The impact of current tax policy on CEO stock option compensation: a quantile analysis

Martin Gritsch

ABSTRACT

The increase in stock option compensation has become controversial as disclosures emerge that executives have abused stock option compensation. The Omnibus Budget Reconciliation Act of 1993 gives stock options preferential tax treatment, allowing options to not be subject to the cap on corporate tax deductions for salaries exceeding $1 million. To examine whether this increases CEO stock option compensation, we use S&P’s ExecuComp data from 1992-2000. We use quantile regressions to examine whether the position in the distribution affects behavior. Findings show that the salary cap has increased stock option compensation; strongest effects are at the bottom of the distribution.

INTRODUCTION

The increase in stock option grants during the 1990s is becoming increasingly controversial as disclosures emerge that senior executives of companies such as Enron Corp. and Global Crossing reaped millions of dollars in profit by exercising their stock options as the public held onto stock that became worthless. For example, Global Crossing reported a negative net income of $10,500,000, while they still issued the CEO, Mr. Annunziata, $182,000,000 worth of stock options. In addressing the corporate scandals, Senator Carl Levin recently suggested that stock option pay encourages firms and executives to push tax law to the limit.

Our current tax code gives stock options preferential tax treatment. The Omnibus Budget Reconciliation Act of 1993, Section 162(m) of the Internal Revenue Code (IRS par. 9001B, sec. 1.162) increases the incentive of firms to pay executives in the form of stock options. Specifically, Section 162(m) allows “performance-based pay,” such as stock options, to not be subject to the limits or the cap on corporate income tax deductions for salaries exceeding $1 million. To ensure the full tax deductibility of top executives’ pay, companies may shift any amount in excess of $1 million to “performance based pay,” such as stock options.

In this paper, we use annual data from Standard & Poor’s ExecuComp on the CEO of 2,412 firms from 1992 to 2000 to examine if the 1993 SEC Section 162(m) cap on salary compensation reform has altered the mix of compensation that firms pay executives. In our analysis, we control for the changes in the stock market, firm size, firm performance and financial structure. Results will help determine to what extent the 1993 salary cap encourages firms to pay executives in stock options to obtain a tax advantage. Since companies may behave differently dependent on their position in the distribution, we then use a quantile regression technique.

STOCK OPTIONS TAX TREATMENT

Currently, a nonqualified stock option is taxed under Code Section 83, stating that an employer is allowed to deduct the full value of the employee’s income from the exercise and sale of stock options. While companies can deduct stock options from their corporate taxes, salary is limited on the amount that it can be deducted from taxes. The Omnibus Budget Reconciliation Act of 1993 (OBRA 1993), Section 162(m) of the Internal Revenue Code (IRS par. 9001B, sec. 1.162), eliminated the deductibility of executive salary compensation in excess of $1 million. This regulation limits the corporate tax deduction for compensation paid to the top five highest-paid executive officers to a 1$ million cap each. However, “qualified performance based pay” was exempt by (IRS par. 9001B, sec. 1.162-7[e][1]). (1)

While this supposedly constrains executive compensation, it may have just altered the mix of compensation towards “performance based pay.” For stock options to qualify for an exemption from this limit, the total number of options and the limit of their reward must be approved by the outside directors. The exemption requirements are much stricter for bonus and long-term incentive plans. As opposed to salary and bonuses, stock options are generally not subject to the limits of Section 162(m). To ensure the full tax deductibility of top executives pay, companies may shift any amounts in excess of $1 million to incentive based forms of compensation, such as stock options.

LITERATURE REVIEW

There have been few studies examining if the recent salary cap imposed by Section 162(m) has altered the mix of CEO compensation and the tax effects of stock option compensation. (2) Among those studies, Rose and Wolfram (2002) and Hall and Liebman (2000) examine if the million dollar cap has decreased the use of salary compensation by affected firms and increased their use of stock options compensation. Both studies find evidence that the salary cap has lowered the potential growth of CEO salaries, however they find differing results regarding if this cap has increased the use of stock options. While Rose and Wolfram find little support for an increase in performance based pay, Hall and Liebman (2000) found support for the salary cap creating an increase in the use of stock option compensation. They suggest that most firms that reduced salaries from above the $1 million cap to below, cite that Section 162 (m) is the reason for such compensation adjustments. Hall and Liebman (1998) find that the median CEO receives 1.1 million in salary and that for executives earning in excess of $1 million, the growth rate in their salary between 1993 -1998 is zero.

