Economic subjects | Investments, Stock exchange » Balsam-Ryan - The Effect of Internal Revenue Code Section 162m on the Issuance of Stock Options

Source: http://www.doksinet The Effect of Internal Revenue Code Section 162(m) on the Issuance of Stock Options Steven Balsam 215-204-5574 steven.balsam@templeedu David Ryan david.ryan@templeedu 215-204-8131 Temple University Fox School of Business Philadelphia, PA 19122 Source: http://www.doksinet The Effect of Internal Revenue Code Section 162(m) on the Issuance of Stock Options Abstract IRC section 162(m) limits tax deductibility of executive compensation to $1 million per covered executive, with an exception for performance based compensation. Consequently, firms that wish to pay an executive more than $1 million either have to forfeit deductions or structure the compensation package so that the excess over $1 million qualifies under the performance based exception. While a variety of compensation forms can qualify as performance based, they vary in the difficulty of qualification and the degree of risk that qualification imposes on the executive. Amounts paid under a bonus

plan, for example, qualify as performance based, if the payout does not exceed that determined based upon objective plan parameters set at the beginning of the year. In contrast, almost any stock option grant qualifies as performance based While there clearly is risk associated with both stock option and annual bonus compensation, the requirements of section 162(m) changed the risk associated with annual bonus compensation relative to what it was prior. Consequently, section 162(m) may have encouraged the use of stock options vis-à-vis other forms of compensation. Using the population of firms available on the Standard & Poor’s ExecuComp database, we find that the propensity to issue stock options has increased for affected executives, both in absolute terms and as a percentage of total compensation. In additional analysis, we find evidence that is consistent with this increase in stock option compensation substituting for lower increases in salary for affected executives. But,

we find no evidence of stock option compensation substituting for annual cash bonuses. We interpret this as indicating firms and 1 Source: http://www.doksinet their executives act in a way consistent with the incentives provided by section 162(m). Keywords: IRC section 162(m), executive compensation, tax deductibility, costs and benefits 2 Source: http://www.doksinet The Effect of Internal Revenue Code Section 162(m) on the Issuance of Stock Options INTRODUCTION The Revenue Reconciliation Act of 1993 added Internal Revenue Code section 162(m), limiting the corporate tax deduction for executive compensation to $1 million per individual for the top five executives of a corporation, providing an exception for compensation in excess of $1 million if it qualifies as "performance-based." Consequently, firms that wish to pay an executive more than $1 million either have to forfeit deductions or structure the compensation package so that the excess over $1 million qualifies

under the performance based exception. While a variety of compensation forms can qualify as performance based, they vary in the difficulty of qualification, the risk qualification imposes on the executive, etc. For example, for amounts paid under a bonus plan to qualify as performance based, the payout must not exceed that determined using objective plan parameters set at the beginning of the year. In contrast stock option plans are relatively easy to qualify, and as long as the exercise price is set at or above the market price on the date of grant, are assumed to be performance based. The Financial Economists Roundtable (2003, 5) has called for the repeal of this provision. The group believes that this tax provision has spurred the increase of stock option compensation and argues that “this rule is a clumsy attempt to regulate the level and structure of executive compensation.” This paper extends the prior research on the effect of Internal Revenue Code section 162(m) on

executive compensation by focusing on the use of stock options. Specifically, we examine the proposition that section 162(m) has led to an increase in stock option compensation. Using the population of firms available on Standard & Poor’s ExecuComp database, we find evidence to support this. Our results show that the propensity to issue stock options has 3 Source: http://www.doksinet increased for affected executives, both in absolute terms and as a percentage of total compensation. In additional analysis, we find evidence that is consistent with this increase in stock option compensation substituting for lower increases in salary for affected executives. But, we find no evidence of stock option compensation substituting for annual cash bonuses. We interpret this as indicating that firms and their executives are acting in a way consistent with the incentives provided by section 162(m). This study continues in section 2 with a discussion of section 162(m), and a review of the

relevant literature in section 3. Section 4 develops our research question and model, while section 5 discusses our sample selection. Section 6 presents the empirical results The findings of the study are summarized in section 7. SECTION 162(m) Internal Revenue Code section 162(m) was a response to the concern about the perceived link between the international competitiveness of United States industry and the substantial salaries paid to United States executives (Brownstein and Panner 1992). Corporate governance critics (eg, Crystal 1992; McCarroll 1993) argued that executive compensation was excessive, both in comparison to that paid lower level employees and that paid overseas executives; and that executives were setting their own pay with no shareholder input. Congress believed that this provision (section 162(m)) would reduce excessive, non-performance based compensation (U.S Congress, House 1993). Section 162(m) places a $1 million cap on the annual deduction for non-performance

based compensation to the top five executives (the chief executive officer (hereafter CEO) and the next four highest compensated officers). Executive compensation generally consists of salary, fringe benefits, annual cash incentives, and long-term cash or stock-based incentives. The 4 Source: http://www.doksinet section 162(m) limit does not apply to (1) commissions, (2) non-taxable fringes and qualified retirement plan contributions, and (3) performance-based compensation. Prior to the imposition of section 162(m), most firms claimed to tie compensation to performance, however compensation committees had substantial discretion in awarding executive bonuses. Specific goals and performance criteria were rarely set in advance and even more rarely made public. Under section 162(m), to qualify bonus plans for the performance-based exception, firms are required to adopt a performance-based plan that is based on the executives attainment of one or more performance goals that were

established ex-ante by a compensation committee composed solely of independent directors. The performance goals must be based on objective formulae and the material terms of the plan must be disclosed to and approved by shareholders. The compensation committee, which has the discretion to award less, but not more than the objectively determined amount, must certify that the performance goals have been met before payment is made. Any compensation awarded by the committee based on discretionary assessments of performance that is in excess of the objectively determined amounts does not qualify. By definition, salary will not qualify as performance-based since it is not contingent on the attainment of any criteria. Thus, any salary amounts earned in excess of $1 million are not deductible unless payment is deferred until after the executives retirement or unless paid under a contract executed prior to February 17, 1993. Annual bonuses will qualify under the performancebased exception as

long as the firm adopts a bonus plan consistent with the section 162(m) requirements discussed above. Employee stock options qualify as performance-based under section 162(m) as long as the options have exercise prices equal to or greater than the market price at the time the award is made and the plan states the maximum number of shares that can be granted during a specified period. LITERATURE REVIEW 5 Source: http://www.doksinet There is a growing body of research that shows section 162(m) has had some impact on firms’ compensation practices, albeit perhaps not the intended impact. For example, Balsam and Ryan (1996) examined the propensity of firms to qualify their short-term bonus plans to meet the requirements of section 162(m), finding that many firms were sensitive to the potential tax and political costs of not qualifying. However, they showed that firms more likely to make the requisite modifications were those where compensation was most related to performance -- a

formalization of existing policy. Further, approximately half of the firms in their sample chose not to modify, and many of those that did, expressly reserved their right to pay nondeductible compensation. Reitenga et al (2002) observed that many firms elected not to qualify their compensation plans on the grounds that executive performance could not be evaluated using a fixed formula and that reserving the use of discretion in determining executive pay was in the best interest of the firm. Prior research (see e.g, Balsam, 2002; Perry and Zenner, 2001) found that all components of the compensation package increased after 1993, with the largest increase coming in the form of stock option grants. This finding that compensation increased post section 162(m) is consistent with the theoretical predictions of Halperin, et al. (2001) However while both show the increase post section 162(m) neither show that the increase in stock options is disproportionate to affected executives and firms.

