Docstoc

CFO Fiduciary Responsibilities a

Document Sample
CFO Fiduciary Responsibilities a Powered By Docstoc
					        CFO Fiduciary Responsibilities and Annual Bonus Incentives




                                              RAFFI INDJEJIKIAN

                                              MICHAL MATĚJKA

                                            Ross School of Business*
                                             University of Michigan




                                                 November, 2008




*
    Contact information: 701 Tappan St., Ann Arbor, MI 48109. e-mail: raffii@umich.edu, matejka@umich.edu.
    We acknowledge the support of the AICPA in this research project. We also wish to thank numerous colleagues
    who provided us with feedback on the survey instrument used to collect data. The paper has benefited from
    comments and suggestions of workshop participants at the University of Chicago Graduate School of Business, the
    Wharton School of the University of Pennsylvania, the University of Michigan Harvey Kapnick Accounting
    Conference, Duke-UNC Accounting Fall Camp, the Rotman School of Management at the University of Toronto,
    the 2008 AAA MAS meeting, and the 2008 AAA annual meeting.
     CFO Fiduciary Responsibilities and Annual Bonus Incentives




                                            ABSTRACT

We examine how firms evaluate and compensate their chief financial officers (CFOs). CFOs

participate in decision making much like other executives, but unlike most other executives they

have fiduciary responsibilities for reporting firms’ financial results and safeguarding the integrity

of financial reporting. Responsibility for financial reporting raises the question of whether it is

appropriate to reward CFOs bonuses contingent on financial performance that is effectively self-

reported. In this paper, we provide a framework that characterizes CFO incentive bonuses as a

tradeoff between CFOs’ decision-making responsibilities and their fiduciary duties over financial

reporting. This framework yields a number of implications which we examine empirically using

a proprietary survey of CFO compensation practices of public and private firms. For instance, we

find that from 2003 to 2007 public entities (relative to private entities) reduced the percentage of

CFO bonuses contingent on financial performance. We interpret this result as evidence that firms

mitigate earnings management or other misreporting practices in part by deemphasizing CFO

incentive compensation.
1   Introduction

In response to a number of prominent corporate failures, recent regulations have called for

reforms aimed at strengthening financial reporting, executive compensation and other

governance practices. Executives most affected by these governance reforms are the chief

executive officer (CEO) and chief financial officer (CFO). Although much is written about the

role of CEOs and their compensation in corporate governance, there is very little research

examining the role of CFOs and their incentives. In this paper, we examine how firms evaluate

and compensate their CFOs.

       What sets CFOs apart from other executives are their significant responsibilities

safeguarding the integrity of financial reporting and other internal organizational controls, areas

most recently targeted for governance reform. Although other top executives have fiduciary

responsibilities for financial reporting as well, CFOs typically have more of an expertise and

capacity to determine what numbers get reported (Mian [2001], Geiger and North [2006]).

However, responsibility for financial reporting raises an important question: Is it is appropriate to

award CFOs incentive compensation based on performance that is effectively self-reported?

Indeed, according to a former IRS commissioner (testifying before the Senate Finance

Committee), CFOs, top corporate attorneys, and board chairs who are charged with ―minding the

cookie jar‖ should have little or no incentives and instead should get ―generous but fixed

compensation‖ (Katz [2006]). While the commissioner’s argument is a prima facie reasonable,

prior evidence as well as the survey findings we present in this paper show that firms routinely

award CFOs substantial incentive compensation tied to financial performance.

       To provide a framework to understand why firms award CFOs compensation tied

specifically to financial performance, we present an agency model that features two executives,



                                                 1
an operating manager (say a CEO) performing a productive task and a CFO entrusted with dual

responsibilities. We assume the CFO performs a productive task much like other executives but

is also charged with fiduciary duties to report financial performance as accurately as possible.

These fiduciary duties imply that CEO and CFO incentives are (positively) linked which is a

feature of our model corroborated by our survey evidence. Further, we find that a CFO’s

incentive compensation reflects the costs and benefits of his dual responsibilities with the key

insight that using reported financial performance to motivate CFOs to be productive necessarily

implies some tolerance for misreporting. 1 While our model has a number of other implications

that we subject to empirical scrutiny, we highlight below two results that have implications for

changes in CFO compensation practices in the post Sarbanes-Oxley (SOX) environment.

           The first result concerns how costs of misreporting borne personally by CFOs affect their

incentive compensation. We find that if CFOs personally bear greater misreporting costs, then

firms offer their CFOs steeper incentives tied to financial performance. The intuition is

straightforward; if CFOs are more conscientious in discharging their fiduciary duties, then firms

are more comfortable offering steeper incentives since rewards for reported performance are less

susceptible to unwarranted overpayments. The second result concerns how costs of misreporting

borne by firms (as opposed to CFOs) affect CFO incentive compensation. Here, we find that as

misreporting becomes more costly, firms are less willing to tolerate misreporting. Hence, firms

offer their CFOs weaker incentives tied to financial performance to expressly motivate them to

focus more on their fiduciary duties.




1
    In our model, we assume that firms rely on annual bonus incentives to alleviate agency conflicts with their
    executives. More generally, firms can rely on a variety of control mechanisms to alleviate agency conflicts
    including internal and external audits, job design and organizational structure.



                                                            2
        In the current post-SOX environment, the empirical implications of our results are as

follows. First, to the extent that SOX substantially increased misreporting penalties for

executives in charge of the financial reporting process or otherwise exogenously improved

financial reporting quality (e.g., Cohen, Dey, and Lys [2007], Zhang [2007]), we expect an

increase in the sensitivity of CFO bonuses to reported financial performance.2 At the same time,

if capital market participants are far more sensitive to firms’ accounting (mis)representations

and/or react negatively to disclosures of internal control weaknesses in the current post-SOX

environment (e.g., Ashbaugh-Skaife et al. [2007], Hammersley, Myers, and Shakespeare [2008]),

then we expect firms to endogenously motivate improvements in financial reporting quality via a

decrease in the sensitivity of CFO bonuses to reported financial performance.

        Given these two countervailing predictions (as well as confounding effects such as time

trends in executive compensation), attributing recent changes in CFO compensation to SOX or

other regulations is an empirical challenge. An ideal test would examine how CFO incentives

changed following SOX for firms complying with the legislation vis-à-vis firms unaffected by

the legislation. Unfortunately, public databases (e.g., Execucomp) used in prior studies typically

cover large public firms all of which had to comply with SOX. 3 In contrast, we rely on a

proprietary survey of CFO performance evaluation and compensation practices of public and

private firms (described below). Since public firms were affected by SOX much more than

private firms, the public versus private distinction allows for an identification strategy of the

SOX-effect on CFO compensation that has not been feasible in prior literature.


2
  There is some recent evidence suggesting that sensitivity of executive compensation to financial performance has
  increased in the post-SOX environment (e.g., Carter, Lynch, and Zechman [2007], Chen, Jeter, and Yang [2007]).
  We briefly survey this literature in section 2.1.
3
  Moreover, public databases rarely provide details concerning the proportion of incentive compensation linked to
  reported financial performance, the measure of interest in our study.



                                                        3
           Specifically, we use a database of CFO compensation practices obtained by surveying

about 30,000 members of the American Institute of Certified Public Accountants who are CFOs,

CEOs, or other executives informed about CFO and CEO compensation.4 In a sample of 1,353

respondents from both public and private entities, we find that annual bonuses are by far the

most common incentive component of CFO compensation plans and that, on average, about 50

percent of CFO bonuses is based on accounting-based financial performance. Moreover,

consistent with our model, we show that CEO and CFO incentive compensation tied to financial

performance are highly correlated. In addition, we find that cross-sectional variation in the

emphasis on financial performance measures in CFO compensation is broadly consistent with the

predictions of our model.

           More importantly, we find that from 2003 to 2007 public entities (relative to private

entities) lowered the percentage of CFO bonuses contingent on financial performance.

Specifically, we compare the bonus weight on financial performance measures that is expected in

2007 with the actual bonus weight in 2003 (indicative of incentives in the pre-SOX environment)

and find marked differences for public versus private entities. For example, predicted values

from one of our regressions suggest that public companies (with median sample characteristics)

lowered the percentage of their CFO’s bonus that depends on financial performance by about six

percent while comparable private companies with similar characteristics increased the percentage

by about three percent.

           Our findings contribute to the literature in at least three ways. First, we provide new

evidence pertaining directly to the importance of accounting-based financial performance

4
    The dataset includes information corresponding to both corporate and business unit entities. In what follows, the
    label CFO refers to corporate chief financial officers (for corporate-level entities) as well as business unit
    controllers (for business unit entities). Similarly, CEO refers to the top executive either at the corporate or business
    unit level.



                                                              4
measures in evaluating and compensating CFOs. In particular, our unique dataset allows us to

focus mostly on private and smaller public entities where bonuses contingent on accounting-

based performance are the primary source of incentives. This complements most prior studies

where the focus is on large public companies with equity-based pay as the dominant form of

incentive compensation. Second, our findings inform the corporate governance debate whether

CFOs should be earning incentive compensation contingent on reported financial performance.

Both our theory and evidence corroborate the notion that CFO fiduciary responsibilities are

inconsistent with high-powered incentive compensation. At the same time, our results suggest

that some incentive compensation is necessary, and more so in settings where CFOs make

productive contributions beyond their fiduciary duties.

       Finally, we contribute to the growing literature on how greater regulatory scrutiny in the

post-SOX environment has affected executive compensation. For instance, we interpret our

finding that public entities (relative to private entities) have decreased the emphasis on financial

performance in CFO bonuses as evidence that firms mitigate earnings management or other

misreporting practices in part by deemphasizing CFO incentive compensation.

       The rest of the paper is organized as follows. In the next section, we briefly outline our

theoretical model and characterize the optimal linear contract in a setting where a CFO self-

reports performance. Subsequently, we discuss empirical implications of our model. In section 3,

we describe our data and empirical measures. In section 4, we present our results and sensitivity

analyses. Finally, in section 5 we offer some concluding remarks.




                                                  5
2      Background and Theory

2.1 BACKGROUND AND RELATED LITERATURE

A key fiduciary responsibility of a CFO is to produce financial statements that fairly represent a

firm’s financial condition. Although overseeing the financial reporting process of a firm is a

collective responsibility of all executives, CFO effort in this regard is likely to be the most

important determinant of financial reporting quality (Mian [2001]). For example, Geiger and

North [2006] find that discretionary accruals decrease significantly following the appointment of

a new CFO (and the change is not driven by a concurrent appointment of a new CEO). Prior

research also underscores the importance of CFO qualifications in improving internal controls

and financial reporting quality (e.g., Aier et al. [2005], Li, Sun, and Ettredge [2007]).

           In addition to fulfilling their fiduciary responsibilities, CFOs may also have significant

decision-making responsibilities since they often serve as members of firms’ senior executive

team (Siegel and Sorensen [1999]).5 Since incentives for financial performance may be

necessary to motivate CFOs to perform their decision-making responsibilities, fiduciary

responsibilities over financial reporting effectively means that CFOs are rewarded based on

performance measures they themselves generate. In section 2.2, we incorporate this tension into

a theoretical model of CFO compensation which we later use to guide our empirical inquiry of

CFO incentives tied to financial performance. In the remainder of this section, we review related

empirical literature on CFO incentive compensation.




5
    There is some evidence that CFOs’ fiduciary and decision-making responsibilities have countervailing
    consequences. For example, Indjejikian and Matějka [2006] show that when business unit CFOs focus more on
    helping local managers (as opposed to focusing on corporate-level fiduciary responsibilities), local decision
    making is improved but at the expense of less control that manifest in the form of local managers enjoying greater
    budgetary slack.



