Selection between earnings management strategies : book-tax conforming versus book-tax nonconforming

(1)

Master's Thesis Accountancy 1

Selection between Earnings Management Strategies:

Book-tax Conforming versus Book-tax Nonconforming

Shuang Geng

10824804

21-June-2015

Supervisor: Dr. W.H.P. Janssen Word count: 11881

MSc Accountancy & Control, variant Accountancy

Amsterdam Business School

(2)

Statement of Originality

This document is written by student Shuang Geng, who declares to take full responsibility for the contents of this document.

I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.

The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

(3)

Abstract

Prior studies have documented two alternative earnings management strategies: book-tax conforming and book-tax nonconforming earnings management. Practically, incentives for these two strategies compete and create a dilemma for managers in selecting between the two. Using data collected from 2006 and 2007 GAO restatement announcement, I examine whether the purchase of tax service is associated with higher rate of nonconforming earning management (i.e. the proportion of nonconforming earnings management to the total amount of earnings management) and if so, to what extent is this association constrained by high quality external auditors.

The result reveals the significantly positive association between the quality of tax service and the likelihood of managers’ selection of nonconforming earnings

management. In addition, it suggests a positive relation between the magnitude of tax service fee and subsequent tax service quality. However, it is unable to detect a

significant relation between the quality of external audit and the rate of nonconforming earnings management.

(4)

Acknowledgement

My sincere gratitude goes to my supervisor, W.H.P. Janssen, for the precious suggestions, patient guidance and timely advices.

To my family and friends, thanks for the continuous trust, understanding and encouragement.

I am grateful to the staff at the University of Amsterdam (UVA), who supported me in accomplishing the study in the UVA, and getting ready for a fresh start.

(5)

1. Introduction

1.1 Background

Earnings management and tax avoidance have attracted increasing emphasis both in practical and in academic categories. A growing number of literatures investigate the causes and results of various earnings management methods. As a result of these studies, many researchers have noticed that different tax effects are generated by different earning management strategies. A lot of studies analyze the accounting choices and earnings management strategies that effect current taxable income and current tax expense (eg. Maydew, 1997; Erickson et al. 2004), which were referred to as book-tax conforming earning management. Other studies (eg. Mills and Newberry, 2001; Frank, 2009) focus on earnings management activities that do not have current tax consequences, which were known as book-tax nonconforming earning management. As managers are motivated by the subsequent benefits from meeting or beating earnings target, some of them tend to increase earnings by engaging in upward earning management strategies. Other managers who are willing to obtain benefits from subsequent stock-option grant would intentionally miss the earnings target, by engaging in downward earning management (McAnally et al, 2008; Cai and Zhang, 2015). In this paper, I focus on upward earnings management and conduct in-depth research regarding the selection of conforming and nonconforming upward earnings management strategies.

There exist both incentives and restrictions for management to engage in different upward earnings management activities. On one hand, since firms are able to obtain immediate cash flow benefits by reducing taxable income (Mills, 1998; Badertscher et al, 2009), the saving result from tax management practices provide strong incentives for managers to engage in nonconforming earning management. On the other hand, firms that have high book-tax discrepancies are very likely to be deemed as aggressive in its tax reporting (Cloyd, 1995; Lilian, 1998), which will lead to public scrutiny and subsequent reputation loss. As aggressive tax planning activities could result in reputational and commercial risks not only to firms but also to their external auditors, firms’

(7)

to select the alternative conforming earnings management strategy to remain a higher level of book-tax conformity.

The incentives and restrictions both exist and compete with each other, which creates a dilemma for managers in selecting conforming and nonconforming upward earning management strategies. Management faces a trade-off between reducing the current tax expense and maintaining low book-tax differences. This study extracts the plentiful achievements of prior researches on earnings management strategies and management incentives, and is expecting to provide deeper understanding of the inter-relationship between these two issues.

In this paper, I identify the purchase of tax service as an incentive of nonconforming earning management, and identify high quality external audit as a restriction of tax management, in other word, an incentive of conforming earnings management. I aim to analyze whether the purchase of tax service is associated with higher rate of

nonconforming earning management and if so, to what extent is this association

constrained by high quality external auditors. Specifically, I examine whether there is a positive relationship between the payment of tax service fee and the preference of

nonconforming earning management, and test the expected negative association between external audit quality and the selection of nonconforming earning management.

An important feature of the research design is that it takes the misstatement United States companies as observation. I collect data from the 2006 US GAO (United States Government Accountability Office) report and the 2007 GAO report. Since the 2006 and 2007 GAO report disclosed a list of firms that restated from 2002 to 2006, I am able to identify the restated firms that originally overstated their earnings. As is disclosed by the 2006 and 2007 GAO report, the restatement announcements between 2002 and 2006 include a list of 1786 firms. I filter out those firms that had underestimated their earnings, delete the observations without adequate information, and restrict the observation to the first year that a firm restated its pre-tax income. In the end there are 151 firms that meet all the requirements.

I identify the rate of nonconforming earnings management as dependent variable, use the adoption of external tax service, and the quality of external auditors as

(8)

different earnings management strategies, I control for other firm characteristics variables, which includes the availability of net operating loss carry forward, leverage, market to book ratio, net cash flow, firm size, the total amount of earnings management and net deferred tax liability at the beginning of the restated year. Through WRDS (Wharton Research Data Services, University of Pennsylvaniato), I collect firm specific variables from Compustat database and auditor related information from Audit Analytics database. The desired information includes both the original and the restated financial statement data, while the above two databases miss some of the information. Hence I hand collect the information that is not available through WRDS.

After the measurement of the variables, I perform a descriptive analysis to provide a preliminary summary of statistics. The result presents a large deviation for the variables between different firms, such as the amount to tax service fee, net deferred tax liability, firm size, the rate of nonconforming earning management and the total amount of earnings management. In particular, there is a high level deviation of tax service fee between different firms. A univariate analysis is conducted to test the correlation of main variables. It shows a generally consistent result with the previous discussion and

deduction, despite the fact that some of the coefficients are not significant enough. To eliminate the confound effect of other firm specific elements, I control for those variables and make a multivariate analysis. In order to take the quality of tax service into

consideration, an additional test is then performed to include the exact amount of tax service fee, which has not been precisely measured during the previous analysis. The result indicates a significantly positive association between the magnitude of tax service fee and the rate of nonconforming earnings management, suggesting that tax service quality is associated with an increased likelihood of the managers’ selection of

nonconforming earnings management. Moreover, the difference between the results of original test and the additional test provides complementary evidence to the discussion that higher tax service fee is related to high tax planning quality. However, the result is unable to detect a significant relation between the quality of external audit and the rate of nonconforming earnings management, while the coefficient is negative.

The remainder of the paper is organized as follows. The next section provides detailed description and explanation of the research question, and puts forward the

(9)

motivation of this study. Section 2 summarizes relevant prior research and develops the hypotheses. Section 3 describes the design and methodology of research, and reports the process of sample selection. Section 4 presents empirical results, sections 5 shows the additional test, while Section 6 discusses implications of these results, concludes the contribution and limitation of this paper.