Harris and Livingstone (2002) examined firms whose CEOs earned less than $1 million, “unaffected firms,” and found it had the perverse effect of raising the compensation of those CEOs. Harris and Livingstone conclude that section 162(m) set a target compensation amount and that firms whose CEOs earned less than that target, raised compensation. While research shows that section 162(m) has not led to a reduction in executive compensation, there is some limited (and mixed) evidence that compensation has become more responsive to firm performance. Examining the sensitivity of pay to performance, Johnson et al 6 Source: http://www.doksinet (2001), Perry and Zenner (2001), and Balsam and Ryan (2004) all found some evidence of an increased sensitivity of compensation to performance after 1993. While Johnson et al (2001) did not attribute this increased sensitivity to section 162(m), Perry and Zenner (2001) did, "especially for firms with million-dollar pay packages."

Similarly, Rose and Wolfram (2000, p 201) provided some evidence that the 162(m) limit "has led firms near the $1 million cap to restrain their salary increases and perhaps to increase the performance components of their pay packages." However in a later paper, Rose and Wolfram (2002, S138) concluded “There is little evidence that the policy significantly increased the performance sensitivity of chief executive officer (CEO) pay at affected firms. We conclude that corporate pay decisions have been relatively insulated from this policy intervention.” A more recent trend is for researchers to examine the details of firms’ responses to section 162(m). Balsam and Yin (2005) examine the actual tax status of executive compensation, finding that almost 40 percent of their sample firms forfeit some tax deductions because of section 162(m). Interestingly, they found that in 90 percent of the firm years in which a forfeiture occurred the firm had at least one plan that met the

requirements of section 162(m), and consistent with our research expectations, they had a qualified stock option plan in the vast majority of cases. RESEARCH QUESTION In firms where the CEO or other top officers are earning in excess of $1 million in annual compensation, the after-tax cost of performance based compensation such as bonuses and stock option grants is reduced relative to other forms of compensation. As discussed above, the firm must take a number of steps and put compensation at risk for an annual cash bonus to qualify for the performance based exception under section 162(m). In contrast, stock option plans can be easily qualified with no change in compensation risk. Option grants are performance-based 7 Source: http://www.doksinet compensation if the options have exercise prices equal to or greater than the market price at the time the award is made and the plan states the maximum number of shares that can be granted during a specified period. Most firms already met

these requirements when section 162(m) was imposed 1 Thus, unlike annual cash bonus plans, section 162(m) required minimal modifications to compensatory option plans. That being said, depending upon the firm, options may still be riskier than annual cash bonuses. However, as illustrated by Reitenga et al (2002), qualifying a bonus plan makes it riskier than it was before. In contrast, qualifying a compensatory option plan has little effect on its risk. Hence, section 162(m) increased the risk of annual cash bonuses relative to options. Consequently, the firm may find it desirable and easier to shift compensation into options if the executive is subject to 162(m) and earns more than $1 million a year. Model We use the following pooled, cross-sectional Tobit regression model to test the hypothesis that section 162(m) has lead to the increased use of stock options in the compensation packages of affected individuals. The dependent variable in our primary analysis, the ratio of stock

option to total compensation, is bounded by zero and one; consequently, we use Tobit. Our formal model is: PERCENTOPT it = α 0 + α 1 DUM1 it + α 2 DUM2 it + α 3-6 RANK it + α 7 VALUE it-1 + α 8 DIVYIELD it + α 9 SIZE it + α 10 TRS it + α 11 ROA it + α 12 VARROA it + α 13 RISK it + α 14 CONSTRAINT it + α 15 FCF it + α 16 BKM it + α∑YEAR + α∑IND + ε it (1) where the dependent variable is: PERCENTOPT it = the Black-Scholes value of options grants to executive i in year t divided by executive is total compensation, where both the Black-Scholes value and total compensation are provided by ExecuComp; 2 and the independent variables are: 8 Source: http://www.doksinet DUM1 it = an indicator variable taking the value of 1 if cash compensation of executive i is greater than $900,000 in year t, 0 otherwise; 3 DUM2 it = an indicator variable taking the value of 1 if cash compensation of executive i is greater than $900,000 in year t and year t is 1994 (1995 if non

December fiscal year end) or later, 0 otherwise; RANK it = a series of indicator variables for executive rank, where rank is based on the level of salary plus bonuses; 4 VALUE it-1 = value of executive i’s shares held plus the intrinsic value of exercisable and unexercisable options deflated by total direct compensation, all measured at the end of year t-1; DIVYIELD it = the dividend yield of executive is firm in year t; SIZE it = the log of assets of executive is firm in year t; TRS it = the return to shareholders of executive is firm in year t; = net income before extraordinary items and discontinued operations deflated ROA it by total assets for executive is firm for year t; 5 VARROA it = variance of net income before extraordinary items and discontinued operations, deflated by total assets for executive i’s firm for year t ; RISK it = the volatility measure (60 month) used by ExecuComp to calculate the Black-Scholes values for executive is firm in year t ; CONSTRAINT it =

indicator variable taking the value of 1 when retained earnings plus the value of cash dividends and stock repurchases in the current year divided by cash dividends and stock repurchases is less than two and 0 otherwise; FCF it = ratio of free cash flow to total assets measured as common and preferred dividends less cash flow from operating and investing activities deflated by total assets for executive i’s firm in year t; BKM it = the ratio of book value to market value for executive i’s firm in year t ; = a series of indicator variables for years, 1 in year t, 0 otherwise for years YEAR 1993 – 2002. 6 IND = a series of indicator variables for two digit SIC codes. Our test variable is DUM2. The coefficient on DUM2 represents the incremental effect of section 162(m) on the percentage of stock options in the compensation package of individuals who are affected by the requirements of section162(m). A positive coefficient on this variable would be consistent with section 162(m)

leading to an increase in stock option compensation for this group. Control Variables The previous literature has shown that executive compensation is related to both executive and firm related factors. Consequently, we include the following control variables in our model: DUM1, RANK, VALUE, DIVYIELD, SIZE, TRS, ROA, VARROA, RISK, 9 Source: http://www.doksinet CONSTRAINT, FCF, BKM, YEAR, and IND. Executive related controls We include DUM1 because, independent of section 162(m), the composition of the compensation package may be more heavily weighted towards options for more highly paid individuals. Consequently we expect a positive coefficient on this variable We include a series of indicator variables for executive rank, RANK, because the composition of an executive’s compensation package has been shown to vary with rank (Balsam, 2002, table 2.11) We also include VALUE as a proxy for the preexisting holdings of managers because there is an optimal level of equity holdings and