Hall and Liebman (2000), along with Goolsbee (2000), examine how current tax policy impacts the use of stock options and shifting forms of executive compensation. Goolsbee shows that the short-run elasticity of marginal tax rates comes exclusively from stock options, suggesting that if one excludes stock options there is no role for tax policy in reducing executive compensation. Hall and Liebman show that the shift in executive compensation towards stock options is caused more by increases in the stock market and the market value of the firm. For example, they find that a 10 percent increase in the market value of a firm leads to an increase in stock options by $1.25 million and that the S&P 500 index alters the mix much more than changes in the tax policy.

ECONOMETRIC TECHNIQUE AND DATA

Due to Section 162 (m), businesses can no longer deduct salaries greater than $1 million from their tax payments. To maximize the tax benefits, firms may pay executives their first million in salary and then through performance-based pay, like stock options. This may increase the share of stock option compensation. To empirically investigate if the salary cap has altered stock option pay, we regress the CEO compensation measures on the salary cap for affected firms. This equation is represented by:

ln([StockOptions.sub.i]) = [[alpha].sub.0] + [[alpha].sub.1] [affected.sub.i] + [[alpha].sub.2] [affected.sub.i] x Cap + [summation over (k)] [[alpha].sub.i] [x.sub.k.i] + [[epsilon].sub.i] (1)

where [StockOptions.sub.i] is the natural logarithm of the Black-Scholes value that a CEO received in a particular year. The [x.sub.i] are regressors to control for factors such as the overall performance of the stock market (we include the mean value of last year’s S&P 500 index), firm performance (a firm’s lagged market value and the return on assets), and the size of the company (we include one dummy variable each for companies which are S&P 500 companies or part of the S&P SmallCap, respectively), the [alpha]’s are parameters and [[epsilon].sub.i] is an error term.

Even after including these control variables, we may not be able to fully capture the factors, which determine how much stock option pay a CEO receives. Thus, we include a variable which takes on the value of one if a CEO, based on pre-cap salary, is predicted to would have received in excess of $1 million in salary had the salary cap not been enacted, zero otherwise. The [affected.sub.i] variable is used to capture the relationship between the level of salary and bonus compensation and the growth rate of salary and bonus. The [affected.sub.i] dummy is one for firms with predicted total CEO compensation exceeding the $1 million cap, zero otherwise. Following Rose and Wolfram (2002), we construct predicted total compensation for the years after the 1994 cap by estimating an AR(1) model using data prior to 1994. (3) Cap is a dummy variable for the years that the $1 million dollar cap is in place. The [affected.sub.i] x Cap interaction term is used to capture the differential growth rate of this compensation after the cap is put into place.

However, it may very well be the case that the effects of the salary cap are not the same for all CEOs. In fact, we suspect that there are pronounced differences based on a CEO’s position in the distribution of stock option compensation. Thus, we estimate the effects of the salary cap–while controlling for a number of other factors–on CEOs’ stock option compensation using a quantile regression technique. That is, we estimate the conditional median (rather than the conditional mean in an OLS regression) for various parts of the distribution of the dependent variable. More specifically, we carry out the same estimation for nine different quantiles of the distribution of stock option compensation to obtain estimates for the conditional median as well as the 10th, 20th, 30th, etc. percentiles.

After examining the effect of the salary cap on CEO’s stock option compensation, we address an additional issue: If it is indeed the case that CEO’s stock option compensation increased in response to the enactment of the salary cap, one might expect that it is simply a shift from salary compensation to stock option compensation, i.e., one would expect a decrease in salary at the time of the stock option compensation increase. In order to investigate this issue, we estimate the following equation:

ln([Salary.sub.i]) = [[beta].sub.0] + [[beta].sub.1] [affected.sub.1] + [[beta].sub.2] [affected.sub.i] x Cap + [summation over (k)] [[beta].sub.i] [x.sub.k.i] + [[eta].sub.i] (2)

where [Salary.sub.i] is the annual salary of the ith CEO and all the regressors are as defined in the estimation for stock option compensation in equation (1) above. Again, we first carry out an OLS regression followed by a quantile regression analysis in order to obtain estimates for the conditional median as well as the 10th, 20th, 30th, etc. percentiles.