1.2 Research question

As is common in both academic and practice categories, tax expense is one of the most significant form of expenses in a firm. The magnitude of tax expense is closely related to firms’ net cash flow and profitability. Specifically, in the United States, the firms' pretax net income has to be charged for up to 35 percent of tax fee. The reduction of taxable income can bring instant cash flow benefits, and boost the short-term

profitability of a firm (Mills, 1998). As there is always a significant positive association between firms’ profitability and management’s bonus, managers have strong motivation to engage in tax planning and tax avoidance activities.

When tax effects are considered as an important aspect in making earnings management decisions, companies are more likely to engage in earnings management activities that have no effect on current taxable income or current tax expense. As higher tax management incentives bring about more expected tax saving benefits, I predict that higher tax avoidance incentives associate with higher rate of nonconforming earnings management (the magnitude of nonconforming earning management scaled by the total amount of earnings management).

Meanwhile, there are other elements that constrain firms’ tax avoidance activities. A large body of prior economic literatures (Cloyd, 1995; stock et al, 1996) provided

evidence for this. According to their analysis, tax service providers and corporate tax executives follow the heuristic that book-tax differences result in red flags for the IRS (Internal Revenue Service, the U.S. government agency that is responsible for tax collection and tax law enforcement), or weaken the position of taxpayer on audit. As a matter of fact, in the United States, to the larger extent that book income exceeds taxable income, the more evidences IRS has that those firms have been aggressive in its tax reporting (Lilian, 1998). Moreover, there are signs that aggressive tax management could result in reputational risks with customers, which may lead to commercial losses. In

(10)

addition, it has been proved that big investors of companies are showing more sensitive reaction to the risks of aggressive tax management at the firms they invest in (Madison, 2014). Hence I predict that high quality external audit helps to create a strict tax reporting environment in which firms are motivated to report financial accounting data of lower book-tax differences. In other word, it is predictable that higher external audit quality associate with higher rate of conforming earnings management (the magnitude of conforming earning management scaled by the total amount of earnings management).

From the above analysis, I aim to resolve the research question of this study: Whether higher tax avoidance incentives associate with higher rate of nonconforming earnings management. And if so, to what extent is this association constrained by external auditors.

1.3 Motivation of this study

By considering the selection of earnings management methods under the framework of tax planning, I am motivated to do an in-depth research based on previous research.

The motivation of this study arises from its expected contribution. As this paper has implications for not only researchers but also government, regulators and investors, the contribution of this study can be analyzed from both scientific and societal point of view.

1.3.1 Contribution to prior literature

This paper extends prior literature in terms of research question, research method and hypotheses design.

First, as mentioned by Fields et al. (2001), examining only one earnings

management technique at a time cannot help to demonstrate the overall consequence of earnings management. Compared to prior studies, this paper contributes to the earnings management literature, by examining the trade-off between two earnings management strategies at the same time. It provides a more comprehensive picture of managers’ selection of conforming and nonconforming earnings management techniques.

Second, this paper introduces, discusses and provides a better understanding of firms’ selection of different earnings management strategies. Most prior researches that focus on the incentives of earning management do not consider the competition between different incentives, or the effect of competed incentives. This paper investigates specific

(11)

management strategies. In particular, it demonstrates the incentives that managers face regarding tax avoidance and high book-tax conformity associated with earnings management practices.

Third, this paper utilizes misstatement sample which is seldom adopted by prior literatures. Restatement announcement shows a clearly and accurately identification of a setting in which earnings management have happened. It helps to better identify and measure the conforming and nonconforming amount of managed earnings.

1.3.2 Contribution from a societal point of view

By studying how managers select between different earnings management strategies, this paper suggests a way to predict firms’ preference of different earnings management activities. That is, by analyzing the operational and accounting environment of a firm, it is possible to deduce managers’ decision about earnings management. As such, the research question has implications about whether enhancing audit quality, might decrease the level of nonconformity earnings management. Moreover, this paper provides

inspiration to investors, regarding their estimation of firms’ earnings performance.

2. Literature review and hypotheses 2.1 Literature review

As a pervasive subject in the accounting literature, earnings management has attracted emphasis by a large number of scholars. Many academic researches have provided evidence on managers’ incentives for upward earning management. Managers engage in earnings management activities when they perceive the benefits of managing earnings to be higher than the costs (Healy and Wahlen, 1999). This interpretation is further supported by other studies, indicating that managers have strong incentives to manage earnings upward, in order to meet the requirement of compensation contracts, stock market pricing, debt covenants, government regulations as well as other public demands (Fields et al., 2001; Phillips et al, 2003; Graham et al ,2005). A large body of researches provide qualitative and quantitative evidences to interpret management’s earnings management behaviors. Phillips et al (2003) discuss three main motivations of earnings management: to meet or beat experts’ earnings forecasts, to gross over a loss, and to avoid reporting a decrease of earnings. Graham et al (2005) note from their survey research that managers manage earnings upward to meet benchmarks. Bergstresser (2006)

(12)

reveals that earnings management activities are more likely to occur in a firm when CEO's potential compensation is more closely related to firm's stock price and earnings performance.

Despite upward earnings management, managers can also be motivated to employ downward earnings management techniques. Matsumoto (2002) observes that managers employ downward earnings management, in order to guide analysts' forecasts downward. Cai and Zhang (2015) further note that managers tend to engage in downward earnings manipulation to reduce the benchmark indicators, so as to meet the requirement of stock option grant.

Prior literature has sought to detect earnings management activities by using various methods. Healy (1985) proposes a model to measure earnings management by using mean total accruals (scaled by total assets) as a proxy for abnormal (discretionary) accruals. Jones (1991) presents a model based on the assumption that nondiscretionary (normal) accruals are constant. She controls for firm-specific parameters that influence a firm’s economic circumstances on nondiscretionary accruals. Dechow et al (1995) document the properties of existing earnings management models and provides a

modified version of the Jones Model. Peasnell et al (1999) provide additional evidence on the performance of existing models and develops a "margin model" to measure the

magnitude of discretionary accruals with higher precision. These researches enhance the ability to detect earnings management and provide important insights regarding the estimation of various earnings management strategies.

There are different scenarios involved in various earnings management techniques. Managers could employ real earnings management, accruals earnings management, or both simultaneously. Specifically, management could either influence operating, investing or financing decisions (i.e. real earnings management) or engage in financial reporting discretion (i.e. accrual earnings management) to mislead information users about the underlying economic performance of the company. Burgstahler and Dichev (1997) conduct preliminary analysis to demonstrate whether managers manipulate real operating activities to meet the zero earnings benchmark. Graham et al (2005) conclude from their survey that managers tend to manipulate real activities to meet earnings targets and forecasts, even though these activities may lead to a decrease in firms’ profitability in

(13)

the long-run. Roychowdhury (2006) focuses on real activities manipulation and presents main techniques of real earnings management. It suggests that managers allow higher price discounts to increase temporary revenue, support overproduction to decrease the costs of goods sold, and reduce discretionary expenses to improve net income. Zang (2012) investigates the underlying mechanism that drives managers’ choice between alternative earnings management strategies: real activities management and accrual earnings management.