                                                            6
           Despite the large body of research on executive compensation issues, empirical research

specifically addressing the design of CFO incentive compensation is only emerging. The

increased interest in CFO incentives stems largely from the anticipated consequences of recent

government initiatives aimed at strengthening financial reporting and other corporate governance

practices (e.g., SOX legislation). For instance, Hoitash, Hoitash, and Johnstone [2007] find that

CFO bonuses are negatively associated with SOX-related disclosures of internal control

weaknesses and that this relationship is moderated by CFO expertise, reputation, and board of

director strength. Burks [2008] on the other hand finds that accounting restatements do not affect

CFO bonuses but do increase the likelihood of CFO turnover in the post-SOX environment. 6

           Two other related papers that focus on CFO incentives are Gore, Matsunaga, and Yeung

[2007] (hereafter GMY) and Wang [2006]. GMY show that CFO compensation arrangements

are correlated with other corporate governance practices. For instance, they find that CFOs’

annual grants of equity as well as their aggregate firm equity holdings are lower in settings where

CEOs have financial expertise and/or where corporate boards have finance subcommittees. At

the same time, they find that the sensitivity of CFO total annual compensation (salary, bonus and

equity) to accounting performance is higher in settings where independent members of audit

committees have financial expertise. Finally, GMY show that the sensitivity of CFO total annual

compensation to accounting performance did not change following the enactment of SOX.

           Wang [2006] finds some limited evidence that CFO incentives were affected by SOX.

For instance, Wang [2006] finds that the sensitivity of CFO cash compensation to stock returns

increased in the post-SOX environment but the sensitivity of CFO total compensation (cash plus

equity) was unaffected by SOX. Wang [2006] also reports a comparable pattern of findings for
6
    Hennes, Leone, and Miller [2007] and Collins et al. [2008] also emphasize the threat of dismissal as an important
    source of CFO incentives.



                                                            7
chief operating officers implying that, on average, CFOs were not differentially affected by

SOX. Finally, Wang [2006] examines whether a differential SOX-effect on CFO incentives is

evident for firms with certain corporate governance characteristics. For instance, she finds (i) a

post-SOX increase in the sensitivity of CFO compensation (cash and total) to stock returns for

firms with a weak board structures, and (ii) a post-SOX decrease in the sensitivity of CFO cash

compensation to stock returns for firms with strong boards that operate in high risk environments

(characterized by frequent securities litigation or internal control deficiencies).

           Several other studies examine the impact of SOX on executive incentives (not necessarily

CFOs). There is evidence that the sensitivity of executive compensation (cash and total) to

earnings increased following the enactment of SOX (Carter, Lynch, and Zechman [2007], Chen,

Jeter, and Yang [2007], Paligorova [2007]). There is also evidence that the sensitivity of CEO

wealth to changes in shareholder wealth decreased following the passage of SOX (Cohen, Dey,

and Lys [2007], Chang, Choy, and Wan [2008]). It is unclear, however, whether these findings

reflect SOX-related effects or a trend to substitute stock options with cash compensation which

started even before the passage of SOX (Cohen, Dey, and Lys [2007]).7

           Collectively then, the prior literature examining the effects of SOX on CFO, CEO or

other executive incentives has documented a mixed set of findings. Data limitations partly

explain such mixed findings since firms rarely disclose sufficient detail about their executive

compensation practices. Consequently, researchers in this literature often use a variety of

competing methods and measures to infer the provision of incentives. For instance, results in the


7
    Many of the studies acknowledge the difficulty of identifying the effect of SOX separately from a general time
    trend and attempt to construct control groups of firms that were arguably unaffected by SOX (e.g., firms with a
    high percentage of independent directors on their boards, firms with low discretionary accruals prior to 2002, and
    regulated firms). Nevertheless, it remains an empirical challenge to construct such control groups using
    Execucomp data on large public firms all of which had to comply with the provisions of SOX.



                                                            8
literature differ depending on a researcher’s choice of compensation component (cash, equity

grants, total compensation or aggregate wealth), performance measure (accounting earnings,

stock returns or both), subsamples, or benchmark control groups (if any).

       Relative to this literature, our study bypasses many of these limitations, albeit perhaps not

all. First, our survey allows us to obtain more detailed information about CFO compensation

practices (e.g., the percentage of annual bonuses tied specifically to financial performance). At a

minimum, this alleviates the need to indirectly infer pay-for-performance sensitivities from

available Execucomp data. Second, our sample includes a large of number of private and smaller

public companies whose compensation practices are rarely examined in the compensation

literature. This allows for an identification strategy of the SOX effect on CFO compensation that

has not been feasible in the prior literature. In particular, we use private companies that were

unaffected (or much less affected) by SOX as a natural control group for the treatment-sample of

public companies that had to comply with SOX. Third, most of our sample firms rely primarily

on bonuses rather than equity-based compensation. This simplifies the measurement of

incentives and accentuates the focus on reported financial performance.

       In the next section, we derive empirically testable predictions based on an agency model

of CEO and CFO compensation.

2.2 THEORETICAL MODEL

To characterize a setting where CFOs’ fiduciary and decision-making responsibilities are

considered jointly, we rely on an agency model similar to those in the linear multitasking

literature (e.g., Holmström and Milgrom [1991], Feltham and Xie [1994]). Specifically, we

assume a risk-neutral principal (hereafter the firm) hires two managers (hereafter the CEO and




                                                 9
the CFO) to run the firm. 8 The firm’s economic profit x is given by x    bCEO eCEO  bCFO eCFO ,

where ei , i  CEO, CFO , is agent i’s unobservable productive effort, bi is the marginal product

of such effort, and  ~ N ( ,  2 ) represents other economic factors affecting firm value.

           Both managers are risk-averse with negative exponential utility and risk aversion

parameter ri and providing effort is costly to the tune of                 1
                                                                           2   ei2 . Next, we assume the two

managers’ contracts are linear in reported profit R and other observable contracting variables Z

as in wi   i  i R   iZ where  i is the fixed component of a manager’s compensation and

(  i ,  i) are the incentive coefficients.

           Let R  x   where  ~ N ( ,  2 ) is independent of x , representing the extent to which

reported profit misstates economic profit. We assume the bias in R is   k (1  q)  0 so that, on

average, R overstates economic profits. In particular, we assume k  0 is a known parameter

that measures the extent to which R is susceptible to overstatements and q  [0,1] is the CFO’s

unobservable effort to reduce such overstatements. 9 We assume that a failure to reduce

overstatements costs the CFO an expected amount of                    1
                                                                      2   c(1  q) 2 , with   1
                                                                                              2   c  0 representing the

maximum costs that can be incurred. Hence, barring incentives to do otherwise, the CFO prefers

unbiased reporting (i.e., q  1 ). Moreover, we suggest that the firm also prefers an unbiased R

since it bears an exogenous cost for such biases in the amount of  2 / 2 (with   0 ).



8
  We can also consider a firm hiring more than two managers as long as one of the managers is entrusted with CFO
  duties as described below. Also, for simplicity, we assume that the CEO and the CFO perform their respective
  tasks simultaneously and preclude the possibility of collusion between them.
9
  In unreported analysis, we also consider a setting where k arises endogenously via the CEO’s strategic choice. For
    instance, if we let k  k0  k1 m and assume that the CEO chooses m at a cost of              1
                                                                                                  2   m 2 , then we can show that
    the comparative statics illustrated in Proposition 1 below largely apply to this setting as well.



                                                             10
       In a similar vein, we represent the measurement error in R as  2   2 (1  h) , where  2

is known and h [0,1] represents the CFO’s unobservable effort to reduce measurement error at

a cost of h /(1  h) . Of course, the firm prefers  2 to be as low as possible since ―noisy‖

performance measures are costly vehicles for evaluating risk-averse agents.

       Finally, we assume Z is the set of all measures (other than R ) that the firm may use to

evaluate its managers. This may include market-based indicators, non-financial measures, or

subjective assessments of managerial performance. However, to simplify the presentation of our

results in this section, we present the model’s key implications below assuming R is the only

contracting variable. In Appendix A, we provide a more complete and detailed illustration where

we let Z  ( z1 , z2 ) with z1 and z 2 informative about the CFO’s choices of eCFO and q

respectively.

       Taken together, our assumptions imply that the principal’s task is to motivate the CEO

and the CFO to perform their productive tasks (eCEO , eCFO ) as well as motivate the CFO to

perform his fiduciary duties q and h . We have,

       LEMMA 1: CEO’s optimal choice of eCEO and CFO’s optimal choice of (eCFO , q, h) are

given by

        eCEO   CEO bCEO ;
         *
                                eCFO   CFO bCFO ;
                                 *
                                                                                                 (1a)

                   k CFO
                                                               
                                                                    1/ 2
        1  q*                 1  h*               CFO 2
                                               rCFO    2
                          ;                      2                          .                    (1b)
                      c

       Lemma 1 suggests a number of observations. First, as expected, the two expressions in

(1a) suggest that incentive pay is necessary to motivate productive effort. Second, the first

expression in (1b) suggests that the CFO’s choice of q is always less than the (first-best)



                                                          11
maximum of one unless R is not susceptible to overstatements (i.e., k  0 ) or the CFO is paid a

flat wage (i.e.,  CFO  0 ). This is a confirmation of the intuition in Dye [1988], Evans and

Sridhar [1996], Goldman and Slezak [2006], and Crocker and Slemrod [2007] among others,

namely that if reported performance used to motivate agents to be productive, then some

overstatements need to be tolerated. Finally, the last expression in (1b) suggests that, CFO

incentive compensation (i.e.,  CFO ) is necessary to motivate h as well as to motivate eCFO .

        To characterize the firm’s selection of optimal linear contracts for its managers, we have:

        LEMMA 2: The CEO’s and CFO’s optimal compensation contracts are characterized by

the incentive coefficients, (  CEO ,  CFO ) as follows

                                  2
                                bCEO
          CEO    2                                  ;                                          (2a)
                  bCEO  rCEO 2  rCEO 2 (1  h* )

                                     rCEO 2 (1  h* )
                         b 2     2

                                         2 CFO
                             CFO   CEO
         CFO                                                                                   (2b)
                   2
                  bCFO    k D  rCFO   rCFO 2 (1  h* )
                            2          2




         where (1  h ) is given by (1b) and D 
                         *                       1  k           2
                                                                       /c   .
                                                               c

        The incentive coefficients in Lemma 2 have many of the usual interpretations illustrated

in the multi-tasking literature with the added feature that the incentives for the two managers are

positively linked via the optimal choice of h . As an illustration, note that  CFO  0 even when

the CFO does not provide productive effort (i.e., when bCFO  0 ). The intuition is

straightforward. Since a less noisy performance measure (i.e., high h ) is more effective in

motivating a CEO, the principal relies on CFO incentive compensation to motivate h .




                                                       12
        In summary, the preceding discussion suggests that, as the need to motivate and reward

various decision-makers in an organization increases, CFO incentives also increase. An

immediate empirical implication is that CFO incentives are positively correlated with CEO

incentives (or other executives’ incentives). However, while strong incentives may be

appropriate for the CEO and other executives, CFOs are the ones primarily in charge of the

financial reporting process and strong incentives could conflict with their role in eliminating

overstatements in R . That is, while a high  CFO motivates both eCFO and h , it also implies

greater overstatements of reported performance (i.e., lower q). The following proposition derives

several implication of this conflict for optimal CFO compensation.

        PROPOSITION 1: The sensitivity of CFO compensation to reported profit is (i) increasing

in both the CFO’s and CEO’s marginal productivity; (ii) decreasing in the extent to which

reported profit is susceptible to misreporting; (iii) increasing in CFO’s personal cost of

misreporting; and (iv) decreasing in the cost of misreporting borne by the firm. That is:

              d CFO      d CFO            d CFO            d                  d CFO
        (i)           0,         0 ; (ii)         0 ; (iii) CFO  0 ; and (iv)        0 .
              dbCFO       dbCEO              dk                dc                  d


        The first result reflects the positive link between the two incentive coefficients discussed

above—when either manager is more productive, CFO incentives are steeper. In contrast, if k is

high and the profit report is more susceptible to misreporting, or if c is low and the CFO is more

inclined to tolerate misreporting, then CFO incentives are muted to mitigate misreporting in the

form of overstated performance (and consequently the payment of excessive compensation).

