CEO Pay and Accounting Performance Measures_

Document Sample
CEO Pay and Accounting Performance Measures_ Powered By Docstoc
					CEO Pay and Accounting Performance Measures: The Role of Earnings Management

Gene Imhoff University of Michigan Business School 701 Tappan Ann Arbor, Michigan 48109-1234 Email: Imhoff@umich.edu

November 2003

Please do not quote, copy or electronically transmit without permission

Keywords:

Executive compensation; accounting performance; earnings management; corporate governance.

I would like to thank the University of Michigan Business School and the Ernst & Young Foundation for their financial support. Comments from Mark Bradshaw, Gerry Lobo, Anwer Ahmed, and workshop participants at Syracuse University are gratefully acknowledged. I appreciate the research assistance of Eleonora Krasteva on this and related projects. Comments are welcome. Please do not quote, copy or transmit without permission.

CEO Pay and Accounting Performance Measures: The Role of Earnings Management

Abstract This research investigates the extent to which accounting based performance measures explain executive compensation in the presence of earnings management. Popular press attention and public outcry over a handful of corporate scandals largely motivated by managerial greed provide the motive for this work. Yet the corporate failures and accounting manipulations reported thus far are but a small sample of companies making up the U.S. capital market. Do these well publicized problems of a handful of public entities warrant broad based public concern regarding accounting, auditing and corporate governance? I focus this investigation on the relation between accounting performance measures and annual CEO incentive compensation (cash bonus and stock incentives) for a sub-sample of companies with a relatively high likelihood of managing earnings. Employing two alternative measures of earnings management, I expect to find a weak or insignificant relation between pay and accounting performance when earnings management is most likely based on the assumption that governance mechanisms are effective. However, I find a relatively strong relation between accounting performance measures and incentive compensation for that sub-sample of companies with the highest likelihood of earnings management. Accounting performance measures appear to explain more of the variation in executive pay for that sub-sample that is most likely to have managed earnings. This is contrary to what might be expected if governance mechanisms were able to identify the incidence of managed earnings, and provides some evidence that broader concern over manipulated earnings aimed at increasing incentive pay of key executives may be warranted.

CEO Pay and Accounting Performance Measures: The Role of Earnings Management
Introduction The issue of concern in this research is whether there appears to be widespread management of the accounting performance measures used to reward corporate executives. The motivation frequently cited as the cause of accounting failures is the abuse of the accounting information system by managers in their efforts to window dress the performance of the entity and in so doing maximize their own wealth. This is not a new theme, as concern over the separation of ownership and management of the corporate entity and the related agency problems that can occur have been discussed in the literature at least since long before public markets and share ownership structures existed. The issue of managerial control to the detriment of shareholders was a focus of business literature as early as the 1930s (Berle and Means, 1932). But recent headlines such as “Deciding on Executive Pay: Lack of Independence Seen” (Henriques and Fabrikant, New York Times); “Before Enron, Greed Helped Sink the Respectability of Accounting” (Dugan, Wall Street Journal); and, “Few Cos In Full Compliance With Sarbanes-Oxley Act” (Dow Jones Newswires) suggest that problems with corporate governance and the abuse of accounting may be getting worse. Within the overall set of potential governance problems a first order effect is the ability of top management to enhance their wealth by manipulating accounting information and/or manipulating the monitoring mechanisms that oversee their actual performance as managers. A recent New York Times study (Editorial, 2002) reported that 420 out of 2,000 major public corporations examined had individuals on their

EM & CEO Pay

1/1/2010

1

compensation committees who were either related to or had ties to the CEO and the company. It is relevant to ask whether this is simply newspaper hype or evidence of a broad based problem with excessive control by managers to determine their own reward structure. In an effort to determine whether earnings management is more common than the handful of accounting scandals reported in the press, I consider the relation between the CEO’s annual incentive pay and the accounting performance of their entity. Specifically, I examine whether a high likelihood of earnings management appears to impact the strength of the relation between executive pay and accounting measures of performance. If mangers successfully manipulate accounting performance measures to achieve personal gain, we might expect a significant positive relation between accounting performance measures and executive pay in those cases where earnings management is considered to be most likely. However, if monitoring mechanisms of corporate boards provide effective oversight of salary related issues, then we should not expect to see earnings related to executive pay in firms where earnings management is considered most likely to have occurred. In cases where earnings has most likely been managed we should be more apt to observe non-earnings measures and balanced scorecard types of performance systems such that, at a minimum, earnings should be less related to managerial rewards when compared with high earnings quality (low earnings management) entities. This research examines the relation between several different accounting based performance measures and two separate measures of annual incentive pay: the annual cash bonus and the annual cash bonus plus stock based incentive compensation. For each

EM & CEO Pay

1/1/2010

2

of these relationships, I examine whether two different measures of earnings management appear to be related to the strength of the association between pay and accounting performance. As in prior research, I find that most commonly referenced measures of accounting performance are significantly related to executive pay for a large sample of public corporations. Once these full sample benchmark results are evaluated, I explore the impact of earnings management on the pay-performance relation. I find evidence suggesting that accounting based performance measures actually explain more of the cross-sectional variations in executive pay in the sub-sample (quintile) where earnings management is most likely to have occurred. These results indicate that managed earnings (as previously defined in our literature) are more closely related to executive pay than when earnings management is unlikely. These findings are consistent with the existence of broad based earnings management, and raise concerns regarding the inability of compensation committees to effectively evaluate managerial performance. The anecdotal evidence of accounting abuses from high profile cases of earnings management may in fact be the tip of the iceberg. Background Performance and Executive Pay In theory, the relative importance of various factors used to measure the performance of agents should be related to how well each measure informs the principal about the agent’s actual performance (i.e., Lambert and Larcker, 1987; Banker and Datar, 1989). For decades accounting measures have been used as primary indicators of managerial performance with prior research documenting a significant relation between accounting based performance and executive compensation (i.e., Antle and Smith, 1986,

EM & CEO Pay

1/1/2010

3

Bloedorn & Chingos, 1991; Ittner, et al., 1997). Moreover, both the annual cash bonus and the sum of the cash bonus plus stock based compensation have been linked to accounting based performance as well as numerous other attributes of the firm’s governance structure (Core, et al., 1999). The compensation literature suggests that most annual cash bonus plans for key executive officers are based in large part on accounting performance measures (Bloedorn & Chingos, 1991; Ittner, et al., 1997). There is also some relation between accounting performance and stock based compensation in many firms since the pool of stock options or stock awards to be distributed each year is often based on annual accounting performance measures. The literature has also documented a high correlation in the total annual inceptive pay amongst the top executives in each firm, and it is commonly assumed that what is observed for the CEO is representative of the incentive pay for the entire top management team for most entities (Antle & Smith, 1986; Gore, et al., 2003; Ittner, et al., 1997). The literature of accounting and finance has also extensively debated the issue of whether accounting information should be used to measure managerial performance (Bushman and Indjejikian, 1993; Kim and Suh, 1993; Sloan, 1993; Lambert, 1993). Because of the shareholders’ interest in maximizing the value of their shares, arguments have been made that only share price should be used to evaluate performance. Managers are typically rewarded most by their share holdings in the employer entity. Recent evidence suggests that the annual cash pay of a majority of CEOs is less than 10% of the annual benefits from their stock options and stock holdings in the employer entity (Core, et al., 2003). However most research suggests that to both provide meaningful incentives for managers and also to then monitor their performance, accounting information is