By taking the effect of tax expense into consideration, managers’ earnings management activities can be classified as two alternative strategies. On the one hand, managers could manage earnings by engaging in accounting choices and business practices that simultaneously affect book income and taxable income. These decisions give rise to current tax expense (i.e. book-tax conforming earnings management). On the other hand, firms could manage earnings without creating current taxable income (i.e. book-tax nonconforming earnings management).

Previous studies on pretax earnings management have separately examined ‘conforming earnings management’ and ‘nonconforming earnings management’. Maydew (1997) adopts firms that switched accounting methods for tax purposes as sample, to investigate the effect of book-tax conformity consideration in firms’

accounting activities. Mills and Newberry (2001) analyze firms’ book-tax discrepancies by comparing the results of public and private manufacturing firms. Erickson et al. (2004) use restated firms that fraudulently overstated their earnings, and provide evidence on whether those firms paid additional taxes on the fraudulent earnings. Frank (2009) investigates the incentives of aggressive tax planning, and discovered its significantly positive relation with subsequent financial reporting. Atwood (2010) measures book-tax conformity in a precise manner, and discovers a strong, positive association between book-tax conformity and future cash flows. These researches help in the understanding of earnings management incentives and their effects on managers’ decision.

There are several literatures that combine earnings management practices with tax avoidance activities to examine the interaction between them. Haw et al. (2004) analyze the impact of high tax compliance on earnings management and find a significant, negative relation. The authors therefore suggested the regulators and government to pay

(14)

more attention to tax enforcement, in order to reduce earning management activities. Desai et al (2009) summarize previous academic findings regarding the association between earnings management and tax management practices. These studies reach the conclusion that book-tax differences as well as tax planning consideration are playing an important role in earnings management decisions.

There are a relatively limited number of studies that provide evidence on the characteristics of accounting standards or transactions that influence managers’ earnings management strategies. Badertscher et al (2009) develop comprehensive measurement of the net present value of tax benefits and the net expected detection costs of

nonconforming earnings management, and provided evidence on the trade-off between earnings management methods. Goncharov et al (2006) focus on Russian private companies to document the impact of tax regulation on earnings management practices. Tang et al (2011) use a sample of Chinese listed companies to investigate the impact of earnings management on book-tax differences. Existing evidence linking earnings management to tax avoidance has questioned the underlying motivation for the selection of alternative earnings management strategies, thus set up a solid foundation for this study. The emphasis of this paper on the selection of earnings management strategies stems from these prior researches and aims to achieve a deeper understanding toward this.

2.2 Hypotheses

Previous literatures have documented a general tendency of companies to manage taxes (Gramlich 1991; Maydew 1997; Keating and Zimmerman, 2000). There are significant and obvious benefits associated with the management of income taxes. The observation of Minnick and Noga (2010) suggests that firms are likely to invest sources into tax planning department, resulting in lower tax expense and better earnings

performance. Moreover, there are evidences that without valid enforcement mechanism (eg. the IRS), firms tend to eliminate tax expense by reporting zero taxable income (Maydew, 1997).

Motivated by the large benefits from tax saving activities, managers tend to obtain tax-specific knowledge not only from internal departments, but also from external expertise. Under this background, tax services provided by external agencies have attracted arising attention of executives. Specifically, external auditors are considered as

(15)

an important option of tax service provider. In 2002, the Sarbanes-Oxley Act (SOX) section 201 resolved uncertainty regarding auditor-provided tax service, by permitting the provision of tax service by firms' external auditors, as long as firms are able to receive a prior approval from the audit committee. In contrast, other major non-audit services by incumbent auditors are restrained. Since then, Auditor-provided tax service (ATS) was one of the few non-audit services that were still allowed by the SOX. A research performed by Krishnan et al (2012) suggests that many firms rely on the incumbent auditor for tax services, while investors perceive the benefits of ATS to be higher than the potential risks.

As proved by prior literature, firms pay larger amount of tax service fee when they have high taxable income, or high tax loss carry forward (Halperin and Lai, 2011).

Furthermore, there is evidence that companies with external tax services employ more tax management techniques (McGuire, 2012), suggesting a positive association between the purchase of tax service and the level of tax avoidance.

In this paper, I consider firms’ tax avoidance practices under the framework of earnings management. As firms that hire tax service agencies exhibit higher level of tax management activities relative to firms that do not purchase external tax services, I expect the purchase of tax service to associate with higher book-tax nonconformity earnings management.

From the above analysis, it is safe to reach the conclusion that, high quality tax service from providers is likely to result in more tax planning opportunities. Under the frame of earnings management, the tax planning opportunities lead to a preference of nonconforming earnings management strategy. Hence I predict that the purchase of tax service brings about high tax service quality, thus increase the extent of tax management and nonconforming earnings management.

To verify whether higher tax avoidance incentive associate with higher rate of nonconforming earnings management, I identify the quality of tax service as a proxy for tax avoidance incentive, and formulate the following hypothesis:

H1: Other things equal, companies that purchase high quality tax service engage in higher rate of nonconforming earnings management.

(16)

Evidence from Cloyd (1995), Lilian (1998), Cho et al (2006) and Madison (2014) indicates that firms with high book-tax differences result in higher IRS audit adjustment rates and the frequency of IRS penalties. Specifically, Lilian (1998) argues that book-tax nonconformity represent firms’ aggressive tax planning strategy. The empirical results of her study show that larger gap between book income and taxable income result in more IRS proposed audit adjustments. Moreover, Cho et al (2006) provide evidence to the generalizability of Lilian’s research. Their study presents a robust, positive association between firm’s book-tax differences and the subsequent IRS audit adjustments, while controlling for industry, time and country specific characteristics.

The potential scrutiny and penalty will cause a reputational damage to not only the firms but also their auditors, thus firms’ external auditors could exhibit low tolerance to firms’ book-tax difference and therefore, help to create a strict financial reporting environment toward this. Prior researches have already investigated the connection between the quality of firms’ external audit and the magnitude of book-tax difference. Empirical results indicate a lower book-tax difference of the firms that audited by Big 4 auditors, compared to those audited by non-Big 4 auditors. In addition, a positive association can be observed between non-audit service firms and the firms’ book-tax difference (Hung-Chieh, 2008). Collectively, auditors face higher reputational and commercial risks that arise from scrutiny if their clients report a high level of book-tax difference, while high quality auditors are more qualified to address this issue and protect their reputational and commercial benefits from being threatened. That is, under the frame of earnings management, firms that audited by high quality auditors may prefer to manage earnings while maintain high book-tax conformity.

