Printer Friendly

An ex-post evaluation of Sarbanes-Oxley act on firms' intrinsic value: a principal-agent framework.


In 1998, the Chairman of the Securities and Exchange Commission (SEC) warned the financial community about aggressively managing reported earnings and making deliberate misstatements regarding firm performance because it was eroding the quality of financial reporting (Levitt 1998). Within a few years, a string of accounting scandals and forced restatements of financial reports emerged by very large companies among the Fortune 500, such as Enron, WorldCom, Tyco, HealthSouth, and Global Crossing who were seemingly oblivious to the warning. Managers had pursued their own self-interests despite the SEC's warning about the numbers game eroding the integrity of the financial reporting system. The problem manifested by material misrepresentations, inaccurate reporting by managers and ineffective auditing by auditors. It resulted in wiping out billions of dollars in shareholder wealth and, consequently, impaired the efficiency and liquidity of the US capital markets. Not only does such action lead to low investor confidence about the reliability of financial statements, it also impacts the intrinsic value of the firm.

In response, the U.S. Congress passed the Public Company Accounting Reform and Investor Protection Act of 2002, commonly called the Sarbanes-Oxley Act of 2002 (SOX), which was signed by the President into law on July 30, 2002, with the objective of restoring investor confidence in the quality of financial reporting (U.S. Congress 2002) and aligning the goals of management and stockholders. We assert that the uncertainty about managers telling the truth about the outcome of their reported actions is a moral hazard problem that makes it difficult for shareholder principals to evaluate the true profitability and risks of their capital investments. An agency problem exists when an agent puts their own self-interest ahead of the principals' (shareholders'), who compensates the agent to make reporting decisions that benefit the firm in the long run (i.e., Berle and Means 1932; Ross 1973; Jensen and Meckling 1976; Jensen and Murphy 1990). In the Pre-SOX period, we assert that two types of agents, the manager and the independent auditor, contribute to the moral hazard agency problem related to inaccurate financial reporting. First, if the manager-agent plays the numbers game and excessively manages earnings, then the quality of financial reporting erodes. (For example, in 2001 and 2002, many headlines linked Fortune 500 companies to a wave of accounting irregularities and securities fraud.) Consequently, such erosive and self-interested actions can destroy billions of dollars of shareholders wealth. Second, failure of the auditor-agent to conduct effective independent monitoring of its large audit clients and not obfuscate their interest with the client's may contribute to auditor-agent's own demise (e.g., Arthur Andersen LLP). Prior to SOX, however, each of the Big 5 accounting firms had several large audit clients under SEC investigation (Scott and Gist 2010).

The Sarbanes-Oxley Act is aimed at re-aligning the behaviors of the manager-agent and auditor-agent with those of the shareholder-principal by providing incentives for accurate financial reporting by managers and reliable assurance services by auditors to the shareholders. Whether SOX was effective in realigning the agents' goals in line with those of the shareholders is an empirical question, requiring ex-post comparison of risk and return data in a behavioral modeling framework. Some studies have focused on surveys to draw implications about the perceived benefits of SOX relative to the expected internal control compliance costs (Carney 2005; Solomon and Bryan-Low 2004 WSJ; Serwer 2006 Fortune; DaVay 2006; Bedard 2006). Other studies have examined the market reaction to SOX-related events (i.e., Zhang 2007, Li, Pincus and Rego 2008); documented changes in firm behavior by adopting strategies to avoid SOX requirements, such as going private or delisting (i.e., Leuz, Triantis and Wang 2008); and examined how SOX has affected managerial decisions related to discretionary accruals and corporate investments (i.e., Cohen Dye and Lys 2007; Bargeron, Lehn and Zutter 2009). The focus on incentives to realign behavior and goals between agents and principal and resolve the bigger moral hazard issues is yet to be addressed.

In this exploratory study, we develop a principal-agent model with risk considerations and evaluate the effectiveness of SOX on the firm's intrinsic value. We assume that audit fees and various forms of management compensation are used by the principal to induce high effort on the part of the agents. We also assume that in the post-SOX era firms will adopt either one of three earnings management strategies: avoid negative return on equity, target industry or market averages, or exceed industry or market expectations. It is important to note that auditors in the U.S. may not require firms to meet specific performance goals, but will evaluate performance vis-a-vis industry performance. A Mann Whitney U-test was used determine if distributions around the three earnings management targets were distinct. A random utility discrete choice model was used to empirically test the theoretical findings. The cluster approach used to examine earnings and evaluate the effectiveness of SOX is intuitive and direct. It does not require estimates of normal or abnormal accruals which are inevitably noisy. It is hypothesized that SOX, on average, will encourage earnings management towards the industry average and provide disincentives to managers to engage in excessive high risk, high return strategies. Such actions will have a positive impact on the intrinsic value of the firm and could be effective in minimizing the moral hazard problem related to the quality of financial reporting. This paper is organized as follows. Section 2 discusses agency theory under SOX and assessing goal congruence with SOX. Section 3 describes the theoretical framework. Section 4 describes the empirical method and data used. The results are presented in Section 5 and we conclude the paper in Section 6.


Agency Theory under SOX

The CEO is an agent of the shareholder-principal. The CEO-agent is hired to make decisions, to manage the corporation and increase the value of the firm. The independent auditor is also an agent of the shareholderprincipal.

The independent auditor-agent performs an audit and verifies the faithfulness of reported numbers from the manager-agent on behalf of the shareholder-principal. The monitoring role of the independent auditor is to give the shareholder-principal a level of assurance about the relevance and reliability of the financial reports prepared by the manager-agent.

The concern about the perception of auditor-agents seemingly acting in their own self-interest is due to (a) earning nontrivial revenues from non-audit services equal to or greater than the respective audit fees, and/or (b) acquiescing to management pressure on reporting issues and allowing manager-agents to aggressively manage earnings to the point of deliberate misstatements. Such uncertainty about manager-agent and auditor-agent behaviors creates a moral hazard problem and makes it difficult for the shareholder-principal to evaluate the true returns and risks of their capital investments. A major goal of Congress, through the implementation of SOX, was to minimize the agency problem that resulted in so many accounting irregularities.

We highlight a few changes to the corporate governance system (principal-agent relationship) required by SOX. (1) To implement SOX, a new federal agency called the Public Company Oversight Board (PCAOB) was created. The PCAOB has oversight responsibility of the independent accounting firms (Section 103). (2) SOX prohibited the independent accounting firms from providing certain non-audit services to their audit clients in an effort to enhance the perception of auditor independence in-fact and in-appearance (Section 201). This specific action reduces the future stream of revenues from non-audit services and lessens the auditors' motivation to give into management pressure. (3) Prior to SOX, the CEO would make the decision to retain the independent auditor and the auditor would report audit results to the CEO. Post SOX, it is the audit committee of the Board of Directors that makes the independent auditor retention decision and to whom the auditor reports (Section 204). (4) In addition, the Board's audit committee must preapprove all audits and permitted non-audit services (Section 202). (5) SOX contains provisions that require both the CEO and CFO to certify that, to the best of their knowledge, the financial report filings with the SEC are accurate and that non-deficient internal financial controls are in place (Section 302). (6) Now, intentional misrepresentation of financial reports and disclosures submitted to the SEC carries substantial criminal penalties of up to 20 years of imprisonment (Section 802). (7) Finally, independent auditors are required to audit the internal controls of the financial reporting system of the company (Section 404).

This study does not assess the adequacy of the internal control requirements of SOX. Rather, we focus on goal congruence, whether SOX effectively realigns the principal-agent relationship to improve the intrinsic value of the firm. Section 404 creates great controversy as executives complain about the current and expected internal control compliance costs to acquire the relative perceived benefits of SOX (i.e., Carney 2005; Solomon and Bryan-Low 2004 WSJ; Serwer 2006 Fortune; DaVay 2006). Bedard (2006) finds that the internal control requirements of Section 302 and Section 404 SOX improve the quality of reported earnings based on the magnitude of unexpected accruals. Various researchers and executives have complained that the expected hefty costs of internal control compliance may lead to higher expenses and lower earnings that may possibly affect their compensation. This study adds to the literature since little or no empirical evidence has been provided about the impact of SOX in resolving the agency problem and on firms' intrinsic value, true profitability and risks.