EM & CEO Pay

1/1/2010

4

essential. The links between accounting performance measures and the current and future market value of the firm justify their use as a target that may be impacted by managerial performance. From the evidence provided to date on executive compensation, I conclude that accounting based performance measures are less affected by externalities unrelated to the performance of the entity than are stock prices. At the same time, it is also observed from the empirical research to date that accounting performance measures are noisy and can not be expected to explain a majority of the variation in either executive pay or firm value. Based on prior work, much of the current executive compensation literature examines the relation between CEO compensation and accounting based performance. In addition, these studies have documented links between CEO pay and other attributes of firms related to their governance structure. These governance related variables have included firm size, number of board members, number of outside directors, number of interlocking directors, whether the CEO is also the Board Chair, and other governance characteristics (Core, et al., 1999; Dechow, Sloan & Sweeny, 1996). And commonly, accounting based performance measures tend to explain much more of the variance in executive pay across firms and time than do the governance characteristics (Core, et al., 1999). This study draws on the results of existing literature to justify the examination of the relation between accounting performance measures and CEO pay. I consider two measures of executive compensation based on the understanding that the Board of Directors is more directly responsible for setting the annual cash bonus than the sum of annual cash bonus plus stock based compensation. While the Board may have much to

EM & CEO Pay

1/1/2010

5

do with stock based compensation, the market value of the firm’s shares and its price stability are key elements in valuing the benefit to the employee, and these factors are beyond the control of the Board. Examining both the cash bonus and the total incentive pay including both cash bonus and stock based compensation will permit an evaluation of whether earnings management has a differential impact on these two measures. Earnings Management Much prior research has considered the question of how to measure the impact of earnings management (Healy, 1985; DeAngelo, 1986; Jones, 1991; Dechow, et al., 1995; DeFond and Subramanyam, 1998, etc.). Jones (1991) developed one of the most popular models of earnings management based on the concept of “abnormal accruals”. Jones documented that firms seeking tariff protection from imports made discretionary earnings decreasing adjustments prior to seeking government protection. Numerous subsequent studies have used the “Jones model” or variations of this model to represent the presence of earnings management (i.e., Dechow, et al., 1995; Dechow, et al., 2002; Holthausen, et al., 1995). Earnings management is a difficult measurement issue in that if it could be identified and measured without ambiguity, it would be simple to guard against. The very notion that earnings management can be measured suggests that it may not be an effective managerial tool. Still, much work has been done to define and identify instances of earnings management and to examine its impact on such things as: whether it is used to help managers meet/beat earnings expectations (Burgstahler and Dichev, 1997), whether it has helped companies obtain relief from imports (Jones, 1991); whether it has been used to create higher expectations for firms’ stock offerings (Teoh, et al.

EM & CEO Pay

1/1/2010

6

1998a, 1998b; Rao, et al. 1998), weather it has been used by property-casualty insurance companies to manipulate their loss reserves (Beaver, et al., 2000); and other economic events that might motivate managers to influence their accounting information (e.g., DeAngelo et al., 1994). In addition to the modified Jones model I examine an income smoothing model. Smoothing has long been suggested as the focus of much earnings management activity, and after the fact many accounting scandals (including Enron) have been blamed on managements efforts to sustain growth trends in sales and earnings. The decade of the 1990s provides one of the most promising for examining broad based earnings management through income smoothing. Empirical research on earnings management often refers to "income smoothing" behavior and suggests that managers seek to present a smooth and somewhat increasing trend in earnings from year to year (Copeland, 1968; White, 1970; Biedleman, 1973; Barnea et al., 1976; Albrecht and Richardson, 1990; Michelson, et al., 1995; DeFond and Park, 1997; Beaver, et al. 2000 ). One of the major disadvantages of research on income smoothing concerns the inability of researchers to convincingly infer earnings were managed to achieve an observed smooth earnings series. Artificially smoothed earnings are a special case of earnings management. But a smoothed earnings series is not a sufficient condition to document earnings management (Imhoff, 1975). A second impediment to the development of the smoothing literature has been the frequent requirement that sample firms have a long enough history of smoothing behavior to permit smoothing to be documented. Often, the period of time required to observe the

EM & CEO Pay

1/1/2010

7

earnings behavior is so great that changes in the composition of the entity or changes in macro economic conditions mask the earnings management that may have occurred. I do not attempt to differentiate between smoothing that is achieved through earnings management or through natural smoothing. Instead, I examine the time series properties of earnings over the prior ten years to determine if the current incentive pay is related to the smoothness of earnings over the prior decade. The decade of the 90s represents one of the longest periods of sustained economic growth in U.S. history, providing an opportunity to examine firm specific smoothing behavior. I employ the modified Jones model and a model of income smoothing in an effort to determine whether earnings management can help explain the strength of the payperformance relationship. Ex-ante identification of earnings management based on publicly available information is considered to be impossible. However, models used in prior research to assess earnings management are assumed to be relatively informative in that they point to cases where it is more likely that earnings may have been managed. In addition to multiple measures of earnings management and multiple measures of accounting performance, I use multiple measures of executive pay in an effort to examine the relation between pay and accounting performance. Accounting Performance Although accounting based performance measures are known to be used in practice, there is some ambiguity regarding which accounting performance measure is relevant across firms. Prior studies have identified a number of different accounting and non-accounting metrics that have been used by companies to measure their annual performance for purposes of setting executive compensation (Bleodorn & Chingos, 1991;

EM & CEO Pay

1/1/2010

8

Ittner, Larcker & Rajan, 1997). The fact that different accounting based performance measures are used by different companies presents a methodological problem for crosssectional studies examining the pay-performance relation. Two approaches have been used to resolve the “earnings” identification problem. The first is to examine firm specific descriptions of their incentive compensation schemes (usually found in proxy statements) in hopes that one or more relevant performance variables can be included for each firm. This approach has at least three important limitations when employed in large sample studies. First, the description of the accounting based performance measures employed by each firm may not eliminate all ambiguity concerning the variable employed.1 Second, when multiple measures are identified there may not be an obvious weighting scheme to determine which variables should be more or less important, and this may create confusion as to which variables to include/exclude in pooled cross-sectional analyses. Finally, when multiple measures are employed, it is not likely that the same set of measures will be relevant to a cross section of firms included in the sample of firms being examined by the research. 2 A second approach to dealing with the uncertainty in accounting performance measures is to rely on one or more of the “most common” measures that appear in incentive pay schemes. This approach relies on obtaining a large enough sample that performance measures are able to explain a statistically significant percentage of the cross-sectional variation in incentive pay. This approach is often used and typically
1

For example, if the proxy statement identifies “operating earnings”, or “net income from operations”, these measures may mean a number of different things to different companies. While terms such as “operating income” or “operating profits” are commonly referred to, there is no uniform definition regarding exactly which components they include/exclude that would eliminate all ambiguity. 2 It is assumed here that including all mentioned variables in a pooled cross section would not be feasible due to the large number of variables mentioned in any good sized sample and the compound effect of the measurement ambiguity of both accounting and non-financial based components.