Drawing upon the related literature, I predict that firms with high quality external auditors have higher demand for book-tax conformity and therefore, are less likely to manage earnings in a nonconforming manner. Since high quality external auditors act as a motivation of high book-tax conformity, it is expected that ceteris paribus, companies with high quality external audit engage in more nonconforming earnings management relative to those without high quality external auditor.

To provide further evidence to the research question, I formulate the following hypothesis:

(17)

H2: Other things equal, companies without high quality external audit engage in nonconforming earnings management to a greater extent than those have high quality external audit.

3. Research methodology

3.1 Sample

In this paper, I identify misstatement companies (i.e., companies that restated their financial statements in the subsequent year) from the July 2006 United States

Government Accountability Office (hereafter US GAO) report and the March 2007 US GAO report.

On 24 July, 2006, the United States GAO published a report in which they disclosed a list of 1390 listed companies that made financial restatement announcements. Those restatements were required as a result of accounting fraud or financial statement errors between 01 July, 2002 and 30 September, 2005.

On March 2007, the US GAO released an updated database of 2006 GAO report, in which it concluded the financial restatement announcements for the period 01 October, 2005 to 30 June, 2006. Up to 396 restated firms are disclosed in this report. In this paper, I combine the 2006 GAO report with updated 2007 GAO report to identify the

restatement announcement between 1 July, 2002 and 30 June, 2006.

The GAO database provides information including the name of each company, the primary listing market of the company, the company’s stock (or ticker) symbol, as well as the dates and reasons of restatement. It excludes announced restatement involving stock splits, changes in accounting rules, and other conditions that were not arise from

accounting irregularities. That is, the restatement announcements in GAO reports only include the correction of previous material misstatements of financial reporting numbers that result from errors in the application of accounting principles. In this way, this study is able to identify earnings management activity in a precise manner.

The 2006 GAO report, together with the 2007 updated report, provides a total amount of 1786 observations. As shown in figure 1 appendix A, the selection of final sample includes 3 steps. First, I delete those observations that restated their earnings in an upward manner, which indicates that they originally underestimated their earnings. Since this paper only focus on firms that engage in upward earnings management, data of those

(18)

firms that engage in downward earnings management should be excluded. Second, in order to obtain a precise result, I filter out the observations without adequate data available. Third, if one firm engages in upward earnings management in more than one year, I only reserve the data of the first year that restatement occurs. As a result of this selection process, I lose 1,635 observations and confirm a sample that includes a total amount of 151 observations.

The collection of data involves several steps. First, I take the Ticker symbol of firms from GAO report and use the Computat and AuditAnalytics identifier Ticker as the empirical definition of a firm. I use Compustat database available through WRDS (Wharton Research Data Services, University of Pennsylvaniato) to identify financial reporting data. Second, I collect information about the name of the auditor of the firm, and the amount of tax service fee that were paid to the auditors. The audit related data comes from the AuditAnalytics database available through WRDS. From these data, I identify the quality of external audit, and a sample of firms that rely on their auditors for tax service provision. Third, I hand collect other firm specific statistics that are not

available through WRDS. In order to start the analysis, the originally and restated income, together with the corresponding tax expenses are required. As the Compustat and

AuditAnalytics database are lack of some of these data, I collect the missing information from the companies’ annual financial statement. Lastly, the hand collected statistics are merged with the Compustat and AuditAnalytics data.

3.2 Measurement of theoretical constructs 3.2.1 Dependent and independent variables

To examine the earnings management methods of firms, I rely on measures of conforming and nonconforming earnings management drawn from the literature (Phillips et al, 2004; Burgstahler et al, 2006; Badertscher, 2006), and calculate the rate (proportion) of different earnings management strategies.

TA_EM, which stands for the total amount of earnings management, is measured by the originally reported pretax net income minus the restated pretax net income. Since firms that made a downward earnings restatement announcement is considered to have originally managed their earnings in a upward manner, the difference between originally reported pretax income and the restated pretax net income could capture the total

(19)

overstated amount, that is, the amount of earnings management. I use pretax net income instead of after-tax net income, as this paper is concerned with pretax earnings

management.

CON_EM stands for the total amount of conforming earnings management. It is measured by the equation CON_EM= (EM_CTE-ΔEM_NOL) / STR. In this equation, EM_CTE stands for the earnings management related to current tax expense, which is calculated by the originally stated current tax expense minus the restated current tax expense. ΔEM_NOL represents earnings management that related to net operating loss carry forward, it is measured by the originally stated change in deferred tax assets minus the restated change in deferred assets. ΔEM_NOL is deducted from EM_CTE to avoid the error of measuring upward conforming earnings management, as net operating carry forward could offset the changes of taxable income in restated firms. STR stands for statutory tax rate. In this paper I only include observations in the United States, the statutory tax rate remains at a stable amount of 35 percent.

NONC_EM represents the total amount of non conforming earnings management. It is measured by the total amount of earnings management divided by conforming earnings management.

RATE_NONC is the total amount of non conforming earnings management scaled by the total amount of earnings management. In this paper, the hypotheses focus on the relative amount of nonconforming earnings management compared to conforming earnings management, therefore, the dependent variable is the proportion rather than the actual amount of nonconforming earnings management. Similarly, RATE_CON

represents the proportion of conforming earnings management.

HQ_AUDIT stands for the quality of external audit. As is common in prior literatures, high quality auditors are defined as auditors from Big 4 audit firms. In this study, external audit quality is measured by a (0, 1) dummy variable assigned with a value of 1 if the firm adopt one of the big 4 firms as external auditor.

TAX_SVS represents the purchase of tax service from external auditor. $0 tax fee indicates that firms did not purchase tax service from their auditors. I include a dummy variable assigned with a value of 1 if the firm pays tax service fee to its external auditor, and 0 otherwise. TAX_SVS is expected to be positively associated with the rate of

(20)

nonconforming earning management. In addition, I introduce TAX_FEE to record the total amount of tax service fee paid to external auditors.

3.2.2 Control variables

Consistent with existing researches (Mills and Newberry, 2001; Anderson and Reeb, 2003; Manzon and Plesko, 2002; Phillips et al., 2004; Badertscher et al, 2009), I control for other factors that may affect the selection between earnings management strategies. The control variables include leverage (LEV), free cash flow (CASH), firm size

(LN_TA), growth (MTB) and the presence of net operating loss carry forward (NOL), net deferred tax liability (ND_TAXL), and the total amount of earnings management

(TA_EM).

I include leverage (LEV) ratio as firms with higher amount of leverage is less desired to engage in tax management practices. Due to the fact that debt may has tax shield effect, it is predictable that higher leverage is associated with lower level of nonconforming earnings management.

CASH, which stands for the free cash flow of the firms, is included as a control variable. As firms with higher free cash flow are in better financial position, they are less likely to take advantages of tax planning activities. I expect that CASH will be negatively associates with the rate of nonconforming earnings management. The level of cash flow is measured by the net cash flow from operating activities plus the dispose of property, plant and equipment (PPE), less capital expenditures and acquisitions, at the beginning of the restatement year.

