Relation between CEO compensation, firm size and firm performance

Ann Lau and Ed Vos. New Zealand Journal of Applied Business Research. Vol 3 No 1. October 2004. pp 51-64. Relation between CEO compensation, firm size and firm performance Authors: Ann Lau and Ed Vos* *Correspondence to: Ed Vos Associate Professor of Finance University of Waikato Department of Finance Private Bag 3105 Hamilton, New Zealand Email: evos@waikato.ac.nz Abstract Studies focused on large companies in the United Sates, Japan and Britain have documented that there is a strong positive relation between firm size and CEO compensation. They also find the relation between firm performance and CEO compensation is weak. Executive compensation of 104 New Zealand firms are examined over the period of 1998 2002, and evidence is obtained consistent with previous studies. Result show that for every percentage increases in the firm s total asset, the cash compensation paid to the average CEO increases by 0.39 percent. This result is very close to the finding of many previous studies that identify a firm-size elasticity between 0.2 to 0.35 for larger U.S., U.K. and Japanese firms. Result also shows that the firm size elasticity of CEO compensation is greater in larger firms. Keywords: CEO, compensation, performance, size, dividends

Relation between CEO compensation, firm size and firm performance 1. Introduction In recent years authors of a number of empirical studies have examined the determinants of managerial pay and its relation to firm performance. These studies have focused on large companies in the United States (e.g.kostiuk 1989; Jensen and Murphy 1990a; Garen 1994; Joskow and Rose 1994), Japan (Kato and Rockel 1992; Kaplan 1994), and Britain (Cosh 1975). Moreover, some very recent studies also document evidence for other countries, including Germany, Spain, Italy and Denmark. It became possible to conduct such studies with New Zealand firms after 1998, when all publicly traded companies in the New Zealand Stock Exchange were required to disclose top executives compensation under the this new Regulation. It is interesting to document the New Zealand evidence concerning executive compensation both because it has not been described in the literature and because it yields results that complement those of earlier studies. In most previous studies large U.S. and Japanese firms were examined, but little work has been done on smaller firms. So, examining the New Zealand data provides an opportunity to determine whether or not the widely documented empirical regularities regarding executive pay schemes hold for relatively small firms as well. The main objective of this paper is to determine whether New Zealand s CEOs compensation is dependent on firm size or the firms performance. The results are compared to previous studies to find out whether it provides consistent evidence that there is a strong relation between managerial pay and firm size and a weak relation between CEO compensation and corporate performance. The executive compensation data used in this study are extracted from the annual reports of each of the 104 companies from Datex during the period of 1998 to 2002. Discussion in this paper is focused on the compensation of chief executive officers. In section 5, the relationship between CEO pay and firm size is examined. It is found 2

that CEO earnings from cash compensation increases by 0.39 per cent for every one per cent increase in the firm s total asset. This elasticity is surprisingly close to that identified in previous studies with large U.S., Japanese, and U.K. firms. It has been a puzzling phenomenon that the elasticity changes little across time and countries. Given the substantial difference in size and industry structure between New Zealand sample and those of previous studies, this finding makes the uniformity of the firmsize elasticity of executive pay even more puzzling: it also holds for mush smaller firms. In section 5, the relation between CEO pay and firm performance is discussed. Sales growth, share price changes and changes in dividend per share are used as proxies for firm performance. Evidence is found to support the hypothesis and is consistent with the findings of the exiting literature that the relation between different performance proxies and managerial compensation is weak. The rest of the paper will be set out as follows; section 2 will discuss the existing literature of managerial compensation and its relation with firm size, firm performance and dividend payout. Section 3 will set out the sample and data collected. In section 4, testable hypothesis and the methodology to test the relation between managerial compensation and different explanatory variables is established. Results will be presented in section 5. Lastly, limitation of this paper will be discussed in section 6 before the paper is concluded in section 7. 2. Literature Review CEO pay and firm size Executive compensation has attracted considerable public attention and academic interest because of both the magnitude of pay and its relation to corporate performance. Early studies of managerial pay schemes are focused on the determinants of pay level, particularly, the role of firm size on CEO earnings. According to the allocation theory of control, in a market equilibrium, the most talented executives occupy top positions in the largest firms, where the marginal productivity of their actions is greatly magnified over the many people below them to whom they are linked (Rosen 1992). This provides the theoretical ground for a 3