Assessing Goal Congruence and Earnings Management with SOX

Goal congruence is the desired outcome from the shareholder-principal's point of view. If the manager-agent is making decisions in the best interest of the shareholder-principal then goal alignment has occurred. There had been much discussion of how and whether corporations adequately solve the issue of motivating the manager-agent to act in the best interest of the principal (Garen 1994). "Management's goal should be to take actions designed to maximize the firm's intrinsic value, not its current market price," (Brigham and Houston 2007 p. 11). Proper action of producing accurate financial reports over the long run increases the credibility of management's financial reporting output over the long run. In the principal-agent relationship, the principal is at an information disadvantage as agents have better information about their actions that influence intrinsic value. The principal is unable to observe and verify every action of the agent. This modeling process views the incentives of two agents, the manager and the independent auditor.

In this study, we assume that both the agents and principal are risk averse, compared to traditional principal-agent models where only the principal is assumed to be risk averse. That is, the principal and agents prefer to take actions that avoid risk and negative outcomes such as managers incurring a criminal penalty for intentional inaccurate reporting or auditors impairing the reputation of their audit service. The underlying assumption is that achieving principal-agent goal congruence will lead to more accurate reporting about performance that, in turn, will enhance the quality of financial reporting. This will increase shareholders' intrinsic value and increase their confidence about the credibility of the reports. Accurate reporting is evaluated empirically by: 1) comparing actual and predicted returns prior, during, and after SOX implementation and 2) determining the marginal contribution of incentives and restatements in achieving the principal's goal, to maximize intrinsic value. If goal congruence is not obtained, then SOX will be ineffective in realigning actions of the agents with those of the principle and, consequently, will not enhance investor confidence about the quality of financial reporting. Managers are paid to manage earnings and for achieving desired operating performance targets such as net income and return on equity (ROE) measures. The successfulness of attaining the desired target measures has economic consequences for the managers in the form of cash, bonuses, and stock compensation. There are two common ways that managers can practice earnings management: accrual-based earnings management and real earnings management. Accrual-based earnings management occurs through the use of accounting estimates, such as bad debt expense and depreciation expense. The accounting estimates use various assumptions, based on the manager's judgment, in the calculation of particular expenses. (An example of a discretionary accrual based expense is the calculation of depreciation expense for a long-term asset. Depreciation expense is based on the manager's assumption of the estimated life and the estimated salvage value at the end of the estimate life for that long term asset and the choice of measurement method such as straight-line or accelerated.) In accordance with generally accepted accounting principles (GAAP), managers have the discretion to choose certain assumptions that may increase (or decrease) current reported earnings to achieve the targeted profitability metrics, such as net income and/or ROE. In contrast, real earnings management occurs through events and transactions that affect reported earnings, such as reporting lower cost of goods sold through increased production, or accelerating the timing of asset sales through increased price discounts. A substantial difference between accrual based and real earnings management is that discretionary estimations can be subsequently revised, whereas once a company sells an asset, it cannot undo this event without creating another real transaction. Aggressive (as opposed to conservative) earnings management tends to maximize current net income and return on equity measures, sometimes to the point of deliberate misstatements that overstate firm performance (i.e., Enron).

Prior studies find empirical evidence that managers are using less accrual based earnings management in favor of more real earnings management transactions after the passage of SOX, relative to the pre-SOX period (Lobo and Zhou 2009; Cohen, Dye and Lys 2007a). The choice of earnings management strategies may be motivated by section 802 of SOX that encourages more truth telling about the firm's actual performance and are dissuaded by the cost of intentional misrepresentation of financial performance of a fine plus up to 20 years of imprisonment. Relative to the pre-SOX period, Bargeron et al. (2009) and Cohen et al. (2007) document that managers choose less risky investment projects in the post-SOX period. The economic consequences of less risky investment projects decrease the potential of the firm to earn high returns relative to more risky investment projects, which may also have dampened performance measures in the post-SOX period. Bargeron et al. (2009) argue that the declines in major, new risky investment projects are partly due to managers' preoccupation with the risks and consequences of making a mistake. The Sarbanes-Oxley Act does not provide direction to the manager-agent in the choice of earnings management strategies or in the selection of alternative investment projects; rather, the goal of SOX is to prevent the issuance of misleading reports about the outcome of those choices (i.e., to reduce the moral hazard problem). However, comparison to industry averages and managing earnings to avoid negative returns are common practices (see Figure 1)

Major Factors affecting Goal Congruence

The main purpose of SOX is to protect current and future investors by improving the accuracy and reliability of corporate disclosures that will lead to restoring investor confidence in the integrity of financial reports prepared by the manager-agent.

Financial Restatements: When management discovers that previously issued financial statements are false and/or misleading, the error is corrected and reflected in the revised financial statements. The Securities and Exchange Commission indicates that managers have a duty to correct previously issued filings of financial statements that are later discovered to be inaccurate, incomplete or misleading (Skinner 1997). Therefore, the intrinsic value or true returns and risk of the firm are at issue. According to Turner and Weirich (2006), the number of financial restatements had increased since the passage of SOX. Eventually, as managers are deterred from overly aggressive management of earnings to the point of misleading investors, the needs to restate financial reports are expected to decrease. Research shows that restatements are, on average, significant economic events (Dechow et al., 1996; Palmrose et al. 2004). Palmrose, Richardson and Scholz (2004) examined the market reaction to a sample of 403 restatements announced from 1995 to 1999. Using a short two-day announcement window, they document a significant negative mean abnormal return of -9 percent. While Hranaiova and Byers (2007) regression results of dampened abnormal returns in the post-SOX period are consistent with Palmrose et al. (2004), Hranaiova and Byers find that the magnitude of negative market abnormal returns are more than twice the magnitude of the positive abnormal returns in both the pre-SOX and postSOX periods. Hence, the market perceives restatements as conveying negative information about the quality of financial reporting, thus eroding firm value.

Although restatements occur when managers correct previous reports and reissue more accurate information, not all restatements are a result of errors or fraud. For example, certain changes in accounting principles require restatement of prior period financial statements and are not considered errors. Therefore, in this study, restatements due to changes in accounting principles are not considered as a restatement, consistent with prior research. In this study, restatements are used as a surrogate variable to access the interests and marginal effects upon the agents' incentives that affect firms' intrinsic value pre- and post-SOX.

Incentive-based Compensation: Incentive based compensation is used to motivate the manager-agent to make decisions on behalf of the shareholder-principle. Compensation often contains two components, short-term and long-term incentives. The short-term component consists of cash and bonus compensation to motivate current performance. The long-term component consists of stock option compensation to motivate managers to make decisions that will improve the firm's long-term performance. The choice of earnings management strategies and risky investment projects in the post-SOX period, relative to the pre-SOX period, may not only dampen net income, but it may also affect certain components of managers' compensation (Cohen et al. 2007).

The Sarbanes-Oxley Act encourages good corporate governance by creating a way of realigning the actions and behavior of manager-agents by increasing their accountability (certification of accurate financial statements) and motivating truth telling (false information allows criminal prosecution). The Sarbanes-Oxley Act encourages good corporate governance by creating a way of realigning the actions and behavior of the auditor-agents by increasing the accountability (report audit results to the audit committee) and reputation of auditor independence (prohibition of certain non-audit services). We assert that the actions of increasing the accountability, truth telling, and reputation of the agents are linked to enhancing the principal-agent relationship, which is linked to long-term value creation or the intrinsic value of the firm.