LN_TA, measured by the natural logarithm of the total amount of asset for firm, controls for the size of a firm. Firms that have a large size could have complex financial reporting environment and accounting situation. As larger firms have more potential opportunity to manage earnings in a nonconforming manner, I predict LN_TA to be positively associated with the rate of nonconforming earnings management.

I use market to book ratio (MTB) to control for the growth of firms. Since fast growing firms are more motivated to delay the recognition of current tax expenses, I predict that higher market to book ratio to associate with higher rate of nonconforming earnings management, that is, a positive association.

(21)

NOL stands for net operating loss (NOL) carry forward, it is available when companies’ allowable tax deductions exceed taxable income, which could lead to a negative taxable income. When a company has certain tax-deductible expenses that are higher than taxable revenues during a taxable year, it has a net operating loss carry forward available at the beginning of the subsequent year. I predict that firms have NOL carry forward have a lower amount of tax benefit associated with nonconforming

earnings management. That is, I expect a negative correlation between NOL carry forward and the rate of nonconforming earnings management.

ND_TAXL, measured by the amount of net deferred tax liability, is used to control for the earnings management result from previous activities. If a firm already has a high amount of net deferred tax liability available on the balance sheet at the beginning of year, this might constrain firm’s nonconforming earnings management activities of the

subsequent year. Therefore, it is predictable that ND_TAXL is negatively associated with the rate of nonconforming earning management.

TA_EM controls for the total amount of earnings management. A large magnitude of earnings management provides the motivation to managers to select the more desired earnings management strategy (nonconforming earning management). Hence I predict that the larger the amount of total earning management is, the more likely that managers employ nonconforming earnings management techniques. In other word, a positive relationship between TA_EM and the rate of nonconforming earnings management.

To test the hypotheses, I estimate a regression model that is presented as follows. RATE _NCON i, t = α0 + α1 HQ_AUDIT i, t + α2 TAX_SVS i, t + α3 NOLi,t-1 + α4

CASH i,t-1 + α5 MTBi,t-1 + α6 LN_TA i,t-1+α7 ND_TAXL i,t-1 +α8 TA_EM i, t +α9 LEV i,t-1 +

ε

To make it easier for reading, I provide the definitions and measurements of all the variables as following:

NCON_RATE i,t = The amount of nonconforming earnings management scaled by

the total amount of earnings management for firm i, at year t;

HQ_AUDIT i,t = (0, 1) dummy variable equals to 1 if the auditor of firm i at year t is

(22)

TAX_SVS i,t = (0, 1) dummy variable valued 1 if firm i pay tax service fee to

external auditor at year t, otherwise equals to 0;

NOL i,t-1 = (0, 1) dummy variable coded as 1 if firm i’s net operating loss carry

forward is positive as of the beginning of the year t, otherwise coded as 0; MTB i,t-1 = Market to book ratio for firm i, at the beginning of year t;

LN_TA i, t-1 = Natural logarithm of the total amount of asset for firm i, at the

beginning of year t;

ND_TAXL i,t-1 = Net deferred tax liability of firm i, measured by the total amount of

deferred tax liability minus the total amount of deferred tax asset, at the beginning of year t;

TA_EM i,t = The total amount of earnings management of firm i, measured by the

originally reported pretax net income less restated pretax net income, at year t;

LEV i, t-1 = The leverage of firm i, measured by the total asset divided by total asset

less total liability, at the beginning of year t;

CASH i,t-1 = The net cash flow of firm i at the beginning of year t. It is measured by

net cash flow from operating activities + the dispose of property, plant and equipment (PPE)- capital expenditures – acquisitions.

4. Empirical Result

To provide answer to the research question and test the hypotheses, I perform a three-step analysis. First, I use descriptive statistics to present a general and direct description of firm specific characteristics and variables. Second, a univariate analysis is performed to analyze the correlation between dependent variables, independent variables and control variables. Third, a multivariate test is included to mitigate the negative impact of noisy information and provide evidence for hypotheses.

4.1 Descriptive Statistics

Descriptive analysis is performed in this paper to provide elaborate description of the tendency and dispersion of a set of statistics.

Table 1 presents the descriptive statistics with respect to the final sample of 151 firms from 2002-2006. Specifically, column 2-5 presents the minimum, means, maximum value, and standard deviation of 151 observations respectively.

(23)

In general, table 1 suggests that the rate of nonconforming management in sample firms tend to be substantially larger than the rate of conforming earnings management. Specifically, the minimum, maximum and mean values of the proportion of

nonconforming earnings management among target firms are 0.17, 0.64 and 0.98 respectively. This is significantly higher relative to those values of conforming earnings management, which are 0.02, 0.36, 0.83 respectively. The standard deviations of the rate of conforming and nonconforming earnings management are similar, amount to 0.13 when the figures are accurate to two decimal places.

Table 1 Descriptive Statistics

Minimum Mean Maximum Standard Deviation

Dependent variable RATE_NONC 0.17 0.64 0.98 0.13 Independent variables HQ_AUDIT 0.00 0.87 1.00 0.33 TAX_FEE 0.00 162901.38 4100000.00 510266.46 TAX_SVS 0.00 0.26 1.00 0.44 Control variables NOL 0.00 95.21 2768.90 377.79 CASH (7.91) 176.84 6625.00 706.02 MTB (3.36) 3.06 75.70 6.41 LN_TA 0.78 2.67 4.62 0.69 ND_TAXL (1607.00) (11.69) 1080.00 203.41 LEV (0.21) 0.53 1.00 0.24 Other variables TA_EM 0.01 38.24 1641.00 192.50 TA_CON 0.01 14.55 647.00 76.23 RATE_CON 0.02 0.36 0.83 0.13

RATE_NONC is the rate of nonconforming earnings management, measured by the amount of

nonconforming earnings management scaled by the total amount of earnings management. HQ_AUDIT is the quality of firm's external audit, it is a dummy variable equals to 1 if the auditor is one of the Big 4 accounting firms, and 0 otherwise. TAX_FEE is the amount of tax service fee. TAX_SVS is the purchase of tax service, it is a dummy variable valued 1 if firm pay tax service fee to external auditor, and 0 otherwise. NOL is the availability of net operating loss carry forward, it is a dummy variable coded as 1 if firm’s net operating loss carry forward is positive at the beginning year, and 0 otherwise. CASH is firm's net cash flow at the beginning of year. MTB is market to book ratio at the beginning of year. LN_TA is natural logarithm of firm's total amount of asset. ND_TAXL is net deferred tax liability of firm at the beginning of year. LEV is the leverage of firm at the beginning of year. TA_EM is the total amount of earnings management. TA_CON is the total amount of conforming earnings management. RATE_CON is the rate of conforming earnings management, measured by 1 minus the rate of nonconforming earnings management.