EM & CEO Pay

1/1/2010

9

results in significant albeit weaker relations between accounting measures and executive pay. However, although the apparent relationship observed by the researcher is usually statistically significant, it may be largely based on the inter-correlation that exists among various accounting based performance measures in large samples of firms and over time. In sum, there are no data collection methods available to external users of company data that are without limitations. In an effort to consider the endogenous nature of the performance measure, I examine the relation between executive pay and a number of different accounting performance measures. Specifically, I consider the following measures of accounting performance to represent “earnings” (Compustat variable identification in parentheses): 1. Operating Income (#170-17-190) 2. Income before Extraordinary Items (#18) 3. Net Income (#172) 4. ROA1 = Operating income over total assets (#178 / #6) 5. ROA2 = Operating income over total assets (#170-17-190/#6) 6. ROA3 = Income before extraordinary over total assets (#18/#6) 7. ROE = Income before extraordinary over common equity (#18/#60) This approach allows the performance measure to vary, and permits an analysis of the impact of earnings management over different performance measures. Prior work, which has opted for treating the accounting performance measure as exogenous, has often used ROA as the accounting performance measure. I use three variations of ROA as well as ROE and three measures of “earnings”.3

3

Additional measures of earnings were examined, including EBITDA, Comprehensive Income, and cash based earnings. None of these other measures have been commonly identified as significant factors in incentive pay and none demonstrated a strong relation with incentive pay as accounting performance measures. Their results are not reported.

EM & CEO Pay

1/1/2010

10

The hypotheses examined by this research are designed to determine whether the significance of the relation between executive pay and accounting based performance measures persists in the presence of a high likelihood of earnings management. If the basic relationship observed between executive pay and accounting performance is the same or stronger in the sub-sample where a high likelihood of earnings management exists, it would suggest that earnings management is profitable for corporate executives and that governance mechanisms fail to detect performance measures that have been managed. Sample and Method The study examines executive pay during the period 1998-1999. During these two years, the market returns on the S&P 500 portfolio gained 26.6% and 19.5% respectively.4 The return for 2000 declined to 7.4% and for 2001 it fell to – 13%, while for 1997 the return was 31%. The two-year period 1998-1999 marks the last two years of a major growth period in the U.S. stock market, and much of this growth in market values was driven by growth in accounting performance measures during the period. Although what we observe in these two years may not be representative of all periods, it was selected because of the increased likelihood of observing the effects of earnings management on the pay-performance relationship. Prior to the realized downturn in the economy, it is predicted that more firms will be managing their earnings to sustain the positive growth trend in their firms and the economy in general. In addition to examining the contemporaneous relation between executive pay and annual accounting performance measures during these two years, the study draws on the historical relationships within the financial statements for the ten-year period 1990-1999
4

These and other statistics regarding returns on the S&P 500 are taken from the internet at Yahoo Finance.

EM & CEO Pay

1/1/2010

11

for measures of earnings management. This time-period provides a decade of relatively constant growth and permits the use of firm specific measures of earnings management observed over a decade ending with the 1998-1999 period. The data required for the study includes information about CEO compensation for 1998 and 1999 that are taken from ExecuComp. To be included in the study the CEO had to be in that position prior to 1997.5 Firm specific measures of income statement and balance sheet data are taken from Compustat. The number of observations for each compensation measure and performance measure combination exceeds 2,000 (1,000 per year). The number of observations for each of the subsequent analyses of the payperformance relations varies between 1,445 and 1,655 depending on further data restrictions imposed from ten consecutive years of usable data as noted in the results. The basic relationship considered in order to provide benchmark results for examining the impact of earnings management is the linear relation between the annual amount of incentive pay (only) as the dependent variable and the change over the prior year in the accounting performance measure as the independent variable. The model also includes total assets as an independent control variable.6 Size is related to a number of governance variables such as board size and number of directors, and provides a useful control for omitted variables. The basic model of pay-performance takes the following form:

5 6

Some sensitivity tests used changes in compensation too, requiring three full years of compensation data. Using size as a control variable influenced the specification of the basic model reported in the paper. For example, you could not use size along with the level of ROA as a performance measure due to the confounding effects. The results reported here were also obtained without using size as a control and using levels of performance measures with levels of annual incentive pay. Using the quantity of annual incentive pay as it relates to the change in the performance measure treats the prior year as a benchmark for the current year’s incentive pay. This is not obviously superior or inferior to other specifications of the payperformance relation.

EM & CEO Pay

1/1/2010

12

BP (ICP) it =  + 1 (%Performance)it + 2 (Size)it + it Where:

(1)

BPit = Bonus Percent defined as annual cash bonus divided by total cash compensation for CEO of firm i in year t. ICPit = Incentive Compensation Percent defined as total compensation less base salary divided by total compensation for CEO of firm i in year t. . % Performance it = % change in performance measure over prior year divided by prior year's measure for firm i in year t, with sample trimmed by top and bottom 1% for each performance measure. Size it = Total assets for firm i in year t. The dependent variable takes on the form of the annual bonus percent (BP) or the more inclusive measure of incentive pay (ICP), which includes stock based compensation plus the cash bonus as a percentage of base salary. The measure of accounting performance is the percentage change in one of the seven accounting performance measures: 1. 2. 3. 4. 5. 6. 7. Operating Income (#170-17-190) Income before Extraordinary Items (#18) Net Income (#172) ROA1 = Operating income over total assets (#178 / #6) ROA2 = Operating income over total assets (#170-17-190/#6) ROA3 = Income before extraordinary over total assets (#18/#6) ROE = Income before extraordinary over common equity (#18/#60)

To control for the effects of size an independent variable is included for the total assets (Compustat item #6) of the entity at the end of the year. 7 The full sample results from the model (1) relation between CEO incentive pay and the change in accounting performance measures provides the benchmark for the study. Once these pay-performance relations have been described, I examine how the likelihood of earnings management affects relation for each of the performance

7

By examining the annual percentage change in the performance measures rather than the level of each measure, the explanatory power of the model is reduced (the variance in the independent variables is increased). However, it permits the use of assets as a control variable. Prior work has demonstrated the importance of size as a control variable and the tradeoff was considered to be necessary in this case.

EM & CEO Pay

1/1/2010

13

measures.8

The method employed seeks to determine whether the pay-performance

relation observed in the full sample differs for that sub-sample of firms where earnings management is most likely to have occurred. Measuring Earnings Management Given that earnings management can not be unambiguously identified, I employ two models to estimate the likelihood of earnings management. The first is the modified Jones model most commonly used to depict earnings management over the past decade of accounting research (Dechow, Richardson & Tuna, 2002). This model measures the ability of revenues, adjusted for the change in accounts receivable, and plant assets (i.e., fixed operating costs) to explain the total accruals in “earnings”. The theory underpinning this model is that the total accruals affecting the measure of periodic income should be related to the change in revenues and the level of fixed assets. The residuals represent “abnormal accrual” behavior and the greater the abnormal accruals, the higher the likelihood of earnings management. I examine the following model: Total Accruals (TA) = α + β1 (∆Sales-∆A/Rec) + β2 (PP&E) + ε (2)

Total accruals are measured as the difference between income before extraordinary items (IBXO) (Compustat #18) and cash from operations (#308). Model (2) is commonly employed by estimating the coefficients for each industry in order to avoid the need for historical data. However, this approach relies on the assumption that the parameter estimates are constant within each industry. I use firm specific data for the ten year period 1990-1999 to estimate model (2) for each firm. The measure of total accruals is constant across all variations of accounting performance measures.
8

It would have been possible to simply introduce the measures of earnings management into the model (1) regression as additional independent variables to examine the impact of earnings management. However, there is no reason to expect that earnings management is linearly related to CEO pay.