As indicated earlier, one direct application of the principal-agent framework is that between stockholders-board of directors and the auditor. The problem may be illustrated by applying a variant of the moral hazard model by Varian (1992) to include both economic and regulatory incentives under conditions of risk and uncertainty. Extensive applications of the moral hazard model have been used by several economists (Prescott, 1999; Cooper and Ross, 1985; Richter et al., 2003; Elbasha and Riggs, 2003).

Consider a firm operating under approved SOX conditions which employs an auditor to perform audit activities. The board of directors (BoDs), (a select group of shareholders representing all shareholders) acting as principal, is assumed to be risk averse by seeking to maximize expected earnings and is willing to pay a risk premium, c(g). The risk premium includes all costs, other than wages, required for effective implementation of SOX. It also includes incentives to set desired optimal earning management targets (e.g., probability that return on equity is within a desired range). The output produced by the manager of the firm, the financial reports, communicates financial information about the performance of the firm that the BoDs use to assess the firm's intrinsic value. The BoDs requires the auditor, acting as an agent, to perform monitoring services that will lead to greater accuracy in financial reporting of output (financial statements), denoted by [DELTA]. Output is assumed to be discrete, and takes a value from within a finite range {[[DELTA].sub.1], ..., [[DELTA].sub.n]}. This does not contradict the profit maximizing or cost minimizing objective of most economic operations because inaccurate financial reporting can wipe out profits and bankrupt the operation, as in the examples of Enron and WorldCom discussed earlier.

Moral hazard problems of financial reporting accuracy produced by the manager and audited by the auditor occurs when BoDs are unable to observe the agents' efforts, thereby creating a welfare loss. The inability of the principal to learn the agents' effort level merely by observing the outcome arises from the fact that outcomes are partly determined by chance, sometimes observed by an associated probability level. According to Weiss (1997), such a model could be applied in analyzing how effectively the principal guards against failures which are subject to chance.

The model by Richter et al. (2003) explains the role of psychic income and current monetary incentives in reducing welfare loss. This model assumes that the effort applied by the agent is discrete, and has only two levels, high effort, [e.sub.H] and low effort, [e.sub.L]. It is assumed that the auditor-agent operates under SOX guidelines, g, prior to starting work, so that an increase in psychic income is induced each time the auditor does a good auditing job. Effort creates disutility in the auditor, so that the disutility of high effort, [d.sup.H] (g), is greater than the disutility of low effort, [d.sup.L] (g). In addition, it is assumed that the manager-agent operates under SOX guidelines, g, prior to starting work, so as to avoid substantial criminal penalties due to intentional misrepresentations to the SEC. Similarly, effort creates disutility in the managers, so that the disutility of high effort, [d.sup.H] (g), is greater than the disutility of low effort, [d.sup.L] (g) .

In the model, SOX regulation is assumed to affect the agents' disutility of effort because it increases the likelihood of accurate financial restatement and strengthens the firms' internal control system and adherence to GAAP. SOX decreases the disutility of exerting high effort and increases the disutility of exerting low effort so that [[partial derivative]d.sup.H] (g)/ [partial derivative]g < 0 and [partial derivative][d.sup.L] (g) / [partial derivative]g > 0 .

To construct the BoDs-principal's benefit function, we let f(r) be the returns from accurate reporting by the firm (including restatements r), and w(r) be the monetary payment to the agent so that [partial derivative]f (r)/ [partial derivative]r > 0 and [partial derivative]w(r)/ [partial derivative]r >0. Therefore the BoDs' benefit function is w(r)- c(g). Also, the agents are assumed to have a utility function defined as U(w(r)) + d(g), with U'(w(r)), a von Neumann Morgenstern utility function strictly decreasing in w.

The model is further simplified by assuming that, although BoDs-principals cannot precisely determine the agents' effort level based on the value of output, given a certain level of effort, the probability distribution of returns or outcome [DELTA] can be estimated using a discrete choice model. The probability distribution of [DELTA] at high effort is defined as {[pi].sub.iL]} and the probability distribution of [DELTA] at low effort as {[[pi].sub.L]}. Since [DELTA] is assumed to be discrete, it follows that both the monetary value of accurate financial reporting f([DELTA]) and monetary payment to the agent w([DELTA]) are also considered discrete. In reality, the principal will only be able to estimate [DELTA], and the higher the monitoring level (independence induced by SOX) the better the financial reporting will be.

If we assume, based on classic agency theory, that effort is not observable, then the problem when SOX is introduced is stated as:


where the first constraint is the participation or individual rationality constraint, and the second constraint is the incentive compatibility constraint. The risk averse BoDs' goal is to maximize long run earnings or intrinsic value. However, the risk averse auditor-agent could choose not to accept the contract and indulge financial restatement or the risk averse manager-agent could choose not to accept do more work for less pay, unless the agent receives at least his reservation utility level [bar.U]. The participation constraint becomes binding at the optimum. If the incentive compatibility constraint is satisfied, then the auditor-agent finds it optimal to exert the effort level that the BoDs-principal wants him to apply.

The problem in equation 1 is analyzed in two stages. First, the problem is solved with low effort and no SOX guidelines implemented. Under such circumstances, the optimum solution is to pay the agent a payment of

2. w = [U.sup.-1] ([bar.U] + [d.sup.L] (0)).

Because payments do not depend on the effort level applied by the agent, the agent will choose the effort level that brings about the lowest disutility, but still lets him earn his reservation utility. Adding SOX implementation guidelines to the low effort case will only lead to an increase in payments and costs since the disutility of low effort increases with SOX, such that only SOX guidelines solve the low effort situation.

The other scenario considers the high effort case with SOX guidelines. The problem is formulated with payment W as the only choice variable and SOX guidelines g as a parameter so that comparative statics can be performed on the optimized function while varying g (pre-SOX era, Sox era, and post-SOX era). This allows for a more intuitive understanding of the problem


where the first constraint is the participation constraint and the second constraint is the incentive compatibility constraint. The optimal payment scheme will maximize stockholders' expected returns by inducing the auditor to exert high effort and recommend restatements accordingly.

Extending the above problem to a multi-agent concept is a necessary but trivial problem. We have management as another agent performing the financial reporting tasks to accurately reflect performance. Both the management prepared financial statements and the auditor's opinion on those financial statements are considered a single joint output. Moral hazard issues arise for multi-agent problems when there is uncertainty in output because agents who cheat cannot be identified if joint output is the only observable indicator of inputs (Holmstrom 1979). The model assumes a risk averse principal and k risk-averse agents (in this case just two agents) producing a single output for which individual contributions cannot be differentiated


The Kuhn-Tucker conditions obtained when the Langrangian of the above problem is differentiated with respect to w show that

5. 1 / [U'.sub.k] = [[lambda].sub.k1] + [[lambda].sub.k2](1--[[pi].sub.iL] / [[pi].sub.iH])

Given [[lambda].sub.k1] > 0, [[lambda].sub.k2] > 0, this suggests that the optimal compensation for management (salaries plus stock option) and auditor (auditor fees) will depend on the likelihood ratio of [[pi].sub.iL]/[[pi].sub.iH] The likelihood ratio is the ratio of probabilities of obtaining accurate financial reporting, i, given low and high efforts. For accurate reporting it implies SOX will induce high effort levels or a desire for the likelihood ratio to be large because a large amount of errors in reporting should be associated with low effort as compared to high effort. SOX will allow obtaining a better estimate of this ratio and induce high efforts from both agents to achieve targeted performance goals and avoid negative returns shareholders.


