Accounting Information and CEO Compensation: The Role of Cash Flow from Operations in the Presence of Earnings*

Transcription

1 See discussions, stats, and author profiles for this publication at: Accounting Information and CEO Compensation: The Role of Cash Flow from Operations in the Presence of Earnings* ARTICLE in CONTEMPORARY ACCOUNTING RESEARCH FEBRUARY 2006 Impact Factor: 1.43 DOI: /BUQJ-8KUQ-X2TF-K7T4 CITATIONS 24 READS AUTHORS, INCLUDING: Simon S.M. Yang Adelphi University 4 PUBLICATIONS 62 CITATIONS SEE PROFILE Available from: Simon S.M. Yang Retrieved on: 08 April 2016

2 Accounting Information and CEO Compensation: The Role of Cash Flow from Operations in the Presence of Earnings EMEKA T. NWAEZE, Rutgers University SIMON S.M. YANG, Adelphi University Q. JENNIFER YIN, University of Texas at San Antonio Accepted by Michel Magnan. We gratefully acknowledge insightful comments and suggestions by the associate editor, Michel Magnan, and two anonymous reviewers. We also benefited from comments by Steve Balsam, John Broussard, Jongsoo Han, Michalis Ioannides, Izzet Kenis, Sungsoo Kim, Sung Kwon, Ayalew Lulseged, and workshop participants at the School of Business, Rutgers University. We also thank John Core for discussing our paper at the 2002 American Accounting Association Conference.

3 Accounting Information and CEO Compensation: The Role of Cash Flow from Operations in the Presence of Earnings Abstract We examine the role of cash flow from operations (CFO) in CEO cash compensation. We test that CFO is contract-relevant in the presence of earnings, and more so when (1) the quality of earnings relative to the quality of CFO as a measure of performance is low and (2) the need for CFO as a financing source is high. Our analysis is motivated principally by normative arguments and anecdotes from financial disclosures linking CFO to managerial effort and contracts, notwithstanding the traditional role of earnings in performance measurement. We find that the weight of CFO in the compensation model is positive and significant in the presence of earnings and stock returns. We also find that the relative quality of CFO compared with that of earnings has a positive (negative) impact on the weight of CFO (earnings). We further find that the relative weight of CFO is enhanced substantially when enterprise activities crucially depend on internally-generated cash flow. These findings are unaltered when we include CEO age, firm size, and risk in the model and allow the coefficients to vary across industries. Key words: compensation, cash flow from operations, earnings JEL classification: J3, L2, M41 1

4 Accounting Information and CEO Compensation: The Role of Cash Flow from Operations in the Presence of Earnings 1. Introduction The use of accounting performance measures in management contracts has a long history (see Pavlik, Scott, and Tiessen 1993 for a survey of the literature). An accounting performance measure is predicted to be contract-relevant when it is incrementally informative about managers stewardship or when its use in a contract facilitates efficient risk-sharing between contracting parties (Holmstrom 1979; Gjesdal 1981; Lambert and Larcker 1987; Banker and Datar 1989). A corollary formally described in several studies is that multiple performance measures facilitate a more efficient contract when each measure is incrementally informative about the desired outcome (e.g., Banker and Datar 1989). Notably, earnings and cash flow from operations (CFO) are the most recognized accounting indicators of financial performance. On the one hand, the literature is replete with studies that show the relevance of earnings in contracts. 1 On the other hand, evidence on the contracting role of CFO is sparse and weak. For example, Kumar, Ghicas, and Patena (1993) test the impact of CFO on compensation and find that, with earnings in the model, the weight of CFO is not significant. Similarly, Natarajan (1996) regresses cash compensation on earnings and CFO measures and finds that earnings and CFO measures have a better association with cash compensation than earnings alone, but the weight of CFO is low and insignificant. The paucity of a consistent body of evidence on the relevance of CFO in contracts is puzzling given several press reports of the increasing use of CFO in compensation contracts. Perry and Zenner (2001), for example, survey a random sample of 100 S&P 500 and 100 Midcap 400 firms and find that 15 percent of the firms use cash flows as a performance measure. Leon (2004) notes that firms are increasingly using CFO-based metrics besides income-statement metrics to measure and reward performance. 2 The trend is also reflected on the incentive plans of several companies that explicitly identify CFO measures in their reward systems (e.g., AT&T, Disney, Duke Energy, General Electric, IBM, Motorola, and Nortel Networks). The Motorola incentive plan (the Motorola 2004 Proxy Statement), for instance, states in part: 2