EM & CEO Pay

1/1/2010

14

The measure of earnings management is the R2 statistic from model (2) for each firm, providing a scale free cross-sectional measure. The R2 statistic provides a relative measure of earnings management in that the higher the R2 the less the abnormal accruals (smaller residuals relative to total accruals). A low R2 statistic would indicate that the independent variables explain less of the variance in total accruals resulting in higher amounts of “abnormal accruals” and hence a higher probability of earnings management. If managers are effective at managing earnings to achieve superior compensation, we expect the explanatory power of the accounting performance measure to be as strong or stronger in the sub-sample with the greatest amount of abnormal accruals (lowest R2 from model 2 ). Alternatively, if high abnormal accruals are observed by corporate boards and their compensation committees, they should lead to less emphasis on accounting performance measures. If such were the case we would expect to see a weak or insignificant relation between pay and accounting performance for the high abnormal accrual sub-sample. In addition to the R2 from the firm specific regression for total accruals, I use a regression model to assess income smoothing behavior. I measure the R2 from the firm specific regression: “Earnings” t = α + β (“Earnings”)t-1 + ε (3)

“Earnings” are defined as one of the seven performance measures noted previously, and is regressed over the ten-year period 1990-1999. When the performance measure is a ratio, the ratio is used as the dependent and independent variable. The R2 statistic provides a cross-sectional measure of the relative stability of the earnings series for each performance measure. Firms with smooth (variable) “earnings” streams over the decade

EM & CEO Pay

1/1/2010

15

will generate high (low) R2 statistics from model (3). In this measure of earnings management, the variable being evaluated for its smoothness over time will be consistent with the performance measure used in model (1). As a result, the composition of each stratum changes with each “earnings” measures. If managers smooth earnings in order to achieve superior compensation, we might expect the relation between accounting performance measures and executive pay to be relatively strong for the sub-sample with the smoothest earnings measure. Alternatively, if smoothing behavior were detected by governance mechanisms to have been achieved through earnings management we would expect the smoothed performance variable to be weakly or insignificantly related to executive pay for the sub-sample with the smoothest earnings. Once the basic relation between pay and accounting performance is examined for each of the accounting performance measures, the two measures of earnings management are then considered. The full sample of firms is stratified based on the degree of earnings management represented by each of the two measures. Examining the strength of the relation between pay and accounting performance within each of the strata will permit us to determine whether accounting performance measures are more, less, or similarly related to executive pay when the relative level of earnings management is high or low. The focus of attention will be on the strata with the highest amount of earnings management. Ex-ante, if the governance process is effective, we might expect to see firms with relatively high earnings management have a weaker relation between accounting performance and executive pay. If the governance process in not effective, we might see accounting performance related to executive pay even when earnings

EM & CEO Pay

1/1/2010

16

management is relatively high. If there is no difference in the pay-performance relation across varying levels of earnings management, it is possible that earnings management does not have an impact on the pay-performance relation. Alternatively, it may simply suggest that the measures of earnings management lack precision and/or that by using only a single accounting performance measure in each observation the pay-performance relation is not sufficiently complete. Results and Analysis Basic Pay-Performance Relation Based on the evidence provided by the body of literature concerned with executive pay and accounting performance, most of the performance measures were expected to be significantly related to executive pay after controlling for firm size. From the results reported in Table 1, it appears that the percentage change in each of the measures of accounting performance is significantly related to each of the two measures of CEO incentive pay (BP and ICP). The size variable is also significant in each of the regressions and provides control for omitted variables that may be associated with firm size. Although Table 1 reveals that all seven accounting performance measures are statistically significant at the p < .01 level, the average explanatory power of the model (1) relation is relatively low. The overall average explanatory power of model (1) reported in Table 1 is 5.8%. The average for model (1) when BP is the dependent variable is 6.1% and the average for model (1) when ICP is the dependent variable is 5.7%. An average of 6.3% of the variation in executive incentive pay is explained by the three measures of “earnings” performance and an average of 5.5% of the variation is explained by the four ratios.

EM & CEO Pay

1/1/2010

17

The ability of the alternative accounting performance measures employed in the model (1) relation to explain the variation in CEOs’ bonus percent (BP) is very comparable to their ability to explain the variation in CEOs’ total incentive compensation (ICP). For each of the alternative performance measures, the ability to explain BP is similar to its ability to explain ICP, with the ROA2 ratio [Operating Income (#170-17190) over Total Assets] resulting in the largest difference within a performance measure (5.24% for BP and 3.87% for ICP). Results regarding the basic pay-performance relation suggest that no one accounting performance measure is significantly better than another for explaining executive pay in large samples. However, from looking at these results it is not clear whether different performance measures are systematically better or worse at explaining executive pay for sub-sets of the overall sample. Any given performance measure may be more appropriate for some companies than for others. To address this issue, and to determine how much more explanatory power might be provided by these accounting performance measures, I examine a subset of firms with the smallest model (1) residuals for each performance measure. Table 2 provides the results of the model (1) regression when applied to the 20% of the sample generating the smallest model (1) residuals from Table 1.9 By design these sub-sample results generate a much higher explained variation in executive pay with the overall average R2 equal to 74%. The explanatory power of accounting performance measures for this sub-sample seems to be somewhat greater for the Bonus Percent (BP: average R2 = 78%) than the overall Incentive Compensation Percent ( ICP: average R2 = 70%). Although both cash

9

The results are provided for only five of the performance measures as the others are qualitatively the same.

EM & CEO Pay

1/1/2010

18

and stock based bonuses are commonly related to accounting performance, it is not unexpected that the cash bonus might be more directly linked to accounting performance more often than stock based compensation. One reason for using a number of different specifications of the accounting performance measures as the independent variable in model (1) was to allow for the possibility that different firms might employ different accounting performance measures. Although it is well documented that different measures of accounting performance in large samples of companies over time are significantly correlated with one another, their correlations are generally less than 1.0 since they are not perfect substitutes. I examine whether the firms included in the sub-sample with the smallest model (1) residuals for each performance measure are the same firms for each performance measure. In other words, do different firms appear to use different performance measures? Examining each pair wise comparison across the five measures reported in Table 2, I find between 85% and 95% overlapping membership. The average of the 10 paired comparisons for BP was 90% overlap; while for ICP it was 92% overlap. The maximum of 15% non-overlapping membership was for BP between the measures of Operating Income and ROE. Although the correlation among the accounting performance measures is clearly significant for the Table 2 data, it is also less than 1.0 as expected. Impact of Earnings Management To examine the impact of earnings management on the basic pay-performance relationship, I examined the sub-sample of companies reported in Tables 1 and 2 that had ten consecutive years of non-missing data for the variables used to measure accounting performance as well as sales, accounts receivable, and plant, property and equipment. I