compensation can be used to adjust for deviation from that optimum (Core and Guay, 1999). The measure that we use is the value of the shares held plus the intrinsic value of both unexercisable and exercisable options deflated by total compensation. We expect a negative coefficient for this variable. Firm related controls We include DIVYIELD as the value of a firm’s stock options is less, all else equal, the higher the dividend yield. Thus, managers in firms with high dividend yields are less likely to prefer stock option compensation (Lambert et al. 1989) We expect a negative coefficient for this variable. We include SIZE, measured as the log of assets, because prior research has shown that the portion of options in an executive compensation package increases with firm size (Balsam, 2002, table 2.6) Thus, we expect a positive coefficient on SIZE We include TRS and ROA because performance may affect the composition of the compensation package. However, the direction of the effect is

not clear While Murphy (1985) finds a negative association between stock option compensation and firm performance, Liang and Weisbenner (2001) find a positive association between stock option compensation and stock price. Consequently we do not predict the direction of the association between stock option 10 Source: http://www.doksinet compensation and firm performance. We include variables to proxy for the relative risk of compensation tied to market and accounting measures. RISK, measured as the 60 month volatility measure used by ExecuComp in calculating the Black-Scholes values, controls for market related risk. We are ambiguous on its effect on the compensation package. That is, while RISK increases the value of an option under the Black-Scholes model, implying a positive coefficient, it also may make the option less desirable to an under-diversified executive. For example, Meulbroek (2001) estimates that for Internet firms, the estimated value of stock options to undiversified

managers is only 53% of their cost to the firm. However, a recent paper (Hodges et al, 2005) shows that managers overvalue options relative to the Black-Scholes model. Which effect predominates is an empirical question. We include VARROA, the variance of ROA as the proxy for the risk associated with accounting measures of performance. We expect VARROA to be positively related to our dependent variable because the greater the volatility of a firm’s income, the greater the risk of compensation tied to accounting measures of performance 7. Prior research (Yermack, 1995; Dechow et al., 1996; Core and Guay; 1999; Carter et al, 2004) shows that firms with less free cash flow are more likely to use equity instead of cash compensation since equity compensation requires no cash payment. Following Core and Guay (1999) and Carter et al. (2004), we include free cash flow, FCF, as a control variable, constructed such that a larger value represents less free cash flow. Consequently, we expect a

positive coefficient on this variable. We also include a proxy for a firm’s investment opportunity set, BKM because firms with greater investment opportunities may be more likely to conserve cash and use stock option compensation instead. (Core and Guay, 1999; Carter et al, 2004) We measure this as the ratio of firm book value to the market value of its equity. We expect a negative coefficient on this since a greater value indicates a lesser opportunity set. 11 Source: http://www.doksinet Core and Guay (1999, 160) also argue that “firms that are constrained with respect to earnings will grant more stock options” because cash compensation is expensed while stock option compensation has, until now, only been required to be disclosed in footnotes to the financial statements. Consequently, consistent with Core and Guay, we include the control variable, CONSTRAINT. We expect a positive coefficient on this variable To account for any macro-economic year-to-year or industry wide

effects, we include indicator variables for each year and industry (2 digit SIC codes) in the sample. Tables 1 & 2 provide the sample distribution by year and industry. Insert Tables 1 & 2 here SAMPLE SELECTION The source for our sample is Standard & Poor’s ExecuComp, which includes the firms in the S&P 500, Mid-Cap, and Small-Cap indexes. We augment the data available on ExecuComp with financial data from Standard & Poor’s Compustat and we hand collected data on bonus plan qualification from corporate proxy statements. As our sample firms are the largest publicly held US corporations, they are the ones most likely to be affected by section 162(m). At the time we conducted our analysis, ExecuComp had compensation data on 125,122 executives over the period 1992 to 2002. Our test sample is reduced to 59,698 observations due to missing data In particular, we lose almost 40,000 observations due to the lagged data required for the VALUE variable 8. However, as we

show in the right hand column of table 4, rerunning the analysis without the VALUE variable and hence on the larger data set does not affect our results. Table 3 provides some descriptive statistics about our sample. In roughly 20 percent (the mean of DUM1 is .20) of executive year observations the individual earned more than $900,000 in cash compensation, making him/her affected (according to our definition) by section 162(m). Options were a significant part of the compensation package (PERCENTOPT), as they comprised a 12 Source: http://www.doksinet mean (median) 53 (28) percent of total compensation and 117 (48) percent of cash compensation, with the mean (median) grant valued at $900,050 ($223,850). The mean (median) dividend yield is 1.36 percent (054 percent) and the mean (median) one-year return to shareholders is 1803 (844) percent. The mean (median) income before extraordinary items and discontinued operations deflated by total assets (ROA) is 4 (5) percent, while the mean

(median) change in ROA (income before extraordinary items and discontinued operations deflated by total assets) was 0 (1) percent. The mean (median) variance of ROA (VARROA) is 1 (0) percent. Sixteen percent of the firm year observations in the sample had a loss in the current year, 37 percent had income lower in the current year than in the prior year, and 62 percent of the sample is earnings constrained (the mean of CONSTRAINT). 9 The mean (median) ratio of free cash flow to total assets is 1 (0) percent and the mean (median) value book to market value is 51 (43) percent. Insert Table 3 here EMPIRICAL RESULTS Table 4 presents the results of the Tobit regression analysis for model (1), as well as a model which excludes VALUE and hence allows us to incorporate 1992 into our analysis. 10, 11 In both regressions the coefficient on DUM2 is, as predicted, positive and significant. 12 This provides support for the hypothesis that there was a positive incremental effect of section 162(m) on

the amount of stock options in the compensation package of affected individuals. The coefficient on the indicator variable, DUM1 is significant, but surprisingly negative. The sign of this coefficient may be driven by our way of defining highly paid executives, which is based upon cash compensation – ceteris paribus the higher cash compensation the lower stock-based compensation. When we redefine highly paid executives based upon total compensation the coefficient on DUM1 becomes positive and significant, while that on DUM2 remains positive and significant. The results for the RANK indicator variables are as expected Higher ranked 13 Source: http://www.doksinet executives receive a higher proportion of their compensation in the form of options; thus we find both RANK1 and RANK2 to be significant and positive, RANK3 to be insignificant and RANK4 to be significant, but negative. The control variable, VALUE is negative as expected, but only marginally significant (p<0.11) All of

the firm related control variables are significant. The coefficients on DIVYIELD and SIZE are negative (p<0.0001) and positive (p<00001) respectively, consistent with the proportion of stock in the compensation package being inversely related to dividend yield and positively associated with firm size. The coefficients on the performance measures are mixed, as the coefficient on ROA is positive and significant (p<0.0001), while the coefficient on TRS is negative and significant (p<0.0001) The coefficient on RISK is positive and significant (p<0.0001) consistent with the positive effect of risk on the value of the option being associated with an increase in the proportion of stock in the compensation package. The other risk related measure, the variability of income, VARROA is also positive and significant (p<0.0001), consistent with an increased use of options when accounting based bonuses are more risky. Consistent with the findings of prior research, FCF and