5 While most employees are rewarded based on sector performance, high-level elected officers have a significant portion of their award based on the [operating earnings] OE and cash flow of the entire Company. The incentive plans of other noted companies similarly justify the link between CFO and executive reward on the logic that CFO is a key driver of success and shareholder return. The Wall Street Journal (1990) reports that when designing bonus formulas, more firms use CFO to supplement profit-based schemes. Emphasis on CFO as a relevant performance measure may also reflect a growing perception that earnings are increasingly noisy as a performance indicator. 3 The financial press often expresses the view that CFO is less susceptible to manipulation than other components of earnings and thus can be expected to be priced in the managerial labor market (Institutional Investor, August 1998, 55). Altogether, the anecdotes in surveys, press reports, and proxy statements suggest general and specific contexts that affect the role of CFO in contracts and provide a compelling motivation for a rigorous analysis of the impact of CFO on contracts. In this study, we describe general and specific contexts that affect the relevance of CFO in contracts, and we test the empirical validity. We predict that CFO is, a priori, contract-relevant in the presence of earnings and more so when (1) the quality of earnings relative to the quality of CFO as a measure of performance is low and (2) the need for CFO as a financing source is high. The prediction reflects the view that CFO has stewardship information that is non-overlapping with what is in earnings. More generally, CFO is a component of earnings that reflects and influences different aspects of CEO actions (e.g., Clinch and Magliolo 1993). Compensation committees are thus expected to use it in contracts to the extent that it has stewardship information beyond what is in earnings. The relative quality hypothesis predicts that CFO quality relative to earnings quality has a positive (negative) impact on the weight of CFO (earnings) in CEO cash compensation, where quality, in a general sense, refers to the extent the signal reflects managers stewardship. The prediction draws from the principal-agent models, which suggest that, when two signals are used in contracts, the relative weight of each signal reflects its precision and the properties of the other signal (e.g., Banker and Datar 1989; Feltham and Xie 1994; Natarajan 1996). Hence, if earnings and CFO both provide information about 3

6 managers actions, the weight of CFO will increase as the quality of earnings relative to the quality of CFO decreases, and vice versa. TThe CFO need hypothesis predicts that CFO need is positively related to the relative weight of CFO in CEO compensation. CFO need in this context reflects a condition in which CFO is crucial for enterprise activities. The prediction is based on anecdotal and research evidence that shows a close link between CFO and corporate activities and the limited ability of earnings to capture CEO performance in certain contexts (Myers 1977; Myers and Majluf 1984; Smith and Watts 1992; Skinner 1993). Several CFO-based incentive plans, in fact, cite the link between corporate activities and CFO as the major impetus for the plans. 4 Based on such evidence, we identify investment opportunity set (IOS) and external financing constraints as two major indicators of CFO need that affect the relevance of CFO in contracts. As we argue later, the two indicators elevate the contract-relevance of CFO in the presence of earnings because the managerial actions directed at exploiting existing investment opportunities or mitigating external financing constraints are highly contingent on CFO but are not captured by earnings. We regress changes in CEO cash compensation on earnings changes and on CFO changes separately and find that earnings and CFO are reliably related to CEO cash compensation, although the earnings-based model has higher explanatory power. When we include both variables in the model, we find that the weight of CFO continues to be reliably different from zero. This finding corroborates the evidence in surveys and the popular press that CFO plays a significant role in determining CEO compensation. Using several stylized proxies of CFO and earnings quality measures, we find strong evidence that CFO quality relative to earnings quality has a positive (negative) impact on the weight of CFO (earnings). Next, we test the impact of CFO need on the weight of CFO using proxy variables that are associated with the degree of CFO need, including IOS and external financing constraints. In the compensation model that includes both earnings and CFO, IOS has a positive (negative) impact on the weight of CFO (earnings); these results are robust across alternative proxies of IOS. Similarly, several stylized measures of external financing constraints have a positive (negative) impact on the weight of CFO (earnings). All the main 4

7 findings are unaltered after we include CEO age, firm size, and risk in the models and allow the coefficients to vary across industries. We interpret the results of our analyses as indicating that compensation committees use CFO-based performance measures to supplement earnings-based measures, especially when earnings quality is low or when CFO availability is a crucial determinant of enterprise activities, or both. In this context, we contribute to the literature by shedding light on contexts in which CFO is useful in determining CEO compensation in the presence of accounting earnings. The empirical methods we employ extend the stewardship-value concept described by Natarajan (1996) to specific contexts, and provide a simple framework for assessing the relative contracting quality of a pair of accounting performance measures. The remainder of the study is organized as follows: Section 2 motivates the relevance of CFO in contracts. Section 3 discusses several determinants of the weight of CFO in CEO compensation. Section 4 describes the empirical methods and sample selection. Section 5 presents the empirical results. Section 6 concludes the study. 2. Is CFO contract-relevant? CFO will play a role in a performance scheme that already includes earnings only if it conveys additional information about managers actions (Holmstrom 1979). Given the preponderance of earnings in contracts, the issue is whether CFO contains additional information for contracting. A priori, there are several reasons CFO can be expected to be contract-relevant in the presence of earnings. As a cash component of earnings, CFO is a crucial financing source, and can be expected to play a significant role in managerial contacts. This point is reflected in one of the predictions of agency theory: CFO availability creates value by giving managers discretion over investments decisions and by allowing a firm to avoid the agency costs of external funding (Myers 1984; Myers and Majluf 1984). Empirical support for this prediction is widespread. Fazzari, Hubbard, and Petersen (1988) and Whited (1992), for example, show that CFO exerts a strong influence over corporate financing and investment decisions: Firms must rely on CFO to repay debt, pay dividends, fund operations, and finance investments (especially when external funds are costly). 5