Finally, when the misreporting costs borne by the firm are high, CFO effort in preventing

misreporting is more important and incentives are muted to avoid compromising CFO’s fiduciary

responsibilities.



                                                   13
2.3 EMPIRICAL IMPLICATIONS

The above results have implications for the extent to which firms emphasize financial

performance measures (R in our model) in CFO incentive compensation. Part (i) of Proposition 1

predicts that CFO incentives tied to financial performance will be positively associated with

proxies for marginal productivity of CFOs and other executives (section 3 describes our proxies

in more detail). Part (ii) predicts a negative association between CFO incentives tied to financial

performance and susceptibility of accounting systems to misreporting. In section 4.2, we test

these two cross-sectional predictions to empirically corroborate our model.

       In addition, in section 4.3 we rely on parts (iii) and (iv) of Proposition 1 to examine the

changes in CFO incentive compensation from 2003 to 2007 (i.e., during the post-SOX

environment). Much of prior literature argues that the integrity of financial reporting has

improved after SOX. However, as explained below, our results suggest that improved financial

reporting can have very different implications for CFO compensation depending on assumptions

about what drove such improvements.

       One of the provisions of SOX makes CEOs and CFOs responsible for establishing and

maintaining internal controls and requires them to certify that financial statements fairly present

results of operations. Other provisions of SOX call for more severe civil and criminal penalties

(incl. imprisonment) for violations of various regulations. Thus, it seems evident that CEOs and

CFOs bear greater personal risk and liability for misreporting the financial performance of their

firms (see also Carter, Lynch, and Zechman [2007], Cohen, Dey, and Lys [2007]). Within the

framework of our model, these SOX provisions that specifically target CFOs can be

characterized as an increase in c (greater CFO loss of reputation or personal cost of




                                                 14
misreporting). Ceteris paribus, an increase in c leads to stronger incentives (part (iii) of

Proposition 1).

       However, in the post-SOX environment misreporting or inadequate internal controls also

impose greater costs on the firm and/or its directors. For example, prior studies find that firms

disclosing internal control weaknesses in the post-SOX environment experienced significant

decreases in market valuation as well as increases in their cost of capital and audit fees (e.g., De

Franco, Guan, and Lu [2005], Ashbaugh-Skaife et al. [2007], Hammersley, Myers, and

Shakespeare [2008], Hogan and Wilkins [2008]). Within the framework of our model, these

effects can be characterized as an increase in  (greater cost of misreporting borne by the firm)

which, ceteris paribus, leads to weaker incentives (part (iv) of Proposition 1).

       The effects of c and  are countervailing. If the impact of higher personal costs for

CFOs (increase in c ) dominates the impact of higher misreporting costs borne by the firm

(increase in  ), then we predict an increase in CFO compensation tied to financial performance

after the passage of SOX. Conversely, if the increase in misreporting costs borne by the firm is

the dominant force, then we predict a decrease in CFO incentive compensation tied to financial

performance.


3   Empirical Design

3.1 SAMPLE SELECTION

We construct our sample from a survey of 29,857 members of the American Institute of Certified

Public Accountants (AICPA). This population is comprised of all AICPA members working in

industry (rather than in public accounting firms) as corporate or business unit (BU) CFOs




                                                 15
(controllers, VP finance) or CEOs (general managers). 10 In March 2007, we sent an e-mail (with

AICPA endorsement) inviting the AICPA members to participate in an on-line survey (by using

an individual log-in key). This invitation was followed by two subsequent e-mails within two

weeks of the initial e-mail to request participation of those who had not responded. By the end of

May, 2,037 AICPA members fully or partially completed our on-line questionnaire yielding a

response rate of 6.8%.

        The on-line survey instrument consisted of 8 questionnaires with similar content that

were customized depending on the type of respondent (there was a corporate- and BU-level

version of the questionnaire for each of the following four types of respondents: CFO, CEO,

other respondent informed about CEO/CFO compensation, other respondent not informed). 11

Most respondents were CFOs (77%) but some were CEOs (5%) and other executives informed

about CFO/CEO compensation (18%). Participation was anonymous and the questionnaire did

not collect any information about firm affiliation. The total number of items on the

questionnaires ranged between 61 and 88 depending on the type of respondent.

        We exclude 684 observations from the initial sample of 2,037 respondents because (i) we

lack data on the type of entity (public vs. private or corporate vs. BU level), or (ii) the respondent

is neither CFO/CEO nor informed about their compensation, or (iii) the entity is a governmental

or not-for-profit institution, or (iv) the SIC-2 industry classification code is greater than 81 (incl.

educational, social services; museums; membership organizations; engineering, accounting,

research services), or (v) sales are less than $10 million. This procedure yields a sample of 1,353

10
   In addition, we also surveyed AICPA members with other titles such as president or chief operating officer who
   are likely to be informed about executive compensation issues.
11
   The survey instrument was extensively pre-tested by the authors, numerous colleagues with survey and/or
   management accounting expertise, and also by a sample of 24 AICPA members and other CFOs/CEOs who
   participated in earlier surveys of the authors (their responses are included in our data; excluding these 24
   observations does not materially affect our results).



                                                        16
entities of which 13% are public companies, 10% are BUs of public companies, 69% are

privately-owned companies, and 9% are BUs of private companies.

3.2 MEASURES

3.2.1      Incentive Weight on Financial Performance Measures.

We ask several questions to collect information about CFO incentives. The first question

(Question 1a in Appendix B) asks respondents to report the percentage of CFO bonus earned in

2006 that can be attributed to the achievement of (i) financial performance targets (denoted

FIN06), (ii) financial performance targets of higher levels such as the firm or business group

(HLF06) where appropriate (e.g., for BU CFOs), (iii) non-financial performance targets

(NFIN06), and (iv) achievements evaluated subjectively without pre-set targets and other

achievements (SUB06). The question is based on an instrument commonly used in prior research

(Gupta and Govindarajan [1984], Abernethy, Bouwens, and van Lent [2004]). Variations of this

question (Question 1b and 1c) ask respondents to report the corresponding percentages in 2003

and the expected percentages in 2007 assuming performance exactly meets the targets on all

measures.12

           Question 2 in Appendix B asks respondents to report the dollar amount of bonus (denoted

as BONUS06), long-term cash compensation, equity-based compensation, and other incentive

compensation earned by CFOs in 2006. Question 3 asks respondents to report CFO annual

salaries for 2006 (SALARY) and Question 4 asks respondents to report the dollar amount of

bonuses expected in 2007 if performance exactly meets targets on all measures (BONUS07).




12
     We use abbreviations similar to those above to denote various components of CFO bonus plans, e.g., FIN07 is the
     percentage of 2007 CFO bonus expected to be based on financial performance targets. We also have information
     on the components of CEO bonus in 2006 which we denote FIN06_CEO, HLF06_CEO, etc.



                                                          17
           Given that annual bonuses represent the primary incentive compensation vehicle for most

CFOs (only about 15% CFOs in our sample are eligible for a long-term cash incentive, and 22%

for equity incentives), in what follows we focus exclusively on the choice of performance

measures underlying annual bonuses. Since our model makes predictions about ex ante incentive

strength, we construct our primary empirical measure of incentives using bonuses that CFOs

expect to earn if financial performance targets are met in 2007. Specifically, we calculate

FINBON07 as the percentage of annual salary CFOs expect to earn as a bonus for meeting

financial performance targets:13

                             FIN07  BONUS07
            FINBON07                                                                                (4)
                                 SALARY

           As an alternative measure of incentives tied to financial performance, we also consider

just the expected percentage of bonus based on financial performance measures, i.e., FIN07 in

the numerator of (4). Finally, as a third alternative, we construct FINBON06, bonuses earned for

financial performance in 2006 as a percentage of annual salary:

                             FIN06  BONUS06
            FINBON06                                                                                (5)
                                 SALARY

           We note that FINBON06 is an ex post measure of incentive strength (e.g., FINBON06

can be zero due to poor performance in 2006 even if the compensation contract at the beginning

of the year allowed for a large bonus contingent on performance). Therefore, our tests relying on

FINBON06 also control for performance in 2006.

3.2.2      Explanatory Variables

To test the implications of our model (see Proposition 1), we assume that the corporate

governance environment in the post-SOX period corresponds to higher CFO misreporting costs

13
     We use 2006 CFO salaries since 2007 salaries are not available.



                                                          18
( c in our model) and higher costs of misreporting borne by the firm (  in our model). In

addition, we construct four cross-sectional proxies for the importance of value-enhancing CFO

and CEO effort ( bCFO , bCEO ) and three proxies for the susceptibility of accounting systems to

misreporting ( k ). We discuss construction of these proxies and additional control variables

below.

Marginal Product of Effort

First, we expect the marginal product of effort of all executives (including the CFO) to be greater

in high-growth settings because of the high opportunity cost of missed investment opportunities

(Smith and Watts [1992], Baber, Janakiraman, and Kang [1996]). We measure GROWTH by

asking respondents about the percentage increase in sales in their organization relative to the

prior year (Question 5 in Appendix B). We winsorize GROWTH at the 1% level to mitigate the

influence of outlier observations.

            Second, we expect the marginal product of CFO effort to be greater for firms and

industries that rely more on external long-term debt financing since CFO input and expertise is

likely to be more critical in these settings. We calculate the median industry ratio of long-term

debt over total assets by averaging the ratio (Data9/Data6) over 2002–2006 for each firm in the

Compustat Industrial Annual data file.14 We then calculate LDEBT as the industry (SIC-2 digit)

median. We assign values of LDEBT to observations in our sample by matching respondents’

descriptions of their industries (Questions 6 in Appendix B) with the SIC industry classification.

We assign missing values whenever a description does not allow for an unambiguous

classification.

14
     We recognize that our sample entities do not necessarily match Compustat firms in size. Therefore, we also
     calculate our industry-level variables using only Compustat firms that match our sample entities in size as closely
     as possible and find qualitatively similar results.



                                                            19
       Third, we expect the marginal product of CFO effort to be greater for firms and industries

with high inventories and/or accounts receivable where CFO involvement in working capital

management has a greater impact on firm profits. We calculate INREC, the median industry ratio

of inventory and receivables over total assets ((Data2+Data3)/Data6) in a similar way as

described above for LDEBT.

       Fourth, we expect the marginal product of CFO effort to increase over his tenure as the

CFO becomes more familiar and knowledgeable about his firm’s operations. We calculate

TENURE as the log of the number of years served as the CFO (we use the logarithm

transformation to reduce deviations from normality).

Susceptibility to Misreporting

Constructing a proxy for susceptibility of accounting systems to misreporting is a challenging

task. Prior literature typically has relied on detailed information extracted from firms’ financial

statements and other corroborating sources to infer misreporting behavior, often with varying

success. Unfortunately, since our respondents’ firm affiliations are unknown, we cannot replicate

misreporting or earnings management proxies commonly employed in the literature. Instead, we

construct three proxies as follows.

       The first proxy is based on information about CEO turnover reported by our respondents.

Prior literature has shown that CEO turnover is often associated with greater frequencies of asset

write-offs, accounting method changes, and recordings of large accruals (e.g., Murphy and

Zimmerman [1993], Pourciau [1993]). In this spirit, we use an indicator variable CEOTURN




                                                 20
(equals one if a new CEO was appointed in the last two years) as a proxy for susceptibility of

accounting systems to misreporting. 15

           Our second proxy is an industry-level measure of total accruals assuming that accounting

systems are more susceptible to misreporting in environments where earnings substantially

deviate from underlying cash flows. Specifically, we calculate ACCRUAL as the industry (SIC-2)

median of the absolute value of total accruals over total assets in 2006 defined as income before

extraordinary items (Data123) minus cash flow from operations (Data308), divided by total

assets (Data6).