When managers have the choice of managing earnings around certain targets, the utility functions specified in the theoretical model can be estimated by a limited discrete choice model. The errors are assumed to be random independent variables and to follow a Weibull distribution. Consequently, the difference between errors is logistic (Domencich and McFadden 1975). With the assumption that the principal is risk averse, the BoD can optimize ROE performance standards. In this study we assume that the BoD has the choice of three alternative earnings management categories: abnormal positive earnings (ROE greater than 10 percent, higher than industry averages), normal earnings (ROE around 10 percent, or aggregate industry average), and negative earnings (ROE less than zero percent). Although these are reasonable earnings management targets, additional tests to ensure that the categories are significantly distinct were necessary. These tests are also used to determine whether aggregation of the data into discrete categories resulted in significant loss of information from observed ROE data. The non-parametric Mann-Whitney test (also known as Mann Whitney U-test when U is calculated) was used to ensure that aggregated data fell into distinct categories. This test is used instead of the parametric t-test because of deviations from normality and differences in the sub-sample sizes in each ROE category. Three test measures were performed for the three multinomial categories (normal versus abnormal, abnormal versus negative ROE and negative ROE versus normal around industry average) with the data assembled into a single set of size N = [n.sub.a] + [n.sub.b] for each pair. The N-size measures are then ranked in ascending order, and the rankings returned to the original samples in place of the raw measures, so that [n.sub.a] is the number of ranks in group A (normal ROE), and [n.sub.b] is the number of ranks in group B (abnormal). In addition, we define [T.sub.A] as the sum of [n.sub.a] ranks in group A, [T.sub.B] as the sum of [n.sub.b] ranks in group B, and [T.sub.AB] as the sum of N ranks in groups A and B. The Mann Whitney test used here is based on the z test which is defined as z = ([T.sub.obs]--[[mu].sub.T])[+ or -] 0.5/ [[sigma].sub.T], where T is the observed value for either [T.sub.A] or [T.sub.B], [[mu].sub.T] is the mean of the corresponding sampling distribution of T, [[sigma].sub.T] is the standard deviation of that sampling distribution, and 0.5 is used as a correction for continuity (with -0.5 used when [T.sub.obs] < [[mu].sub.T] and + 0.5 used when [T.sub.obs] > [[mu].sub.T]). With a calculated symmetric z-value of 234.71 and a p-value of 0.0001, we conclude that data for abnormal ROE and normal ROE can be grouped into separate discrete categories without significant loss of information.

To test the parameters of the principal-agent model (equation 5) and estimate the probability of returns for these three categories, the multinomial logit model is used. (However, results were compared with the ordered probit model results to assess robustness of parameter estimates.) In this model, the probability distribution of firm returns falling in one of the three ROE categories is a function of audit fees; non-audit fees; management compensation, with and without stock options (Stock options cashed over a period of time is identified in the literature as one strategy that can be used to align BoDs and managements' goal to maximize firm intrinsic value.); dummy variables for SOX and post SOX regulation; a dummy variable indicating restatements; a quadratic interaction term indicating the combined effects of the magnitude of restatement and stock compensation, incentives for both agents; and all three other arguments used to ensure ROE estimations, as in the DuPont analysis framework, are exhaustive were included (profit margin, asset turnover ratio, and the equity multiplier or a liability ratio capturing financing risk). The multinomial logit model is used because the dependent variable, the ROE category, is qualitative in nature and it is classified in more than two alternatives. The marginal impacts of the independent variables are used to evaluate transition probabilities for firms moving from one category to the next for the different years and as a result of SOX.

It is assumed that the negative ROE category is the base category (relating to firms managing their earnings to avoid loss) and is chosen outside of the modeling framework. Therefore, the probability of falling in the base category is indeterminate in the present choice set. Nevertheless, in normalizing the coefficients for 'negative ROE category' to zero, the problem is resolved (Amemiya and Nold, 1975). Full information maximum likelihood (FIML) is utilized to estimate the coefficients for the other 'normal ROE' and 'abnormal ROE' categories. The probability of the jth firm falling in the ith ROE category can be calculated as in equation 6 (Greene, 1995).

6. Prob[choice j] = exp([[beta]'.sub.j][X.sub.jt] / [[summation].sup.J.sub.i] exp ([[beta].sub.m][]), j = 0, 1, & 2.

The marginal effects of the variables, [X.sub.j], can also be estimated. In the discrete choice model, the effect of a change in attribute m (such as audit fees, restatement, etc.) of the alternative j on the probability that the firm would choose alternative k (where k may or may not equal j) is estimated as in equation 7 (Greene, 1995).

7. [[partial derivative].sub.jk] (m) = [partial derivative]Prob[[y.sub.i]=k]/ [partial derivative][x.sub.ij](m) = [1(j=k)--[P.sub.j] [P.sub.k]][[beta].sub.m]

It is essential to evaluate the differential impacts of SOX on the quality of financial reporting, profitability and risk. The probabilities and marginal effects from the multinomial logit model are estimated by the LIMDEP econometrics software package (Nlogit version 3.0).

To assess the effectiveness of SOX, we obtain firm financial data, auditor fees, and executive cash and stock compensation data from the University of Pennsylvania's Wharton Research Database (WRDS) for years 2000 through 2005. We begin with year 2000 because that is the first year auditor fee data had become publicly available. We obtained 8,828 observations across 14 industries based on four-digit SIC codes. After the deletion of observations with significant missing data, and 100 outlier observations, our final sample consists of 7,497 observations, an average of approximately 1,065 firms per year. Firms in the durable manufacturing industry represent 26 percent of the sample. The computer and retail-other industries represent 15 and 10 percent, respectively, of the sample firms. Table 1 shows the distribution of all sampled firms by industry membership based on four-digit SIC codes.

Since restatements that most concern investors involve overstatement of revenues and accounts receivable as well as improper recognition of expenses, we first identify financial restatements that involve total assets (Compustat data6), sales (Compustat data12), and net income (Compustat data172). Next, we compare the originally stated values to the respective restated values (Compustat data120, data117, and data177) to obtain differential values. Restatements were identified if the differential value is nonzero. Restatements due to discontinued operations, changes in accounting principles, and acquisitions/mergers are not counted as restatements, consistent with prior studies. Then, if one or all three differential values obtain a nonzero value per firm in a given year, we consider it as a single restatement item for that year for that firm to avoid double counting. In addition, we divide our sample into three sample periods, Pre-SOX (years 2000--2001), DSOX or year of SOX implementation (year 2002), and Post-SOX (years 2003--2006). This allows us to assess the marginal effects of restatements upon intrinsic value and evaluate the differences between actual and predicted earnings for those periods. Table 2 shows the number of restatements per year for pre-, during, and post- SOX periods. In our multinomial logistic model, we use indicator variables to control for the sample periods. Pre-SOX is an indicator variable that equals 1 if the year is 2000 or 2001 and 0 otherwise. DSOX is an indicator variable that equals 1 if the year is 2002 and 0 otherwise. DPOSTSOX is an indicator variable that equals 1 if the year equals 2003, 2004, 2005, or 2006 and 0 otherwise.

To obtain goal congruence on increasing intrinsic value, agency theory purports that the principal motivates the agent though compensation incentives. The independent auditor-agent plays a significant role in monitoring the faithful representation of financial reports produced by the manager-agent. Total auditor fees paid to the auditor consist of two basic components, audit fees and non-audit service fees. We use two variables, AFEAS and NFEAS, to represent compensation payment to the independent auditor-agents. AFEAS equals audit fees received by the auditor paid by the client firm divided by total assets. The agreed upon compensation for providing audit services, AFEAS, is positively correlated to the level of expected auditor effort (Gist 1994; Davidson and Gist 1996). NFEAS equals nonaudit service fees received by the auditor paid by the client firm divided by total assets. Since the SOX Act has limited the type of non-audit services that an auditor can provide to their audit client, we expect NFEAS to play an insignificant role in motivating principal-agent goal congruence.