8 Moreover, CFO is a major component of earnings for which detailed reports are required and are costlessly available at the end of every accounting period. Thus, rational shareholders can be expected to consider it in performance evaluation because finer information is preferable to coarser information that is equally costly (Natarajan 1996). Earnings are also easier to manage than CFO and are less informative than CFO about resources available for enterprise activities. Analysts often cite this feature as the basis for preferring CFO over earning. As stated in Chemical Week (May 8, 1991, 28), Many financial analysts regard operating cash flow as a better gauge of corporate financial performance than net income, since it is less subject to distortion from differing accounting practices. To capture aspects of managers effort that are reflected in current compensation but are missing or poorly reflected in current earnings, several prior studies include security returns in the compensation model (e.g., Clinch 1991; Lambert and Larcker 1987). The inclusion of a stock price measure, however, does not purge CFO of contract-relevant information. Stock prices are susceptible to vagaries of the financial markets that are unrelated to managers actions. For example, movements in stock price may reflect world-wide events over which managers have no control (e.g., changes in the world oil market). CFO, by contrast, is the result of concerted credit policy and of overall operating decisions. Bushman, Indjejikian, and Smith (1996) observe that the stock price itself may not adequately reflect all valuable contracting information, especially when proprietary information about managers efforts is directed at increasing firm value in the long run. Consequently, firms may find it optimal to incorporate a CFO measure in its compensation plans (e.g., see the 2003 Proxy Statement for Nortel Networks, Inc.). CFO is also contract-relevant in contexts where its availability is critical to enterprise success. For example, among growth firms, highly leveraged firms, and distressed firms, the need for internal funding and the agency costs of raising funds externally elevate the importance of CFO in performance evaluation (New York Times, February 25, 1990, 3:29). Agency theory explains that growth firms are less likely to rely on earnings-based bonus plans because accounting profits are less informative about managers actions when growth is a large portion of a firm s value (Lambert and Larcker 1987; Smith and Watts 1992; Gaver and Gaver 1993; Skinner 1993; Baber, Janakiraman, and Kang 1996). For such firms, CFO is needed to 6

9 support profitable growth. For highly leveraged and distressed firms, incentive plans can be expected to focus on outcome measures, such as CFO, that relax external financing constraints (Whited 1992) and alleviate financial distress (Ittner, Larcker, and Rajan 1997). 3. Determinants of the weight of CFO and earnings in compensation The principal-agent models show that performance evaluation based on multiple signals facilitates more efficient contracts (e.g., Holmstrom 1979; Banker and Datar 1989); the contract weight of each signal reflects the precision of the signal and the properties of the other signals with respect to managers stewardship. 5 Already, there is evidence that the weight of earnings in compensation is related to earnings persistence (Baber, Kang, and Kumar 1998), earnings variability (Banker and Datar 1989), size of transitory earnings (Gaver and Gaver 1998), and IOS (Natarajan 1996). In this study, we focus on general and contextual properties of CFO and earnings that are expected to affect their relative stewardship information and thus their relative weight in compensation. We describe their relative stewardship information in the context of two broad factors: the CFO-earnings relative quality and CFO need. CFO-earnings relative quality The CFO-earnings relative quality hypothesis predicts that CFO quality relative to earnings quality has a positive (negative) impact on the weight of CFO (earnings) in CEO cash pay. CFO quality or earnings quality, in this context, refers to the extent to which the signal is informative about managers stewardship. On the premise that CFO and earnings are both contract-relevant, we extrapolate from the studies by Holmstrom (1979), Lambert and Larcker (1987), Banker and Datar (1989), Sloan (1993), Feltham and Xie (1994), Natarajan (1996), Bushman, Indjejikian, and Smith (1996), Ittner, Larcker, and Rajan (1997) and Hayes and Schaefer (2000) and posit that, a decrease in earnings quality relative CFO quality reduces (increases) the weight of earnings (CFO), and vice versa. The principal-agent model identifies certain attributes of a performance measure that affect the contracting quality of the measure. Banker and Datar (1989), Feltham and Xie (1994), and Natarajan (1996), for example, show that pay-performance sensitivity is related to the precision with which a measure 7

10 reveals managers actions. Aggarwal and Samwick (1999) find that the variance of a performance measure is an important determinant of pay-performance sensitivity. Baber, Kang, and Kumar (1998, 1999) and Bushman, Engel, Milliron, and Smith (2000) suggest that certain indicators of earnings quality that have successfully explained variations in stock price metrics are also reflected in CEO compensation. We use four such indicators to construct CFO-earnings relative quality the ratios of (1) CFO persistence to earnings persistence, (2) earnings variance to CFO variance, (3) absolute size of earnings change to CFO change, and (4) total accruals to book value of equity and explore their impact on the weight of earnings and CFO in CEO compensation. CFO persistence relative to earnings persistence Existing research shows a positive link between earnings persistence and the compensation weight of earnings. Baber, Kang, and Kumar (1998, 1999), for example, show that the relation between executive pay and earnings varies positively with earnings persistence. Bushman, Engel, Milliron, and Smith (2000) also show that the link between executive pay and earnings is sensitive to the underlying value in earnings and to the determinants of the earnings response coefficient, including earnings persistence. When earnings persistence is low, earnings are a less reliable index of performance and become less suitable for determining incentive awards. In such contexts, the relative importance of CFO in contracting will depend on its ability to credibly fill or narrow the information gap created by low earnings persistence. As noted earlier, CFO is contract-relevant a priori, because it is a measure of operating results available for funding enterprise activities and it is less vulnerable to the effects of managers discretionary reporting actions. As Revell (2001) remarks, Cash flows are the closest you can get to economic reality. Its importance in contracts increases in its persistence relative to earnings. Earnings variability relative to CFO variability Earnings (CFO) variability reflects the volatility of earnings (CFO) over time and is generally operationalized as the time-series variance of earnings (CFO) (e.g., Lipe (1990). Banker and Datar (1989) and Natarajan (1996) show that, for contracts based on two signals, the weight of each signal is directly 8

