]Time) + [u.sub.it]

where the operator [Delta] on any variable X is simply current value X minus last period value (that is, [Delta]X = [X.sub.t] - [X.sub.t-1]) and u is an error term(8).

As well as netting out firms specific characteristics we consider five separate groups of variables that may act as signals. These are product market, debt holding, acquisitions, company ownership status, and union presence.

Within the product market group we consider industry sales growth or industry average shareholder returns within the 2 digit SIC industry as a measure of relative performance evaluation. These are derived from other companies in our sample and those available from the original 558 firms but excluded for reasons of other missing data; a total of 260 firms were used. We also use a variable capturing managers' assessment of changing product market competition. Increase competition is defined by a dummy variable which takes a value one (zero otherwise) if, in the view of management, the firm faced increasing product market competition between 1985 and 1990.

A debt measure is included to isolate the potential disciplining effects that debt holding may have on managerial discretion (see Tirole, 1988, or Nickell, Wadhwani and Wall, 1992). It has been suggested (see Diamond, 1984) that debt holders, and banks in particular, may perform a useful policing role for shareholders. Thus, the pay for performance link need not be as important where debt discipline applies. The debt variable used is cash to total current liabilities ratio (CLR). The level of CLR is included lagged twice to remove an potential biases from contemporaneous observation. This is, in effect, an inverse measure of short-term cash constraints.

Acquisitions by firms represent a radically different method of firm growth tha internal or organic expansion. For example, Cowling et al (1980) conclude that the effects of merger on subsequent performance are generally adverse?(9) The fact that delegated authority results in aspects of acquisitive behaviour is clearly of interest to shareholders in assessing firm performance. In our analysis we include two specifications of acquisitions. The first is a dummy variable equal to one (zero otherwise) if the firm engaged in one take-over in the previous three time periods. The second dummy variable indicates whether th firm engaged in two or more take-overs in the last three years. The time lag is to allow for the net impact on pay to become fully apparent and is a statistically accepted aggregation of the impact on pay over the three separate years.

If the firm observed is not a parent but a subsidiary of another (often foreign firm) then the control relationship is somewhat different as another layer of management is in place to monitor effort and authority to enact changes in management personnel or structure. Hence, the pay-performance relationship is unlikely to be as great in these firms. This variable is a dummy equal to one (zero otherwise) if the company is a subsidiary. Whether the company divested itself of part of its assets is also included as a control variable. This is defined as for acquisitions above, except it applies to divestment.

The 1980s were a period of rapid change in industrial relations. The reorganisation of the corporate industrial relations machinery may act as a signal of increased managerial prerogative. The nature of the industrial relations system is a major influence on corporate performance (for example, se Machin and Stewart, 1990, Gregg and Machin, 1991, and Nickell, Wadhwani, and Wall, 1992). Specifically, union recognition acts to raise wages and adversely affect profits. Our empirical model, therefore, includes a variable reflecting firms where union presence was diminished in this period. If management reporte that at least one union was derecognised in any plant within the firm between 1985 and 1990, a derecognition dummy is set to unity. Since the derecognition observed is normally partial (only 5 out of 30 firms had total derecognition) then this dummy variable is likely to pick-up firms where management are re-asserting their prerogative to manage rather than the effect of union exclusion. A levels union control for all firms is included to provide a test o the appropriate benchmark against which to view derecognition.

Data sources

The basic data set that we use is drawn from a survey of companies carried out in 1990 at NIESR. This produced 558 participating companies and gives data on the firms' industrial relations conditions and product market pressures. Datastream International and EXSTAT provide the firm-level economic variables o interest (for example sales, profits and so on). EXSTAT was the original sampling frame for the NIESR survey and covers all industries except banking an insurance but consists of mainly larger publicly quoted companies. Our top director pay variable is drawn from Datastream International and is simply the total annual remuneration package of the highest paid director of the company. This will include salary and bonuses but not share options. Measures of company performance including shareholder returns (that is, we need share price data) i also only available from Datastream. Thus, the need to use both of these compan based datasets, together with the requirement that there are at least four continuous records in the period under consideration for each company, reduces the available sample size sharply from 558 to 169. The basic data set is augmented by supplementary information on take-over activity. This was derived from EXSTAT data tapes directly and from the monthly records of mergers and acquisitions provided by EXTEL Financial for quoted companies.