Executive compensation consists of two basic components to motivate managers to enhance the value of the firm in both the short and long run. TCOMP and COMP are variables that indicate compensation to the executive managers in the form of executive compensation. TCOMP equals the sum of cash and bonus compensation paid to executive managers for the recent performance. COMP, the long run incentive, represents the stock option compensation granted to executive managers. DRESTAT is an indicator variable that equals 1 if the year the firm issued a restatement that involves either sales, net income or total assets as reported by Compustat, and 0 otherwise. QRSCOMP is a quadratic interaction term that equals DRESTAT multiplied by COMP.

We also control for financial risk (LIABR variable) and profitability (PMR variable). LIABR equals total liabilities divided by total assets. PMR represents the profit margin ratio calculated as net income divided by sales. Table 3 lists the variables used in this study.

Descriptive Statistics

Pearson correlation analysis identified correlation between the variables and are shown in Table 4. AFEAS is positively correlated with NFEAS (0.327), COMP (0.577) is negatively correlated with PMR (-0.445). COMP is positively correlated with TCOMP (0.439), NFEAS (0.277), QRSCOMP (0.350) is negatively correlated with PMR (-0.416). PMR is negatively and significantly correlated with LIABR (-0.514) and positively correlated with ROE (0.289). QRSCOMP is correlated with DRESTAT (0.422). As a result, a choice-based sampling procedure was used during the empirical analysis to ensure that the estimated parameters are robust and to correct potential multicollinearity problems. Thus, the estimation errors are minimized but the coefficients are not affected.

We present the descriptive statistics in Table 5. Analysis of variance indicates that net income, AFEAS, NFEAS, TCOMP, QRSCOMP, and PMR are significantly different across periods. The mean Pre-SOX ROE of 8.18 percent decreases significantly to 3.05 percent in year 2002 then significantly increases to 9.81 percent in the Post-SOX period. Although the mean Post-SOX ROE is higher than the mean Pre-SOX ROE, the difference is not significant. Meanwhile, the median ROE shows less variation across periods. Detail empirical results from the multinomial logit analysis are presented next. Figure 1 presents further evidence that earnings management strategies target toward the aggregate industry average performance measure.



Multinomial Logit Analysis

Three goodness-of-fit measures are used to evaluate the overall fit of the model. The estimated value of the McFadden [R.sup.2] is 0.206, which is appropriate for categorical and limited dependent variable models. The estimated chi-square test statistic is equal to 3,190.234 with 18 degrees of freedom and it is significant at the 0.001 percent level. The percentage of correct predictions is equal to 70.37 percent. The three measures-of-fit together indicate that the explanatory power of the model is good. To ensure that parameter estimates are efficient and unbiased the full information maximum likelihood estimator procedure and choice based sampling procedure is used to estimate the multinomial logit model.

As the model is a multinomial logit, two sets of coefficients are mentioned, one for the 'normal ROE' category and one for the 'abnormal ROE' category. Whatever the set, most of the coefficients are significant at the 5 percent and 1 percent level of significance. The explanation of the multinomial logit results follow the recommendation by Greene (2003, 722), as he noted that these coefficients should not be explained as those from a continuous variable model. Table 6 shows the estimated results. Overall, the coefficients from the normal and abnormal ROE reveal very interesting results relating to the hypothesis stated earlier.

See Table 3 for list of variable definitions.

As expected, audit and non-audit fees (AFEAS and NFEAS) will reduce ROE when these are significant. NFEAS was not significant. However, AFEAS played a significant role at the 1 percent level in deterring management from engaging in higher risk higher return behaviour. This suggests that increasing audits (e.g., audit fees) will discourage aggressive earnings management behaviour by firms that may yield higher than industry average ROE but also increase the likelihood of loss. Higher total compensation package (TCOMP) paid to management is associated with the likelihood of higher abnormal returns and is significant at the 10 percent level. This suggests that higher than average ROE may result in bonuses for the manager-agent. Stock compensation (COMP) on the other hand tells a very interesting story. COMP is not significant for the abnormal ROE category but it is positive and significant for the normal ROE category at the 1 percent level. This is a good indication that stock compensation will encourage management to adopt a solid earnings management strategy around industry averages that may lead firms to maximize long-run intrinsic value of the firm. Another interesting observation was the effect of reinstatement (DRESTAT) in goal congruent. As with the impact of stock compensation, reinstatement will favour a solid earnings management strategy around the industry average that discourages behaviour to engage in pursuing abnormal high returns that may also imply high risk. An interesting observation is that the quadratic interaction term of the magnitude of restatement and stock compensation option (QRSCOMP) is significant and negative, indicating that reinstatements may be viewed negatively by investors even when firms are employing strategies to maximized long-run intrinsic value of the firm.

A major focus of our analysis is to investigate the impact of SOX in aligning principal and agent behaviour. The results reveal that SOX guidelines had significant positive impacts in aligning management's goal around solid earnings management strategies (industry averages) and discourage behaviour to pursue risky behaviour that may increase returns but also exposes the firm to loss. The results suggest that SOX guidelines have a significant effect on reconciling moral hazard problems between BoDs-principal and the manager-agent, as it relates to accurate financial reporting. The marginal impact of SOX guidelines will be discussed later.

Finally, in the last set of variables, the liability ratio (LIABR) and profit margin (PMR) are analyzed. The estimated coefficients are all positive and significant, except for the LIABR of the normal ROE category. These results are as expected, as leverage and profitability increase, ROE will increase. However, as a firm takes on increasing amounts of debt, it leads to an increase in financial leverage and financial risk that may, at very high levels, have an adverse impact on the earning management strategy.

Marginal Impact Analysis

The predicted probabilities for all three categories (abnormal ROE, normal ROE and negative ROE) are 44.1 percent, 41.9 percent and 14.0 percent, respectively. In order to understand how each independent variable impacts the results of the model, the marginal probability of the variable must be determinate. Table 7 reports the marginal impacts of selected variables averaged over firms on the probability to be classified as 'abnormal ROE,' 'normal ROE,' and 'negative ROE.' The entire results of the marginal impacts and elasticities are included in Appendix A.

Restatements (DRESTAT) have a negative and significant marginal impact on firms that have abnormal ROE, firms that might have overstated earnings. DRESTAT has a positive significant marginal impact on firms with normal ROE. Interestingly, QRSCOMP, the quadratic interaction term of restatement and stock compensation, has a positive and significant marginal impact on firms that have negative ROE. As seen in Appendix A, stock option compensation by itself does not have the greatest impact on accurate financial reporting or earnings management strategy. The results suggest firms may issue more accurate financial reporting by using a joint strategy of providing the auditor-agent with greater independence by reporting directly to the audit committee of the BoDs-principal, while continuing to provide the manager-agent with stock compensation options. Focusing solely on management incentives, the stock compensation option without auditor incentives may not accurately correct the problems of accurate financial reporting and effective earnings management.



The main objective of this study is to evaluate the effectiveness of SOX as the regulatory solution to accounting scandals, by creating a way of re-aligning the actions and behavior of the manager-agent and auditoragent with those of the shareholder-principal to maximize the firm's intrinsic value. A principal-agent model with multiple agents is developed based with risk considerations. In this model, the probability of being in one of the three ROE categories ('abnormal,' 'normal,' and 'negative') is a function of audit fees, management compensation, and other control variables are used to better simulate actual firm behavior in the analysis and improve the model's predicting power. In traditional principal-agent models, the principal is assumed to be risk neutral. However, in reality and in our paper, the principal often pays a risk premium, c(g), to its manager-agent to achieve desirable profit margin (PMR) and manage financial risk (LIABR) that, in turn, directly affects the choice of ROE categories. That risk premium can be in the form of employee compensation, additional employees, technology or other costs necessary to achieve the target (e.g., ROE falling within normal category).

The multinomial logit procedure is used to empirically evaluate and test the theoretical findings, and the principal-agent framework seems to explain auditor-agent and manager-agent behavior in issuing accurately stated financial statements, implying the use of less aggressive earnings management measures. Relative to the pre-SOX period, the principal-agent relationship appears to obtain greater goal congruence in the post-SOX period. The evidence indicates that SOX provides disincentives for the agents to engage in excessive high risk high return strategies.