11 proportional to the precision of the signal. Holmstrom and Milgrom (1987) similarly contend that payperformance sensitivity is decreasing in the volatility of the performance measure in that a volatile index is a noisy indicator of underlying performance. On this premise, we expect CFO to play an increased role in compensation when earnings variability is high relative to CFO variability. Earnings change relative to CFO change Extant research shows that extreme innovations in earnings are less informative about performance (Freeman and Tse 1992; Das and Lev 1994; Subramanyam 1996) and that earnings quality is decreasing in the absolute size of earnings innovation (Cheng, Liu, and Schaefer 1996). Large innovations in earnings or CFO contain relatively more transitory items (Cheng, Liu, and Schaefer 1996; Ali 1994). Subramanyam (1996) notes that the capital market, on average, associates extreme signals with low precision and weights such signals accordingly. If compensation committees similarly view extreme earnings changes as being less informative about performance, extreme earnings innovations will receive lower weight in compensation. By contrast, CFO change is informative about operating results available for funding enterprise activities. Moreover, CFO performance is often used to assess earnings quality. In particular, analysts routinely compare earnings innovation with CFO innovation to assess the extent that reported earnings match tangible operating results. A close match suggests a credible earnings report; a mismatch diminishes the reliability of earnings and elevates the importance of CFO in performance evaluation. We thus predict that CFO will play an increased role in incentive contracts when the absolute change in earnings relative to change in CFO is large. Accruals to book value Total accrual-to-book value ratio (AC/BV) is the accrual portion of accounting return on equity, which reflects the extent earnings diverge from CFO (noting that AC/BV (earnings CFO) / BV). The larger the ratio in absolute terms, the larger is the gap between reported earnings and CFO. The literature on earnings quality generally predicts that earnings quality is decreasing (increasing) in the accrual (CFO) component of earnings (see Harris, Huh, and Fairfield 2000; Raj, Hawkins, Bernstein, and Redlich 2002). 9

12 This relation occurs because large accruals generally arise from transitory and non-recurring items (e.g., special charges, managed accruals, etc.). Gaver and Gaver (1998) present evidence that transitory items reduce the contracting quality of earnings. In fact, it is typical for compensation plans to include provisions that authorize the compensation committee to revise the bonus criteria when earnings are unduly influenced by special or non-recurring items. On this general premise, we expect the relative contracting role of CFO to be positively related to the accrual component of earnings. CFO need The CFO need argument focuses on the contracting relevance of CFO in the presence of earnings when CFO availability is crucial to enterprise activities. We use two broad economic constructs IOS and external financing constraints as indicators of the sensitivity of enterprise activities to CFO availability. There is evidence in the literature that both factors often compel managers to favor CFO over external funds (see Donaldson 1961; Pinegar and Wilbrich 1989). We contend next that both factors elevate the relevance of CFO in contracts in that they typify conditions in which (1) CFO shortfall restricts financing and investment actions and (2) the value-enhancing actions of managers that depend crucially on CFO are imperfectly reflected in earnings. High IOS High-IOS firms expand rapidly and require large cash flow to fund growth. Several studies further show that high IOS firms pay high cash compensation (Smith and Watts 1992; Gaver and Gaver 1993), which adds to their cash needs. However, there is evidence that such firms are less likely to use external funding for their cash needs. One argument is that external funding costs are prohibitive and create a disincentive for valuable investments (Myers 1977). New debt, for example, creates restrictions on further investments; new equity may be under-priced because of information asymmetry (see Myers and Majluf 1984; Jensen 1988; Stulz 1990). The related agency costs can cause managers to shelve lucrative projects when cash is not available internally (Lamont 1997). By contrast, CFO is valuable to growth firms not only because it is a source of low-cost funds for new projects, but also because it mitigates the risks inherent in using 10

13 external funding and gives managers discretion in selecting projects (see Myers 1984; Myers and Majluf 1984). 6 For this reason, growth firms have a strong impetus to emphasize CFO performance. In addition, Smith and Watts (1992), Gaver and Gaver (1993), Skinner (1993), and Baber, Janakiraman, and Kang (1996) argue that high IOS firms are less likely to rely on earnings-based incentive plans. They point out that accounting profits are less informative about managers actions when IOS is a large portion of a firm s value. For example, growth in capital expenditures that enhance firm value leads to increased depreciation which in turn reduces current earnings. 7 Reported profits in such contexts omit aspects of managers performance directed at valuable investments and poorly reflect managers stewardship (Lambert and Larcker 1987). In concert with this view, Natarajan (1996) finds that IOS is negatively related to the weight of earnings in compensation. By contrast, high IOS firms are likely to elevate the relative contracting role of CFO because the investment and financing actions of managers, which are poorly reflected in earnings, are highly contingent on and correlated with CFO availability (Vogt 1997; Shyam-Sunder and Myers 1999; Minton and Schrand 1999; Opler, Pinkowitz, Stulz, and Williamson 1999). 8 High-IOS firms may also adopt a CFO-based incentive plan jointly with, or in lieu of earningsbased plans to reduce the risk to managers of earnings fluctuations caused by managers valuable investment actions. 9 Moreover, CFO is less vulnerable than earnings to the noise inherent in accrual accounting practices. External financing constraints External financing constraints, CON, refer to the difficulties firms face in obtaining external capital at reasonable costs. Difficulties in obtaining external financing are predicted to elevate the role of CFO in incentive compensation on the logic that constrained access to external capital compels affected firms to emphasize CFO for their strategic and operational cash needs (Donaldson 1961, Pinegar and Wilbrich 1989; Ittner, Larcker, and Rajan 1997). Although a firm s ability to access external capital depends on its specific circumstances, certain economic factors are widely known to raise the cost of obtaining external capital. We focus on two such factors: High leverage and financial distress. In addition, we identify 11