(24)

There is a highly deviation with respect to tax service fee. Statistically, the minimum amount of tax service fee is 0, while the maximum amount reaches 4,100,000. The

standard deviation is 510,266.46 with respect to the tax service fee, makes it the variable with the highest deviation. The result represents that some firms did not rely on auditor-provided tax service, while a number of firms invest a large sum of money in tax planning practices.

The mean value of HQ_AUDIT amount to 0.8742, suggesting a large proportion of the observation films hire big 4 external auditors, who are expected to provide high quality audit services. The minimum, maximum and mean values are significantly different with respect to the total amount of earnings management, the magnitude of conforming earnings management, market-to-book ratio, net cash flow, leverage, net operating loss carry forward, and net deferred tax liability. The standard deviation result shows a high discrepancy as well. The remarkably differences between observations indicate a large diversity between not only firm specific characteristics, but also the preference towards tax planning and earnings management activities.

In conclusion, table 1 reports minimum, maximum, means and standard deviation of audit quality, purchase of tax service, amount and rate of different earnings management, net operating loss carry forward, net cash flow, leverage, market-to-book ratio and net deferred tax liability with respect to 151 sample firms. It helps describe and summarize about the sample, provides a general summary of the statistics and set up a solid base for the upcoming analysis.

4.2 Univariate analysis

In this paper, I aim to test the association between different incentives and the selection of different earnings management strategies. In this setting, the univariate analysis is incorporated to check if the variables are related to one another.

The result indicates a negative association between RATE_NONC and HQ_AUDIT, and a positive correlation between RATE_NONC and TAX_SVS. The coefficients are -0.083 and 0.023, respectively. Statistically, the correlations are not significant.

It is worthwhile to note that RATE_NONC is negatively associated with CASH, NOL and ND_TAXL, which is consistent with the analysis in the research methodology section. The coefficients of association between RATE_NONC and CASH, NOL,

(25)

ND_TAXL are -0.025, -0.02 and -0.003 respectively. The univariate analysis also

produces other signs that essentially in agreement with the previous discussion, while the correlations are not significant enough. Specifically, it shows a positive correlation between MTB and the rate of nonconforming earnings management, and provides evidence regarding the negative association between NOL carry forward and firms’ preference of nonconforming earnings management.

Table 2 Univariate analysis HQ_ AUD IT TAX _SVS NOL CAS H MTB LN_TA ND_TA XL LEV TA_ EM RATE _NON C HQ_ AUDIT 1 0.23** 0.09 0.03 0.01 0.33** (0.03) (0.12) 0.07 (0.08) TAX_ SVS 0.23** 1 0.11 0.23** 0.17* 0.27** (0.13) (0.12) 0.21** 0.02 NOL 0.09 0.11 1 0.11 (0.01) 0.34** 0.09 (0.13) 0.08 (0.02) CASH 0.08 0.23** 0.11 1 (0.01) 0.43** (0.72)** (0.08) 0.97** (0.03) MTB 0.01 0.17* (0.01) (0.01) 1 (0.01) 0.07 (0.04) -.026 0.04 LN_TA 0.33** 0.27** 0.34** 0.43** (0.01) 1 (0.19)* (0.49)** 0.39** (0.14) * ND_ TAXL (0.03) (0.13) 0.09 (0.72) ** 0.07 (0.19)* 1 (0.01) (0.65)** (0.01) LEV (0.12) (0.12) (0.13) (0.08) (0.04) (0.49)** (0.01) 1 (0.08) 0.02 TA_EM 0.07 0.21** 0.08 0.97** (0.03) 0.39** (0.65)** (0.08) 1 (0.03) * RATE_ NONC (0.08) 0.02 (0.02) (0.03) 0.04 (0.14) * (0.01) 0.02 (0.03) * 1 *: Correlation is significant at 0.01≤α≤0.05. **: Correlation is significant at α≤0.01.

T statistics and Pearson significant levels are based on standard errors that clustered by firm. RATE_NONC is the rate of nonconforming earnings management, measured by the amount of

nonconforming earnings management scaled by the total amount of earnings management. HQ_AUDIT is the quality of firm's external audit, it is a dummy variable equals to 1 if the auditor is one of the Big 4 accounting firms, and 0 otherwise. TAX_FEE is the amount of tax service fee. TAX_SVS is the purchase of tax service, it is a dummy variable valued 1 if firm pay tax service fee to external auditor, and 0 otherwise. NOL is the availability of net operating loss carry forward, it is a dummy variable coded as 1 if firm’s net operating loss carry forward is positive at the beginning year, and 0 otherwise. CASH is firm's net cash flow at the beginning of year. MTB is market to book ratio at the beginning of year. LN_TA is natural logarithm of firm's total amount of asset. ND_TAXL is net deferred tax liability of firm at the beginning of year. LEV is the leverage of firm at the beginning of year. TA_EM is the total amount of earnings management.

The number of observation is 151.

The correlation tests indicate that RATE_NONC is significantly negatively associated with LN_TA and TA_EM, which is not consistent with the expectation. Statistically, the coefficient of the correlation between RATE_NONC and LN_TA is -0.14. That of TA_EM is -0.03. Despite this, the univariate test is, in general, consistent with the expectation.

(26)

3.3 Multivariate analysis

In addition to the univariate test of the hypotheses, I perform a multivariate analysis that includes RATE_NONC as dependent variable. I include HQ_AUDIT and TAX_SVS as independent variables. In addition, I control for potential differences across the sample firms that may confound the multivariate regression, the control variables include NOL, CASH, MTB, LN_TA, ND_TAXL, LEV and TA_EM.

Table 3 Multivariate analysis

Expected sign Coefficient T statistic Pearson value

Dependent variable: RATE_NONC 10.285 .000 Independent variables: HQ_AUDIT - -.048 -.532 .595 TAX_SVS + .060 .670 .504 Control variables: NOL - .031 .342 .733 CASH - .103 .240 .810 MTB + .027 .319 .750 LN_TA + -.200 -1.669 .097 ND_TAXL - -.014 -.108 .914 LEV - -.073 -.750 .455 TA_EM + -.074 -.197 .844

This multivariate analysis aims to test the following regression:

RATE _NCON i, t = α0 + α1 HQ_AUDIT i, t + α2 TAX_SVS i, t + α3 NOLi,t-1 + α4 CASH i,t-1 + α5 MTBi,t-1 + α6

LN_TA i,t-1+α7 ND_TAXL i,t-1 +α8 TA_EM i, t +α9 LEV i,t-1 + ε

In general, the multivariate analysis shows a consistent result with the expected correlation sign. However, the coefficients are not statistically significant at less than 0.1 level. An exceptional is the coefficient of LN_TA, which indicates a significant, negative association between the natural logarithm of firms’ total asset and the rate of

(27)

Collectively, one potential explanation for the multivariate result is the inexact measurement of variables. To address this issue, I perform an additional analysis to test the sensitivity of the results.