EM & CEO Pay

1/1/2010

19

then computed the R2 statistic for each company for each of the two models of earnings management. Companies were then stratified into quintiles based on the rank of their R2 statistic from each of the earnings management models. The basic model (1) regression was then computed for each quintile of firms. This approach permits an analysis of how the pay-performance relation for the quintile where earnings management is most likely differs from the full sample of firms. The two measures of earnings management combined with seven measures of accounting based performance and two different measures of incentive pay resulted in 140 model (1) regressions. In an effort to summarize these 140 regression results, I focus on the T-statistics for β1 (the coefficient on the accounting performance measure) and the overall explanatory power of the model (1) regression for each quintile. Table 3 provides a summary of the 70 regressions for the total accruals measure of earnings management (Model 2), and Table 4 provides a summary of the 70 regressions for the earnings smoothness measure from model (3). Panel A in both Tables 3 & 4 provides the results for the cash bonus incentive measure while Panel B in both tables provides the results for the total incentive compensation measure. Abnormal Accruals When examining the quintile with the highest amount of abnormal accruals, Table 3 provides what is perhaps the most interesting set of results. The highest abnormal accruals are found in the lowest quintile (where the R2 statistic from model 2 was the lowest). In both Panel A and Panel B of Table 3 we find that the average t-statistic on β1 across the seven measures of “earnings” is highest for that quintile where earnings management is most likely. In Panel A, the average t-statistic is 3.5 while in Panel B it is

EM & CEO Pay

1/1/2010

20

3.13. For both measures of annual CEO incentive pay the relation between earnings and pay is strongest in the sub-sample with the greatest amount of “abnormal accrual” behavior. For the BP measure of incentive pay, only ROA2 (Operating income defined as Income before interest expense and special items deflated by total assets) was not significant at the p<.05 level (significance level was p = 0.058). The average R2 statistic from the model (1) regression for BP in the quintile of total accruals with the largest amount of “abnormal accruals” was 7.9%, which is higher than the average from Table 1 of 6.1% for BP (only). In Table 3 Panel B, all of the accounting performance measures for the lowest quintile (highest abnormal accruals) are significantly related to the CEO’s total Incentive Compensation Percent. The average t-statistic of 3.13 is the highest of any quintile. The average R2 for the ICO measure was 6.9% compared to the average of 5.7% from Table 1 results for ICP (only). In addition to the overall BP and ICP analyses, it was noted that the average t-statistics on β1 for the three “earnings” performance measures (ignoring the four performance ratios) was 4.15 for BP and 3.57 for ICP for the high earnings management quintile. Moreover, the t-statistic on β1 for the high earnings management quintile was higher than the t-statistic in any other quintile for each of these three earnings measures for both BP and ICP with only two exceptions out of 24 comparisons.10 The Table 3 results are viewed to stand on their own without further statistical tests. These results are essentially descriptive of a population as defined by the full sample. The Table 3 results are simply descriptive data on that portion of the full sample

10

In Panel B of Table 3, only quintile 3 for Operating income and quintile 4 for Net income generated a higher t-statistic for β1 than that in the high abnormal accrual quintile.

EM & CEO Pay

1/1/2010

21

with sufficient data available for the model 2 regressions. As such, any differences between quintiles may be viewed as “real” differences. By design, it is unclear whether the outcomes of the tests for the 1998-1999 period would hold in any other periods without further data collection and analysis. In general, the Table 3 results indicate that the accounting performance measures are not only significantly related to both measures of annual CEO incentive pay for the high abnormal accruals sub-sample, but the accounting performance measures seem to be most important in explaining annual incentive pay for this sub-sample of firms. These results are consistent with the premise that extreme cases of abnormal accruals (earnings management) are effective at fooling compensation committees into granting annual cash bonuses based on managed measures of accounting “earnings”. This inference holds for both cash based incentive pay and for incentive pay that includes stock based compensation. Income Smoothing The Table 4 results are based on the income smoothing measure of earnings management. The quintiles measure the stability or smoothness of the earnings stream over the period 1990-1999, and firms with the smoothest series are found in the highest quintile. If smoothing is achieved through earnings management practices that were observed by governance mechanisms, we might expect that accounting performance measures of firms with the smoothest earnings quintile would not be related to CEO incentive pay. Alternatively, we might expect the pay-performance relation to be significant for the smoothest quintile if smoothing achieved through earnings management had been effective at fooling governance mechanisms.

EM & CEO Pay

1/1/2010

22

The results reported in Table 4 are mixed. In Panel A of Table 4 where the cash bonus is the measure of incentive pay the average t-statistic for the smoothest quintile is 3.05 which is higher than any other quintile average in Panel A. However, when the tstatistics for the smoothest quintile are compared to those in other quintiles for each of the “earnings” measures they are not as dominant as they were in Panel A of Table 3. One potential limitation of the Table 4 results is that it may not be as relevant to consider the smoothness of ratios (ROA and ROE) as the smoothness of various measures of (undeflated) “earnings”. In this regard, the last row of the two panels of data in Table 4 report the average t-statistics for the three earnings measures only (the unshaded area). For the three undeflated measures of performance the t-statistics are highest for the smoothest and the least smooth earnings series. The data suggest a U shaped relation between cash bonus percent and the three measures of earnings performance, with accounting performance most relevant for the smoothest and the most volatile quintiles respectively. For the ICP measure of incentive pay reported in Panel B of Table 4, the tstatistics for β1 are not statistically greater than zero for 4 of the 7 accounting performance measures in the smoothest quintile. The average t-statistic of 1.22 for the smoothest quintile in Panel B of Table 4 is the lowest average t-statistic of any quintile. The Panel B results in Table 4 are generally not supportive of the claim that smoothed earnings are used to boost the CEO’s annual incentive pay that includes stock based compensation. The discrepancy between Panel A and Panel B of Table 4 could have several explanations. For one, it remains problematic that smoothed earnings might not imply

EM & CEO Pay

1/1/2010

23

manipulation. It is feasible that a smooth earnings series is simply the result of good business management rather than earnings management. There is not obvious way to determine whether the smooth series was achieved through earnings management.11 Another possible explanation of the difference between Panels A & B in Table 4 is that markets see through the management to smooth earnings and market prices do not respond favorably to such smoothing. This could explain why the cash bonus aspect of pay is related to smoothed earnings while the inceptive pay that includes stock based compensation is not related to earnings.12 The examination of the Table 4 results lends some support to the premise that managers can achieve superior cash bonuses by reporting smoothed “earnings”. These results must also be tempered by the possibility that some of the extremely smooth earnings streams are “natural smoothers”, having achieved the smoothed earnings through real operating stability and growth rather than earnings management. Summary The results observed from both measures of earnings management offer some basis for broad concern regarding the ability of managers to manipulate accounting based performance measures. The results reported in Table 3 and, to a lesser extent, Table 4 suggest that effective earnings management may be relatively common within the subsamples of firms considered most likely to be managing earnings. For these subsets of firms, the pay-performance relation remained statistically significant and relatively strong
11

Interestingly, the firm-year observations in the high abnormal accruals quintile do not overlap by more than a few observations with the firm-year observation in the smoothest earnings quintile. These two models (2 & 3) are far from substitutes for one another, and represent two distinctively different proxies for earnings management. 12 Presumably the stock component would dominate the cash bonus component. This interpretation also suggests that markets may be better at detecting earnings management to smooth than they are earnings management via abnormal accruals.