14 negative free cash flow as an outcome measure that exacerbates the constraints among highly-leveraged and financially-distressed firms. Leverage: Firms that have debt require cash flow to service/repay their debt or fund new debt. Highly leveraged firms favor CFO in lieu of other funding sources, primarily because of the adverseselection costs of high debt noted by Myers and Majluf (1984). High debt is typically associated with restrictive covenants that make it harder for firms to raise external funds at reasonable costs (Myers 1977), compels firms to delay or forgo lucrative projects when funds are not available internally (Myers and Majluf 1984; Jensen 1988; Sch), and puts pressure on managers to generate cash internally to meet debt repayments (Baker and Wruck 1989; Murphy 2001). Not surprisingly, leveraged firms stress CFO as a device for relaxing the constraints on firm activities imposed by debt (Fazzari, Hubbard, and Petersen 1988) and as an important performance measure (Baker and Wruck 1989). By contrast, earnings play a limited role in reducing external financing constraints, mainly because they are an imperfect measure of fund availability. This is especially true for highly leveraged firms whose debt overhang gives creditors an incentive to stipulate stringent liquidity requirements (as opposed to profit goals) for new debt. This further implies a diminished role for earnings in the incentive contracts of high-leverage firms. 10 Press reports also note that highly leveraged firms stress CFO in lieu of earnings in incentive plans (e.g., The Wall Street Journal 1990), mainly because CFO is a more reliable indicator of liquidity. Thus, we expect leverage to have a positive (negative) effect on the weight of CFO (earnings) in incentive pay plans. Financial distress: A distressed firm is characterized by deteriorating financial performance and increased probability of bankruptcy. Firms in distress typically take actions intended to bring them out of distress, including restructuring and redeployment of resources (John, Lang, and Netter 1992; Blackwell, Marr, and Spivey 1990; Lang, Poulsen, and Stulz 1995). Gilson and Vetsuypens (1993) find that compensation policy is an important strategy for dealing with financial distress and that earnings performance poorly explains the variation in CEO cash compensation for distressed firms. Ittner, Larcker, and Rajan (1997) observe that firms in distress often tie CEO bonuses to short-term achievements, such as increases in CFO, that are related to bringing them out of distress. Whited (1992) finds also that distressed 12

15 firms rely on CFO in lieu of external funds to support their operations. Their reliance on CFO funding is expected because they face increased agency costs of external capital. Meanwhile, earnings of firms in distress are unlikely to reflect the managers actions directed at bringing the firms out of distress. To the extent that actions directed at combating distress depend on CFO availability and less on earnings, we expect financial distress to have a positive (negative) impact on the weight of CFO (earnings) in CEO compensation. Negative Free Cash Flow: Free cash flow (FCF) is the residual CFO beyond the amount needed for strategic investments; it is the CFO available to the shareholders after a firm pays for expenses and invests in capacity maintenance/growth. Ordinarily, FCF affects a firm s capacity to service debt, exploit new profit opportunities, pay cash dividends, etc. Positive FCF gives managers discretion over financing, investment, and dividend decisions. Negative FCF has the reverse consequences: It indicates a CFO deficiency and portends future constraints on debt repayment, new debt financing, strategic investments, and/or dividend payment. It raises liquidity concerns and credit problems particularly for poor performers and highly leveraged firms that have poor access to external capital a priori. To the extent that an observed negative FCF is associated with systemic or worsening financial conditions, external financing will be more difficult and lead to grater emphasis on CFO performance Empirical methods and data definition Earnings and CFO in CEO compensation We relate changes in earnings and CFO to change in cash compensation defined as the change in salary and bonus. As in several prior studies, we use changes in cash compensation as the dependent variable rather than compensation levels or changes in total compensation to (1) control for certain factors that influence compensation levels, but are unrelated to managers performance (e.g., firm size, CEO tenure, CEO age, etc.) and (2) focus on executive pay-performance sensitivities that are related to innovations in earnings and CFO performances (e.g., Lambert and Larcker 1987; Baber, Janakiraman, and Kang 1996; Baber, Kang, and Kumar 1998, 1999). 12 We also include security returns as a proxy for 13