The survey into firms' trade union and product market conditions asked for details of changes that occurred in two sub-periods through the 1980s, 1980-4 and 1985-90. Furthermore, as sample attrition for details of directors' remuneration is high prior to 1984 this makes 1985 a sensible starting point fo our empirical analysis.

4. Results

Table 1 provides summary information on our key variables. It details average total remuneration (salary plus bonus in thousand pounds) for top directors for the period 1985 to 1990 and average real growth for the same period. Average real wage growth (wage bill per employee) for employees in the same companies i also shown by way of comparison. It is revealed that real wage growth for top directors outstripped the average pay of all employees by a factor of approximately 3.7. Not unsurprisingly, however, the standard deviation of directors' pay was also substantially higher. In the lower half of the table some measures of current performance are given. Profits' growth over this perio was extremely rapid and it broadly corresponds to the considerable cyclical upswing of the mid-1980s. Sales' growth is much more modest but the similarity of the growth rates of sales and directors' pay and their standard deviations i quite striking. Particularly when deviations across the two sub-periods in the sample are considered, 1985 to 1987 and 1988 to 1990.

Table 2 presents estimates of the effect of the lagged firm performance measure on top director pay. Column 1 includes the two direct performance measures (proportional) growth in operating profits and shareholder returns. All estimated models report Huber (1967) standard errors and are robust to arbitrar heteroscedasticity. The result in column 1 illustrates that the profits' term i not significant but shareholder returns has a small but significant impact. The importance of such long-term measures of performance is now well documented (se Gibbons and Murphy, 1990, Gregg et al, 1993 and Conyon and Leech, 1994). The co-efficient estimate reported here is broadly in line with available UK evidence. It implies that a doubling of shareholder returns raises directors' pay by just 6 per cent. Indeed, when evaluated at median executive earnings it implies that a TABULAR DATA OMITTED 10 per cent increase in shareholder returns raises top pay by only 491 pounds. The inclusion of a change in log sales' term (column 2) does not change this pattern much but note that the sales' growth term is much larger than that for shareholder returns.(10) The result here is i line with other recent UK work (Main, 1992, Gregg, Machin and Szymanski, 1993 and Conyon and Leech, 1994 and Conyon, 1994). These authors report top pay equations in which the pay-performance link is weak and the relatively stronger variable is sales' growth. Column 3 splits the shareholder return result betwee 1985 and 1987 and 1988 to 1990. This is purely to replicate Gregg et al, 1993, who demonstrated that for their sample of the FT top 500 firms the pay-performance link diminished between the early and late-1980s. Our sample would include some of these large firms but will also include a TABULAR DATA OMITTED large number of smaller quoted companies (broadly drawn from the top 2000 companies rather than the top 500). The evidence in column 3 is consistent with the notion that there is heterogeneity in the pay for performance link since there is no evidence of a pay and performance link in the post-1987 period. This heterogeneity result is also observed in Conyon, 1994, for a sampl of 214 firms from 1983 to 1987 and 1988 to 1993.

Column 4 is more central to our analysis. It introduces industry benchmarks for the two key measures of company performance identified. As would be expected from benchmark comparisons the industry sales growth term is negative and insignificantly different from the (negative) companies own sales growth. In other words sales growth relative to competitors appears the correct specification. However, industry shareholder returns is positively signed and insignificant. Hence benchmark comparisons of share price and dividend performance of other companies does not appear to be used in assessing managerial pay.

In column 1 of Table 3 we show the direct correlation between the available corporate shock information and directors' pay without including the performanc measures. The product market, debt, ownership, union and take-over controls are now introduced. Only the debt measure, the take-over dummies and whether the firms are subsidiaries have direct and significant effects. All of these information sources are really concerned with interpreting performance indicators in addition to any direct effect on pay of top managers. Hence, column 2 introduces the performance measures (relative sales growth and shareholder returns as preferred from column 4 of Table 2. The derecognition term is now weakly significant (at the 10 per cent significance level). The results are somewhat curious as firms engaging in take-over activity, pay a substantial premium to their top directors. A single take-over results in highe pay growth of 2.7 per cent per annum for three years or just under 8 per cent. Two or more take-overs attracts a coefficient of just over twice the size at 6. per cent per annum. Firms which were became more cash constrained raised pay levels. This could reflect the increase of risk attached to continuing to work for liquidity constrained companies. It implies that managers, engaging in cash financed take-over activity, receive a considerable pay premium. Firms engaging in such a strategy have been identified as highly vulnerable to the subsequent 1990-3 recession (see Geroski and Gregg, 1994).

