Recent firm performance and the extent of tax aggressiveness as a practice to manage earnings

(1)

1

Recent firm performance and the extent of tax aggressiveness as a

practice to manage earnings

Name:

Martijn van Zanten

Student Number:

10450831

Thesis Supervisor:

Reka Felleg

Date:

25-06-2017

Word Count:

14,908

Study:

MSc Accountancy & Control, accountancy track

Education Institution: University of Amsterdam, Amsterdam Business

(2)

2

Statement of Originality

This document is written by Student Martijn van Zanten 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)

3

Abstract

This study provides evidence on the relationship between recent firm performance and the extent of earnings management by applying aggressive tax planning strategies. Specifically, this study investigates whether recent firm performance cause lower or higher effective tax rates in the near future. The results suggest that a significant negative relationship exists between recent firm performance and the effective tax rate. This implies that managers are more risk averse in their choice of earnings management practices after a poor recent firm performance and less risk averse after a good recent firm performance, since aggressive tax planning strategies are considered to be a risky tool to manage earnings, because of the reputational costs and fines that firms could receive if the tax laws are broken.

Keywords: earnings management, tax planning, tax aggressiveness, effective tax rate, recent firm performance

(4)

4

1. Introduction

This study examines whether the use of aggressive tax planning strategies as a practice of earnings management is determined by recent firm performances. Specifically, I examine whether a negative relationship exists between recent firm performance and the effective tax rate, which would imply a positive relationship between recent firm performance and the extent of earnings management by applying aggressive tax planning strategies.

Most prior literature on earnings management focuses on accrual-based earnings management and real earnings management, which are widely recognized as the two main categories of earnings management (Zang, 2012; Cohen, Dey, & Lys, 2008). A big part of this extant literature focuses on the trade-off between accrual-based earnings management and real activities earnings management. This literature shows that since the beginning of this century firms tend to use more real earnings management instead of accrual-based earnings management. The underlying reasoning is that society has become more skeptical towards accrual-based earnings management as the consequence of the numerous accountings scandals at the end of the previous century. Moreover, the passing of the Sarbanes-Oxley in 2002 means that managers of firms can get heavier punishments if the financial reporting rules are broken (Cohen, Dey, & Lys, 2008).

In addition to accrual-based earnings management and real earnings management, income taxes can be used to manage earnings. Since the tax rate in the United States equals 35 percent, income taxes is a big line in the financial statement, which means that it can be well used to manage earnings (Dhaliwal, Gleason, & Mills, 2004). However, this is not the most known practice of tax planning, since it can also be used to save money by preventing or delaying a shift of resources from the firms to the government. Academic literature, but also media, like newspapers and television shows, focus mainly on the money-saving

explanation of tax aggressiveness. This motivates me to highlight in this study that aggressive tax planning strategies are also used to manage earnings and under what circumstances this is done by firms and their managers.

The aim of this study is to investigate whether a relationship can be found between recent firm performance and the use of aggressive tax planning strategies to manage earnings. By fulfilling this aim this article contributes to extant literature on the traditional types of earnings management by giving insights in how and under what circumstances aggressive tax planning strategies are applied as a practice of earnings management. Specifically, valuable insights about how a change in risk appetite decrease or increase the use of aggressive tax

(7)

7 planning strategies to manage earnings.

An important characteristic of aggressive tax planning as a tool of earnings

management is that with this type of earnings management firms actually save money, since tax aggressiveness prevents or delays a shift of resources from the firm and its investors to the government (Kirchler, Maciejovsky, & Schneider, 2003). However, next to the advantage of saving money, firms can also face disadvantages from being too aggressive in applying a tax planning strategy, since reputational costs and fines can be a serious risk.

I used the concept of managing earnings by applying an aggressive tax planning strategy to investigate whether the risk appetite of a firm and its managers changes after a good or bad recent firm performance. A change in risk appetite can possibly result in an increasing or decreasing use of aggressive tax planning strategies as a tool to manage earnings, since the changed risk appetite can influence a managers’ willingness to face the risks of applying an aggressive tax planning strategy.

In order to contribute to the existing literature, the following research question is answered in this article:

“Does recent firm performance cause a change in the extent of tax aggressiveness that is applied as a tool to manage earnings?’’

To give a well-structured answer on the research question, I use a sample consisting of 29,860 firm-year observations of 6,136 firms headquartered in the United States for the period from 2003 until 2016. By comparing a firms’ performance to the industry-adjusted ROA, as is also done in the article of Eisenberg, Sundgren and Wells (1998), the half of the observations are qualified as good-performing and the other half as bad-performing.