positive relationship between executive pay and firm size. Evidence has been reported that unanimously supports a strong positive pay-size relation (Roberts 1956; Cosh 1975; Murphy 1985; Kostiuk 1989). Another argument has put forward for the positive relation between firm size and CEO pay as larger firms may employ betterqualified and better-paid managers (Rosen 1982; Kostiuk 1990). CEO pay and firm performance Agency theory predicts that a manager s compensation should be positively correlated with the firm s performance. What agency theory does not do, however, is tell us how strongly correlated. Consequently, a sizeable empirical literature has emerged that seeks to measure this relationship. The most notable example of which is arguably Jensen and Murphy (1990), which found that CEO compensation in many industries seemed fairly insensitive to firm performance (there is statistically significant positive correlation but the coefficient are small). This has also been found in papers that have focussed on regulated firms (Corroll and Cisel, 1982; Joskow 1993). As a consequence of these studies, some have argued that pay for performance is too small to be economically significant. Others (e.g. Haubrich 1994) have argued that Jensen and Murphy s results are actually consistent with plausibly parameterised principalagent model. CEO pay and dividend policy Although none of the dividend theories in the finance literature have explicitly linked managerial compensation and dividends, some studies have found empirical evidence of such a connection. White (1996) found that dividend payouts and dividend yields are both influenced by the presence of stock options in managerial compensation contracts. Fenn and Liang (2001) also found a negative association between stock options and dividends. Bhattacharyya (2000) develops a model for dividend payout that is based in the principal-agent paradigm. In his model, uninformed principals (shareholders) set up a menu of contracts to screen agents according to productivity type (which is known to the agent). Higher quality agents are those who have access to more positive net present value (NPV) projects. These agents are induced to invest the firm s cash rather than pay out dividends. Lower quality agents do not have the same access to 4

positive NPV projects, and the compensation contract they choose induces them to pay out higher dividends. In equilibrium, high quality managers receive higher compensation than do low quality mangers and pay out lower dividens. Empirically, Bhattacharyya s model predicts that dividend payout and managerial compensation are negatively correlated. 3. Data The CEO compensation data are compiled from the firms annual report filed with Datex. Since 1998, all companies publicly traded in New Zealand Stock Exchange have been required to disclose the payments made to their CEO and other directors whose remuneration is greater NZ$100,000. The data covered all companies that submitted this information on their annual reports during 1998 to 2002, with 104 firms reporting executive compensation for three consecutive years. In a broad sense, the components of CEO compensation can be classified into four categories: salary, bonus, long-term incentive rewards, and benefits. Salary is the major component of executive pay in New Zealand firms. Bonus is the annual variable component of remuneration or short-term incentive pay. Long-term incentive rewards include stock options, restricted shares, and long-term incentive plan payouts. Annual bonus and option grants are the two important schemes that are commonly used in large firms. Benefits include all other payments that can not be adequately reported under any of the above pay components, such as payments for life insurance, contributions to a pension plan, imputed interest benefits for debt, tax, subsidy, car and housing allowance, and so on. However, in many of the annual reports, firms only reported the cash component of the CEO pay for 2 reasons. First, stock options as remuneration is not a common way of compensation in New Zealand. And secondly, even a firm uses stock options as remuneration, it is not required by law that they disclose this information to the public. Throughout the paper we refer the sum of all four components as total pay and to total pay excluding stock options as cash compensation. Moreover, our study is based on the cash component of the CEO pay as statistics of the pay variables show 5

that cash component is the most important pay component of CEO pay in New Zealand. In order to look at the relation between firm size and CEO pay, total assets is used in this study. To study the relation between firm performance and CEO pay, three proxies were chosen: sales growth, share price changes, and growth in dividend payout. Among all 202 firms listed in New Zealand Stock Exchange, only the companies with at least 3 years of consecutive data are used, otherwise variables like sales growth cannot be calculated. So altogether 104 firms listed in the New Zealand stock exchanged are used in this study to examine the relation between CEO pay, firm size and firm performance. 4. Methodology As the literature suggests that there is a strong positive relation between firm size and managerial compensation as larger firms could employ better-qualified managers. Our primary hypothesis is to test whether such relation exist in New Zealand firms. The following OLS regression model is estimated: LComp jt = α + β 1 LogAsset jt + ε where LComp jt = Log of the cash component of CEO pay of firm j in period t; and LogAsset jt = Log of total asset of firm j in period t The reason for taking log of both variables is that β 1 is the elasticity of CEO compensation with respect to the firm size, measured by total asset of the firm. That means, β 1 measures the percentage change in CEO compensation for a percentage change in the firm size, where firm size is measured by the firm s total asset. If β 1 from the above regression model is positive and statistically significant, the support could be found for this hypothesis with New Zealand firms, such that there is a strong positive relation between firm size and managerial compensation in New Zealand, Results are discussed in the next section. 6