           Finally, as a third proxy, we consider an indicator of board of directors’ influence on

compensation of corporate CFOs (alternatively, higher-level executives’ influence on

compensation of BU CFOs). Prior literature suggests that accounting systems are less susceptible

to misreporting when boards are actively engaged in corporate governance (Beasley [1996],

Dechow, Sloan, and Sweeney [1996], Klein [2002]). Our measure of board involvement,

INVOLVE, asks respondents to indicate who selects performance measures for CFOs on a 5 -item

fully anchored Likert scale ranging from ―CEO decision‖ through ―Joint decision‖ to ―Board

decision‖ (Question 7 in Appendix B also includes a similar scale for BU-level entities).

Additional Control Variables

We also include a number of control variables in our tests. To control for ownership status and

the organizational level employing the CFO (corporate vs. BU), we use the following indicator

variables: (i) PUBLIC_CORP for publicly listed (independent) companies, (ii) PRIVATE_CORP

for independent companies that are privately owned, (iii) PUBLIC_BU for business units of

publicly listed companies, and (iv) PRIVATE_BU for business units of privately-owned
15
     Given that respondents answer the question between March and May 2007, CEOTURN equals one for all entities
     with a new CEO during 2006.



                                                        21
companies. To control for the size of the entities in our sample, we use the log of the number of

employees employed by a company or BU, labeled SIZE. Of course, we recognize that size and

ownership status can also serve as proxies for marginal product of CFO/CEO effort.

Nonetheless, given that size and ownership status may also reflect susceptibility of accounting

systems to misreporting, the cost of misreporting borne by the firm, or other unmodeled effects,

we make no specific predictions regarding the effects of these variables in our tests.


4   Results

4.1 DESCRIPTIVE STATISTICS

Table 1 presents descriptive statistics for our sample of 1,353 observations divided into four

panels by entity type. As expected, the size of the entities in our sample (measured by the median

number of employees) varies depending on the type of entity: 1,160 in public companies, 500 in

BUs of public companies, 175 in private companies, and 111 in BUs of private companies. All

entities exhibit solid growth in sales (the median increase in 2006 sales as compared to the year

before is 10 to 11% for all types of entities). The median industry ratio of long-term debt over

total assets is 12 to 16%, median inventory and receivables over total assets is 28 to 33%, and

median (absolute value of) total accruals over total assets is 7 to 8%.

       The median tenure of a CFO is about 6 years for most entities except for CFOs of private

companies where tenure is significantly higher at about 8 years. Private companies are also less

likely to experience CEO turnover (CEOTURN in Table 1). For instance, panel C of Table 1

reports that 11% of corporate CEOs of private companies have been appointed in the last two

years which is significantly less than the 24% reported in panel A of Table 1 for corporate CEOs

of public companies (the corresponding percentages for BU-level CEOs are 33% for public

companies and 21% for private companies). Finally, the modest median values for INVOLVE in


                                                 22
Table 1 suggest that board of directors’ are not intricately involved in evaluating CFO

performance.

                                          [Insert Table 1]

       Table 1 also shows that median 2006 CFO salary and bonus in public companies are

$246,403 and $88,000, respectively (which is comparable to the 2006 median cash compensation

of CFOs in Execucomp, $395,500). Not surprisingly, this is significantly more than median CFO

salaries of $145,000 in BUs of public companies, $135,000 in private companies, and $110,000

in BUs of private companies. We also find that both long-term cash and equity-based

compensation are insignificant in our sample entities with the exception of corporate CFOs in

public companies where the median 2006 equity-based compensation is $95,800.

       Table 2 provides information on the types of performance measures used in CFO (and

CEO) bonus plans. For each of the different types of entities, Panels A through D report

percentages of CFO bonus tied to various performance measures in 2006 and 2003 as well as the

percentages expected if 2007 performance exactly meets all targets (CEO-specific percentages

are available only for 2006). The key highlight of Table 2 is that financial performance measures

are by far the most important determinants of CFO bonuses. For instance, we find that bonuses

based on financial performance are, on average, expected to be about 67% of annual bonuses for

corporate-level CFOs of public companies (FIN07 in panel A) and 51% of annual bonuses for

corporate-level CFOs of private companies (FIN07 in panel C). The corresponding percentages

for BU-level CFOs of public and private companies are 57 and 37% respectively (FIN07 in

panels B and D). The percentages actually earned in 2006 and 2003 (FIN06 and FIN03 in panels

A through D) exhibit a similar pattern.

                                          [Insert Table 2]



                                                23
            Table 2 also shows that CFO bonuses are sensitive to non-financial measures, subjective

evaluations, and for BU CFOs to higher-level (corporate or business group) financial

performance. For instance, we find that corporate CFOs of public companies expect to earn 15%

of their 2007 bonus based on explicit non-financial measures and about 17% based on subjective

evaluations (panel A of Table 2). The corresponding percentages for corporate CFOs of private

companies are 14% and 35% (panel C of Table 2) suggesting that private entities place a

significantly greater weight on subjective evaluations of CFOs.

            Finally, Table 2 highlights how CFO bonuses linked to financial performance compare

with similar incentives for CEOs. Combining all types of entities into a sample of 870 entities

where information for both executives is available, we estimate that CFO bonuses linked to

financial performance (as a percentage of the 2006 annual bonus) is lower than the

corresponding CEO bonus; that is, FIN06 is lower than FIN06_CEO by an average of 4.4%

(p<0.001). We also find that FIN06 is highly correlated with FIN06_CEO (r=.80, p<.001;

untabulated). This is consistent with our model predicting a positive link between incentive

coefficients of CFOs and CEOs.

            For a subset of the observations that report the use of non-financial measures in Table 2,

we also provide additional detail about the specific types of non-financial measures that underlie

CFO bonuses. These are tabulated in Table 3 into four broad categories labeled General

Management, Reporting, Financing, and Communications and Teamwork. 16 For this subsample,

we make two observations that further validate the two types of CFO duties contemplated by our

theoretical model. First, the frequency of use of non-financial measures linked to General


16
     Table 3 is based on a survey item that asks respondents to describe in words the types of non-financial
     performance measures used in their organizations. Although it is generally difficult to obtain responses to such
     open-ended survey questions, 278 respondents from our sample of 1,353 provided such descriptions.



                                                            24
Management duties (51% of respondents) and/or Financing duties (31% of respondents) suggest

a first-order concern for motivating CFOs to perform value-enhancing tasks. Second, the use of

non-financial measures linked to Reporting duties (36% of respondents) suggests a non-trivial

concern also for motivating CFOs to perform their fiduciary duties. In particular, we note that the

frequency of using at least one compliance and control objective (including objectives related to

clean audit, no restatements, external/tax reporting, SOX compliance, documentation of

procedures, risk management, or regulatory reporting) is highest (26% of respondents) in public

corporate-level entities where CFO fiduciary responsibilities are arguably most important and

lowest (8%) in private corporate-level entities (the difference is highly significant, p=.002).

                                          [Insert Table 3]

4.2 CROSS-SECTIONAL DETERMINANTS OF CFO INCENTIVE WEIGHTS

A key feature of our theoretical model is that incentives tied to financial performance divert

CFOs’ attention from their fiduciary responsibilities, yet at the same time such incentives are

necessary to motivate productive effort. Reflecting this trade-off, our model predicts that the

weight on financial performance measures in CFO incentive compensation increases with

marginal product of CFO and CEO effort and decreases with susceptibility of accounting

systems to misreporting. In this section, we empirically examine the extent to which firms take

into account the relative importance of CFO fiduciary and productive responsibilities when

designing CFO incentive compensation.

       Specifically, we estimate the relation between the weight on financial performance

measures in CFO bonuses and several proxies for marginal product of effort and susceptibility of

accounting systems to misreporting described in section 3.2. We use three measures of the

weight on financial performance: (i) FINBON07, the proportion of annual salary CFOs expect to



                                                 25
earn as bonus if financial performance targets are met in 2007 (expression (4) in section 3.2), (ii)

FIN07, the expected percentage of bonus based on financial performance measures, and (iii)

FINBON06, the proportion of annual salary CFOs earned as bonus based on financial

performance measures in 2006 (expression (5) in section 3.2).

           Table 4 presents the results of our Tobit estimates which take into account that our

dependent variables are continuous variables with a positive probability mass at zero (and at 100

in case of FIN07).17 We impose an additional sampling criterion that CFOs be in their current

position for at least one year before participating in our survey during March–May 2007.

Consistent with our prediction, the results in Column 1 suggest that the 2007 (ex ante) incentive

weight on financial performance measures in CFO bonus plans is positively associated with three

of our proxies for high CFO impact on firm value—growth (p=.083), high industry-level

importance of long-term debt (p=.058) and CFO tenure (p=.042).

                                                     [Insert Table 4]

           The results in Column 1 also suggest that the incentive weight on financial performance

measures in CFO bonus plans is lower (p=.028) in the period surrounding CEO turnover.

Specifically, the 2007 weight on financial performance measures is lower by about 9% in entities

with recent CEO turnover as compared to entities without CEO turnover (this comparison relies

on coefficient estimates in Column 1 and assumes all other predictors are constant at their

sample medians). The remaining significant results in Column 1 show that the incentive weight

on financial performance measures is positively associated with size as measured by the number

of employees (p<.001) and higher by about 12% (p=.080) in public corporate entities than in

private corporate entities.

17
     OLS estimations yield qualitatively similar results.



                                                            26
       Column 2 of Table 4 presents the results of estimating a model where the dependent

variable is FIN07 which does not depend on the amount of bonus expected in 2007 and allows

for a larger sample size (since BONUS07 and consequently FINBON07 have missing values). As

in Column 1, the relative emphasis on financial performance measures is positively associated

with CFO tenure (p=.001) and with one of the industry-level proxies for marginal product of

CFO effort (the effect of INREC is significantly positive, p=.040, while the effect of LDEBT is

not significant). Consistent with Column 1, we also find that that the relative emphasis on

financial performance measures is lower in the period surrounding CEO turnover (p=.071).

       Relative to Column 1, the additional insight from Column 2 is that board (higher-level)

involvement in selecting CFO performance measures is a strong predictor of the emphasis on

financial performance measures (p=.002). Specifically, when boards are actively involved (one

of our proxies for lower susceptibility to misreporting) in evaluating CFOs, firms rely on

financial performance measures more than other measures. In addition, we also find that BUs of

private companies emphasize financial performance measures in CFO incentive compensation

much less than private corporate entities (p=.008).

       Finally, the last column of Table 4 presents the results for FINBON06, the ex post

incentive weights on financial performance measures in 2006. To alleviate the concern that

variation in FINBON06 is largely due to random shocks to financial performance, we control for

2006 performance relative to budget (see Question 8 in Appendix B). While we do not interpret

the coefficients estimates directly (given endogeneity biases arising from including performance

as a regressor), we note that that the results are largely similar those in Columns 1 and 2. The




                                                27
only additional finding of note is that ACCRUAL is negatively associated with ex post incentive

weights on financial performance measures in 2006.18

            In summary, we find that ex ante incentive weights on financial performance measures

are strongly positively associated with CFO tenure and size which likely proxy for greater

marginal product of CFO value-enhancing effort. Other proxies for marginal product of effort—

growth, industry importance of long-term debt, and industry importance of inventory and

receivables—are also significant predictors in at least one of the specifications. At the same time,

we find that incentive weights on financial performance measures are lower when accounting

systems are more susceptible to misreporting, i.e., in periods surrounding CEO turnover and/or

when corporate boards have lower influence on CFO compensation.

            We interpret these findings as consistent with the key feature in our model that firms

balance the need to motivate productive tasks with the importance of CFO fiduciary

responsibilities when designing CFO incentive compensation. At the same time, we

acknowledge that our proxies based on limited survey data are imperfect (in particular, proxies

for susceptibility of accounting systems to misreporting) and that we have limited ability to rule

out various alternative plausible explanations of these cross-sectional results.