The policy implications of these findings are multiple. First, an emphasis on firm intrinsic value that targets management and auditor incentives may be most effective for accurate financial reporting. Second, SOX guidelines may be ineffective if auditors are provided with fixed audit fee compensation, as that may adversely impact the auditor's incentive to require managers to report restatements. Third, increasing audit fees, as a proxy for increasing auditor effort, contributes significantly to lowering abnormal ROE or more accurate financial reporting. Further research that explicitly incorporates the cost of SOX implementation is warranted. Also, similar analyses should be conducted for different industries.

Supporters of SOX say that it has helped protect investors while critics complain about the cost of compliance. As plaintiffs challenge the creation of the PCAOB and its power to set auditing standards and investigate suspected wrongdoing by audit firms to the Supreme Court (Hilzenrath 2009), maybe they should consider the role that the PCAOB plays in enhancing the effectiveness of SOX in realigning the principal-agent interests.
A Summary of Marginal Effects and Elasticities

Variable      Coefficient      Standard        b/St.Er.

Marginal effects on Prob[Y = Negative ROE]

AFEAS            11.45992         5.24759           2.184
NFEAS             -.88694        13.26057           -.067
TCOMP            -1.22927         1.24893           -.984
COMP             -12.1746         7.72079          -1.577
DRESTAT            .00699          .01047            .668
LIABR             -.08315          .02410           3.450
DSOX               .01697          .01221           1.389
DPOSTSOX          -.03441          .01089          -3.160
PMR              -2.40846          .48563          -4.960
QRSCOMP           11.6784         4.22440           2.765

Variable      P[[absolute     Elasticity
              value of Z]

AFEAS               .0290         .095219
NFEAS               .9467        -.003250
TCOMP               .3250        -.029947
COMP                .1148        -.110945
DRESTAT             .5044         .023027
LIABR               .0006        -.312416
DSOX                .1647         .018502
DPOSTSOX            .0016        -.140954
PMR                 .0000        -.570069
QRSCOMP             .0057         .065297

Variable      Coefficient      Standard        b/St.Er.

Marginal effects on Prob[Y = Normal ROE]

AFEAS            12.83315         6.17592           2.078
NFEAS            21.48742         9.26649           2.319
TCOMP            -2.87941         1.61849          -1.779
COMP             10.47398        11.12596            .941
DRESTAT            .09158          .01638           5.590
LIABR             -.15390          .02329          -6.609
DSOX               .16867          .02594           6.502
DPOSTSOX           .17636          .02456           7.181
PMR              -1.69910          .45274          -3.753
QRSCOMP          -7.72638         5.27173          -1.466

Variable      P[[absolute     Elasticity
              value of Z]

AFEAS               .0377         .035584
NFEAS               .0204         .026279
TCOMP               .0752        -.023409
COMP                .3465         .031852
DRESTAT             .0000         .100707
LIABR               .0000        -.192961
DSOX                .0000         .061365
DPOSTSOX            .0000         .241119
PMR                 .0002        -.134213
QRSCOMP             .1428        -.014417

Variable      Coefficient      Standard        b/St.Er.

Marginal effects on Prob[Y = Abnormal ROE]

AFEAS           -24.29307         9.21698          -2.636
NFEAS           -20.60048        17.63131          -1.168
TCOMP             4.10868         1.74697           2.352
COMP              1.70066        17.18502            .099
DRESTAT           -.09856          .01667          -5.911
LIABR              .23705          .03669           6.461
DSOX              -.18564          .02270          -8.178
DPOSTSOX          -.14196          .02753          -5.157
PMR               4.10757          .93007           4.416
QRSCOMP          -3.95199         7.37353           -.536

Variable      P[[absolute     Elasticity
              value of Z]

AFEAS               .0084        -.064123
NFEAS               .2426        -.023983
TCOMP               .0187         .031798
COMP                .9212         .004923
DRESTAT             .0000        -.103182
LIABR               .0000         .282935
DSOX                .0000        -.064293
DPOSTSOX            .0000        -.184750
PMR                 .0000         .308863
QRSCOMP             .5920        -.007019

See Table 3 for list of variable definitions.


Amemiya, T., & F. Nold (1975). A Modified Logit Model. The Review of Economics and Statistics 57:2 (May) 255-257.

Bargeron, K.M. Lehn, and C.J. Zutter (2009). Sarbanes-Oxley and Corporate Risk-Taking. Journal of Accounting and Economics. Retrieved from doi:10.1016/jjacceco.2009.05.001

Berle, A. & G. Means (1932). The Modern Corporation and Private Property. New York.

Bedard, J. (2006). Sarbanes Oxley Internal Control Requirements and earnings Quality. Working paper retrieved from

Brigham, E.F. &J.F. Houston (2007). Essentials of Financial Management. Essentials of Financial Management (11th Edition). Thomson South-Western.

Carney, W.J. (2005). The Cost of Being Public After Sarbanes-Oxley: The Irony of Going Public. Emory Law and Economics Research Paper No. 05-4. Retrieved from

Cohen, D.A., A.D. Dye, & T.Z. Lys (2007a). Real and Accrual-Based Earnings Management in the Pre- and Post Sarbanes Oxley Periods. Working paper. Retrieved from

Cohen, D.A., D.A. Dye, & T.Z. Lys (2007b). The Sarbanes-Oxley Act of 2002: Implications for Compensation Contracts and Managerial Risk-Taking. Working paper. Retrieved from

Cooper, R. & T.W. Ross (1985). Product Warranties and Double Moral Hazard. The RAND Journal of Economics 16: 1 (Spring) 103-113.

DaVay, D.L. (2006). The Effectiveness of Sarbanes-Oxley Act of 2002 in Preventing and Detecting Fraud in Financial Statements. Argosy University of Orange County. Dissertation. Com. Boca Raton, Florida.

Davidson, R. A. & W.E. Gist (1996). Empirical evidence on the functional relation between audit planning and total audit effort. Journal of Accounting Research 34 (Spring): 111- 124.

Dechow, P., R. Sloan, & A. Sweeney (1996). Causes and consequences of earnings manipulation: an analysis of firms subject to enforcement actions by the SEC. Contemporary Accounting Research 13 (1), 1-36.

Dey, A. (2009). The chilling effect of Sarbanes-Oxley: A discussion of Sarbanes-Oxley and corporate risk-taking. Journal of Accounting and Economics. Doi:10.1016/j/jacceco.2009.06.003.

Domencich, T.A., & D. McFadden (1975). Urban Travel Demand. A Behavioral Analysis. North-Holland, Amsterdam.

Elbasha, E.H., & T.L. Riggs (2003). The effects of information on producer and consumer incentives to undertake food safety efforts: A theoretical model and policy implications Agribusiness, An International Journal, 19:1 29-42.

Frankel, R. M., M.F. Johnson & R.K. Nelson (2002). The relation between auditors' fees for non-audit services and earnings management. The Accounting Review 77 (Supplement): 71-105.

Gist, W.E. (1994). Empirical evidence on the effect of audit structure on audit pricing. Auditing: A Journal of Practice & Theory 13 (Fall): 25-40.

Garen, J.E. (1994). Executive Compensation and Principal-Agent Theory. The Journal of Political Economy, 102: 6 (Dec.): 1175-1199.

Greene, W.H. (2003). Econometric Analysis (Fifth Edition). Upper Saddle River, NJ: Prentice-Hall.

Hilzenrath, D.S. (2009). Supreme Court Will Hear Sarbanes-Oxley Challenge. Washington Post (May 19). Retrieved from article/2009/05/18/ AR2009051803107. html?hpid=topnews

Holmstrom, B. (1979). Moral Hazard and Observability. Bell Journal of Economics 10: 74-91.