EM & CEO Pay

1/1/2010

24

compared to other sub-samples, particularly for the cash bonus measure of incentive pay. These results suggest that the corporate board and its compensation committee, which is directly responsible for setting the annual cash bonus, appears to be unaffected by and/or unaware of the higher likelihood that earnings have been managed. Implications and Limitations This study focuses on the relation between accounting performance measures and annual executive incentive pay. While other factors involving governance and legal climate have been found to explain executive pay in both domestic and international studies, it is widely accepted that accounting based performance is an important link to understanding management compensation. Since many accounting failures in public corporations are linked to managerial self-interests, the first order relation between management pay and accounting performance is a relevant focus. The question of interest here centers on whether earnings management appears to be commonly practiced in companies where earnings based performance are linked to executive pay. I find that, for the quintile of firms most likely to be managing earnings, the change in the earnings performance measure explains more of the variations in annual incentive pay measures than in any other quintile. These results are somewhat stronger for measures of the annual cash bonus, which are largely determined by the board’s compensation committee than for measures of annual cash bonus plus stock compensation, which is subject to market influences. Perhaps the most relevant implication from these results is that the management of earnings measures used as a basis for incentive pay appears to be anything but uncommon.

EM & CEO Pay

1/1/2010

25

While I make the claim of earnings management for pay being common, I also admit that measures of earnings management are far deterministic. Researchers are not able to prove the incidence of earnings management and may not be able to do so anytime soon. Both measures of earnings management employed here may result in specification errors. Unexplained accruals do not necessarily indicate earnings management. Smoothed earnings do not necessarily mean earnings have been manipulated. Of course if a measure of earnings management were found to be deterministic it would quickly lead to an elimination of that form of earnings management. And presumably other means of achieving managed earnings would still remain available to creative managers. Given these limitations, the measures of identifying earnings management employed here suggest that more than a handful of top executives may be manipulating earnings for personal gain. To the extent that these measures of earnings management are relevant the results imply that governance mechanisms may not be effective at monitoring executive performance. Future work might explore governance and other characteristics of firms that appear to be most likely to be managing earnings. Future work might also examine whether the results reported here hold for other time periods. And similar tests of the pay-performance relation might be applied to other forms of evidence that earnings have been managed, including those firms who just meet or beat earnings forecasts or those firms who have had to restate prior years’ results. By focusing attention on that subset of firms where earnings management is most likely to occur, a better understanding might be gained into the role of earnings management in the governance of the firm.

EM & CEO Pay

1/1/2010

26

References
Abowd, J.M. and D.S. Kaplan, 1999. Executive Compensation: Six Questions that Need Answering. Journal of Economic Perspectives XIII: 145-168. Albrecht, W. and F. Richardson. 1990. Income Smoothing by Economic Sector. Journal of Business Finance & Accounting. December. 17(5): pp. 713-730. Antle, R. and A. Smith, 1986. An Empirical Investigation of the Relative Performance Evaluation of Corporate Executives. Journal of Accounting Research, Spring, 24: pp. 139. Banker, r. and S. Datar. 1989. Sensitivity, Precision, and Linear Aggregation of Signals for Performance Evaluation. Journal of Accounting Research. Spring: pp. 21-39. Barnea, A., J. Ronen and S. Sadan, 1976. Classificatory Smoothing of Income with Extraordinary Items. The Accounting Review, January, pp. 110-122. Beaver, W., M. McNichols and K. Nelson. 2000. Management of the Loss Reserve Accrual and the Distribution of Earnings on the Property-Casualty Insurance Industry. Stanford University Working Paper. Berle, A. and G. Means. 1932. The Modern Corporation and Private Property. New York, NY and Chicago, IL: Commerce Clearing House. Bertrand, M., and S. Mullainathan. 2001. Are CEOs Rewarded for Luck? The Ones Without Principals Are. The Quarterly Journal of Economics CXVI: 901-931. Biedleman, C., 1973. Income Smoothing: The Role of Management. The Accounting Review, October, pp. 653-68. Bloedorn, J.D. and P.T. Chingos. 1991. Executive Pay and Company Performance. Scottsdale, Arizona: American Compensation Association. Burgstahler, D. and I. Dichev. 1997. Earnings management to avoid earnings decreases and losses. Journal of Accounting and Economics 24: 99-126. Bushman, r. and R. Indjejikian. 1993. Accounting income, stock price, and managerial compensation. Journal of Accounting and Economics. 16 January/April/July: pp. 3-24 Copeland, R., 1968. Income Smoothing. Journal of Accounting Research, Supplement, pp. 101-16.

EM & CEO Pay

1/1/2010

27

Core, J., W Guay and R Verrecchia. 2003. Price versus Non-Price Performance Measures in Optimal CEO Compensation Contracts. The Accounting Review. 78 October: pp. 957-982. Core, J., R. Holthausen and D. Larcker. 1999. Corporate governance, chief executive officer compensation, and firm performance. Journal of Financial Economics, 51: pp.371-406. DeAngelo, L. 1988. Managerial competition, information costs, and corporate governance: The use of accounting performance measures in proxy contests. Journal of Accounting and Economics 10: 3-36. DeAngelo, H., L. DeAngelo, and D. Skinner, 1994. Accounting Choice in Troubled Companies. Journal of Accounting and Economics, January, pp. 113-144. Dechow, P.M., R.G. Sloan and A.P. Sweeney. 1995. Detecting Earnings Management. The Accounting Review 70: 193-226. _________,___________,_________, 1996. Causes and Consequences of Earnings Manipulation: An Analysis of Firms Subject to Enforcement Actions by the SEC. Contemporary Accounting Research, Spring, 13:1, pp.1-36. Dechow, P.M., S.A. Richardson and I.A. Tuna. 2002. Why are earnings kinky? University of Michigan working paper, July. DeFond, M. and K. Subramanyam, 1998. Auditor changes and discretionary accruals. Journal of Accounting and Economics 25: pp. 35-67. ________, and C. Park. 1997. Smoothing Income in Anticipation of Future Earnings. Journal of Accounting and Economics. July, 23-2: pp. 115-39. Dow Jones Newswires. 2003. Few Cos In Full Compliance With Sarbanes-Oxley Act: Study. The Wall Street Journal Online. April 22. Dugan, I. 2002. Before Enron, Greed Helped Sink The Respectability of Accounting. The Wall Street Journal. March 14. Editorial. 2002. C.E.O.’s and Their Paymasters. The New York Times. December 23. Gore, A., S. Matsunaga and E. Yeung. 2003. Does the Financial Expertise of Monitors Matter? Evidence from the Cash Compensation of Chief Financial Officers. Working paper, University of Oregon. Healy, P.M. 1985. The effect of bonus schemes on accounting decisions. Journal of Accounting and Economics 7: 85-107.