We collect annual compensation data from Standard and Poor’s ExecuComp database from 1992 to 2000. Thus, we can examine the changes in tax law, which took place after the Section 162 (m) tax law change in 1993 that created a salary cap limit, favoring “performance based pay” over salary from a tax perspective. One advantage of the ExecuComp database is its large size. It follows a total of 2,412 companies over time, which are or were a member of the S&P 1,500 (consisting of the S&P 500, the S&P MidCap 400, or the S&P SmallCap 600). Since each company must provide information about the top five executives in each year, the overall number of records is substantial. After imposing some restrictions (most notably the restriction to CEOs as well as the exclusion of firms whose fiscal year does not end in December in order to properly account for any tax change effects and the exclusion of executives whose reported value of stock options granted in a certain year is missing), there are 6,062 individual-year observations that are included in our estimations of equation (1) and 8,235 observations in our estimations of equation (2). (4)

RESULTS

Estimation results from the OLS regression of the stock option compensation equation (equation (1)) are shown in Table 2. The two variables of main interest show the following patterns: Before the salary cap went into effect, “affected” firms paid their CEOs approximately 15 percent less in stock options, ceteris paribus. After the cap went into effect, CEO stock option compensation was substantially higher in affected firms after controlling for a number of other factors. The average affected firm issued approximately 33 percent (48.3 percent minus 15.3 percent) more in stock options to their CEO than the unaffected firms. That is, in contrast to results found in Rose and Wolfram (2002), our results suggest a significant increase in the use of “performance based compensation” for affected firms. Our results are consistent with those found in Hall and Liebman (2000). With the exception of the dummy variable indicating an S&P500 firm, all estimates of the control variables are highly statistically significant. The effect of return on assets is negative which may have to do with low-performing firms that try to attract a CEO who can turn a company around by offering a substantial amount of stock options. S&P500 companies issue their CEOs close to 3 percent more in stock options, all else equal, but this is the one estimate with a large p-value. Small companies, on average, issue approximately 16 percent less in stock options to their CEOs than do their larger peers. Our results support Hall and Liebman’s (2000) findings that past stock market performance has a positive impact on the average CEO’s stock option compensation. Similarly, there is also a positive relationship between a firm’s market value and the value of stock options granted. Our results indicate that a 10-percent increase in the market value of the firm is associated with a 37-percent increase in CEO stock option compensation. Evaluated at the mean of the data, this result corresponds with an increase of approximately $1 million, a finding that is remarkably close to Hall and Liebman’s result of $1.25 million. Unlike Hall and Liebman, our results continue to show that the salary cap has increased the use of stock option compensation even after including these control variables.

We now take a closer look at the two variables of main interest in our study using a quantile analysis. As previously discussed, estimation via OLS may obscure some important differences in firm/CEO behavior. More specifically, firms at the bottom of the distribution may show different results from those around the median or at the top of the distribution. These issues can be addressed with the use of quantile regression. To indicate pre-cap behavior, the dummy variable (“affected”) predicts salary compensation in excess of $1 million had the salary cap not been instituted. Estimates are presented in the second column of Table 3. The estimates for the first five quantiles are small in magnitude and not statistically significant at any conventional level. The results for the upper four quantiles are quite different: For that top part of the distribution, we find that affected firms (i.e., those whose predicted post-cap CEO salary is in excess of $1 million), on average, pay their CEOs approximately 21 to 26 percent less in stock options, all else equal. This may be due to the fact that before the salary cap was enacted, a larger share of CEO compensation was paid in the form of salary.

An even more interesting finding, in our opinion, emerges when the behavior of CEOs of affected firms in response to the enactment of the salary cap is examined. The full effect is captured by the sum of the estimates of the dummy variable for affected firms and the interaction term of that dummy with the salary cap term, which indicates whether the cap was in or not in effect. The total impact is presented in column 3 of Table 3 as well as in Figure 1. (5)

The full effect exhibits a very clear downward trend over almost the entire distribution. The largest effect is shown for CEOs in the bottom quantile. The combined effect is estimated to be 0.61, i.e., our results indicate that after the salary cap went into effect, the average CEO of an “affected” firm received 61 percent more in stock options than a CEO at a firm that was not affected by the salary cap. The estimated effect is strictly monotonically declining over the entire distribution with the exception of the top quantile.

The statistical precision of the estimates of the interaction term between the dummy variable showing affected firms and the dummy variable which indicates whether the cap was in or not in effect is noteworthy: All nine estimates have p-values of less than 0.01. Keeping in mind that the (negative) estimates of the “affected” dummy have high p-values for the bottom five quantiles, we can conclude that the effects shown in Figure 1 for the bottom half of the distribution are in fact lower bounds; they might well be larger than what is shown in the table and the figure.