Hranaiova, J. & S.L. Byers (2007). Changes in Market responses to financial statement restatement announcements in the Sarbanes-Oxley era. Working-paper 2007-001; PCAOB.

Jensen, M. C. & W.H. Meckling (1976). Theory of the Firm: Managerial Behavior, Agency Costs, and Ownership Structure. Journal of Financial Economics 3: 305-350.

--& K.J. Murphy (1990). Performance Pay and Top-Management Incentives. Journal of Political Economy 98:225-264.

Leuz, C., A.J. Triantis & T.Y. Wang (2008). Why do firms go dark? Causes and economic consequences of voluntary SEC deregistration. Journal of Accounting and Economics 45, 181-208.

Levitt, A. (1998). The numbers game. Speech delivered at the NYU Center for Law and business, New York, NY, September 28.

Li, H., M. Pincus & S. Rego(2008). Market reaction to events surrounding the Sarbanes-Oxley Act of 2002 and earnings management. Journal of Law and Economics 51, 111--134.

Palmrose, Z., V.J. Richardson & S. Scholz (2004). Determinants of market reactions to restatement announcements. Journal of Accounting & Economics 37: 59-89.

Prescott, E. S. (1999). A Primer on Moral-Hazard Models. Federal Reserve Bank of Richmond Economic Quarterly 85:47-77.

Richter, G. C., E.F. Francisca, P.B. Diaz, W. Brorsen &K. Currier (2003). Using Both Sociological and Economic Incentives to Reduce Moral Hazard. Journal of Agricultural and Resource Economics 28(2):364-373.

Ross, S. (1973). The Economic Theory of Agency: The Principal's Problem. American Economic Review 63:134139.

Scott, W.D., & W.E. Gist (2009). Are Higher Audit Fees another Cost Consequence Associated with a U.S.

Securities and Exchange Commission Investigation? Working Paper. Securities and Exchange Commission (SEC) (2003). AEER No. 1744. Litigation Release No. 18044. SEC Charges HealthSouth Corp., CEO Richard Scrushy with $1.4 Billion Accounting Fraud. (2002). AEER No. 1658. Litigation Release No. 17829. SEC Files Amended Complaint Against WorldCom To Add Additional Fraud Charge and Two Other New Charges, and Expand the Scope of the Alleged Fraud.

Serwer, A. (2006). Stop Whining about SARBOX! Fortune. magazines/ fortune/fortune_archive/2006/08/07/8382589/index.htm.

Skinner, D. J. (1997). Earnings disclosures and stockholder lawsuits. Journal of Accounting and Economics 23:3 (November): 249-282.

Solomon, D., & C. Bryan-Low (February 10, 2004). Companies Complain About Cost of Corporate-Governance Rules. http://online.wsj .com/ad/article/ironmountain/SB 107636732884524922.html?mod=sponsored_by_ ironmountain.

Stanley, J.D. & F. T. DeZoort (2007). Audit firm tenure and financial restatements: An analysis of industry specialization and fee effects. Journal of Accounting and Public Policy 26: 131-159

Turner, L.E. & T.R. Weirich (2006). A closer look at financial statement restatements: analyzing the reasons behind the trend. The CPA Journal. (December) cpajournal/2006/1206/index.htm

U.S. Congress (2002). HR 3763 Sarbanes-Oxley Act 2002. The Public Company Accounting Reform and Investor Protection Act.

Varian, H. R. (1992). Microeconomic Analysis (Third Edition). New York: W.W. Norton &Company, Inc. Weiss, A. (1990). Efficiency wages: Models of unemployment, layoffs, and wage dispersion. Princeton, NJ: Princeton University Press.

Zhang, I. (2007). Economic Consequences of the Sarbanes-Oxley Act of 2002 (February 1). Retrieved from SSRN:

Zhou, J. & G. J. Lobo (2006). Did Conservatism in Financial Reporting Increase after the Sarbanes-Oxley Act? Initial Evidence. Accounting Horizons 20:1 (March) 57-73.

Winifred D. Scott, Arizona State University

William Nganje, Arizona State University
Table 1
Industry membership

Industry                        Pre-SOX      SOX     Post-SOX    Total

Other                                17        10         32        59
Mining & Construction                53        30        123       206
Food                                 60        35        134       229
Textiles                            163        90        333       586
Chemicals                            76        43        162       281
Biotechnology/Pharmaceuticals       101        54        216       371
Extractive                          100        53        214       367
Durable Manufacturers               516       300      1,112     1,928
Computers                           307       177        618     1,102
Transportation                      123        63        238       424
Retail--Wholesale                    81        47        175       303
Services                            159        92        335       586
Utilities Electric & Gas             99        56        214       369
Retail--Other                       223       122        441       786
TOTAL                             2,078     1,172      4,347     7,597

Table 2
Frequency of Restatements

                       Year           2000       2001       2002

None                     0            377        581        684
Restatements             1            547        549        460
Total # of Restatements               924       1130       1144
Percentage of Observations with
  Restatements per Year              59.2%      48.6%      40.2%

                       Year           2003       2004       2005

None                     0            718        763        914
Restatements             1            461        441        240
Total # of Restatements              1179       1204       1154
Percentage of Observations with
  Restatements per Year              39.1%      36.6%      20.8%

                       Year           2006      Total

None                     0            747        4784
Restatements             1             15        2713
Total # of Restatements               762        7497
Percentage of Observations with
  Restatements per Year               2.0%       36.2%

Sample Period     Observations    Observations    Total   Percentage of
                     without          with                Observations
                  Restatements    Restatements                with
                                                           per Sample

PRESOX                 958            1,096       2,054       53.4%
DSOX                   684             460        1,144       40.2%
POSTSOX               3,142           1,157       4,299       26.9%
Totals                4,784           2,713       7,497       36.2%

Table 3
List of Variables

ROE        the independent variable that equals OPM x ATR x EM


           OPM     = Operating Profit Margin Ratio = Income from
                   Operations + Interest Expense divided by Gross

           ATR     = Asset Turnover Ratio = Gross revenue divided
                   by average Total assets

           EM      = Equity Multiplier = Average Total Assets
                   divided by Average Shareholder's Equity

AFEAS      audit fees received by the auditor paid by the client
           firm divided by total assets.

NFEAS      non-audit service fees received by the auditor paid by
           the client firm. divided by total assets

TCOMP      the sum of cash and bonus compensation paid to
           executive managers for the recent performance.

COMP       the long run incentive, represents the stock option
           compensation granted to executive managers.

DRESTAT    an indicator variable that equals 1 if the year the
           firm issued a restatement that involves either sales,
           net income or total assets as reported by Compustat,
           and 0 otherwise.

LIABR      total liabilities divided by total assets; to control
           for financial risk.

Pre-SOX    an indicator variable if year is from 2000 to 2001 or
           zero otherwise.

DSOX       an indicator variable if year equals 2002 or zero

DPOSTSOX   an indicator variable if year is from 2003 to 2005 or
           zero otherwise.