16 market-based performance used in CEO compensation (Lambert and Larcker 1987; Jensen and Murphy 1990; Baber, Kang, and Kumar 1998, 1999). Sample and data definition Financial data are obtained from Compustat and CEO compensation data are obtained from ExecuComp from 1992 (the first year CEO compensation data are available in the database) to Firmyears in which CEOs changed are deleted to mitigate potential biases due to partial-year compensation not specified in the model. Panel A of Table 1 describes the data screening steps and the final data used in the analyses. The initial sample consists of 15,314 firm-years and 2,474 firms obtained from ExecuComp. We exclude observations with missing or invalid values of CFO change, earnings change, stock return, or book value of equity. Missing values of salaries or changes in cash compensation are also excluded. We also delete the top and bottom 1 percent of each variable to reduce the influence of outliers on results. Because the models require changes in earnings and CFO, 1993 is the initial year for which usable data are available. The final sample consists of 9,887 firm-years and 2,127 firms. Panel B reports the first and third quartiles, means, medians, and standard deviations of various attributes of the sample firms. The mean CEO cash compensation is $1.1 million. On average, cash compensation increases by $87,022 over the prior year, equivalent to a 16.7 percent increase over the prior year s salary. [Table 1 about here] 5. Regression results Earnings and CFO in CEO compensation Initially, we estimate the following three models to assess the separate and incremental roles of earnings and CFO in determining changes in CEO compensation: COMPB itb = αb B+ αb 1B EB itb + λretb it + B εb 1t B (1) COMPB itb = αb B+ αb 2 B CFOB itb + λretb itb + εb 2t B (2) COMPB itb = αb B+ αb 1B EB itb + αb 2B CFOB itb + λretb itb + εb 3t B (3) where, for firm i in year t: 14

17 COMP = change in total cash compensation (salary + bonus) deflated by lagged salary; E = change in earnings deflated by beginning-period book value of equity; CFO = change in CFO, deflated by beginning-period book equity; 13 RET = raw returns; and ε = error term. We deflate the dependent variable by the lagged salary, to align the metric with how the rewards are reported by companies; 14 we deflate the independent variables by the beginning-period book equity, to reduce heteroscedasticity. In (1) and (3), αb 1B assesses the weight of earnings change. In (2), αb 2 assess the weight of CFO change in the absence of earnings. In (3), we test the impact of CFO change on compensation in the presence of earnings and stock returns. To reduce the bias in parameters caused by pooling data across time, we also estimate each model annually (1993 to 2001). We then average the coefficients for each variable and obtain the t-value as the ratio of the mean coefficient to the standard error of the coefficients. The results are presented in Table 2. [Table 2 about here] As expected, earnings change has a positive and significant impact on CEO cash compensation. The adjusted RP 2 P for (1) that excludes CFO change is 12.7 percent. Results for (2) and (3) show that the coefficient for CFO change, αb 2,B is also positive and significant (columns 3 through 6). The adjusted RP 2 P for (2) that excludes earnings change is 9.5 percent for the pooled results and 9.9 percent for the results based on annual regressions. The results indicate that earnings and CFO are each associated with CEO cash compensation; The CFO model, however, has lower adjusted RP 2P. The results for (3) that includes both earnings and CFO show that the coefficient for earnings change is positive and significant and the coefficient for CFO change is also positive and significant, suggesting that CFO complements earnings in explaining changes in CEO compensation. The CFO results reported here are generally stronger than those reported by Kumar, Ghicas, and Pastena (1993, Table 4) and Natarajan (1996, Table 3); we report higher t-values and adjusted RP 2P s for the CFO models. The results, perhaps, reflect our use of more reliable CFO measures obtained directly from the statement of cash flows in lieu of estimates from working capital accounts in the balance sheet. 15

18 Estimates of CFO from the balance sheet are prone to errors related to the effects of mergers, acquisitions, divestitures, etc., on working-capital balances that do not affect the reported CFO. It is also possible that CFO from the statement of cash flows has higher information content than estimated CFO (Cheng, Liu, and Schaefer 1996). Furthermore, Natarajan (1996) uses compensation level rather than compensation change, increasing the possibility of an omitted variables problem. 15 CFO-earnings relative quality and the compensation weight of CFO and earnings In this section we test our prediction that, in the presence of earnings and stock returns, the CFO quality relative to earnings quality has a positive (negative) impact on the weight of CFO (earnings) in CEO compensation, where quality reflects the extent to which the measure is informative about managers stewardship. For the analysis we construct four indicators of CFO quality relative to earnings quality: (1) ratio of CFO persistence to earnings persistence, (2) ratio of earnings variability to CFO variability, (3) absolute value of the ratio of earnings change to CFO change, and (4) ratio of total accruals to book value of equity. 16 Earnings (CFO) persistence is estimated for each firm that has at least 13 years of consecutive data by regressing earnings before extraordinary items and discontinued operations (CFO) on their past values, based on the assumption that earnings (CFO) follow an IMA(1,1) process (e.g., Beaver 1970; Beaver, Lambert, and Morse 1980; Ali and Zarowin 1992; Ohlson and Schroff 1992). 17 Earnings (CFO) variability is estimated as the variance of the most recent five years of earnings (CFO), scaled by the mean of the earnings (CFO). Each firm-year ratio of total accruals to book value is adjusted by the industry median ratio to mitigate the effects of industry-related variations in accruals. Moreover, the ratio is usable only if the book value of equity is positive. We then combine the four indexes in a factor analysis to obtain a composite index of the relative quality, RQfac. 18 Panel A of Table 3 presents the correlations among the indicators and the factor loading of each indicator (i.e., the correlation between the indicator and the common factor). The indicators are positively correlated. The correlations range from to The persistence ratio has the highest factor loading of 0.731, followed by the ratio of earnings change to CFO change, with a factor loading of The 16