4.3 CHANGES IN CFO INCENTIVE WEIGHTS

In this section, we examine whether firms changed the percentage of CFO bonuses based on

financial performance measures between 2003 and 2007. In particular, we compare FIN03, the

percentage weight on financial performance measures in 2003, to FIN07, the percentage weight


18
      This result has to be interpreted with caution for two reasons. First, the specification in Column 3 of Table 4
     suffers from endogeneity biases since both performance and accruals are used as regressors. Second, the result is
     driven by a few industries (SIC-2 categories) that exhibit low average incentive weights on financial performance
     measures and have total accruals that are much more negative than other industries. Thus, for example, if we use
     raw total accruals rather than their absolute values, the negative association is no longer apparent.



                                                           28
on financial performance measures expected in 2007. Comparing FINBON07 to its 2003

equivalent is not feasible since we lack data on CFO bonuses paid in 2003.

            Given that FIN03 was established in 2002 or earlier, the difference between FIN07 and

FIN03 likely reflects firms’ reactions to SOX which was enacted around the beginning of this

time period. Since SOX was targeted first and foremost at public entities, we identify the impact

of SOX as the difference between FIN07 and FIN03 in public entities relative to the difference in

private entities. We note that this difference-in-differences design dominates a simple before-

and-after design without an untreated comparison group (Meyer [1995], Bertrand, Duflo, and

Mullainathan [2004]). In particular, this design allows us to isolate trends in compensation

practices affecting all firms comparably during this time period from the effects attributable to

SOX.19

            Based on the above discussion, we estimate the following OLS model:

             yij   0  1 ×Y07   2 × BU  3 × PUBLIC   4 ×Y07× PUBLIC   ij ,                                     (6)

where the dependent variable is the percentage of bonus based on financial performance,

i represents entities, j represents year 2007 or 2003, Y07 is an indicator variable for 2007, BU is

an indicator variable for BU-level entities, PUBLIC is an indicator variable for public entities,

and  4 represents the difference-in-differences estimator of the effects of SOX.

            Table 5 presents the results of estimating (6) in three columns. Column 1 reports results

for all available observations whereas, more in the spirit of our theoretical model, Columns 2

and 3 focus on sample entities where the annual bonus rather than equity-based pay is the


19
     Later in this section, we assess the sensitivity of our results to potential misclassification of treatment and control
     observations based on the public/private distinction. We also note that the difference-in-differences design does
     not account for other events during this time period that may have affected public and private entities
     differentially.



                                                              29
dominant form of incentive compensation. In particular, Column 2 presents the results for a

sample excluding observations where CFOs earn significant amount of equity compensation

(defined as $50,000 or more in 2006).20 Column 3 further excludes observations where CFO

bonuses are equal or less than 30% of salary.

                                                     [Insert Table 5]

           Column 1 of Table 5 suggests that CFOs in private companies have, on average, 48% of

their bonus tied to financial performance during this period. The corresponding percentages for

public corporate CFOs and BU CFOs (of both public and private companies) are higher by 24%

(p<.001) and 9% (p=.002) respectively. More importantly, we find that public entities (relative to

private entities) reduced CFO bonus weights on financial performance by about 8% (p=.001)

from 2003 to 2007 (i.e., in the post-SOX environment). For the alternative samples in Columns 2

and 3, we find similar results. For instance, we find that public entities reduced CFO bonus

weights on financial performance by about 9% in Column 2 and by about 17% in Column 3.

           The analysis in Table 5 rests on two important design features worthy of further

comment. First, our empirical specification in (6) considers a firm-year observation as the unit of

analysis without requiring matching entities for the 2003 and 2007 subperiods. Thus, to the

extent that the 2003 sample entities differ from the 2007 entities (due to missing observations),

our estimates of  4 in Table 5 reflect both sample heterogeneity and the effects of interest

attributable to SOX. Second, our empirical specification in (6) assumes that the difference

between FIN07 and FIN03 in public entities relative to the difference in private entities (i.e., the

SOX effect) is independent of the organizational level (i.e., corporate versus BU) employing the

CFO. To assess the impact of these two features, we reestimate (6) as follows:

20
     Using $70,000 or $30,000 as alternative cut-offs yields similar results.



                                                             30
            FIN07 -03i   0   1 PUBLIC _ CORP   2 PUBLIC _ BU   3 PRIVATE _ BU   i ,                (7)

where FIN07-03 is the difference between FIN07 and FIN03 for observations with non-missing

values on both FIN07 and FIN03,  0 corresponds to  1 in (6), and the three coefficients (  1 ,  2 ,

 3 ) for each type of entity are analogous to  4 in (6).21 To ensure that FIN07-03 is not affected

by CFO turnover, we also exclude entities with CFO tenure smaller than or equal to 3 years.

                                                      [Insert Table 6]

           Table 6 presents the results in three columns similar to those in Table 5. For example, the

sample of 647 entities in Column 1 of Table 6 corresponds to the 1,898 observations in

Column 1 of Table 5 after removing 604 entities with missing values on either FIN03 or FIN07.

Columns 2 and 3 of Table 6 impose the same sample restrictions (on the importance of bonu s

incentives) as the corresponding columns in Table 5. The results in Column 1 of Table 6 show

that from 2003 to 2007 private companies increased the bonus weights on financial performance

for corporate CFOs by about 3% (p=.008). This change may reflect the increased importance of

CFO value-enhancing activities during this period. Relative to private companies, BUs of public

companies reduced the weights on financial performance measures in CFO bonuses by about 8%

on average (p=.004), while a comparable estimate specific to corporate-level CFOs in public

companies is not significantly different from zero. 22

           Comparing Tables 5 and 6, we offer two explanations for our finding that the decrease in

bonus weights on financial performance for public entities is most evident for BU CFOs rather

than corporate CFOs. First, our model (illustrated in Section 2.3) predicts a decrease in

incentives tied to financial performance if the increase in misreporting costs borne by a firm

21
      0 ,  2 , and  3 in (6) are differenced out since they are constant over time and across entities.
22
     These results are robust to including many of the control variables considered earlier in Table 4.



                                                               31
dominates the increase in a CFO’s personal costs of misreporting. In this spirit, it is plausible that

CFOs in public BUs did not experience a substantial increase in personal liability for

misreporting during this time period since, for example, they do not have to publicly certify the

accuracy of financial statements (a requirement SOX introduced only for corporate CFOs). Thus,

to the extent that PUBLIC_BU proxies more for cost of misreporting borne by the firm (or

corporate executives) than personal costs borne by a CFO, the results in both Tables 5 and 6 are

consistent with our model.

       The second explanation for our finding that the decrease in bonus weights on financial

performance for public entities is most apparent for BU CFOs revolves around the likely

confounding effect of equity compensation. Consistent with this explanation, Column 2 of

Table 6 shows that (after removing entities with substantial equity compensation) public entities

relative to private entities reduced the bonus weights on financial performance by about 8% for

corporate and by about 6% for BU CFOs (p=.013 and p=.036, respectively). Finally, Column 3

of Table 6 illustrates a pattern similar to Column 2; we find public entities reduced the bonus

weights on financial performance for both corporate CFOs and BU CFOs.

       Taken together, the evidence in Tables 5 and 6 suggests that from 2003 to 2007 public

entities (relative to private entities) reduced the percentage of CFO bonuses tied to financial

performance. Although we attribute this decrease to factors associated with SOX, we

acknowledge that the public/private distinction imperfectly isolates the effects attributable to

SOX. For instance, given the charged corporate governance atmosphere around this time period,

it is conceivable that some private entities in our control sample may have been affected by SOX

indirectly. Hence, some of these private entities may be viewed more appropriately as treatment

observations rather than control observations. Conversely, some of our public entities in our



                                                 32
treatment sample may be viewed as control observations if the effects of SOX during this time

period were weakened due to its delayed implementation in smaller public companies.

       To assess the potential for misclassification, we analyze an item in our survey initially

intended to gauge respondents’ focus on their unit’s financial reporting function in the post-SOX

period. The survey item asks respondents to indicate the extent to which they agree with the

following statement: ―Since the enactment of the Sarbanes-Oxley legislation, we have greatly

increased the independence of the financial reporting function in our firm [BU].‖ Table 7 shows

the distribution of responses for a sample of 1,188 observations with non-missing values. For

private companies, we find that about 40% of respondents viewed the statement as not applicable

while 20% agreed with the statement. In contrast, for public companies, we find that about 4% of

respondents viewed the statement as not applicable while 46% agreed with the statement. A

comparison of private BUs and public BUs illustrates a similar pattern.

                                          [Insert Table 7]

       Overall, and as expected, the responses in Table 7 suggest that public entities exhibited a

greater concern for the independence of financial reporting function than private entities during

the post-SOX time period. However, Table 7 also hints at the possibility of misclassification

based on the public/private distinction. For instance, if responses of ―not applicable‖ effectively

imply that the respondents are not affected by SOX, then about 4% of public companies and 8%

of public BUs are misclassified as treatment observations. Similarly, if responses of ―agree‖

effectively imply that the respondents are directly or indirectly affected by SOX, then 20% of

private companies and 28% of private BUs are misclassified as control observations.

       To assess the sensitivity of our results to these misclassifications, we present Table 8

where we reestimate our earlier results from Column 1 of Table 6 (reproduced in Column 1 of



                                                 33
Table 8). In Column 2, we eliminate 99 private entities that agreed with the SOX-related

statement in the survey and 9 public entities that viewed the same statement as not applicable. In

Column 3, we reclassify the 99 private entities as part of the treatment group and the 9 public

entities as part of the control group. We note that our results are effectively identical across all

three columns of Table 8 suggesting that the misclassification of treatment and control

observations is not a salient concern.

                                           [Insert Table 8]


5   Summary

In this paper, we use survey evidence of CFO compensation practices to examine how firms

provide incentives and measure performance of their chief financial officers. Based on a

principal-agent characterization of the firm-CFO relationship, we predict that firms’ reliance on

financial performance measures in evaluating CFOs reflects the costs and benefits of CFOs’

decision-making responsibilities as well as their fiduciary duties to oversee the financial

reporting process. In particular, we find that firms rely less on financial performance measures in

settings where CFOs’ fiduciary duties are important, but nevertheless continue to rely on such

measures when they are useful to motivate better decision-making.

       Our cross-sectional findings are broadly consistent with our theoretical predictions. For

instance, we find some evidence that the incentive weight on financial performance measures in

CFO bonus plans is positively (negatively) associated with proxies that capture the importance of

CFO decision-making (fiduciary) responsibilities. We also find that CFO and CEO incentives are

positively correlated, a finding that in part can be attributed to CFOs’ fiduciary duties to oversee

the financial reporting process.




                                                  34
       We also examine how firms’ reliance on financial performance measures in evaluating

CFOs changed from 2003 to 2007 (i.e., the post-SOX environment). In contrast to some prior

studies, we find that public companies (relative to private companies) have reduced the

percentage of CFO bonuses contingent on financial performance during this time period. To the

extent that SOX has made firms much more concerned about the integrity of their financial

reports and increased the costs of non-compliance, this finding is consistent with our model

which predicts that firms offer their CFOs muted incentives in order to motivate them to focus

more on their fiduciary responsibilities. In this sense, our study suggests that the design of CFO

compensation is an important part of internal controls firms rely on to limit earnings

management and other misreporting practices.

       Our study is subject to a number of limitations. First, we are unable to augment or

validate our survey measures with additional public data since the identity of participating firms

is unknown. Second, our sample is not random and the profile of the surveyed entities is different

from those considered in prior literature; hence our results need not easily generalize. Third, by

the very nature of survey data, we rely on our respondents to recall information from the past and

to accurately communicate sensitive information about their own and other executives’

compensation. Although relying on survey responses invariably introduces some measurement

error, it is unlikely to be the cause of a systematic bias. Despite these limitations, we believe our

study provides unique evidence on the determinants and evolution of CFO evaluation and

compensation practices.




                                                 35
References

ABERNETHY, M. A., J. BOUWENS, and L. VAN LENT. "Determinants of Control System Design in
       Divisionalized Firms." Accounting Review 79 (2004): 545–570.