5. Additional test

To provide evidence to hypothesis 1, I use the purchase of tax service as a proxy for tax planning quality, and test the association between the adoption of tax service and the rate of nonconforming earnings management. The results indicate an insignificant, positive relationship between the two. For the above analysis, TAX_SVS is measured as (0, 1) dummy variable coded as 1 if the firm purchase tax service from external auditors, and 0 otherwise.

As discussed in the descriptive statistics section, the tax service fees paid to external auditors are significantly different between firms. As the level of auditor effort is

positively associated with audit fee (Hanlon, 2012), it is reasonable to deduce that different amount of tax fee could affect tax service quality and result in different tax planning consequences.

To mitigate the effect of firms’ tax service quality differences, I include TAX_FEE instead of TAX_SVS to capture the quality of tax service in the univariate test and multivariate analysis. In this setting, the actual amount of tax service fee is considered in the tests.

5.1 Univariate test

Table 4 shows the univariate analysis result of the additional test. It is observable that TAX_FEE shows a positive association with RATE_NONC, with a coefficient of 0.07. The correlation is not statistically significant.

Previous univariate analysis has also presented insignificant positive association between TAX_SVS and the rate of nonconforming earnings management, with a coefficient of 0.02. By comparing the result of previous analysis with that of additional univariate analysis, it is deducible that the magnitude of tax service fee does affect the rate of nonconforming earnings management, as the coefficient between TAX_FEE and RATE_NONC is statistically different from that between TAX_SVS and RATE_NONC. Moreover, the larger coefficient for TAX_FEE indicates a positive influence of tax service fee upon the selection of nonconforming earnings management.

(28)

Table 4 Univariate analysis HQ_

AUD IT

TAX _FEE

NOL CASH MTB LN_TA ND_T

AXL LEV TA_E M RATE_ NONC HQ_A UDIT 1 0.12 0.09 0.08 .014 0.33** (0.03) (0.12) 0.07 (0.08) TAX_F EE 0.12 1 0.11 0.39** 0.09 0.37** 0.02 (0.16) 0.45** 0.07 NOL 0.09 0.11 1 0.11 (0.01) 0.34** 0.09 (0.13) 0.08 (0.02) CASH 0.08 0.39** 0.11 1 (0.01) 0.43** (0.72)** (0.08) 0.97** (0.03) MTB 0.01 0.09 (0.01) (0.01) 1 (0.01) 0.07 (0.04) (0.03) 0.04 LN_TA 0.33** 0.37** 0.34** 0.43** (0.01) 1 (0.19)* (0.49)** 0.39** (0.14) ND_T AXL (0.03) 0.02 0.09 (0.72)** 0.07 (0.19)* 1 (0.01) (0.65) ** (0.01) LEV (0.12) (0.16) (0.13) (0.08) (0.04) (0.49)** (0.01) 1 (0.08) .016 TA_E M 0.07 0.45** 0.08 0.97** (0.03) 0.39** (0.65)** (0.08) 1 (0.03) RATE_ NONC (0.08) 0.07 (0.02) (0.03) 0.04 (0.14) (0.01) 0.02 (0.03) 1 *: Correlation is significant at 0.01≤α≤0.05. **: Correlation is significant at α≤0.01.

5.2 Multivariate analysis

In the next step, I perform a multivariate analysis by estimating the coefficients with the following regression model:

RATE _NCON i, t = α0 + α1 HQ_AUDIT i, t + α2 TAX_FEE i, t + α3 NOLi,t-1 + α4

CASH i,t-1 + α5 MTBi,t-1 + α6 LN_TA i,t-1+α7 ND_TAXL i,t-1 +α8 TA_EM i, t +α9 LEV i,t-1 +

ε

TAX_FEE i,t stands for the total amount of tax service fee paid to external auditors

for firm i, at year t. In the regression estimation, RATE_NONC, HQ_AUDIT, NOL, CASH, MTB, LN_TA, ND_TAXL, and TA_EM are measured as described earlier in previous section.

(29)

In Table 5 I present the result of the multivariate test with a number of variables that had already been discussed in the research design section.

Table 5 Multivariate analysis

Expected sign Coefficient T statistic Pearson value

Dependent variable: RATE_NONC 10.539 .000 Independent variables: HQ_AUDIT - -.042 -.480 .632 TAX_FEE + .202 1.886 .061 Control variables: NOL - .043 .478 .634 CASH - .189 .445 .657 MTB + .019 .224 .823 LN_TA + -.240 -1.989 .049 ND_TAXL - -.107 -.761 .448 LEV - -.075 -.783 .435 TA_EM + -.282 -.726 .469

This multivariate analysis aims to test the following regression:

RATE _NCON i, t = α0 + α1 HQ_AUDIT i, t + α2 TAX_FEE i, t + α3 NOLi,t-1 + α4 CASH i,t-1 + α5 MTBi,t-1 + α6

LN_TA i,t-1+α7 ND_TAXL i,t-1 +α8 TA_EM i, t +α9 LEV i,t-1 + ε

T statistics and Pearson significant levels are based on standard errors that clustered by firm. RATE_NONC is the rate of nonconforming earnings management, measured by the amount of nonconforming earnings management scaled by the total amount of earnings management.

HQ_AUDIT is the quality of firm's external audit, it is a dummy variable equals to 1 if the auditor is one of the Big 4 accounting firms, and 0 otherwise. TAX_FEE is the amount of tax service fee. TAX_SVS is the purchase of tax service, it is a dummy variable valued 1 if firm pay tax service fee to external auditor, and 0 otherwise. NOL is the availability of net operating loss carry forward, it is a dummy variable coded as 1 if firm’s net operating loss carry forward is positive at the beginning year, and 0 otherwise. CASH is firm's net cash flow at the beginning of year. MTB is market to book ratio at the beginning of year. LN_TA is natural logarithm of firm's total amount of asset. ND_TAXL is net deferred tax liability of firm at the beginning of year. LEV is the leverage of firm at the beginning of year. TA_EM is the total amount of earnings management.

For the correlation between HQ_AUDIT and RATE_NONC, the sensitivity tests generate similar inferences. There is a negative association between HQ_AUDIT and RATE_NONC. The coefficient is not significant, while with an amount of -0.042, it reaches an agreement with the expected sign.

In consistence with the previous test, the additional test presents positive coefficients for NOL, CASH and MTB, while the coefficients for ND_TAXL, LEV and TA_EM are negative, at -0.107, -0.075 and -0.282 respectively.

It is noteworthy that TAX_FEE shows a positive correlation with the rate of nonconforming earnings management. With a coefficient of 0.202, the correlation is significant at less than 0.1 level. This result is consistent with hypothesis 1, suggesting

(30)

the positive influence of tax service upon managers’ preference of nonconforming earnings management. The test also provides evidence to the association between tax service fee and the level of tax avoidance. In the previous test, I examine the association between TAX_SVS and RATE_NONC, and discover a correlation which is not

statistically significant. In this additional test, however, a statistically significant relation is observed between the amount of tax service fee and the level of nonconforming earnings management. The different outcome of the additional test also indicates a positive association between the magnitude of tax service fee and the extent of tax avoidance.