Existing literature also documented that there is a weak relation between firm performance and CEO pay. In order to see whether such relation exists in New Zealand firms, four different regression models have been estimated to study the relation between CEO pay and different performance variables. Dividend payout is used as the first performance variable. Bhattacharyya (2000) suggested that dividend payout should be negatively related to managerial compensation, as higher quality managers are those who have access to more positive NPV projects and therefore they are less likely to pay out dividends. In other words the higher the dividend payout ratio, the lower the CEO compensation. Therefore, another testable hypothesis for this study is to examine whether such negative relation between dividend payout and CEO pay exists and the following OLS regression model is estimated: LComp jt = α + β 1 LogAsset jt + β 2 DivChange jt + ε where LComp jt = Log of the cash component of CEO pay of firm j in period t; LogAsset jt = Log of total asset of firm j in period t; and DivChange jt = Change in dividend payout between period t and period t-1 The reason of including LogAsset in the regression model above is to control the size effect and β 1 still measures the elasticity of CEO compensation with respect to firm size, with DivChange jt being held constant. The elasticity of CEO compensation with respect to the change in dividend payout is measure by β 2 in the above regression model, i.e. β 2 measures the percentage change in CEO compensation for a percentage change in dividend payout. If β 2 from the above regression model is negative and statistically insignificant, then support could be found for the hypothesis that there is a weak negative relation between dividend payout and managerial compensation in New Zealand. Results are discussed in the next section. Relation between managerial compensation and two other performance variables, sales growth and changes in share price, are also investigated. Existing literature does not have extensive studies on these two variables. Therefore it is worth to take a look at these two performance variables. It is reasonable to consider that there should be a 7

positive relation between these two variables and managerial compensation. Therefore another testable hypothesis is that there should be a positive relation between these performance variables and CEO compensation, however this relation should be weak according to existing literature. The following two regression models are estimated: LComp jt = α + β 1 LogAsset jt + β 2 SalesChange jt + ε where LComp jt = Log of the cash component of CEO pay of firm j in period t; LogAsset jt = Log of total asset of firm j in period t; and SalesChange jt = Change in sales between period t and period t-1 and LComp jt = α + β 1 LogAsset jt + β 2 PriceChange jt + ε where LComp jt = Log of the cash component of CEO pay of firm j in period t; LogAsset jt = Log of total asset of firm j in period t; and PriceChange jt = Change in share price between period t and period t-1 The elasticity of CEO compensation with respect to change in sales is measure by β 2 in the above regression model, i.e. β 2 measures the percentage change in CEO compensation for a percentage change in sales. If the hypothesis is true with New Zealand firms, such that there is a weak positive relation between firm performance and managerial compensation in New Zealand, β 2 from both of the above regression model should be positive but statistically insignificant. Results are discussed in the next section. After examining the firm performance variables separately, it is worth to put all these variables into one regression model. Estimated coefficients might be incorrectly specified if relevant variables are omitted from the regression model and this can lead to very serious consequences. In this case, it is very unlikely that CEO compensation just depends on one the three performance variables, namely sales growth, change in share prices, and change in dividend payout. Due to underfitting the model, i.e. omitting relevant variables, coefficients estimated from the above regression models 8