CONSTRAINT are positive and significant and BKM is significant and negative (p<0.0001) These results indicate that firms are more likely to use stock options to compensate managers when they have less free cash flow, are constrained with respect to earnings, or have greater investment opportunities. The year and industry indicator variables provide a control for industry wide and macro economic effects. While we omit the coefficients on these indicator variables for brevity, we discuss them here. The coefficients associated with the years 1994 and 1995 are insignificant, for the years 1996 through 2001 are significant and positive and for the year 2002 is significant and negative. With a lag, these coefficients seem to track overall market stock price movements That is, while stock returns were flat in 1994, beginning in 1995 stock prices increased through 14 Source: http://www.doksinet the beginning of 2000, at which point a bear market began. The results are consistent with

overall market performance affecting the desirability of stock options to executives and their use by corporations. While most of the coefficients on the industry controls are significant and negative; consistent with expectations, those associated with high technology industries are significant and positive. Insert Table 4 here Sensitivity Analysis Murphy (1998) notes that about 40 percent of firms grant a fixed number of options each year, while another 40 percent of firms grant options with fixed value each year. In the former situation there would be a mechanical relation between the Black-Scholes value of option grants and share price. Consequently in a rising market, independent of any other incentives we would observe, on average, an increase in BLK VALU and PERCENTOPT. For that reason we rerun our analysis in Table 5 using three alternative dependent variables; PERCENTOPT2, BLK VALU, and SOPTGRT. PERCENTOPT2 is the Black-Scholes value of options granted divided by the

executive’s total cash compensation, while BLK VALU is simply the BlackScholes value of the options granted, and SOPTGRT is the total number of options granted to the executive. While the first two alternative measures are subject to the same mechanical relation between option value and share price, the last is unaffected by it. That is, any changes observed in SOPTGRT are the result of a conscious decision by the compensation committee to increase or decrease the number of options granted. The results for all three alternative dependent variables are consistent with our original model. In all of the analyses, the coefficient on DUM2 is positive and significant (p<0.0001) Except for the variable, VALUE, the control variables remain significant; although some signs change. For instance, the coefficient on the DUM1 15 Source: http://www.doksinet variable is positive and significant, in the regressions with BLK VALU and SOPTGRNT as dependent variables. Insert Table 5 here We also

examine the effect of some alternative performance measures, which we report in table 6. For accounting based performance measures, we replace ROA first with LESS (an indicator variable equal to one if net income before extraordinary items and discontinued operations is less than in the prior year) and then with LOSS (an indicator variable taking the value of one if net income before extraordinary items and discontinued operations was less than zero). For stock based performance measures, we replace total return to shareholders over one year with total returns to shareholders over three (TRS3YR) and five (TRS5YR) year periods. In all permutations, the coefficient on DUM2 is positive and significant (p<0.0001) The major difference when we vary our accounting performance variable is the coefficient on that variable. That is, in the base model, the coefficient on ROA is positive and significant, as it is when we use LESS as our performance measure. But when we use LOSS, the coefficient

becomes negative, but not significant. The major difference when we vary our market variable is that, while in the one year window, the coefficient on TRS is negative and significant, in the longer windows, TRS3YR and TRS5YR, it is positive and significant. Insert Table 6 here Additional Analysis Our tests of model (1) show that affected executives are receiving a greater portion of their compensation in the form of stock options in the post 162(m) period. Our finding of increased stock option compensation post-section 162(m) may have occurred for two reasons. The first possibility is that stock options increased because section 162(m) gave it additional imprimatur and consequently compensation committees simply added more options to 16 Source: http://www.doksinet compensation packages without any offsetting reduction elsewhere in the package. In fact, this theory is consistent with the pattern observed by Balsam (2002, table 2.7), whereby stock option grants increased over time,

but so did the other components of the compensation package. The other alternative is that the increase in stock options was offset by reductions, or if not reductions, lesser increases in the other components of the compensation package than would have been observed in the absence of section 162(m). In effect, did firms substitute stock option compensation for other forms of compensation in the pay packages of affected executives? Risk and taxes provide opposing incentives to substitute options for bonuses. If the plan is non qualified then there is no change in the risk of the bonus, hence there is no reason from a risk perspective to shift from bonuses to options. However if the plan is non qualified the firm may shift compensation from bonuses to options to preserve deductions. Alternatively if the plan is qualified, then the risk of the bonus has increased relative to the risk of the options (which may still be riskier). Consequently for risk reasons we may observe a shift from

bonuses to options However since both are deductible there is no tax reason to expect a shift. To test which of the two alternatives are more likely, we run the following modified version of model (1) focusing on changes in compensation from the pre to post section 162(m) period. ∆BSV it = α 0 + α 1 ∆SAL i + α 2 ∆BONUS i + α 3 DUM2 i + α 4 DUM3 i + α 5 ∆SAL*DUM2 i + α 6 ∆BONUS*DUM2 i + α 7 ∆BONUSDUM3 i + α 8-11 RANK i + α 12 ∆RANK i + α 13 ∆DIVYIELD i + α 14 ∆SIZE i + α 15 ∆TRS i + α 16 ∆ROA i + α 17 ∆VARROA i + α 18 ∆RISK i + α 19 ∆CONSTRAINT i + α 20 ∆FCF i + α 21 ∆BKM i + α 22 ∆VALUE it-1 + α∑IND + ε it (2) One difference between models (1) and (2) is that while model (1) is a levels regression, model (2) is a changes regression. 13 Further since the change can be either positive or negative, we use ordinary least squares rather than the Tobit model used above. Instead of using the percentage of the compensation package

comprised of stock options as the dependent variable 17 Source: http://www.doksinet we use ∆BSV, the percentage change in the Black-Scholes value of the option grants as the dependent variable. This change allows us to include as independent variables ∆SAL, the percentage change in salary, i.e, (salary-lag(salary))/lag(salary) and ∆BONUS, the percentage change in bonus and their interactions with DUM2 as independent variables. That is, we could not use the change in the percentage of compensation package comprised of stock options as the dependent variable and the corresponding changes in percentage of compensation package comprised of salary and bonus as independent variables because, by definition, there would have to be a negative relation between them. The coefficient on DUM2 indicates the incremental change in the Black-Scholes value of the options granted to affected executives independent of any changes in the salary and bonus components of compensation, while the