Since we estimate a separate regression for each of the nine quantiles, we do not present all the estimates for the control variables. Instead, we briefly summarize the results for our control variables. In the quantile regressions for stock option compensation, none of the estimates of the dummy for S&P 500 companies is statistically significant at any conventional level, and they are small in magnitude. The estimates for the effect of being a S&P SmallCap company have p-values from 0.26 to 0.85 for the lowest four quantiles. The p-values for the other five quantiles are less than 0.01. These precisely estimated effects for the top half of the distribution are all negative and strictly monotonically increase (in absolute value) from -0.14 for the median quantile to -0.36 for the top quantile. The estimated effects of the lagged S&P 500 and a firm’s market value are positive and highly statistically significant in all nine regressions. The coefficient estimate (rounded to the third decimal place) is 0.001 in all nine regressions. The estimate of a firm’s return on assets is negative, but relatively small in size and, in the case of the lowest two quantiles, not statistically significant.

[FIGURE 1 OMITTED]

In order to see whether the increase in stock option compensation coincided with a decrease in CEO salaries, we then estimate the salary compensation equation (equation (2)). Results of the OLS regression are presented in Table 4. The estimate for the “affected” dummy variable (which captures whether a firm is predicted to be affected by the salary cap) shows that the CEOs of affected firms, on average, receive a salary, which is almost 50 percent higher than the salary of CEOs at unaffected firms. This large positive (and highly statistically significant) estimate is to be expected since the “affected” dummy variable takes on the value “one” whenever we predict that a CEO receives a large salary (based on our auxiliary AR(1) regression). Examination of the estimate for the interaction term reveals that, interestingly, the enactment of the salary cap had virtually no effect on salary compensation. The estimated reduction of approximately one percent is clearly not statistically significant (p-value = 0.82). That is, it is not the case that the increase in stock option compensation shown previously occurred in conjunction with a decrease in salary compensation. Instead, stock option compensation rose substantially in response to the enactment of the salary cap while mean salaries were virtually unaffected. Unlike, Rose and Wolfram (2002), who found that the salary cap greatly reduced CEO salaries, our results show that the salary cap has had very little impact on CEO salary compensation. This is consistent with Hall and Liebman’s (1998) finding that the median CEO salary between 1993 and 1998 was close to $1 million. Interestingly, this corresponds perfectly with the $1 million salary cap creating no incentive to lower the majority of CEO salaries. Results also support Hall and Liebman (2000), again showing that a firm’s market value has a highly statistically significant, positive impact on a CEO’s compensation. As they suggest, this is due to CEO compensation being based on firm performance, with good performance reaping higher compensation. CEOs of small companies, ceteris paribus, receive approximately 8 percent lower salaries than their counterparts in medium-sized corporations (the omitted category). (6)

We now estimate equation (2) using quantile regressions in order to examine whether there are distinctly different patterns dependent on the position in the distribution. As can be seen from the second column in Table 5, (7) the effect of an affected firm is most pronounced for the lowest quantiles, then decreases somewhat and is virtually constant for the top six quantiles. In order to examine the effect of the enactment of the salary cap, the combined impact of the dummy variable and the interaction term must be considered. As can be seen from Table 5, the second and the third columns of numbers are almost identical. This implies that the estimates for the interaction term are very small. In fact, the range of estimates is from -0.029 to +0.019. Not a single one of these nine estimates is statistically significant (p-values range from 0.15 to 0.98.) This means that the post-cap behavior in terms of salary is, for all practical purposes, the same as the pre-cap behavior. Thus, our previous finding that stock option compensation increased substantially while salary compensation was virtually unaffected by the enactment of the salary cap approximately applies to all quantiles of the distribution.

To summarize the results, we find that overall, the salary cap has increased the use of stock option compensation for affected firms. More importantly, however, we have been able to show that there exist very distinct patterns dependent on the part of the distribution that is being estimated. For example, the total effect of the salary cap is more than five times as high for the bottom quantile as it is for the second-highest quantile. Moreover, this result may even understate the true effect because of the varying degrees of precision of the estimates as discussed above. Additionally, we were able to demonstrate that the increase in stock option compensation due to the salary cap did not occur simultaneously with a decrease in salaries. Instead, salary compensation, surprisingly, remained virtually unaffected by the salary cap.