Table 3. The responsiveness of top directors' pay to product market, debt,
merger and industrial relations shocks: 1985-90

Constant .1375 .1195 .1123
 (.0169) (.0177) (.0135)
[Delta]Competition -.0107 -.0010
 (.0169) (.0168)
CLR(t-2) -.0493 -.0467 -.0467
 (.0191) (.0190) (.0189)
One acquisition in previous three years .0308 .0266 .0263
 (.0188) (.0189) (.0186)
Two acquisitions in previous three years .0689 .0641 .0648
 (.0281) (.0272) (.0272)
Assets sold -.0089 -.0062
 (.0268) (.0272)
Subsidiary in 1990 -.0381 -.0354 -.0370
 (.0203) (.0204) (.0203)
Trade union derecognition .0266 .0322 .0305
 (.0192) (.0192) (.0176)
Recognised trade union -.0062 -.0011
 (.0182) (.0179)
[Delta]log relative sales (t-1) .0851 .0873
 (.0293) (.0286)
Shareholder returns (t-l) .0468 .0458
 (.0241) (.0241)
Time dummies Yes Yes Yes
 F=1.25 F=0.80 F=0.80
 (5,677) (5,675) (5,678)
[R.sup.2] 0.037 0.061 0.061
Observations 691 691 691
Companies 169 169 169

1. Variable definitions are fully described in the data appendix. Note that CLR
is the cash liability ratio. An F test of joint significance of the time dummie
is presented.

However, these results do not allow the pay for performance link to vary according to whether performance is attributable to the shocks described above. For instance an increase in firm size through acquisitive behaviour is effectively a size shift rather than an indicator of performance. We might expect the alternative methods of growth to have differing relations with pay increases. We therefore interact the two performance measures (relative sales' growth and shareholder returns) with the shocks described above. However, to keep the specification as parsimonious as possible given the small sample sizes we drop the insignificant terms in column 2.

Table 4 reports these interactions separately for each type of shock or signal. Column 1 is for take-over TABULAR DATA OMITTED activity. Both interaction terms were insignificantly (and positive for sales but negative for shareholder returns) different from the base group (of organic growth). This implies that sales increases generated at times of take-over activity are not less related t directors' pay growth than at other times nor are the returns to shareholders despite the direct involvement of chief executives in these decisions.

Column 2 of Table 4 looks at relative sales' growth and shareholder returns where the lagged cash liabilities ratio is above average for the sample (that is, a high CLR/low CLR dummy set equal to unity). There is again no significant variation in terms of pay-performance links where debt is high. The CLR(-2) measure itself remains strongly significant.

Column 3 splits the performance measures according to whether the firm is a subsidiary and hence where higher echelons of management will be in place to monitor performance. Sales' growth in subsidiaries proves unrelated to pay (wit a negative coefficient) and significantly different from that relation in paren companies. There is no important variation in shareholder returns, however.

Column 4 looks at performance where derecognition has occurred. The derecognition term itself is better determined but the sales' growth and shareholder returns offer no significant variation.

Thus, the only important variation occurs where control has not been fully delegated and here the sales' growth term is unimportant. However, the subsidiaries in our data are predominately not engaging in merger/take-over activity. 17 per cent of subsidiaries engaged in a take-over as opposed to 36 per cent of non-subsidiary firms (t test of unequal means is 3.47). In column 5 we, therefore, re-evaluate the acquisition/performance interaction with subsidiaries excluded. The step returns to a take-over are now much greater at 4.6 per cent for the three years after a take-over and 8.1 per cent if two or more take-overs have occurred. The sales' growth term is nearly double that in the absence of a take-over, although still not well determined. However, the shareholder returns are clearly' unrelated to pay, unlike where take-overs have not occurred. This is important because sales' growth is higher after a take-over and shareholder returns lower (where there has not been a take-over relative sales' growth is 6.5 per cent, and where there has been a take-over it is 10.7 per cent. The t test of the means is 1.93; shareholder returns in the absence of take-overs is 23.9 per cent and where there has been a take-over it is 16.6 per cent; the t test rejecting equivalence is 2.73). It seems perverse that top directors are rewarded for organising take-overs, benefit again from any increase in sales associated with such a take-over but are isolated from an costs in terms of shareholder returns which are thus wholly born by the shareholders.