19 variance ratio has a loading of The ratio of total accruals to book value has the smallest loading of The total variance explained is percent. [Table 3 about here] To examine the impact of the CFO-earnings relative quality, RQ, on the weight of CFO and earnings in CEO compensation, we estimate the following regression model: COMPB itb = αb 0B + αb 1B EB it B +αb 2B CFOB it B +αb 3B RQB itb EB itb +αb 4B RQB itb CFOB itb +αb 5 B RETB itb +αb 6B RQB itb + εb it B (4) The coefficient αb 3B estimates the effect of RQ on the weight of earnings. We expect the coefficient to be negative on the premise that low earnings quality diminishes the role of earnings in CEO compensation. The coefficient αb 4B tests our prediction that low earnings quality relative to CFO quality improves the weight of CFO in CEO compensation. The coefficient is expected to be positive. 19 As in the preceding sections, (4) is estimated by pooling the data and annually from 1993 to In Panel B of Table 3 we first present results in which the ratio of CFO persistence to earnings persistence is used as a proxy for CFO-earnings relative quality. This aspect of the analysis is motivated by the wide use of earnings persistence as an indicator of quality (we note also that the persistence measure has the highest loading on the RQfac). Columns 1 and 2 show the results based on relative persistence. The coefficient αb 3B is negative and significant, both in the pooled results (αb 3B = 0.075, t= 2.27) and in the results based on annual regressions (αb 3 B = 0.079, t= 3.02); αb 4B is positive and significant at the 0.10 level in the pooled results (αb 4B =0.047, t=1.57) and in the results based on annual regressions (αb 4B =0.046, t=1.45). These results are consistent with our prediction that, in the presence of earnings, the impact of CFO (earnings) in CEO compensation will increase (decrease) as the quality of earnings relative to the quality of CFO decreases. The next set of results in Table 3 examines the impact of the composite relative quality index, RQfac, on the weight of earnings and CFO. The results parallel those based on the relative persistence measure except that the absolute magnitudes of αb 3 B and αb 4B are larger. The coefficient, αb 3B is in the pooled result and in the results based on annual regressions. The t-ratio is significant at the

20 level in each case. The coefficient, αb 4B, is in the pooled results and in the results based on annual regressions. The coefficient is significant at the 0.05 level for the pooled results although it is not significant in the results based on annual regressions. Consistent with our predictions, these results suggest that CFO quality relative to earnings quality has a positive (negative) impact on the compensation weight of CFO (earnings). In an additional test, we describe the relative contracting quality of CFO to earnings in terms of the relative stewardship values developed by Natarajan (1996). Specifically, we estimate the relative stewardship value of CFO to earnings as: RSV = [SV(CFO) SV(E)]/ SV(CFO). The stewardship value of CFO, SV(CFO), is the product of the mean and precision of CFO, adjusted for the information shared with earnings; the stewardship value of earnings, SV(E), is interpreted analogously (see the bottom of Table 4). The RSV defined here is a modification of the measure in Natarajan (1996), intended to mitigate cases in which negative stewardship values confound interpretation of RSV variable. Panel A of Table 4 presents the descriptive statistics for SV(E), SV(CFO), and RSV. The mean stewardship value of earnings (CFO) is (0.076). The mean relative stewardship value is [Table 4 about here] To test the effect of relative stewardship value on the compensation weight of CFO and earnings, we replace RQ in (4) with RSV. We expect the weight of CFO (earnings) to be increasing (decreasing) in the relative stewardship value of CFO. Hence, αb 3B in the revised model is expected to be negative, whereas αb 4B is expected to be positive. Panel B of Table 4 presents the results based on the pooled sample and annual regressions. Across both results, αb 3B is negative and significant (pooled results: αb 3B equals 0.091; results based on annual regressions: αb 3B = 0.101). By contrast, αb 4B is positive and significant across both results (pooled results: αb 4B = 0.028; results based on annual regressions: αb 4B = 0.025). Thus, consistent with the predictions of agency theory, higher stewardship measure of CFO relative to that of earnings increases (decreases) the compensation weight of CFO (earnings)