coefficients on ∆SAL and ∆BONUS will indicate the unconditional association of the change in salary and bonus on the change in the Black-Scholes value of the options granted to executives – which we would normally expect to be positive. We collect data on whether our sample firms qualified their annual bonus plans 14 to include two additional variables – an additional indicator variable taking the value of one if the executive is affected and if the firm qualified its short term bonus plan (DUM3), and an interaction variable, ∆BONUS*DUM3. The coefficients of primary interest are those on the interactions between DUM2 and ∆SAL, DUM2 and ∆BONUS, and DUM3 and ∆BONUS, which are the incremental effects of the change in salary and change in bonus on change in option compensation for affected executives. We conduct our analysis using the change between the last year pre-section 162(m), and the first year post-section 162(m). For December fiscal year end companies, the last

year prior to (first year after) section 162(m) would be 1993 (1994), while for non-December fiscal year end companies, the last year prior to (first year after) section 162(m) would be 1994 (1995). 18 Source: http://www.doksinet The results of this analysis are presented in table 7. We first run the model without the qualification variable to serve as a baseline. We find positive coefficients on ∆SAL and ∆BONUS, indicating that the change in option compensation is positively associated with the change in salary and bonus compensation. Interestingly while the coefficient on the change in bonus (09531) is significantly greater than zero, it is insignificantly different from one. This indicates a one dollar increase in bonus is associated with a one dollar increase in option compensation. In contrast the coefficient on the change in salary (73.0343) is significantly greater than one, indicating a one dollar increase in salary is associated with a 73 dollar increase in option

compensation. We then find that the coefficient on DUM2 is positive and significant, indicating an increase in option compensation independent of any other changes in the compensation package for affected executives. The coefficient on ∆SAL*DUM2 (12.1554) is significantly greater than zero, while that on ∆BONUS*DUM2 (-0.883) is significantly less than zero The former indicates the multiplier on a dollar increase in salary increases from 73.0343 for unaffected executives to 851897 for affected executives, and is consistent with firms substituting larger increases in stock options for increases in salary for affected executives. The latter indicates the coefficient on bonus for affected executives, the sum of 0.9531 and -08883, is insignificantly different from zero The results of the regression including the qualification variable (DUM3) are in the right most columns of table 7. As with the first regression in table 7, we find positive coefficients on ∆SAL, ∆BONUS, DUM2, and

∆SAL* DUM2, and a negative coefficient on ∆BONUS DUM2, with the magnitudes of the coefficients almost identical to those in the earlier model. Of the additional variables, the coefficient on DUM3 is positive, but insignificant, while the interaction, ∆BONUS* DUM3, is positive and significant. The latter, when summed with the coefficients on the level variable, ∆BONUS, and the interaction, ∆BONUS*DUM2, is also positive and significant, indicating that for affected executives at firms which qualified their annual bonus plans, there is a 19 Source: http://www.doksinet positive and significant association between the change in bonus and change in option compensation. Focusing on the association between change in option compensation and change in bonus, we observe that it is positive and significant, and the coefficient is insignificantly different from one for unaffected executives and affected executives in firms with qualified annual bonus plans. This is consistent with both

bonuses and option grants varying with performance, i.e, assuming that bonuses are based upon performance, which seems reasonable for at least the firms with qualified bonus plans. In contrast, when the firm has not qualified its bonus plan, the association between the change in option compensation and the change in bonus is insignificantly different from zero. That is, for affected executives changes in options appear to be independent of changes in bonuses. We do not see any evidence that firms substitute options for bonuses. SUMMARY This paper extends the prior research on the effect of Internal Revenue Code section 162(m) on executive compensation by focusing on the use of stock options. Under the requirements of section 162(m), firms must put the annual cash bonus compensation at risk for the bonus to qualify as deductible. This is a significant change from most firms’ practices in previous years In contrast, section 162(m) required little or no change in compensatory stock

option plans because such plans generally met the definition of ‘performance-based’ under section 162(m). As a consequence, section 162(m) increased the risk of annual cash bonuses relative to option compensation. In this paper, we posit that this may provide an incentive for firms and their executives to shift compensation to stock options. Our results provide evidence that section 162(m) has led to increases in the use of stock options by affected firms, presumably to maximize their deductible compensation. In addition, we find evidence of a substitution effect for salary increases for affected executives, but no evidence for annual cash bonuses. 20 Source: http://www.doksinet Our analysis controls for a variety of other factors that may affect executive stock option grants, including the dividend yield, firm size, firm performance (both accounting and market), the riskiness of the firm’s shares and volatility of its accounting income as well as the firm’s cash constraints

and its investment opportunities. We also include controls for existing executive equity ownership, executive rank, year and industry. We interpret our results to mean that firms and their executives are acting in a way consistent with the incentives provided by section 162(m). References Balsam, S., 2002 An Introduction to Executive Compensation, Academic Press, San Diego: California. Balsam, S., and D Ryan 1996 Response to tax law changes involving the deductibility of executive compensation: A model explaining behavior. Journal of the American Taxation Association 18: 1-12. Balsam, S., and D Ryan 2004 Social engineering and the Internal Revenue Code: The case of CEOs hired after the imposition of section 162(m), the million-dollar cap on executive compensation. Working paper, Temple University Balsam, S., and J Yin 2005 Explaining Firm Willingness to Forfeit Tax Deduction Under Internal Revenue Code Section 162(m): The Million-dollar Cap, forthcoming Journal of Accounting and

Public Policy. Brownstein, A., and M Panner 1992 Who should set CEO pay? The press? Congress? Shareholders? The Harvard Business Review (May-June): 28-38. Carter, M.E, LJ Lynch, and I Tuna 2004 The role of incentives and accounting in the design of executive compensation packages. Working Paper, The University of Pennsylvania Core, J and W. Guay 1999 The use of equity grants to manage optimal equity incentive levels Journal of Accounting and Economics 28: 151-184. Crystal, G. 1992 In Search of Excess: The Overcompensation of the American Executive New York, NY: W.W Norton Dechow, P., A Hutton, and R Sloan 1996 The economic consequences of accounting for stockbased compensation Journal of Accounting Research 34: 1-20: Financial Economists Roundtable. 2003 Statement on the controversy over executive compensation. Website url: http://wwwlucedu/orgs/finroundtable 21 Source: http://www.doksinet Halperin, R., Y Kwon, and S Rhoades-Catanach 2001 The impact of deductibility limits on

compensation contracts: A theoretical examination. Journal of American Taxation Association 23 (supplement): 52-65. Harris, D., and J Livingstone 2002 Federal tax legislation as a political cost benchmark The Accounting Review 77 (October): 997-1018. Hodges, F., S Rajgopal, and T Shevlin 2005 How do managers value stock options and restricted stock. Working Paper, University of Washington Johnson, M., S Porter and M Shackell 2001 Stakeholder pressure and the structure of executive compensation. Working Paper, University of Michigan Lambert, R. A, W N Lanen, and D F Larcker, 1989, Executive stock option plans and corporate dividend policy, Journal of Financial and Quantitative Analysis 24: 409-425. Liang, N. and S J Weisbenner 2001 Who Benefits from a Bull Market? An Analysis of Employee Stock Option Grants and Stock Prices. FEDS Working Paper No 2001-57 Matsunaga, S. R 1995 The effects of financial reporting costs on the use of employee stock options, The Accounting Review 70(1): 1-26.