AIER, J. K., J. COMPRIX, M. T. GUNLOCK, and D. LEE. "The Financial Expertise of CFOs and
       Accounting Restatements." Accounting Horizons 19 (2005): 123–135.

ASHBAUGH-SKAIFE, H., D. W. COLLINS, W. R. KINNEY, and R. LAFOND. "The Effect of SOX Internal
       Control Deficiencies on Firm Risk and Cost of Equity." Working paper, University of
       Iowa, 2007.

BABER, W. R., S. N. JANAKIRAMAN, and S. H. KANG. "Investment Opportunities and the Structure of
       Executive Compensation." Journal of Accounting & Economics 21 (1996): 297–318.

BEASLEY, M. S. "An Empirical Analysis of the Relation between the Board of Director
       Composition and Financial Statement Fraud." The Accounting Review 71 (1996): 443–
       465.

BERTRAND, M., E. DUFLO, and S. MULLAINATHAN. "How Much Should We Trust Differences-in-
       Differences Estimates?" Quarterly Journal of Economics 119 (2004): 249–275.

BURKS, J. J. "Disciplinary Measures in Response to Restatements after Sarbanes-Oxley." Working
       paper, University of Notre Dame, 2008.

CARTER, M. E., L. J. LYNCH, and S. L. ZECHMAN. "Changes in Bonus Contracts in the Post-
       Sarbanes-Oxley Era." Review of Accounting Studies (2007): forthcoming.

CHANG, H., H. L. CHOY, and K.-M. WAN. "Effect of the Sarbanes-Oxley Act on CEOs' Stock
       Ownership and Pay-Performance Sensitivity." Working paper, Drexel University, 2008.

CHEN, H., D. JETER, and Y.-W. YANG. "Regulatory Commitment to Auditing: The Impact of
       Sarbanes-Oxley on Pay-Performance Sensitivity and Managerial Effort." Working paper,
       Venderbilt University, 2007.

COHEN, D. A., A. DEY, and T. Z. LYS. "The Sarbanes Oxley Act of 2002: Implications for
       Compensation Contracts and Managerial Risk-Taking." Working paper, Northwestern
       University, 2007.

COLLINS, D., A. MASLI, A. REITENGA, and J. M. SANCHEZ. "Earnings Restatements, the Sarbanes-
       Oxley Act and the Disciplining of Chief Financial Officers." Journal of Accounting,
       Auditing, and Finance (2008): forthcoming.

CROCKER, K. J., and J. SLEMROD. "The Economics of Earnings Manipulation and Managerial
       Compensation." RAND Journal of Economics 38 (2007): 698–713.




                                                36
DE FRANCO, G., Y. GUAN, and H. LU. "The Wealth Change and Redistribution Effects of Sarbanes-
       Oxley Internal Control Disclosures." Working paper, University of Toronto, 2005.

DECHOW, P. M., R. G. SLOAN, and A. P. SWEENEY. "Causes and Consequences of Earnings
       Manipulations: An Analysis of Firms Subject to Enforcement Actions by the SEC."
       Contemporary Accounting Research 13 (1996): 1–36.

DYE, R. A. "Earnings Management in an Overlapping Generations Model." Journal of Accounting
       Research 26 (1988): 195–235.

EVANS, J. H., and S. SRIDHAR. "Multiple Control Systems, Accrual Accounting, and Earnings
       Management." Journal of Accounting Research 34 (1996): 45–65.

FELTHAM, G. A., and J. XIE. "Performance Measure Congruity and Diversity in Multi-Task
       Principal-Agent Relations." Accounting Review 69 (1994): 429–453.

GEIGER, M. A., and D. S. NORTH. "Does Hiring a New CFO Change Things? An Investigation of
       Changes in Discretionary Accruals." Accounting Review 81 (2006): 781–809.

GOLDMAN, E., and S. L. SLEZAK. "An Equilibrium Model of Incentive Contracts in the Presence of
       Information Manipulation." Journal of Financial Economics 80 (2006): 603–626.

GORE, A. K., S. MATSUNAGA, and P. E. YEUNG. "The Relation between Financial Monitoring and
       Incentives for Chief Financial Officers." Working paper, University of Oregon, 2007.

GUPTA, A. K., and V. GOVINDARAJAN. "Business Unit Strategy, Managerial Characteristics, and
       Business Unit Effectiveness at Strategy Implementation." Academy of Management
       Journal 27 (1984): 25–41.

HAMMERSLEY, J. S., L. A. MYERS, and C. SHAKESPEARE. "Market Reactions to the Disclosure of
       Internal Control Weaknesses and to the Characteristics of Those Weaknesses under
       Section 302 of the Sarbanes Oxley Act of 2002." Review of Accounting Studies 13
       (2008): 141–165.

HENNES, K., A. LEONE, and B. MILLER. "Management Intent and CEO and CFO Turnover around
       Earnings Restatements: Evidence from the Post-Enron Era." Working paper,
       Pennsylvania State University, 2007.

HOGAN, C. E., and M. S. WILKINS. "Evidence on the Audit Risk Model: Do Auditors Increase Audit
       Fees in the Presence of Internal Control Deficiencies?" Contemporary Accounting
       Research 25 (2008): 219–242.

HOITASH, U., R. HOITASH, and K. M. JOHNSTONE. "The Effects of Internal Control Quality, CFO
       Characteristics, and Board of Director Strength on CFO Annual Compensation."
       Working paper, Rutgers University, 2007.




                                              37
HOLMSTRÖM, B., and P. MILGROM. "Multitask Principal Agent Analyses - Incentive Contracts,
       Asset Ownership, and Job Design." Journal of Law Economics & Organization 7 (1991):
       24–52.

INDJEJIKIAN, R. J., and M. MATĚJKA. "Organizational Slack in Decentralized Firms: The Role of
       Business Unit Controllers." Accounting Review 81 (2006): 849–872.

KLEIN, A. "Audit Committee, Board of Director Characteristics, and Earnings Management."
       Journal of Accounting & Economics 33 (2002): 375–400.

LI, C., L. SUN, and M. ETTREDGE. "Financial Executive Quality, Financial Executive Turnover, and
       Adverse SOX 404 Opinions." Working paper, University of Kansas, 2007.

MEYER, B. D. "Natural and Quasi-Experiments in Economics." Journal of Business & Economic
       Statistics 13 (1995): 151–161.

MIAN, S. "On the Choice and Replacement of Chief Financial Officers." Journal of Financial
       Economics 60 (2001): 143–175.

MURPHY, K. J., and J. L. ZIMMERMAN. "Financial Performance Surrounding CEO Turnover."
       Journal of Accounting & Economics 16 (1993): 273–315.

PALIGOROVA, T. "Corporate Governance and Executive Pay: Evidence from a Recent Reform."
       Working paper, CERGE-EI, 2007.

POURCIAU, S. "Earnings Management and Nonroutine Executive Changes." Journal of
       Accounting & Economics 16 (1993): 317–336.

SIEGEL, G., and J. E. SORENSEN. Counting More, Counting Less: Transformation in the
       Management Accounting Profession: The 1999 Practice Analysis of Management
       Accounting. Montvale, N. J.: Institute of Management Accountants, 1999.

SMITH, C. W., and R. L. WATTS. "The Investment Opportunity Set and Corporate Financing,
       Dividend, and Compensation Policies." Journal of Financial Economics 32 (1992): 263–
       292.

WANG, X. "Risk, Board Monitoring and Pay-for-Performance Relation: Evidence from Chief
       Financial Officer Compensation before and after the Corporate Governance Reform."
       Working paper, Emory University, 2006.

WOOLDRIDGE, J. M. Econometric Analysis of Cross Section and Panel Data. Cambridge: The MIT
       Press, 2002.

ZHANG, I. X. "Economic Consequences of the Sarbanes-Oxley Act of 2002." Journal of
       Accounting & Economics 44 (2007): 74–115.




                                               38
Table 1—Descriptive Statistics

                      N          Mean          St. Dev.        25th Pct.        Median         75th Pct.

Panel A: Public companies (PUBLIC_CORP)
SALARY                 151        277,764         147,780          180,000         246,403         340,000
BONUS06                143        170,171         210,212           30,000          88,000         247,653
BONUS07                 91        184,205         177,908           55,000         120,000         243,750
LONGTR06               151         40,083         182,063                0               0                0
EQUITY06               151        264,161         464,656                0          95,800         267,000
GROWTH                 134          16.31           22.62             4.00            11.0            20.0
LDEBT                  130           0.12            0.10             0.06            0.09            0.19
INREC                  130           0.28            0.20             0.15            0.21            0.39
TENURE                 169           5.91            4.98             2.00            4.00              8.0
CEOTURN                170           0.24            0.43             0.00            0.00            0.00
ACCRUAL                130           0.08            0.05             0.06            0.07            0.09
INVOLVE                141           2.80            1.06             2.00            3.00            3.00
SIZE (unlogged)        170          7,642          22,041              240           1,160           5,400

Panel B: BUs of public companies (PUBLIC_BU)
SALARY                 126        158,471          69,438          110,000         145,000         190,000
BONUS06                123         69,967          87,728           22,500          35,000          90,000
BONUS07                 86         62,744          70,165           25,000          42,500          80,000
LONGTR06               127         14,130          52,002                0               0                0
EQUITY06               127         44,650         184,629                0           3,400          25,000
GROWTH                 130          11.71           15.11             4.00            10.0            18.0
LDEBT                  112           0.13            0.09             0.06            0.09            0.19
INREC                  112           0.31            0.19             0.16            0.30            0.41
TENURE                 132           5.93            5.70             2.00            4.00              7.3
CEOTURN                132           0.33            0.47             0.00            0.00            1.00
ACCRUAL                112           0.07            0.04             0.05            0.07            0.09
INVOLVE                127           3.16            1.56             2.00            3.00            5.00
SIZE (unlogged)        132          1,637           2,870              170             500           1,775
SALARY—annual CFO salary in 2006, BONUS06—annual CFO bonus earned in 2006, BONUS07—CFO bonus
expected in 2007 if performance exactly meets targets on all measures, LONGTR06—long-term cash compensation
earned in 2006, EQUITY06—equity-based compensation earned in 2006, GROWTH—percentage increase in
annual sales in 2006, LDEBT—average long-term debt over total assets during 2002–2006 (SIC-2 industry median);
INREC—average of inventory plus receivables over total assets during 2002–2006 (SIC-2 industry median),
TENURE—number of years CFO has been on the job, CEOTURN—equals one if a new CEO has been appointed in
the last two years, ACCRUAL—absolute value of total accruals in 2006 over total assets (SIC-2 industry median),
INVOLVE—higher-level involvement in selecting CFO performance measures, SIZE—the log of the number of
employees in a company or BU.


                                                      39
Table 1 (Continued)

                         N          Mean            St. Dev.           25th Pct.    Median     75th Pct.

Panel C: Private companies (PRIVATE_CORP)
SALARY                   835         153,064            74,185            100,000    135,000      185,000
BONUS06                  737          63,522           191,273             10,000     30,000       70,000
BONUS07                  498          72,402            94,248             20,000     40,500       85,000
LONGTR06                 840          16,622           104,774                  0          0            0
EQUITY06                 840          32,025           230,389                  0          0            0
GROWTH                   790           15.49             26.13               4.00       10.0         18.0
LDEBT                    751            0.16              0.11               0.07       0.14         0.22
INREC                    751            0.33              0.22               0.16       0.29         0.49
TENURE                   929            7.97              6.49               3.00       6.00         11.0
CEOTURN                  935            0.11              0.32               0.00       0.00         0.00
ACCRUAL                  748            0.08              0.05               0.06       0.07         0.09
INVOLVE                  755            1.97              1.08               1.00       2.00         3.00
SIZE (unlogged)          922             654             2,041                 78        175          450

Panel D: BUs of private companies (PRIVATE_BU)
SALARY                   113         125,204            63,936             86,000    110,000      140,000
BONUS06                  104          33,654            42,597              7,150     20,000       39,950
BONUS07                   70          43,715            45,777             15,000     31,000       50,000
LONGTR06                 115           5,662            24,702                  0          0             0
EQUITY06                 115           3,941            16,498                  0          0             0
GROWTH                   107           15.88             26.52               2.30       10.0         20.0
LDEBT                     93            0.15              0.10               0.07       0.14         0.20
INREC                     93            0.30              0.19               0.18       0.22         0.41
TENURE                   115            5.64              4.20               2.00       4.50           8.0
CEOTURN                  116            0.21              0.41               0.00       0.00         0.00
ACCRUAL                   93            0.07              0.03               0.05       0.07         0.09
INVOLVE                  104            3.07              1.53               1.00       3.00         5.00
SIZE (unlogged)          114             309               569                 33        111          365
See notes to Panels A and B of this table for variable descriptions.