would be biases and inconsistent. It is believed that all three variables will interactively affect managerial compensation. To capture this effect, a regression model that incorporated all three performance variables is estimated: LComp jt = α + β 1 LogAsset jt + β 2 DivChange jt + β 3 SalesChange jt + β 3 PriceChange jt + ε where LComp jt = Log of the cash component of CEO pay of firm j in period t; LogAsset jt = Log of total asset of firm j in period t; DivChange jt = Change in dividend payout between period t and period t-1; SalesChange jt = Change in sales between period t and period t-1; and PriceChange jt = Change in share price between period t and period t-1 The signs and statistic significance of each of the coefficients from the above regression model should remain the same as their respective regression models, i.e. β 1 should still be positive and statistically significant, β 2 be negative and statistically insignificant, and β 3 and β 4 be negative and statistically insignificant. Although the same result is expected, this model should be a better one as it incorporated more variables and is better to explain the relation between managerial compensation and the performance variables. Given the wide dispersion of firm size distribution and the particular industry structure, the sample provides an opportunity to examine further possible variations of the firm-size elasticity of executive compensation. To capture the difference between firms of different sizes, the total sample is divided into large firms and small firms at the sample median of the firm s total asset, and the same regression models are used to estimate to each of these subsamples. New Zealand has a different industry structure from US firms. In order to compare with existing U.S. studies of managerial compensation, we follow Murphy (1998) and divide the total sample into five industry groups: resources (including firms in agriculture, mining, and oil and gas extraction), manufacturing, utilities, financial services, and other industries (including wholesales, and retail trade, and services). Two things are worth mentioning. First, both recourses and financial services are 9

important industries in New Zealand, constituting about 30 and 22 percent respectively, of the total sample. Second, considerable differences in firm size exist across industries; the average resources firms are roughly one-quarter as large as utilities and financial services companies. Regression models are estimated for each of the five industries to see whether the same relation exist between managerial compensation and different explanatory variables. 5. Results A brief comparison of New Zealand firms with U.S. firms shows a marked difference between the two countries in the level of CEO compensation. New Zealand CEOs earn less pay than their U.S. counterparts. For example, Murphy (1998) reports that the cash component of CEO pay of about US$ 0.83 million to the median CEO of the S&P500 Industrials over the years 1993-1995. For the top 25 per cent New Zealand firms in our sample, the median CEO pay during the 1998 to 2003 is about NZ$0.45 million, or US$0.21 million when changed into U.S. dollars by the annual exchange rates. This pay level is even lower than the median CEO pay of US$0.43 million for the S&P400 Mid-Cap Industrials and lower than $US$0.26 million for the S&P600 Small-Cap Industrials. Table 1 presents the OLS estimates of the firm-size elasticity of CEO pay for the total sample. The estimated coefficient of LOGASSET, 0.39887, is the result for the elasticity of CEO compensation with respect to the firm size. For every percentage increases in the firm s total asset, the cash compensation paid to the average CEO increases by 0.39 percent. This result is very close to the finding of many previous studies that identify a firm-size elasticity between 0.2 to 0.35 for larger U.S., U.K. and Japanese firms. Moreover, the p value given by the regression model is 0, suggesting that the result confirms with our hypothesis (and existing literature) that there is a strong positive relation Table 1 Regression of CEO compensation and firm size (whole sample) NAME COEFFICIENT ERROR 273 DF P-VALUE LOGASSET 0.39887 0.04804 8.303 0.000 CONSTANT 9.1315 0.3942 23.16 0.000 10

After reviewing the evidence documented by previous studies, Rosen concludes that the relative uniformity of the elasticity of executive pay with respect to scale across firms, industries, countries and periods of time is notable and puzzling because the technology that sustains control and scale should vary across these disparate units of comparison (1992, 86). Given that the average firm size of our sample is substantially smaller than that of other studies, the above results make the uniformity of the elasticity even more puzzling: it also holds for much smaller firms. Table 2 presents the OLS estimates of the firm-performance elasticity of CEO pay for the total sample, where dividend payout is used as a proxy for firm performance. The estimated coefficient of DIVCHANGE, 0.04527, is the result for the elasticity of CEO compensation with respect to change in dividend payout. For every percentage increases in the dividend payout, the cash compensation paid to the average CEO increases by 0.05 percentage. This result contradicts with existing literature and our hypothesis that dividend payout should be negatively related to managerial compensation. However, as seen in table 2 that the p value given by our regression model is extremely high, suggesting that this result is not statistically significant. In other words, the positive relation estimated by our regression model is very weak. This confirms with the literature that the relation between firm performance and CEO compensation is weak. One thing to notice that firm size variable is still highly significant. Table 2 Regression of dividend payout and CEO compensation (whole sample) NAME COEFFICIENT ERROR 272 DF P-VALUE DIVCHANGE 0.045274 0.08702 0.5203 0.603 LOGASSET 0.39852 0.04811 8.284 0.000 CONSTANT 9.1292 0.3948 23.13 0.000 Table 3 and Table 4 present the OLS estimates of the firm-performance elasticity of CEO pay for the total sample. Table 3 uses sales growth as a measure of firm performance and table 4 uses change in share prices as a proxy. The estimated 11