McCarroll, T. 1993 Rolling back executive pay Time (March 1): 49-50 Meulbroek, Lisa K. 2001 The efficiency of equity-linked compensation: Understanding the full cost of awarding executive stock options, Financial Management 30(2): 5-44. Murphy, K. J 1985 Corporate Performance and Management Remuneration: An Empirical Analysis, Journal of Accounting and Economics 7: 11-42. Murphy, K. J 1998 Executive compensation, Working Paper, University of Southern California Perry, T., and M Zenner 2001 Pay for performance? Government regulation and the structure of compensation contracts. Journal of Financial Economics 62(3): 453-488 Reitenga, A., S Buchheit, J Yin, and T Baker 2002 CEO bonus pay, tax policy, and earnings management. Journal of American Accounting Association (supplement), 1-23 Rose, N., and C Wolfram 2000 Has the “million-dollar cap” affected CEO pay? The American Economic Review 90(2): 197-202. Rose, N., and C Wolfram 2002 Regulating executive pay: Using the tax code to

influence chief executive compensation. Journal of Labor Economics 20(2): S138-175 U.S Congress, House 1993 Fiscal year budget reconciliation: Recommendations of the Committee on Ways and Means. (May 18) Washington, DC: US Government Printing Office 22 Source: http://www.doksinet Yermack, D. 1995 Do corporations award CEO stock options effectively? Journal of Financial Economics 39: 237-269. 23 Source: http://www.doksinet TABLE 1 Industry Distribution Two-digit SIC Code Number of Observations 1 7 10 12 13 14 15 16 17 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 44 45 47 48 49 50 51 52 188 18 442 15 2,000 140 365 215 74 1,609 85 516 586 470 399 1,174 1,268 4,633 624 613 199 396 1,509 959 3,775 4,289 1,931 2,707 450 283 29 425 241 579 191 1,397 4,608 1,241 733 289 24 Percentage of Observations 0.30 0.00 0.70 0.00 3.40 0.20 0.60 0.40 0.10 2.70 0.10 0.90 1.00 0.80 0.70 2.00 2.10 7.80 1.00 1.00 0.30 0.70 2.50 1.60 6.30 7.20 3.20 4.50 0.80 0.50 0.00 0.70

0.40 1.00 0.30 2.30 7.70 2.10 1.20 0.50 Source: http://www.doksinet 53 54 55 56 57 58 59 60 61 62 63 64 67 70 72 73 75 78 79 80 82 83 87 99 856 493 256 960 472 1,075 1,171 58 557 882 2,501 386 433 202 308 4,643 119 188 447 973 155 14 612 272 25 1.40 0.80 0.40 1.60 0.80 1.80 2.00 0.10 0.90 1.50 4.20 0.60 0.70 0.30 0.50 7.80 0.20 0.30 0.70 1.60 0.30 0.00 1.00 0.50 Source: http://www.doksinet TABLE 2 Year Distribution Fiscal Year Ending Number of Observations 1993 1994 (Pre December) 1994 (December) 1995 1996 1997 1998 1999 2000 2001 2002 4,526 2,280 3,966 6,253 6,339 6,562 6,170 6,301 6,090 5,763 5,448 59,698 26 Percentage of Observations 7.60 3.82 6.64 10.50 10.60 11.00 10.30 10.60 10.20 9.70 9.10 Source: http://www.doksinet TABLE 3 Descriptive Statistics Variables PERCENTOPT PERCENTOPT2 BSVAL NUMGRT DUM1 DUM2 DIVYIELD SIZE TRS TRS3YR TRS5YR ROA PCROA LOSS LESS VARROA RISK CONSTRAINT FCF BKM VALUE Observations 59698 59619 59698 59698 59698 59698 59698 59698 59698

59698 59698 59698 59698 59698 59666 59698 59698 59698 59698 59698 59698 Mean 0.53 1.17 900.05 74.87 0.20 0.19 1.36 7.15 18.03 9.16 7.41 0.04 0.00 0.16 0.37 0.01 0.40 0.62 0.01 0.51 584.07 Std Dev 1.09 1.94 1,946.03 135.52 0.40 0.39 2.21 1.58 69.33 27.56 21.37 0.09 0.08 0.36 0.48 0.05 0.18 0.48 0.10 0.61 87,217.67 1st Quartile 0.05 0.06 19.62 5.00 0.00 0.00 0.00 5.97 -15.83 -5.72 -5.42 0.02 -0.01 0.00 0.00 0.00 0.27 0.00 -0.05 0.26 0.83 Median 0.28 0.48 223.85 25.00 0.00 0.00 0.54 7.01 8.44 9.39 9.21 0.05 0.01 0.00 0.00 0.00 0.36 1.00 0.00 0.43 2.25 3rd Quartile 0.51 1.28 784.62 75.00 0.00 0.00 2.09 8.22 35.48 24.56 20.46 0.08 0.03 0.00 1.00 0.00 0.50 1.00 0.06 0.66 6.11 Where PERCENTOPT it = the Black-Scholes value of option grants to executive i in year t divided by executive is total compensation, where both the Black-Scholes value and total compensation are provided by ExecuComp; PERCENTOPT2 it = is the Black-Scholes value of option grants to executive i in year t divided by

executive i’s total cash compensation, where both the BlackScholes value and total cash compensation are provided by ExecuComp; BSVAL it = is the Black-Scholes value of the options granted to executive i in year t as provided by ExecuComp; NUMGRT it =the total number of options granted to executive i in year t; DUM1 it =is an indicator variable taking the value of 1 if cash compensation of executive i is greater than $900,000 in year t, 0 otherwise; DUM2 it =is an indicator variable taking the value of 1 if cash compensation of executive i is greater than $900,000 in year t and year t is 1994 or later, 0 otherwise; DIVYIELD it = the dividend yield of executive is firm in year t; SIZE it = the log of assets of executive is firm in year t; TRS it = the return to shareholders of executive is firm in year t; 27 Source: http://www.doksinet TRS3YR it = the return to shareholders of executive is firm for three years ending with year t; TRS5YR it = the return to shareholders of executive

is firm for three years ending with year t; ROA it = net income before extraordinary items and discontinued operations deflated by total assets for executive is firm for year t; PCROA it = change in net income before extraordinary items and discontinued operations deflated by total assets for executive is firm in year t; = indicator variables taking the value of one if net income before LESS it extraordinary items and discontinued operations was less than prior year, and zero otherwise for executive is firm in year t; LOSS it = indicator variables taking the value of one if net income before extraordinary items and discontinued operations was less than zero, and zero otherwise for executive is firm in year t; VARROA it = variance of NIBEX using all available observations for company i. Minimum number of observations is 6. RISK it =the volatility measure (60 month) used by ExecuComp to calculate the Black-Scholes values for executive is firm in year t ; CONSTRAINT it = indicator