                                                          40
Table 2—Percentage of CFO and CEO Bonus Contingent on Different Types of
        Performance Measures

                       N         Mean           St. Dev.        25th Pct.        Median          75th Pct.

Panel A: Public companies (PUBLIC_CORP)
FIN07                  138            67.34           30.76             50.0            75.0             100
NFIN07                 138            15.45           20.87             0.00            0.00           25.00
SUB07                  138            17.21           29.30             0.00            0.00           25.00
FIN06                  129            64.32           35.37             50.0            75.0             100
NFIN06                 129            14.60           22.85             0.00            0.00           25.00
SUB06                  129            21.07           34.20             0.00            0.00           25.00
FIN03                   88            67.46           35.85             50.0            75.0             100
NFIN03                  88            11.59           21.08             0.00            0.00            22.5
SUB03                   88            20.95           35.52             0.00            0.00            25.0
FIN06_CEO              120            66.44           35.61             50.0            75.0             100
NFIN06_CEO             120            15.07           24.77             0.00            0.00            25.0
SUB06_CEO              120            18.49           32.17             0.00            0.00            25.0
Panel B: BUs of public companies (PUBLIC_BU)
FIN07             127         56.63      35.20                          25.0            50.0            90.0
HLF07             127         14.69      23.72                          0.00            0.00            25.0
NFIN07            127         13.31      20.49                          0.00            0.00            20.0
SUB07             127         15.37      26.86                          0.00            0.00            25.0
FIN06             125         54.02      37.09                          20.0            50.0            90.0
HLF06             125         15.72      25.83                          0.00            0.00            25.0
NFIN06            125         12.70      20.77                          0.00            0.00            20.0
SUB06             125         17.56      29.41                          0.00            0.00            25.0
FIN03               87        68.27      35.89                          50.0            80.0             100
HLF03               87         7.82      20.14                          0.00            0.00            0.00
NFIN03              87         9.57      19.64                          0.00            0.00            10.0
SUB03               87        14.35      29.29                          0.00            0.00            10.0
FIN06_CEO         104         63.98      32.87                          40.0            72.5             100
HLF06_CEO         104         16.54      23.63                          0.00            0.00            25.0
NFIN06_CEO        104         11.77      18.86                          0.00            0.00            20.0
SUB06_CEO         104          7.70      18.12                          0.00            0.00            0.00
Percentages of CFO and CEO bonuses linked to financial performance targets (FIN), higher-level financial
performance (HLF), non-financial performance targets (NFIN), and subjective evaluation (SUB). FIN07, HLF07,
NFIN07, SUB07 are percentages expected if 2007 performance exactly meets all targets, the other percentages
reflect performance in 2006 and in 2003. For example, FIN06 is the percentage of CFO bonus in 2006 earned for
meeting financial performance targets (FIN06_CEO is the equivalent for CEOs).


                                                      41
Table 2 (Continued)

                         N          Mean            St. Dev.           25th Pct.    Median     75th Pct.

Panel C: Private companies (PRIVATE_CORP)
FIN07                    744            50.78             39.43              0.00       50.0          100
NFIN07                   744            13.89             22.21              0.00       0.00         25.0
SUB07                    744            35.33             40.60              0.00       20.0         75.0
FIN06                    714            47.48             40.82              0.00       50.0          100
NFIN06                   714            13.32             23.35              0.00       0.00         25.0
SUB06                    714            39.20             42.50              0.00       24.0          100
FIN03                    549            48.25             41.88              0.00       50.0          100
NFIN03                   549            11.57             22.66              0.00       0.00         20.0
SUB03                    549            40.19             43.78              0.00       25.0          100
FIN06_CEO                611            53.94             42.26              0.00       55.0          100
NFIN06_CEO               611            11.27             21.60              0.00       0.00         20.0
SUB06_CEO                611            34.79             42.86              0.00       0.00          100
Panel D: BUs of private companies (PRIVATE_BU)
FIN07                    101            37.44             35.32              0.00       30.0         60.0
HLF07                    101            15.41             28.22              0.00       0.00         25.0
NFIN07                   101            19.56             28.31              0.00       0.00         25.0
SUB07                    101            27.58             36.64              0.00       0.00         50.0
FIN06                     95            34.59             36.14              0.00       25.0         55.0
HLF06                     95            15.38             27.00              0.00       0.00         25.0
NFIN06                    95            18.41             29.10              0.00       0.00         25.0
SUB06                     95            31.61             38.54              0.00       20.0         55.0
FIN03                     64            40.05             41.38              0.00       27.5         80.0
HLF03                     64             8.80             23.29              0.00       0.00          0.0
NFIN03                    64            16.41             28.53              0.00       0.00         25.0
SUB03                     64            34.75             42.13              0.00        5.0         87.5
FIN06_CEO                 83            42.22             38.18              0.00       40.0         75.0
HLF06_CEO                 83            16.24             27.87              0.00       0.00         25.0
NFIN06_CEO                83            15.75             25.91              0.00       0.00         25.0
SUB06_CEO                 83
                           0            25.80
                                         0.00             36.51
                                                           0.00              0.00       0.00         30.0
                                                                                                     0.00
See notes to Panels A and B of this table for variable descriptions.




                                                          42
   Table 3—Frequency of Different Types of Non-Financial Performance Measures in a
           Subsample of CFO Bonus Plan

                                                 PUBLIC            PUBLIC           PRIVATE          PRIVATE
                                                  CORP               BU              CORP               BU
                                                   N=39              N=41             N=167             N=31
  I) General management targets                    51.3%            65.9%             56.9%             71.0%

    Operations                                      7.7%             31.7%            23.4%             38.7%
    People management                              10.3%             14.6%            17.4%             22.6%
    Strategic management                           25.6%             14.6%            12.6%              6.5%
    Customer orientation                           12.8%              9.8%            15.0%             16.1%

  II) Reporting targets                            35.9%            31.7%             35.9%             35.5%
    Compliance & control                           25.6%             17.1%             8.4%             19.4%
    Efficiency of reporting                         5.1%             12.2%            16.8%              6.5%
    IT & systems                                    7.7%              9.8%            12.0%              9.7%
    Internal decision support                       2.6%              2.4%             7.2%              3.2%

  III) Financing targets                           30.8%            22.0%             16.2%             16.1%
    Short-term financial management                12.8%              9.8%             5.4%              6.5%
    Mergers & acquisitions                          5.1%             12.2%             6.0%              6.5%
    Obtaining capital                              15.4%              0.0%             5.4%              3.2%

  IV) Communication & teamwork                     5.1%              4.9%              7.8%             6.5%
Tabulated categories include the following examples:
I) General management: Operations—efficiency, quality, safety, process improvement, or cost control; People
management—employee turnover, staff development, recruiting, contribution to performance evaluations, or
corporate culture; Strategic management—market share, business development milestones, business growth, R&D
goals, implementation of profitability strategies, restructuring, organizational structure, or design of incentives;
Customer orientation—customer satisfaction, on-time delivery, community involvement, supplier relations, or PR
initiatives.
II) Reporting targets: Compliance & control—clean audit, no restatements, external/tax reporting, SOX
compliance, documentation of procedures, risk management, or regulatory reporting; Efficiency of reporting—
timeliness, accuracy, lean accounting, reporting in new acquisitions, or accounting department efficiency; IT &
systems—ERP implementation, software upgrades, or IT projects; Internal decision support—working on projects
for BUs, useful information for management, management satisfaction, or setting BU performance targets.
III) Financing targets: Short-term financial management—collections, cash-flow, working capital, financial
condition, debt covenants, or forecasting; Mergers & acquisitions—divestitures, deals, or capital investment
evaluations; Obtaining capital—banking/investor relations, communication with analysts, or IPO.
IV) Communication & teamwork—leadership, interpersonal skills, interaction with managers or board members,
loyalty, or work ethics.




                                                        43
Table 4—Tobit Models of the Incentive Weight on Financial Performance Measures

                           Expected               Column 1               Column 2             Column 3
  Variables                  Sign                 FINBON07                FIN07               FINBON06

  Intercept                                             -28.59                  -6.84               -78.10
                                                         (.009)                (.660)                (.000)
  PUBLIC_CORP                                            11.84                 11.66                 12.08
                                                         (.080)                (.160)                (.012)
  PUBLIC_BU                                               -4.10                 -1.62                  5.42
                                                         (.478)                (.827)                (.296)
  PRIVATE_BU                                              -5.43               -20.05                  -3.41
                                                         (.451)                (.008)                (.493)
  GROWTH                        +                         0.12                  -0.14                  0.11
                                                        (.083)                 (.270)                (.164)
  LDEBT                         +                       47.30                 -19.60                 16.85
                                                        (.058)                 (.413)                (.390)
  INREC                         +                         6.90                 20.09                   9.57
                                                        (.606)                 (.040)                (.229)
  TENURE                        +                         6.25                   9.01                  5.37
                                                        (.042)                 (.001)                (.013)
  CEOTURN                       -                         -9.21               -11.32                  -8.77
                                                         (.028)                (.071)                (.004)
  ACCRUAL                       -                       -23.42                -48.75                -66.58
                                                         (.344)                (.178)                (.000)
  INVOLVE                       +                          0.92                  8.15                  2.68
                                                         (.591)                (.002)                (.032)
  SIZE                                                    5.60                   6.38                  7.54
                                                        (.000)                 (.000)                (.000)
  PERF06                                                                                               8.18
                                                                                                     (.000)
  Log-likelihood value                                  -2,300                -2,642                -2,837
  R-squared                                               .085                  .075                   .185
  ˆ                                                     41.28                 63.41                 37.95
  N                                                        529                   790                    792
Two-sided p-values in brackets (based on Huber-White standard errors adjusted for clustering by SIC-2
industry categories). R-squared is the square of the correlation coefficient between actual and fitted values of
the dependent variables (Wooldridge [2002]).
FINBON07—CFO 2007 bonus based on financial performance measures as a percentage of CFO salary in
2006 (as expected if all targets are met); FIN07—percentage of CFO 2007 bonus based on financial
measures; FINBON06—CFO 2006 bonus based on financial measures as a percentage of CFO salary in
2006; PERF06—2006 performance relative to budget. Other variables defined in Table 1.


                                                     44
Table 5—OLS Estimation of the Effect of SOX on Incentive Weights on Financial
        Performance Measures

              yij   0  1 ×Y07   2 × BU  3 × PUBLIC   4 ×Y07× PUBLIC   ij


                                                                            FIN
     Variables                               Column 1                  Column 2                 Column 3

     Intercept                                    48.34                    48.12                     53.37
                                                  (.000)                   (.000)                    (.000)
     Y07                                            1.94                     1.88                      4.52
                                                  (.111)                   (.138)                    (.015)
     BU                                            -9.09                   -10.56                     -3.76
                                                  (.002)                    (.002)                   (.550)
     PUBLIC                                       24.04                    25.62                     20.58
                                                  (.000)                   (.000)                    (.004)
     Y07 · PUBLIC                                  -7.75                    -8.88                   -16.76
                                                  (.001)                   (.003)                    (.000)

     Adjusted R2                                    .036                     .025                      .008
     N                                             1898                     1620                        663

   Two-sided p-values in brackets (based on Huber-White standard errors adjusted for clustering by entities).
   FIN—weight on financial performance measures in CFO bonus (pooled FIN03 and FIN07 observations),
   Y07—indicator variable for 2007, BU—indicator variable for BU-level entities (of both public and private
   companies), PUBLIC—public entities (both at the corporate and BU level).
   Column 1 reports results for all available observations.
   Column 2 reports results for a sample excluding observations where CFOs earn significant amount of
   equity compensation (defined as $50,000 or more in 2006).
   Column 3 further excludes observations where CFO bonuses are equal or less than 30% of salary.