coefficient of SALESCHA from table 3, -0.0019557, is the result for the elasticity of CEO compensation with respect to sales growth. For every percentage increases in the sales, the cash compensation paid to the average CEO decreases by 0.002 percentage. This result is different from our prior expectation that sales growth is positively related to CEO compensation. However, the p value given by this regression model is large, suggesting that this relation is statistically insignificant and therefore weak. The estimated coefficient of PRICECHA from table 4, 0.0197, is the result for the elasticity of CEO compensation with respect to changes in share price. For every percentage increases in the share price, the cash compensation paid to the average CEO increases by 0.02 percentage. Moreover, the p value given by this model is again large, meaning it is statistically insignificant. This result confirms with our prior expectation that changes in share price is positively related to CEO compensation, but the relation is weak. One thing to notice that firm size variable is still highly significant. Table 3 Regression of sales growth and CEO compensation (whole sample) NAME COEFFICIENT ERROR 272 DF P-VALUE SALESCHANGE -0.0019557 0.001348-1.450 0.148 LOGASSET 0.40121 0.04797 8.364 0.000 CONSTANT 9.1192 0.3935 23.17 0.000 Table 4 Regression of share prices changes and CEO compensation (whole sample) NAME COEFFICIENT ERROR 272 DF P-VALUE PRICECHANCE 0.019700 0.05354 0.3680 0.713 LOGASSET 0.40149 0.04864 8.255 0.000 CONSTANT 9.1073 0.4003 22.75 0.000 It is very unlikely that CEO compensation just depends on one of three performance variables. Estimated coefficients might be incorrectly specified if relevant variables are omitted from the regression model and this can lead to very serious consequences such as incorrect signs of coefficients and level of significance. It is believed that all three performance variables, namely dividend payout, sales growth, and changes in share price, will interactively affect managerial compensation. Table 5 present the 12

OLS estimates for the regression model that capture interactive effect of all three performance variables on CEO compensation. The result is very similar to the individual regression models estimated above. The coefficient of each variable is the elasticity of CEO compensation with the other variables held constant. For example, the elasticity of CEO compensation with respect to the firm size is 0.4047, with sales growth, dividend payout, and share price being held constant. There is still strong evidence showing that there is a strong positive relation between firm size and managerial compensation. However, the p values for the performance variables are still very large, suggesting that the relation between managerial compensation and firm performance is still weak. Table 5 Regression of CEO compensation with all explanatory variables (whole sample) NAME COEFFICIENT ERROR 270 DF P-VALUE SALESCHA -0.0020277 0.001364-1.487 0.138 PRICECHA 0.028357 0.05401 0.525 0.600 DIVCHANG 0.038897 0.08716 0.4463 0.656 LOGASSET 0.40477 0.04868 8.316 0.000 CONSTANT 9.0818 0.4003 22.69 0.000 In summary, from the total sample, all regression model shows that there is a strong positive relation between firm size and managerial compensation, which is consistent with the findings of the existing literature. Results from all the firm-performance regression models also provide evidence that the relation between managerial compensation and firm performance is weak. The significance level of the coefficients is quite low in most of the regression models discussed so far. A possible explanation is that the specification of the models may not adequately capture the sample s intra-firm relationship of pay to performance, owing to a strong cross-sectional effect of firm size. Therefore we estimate the same set of regression models on two subsamples, where one subsample consists of firms with total asset greater than the sample median of the firm s total asset and the other subsample consists of firms with total asset less than the sample median. 13

Table 6a and 6b presents the results of the regression of firm size and CEO compensation of the two subsamples. The strong positive relation between firm size and CEO compensation still exists in both subsample. However, the firm size elasticity of CEO compensation is greater in larger firms: for every one percent increases in the firm s total asset, there is a 0.36740 percent increase in average CEO compensation of larger firms, where there is only a 0.23806 percent increase in average CEO compensation of smaller firms. Table 7a to table 9b show the result of different performance variables and CEO compensation. For performance variables, the sign and magnitude of the coefficients of these variables and the p values are similar to the individual regression models. The elasticity of CEO compensation of large and small firms are similar and the relation is still weak. Table 6a Regression of firm size and CEO compensation (smaller firms) NAME COEFFICIENT ERROR 136 DF P-VALUE LOGASSET 0.23806 0.09360 2.543 0.012 CONSTANT 10.593 0.8247 12.84 0.000 14