variable taking the value of 1 when retained earnings plus the value of cash dividends and stock repurchases in the current year divided by cash dividends and stock repurchases in prior year is less than two and 0 otherwise; = ratio of free cash flow to total assets; FCF it = book to market value of equity; and BKM it VALUE it = value of shares owned, plus intrinsic value of options held, deflated by total direct compensation – all measured at end of previous year. 28 Source: http://www.doksinet TABLE 4 Tobit Regression Results PERCENTOPT it = α 0 + α 1 DUM1 it + α 2 DUM2 it + α 3 DIVYIELD it + α 4 SIZE it + α 5 TRS it + α 6 ROA it + α 7 VARROA it + α 8 RISK it + α 9 CONSTRAINT it + α 10 FCF it + α 11 BKM it + α 12 VALUE it-1 + α 13-16 RANK it + α∑YEAR + α∑IND + ε it Variable name Model w/VALUE COEFFICIENT (Chi-square) Model w/o VALUE COEFFICIENT (Chi-square) -2.3394 INTERCEPT -2.1014 *(545.19) * (370.22) 0.2805 DUM2 0.2698 *(43.54) * (24.66)

Executive related control variables DUM1 -0.3774 -0.2504 *(41.09) 0.5613 *(1410.55) 0.2154 *(181.93) 0.0097 (0.34) -0.0867 *( 24.56) * (47.02) 0.4339 RANK1 * (576.95) 0.1800 RANK2 * (80.61) -0.0197 (0.82) -0.0898 * (13.79) -0.0000 (2.49) RANK3 RANK4 VALUE Firm related control variables DIVYIELD -0.0343 -0.0342 *(226.32) 0.2536 *( 3792.76) -0.0003 *(31.00) 1.0604 *(293.55) 3.2913 *(5563.75) 1.0082 *(103.51) 0.1969 *(266.66) 0.7376 *(242.77) -0.2570 *( 771.89) * (142.80) 0.3335 SIZE * (3665.69) -0.0004 TRS * (21.25) 1.7415 ROA * (419.81) 3.6876 RISK * (3951.78) 1.4143 VARROA * (115.85) 0.2349 CONSTRAINT * (224.03) 0.8592 FCF * (182.43) -0.3168 BKM * (624.22) YEAR IND NR NR * denotes significance at p value < .01 Variable definitions are provided at the bottom of table 3 We omit the coefficients for the year and industry dummies for brevity. 29 NR NR Source: http://www.doksinet TABLE 5 Tobit Regression Results for Alternative Dependent

Variables DEPEND it = α 0 + α 1 DUM1 it + α 2 DUM2 it + α 3 DIVYIELD it + α 4 SIZE it + α 5 TRS it + α 6 ROA it + α 7 VARROA it + α 8 RISK it + α 9 CONSTRAINT it + α 10 FCF it + α 11 BKM it + α 12 VALUE it-1 + α 13-16 RANK it + α∑YEAR + α∑IND + ε it Dependent variable Variable name PERCENTOPT2 BLK VALU SOPTGRNT COEFFICIENT (Chi-square) COEFFICIENT Chi-square) COEFFICIENT (Chi-square) -5.3058 -7529.7000 -396.3115 * (729.65) * (1359.18) * (810.10) INTERCEPT 0.4269 1833.5000 79.9039 * (21.51) * (363.12) * (135.21) DUM2 Executive related control variables -0.7031 DUM1 * (61.64) 0.8263 RANK1 * (787.72) 0.3385 RANK2 * (107.63) -0.0428 RANK3 (1.45) -0.1504 RANK4 * (14.55) -0.0000 VALUE (2.43) Firm related control variables -0.0632 DIVYIELD * (189.73) 0.6313 SIZE * (4969.29) -0.0005 TRS * (11.10) 2.8155 ROA * (423.01) 7.0420 RISK * (5388.19) 2.1850 VARROA * (104.63) 0.3581 CONSTRAINT * (195.02) 1.5787 FCF * (230.53) -0.5865 BKM * (832.23) YEAR IND NR

NR 317.3985 71.0117 * (11.52) * (125.33) 1184.0000 95.9944 * (1461.85) * (2093.52) 470.0726 32.6146 * (190.95) * (198.41) -122.5160 -20.4406 * (10.95) * (65.89) -116.3133 * (8.06) -0.0000 (0.06) -0.0000 (0.07) -56.8137 -1.2296 * (128.81) * (13.95) 824.8847 42.4815 * (7966.08) * (4656.33) 1.0602 -0.0346 * (46.73) * (10.23) 2953.6000 -48.6752 * (416.50) * (23.15) 5415.7000 338.5982 * (2976.40) * (2530.59) 2609.3000 325.73 * (135.12) * (437.10) 335.5600 23.1464 * (159.34) * (166.05) 557.1492 -29.7300 * (25.83) * (15.75) -425.1669 -19.1967 * (420.64) * (197.86) NR NR Notes: * denotes significance at p value < .01 Variable definitions are provided at the bottom of table 3 We omit the coefficients for the year and industry dummies for brevity. 30 -17.2007 * (37.88) NR NR Source: http://www.doksinet TABLE 6 Tobit Regression Results with Alternative Performance Measures The dependent variable is PERCENTOPT Variable name

COEFFICIENT (Chi-square) -2.2663 *(346.81) 0.2794 DUM2 *(24.26) Executive related control variables -0.3440 DUM1 *(38.74) 0.4276 RANK1 *(556.75) 0.1866 RANK2 *(85.91) -0.0207 RANK3 (0.89) -0.0810 RANK4 * (11.13) -0.0000 VALUE (1.91) Firm related control variables -0.0333 DIVYIELD * (141.40) 0.3144 SIZE * (3261.42) -0.00 TRS (0.10) INTERCEPT -2.2222 *(322.90) 0.2831 *(24.94) -2.4100 *(397.59) 0.2901 *(26.38) -2.3875 *(390.39) 0.2856 *(25.61) -0.3580 *(42.00) 0.4286 *(559.16) 0.1832 *(82.81) -0.0220 (1.01) -0.0837 * (11.89) -0.0000 (2.09) -0.3990 *(52.55) 0.4320 *(573.01) 0.1754 *(76.58) -0.0271 (1.54) -0.0964 * (15.91) -0.0000 (2.65) -0.3962 *(51.90) 0.4344 *(579.74) 0.1809 * (81.64) -0.0281 (1.66) -0.0967 * (16.02) -0.0000 (2.44) -0.0327 * (133.24) 0.3182 * (3346.5) -0.0002 (3.51) -0.0320 * (118.30) 0.3304 * (3595.5) -0.0311 * (111.69) 0.3254 * (3448.9) 0.0028 * (134.21) TRS 3 yr 1.4487 * (279.70) 0.0041 * (174.21) 1.4821 * (300.35) 3.3662 * (3471.70) 0.8644 * (45.76)