In contrast to the expectation, the multivariate test presents a coefficient of -0.24 for the relation between LN_TA and RATE_NONC. The correlation is significant at less than 0.05 level. In the previous discussion section, I expect firms’ total asset to be positively related to the preference of nonconforming earnings management, as larger firms with more complex financial reporting environment have more opportunities to engage in book-tax nonconforming earnings management. However, the additional multivariate test indicates a significant negative relationship between firm size and the preference of nonconforming earnings management. One reasonable explanation is the avoidance of reputation damage. Since larger firm could face more strict public scrutiny, and suffer stronger pressure once disclosed as tax manipulator, the detection cost of nonconforming earnings management may constrain them from engaging in such activities.

Taken together, the empirical results provide support to hypothesis 1. They add to the literature on tax management and earnings management by demonstrating a

significant relation between tax service fee and firms’ preference of nonconforming earnings management.

6. Conclusion

While a considerable number of research has separately examined managers’ incentives to engage in tax planning and earnings management activities, relatively little work has been done to examine factors that influence the selection between different earnings management strategies. This paper focuses on two such factors: the purchase of tax service, and the quality of firms’ external auditors. The first hypothesis is that ceteris

(31)

paribus, companies that purchase high quality tax service engage in higher rate of nonconforming earnings management. The second hypothesis is, ceteris paribus, companies without high quality external audit engage in nonconforming earnings management to a greater extent than those have high quality external audit.

The sample of this paper, developed from the 2006 and 2007 US GAO reports, covers the 2002-2006 time period. Specifically, it consists of 151 observations for which the originally reported pretax net income is higher than the amount of restated pretax net income. Since the earnings of these firms were downward adjusted in the restatement report, it is reasonable to deduce that their managers originally engaged in upward earnings management techniques.

I perform descriptive analysis, univariate analysis and multivariate analysis to provide support for the hypotheses. Moreover, I supplement the analysis by including an additional test in which the magnitude of tax service fee is included as an independent variable. The empirical results indicate a robust, positive association between the amount of tax service fee and the rate of firms’ nonconforming earnings management, suggesting a positive effect of high quality tax service on the level of tax avoidance techniques. The findings hold after accounting for other variables identified in the prior literature to affect the selection between alternative earnings management techniques. There is a negative coefficient for the relation between the quality of external auditor and the rate of nonconforming earnings management, yet the association is not significant enough.

This paper complements the existing literature on earnings management practices in the following ways. First, Instead of examining one earnings management strategy, I investigate two alternative earnings management at the same time, thus provides the comprehensive picture of earnings management practices. By measuring the relative amount of each earnings management strategy, this paper goes deeper into managers’ decision making process to detect the incentives of their behaviors. Second, with the restatement announcement in GAO report, I identify the firms’ that engage in upward earnings management activities and subsequently restate their earnings in a downward manner. Restatement announcement shows a clearly and accurately identification of a setting in which earnings management have happened. It helps to better identify the firms that manage earnings. With a list of firms that engage in upward earnings management, I

(32)

am able to employ proxies that capture the overall earnings management and the proportion of different earnings management practices.

Despite the contributions, this research is subject to limitation. First, in this paper, the amount of tax fee paid to auditors is utilized as a proxy of tax service quality. In this way, it is indicated that all the tax services are provided by auditors. In practice, however, firms could purchase tax service from non-auditors (eg. tax service agency). Due to the unavailability of data regarding the tax fees paid to non-auditors, the proxy is not precise enough to capture the availability of tax service adopted by companies.

Second, the total amount of nonconforming earnings management is calculated using the equation CON_EM= (EM_CTE-ΔEM_NOL) / STR. In this equation, STR stands for the statutory tax rate of United States and was evaluated by 35 percent, thus all the changes in firm’s current tax expense is considered to be attributable to the

differences between the originally reported and the restated U.S. pretax income. In this setting, all the foreign earnings of U.S. multinationals are assumed to be repatriated back to the U.S.. This assumption is based on the U.S. tax law, by which U.S. multinationals are required to pay U.S. income taxes upon repatriation of foreign affiliates earnings earned in lower tax jurisdictions. Practically, however, much of the earnings derived from foreign subsidiaries of U.S. multinationals were not transferred back to the U.S.. It is revealed by a recent analysis performed by JP Morgan Analysts that, the statutory tax in the United States is one of the highest among developed countries. As a result of the high corporate tax rate in U.S., U.S. multinational companies are likely to avoid or at least defer, the tax liability arises from earnings' repatriation. Hence the restated changes in firm’s current tax expense are attributable not only to the differences between the

originally reported and the restated U.S. pretax income, but also to that of foreign pretax income. That is, the tax rate, measured by the statutory tax rate of United States, is overestimated, which could lead to a potential underestimation of conforming earnings management.

Third, it is revealed by the empirical result that, the magnitude of firm’s tax service fee is positively associated with the rate of nonconforming earnings management. With respect to this result, one explanation is the influence of tax service upon the level of tax avoidance, which has been discussed in the hypothesis section of this paper. However,

(33)

there exists another explanation for the association, that is, the effect of managers’ tax avoidance preference on the charge of auditor-provided tax service. As is presented in the introduction section, firms with higher rate of nonconforming earnings management are more likely to be detected as aggressive tax planner, which may cause reputational risk to the auditors. Since it is widely accepted that managers engage in earnings management activities when they perceive the benefits of managing earnings to be higher than the costs (eg. Healy and Wahlen, 1999), it is questionable whether auditors, act as tax service provider, engage in tax avoidance activities when they perceive the margin benefits of tax planning to be higher than the margin costs. As such, auditors also face a trade-off, between providing high quality tax service and maintaining low book-tax difference. Hence a higher amount of tax service fee can be seemed as an offset of the potential subsequent reputation loss. Collectively, the interaction between tax service fee and managers’ preference of nonconforming earnings management is an open topic for future researchers.

Fourth, due to data unavailability, this paper is not able to consider private firms, or make a comparison between the empirical result of private and public listed firms. As public scrutiny is stricter for public firms compared to that of private firm, it is

predictable that ceteris paribus, private firms show higher preference of nonconforming earnings management. Future study on non-listed firms would further complement this research.

Overall, this paper has implications for regulators, stakeholders and investors. By considering the influence of different incentives upon managers’ preferences, regulators could link firm characteristics with the expected implementation of regulations. For instance, tax authorities could impose more scrutiny on firms with higher amount of tax service fee, as managers of those firms tend to manage earnings without creating current tax expense. By analyzing firm specific characteristics that are discussed in this paper, both analysts and investors could get inspiration in predicting managers’ behavior and the subsequent earnings performance. Furthermore, relying on the theoretical and empirical results of this paper, future researchers could extend their understanding of this subject and confirm an objective to work on.

Selection between earnings management strategies : book-tax conforming versus book-tax nonconforming