Table 6b Regression of firm size and CEO compensation (larger firms) NAME COEFFICIENT ERROR 135 DF P-VALUE LOGASSET 0.36740 0.1166 3.150 0.002 CONSTANT 9.3205 0.8805 10.59 0.000 Table 7a Regression of dividend payout and CEO compensation (smaller firms) NAME COEFFICIENT ERROR 135 DF P-VALUE DIVCHANG 0.026105 0.1012 0.2580 0.797 LOGASSET 0.24160 0.09492 2.545 0.012 CONSTANT 10.558 0.8385 12.59 0.000 Table 7b Regression of dividend payout and CEO compensation (larger firms) NAME COEFFICIENT ERROR 134 DF P-VALUE DIVCHANG 0.019375 0.1635 0.1185 0.906 LOGASSET 0.36588 0.1178 3.107 0.002 CONSTANT 9.3303 0.8876 10.51 0.000 Table 8a Regression of sales growth and CEO compensation (smaller firms) NAME COEFFICIENT ERROR 135 DF P-VALUE SALESCHA -0.0027714 0.001806-1.535 0.127 LOGASSET 0.25631 0.09390 2.730 0.007 CONSTANT 10.441 0.8266 12.63 0.000 Table 8b Regression of sales growth and CEO compensation (larger firms) NAME COEFFICIENT ERROR 134 DF P-VALUE SALESCHA -0.00069791 0.002009-0.3474 0.729 LOGASSET 0.36755 0.117 3.141 0.002 CONSTANT 9.3220 0.8834 10.55 0.000 15

Table 9a Regression of changes in share price and CEO compensation (smaller firms) NAME COEFFICIENT ERROR 135 DF P-VALUE PRICECHA 0.0017651 0.1464 0.01205 0.990 LOGASSET 0.23830 0.09603 2.482 0.014 CONSTANT 10.591 0.8487 12.48 0.000 Table 9b Regression of changes in share price and CEO compensation (larger firms) NAME COEFFICIENT ERROR 134 DF P-VALUE PRICECHA 0.019280 0.05959 0.3236 0.747 LOGASSET 0.37186 0.1178 3.156 0.002 CONSTANT 9.2828 0.8910 10.42 0.000 By decomposing the sample into five different industries, we can see that the average managerial compensation for these industries is consistent with the stylised facts: executives in financial services firms earn higher pay, and in utilities earn lower pay, than their counterparts in other industries. However, there are only very few firms in each category and the statistics are not significant, therefore the numbers are not shown here and none of the estimated regression models are statistically significant as the number of observations are too few. 6. Future Research This paper has been focused on the evidence parallel to those documented by previous studies. There are important issues to be addressed concerning the New Zealand executive compensation system. For example, how are managerial pay schemes in New Zealand different from those in other countries, and what is the impact of the pay schemes on New Zealand firms performance? The difference between New Zealand and the U.S. is of particular interest. Despite extensive linkages between the two countries, there are a number of institutional and market differences between New Zealand and the United States, and so the pay system and their effectiveness are expected to be different. In this paper, comparisons have frequently been made 16

between the two countries by referring to earlier studies. A more careful intersample comparison is needed, however, to obtain a complete and formal evaluation of the New Zealand-U.S. difference; as sample heterogeneity often plays an important role in estimating pay-performance sensitivities, such issues can be properly addressed in an intersample analysis based on closely comparable samples. 7. Conclusion This paper provides a systematic examination of the relationship between CEO compensation, and firm size and corporate performance for New Zealand companies. The evidence is consistent with, and largely similar to, the findings of previous studies for other countries. It is found that CEO pay rises with firm size. It is found that the elasticity of CEO compensation with respect to the firm size is 0.39. For every percentage increases in the firm s total asset, the cash compensation paid to the average CEO increases by 0.39 percent. This result is very close to the finding of many previous studies that identify a firm-size elasticity between 0.2 to 0.35 for larger U.S., U.K. and Japanese firms. As for firm performance, it has proven that it has a very weak relation with CEO compensation, which is also consistent with the existing literature. 17