PMR        net income divided by sales; to control for

QRSCOMP    equals DRESTAT multiplied by COMP; a quadratic
           interaction term of the magnitude of restatement and
           stock compensation option

Table 4
Pearson Correlations

                NET          ROE         AFEAS        NFEAS

ROE            .167 **
AFEAS         -.074 **     -.096 **
                 0.000        0.000
NFEAS         -.049 **        0.003      .327 **
                 0.000        0.816        0.000
TCOMP         -.039 **      .034 **      .217 **      .140 **
                 0.001        0.004        0.000        0.000
COMP          -.071 **     -.077 **       577 **       277 **
                 0.000        0.000        0.000        0.000
DRESTAT          0.005       -0.002     -.118 **        0.011
                 0.668        0.838        0.000        0.355
LIABR            0.012      .064 **      .213 **      .161 **
                 0.280        0.000        0.000        0.000
DSOX          -.060 **     -.066 **     -.072 **        0.012
                 0.000        0.000        0.000        0.306
DPOSTSOX       .067 **      .050 **      .213 **      -199 **
                 0.000        0.000        0.000        0.000
PMR            .140 **      .289 **      -445 **     -.207 **
                 0.000        0.000        0.000        0.000
QRSCOMP       -.050 **       -0.002      .084 **       117 **
                 0.000        0.851        0.000        0.000

               TCOMP         COMP       DRESTAT       LIABR





COMP           .439 **
DRESTAT        .059 **      .077 **
                 0.000        0.000
LIABR         -.072 **      .104 **      .087 **
                 0.000        0.000        0.000
DSOX           -.024 *        0.012       171 **        0.004
                 0.038        0.302        0.000        0.748
DPOSTSOX         0.004       -0.006     -.164 **       -0.008
                 0.709        0.610        0.000        0.493
PMR           -.056 **     -.416 **        0.001      -514 **
                 0.000        0.000        0.945        0.000
QRSCOMP        .261 **      .350 **      .422 **     -.055 **
                 0.000        0.000        0.000        0.000

                DSOX       DPOSTSOX       PMR









DPOSTSOX      -.492 **
PMR           -.069 **      .047 **
                 0.000        0.000
QRSCOMP        .081 **     -.102 **        0.022
                 0.000        0.000        0.056

** Correlation is significant at the 0.01 level (2-tailed).

* Correlation is significant at the 0.05 level (2-tailed).

Table 5 Descriptive Statistics

                              NET INCOME        ROE          AFEAS

PreSOX      Mean                  232.98         0.082         0.001
n = 2054    Median                 46.18         0.108         0.000
            Std. Deviation      1,164.47         0.345         0.001
            Minimum           -16,198.00        -2.747         0.000
            Maximum            17,720.00         2.844         0.016

SOX         Mean                   48.16         0.031         0.001
n = 1144    Median                 37.98         0.088         0.001
            Std. Deviation      3,144.28         0.396         0.001
            Minimum           -98,696.00        -2.862         0.000
            Maximum            14,118.00         2.925         0.026

PostSOX     Mean                  421.67         0.098         0.002
n = 4299    Median                 76.47         0.118         0.001
            Std. Deviation      1,675.32         0.319         0.002
            Minimum           -17,462.20        -2.847         0.000
            Maximum            39,500.00         2.917         0.051

Total       Mean                  312.98         0.083         0.001
n = 7497    Median                 60.43         0.111         0.001
            Std. Deviation      1,872.80         0.340         0.002
            Minimum           -98,696.00        -2.862         0.000
            Maximum            39,500.00         2.925         0.051

                                 NFEAS         TCOMP         COMP

PreSOX      Mean                   0.001         0.004         0.001
n = 2054    Median                 0.001         0.001         0.001
            Std. Deviation         0.001         0.009         0.002
            Minimum                0.000         0.000         0.000
            Maximum                0.024         0.143         0.054

SOX         Mean                   0.001         0.003         0.001
n = 1144    Median                 0.000         0.001         0.001
            Std. Deviation         0.001         0.007         0.003
            Minimum                0.000         0.000         0.000
            Maximum                0.015         0.129         0.067

PostSOX     Mean                   0.000         0.003         0.001
n = 4299    Median                 0.000         0.002         0.001
            Std. Deviation         0.001         0.009         0.002
            Minimum                0.000         0.000         0.000
            Maximum                0.045         0.389         0.065

Total       Mean                   0.001         0.003         0.001
n = 7497    Median                 0.000         0.001         0.001
            Std. Deviation         0.001         0.009         0.002
            Minimum                0.000         0.000         0.000
            Maximum                0.045         0.389         0.067

                                 LIABR          PMR         QRSCOMP

PreSOX      Mean                   0.529         0.034         0.001
n = 2054    Median                 0.543         0.044         0.000
            Std. Deviation         0.232         0.127         0.003
            Minimum                0.023        -1.749         0.000
            Maximum                1.879         0.985         0.054

SOX         Mean                   0.529         0.003         0.001
n = 1144    Median                 0.535         0.036         0.000
            Std. Deviation         0.263         0.180         0.002
            Minimum                0.026        -2.877         0.000
            Maximum                2.958         0.853         0.020

PostSOX     Mean                   0.525         0.041         0.001
n = 4299    Median                 0.517         0.053         0.000
            Std. Deviation         0.316         0.213         0.002
            Minimum                0.000       -10.669         0.000
            Maximum               11.383         0.642         0.029

Total       Mean                   0.527         0.033         0.001
n = 7497    Median                 0.525         0.049         0.000
            Std. Deviation         0.287         0.189         0.002
            Minimum                0.000       -10.669         0.000
            Maximum               11.383         0.985         0.054

Table 6
Results of Multinomial Logit Model

Variable          Coefficient        Standard         b/St.Er.

Characteristics in numerator of Prob[Y = Normal ROE]

AFEAS                -51.27606         34.21355           -1.499
NFEAS                 57.56726         93.09157             .618
TCOMP                  1.91715         10.70605
COMP                 111.95140         42.82958            2.614
DRESTAT                 .16841           .09088            1.853
LIABR                   .22714           .15350            1.480
DSOX                    .28090           .11303            2.485
DPOSTSOX                .66629           .08935            7.457
PMR                   13.15575          3.43785            3.827
QRSCOMP             -101.85506         30.65802           -3.322

Variable          P[[absolute       Mean of X
                 value of Z]>z]

AFEAS                    .1339           .00116
NFEAS                    .5363          .000513
TCOMP                     .179            .8579
COMP                     .0090          .001276
DRESTAT                  .0639          .461251
LIABR                    .1389          .525896
DSOX                     .0129          .152594
DPOSTSOX                 .0000          .573429
PMR                      .0001          .033131
QRSCOMP                  .0009          .000783

Variable          Coefficient        Standard         b/St.Er.

Characteristics in numerator of Prob[Y = Abnormal ROE]

AFEAS               -137.00901         54.46262           -2.516
NFEAS                -40.41858        131.42663            -.308
TCOMP                 18.10742         10.99490            1.647
COMP                  90.83900         88.58331            1.025
DRESTAT                -.27362           .09640           -2.838
LIABR               1.13206950           .21017            5.386
DSOX                   -.54258           .10114           -5.364
DPOSTSOX               -.07638           .11857            -.644
PMR                   26.52937          6.74223            3.935
QRSCOMP              -92.40322        42.3059 7           -2.184

Variable          P[[absolute       Mean of X
                 value of Z]>z]

AFEAS                    .0119          .001163
NFEAS                    .7584          .000513
TCOMP                    .0996          .003410
COMP                     .3051          .001276
DRESTAT                  .0045          .461251
LIABR                    .0000          .525896
DSOX                     .0000          .152594
DPOSTSOX                 .5195          .573429
PMR                      .0001          .033131
QRSCOMP                  .0290          .000783

Table 7
Summary of Marginal Effects of Selected Variables Averaged Over

              Y = Negative     Y = Normal     Y = Abnormal

Variable           ROE             ROE             ROE

DRESTAT             0.0028          0.0757         -0.0785
LIABR              -0.0617         -0.1240          0.1857
DSOX                0.0085          0.1392         -0.1477
DPOSTSOX           -0.0328          0.1480         -0.1152
QRSCOMP             9.5551          0.0000          0.0000

* See Table 3 for list of variable definitions.
COPYRIGHT 2011 The DreamCatchers Group, LLC
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2011 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Author:Scott, Winifred D.; Nganje, William
Publication:Academy of Accounting and Financial Studies Journal
Geographic Code:1USA
Date:Jul 1, 2011
Previous Article:Are companies that report material weaknesses in internal control more likely to restate their financial statements?
Next Article:Managed earnings: a closer look at pension expense.

Terms of use | Privacy policy | Copyright © 2018 Farlex, Inc. | Feedback | For webmasters