Following Stickney and McGee (1982) and Gupta and Newberry (1997) I ran an ordinary least squares regression on the sample to gather empirical evidence to answer my research question. Based on the result of this regression, I find that a negative association exists between recent firm performance and the effective tax rate, which means that firms and its managers are less likely to use aggressive tax planning strategies to manage earnings after the firm performance was poor in recent history and more likely to do so if the recent firm performance was good.

These findings imply that the risk appetite of a firm’s manager influences the decision about the types of earnings management to use. Results further imply that recent firm

performance changes the managers’ willingness to take risks. Poor recent performance leads to more risk averse behavior. Since aggressive tax planning strategies are considered to be a

(8)

8 risky way of managing earnings, less earnings management by tax aggressiveness is applied after a firm have performed poorly in the recent history.

A practical implication of these findings is that it can help national tax authorities in the fulfillment of their enforcement duty. Good firm performances can be an indication for tax authorities that firms are more likely to apply an aggressive tax planning strategy in the near future.

The remainder of this thesis will be as follows. The second section contains a theoretical framework, where the costs and benefits of different types of earnings

management will be further elaborated. Furthermore, the prospect theory of Kahneman and Tversky (1979) will be elaborated to state why risk appetites of managers can possibly

change if firms are performing poor. With the help of this theoretical framework a hypothesis is constructed in the third section of this article. Section four is the methodology section. In this section the way of doing research will be explained. Special attention will be given to the sample selection and the regression that is used. After the methodology section, the results are presented in the fifth section. This section contains the descriptive statistics, results on the OLS regression, robustness tests and additional analyses. The last section of this thesis is the conclusion, which also include a discussion and some recommendations for subsequent research.

(9)

9

2. Literature

2.1 Earnings management

Earnings management is an extensively investigated topic in the accounting literature. A definition of earnings management that is given in the article of Paul Healy and James Wahlen (1999, p.368) is the following:

“Earnings management occurs when financial managers use judgment in financial reporting and in structuring transactions to alter financial reports to either mislead some stakeholders about the underlying performance of the company or to influence contractual outcomes that depend on reported accounting numbers.’’

As can be derived from this definition, two criteria have to be met to be involved in an earnings management practice. Firstly, the financial management have to use its own

judgment to influence the way the underlying business activities are reflected in the financial reports. This criterion is actually a criterion that is often met, since most accounting standards oblige the financial managers to use their own judgment in the financial statements.

The second criterion is that the earnings management practices are used to mislead stakeholders or influence several contractual incomes. This means that the financial managers aim to use their discretion to gain a benefit for their selves or the shareholders as a result from the earnings management practices.

2.1.1 Accrual-based earnings management

Accrual-based earnings management means that managers use their own judgement to influence the accounting choices in a way that stakeholders are misled or contractual outcomes are influenced in their own advantage or in the advantage of the firm. The consequence is that the presentation of the underlying activities are changed, but not the underlying activities themselves (Kothari, Mizik, & Roychowdhury, 2016).

Two types of accruals are distinguished within prior literature: discretionary accruals and non-discretionary accruals (Geiger & North, 2006). The non-discretionary accruals are the accruals that can normally be expected to be present in the financial reports if the accounting standards, size of the firm, operating industry and other firm characteristics are taken into account.

On the other hand, the discretionary accruals are the accruals that exceed the expected level of non-discretionary accruals. A characteristic of discretionary accruals is that they are

(10)

10 subject to more financial reporting judgment and consequently more used to manage earnings (Geiger & North, 2006). Depreciation costs and the write-off of inventories are examples of discretionary accruals.

2.1.2 Real activities earnings management

Another type of earnings management is real activities earnings management. This means that a firm times or structures its real activities differently with the consequence that the financial results differ negatively from the financial results that would have been the result of normal, optimally structured business operations. The goal is again to mislead stakeholders or influence contractual outcomes (Kothari, Mizik, & Roychowdhury, 2016). The difference of manipulated activities with respect to normal business activities is that the business

consequences are suboptimal (Zang, 2012).

Examples of real activities manipulation are a decrease in research and development costs and a decrease in advertising costs. This type of earnings management is hard to determine, since there are no guidelines for how normal business operations should be exercised.

2.2 Incentives for earnings management

During the last decades many research has been done on the factors that are an incentive for firms and its managers to be engaged in earnings management practices. As seen in the definition stated by Healy and Wahlen (1999) earnings management is used to mislead stakeholders or to influence the outcome of contracts.