EM & CEO Pay

1/1/2010

28

Henriques, D. and G. Fabrikant. 2002. Deciding on Executive Pay: Lack of Independence Seen. The New York Times. December 18. Imhoff, E., 1975. Income Smoothing: The Role of Management: A Comment. The Accounting Review, January, L:1, pp. 118-121. Ittner, C.D., D.F. Larcker and M.V. Rajan. 1997. The Choice of Performance Measures in Annual Bonus Contracts. The Accounting Review 72: 231-255. Kim, O., and Y. Suh. 1993. Incentive efficiency of compensation based on accounting and market performance. Journal of Accounting and Economics. 16 January/April/July: pp. 25-54. Kothari, S.P., A.J. Leone and C.E. Wasley. 2001. Performance Matched Discretionary Accrual Measures. M.I.T. working paper, November. Jones, J. 1991. Earnings management during import relief investigations. Journal of Accounting Research 29: 193-228. Lambert, R. 1993. The use of accounting and security price measures of performance in managerial compensation contracts: A discussion. Journal of Accounting and Economics. 16 January/April/July: pp. 101-124. __________ and D. Larcker. 1987. An analysis of the use of accounting and market measures of performance in executive compensation contracts. Journal of Accounting Research. 25 Supplement: pp. 85-125 McNichols, M., and G.P. Wilson. 1988. Evidence of earnings management from the provisions for bad debts. Journal of Accounting Research 26 (Supplement): 1-31. Michelson, S., J. Jordan-Wagner, and C. Wooton. 1995. A Market Based Analysis of Income Smoothing. Journal of Business Finance and Accounting. December, 22-8: pp. 1179-93. Murphy, K.J. and P. Oyer. 2001. Discretion in Executive Incentive Contracts: Theory and Evidence. USC working paper December. Murphy, K.J. 1999. Executive Compensation. In Handbook of Labor Economics, Vol III (North Holland, Ashenfelter and Card eds.) Parfet, W.U. 2000. Accounting Subjectivity and Earnings Management: A Preparer Perspective. Accounting Horizons 14: 481-488. Rao, G., S. Teoh and T Wong, 1998. Are the Accruals During an Initial Public Offering Opportunistic? Review of Accounting Studies, 3: pp.175-208.

EM & CEO Pay

1/1/2010

29

Sloan, R. 1993. Accounting earnings and top executive compensation. Journal of Accounting and Economics. 16 January/April/July: pp. 55-100 Teoh, S., I Welch and T. Wong, 1998a. Earnings Management and the Post-Issue Performance of Seasoned Equity Offerings, Journal of Financial Economics. October, 50:pp. 63-99. _________,_______,______, 1998b. Earnings Management and the Long-term Market Performance of Initial Public Offerings, Journal of Finance. December, 53: pp. 19351974. White, G., 1970. Discretionary Accounting Decisions and Income Normalization. Journal of Accounting Research. Autumn. pp. 260-73.

EM & CEO Pay

1/1/2010

30

Table 1 Contemporaneous Relation Between CEO Pay and Change in Accounting Performance Measures
Equation: BP(ICP)it =  + 1 (%Performance)it + 2 (Size)it + it Where: (1)

BPit = Bonus Percent defined as annual cash bonus divided by total cash compensation for CEO of firm i in year t. ICPit = Incentive Compensation Percent defined as total compensation less base salary divided by total compensation for CEO of firm i in year t. % Performance it = % change in performance measure over prior year divided by prior year's measure for firm i in year t, with sample trimmed by top and bottom 1% for each performance measure. Size it = Total assets for firm i in year t.
Performance Measure Operating Income #(170-17-190) Dependent Variable (BP/ICP) BP ICP Income before X/O #18 BP ICP Net Income #(172) BP ICP OI*/Assets #[(178)/6] BP ICP OI/Assets #(170-17-190/6) BP ICP Income before X/O/Assets #(18/6) Income before X/O/Common Equity #(18/60) BP ICP BP ICP Number of observations 2408 2420 2407 2419 2407 2419 2407 2419 2143 2155 2404 2416 2407 2419 Coefficient on Performance (t-stat) 0.027 (7.26)** 0.027 (6.10)** 0.014 (6.33)** 0.020 (7.42)** 0.013 (5.97)** 0.017 (6.85)** 0.037 (5.79)** 0.027 (3.57)** 0.032 (6.35)** 0.026 (4.49)** 0.018 (6.27)** 0.021 (6.31)** 0.016 (6.14)** 0.021 (7.05)** Coefficient on Total Assets (t-stat) 0.190 (11.07)** 0.213 (10.64)** 0.190 (11.05)** 0.215 (10.72)** 0.190 (11.07)** 0.215 (10.73)** 0.190 (11.08)** 0.213 (10.60)** 0.183 (9.03)** 0.199 (8.35)** 0.189 (11.03)** 0.213 (10.64)** 0.189 (11.02)** 0.213 (10.61)** Adjusted R2 6.68% 5.74% 6.19% 6.89% 6.06% 6.17% 5.93% 4.78% 5.24% 3.87% 6.18% 5.87% 6.15% 6.24%

EM & CEO Pay

1/1/2010

31

Table 2 Contemporaneous Relation Between CEO Pay and Change in Accounting Performance Measures Top 20% of Full Sample
Equation: BP(ICP)it =  + 1 (%Performance)it + 2 (Size)it + it Where: (1)

BPit = Bonus Percent defined as annual cash bonus divided by total cash compensation for CEO of firm i in year t. ICPit = Incentive Compensation Percent defined as total compensation less base salary divided by total compensation for CEO of firm i in year t. % Performance it = % change in performance measure over prior year divided by prior year's measure for firm i in year t, with sample trimmed by top and bottom 1% for each performance measure. Size it = Total assets for firm i in year t.
Performance Measure Operating Income #(170-17-190) Dependent Variable (BP/ICP) BP ICP Income before X/O #18 BP ICP Net Income #(172) BP ICP OI/Assets #(170-17-190/6) BP ICP Income before X/O/Common Equity #(18/60) BP ICP Number of observations 481 483 480 483 480 483 481 483 480 483 Coefficient on Performance (t-stat) 0.029 (20.97)** 0.026 (15.03)** 0.014 (18.13)** 0.019 (20.95)** 0.012 (15.90)** 0.018 (17.77)** 0.031 (20.79)** 0.025 (10.51)** 0.016 (19.80)** 0.024 (23.21)** Coefficient on Total Assets (t-stat) 0.184 (36.70)** 0.213 (27.85)** 0.185 (36.26)** 0.213 (28.41)** 0.184 (35.51)** 0.213 (28.02)** 0.184 (37.16)** 0.211 (27.72)** 0.184 (36.58)** 0.214 (27.79)** Adjusted R2 78.65% 67.35% 77.09% 72.44% 75.80% 70.11% 78.81% 64.46% 78.11% 73.41%

EM & CEO Pay

1/1/2010

32

Table 3 Significance of the Pay-Accounting Performance Relation Across Quintiles Representing the Degree of Abnormal Accruals: BP(ICP)it =  + 1 (%Performance)it + 2 (Size)it + it Where:

(1)

BPit = Bonus Percent defined as annual cash bonus divided by total cash compensation for CEO of firm i in year t. ICPit = Incentive Compensation Percent defined as total compensation less base salary divided by total compensation for CEO of firm i in year t. . % Performance it = % change in performance measure over prior year divided by prior year's measure for firm i in year t, with sample trimmed by top and bottom 1% for each performance measure. Size it = Total assets for firm i in year t. Panel A: Cash Bonus as a Percent of Cash Compensation

Ranked R2 From Total Accruals Model 2
Accounting Performance Measure
Operating Income (#170-17-190) Statistics From Model 1 Highest Abnormal Accruals Quintile #2 Quintile #3 Quintile #4 Lowest Abnormal Accruals