21 CFO need and the compensation weight of CFO and earnings In this section we test the hypothesis that IOS and external financing constraints elevate the contracting relevance of CFO on the premise that both factors are associated with CFO need. For the empirical tests, we examine separately how each factor affects the weight of CFO and earnings in compensation. We favor separate tests of the effects of each factor because each factor poses a different set of economic (and stewardship) challenges that are likely to create differences in the manner CFO is used in contracts across the contexts. 21 IOS indicator of CFO need We first investigate the impact of IOS on the weight of CFO and earnings in CEO cash compensation. Our prediction is that IOS is positively (negatively) related to the weight of CFO (earnings) in CEO cash compensation. To test the prediction, we identify a set of alternative proxies of IOS, including market-to-book ratio, investment intensity, geometric mean of annual growth rate of market value of total assets, and R&D-to-asset ratio (see, Baber, Janakiraman, and Kang 1996; Baber, Kang, and Kumar 1999). Similar to the approach used by Baber, Janakiraman, and Kang (1996), we use factor analysis to extract a composite measure of IOS denoted, IOSfac, from the proxies. As shown in panel A of Table 5, the proxies are positively correlated. The correlations range from for R&D-assets ratio and annual growth rate of the market value of assets to for investment intensity and R&D-assets ratio. The factor loadings are high for all four proxies and range from for R&D-to-asset ratio to for market-to-book ratio. Total variance explained by IOSfac is 52.1percent. [Table 5 about here] We estimate the effect of IOS on the weight of CFO and earnings in CEO cash compensation using the following model: COMPB itb = αb 0B +αb 1B EB it B +αb 2B CFOB itb +αb 3B IOSB itb EB itb +αb 4B IOSB itb CFOB itb +αb 5B RETB itb +αb 6B IOSB itb + εb it B (5) The coefficient, αb 3, B evaluates the effect of IOS on the weight of earnings; it is expected to be negative; αb 4B tests our prediction that, in the presence of earnings, IOS has a positive impact on the weight 19

22 of CFO. We estimate (5) using pooled data; we also estimate the model annually and use the results to estimate the mean of each coefficient and the corresponding t-value. In panel B of Table 5, we first present results in which we use the market-to-book ratio as a proxy for IOS because it is used widely as an indicator of IOS (we note also that the market-to-book ratio has the highest loading on IOSfac). For the market-to-book indicator of IOS, αb 3B is negative and highly significant in both the pooled results and the results based on annual regressions. This result is consistent with evidence in prior studies which shows that high IOS lowers the weight of earnings in CEO compensation. The coefficient αb 4B, which evaluates the effect of market-to-book ratio on the weight of CFO, is positive and significant in the pooled results and in the results based on annual regressions. This result supports our prediction that IOS has a positive impact on the weight of CFO. The results using the composite index of IOS, IOSfac, are presented in columns 4 and 5. The results parallel those based on the market-to-book value proxy. 22 With IOSfac in the model, αb 3B is negative and highly significant both in the pooled results and in the results based on annual regressions; by contrast, αb 4B is positive and significant in both results. Altogether, the results indicate that IOS has a negative (positive) impact on the weight of earnings (CFO) in CEO cash compensation. External financing constraint indicator of CFO need Next, we test the effect of external financing constraints on the weight of CFO and earnings. For the test, we construct an index of external financing constraints using leverage, financial distress, and FCF measures. Leverage (Lev) is the ratio of long-term debt to beginning-period book value of equity. To reduce cross-sectional differences in leverage caused by industry variations in capital structure patterns, we deduct the industry median leverage from each firm-year leverage. Distress is the Altman s Z-score of financial health (Altman 1968). A firm-year for which the score is less than 1.81 is classified as distressed and assigned a value of 1; else, the firm-year is classified as healthy and assigned a value of 0. As another proxy for distress, we include the five-year moving average ratio of earnings before extraordinary items to book value (E/B). A firm-year is classified as distressed and assigned a value of 1 if E/B is less than or equal to 0; otherwise the firm-year is classified as healthy and assigned a value of 0. E/B captures the 20

23 intuition that extreme low (high) earnings relative to book value is correlated with poor (good) financial health (Burgstahler and Dichev 1997; Nwaeze 2004). FCF is an indicator that takes the value of 1 if a firmyear free cash flow is negative and 0 otherwise. We then apply factor analysis to the variables to derive a composite score of external financing constraints, CONfac. As shown in Panel A of Table 6, the variables are positively correlated. The correlations range from for FCF and Lev to for Z-score and Lev. The variables load reasonably on CONfac, ranging from for E/B to for Z-score. The total variance explained by CONfac is 34.5 percent. [Table 6 about here] To test the impact of external financing constraints on the weight of CFO and earnings, we replace IOS with CON in (5). αb 3 and αb 4 B test the effect of external financing constraints on the weights of earnings and CFO, respectively; αb 3 is predicted to be negative, whereas αb 4 is predicted to be positive. We estimate the model using pooled data. We also obtain annual results from which we estimate the mean of each coefficient and the corresponding t-value. In panel B of Table 6, we present results in which we use leverage and Z-score proxies of external financing constraints (because both indicators are highly correlated with external financing constraints) and results based on the composite index. 23 The results in columns 1 and 2 show that leverage has a negative effect on the weight of earnings: αb 3B = for the pooled results and for the results based on annual regressions; the t-value is significant in each case. By contrast, leverage has a positive and significant effect on the weight of CFO: αb 4B = (P < 0.01) in the pooled results and (P < 0.01) in the results based on annual regressions. 24 The results based on the Z-score proxy (columns 3 and 4) are qualitatively similar to those based on the leverage proxy: The effect of Z-score on the weight of earnings is in the pooled results and (although insignificant) in the results based on annual regressions. On the other hand, its effect on the weight of CFO is positive and significant in both the pooled results and the results based on annual regressions. The results based on the composite index of external financing constraints, CONfac, (columns 5 and 6) are similar to the results based on leverage and Z-score proxies of external financing constraints. 21