5. Conclusions

This article has addressed a number of issues concerning the determination of top directors' pay. We have argued that as well as corporate performance acting as a signal for managerial effort there are also other important corporate specific information available to shareholders (signals) which are likely to be taken into consideration in pay setting. Indeed, we argued that signals emanating from product market, debt holding, acquisitive behaviour and union presence all potentially reveal information about the extent to which corporate performance is due to managerial effort. In addition, these company specific signals may reveal information about managers outside career options which will also be reflected in managerial remuneration.

The empirical analysis suggests that all the types of signal considered are important in shaping pay. Relative performance evaluation or the use of the performance of other firms is taken into account in determining pay and the coefficient is consistent with relative sales growth being the appropriate measure of performance used. However, the same result does not appear to apply to shareholder returns. Reduced union presence results in higher pay for top executives but heading a subsidiary lowered pay in this period. Most surprisingly however, was the results of cash holding and acquisitive behaviour by firms. The results indicated that lower cash holdings relative to current liabilities raises pay, as does expansion through take-overs. Firm growth by take-over which results in the firm being cash poor is a strategy for managers which raises pay considerably, despite other evidence which suggests that such behaviour does not enhance firm performance. Indeed such a strategy left firms highly vulnerable to the subsequent recession. This may be explicable if increasing firm size improves managers outside options. Either way this result raises questions about the degree of effective control of top managers pay and over decision making concerning who controls and benefits from take-over decisions.

Overall the results are consistent with other UK and US evidence that corporate performance explains little of the very large pay growth achieved by top executives in the late-1980s. Even after performance effects and those associated with take-over activity and the like are netted out (what can be described as the going rate) annual pay rises of 13 to 15 per cent per year wer achieved between 1985 and 1990. This would imply that either managerial skills were becoming increasingly valued by shareholders or that top executives were increasingly able to influence their own pay levels.

TABULAR DATA OMITTED

NOTES

(1) The considerable US literature on the determinants of executive compensatio includes Gibbons and Murphy (1992, 1990), Jensen and Murphy (1990), Leonard (1990) and Murphy (1985). Recent UK research includes Cosh (1975), Main (1991, 1992), Gregg et al (1993) and Conyon and Leech (1994).

(2) The principal agent model, and the theoretical design of optimal contracts, is considered formally by Grossman and Hart (1983). Typically, such models examine how incentives schemes are constructed in the presence of moral hazard. Tirole (1988) provides a formal review.

(3) This is augmented from the data set employed by Gregg and Yates (1991).

(4) We have in mind that a managers' outside option is positively related to th size of the firm they control. Hence growth via takeover activity can raise outside options in the managerial labour market, but takeovers are not necessarily consistent with shareholder interests.

(5) Other authors have dealt at length with the appropriate formulation of the dependent variable. For example, should the salary component be separated from the bonus element? How should directors' share holdings in the company be value prior to their realisation? The empirical treatment of these important issues i dealt with by Antle and Smith (1985). A partial justification of the salary plu bonus measure is provided by Lewellen and Huntsman (1970) who demonstrate a strong cross sectional association between a wider more comprehensive measure o executive compensation and the salary plus bonus variable.

(6) Many studies use market valuation type measures of corporate performance rather than operating profits or sales. The results are mixed. Conyon and Leech (1994) use market value measures and find only small performance effects. Decko (1988) finds that accounting based measures are important. Abowd (1990), however, discusses alternative accounting and market based performance measures and concludes in favour of market based measures.

(7) This follows since we are implicitly including individual firm specific dummies in our equation as well as netting out aggregate factors.

(8) In the case of profits it is the proportional growth rate:[Delta]X = ([X.sub.t] - [X.sub.t-1])/[X.sub.t-1] since the profit series contains some negative values.

(9) This result seems quite general. Acquisitions appear to do little for post merger performance. A review is provided in Hay and Morris (1991) which also considers alternative methods of assessing these effects.

(10) We report the [R.sup.2] values for all models but note that by the very nature of difference equations these values usually turn out to be small in magnitude.