β1 t-stat R2 Income before X/O (#18) β1 t-stat R2 Net Income (#172) β1 t-stat R2 OI/A (#178/ #6) β1 t-stat R2 OI/A (#170-17-190/ #6) β1 t-stat R2 IBXO / A (#18 / #6) β1 t-stat R2 IBXO / CE (#18 / #60) β1 t-stat R2 2 Average R for Quintile Average t-Stat on β1 for Quintile

3.44 7.5% 5.31 12.2% 3.71 8.1% 2.36 5.5% 1.58 4.4% 4.25 9.3% 3.87 8.4% 7.9% 3.50

2.24 5.5% 1.58 4.8% 2.17 5.6% 1.17 4.5% 1.39 5.8% 1.52 4.8% 0.88 4.4% 5.1% 1.56

2.77 6.5% 3.16 7.2% 1.35 4.7% 2.13 5.6% 3.69 8.8% 3.12 7.2% 1.56 4.9% 6.4% 2.54

1.88 5.1% 1.21 4.3% 3.08 7.1% 1.07 4.3% 1.89 5.2% 1.03 4.4% 1.12 4.4% 5.0% 1.61

3.26 10.1% 2.45 9.2% 2.64 9.5% 3.72 11.6% 2.27 7.8% 2.28 9.0% 2.03 8.6% 9.3% 2.66

EM & CEO Pay

1/1/2010

33

Panel B: Total Compensation less Salary as a Percentage of Total Compensation

Ranked R2 From Total Accruals Model 2
Accounting Performance Measure
Operating Income (#170-17-190) Statistics From Model 1 Highest Abnormal Accruals Quintile #2 Quintile #3 Quintile #4 Lowest Abnormal Accruals

β1 t-stat R2 Income before X/O (#18) β1 t-stat R2 Net Income (#172) β1 t-stat R2 OI/A (#178/ #6) β1 t-stat R2 OI/A (#170-17-190/ #6) β1 t-stat R2 IBXO / A (#18 / #6) β1 t-stat R2 IBXO / CE (#18 / #60) β1 t-stat R2 Average R2 for Quintile Average t-Stat on β1 for Quintile

3.00 6.5% 4.51 10.0% 3.20 6.9% 2.96 6.4% 2.17 5.0% 3.20 7.0% 2.87 6.3% 6.9% 3.13

1.81 4.9% 0.89 4.0% 2.84 6.4% 0.89 4.1% 0.99 4.1% 2.58 6.0% 1.99 5.1% 4.9% 1.71

3.10 6.4% 3.45 7.1% 1.86 4.6% 1.37 4.0% 3.06 6.6% 3.08 6.5% 2.06 4.8% 5.7% 2.57

2.41 4.0% 3.80 6.7% 3.67 6.4% 0.96 2.4% 1.90 3.3% 2.39 4.0% 2.97 5.0% 4.5% 2.59

1.56 9.0% 1.30 9.0% 1.80 9.5% 1.22 8.9% 0.48 7.0% 0.95 8.7% 0.93 8.7% 8.7% 1.18

EM & CEO Pay

1/1/2010

34

Table 4 Significance of the Pay-Accounting Performance Relation Across Quintiles Representing the Degree Earnings Smoothness: BP(ICP)it =  + 1 (%Performance)it + 2 (Size)it + it Where:

(1)

BPit = Bonus Percent defined as annual cash bonus divided by total cash compensation for CEO of firm i in year t. ICPit = Incentive Compensation Percent defined as total compensation less base salary divided by total compensation for CEO of firm i in year t. . % Performance it = % change in performance measure over prior year divided by prior year's measure for firm i in year t, with sample trimmed by top and bottom 1% for each performance measure. Size it = Total assets for firm i in year t. Panel A: Cash Bonus as a Percent of Cash Compensation

Ranked R2 From Smoothing Model 3
Accounting Performance Measure
Operating Income (#170-17-190) Statistics From Model 1 Least Smooth Earnings Series Quintile #2 Quintile #3 Quintile #4 Smoothest Earnings Series

β1 t-stat R2 Income before X/O (#18) β1 t-stat R2 Net Income (#172) β1 t-stat R2 OI/A (#178/ #6) β1 t-stat R2 OI/A (#170-17-190/ #6) β1 t-stat R2 IBXO / A (#18 / #6) β1 t-stat R2 IBXO / CE (#18 / #60) β1 t-stat R2 2 Average R for Quintile Average t-Stat on β1 for Quintile Average t-Stat on 3 Earnings variables

3.40 8.4% 5.96 13.3% 3.05 5.8% 2.55 5.3% 1.76 6.9% 1.39 7.9% 1.89 6.4% 7.7% 2.86 4.14

3.30 6.7% 2.20 6.9% 3.40 5.7% 2.55 2.5% 0.92 4.6% 2.14 5.0% 2.61 4.3% 5.1% 2.45 2.97

2.78 5.25 6.3% 13.3% 3.00 0.55 6.8% 10.0% 3.19 3.29 10.7% 9.8% 3.00 1.82 8.0% 10.1% 3.36 -0.29 9.3% -0.6% 3.86 3.49 11.6% 9.9% 1.47 1.98 7.0% 6.4% 8.5% 8.6% 2.95 2.30 2.99 3.03

3.71 7.6% 5.43 12.6% 3.45 8.5% 0.31 7.3% 3.69 10.6% 1.73 2.4% 3.05 6.7% 8.0% 3.05 4.20

EM & CEO Pay

1/1/2010

35

Panel B: Total Compensation less Salary as a Percentage of Total Compensation

Ranked R2 From Smoothing Model 3
Accounting Performance Measure
Operating Income (#170-17-190) Statistics From Model 1 Least Smooth Earnings Series Quintile #2 Quintile #3 Quintile #4 Smoothest Earnings Series

β1 t-stat R2 Income before X/O (#18) β1 t-stat R2 Net Income (#172) β1 t-stat R2 OI/A (#178/ #6) β1 t-stat R2 OI/A (#170-17-190/ #6) β1 t-stat R2 IBXO / A (#18 / #6) β1 t-stat R2 IBXO / CE (#18 / #60) β1 t-stat R2 Average R2 for Quintile Average t-Stat on β1 for Quintile Average t-Stat on 3 Earnings variables

3.52 9.6% 5.30 11.9% 3.90 12.1% 2.19 5.9% 2.60 8.2% 2.07 7.9% 2.47 8.8% 9.2% 3.15 4.24

2.59 6.2% 1.52 9.7% 2.57 5.6% 3.30 5.5% 2.01 6.5% 2.90 6.8% 2.98 5.9% 6.6% 2.55 2.23

3.11 7.0% 3.81 9.0% 3.37 10.9% 1.29 7.3% 1.30 3.1% 2.38 10.1% 1.35 6.2% 7.7% 2.37 3.43

3.66 8.1% 2.97 8.9% 3.68 7.9% 1.45 7.0% -0.83 1.1% 2.88 6.9% 2.41 5.0% 6.4% 2.31 3.44

0.38 4.2% 2.81 8.0% 1.33 6.4% -1.96 5.4% 1.19 7.6% 1.85 3.1% 2.97 6.6% 5.9% 1.22 1.51

EM & CEO Pay

1/1/2010

36