Earnings management is a very effective way to mislead the stakeholders and to influence the outcomes of contracts, since earnings are considered to be one of the most important lines in the financial statements by analysts, investors, senior executives, and boards of directors (Degeorge, Patel, & Zeckhauser, 1999). First of all, reported earnings are often used as a part of the screening process by investors to get an understanding of the risk they are facing by investing in a company. Since the earnings line plays such an important role in the screening process, managers face the incentive to give a good signal to the investors and analysts that also play an important role in the screening process of firms. A good signal can be given by exceeding thresholds like last years’ earnings, analysts’ forecasts and the zero-earnings line. According to Burgstahler and Dichev (1997), Degeorge, Patel and Zeckhauser (1999) and Hayn (1995) firms do indeed report relatively often earnings that slightly exceed those thresholds. This is the result of managers managing earnings in a way

(11)

11 that those thresholds are exceeded.

Furthermore, contractual outcomes are often dependent on accounting numbers, for instance the reported earnings. For example, a contractual outcome that is often influenced by accounting numbers is the contract between the firm and employees. Examples are the stock-based and option-stock-based compensation plans for executives of a firm. Many of the extant literature focuses on the relationship between performance-based compensation contracts and earnings management. The last decades an enormous increase in stock-based and option-based compensation can be noticed. The underlying reason is that this method of

compensation reduces agency costs and aligns the incentives of the managers and the shareholders of a firm (Bergstresser & Philippon, 2006; Cheng & Warfield, 2005).

Bergstresser and Phillipon (2006) found that those options are relatively often exercised by executives in years that high accruals are reported, which consequently means that the stock price increases on the short term. These findings imply that earnings

management practices are used to fulfil the incentive that arises from the stock-based and option-based compensation contracts to gain a personal benefit.

Comparable results are found by Cheng and Warfield (2005). They find that managers with high equity incentives are more likely to report earnings that exceed thresholds, for instance last years’ earnings and analyst forecasts. Besides, managers that face equity incentives are less likely to report big positive earnings surprises, since those managers are more sensitive to future stock-price fluctuations, which lead to earnings smoothing, an increased reserving of current earnings, to avoid earnings disappointments in the future, which would have a negative effect on their own compensation.

The results in the above-mentioned studies suggest that managers of firms are

incentivized to apply practices of earnings by the process investors and analysts screen firms to understand the riskiness of investing in that firm. Furthermore, the managers apply

earnings management to influence contractual outcomes, for instance the outcome of stock-based and option-stock-based compensation plans to gain a personal benefit.

2.3 The economic consequences of earnings management

With the evidence that incentives for earnings management exist and are fulfilled within firms, the next logical question to answer is whether the managers and shareholders of firms face indeed positive consequences by managing earnings. In the extant literature a shift of economic consequences of earnings management can be found over time.

(12)

12 Barth, Elliot and Finn (1999) have examined the association between market rewards and increasing earnings patterns. They have shown that firms that have patterns of increasing earnings have bigger price-earnings multiples. Price-earnings multiples are higher if the market gives the firm a high value. This means that firms with patterns of increasing earnings have higher market values than firms that have patterns of non-increasing earnings. The longer the series of increasing earnings lasts, the higher the positive effect on the market value becomes. These findings indicate that earnings management practices to exceed the threshold of last year’s earnings have an increasing effect on the value of the firm.

In the beginning of this century the attitude towards earnings management and financial reporting in general has changed as the consequence of the numerous accounting scandals and the Sarbanes-Oxley act that was passed in response to these scandals (Grasso, Tilley, & White, 2009). The main implications of the Sarbanes-Oxley regulations is that the requirements for proper financial reporting became much tighter and that violations of these requirements can lead to penalties up to twenty years imprisonment.

These changes show that society’s attitude towards accruals-based earnings

management methods became more skeptical in comparison to real earnings management. This statement is supported by the research of Keung, Lin and Shih (2010). They found that investors and analysts tend to associate earnings that are slightly above last year’s earnings or the zero-earnings benchmark with manipulation. The result is that investors and analysts tend to give those firms that are suspect to earnings management a lower value.

To mitigate society’s concerns regarding accruals-based earnings management, managers tend to use more real earnings management practices, since those practices are less visible for the users of financial statements than practices of accruals-based earnings

management. This is the consequence of the fact that accruals-based earnings management is a discretionary change of the reported numbers, while real earnings management means that that underlying business decisions are changed, but the way of reporting is not affected by managerial discretion. This means that real earnings management is less easy to detect by the users of financial statements (Cohen & Zarowin, 2008).

However, the disadvantage of real earnings management is that managers have to deviate from optimal business decisions (Zang, 2012). Most researches state that real earnings management practices has the consequence that operating performance is lower in years after the real earnings management had taken place (Tabassum, Kaleem, & Nazir, 2015). These findings imply that the choice to make use of real earnings management practices are not in line with the interest of shareholders. Despite the evidence that deviating

(13)

13 from optimal business decision is not in the best interest of shareholders, managers tend to use more real earnings management practices, since the risks of accruals-based earnings management have increased after the passage of the Sarbanes-Oxley act (Cohen, Dey, & Lys, 2008).