REFERENCES

Abowd, JM, (1990), 'Does performance-based managerial compensation affect corporate performance?', Industrial and Labor Relations Review, no 43, pp 52-73

Antle, R and A Smith, (1985), 'Measuring executive compensation: methods and an application', Journal of Accounting Research, no 23, pp 296-325.

Conyon, MJ, (1994), 'Corporate governance and executive compensation', Oxford University mimeograph.

Conyon, MJ and D Leech, (1994), 'Executive compensation, corporate performance and ownership structure', Oxford Bulletin of Economics and Statistics, no 56, p 229-247.

Cosh, A, (1975), 'The remuneration of chief executives in the United Kingdom', Economic Journal, pp 75-94.

Cowling, K .... et al, (1980), Mergers and Economic Performance, Cambridge University Press.

Deckop, JR, (1988), 'Determinants of chief executive officer compensation', Industrial and Labor Relations Review, no 41, pp 215-226.

Diamond, DW, (1984), 'Financial intermediation and delegated monitoring', Revie of Economic Studies, no 51, pp 393-414.

Fama, E, (1980), 'Agency problems and the theory of the firm', Journal of Political Economy, pp 288-307.

Geroski, P and P Gregg, (1993), 'Coping with recession', National Institute Economic Review, no 146, pp 64-75.

Gibbons, R and KJ Murphy, (1990), 'Relative performance evaluation for chief executive officers', Industrial and Labor Relations Review, no 41, pp 30-51.

Gibbons, R and KJ Murphy, (1992), 'Optimal incentive contracts in the presence of career concerns, theory and evidence', Journal of Political Economy, no 100, pp 468-505.

Gregg, P and Yates, A (1991), 'Changes in wage setting arrangements and trade union presence in the 1980s', British Journal of Industrial Relations, no 29, p 361-377.

Gregg, P, Machin, S and Szymanski, S, (1993) 'The disappearing relationship between directors' pay and corporate performance', British Journal of Industria Relations, no 31, pp 1-9.

Grossman, S and O Hart, (1980), 'Take-over bids, the free rider problem and the corporation', Bell Journal of Economics, no 11, pp 42-64.

Grossman, SJ and OD Hart, (1983), 'An analysis of the principal agent problem', Econometrica, no 51, pp 7-45.

Hay D and D Morris, (1991), Industrial Economics and Organisation, Oxford University Press.

Holmstrom, B, (1982), 'Managerial incentive problems: a dynamic perspective', i Essays in Economics and Management in Honour of Lars Wahlbeck.

Huber, PJ, (1967), 'The behavior of maximum likelihood estimates under non-standard conditions', Proceedings of the Fifth Berkely Symposium on Mathematical Statistics and Probability, no 1, pp 221-233.

Jensen MC and KJ Murphy, (1990), 'Performance pay and top management incentives', Journal of Political Economy, no 98, pp 225-264.

Lazear, E and S Rosen, (1981), 'Rank-order tournaments as optimal labour contracts', Journal of Political Economy, no 89, pp 841-864. Leonard, J, (1990) 'Executive pay and corporate performance', Industrial and Labor Relations Review, no 43, pp 13-29.

Lewellen, WG and B Huntsman, (1970), 'Managerial pay and corporate performance' American Economic Review, no 60, pp 710-720.

Machin, S and M Stewart, (1990), 'Unions and the financial performance of British private sector establishments', Journal of Applied Econometrics, no 5, pp 327-350.

Main, B, (1991), 'Top executive pay and performance', Managerial and Decision Economics, no 12, pp 219-229.

Main, B, (1992), 'Top executive pay and company performance', University of Edinburgh, mimeograph.

Murphy, K (1985), 'Corporate performance and managerial incentives', Journal of Accounting and Economics, no 7, pp 11-42.

Nickell, S, S Wadhwani and M Wall, (1992), 'Productivity growth in UK companies 1975-1986', European Economic Review, no36, pp 1055-85.

Rosen, S, (1990), 'Contracts and the market for executives', NBER working paper 3542.

Stewart, MB, (1990), 'Union wage differentials, product market influences and the division of rents', Economic Journal, no 100, pp 1122-1137.

Strong, N and M Waterson, (1987), 'Principals, agents and information', in The Economics of the Firm, edited R Clarke and T McGuinness.

Tirole, J, (1988), The Theory of Industrial Organisation, MIT press.