0.2454 * (243.14) 0.4953 * (64.28) -0.3840 * (987.80) -0.0248 (1.58) 3.4184 * (3378.6) 0.9039 * (49.37) 0.2422 * (236.82) 0.5241 * (71.79) -0.3729 * (896.27) 3.6366 * (3892.5) 1.4714 * (123.91) 0.2449 * (242.58) 0.8306 * (170.64) -0.2712 * (441.05) 3.6571 * (3956.5) 1.4960 * (128.68) 0.2623 * (274.15) 0.8568 * (182.48) -0.2767 * (472.19) NR NR NR NR NR NR NR NR TRS 5 yr ROA LESS 0.0889 * (41.62) LOSS RISK VARROA CONSTRAINT FCF BKM YEAR IND * denotes significance at p < 0.01 Variable definitions are provided at the bottom of table 3. We omit the coefficients for the year and industry dummies for brevity. 31 Source: http://www.doksinet TABLE 7 OLS Regression Results for Change Model The dependent variable is the change in the Black Scholes value of options ∆BSV it = α 0 + α 1 ∆SAL i + α 2 ∆BONUS i + α 3 DUM2 i + α 4 DUM3 i + α 5 ∆SAL*DUM2 i + α 6 ∆BONUSDUM2 i + α 7 ∆BONUS*DUM3 i + α 8-11 RANK i + α 12 ∆RANK i + α 13 ∆DIVYIELD i + α 14 ∆SIZE

i + α 15 ∆TRS i + α 16 ∆ROA i + α 17 ∆VARROA i + α 18 ∆RISK i + α 19 ∆CONSTRAINT i + α 20 ∆FCF i + α 21 ∆BKM i + α 22 ∆VALUE it-1 + α∑IND + ε it (2) Variable INTERCEPT ∆SAL ∆BONUS DUM2 DUM3 ∆SAL*DUM2 ∆BONUS*DUM2 ∆BONUS*DUM3 RANK1 RANK2 RANK3 RANK4 ∆RANK ∆DIVYIELD ∆SIZE ∆TRS ∆ROA ∆RISK ∆VARROA ∆CONSTRAINT ∆FCF ∆BKM IND Coeff (t-stat) -12.1822 73.0343 0.9531 23.2195 -0.24 *61.40 *9.27 *2.96 12.1554 -0.8883 *4.16 *-4.20 19.0852 8.0037 -2.1867 5.1479 -1.3885 -11.1006 59.0416 -18.2960 1.6451 0.0000 -0.2094 -57.1407 0.0579 12.3386 *2.17 0.79 -0.20 0.44 -0.41 *-5.15 *6.88 -1.01 1.49 *4.12 -0.80 -1.11 0.20 1.55 NR Coeff -12.6011 73.0450 0.9524 20.9733 12.9285 11.5066 -0.9394 0.7277 18.8834 7.8686 -2.0851 5.1350 -1.3927 -11.0634 58.8093 -17.9035 1.6486 0.0000 -0.2149 -55.0935 0.0491 12.0442 (t-stat) -0.25 *61.41 *9.27 *2.51 0.74 *3.92 *-4.40 1.96 *2.15 0.77 -0.19 0.44 -0.41 *-5.14 *6.85 -0.99 1.49 *4.16 -0.82 -1.07 0.17 1.51

NR Notes: * denotes significance at p < 0.01 Where DUM3 is an indicator variable that takes the value of 1 if the executive is defined as affected (i.e, DUM2=1) and the firm has qualified its short-term bonus plan and all other variables are defined at the bottom of table 3. The symbol ∆ in a variable name denotes a change in the value of the variable from the last year pre-section 162(m) to the first year post-section 162(m). For December fiscal year end companies, the last year prior to (first year after) section 162(m) would be 1993 (1994), while for non-December fiscal year end companies, the last year prior to (first year after) section 162(m) would be 1994 (1995). We omit the coefficients for the industry dummies for brevity. 32 Source: http://www.doksinet Notes 1 Both Matsunaga (1995, note 6) and Murphy (1998) find about 95 percent of corporations granting options with an exercise price equal to grant-date fair market value . 2 As noted in the sensitivity analysis,

we alternatively define the dependent variable as Black-Scholes value of options granted divided by total cash compensation, as the un-deflated Black-Scholes value of options granted, and as the number of options granted, with no change in results. 3 We use cash compensation rather than total compensation to define affected executives following the previous literature, e.g, Perry and Zenner (2001), Balsam and Ryan (2004), Balsam and Yin (2005) We use a cutoff of $900,000 to avoid missing firms that reduced compensation because of section 162(m). Our results are not affected by the use of other cutoffs, i.e, $950,000 or $1,000,000 In addition, our results are qualitatively the same when we also use salary or total direct compensation to define affected executives. Consequently while our choice of affected executives could potentially bias our results, the fact that the results are robust to alternative cutoffs and ways of measuring affected executives provides reassurance that this is

not the case. 4 Rank 5 is incorporated into the intercept. The results are the same if we use rank 1 as the intercept 5 In place of ROA, we tried other performance measures, indicator variables indicating if net income was less than the prior year or the existence of a loss. As shown in the sensitivity analysis, our conclusion is unchanged. 6 The year 1993 is incorporated into the intercept. The results are the same if we use 2002 as the intercept 7 While the bonus has to be based on objective performance measures, there is no requirement that these measures be accounting based. 8 We need information on each executive’s prior year stock and option holdings to compute the variable, VALUE. Consequently, we lose observations in those instances where prior year information on the executive’s holdings is not available because the executive was not a listed officer in the company in the prior year. 9 The difference between the 62 percent we report and the 44 percent reported in

Core and Guay (1999) is driven by our decision to classify those firms that do not pay dividends or repurchase shares as constrained. If we reclassify those firms as unconstrained the percentage of firms classified as constrained drops to 47 percent. Most importantly, reclassifying those firms does not affect our results 10 Recall that in our model, this period’s equity compensation is based in part, on the equity and option holdings at the end of the period, i.e, VALUE is lagged, causing us to lose the earliest year for which data is available. 11 One advantage of incorporating 1992 into the analysis is that the percentage of the observations in the pre section 162(m) period increases from 11.42 percent to 1890 percent of the sample 12 The coefficient on VALUE is insignificant and does not affect the significance of the coefficients on any of the other variables in the model. 33 Source: http://www.doksinet 13 We include the current level of executive rank as an independent

variable as the rate of compensation change can differ across executive ranks, and we include the change in executive rank as an additional independent variable as a promotion (or demotion) will affect the rate of compensation change. 14 We also reran model one including an additional indicator variable for whether on not the firm qualified its annual bonus plan. The indicator variable was insignificant and our variable of interest, DUM2 remained positive and significant. 34