8. References Baker, George P., Michael C, Jensen, and Kevin J. Murphy (1998) Compensation and incentives: practice vs. theory, Journal of Finance 43, 593-616 Black, Fischer, and Myron Scholes (1973) The pricing of options and corporate liabilities, Journal of Political Economics 81, 637-54 Chaykowski, Richard P., and Brain Lewis (1995) Compensation Practice and Outcomes in Canada and the United States (Kingon: IRC press, Queen s Univerisity) Ciscel, David H., and Thomas M. Carroll (1980) The determinants of executive salaries: an econometric survey, Review of Economics and Statistics 62, 6-13 Cosh, A (1975) The remuneration of chief executives in the United Kingdom, Economic Journal 85, 75-94 Coughlan, Anne T., and Ronald M. Schmidt (1985) Executive compensation, managerial turnover, and firm performance, Journal of Accounting and Economics 7, 67-84 Garen, John E. (1994) Executive compensation and principal-agent theory, Journal of Political Economics 102, 1175-99 Gibbons, Robert, and Kevin Murphy (1992) Optimal incentive contracts in the presence of career concerns: theory and evidence, Journal of Political Economiy 100, 468-505 Hall, Brain J., and Jeffrey B. Liebman (1997) Are CEOs really paid like bureaucrats? Discussion Paper No. 1789, Harvard Institute of Economic Research Haubrich, Joseph G. (1994) Risk aversion, performance pay, and the principal-agent problem, Journal of Political Economy 102, 258-76 Holmstrom, Bengt (1979) Moral hazard and obervability, Bell Journal of Economics 10, 74-91 Hubbard, R. Glenn, and Darius Palia (1995) Executive pay and performance evidence form the U.S. banking industry, Journal of Financial Economics 39, 105-30 Jensen, Michael C., and Kevin J. Murphy (1990a) Performance pay and topmanagement incentives, Journal of Political Economy 98, 225-64 - (1990b) CEO incentives it s not how much you pay, but how, Harvard Business Review (May/June), 138-49 Joskow, Paul, and Nancy Rose (1994) CEO pay and firm performance: dynamics, asymmetries, and alternative measures of performance, NBER Working Paper No. 4976 Joskow, Paul, Nancy Rose, and Andrea Shepard (1993) Regulatory constraints on CEO compensation, Brookings Papers on Economic Activity Microeconomics, 1-72 Kaplan, Steven N. (1994) Top executive rewards and firm performance: a comparison of Japan and the United States, Journal of Political Economy102, 510-46 Kato, Takao, and Mark Rockel (1992) Experiences, credentials, and compensaton in the Japanese and U.S managerial labor markets: evidence from new micro data, Journal of the Japanese and International Economics 6, 30-51 Kostiuk, Perter F. (1989) Firm size and executive compensation, Journal of Human Resources 25, 91-105 18

Leonard, J. (1990) Executive pay and firm performance, Industrial and Labor Relations Review 43, 13S-29S Lewellen, Wilber G., and Blaine Huntsman (1970) Managerial pay and corporate performance, American Economic Review 60, 710-20 Masson, Robert (1971) Executive motivations, earnings, and consequent equity performance, Journal of Political Economy 79, 1278-92 McLaughlin, Kenneth J. (1994) Individual compensation and firm performance: the economics of team incentives, Working Paper No. 104, Center for the Study of the Economy and the State, University of Chicago Mirrlees, J.A. (1976) The optimal structure of incentives and authority within an organization, Bell Journal of Economics 7, 105-31 Murphy, Kevin J. (1985) Corporate performance and managerial remuneration, an empirical analysis, Journal of Accounting and Economics 7, 11-42 -(1986) Incentives, learning, and compensation: a theoretical and empirical investigation of managerial labor contracts, Rand Journal of Economics 17, 69-77 -(1998) Executive compensation, forthcoming in Handbook of Labor Economics, Vol. 3 ed Orley Ashenfelter and David Card (Amsterdam North-Holland) Roberts, D.R. (1956) A general theory of executive compensation based on statistically tested prepositions, Quarterly Journal of Economics 70, 270-94 Rosen, Sherwin (1992) Contracts and the market for executives, in Contract Economics, ed. Lars Werin and Hans Wijkander (Malden, MA: Blackwell) Schaefer, Scott (1995) The dependence of CEO pay-performance sensitivity on the value of the firm. Manuscript, Kellogg Graduate School of Management, Northwestern University Sloan, R. (1993) An examination of the role of accounting earnings in top executive compensation contracts, Journal of Accounting and Economics 16, 55-100 Zhou, Xianming (1998) Executive compensation and managerial incentives: a comparison between Canada and United Sates, manuscript, University of Sydney 19