[FIGURE 2 OMITTED]

CONCLUSIONS AND POLICY IMPLICATIONS

There has been a tremendous increase in executive stock option compensation during the 1990s. During the same time period, a tax law change created a salary cap for tax deductions. By analyzing the effects of the tax cap, we can consider if tax treatment of compensation influences stock option compensation. Using a quantile analysis allows us to further examine if CEO compensation differences based on a CEO’s position in the compensation distribution.

Our results suggest that the salary cap has increased the use of stock options as a form of executive compensation. This is especially true for companies and their CEOs at the bottom of the distribution. However, our results show very little change in salary compensation, regardless of the CEO’s position in the respective distribution. This suggests that the salary cap has not reduced CEO salary compensation, but has greatly increased their stock option compensation, especially at the lower end of the distribution. This also suggests that since the implementation of the salary cap in 1993, most increases in CEO compensation have taken the form of stock option compensation. While CEOs may not be representative for the general public, their response to current tax policy may be interesting in its own right, especially considering the magnitude of their overall incomes. Additionally, CEO compensation has received substantial attention in light of the current scandals. Presumably, the government’s intention in 1993 was to limit executive compensation. However, we find that CEO salary compensation was not affected by the enactment of the salary cap. Moreover, stock option compensation increased substantially after the salary cap went into effect. In summary, the salary cap was not only ineffective in curbing executive compensation; it may actually have led to the opposite effect, i.e., a substantial increase in overall CEO compensation.

REFERENCES

Brookfield, David and Ormrod Phillip (2000) “Executive Stock Options: Volatility, Managerial Decisions and Agency Costs,” Journal of Multinational Financial Management, v10, n3-4, Sept.-Dec., pp. 275-95.

Bryan, Stephen, LeeSeok Hwang and Steven Lilien (2000) “CEO Stock-Based Compensation: An Empirical Analysis of Incentive-Intensity, Relative Mix, and Economic Determinants,” Journal of Business, v73, n4, pp. 661-694.

Dhillon, Upinder S. and Gabriel G. Ramirez (1994) “Employee Stock Ownership and Corporate Control: An Empirical Study,” Journal of Banking and Finance, v18, n1 January, pp. 9-26.

Feldstein, M. and D. Feenberg (1996) “The Effects of Increased Tax Rates on Taxable Income and Economic Efficiency: A Preliminary Analysis of the 1993 Tax Rate Increases,” Tax Policy and the Economy, v10, pp. 89-117, Cambridge: MIT Press for the National Bureau of Economic Research.

Goolsbee, Austan (2000) “What Happens When You Tax the Rich? Evidence from Executive Compensation,” Journal of Political Economy, v108, April, pp. 352-78.

Goolsbee, Austan (2000), “Taxes, High-Income Executives, and the Perils of Revenue Estimation in the New Economy”, American Economic Review Papers and Proceedings, (May), pp. 271-275.

Graham, John R., and Michael Lemmon, (1998), “Measuring Corporate Tax Rates and Tax Incentives: A New Approach,” Journal of Applied Corporate Finance, 11, pp. 54-65.

Hall, Brian and Jeffrey Liebman (1998), “Are CEOs Really Paid Like Bureaucrats?” The Quarterly Journal of Economics, v113, n3, pp. 653-91.

Hall, Brian and Jeffrey Liebman (2000), “The Taxation of Executive Compensation”, National Bureau of Economic Research, March, Working Paper 7596.

Jensen, M.C. and W. Meckling (1976) “Theory of the Firm: Managerial Behavior, Agency Costs and Ownership Structure,” Journal of Financial Economics, v3, n4 October, pp. 305-60.

Mehran, H. (1995) “Executive Compensation Structure, Ownership, and Firm Performance,” Journal of Financial Economics, v38, pp. 163-184.

Mehran, H., and J. Tracy (2001) “The Effect of Employee Stock Options on the Evolution of Compensation in the 1990sof Employee Stock Options on the Evolution of Compensation in the The Effect of Employee Stock Options on the Evolution of Compensation in the 1990s,” Federal Reserve Bank of New York Economic Policy Review, v7, n3, December, pp.17-34.

Rose, Nancy L. and Catherine Wolfram (2002), “Regulating Executive Pay: Using the Tax Code to Influence Chief Executive Officer Compensation” Journal of Labor Economics, v. 20, n. 2, pp. 138-175.

Rosen, Corey (1990) “The Record of Employee Ownership,” Financial Management, 19,1, Spring, pp. 39-47.

Sanders, G (2001) “Behavioral Responses of CEOs to Stock Ownership and Stock Options,” Academy of Management Journal, v44, n3, pp. 473-492.