                                                       45
Table 6—OLS Regressions of Changes in Weights on Financial Performance Measures
        between 2007 and 2003

                                                                FIN07-03
     Variables                            Column 1               Column 2                Column 3

     Intercept                                  2.90                    2.78                   3.47
                                              (.008)                  (.014)                 (.043)
     PUBLIC_CORP                               -0.89                   -8.34                 -13.26
                                              (.746)                  (.013)                  (.007)
     PUBLIC_BU                                 -7.62                   -5.65                 -13.70
                                              (.004)                  (.036)                  (.001)
     PRIVATE_BU                                -1.62                   -1.35                  -1.47
                                              (.681)                  (.742)                 (.726)

     Adjusted R2                                .005                    .005                   .032
     N                                           647                    560                     254

   Two-sided p-values in brackets (based on White heteroskedasticity-adjusted standard errors).
   FIN07-03—2007 weight on financial performance measures in CFO bonus minus the weight in 2003. Other
   variables defined in Table 1.
   Column 1 reports results for all available observations.
   Column 2 reports results for a sample excluding observations where CFOs earn significant amount of
   equity compensation (defined as $50,000 or more in 2006).
   Column 3 further excludes observations where CFO bonuses are equal or less than 30% of salary.




                                                   46
Table 7—The Effect of SOX as Perceived by Different Types of Entities


                                                  N                   Frequency of responses
                                                                  (i)          (ii)          (iii)

 PUBLIC_CORP                                     140             3.6%            46.4%            50.0%
 PUBLIC_BU                                       131             8.4%            52.7%            38.9%
 PRIVATE_CORP                                    802            39.7%            19.6%            40.8%
 PRIVATE_BU                                      115            35.7%            27.8%            36.5%

Tabulated is the distribution of responses indicating whether respondents agree with the following statement:
―Since the enactment of the Sarbanes-Oxley legislation, we have greatly increased the independence of the
financial reporting function in our firm [BU].‖ Column (i) shows the percentage of respondents answering ―not
applicable.‖ Column (ii) shows the percentage of respondents who agree with the statement. Column (iii) shows
the percentage of other responses.




                                                      47
Table 8—Changes in Incentive Weights between 2007 and 2003 (Robustness Check)


                                                                   FIN07-03
   Variables                              Column 1                  Column 2                  Column 3

   Intercept                                     2.90                      2.78                     2.64
                                               (.008)                    (.021)                   (.025)
   PUBLIC_CORP                                  -0.89                     -0.74                     0.57
                                               (.746)                    (.793)                   (.799)
   PUBLIC_BU                                    -7.62                     -8.94                    -6.64
                                               (.004)                    (.001)                   (.027)
   PRIVATE_BU                                   -1.62                     -2.67                    -2.56
                                               (.681)                    (.541)                   (.470)

   Adjusted R2                                   .005                      .009                     .009
   N                                              647                      539                       647

Two-sided p-values in brackets (based on White heteroskedasticity-adjusted standard errors).
FIN07-03—2007 weight on financial performance measures in CFO bonus minus the weight in 2003.
Column 1 reproduces the results from Column 1a of Table 7.
Column 2 excludes 99 private entities that agreed with the SOX-related statement (see Table 7) and
9 public entities that did not consider the statement applicable.
Column 3 reclassifies the 99 private entities as the treatment group, PUBLIC, and the 9 public entities as
the control group, PRIVATE (thus, to economize on space, PUBLIC in Column 3 stands both for public
entities and private entities that agreed with the SOX-related statement; similarly, PRIVATE also includes
some public entities that did not consider the statement applicable).




                                                    48
Appendix A—Detailed Derivations and Proofs

For the model described in section 2, the firm’s optimization program can be stated as

                                                      
           MAX                   E  x  1  2   wi 
                                         2
                                                                                                       (A1)
           i , i , i ,i
           i  CEO , CFO                         i    

subject to (1) the incentive compatibility constraints that CEO ={eCEO } and CFO ={eCFO , q, h}

maximize the expected utilities of the CEO and CFO respectively, and (2) the incentive

rationality constraints that the contracts be acceptable to both managers.

        To characterize the incentive compatibility constraints, we write the managers’ wages as

        wi   i  i R   iZ                      i  CEO, CFO                                      (A2)

where  i is the fixed component of a manager’s compensation, (  i ,  i) are the incentive

coefficients, R is reported profit as described earlier in section 2, and Z is the set of all

measures (other than R ) that the firm may use to evaluate its managers. To focus the analysis on

CFO compensation, we assume that Z  ( z1 , z2 ) is informative about CFO performance with z1

informative about eCFO and z 2 informative about q ; in particular, we assume z1 and z 2 are

independent normally distributed variables with means (eCFO , q) and variances of ( z21 ,  z22 ) .

        Since the CEO solves

                       MAX E ( wCEO )  1 eCEO  rCEO Var ( wCEO )
                                        2
                                           2
                                                   2
                                                                                                       (A3)
                              eCEO



and the CFO solves

                       MAX E ( wCFO )  1 eCFO  1 c(1  q) 2  h /(1  h)  rCFO Var ( wCFO ) ,
                                        2
                                           2
                                                 2                             2
                                                                                                       (A4)
                        eCFO , q , h



we have,

                       eCEO   CEO bCEO and eCFO  CFObCFO   CFO1
                        *                     *
                                                                                                       (A5)


                                                             49
                                     k CFO   CFO 2
                                                                                                                   
                                                                                                                        1/ 2
                     1  q*                                        and             1  h*        rCFO
                                                                                                     2    CFO 2
                                                                                                           2
                                                                                                                                     (A6)
                                               c

with the second-order conditions easily satisfied.

           To characterize the incentive rationality constraints, we assume (without loss of

generality) both managers’ reservation certainty equivalents are zero which allows us to write the

managers’ aggregate wages as


             w   e
             i
                 i
                     1
                     2
                          i
                                 2
                                 i    ri i2[2   2 (1  h)]  ri i2 z21  ri i22 z22  1 c(1  q)2  h /(1  h) .
                                                                        1                        2                                  (A7)


Solving (A1) subject to (A5), (A6), and (A7), the first-order conditions are:


            bCEO  CEO bCEO  rCEO2  rCEO 2 (1  h* )  0
             2
                        
                          2
                                                                                                                                   (A8a)


                                                         rCEO 2 (1  h* ) CEO
                                                                            2
            bCFO  bCFO CFO 1   CFO 2 kD 
             2

                                                                2CFO                                                               (A8b)
                                                 CFO bCFO  rCFO 2  k 2 D  rCFO 2 (1  h* )   0
                                                       
                                                         2
                                                                                                    


            bCFO  bCFO CFO   CFO1 (1  rCFO z21 )  0                                                                          (A8c)


            CFOkD   CFO 2 (D  rCFO z22 )  0                                                                                   (A8d)


                                                              1  k             and 
                                          
                                               1/ 2                    2
                                                                            /c
where (1  h* ) 
                          rCFO
                            2
                                 CFO 2
                                  2
                                                       , D                                CEO 1     CEO 2  0 . Further substitution
                                                                   c

and simplification yields the following two conditions: 23


                              2                   rCEO      
            J1  bCEO  CEO bCEO  rCEO 2 
                  2
                                                              0                                                                   (A9a)
                             
                                               CFO rCFO / 2 
                                                              



23
     The subscripts denote partial derivatives with respect to CEO and CFO respectively.



                                                                            50
                     2
                    bCFO          rCEOCEO
                                        2
         J2                                   2rCFO 2
              1  (rCFO z1 )
                         2 1
                                2CFO rCFO / 2
                                   2

                                                                                                     .    (A9b)
                                                   2
                                                   bCFO            k2                        
                                      CFO                  1                  rCFO 2   0
                                            1  (rCFO z1 )  D  (rCFO z22 ) 1
                                                        2 1
                                                                                             


We note that the second-order conditions for the program are also easily satisfied since J11  0 ,

J 22  0 and J11 J 22  J12  0 .
                          2




PROOF OF PROPOSITION 1

The comparative statics presented in Proposition 1 can be expressed as:


         d CFO  J11 J 2bCFO  J12 J1bCFO
                                           0 , since J 2bCFO  0 .                                     (A10a)
         dbCFO        J11 J 22  J12
                                   2




         d CFO  J11 J 2bCEO  J12 J1bCEO
                                           0 , since J1bCEO  0 and J12  0 .                          (A10b)
         dbCEO        J11 J 22  J12
                                   2




         d CFO  J11 J 2 k  J12 J1k
                                     0,         since J 2 k  0 .                                      (A10c)
           dk      J11 J 22  J12
                                2




         d  CFO  J11 J 2 c  J12 J1c
                                       0,       since J 2 c  0 .                                      (A10d)
           dc       J11 J 22  J12
                                 2




         d  CFO  J11 J 2   J12 J1
                                       0.       since J 2   0 .                                      (A10e)
           d       J11 J 22  J12
                                 2




These results hold when both z1 and z 2 are useful contracting variables as well as in the special

case considered in section 2 where z1 and z 2 are not used at all (i.e., when  z21   and

 z22   ). Finally, we can show that the comparative statics for the relative weights, CFO /  CFO 1

and CFO /  CFO 2 , are qualitatively similar.




                                                        51
Appendix B—Questionnaire Items

1a. What percentage of the 2006 annual bonus incentive component did you earn based on
    the achievement of
1b. If performance exactly meets targets on all measures, what percentage of your 2007
    annual bonus do you expect to be based on the achievement of
1c. What percentage of the 2003 annual bonus did you earn based on the achievement of

   Financial performance targets of the division

   Financial performance targets of higher levels
   (e.g., the firm, business group)

   Non-financial performance targets

   Achievements evaluated subjectively
   (i.e., without pre-set targets)

   Other, please specify below


2. How much did you earn through the following incentive components in 2006?
   (Fill in zero if an incentive opportunity was offered but did not pay out.)
   (Leave blank if no such incentive opportunity offered.)

   Annual bonus

   Long-term cash compensation plan

   Equity-based compensation

   Other, please specify below


3. Your annual salary in 2006 was

4. If performance exactly meets targets on all measures, your 2007 annual bonus should
   be about

5. Last year's annual sales growth was about         %

6. Please describe your industry.
   (For example: manufacturing—components for the auto industry; wholesale—industrial
   packaging products; management, IT, and strategy consulting)


                                                52
7. Who decides the following aspects of the CFO’s incentives?

   Selection of performance measures

Version for corporate-level respondents
   (CEO decision = The CEO decides with little or no consulting of the board of directors.
   Board decision = The board of directors decides with little or no consulting of the CEO.)
   Scale:
   CEO decision
   Largely CEO
   Joint decision
   Largely the Board
   Board decision
   N/A

Version for BU-level respondents
   (Division manager = The general manager at the division decides with little or no consulting
                        of the higher-level CFO/controller.
   Higher-level CFO = Higher-level CFO/controller decides with little or no consulting of the
                        general manager at the division.)
   Scale:
   Division manager
   Largely division manager
   Joint decision
   Largely higher-level CFO
   Higher-level CFO
   N/A

8. How well did the firm perform in the last three years relative to the budget?

   2006 performance
   2005 performance
   2004 performance

   Scale:
   Far below budget
   Below budget
   About the same as budget
   Above budget
   Far above budget
   N/A

9. What were the most important non-financial targets in your 2006 annual bonus?




                                              53

				
DOCUMENT INFO