ENDNOTES

(1) See Rose and Wolfram (2002).

(2) It should be noted that agency theorists such as Jensen and Meckling (1976), Rosen (1990), and Brookfield and Phillip (2000) have addressed how stock options impact the principal-agent dilemma, while Sanders (2001), Mehran (1995), and Dhillon and Ramirez (1994), addressed whether the growing number of stock options impact firm performance. However, they all neglect to address whether current federal tax policy limiting tax deductible salary compensation alters the mix of executive compensation.

(3) Section 162 (m) states that only predetermined “performance based pay” that is voted on by outside members of the board are qualified to be exempt for the million dollar cap. Rose and Wolfram found that while almost all stock option grants are qualified, the majority of bonuses are not predetermined and thus are not qualified. Results are robust if we only examine affected firms based on salary compensation and are available upon request

(4) The difference in the number of observations in the two sets of regressions is due to the fact that a number of CEOs received zero stock options in a particular year which results in a missing value for the constructed natural logarithm of that variable.

(5) While the second column presents estimates of a1 from the nine quantile regressions, the third column shows the sum of a1 and a2, i.e., the a2 can be obtained by forming the difference between the third column and the second column.

(6) The estimates of the other control variables are not statistically significant at any conventional level.

(7) Analogously to the earlier set of quantile regressions, the second and third columns are the estimates of [[beta].sub.1] as well as the sum of [[beta].sub.1] plus [[beta].sub.2] so that [[beta].sub.2] can be obtained by forming the difference between the two columns.

Martin Gritsch, William Patterson University

Tricia Coxwell Snyder, William Paterson University

Table 1: Summary Statistics of Variables Included in Estimations

Variable Mean Standard Deviation

Affected 0.424 0.494

S&P 500 index (lagged) 752.4 309.4

Market Value ($ million) 6,162.9 18,721.0

ROA 3.1 13.3

S&P500 firm 0.322 0.467

S&P SmallCap firm 0.201 0.401

Table 2: OLS Regression Results for Stock Option Compensation

(Standard Errors in Parentheses)

Ln(Black-Scholes value of SO)

Affected -0.153

(0.085)

Affected * cap 0.483

(0.083) **

S&P 500 index (lagged) 0.001

(0.000) **

Ln(MV) ($ million) 0.367

(0.015) **

ROA -0.004

(0.001) **

S&P500 firm 0.029

(0.042)

S&P SmallCap firm -0.165

(0.045) **

Constant -3.691

(0.101) **

Observations 6,062

R-Squared 0.31

* significant at 5%

** significant at 1%

Table 3: Effect of the Salary Cap on Stock Option Compensation

by Quantile

Pre-Cap Effect Post-Cap Effect

on Stock Options on Stock Options

Quantile (Affected Firms) (Affected Firms) Pseudo R-Squared

1 -0.106 0.606 0.18

2 -0.065 0.493 0.19

3 -0.036 0.389 0.19

4 -0.027 0.343 0.20

5 -0.043 0.286 0.20

6 -0.240 0.223 0.19

7 -0.244 0.134 0.19

8 -0.214 0.115 0.18

9 -0.256 0.196 0.16

Table 4: OLS Regression Results for Salary Compensation

(Standard Errors in Parentheses)

Ln(Salary)

Affected 0.492

(0.044) **

Affected * cap -0.010

(0.044)

S&P 500 index (lagged) 0.001

(0.001)

Ln(MV) ($ million) 0.104

(0.007) **

ROA 0.001

(0.001)

S&P 500 firm 0.035

(0.022)

S&P SmallCap firm -0.079

(0.022) **

Constant -1.658

(0.051) **

Observations 8,235

“R-Squared” 0.22

* significant at 5%

** significant at 1%

Table 5: Effect of the Salary Cap on Salary Compensation

by Quantile

Pre-Cap Effect Post-Cap Effect

on Salary on Salary

Quantile (Affected Firms) (Affected Firms) Pseudo R-Squared

1 0.528 0.530 0.24

2 0.451 0.470 0.31

3 0.418 0.421 0.34

4 0.389 0.394 0.36

5 0.377 0.378 0.37

6 0.375 0.364 0.37

7 0.380 0.355 0.37

8 0.387 0.358 0.35

9 0.372 0.362 0.31

COPYRIGHT 2005 The DreamCatchers Group, LLC

COPYRIGHT 2008 Gale, Cengage Learning