2.4 Tax aggressiveness, tax avoidance and tax evasion

Another tool that can be used to manage earnings is the tax planning strategy of a firm (Cook, Huston, & Omer, 2008). Taxation is an important tool to manage earnings, since it is a large line item in the financial statements of the firm. Taxes are used by the government to finance the governmental investments. Despite that firms can benefit from the infrastructure, military force, law enforcement and other governmental investments, an incentive exists to free-ride by preventing or delaying the shift of resources to the government in the form of tax

payments (Christensen & Murphy, 2004). This can be done by being aggressive in the application of a tax planning strategy. Being tax aggressive means that a firm uses practices to avoid or even evade taxes.

Both, tax avoidance and tax evasion, have the aim to reduce tax payments, which has the economic consequence that resources are not transferred from the firms to the

government. This means that economic consequences are not the decisive criterion to make a distinction between tax avoidance and tax evasion (Kirchler, Maciejovsky, & Schneider, 2003). A better distinction between tax evasion and tax avoidance can be made by taking the legal consequences into account.

Tax avoidance is reducing the tax payments by legal means, for example exploiting loopholes in the tax laws or making extensive use of the facilities that are included in the tax code. On the other hand, tax evasion is the reduction of tax payments by illegal means. This means that tax laws are being broken, when a practice of tax evasion is applied (Kirchler, Maciejovsky, & Schneider, 2003).

2.5 Economic consequences of tax aggressiveness

Since the arising scepticism towards financial reporting within society and the passing of the Sarbanes-Oxley-act in the beginning of this century a shift is ongoing from accrual-based earnings management towards real earnings management (Keung, Lin, & Shih, 2010). An alternative for these practices of earnings management is earnings management by taxation. However, the economic consequences of taxation as a tool to manage earnings management are not the same as the economic consequences of accrual-based earnings management or

(14)

14 real earnings management.

A successful practice of earnings management by tax aggressiveness prevents a transfer of resources from the firm to the government (Christensen & Murphy, 2004). This is a consequence that is in line with the interests of both the shareholders and the manager. This differs from practices of real earnings management that often goes together with suboptimal business decisions. Those suboptimal business decision are not in line with the interests of the shareholders (Zang, 2012). Those suboptimal business decisions or actual savings of money are both not related to accruals-based earnings management. This means that from a shareholders’ perspective earnings management by using an aggressive tax planning strategy can be the most advantageous before accruals-based earnings management, while real earnings management is the least beneficial.

As can be derived from the articles of Keung et al. (2010) and Grasso et al. (2009) practices of accruals-based earnings management are often considered as unethical in the period after the accounting scandals in the beginning of this century. Acting in opposition to society’s ethical norms and values means that practices of earnings management of a firm are considered to be unethical by a huge amount of people. This social scepticism also holds for firms that apply an aggressive tax planning strategy. The more aggressive the practices of accrual-based earnings management or tax planning are, the more people think that a firm is acting unethically, which leads to higher reputational costs (Fisher, 2014; Gallemore,

Maydew, & Thornock, 2014; Hanlon & Slemrod, 2009). The reputational costs can consist of potential customers that are less likely to become a client, since they do not want to cooperate with firms that are acting unethically in their perception. Besides, firms can face difficulties on the capital market to collect money, since many potential investors do not want to invest in firms that are involved in a tax shelter or an accounting scandal. Another component of reputational costs is that firms that are involved in tax shelters or accounting scandals will be investigated by the enforcement agencies with much more suspicion than before the tax shelter or accounting scandal.

Furthermore, earnings management by applying an aggressive tax planning strategy can result in a fine by the Internal Revenue Service (Baldry, 1986). The possibility to get fined by the tax authority as a punishment for illegal tax practices is a risk that firms are facing if an aggressive tax planning strategy is applied.

A clarification of this risk can be given by the use of the crime and punishment model of Becker (1968). This model states that from an economic point of view crimes are only committed if the benefits of the crime exceed the probability that a crime is detected times the

(15)

15 height of the fine that one can potentially receive as the consequence of the crime.

Allingham and Sandmo (1972) have applied the above-mentioned model on the decision-making setting around the tax planning strategy. The risky factor in this setting is the probability that faults in the tax returns are detected by the tax authority, the Internal Revenue Service. This means that this factor is determined by the tightness of the inspections by the Internal Revenue Service.

Comparable risks are related to accruals-based earnings management, since

reputational are a possible consequence of accruals-based earnings management, especially after the scepticism in society has increased as a result of the numerous accounting scandals. Besides, the Sarbanes-Oxley act has increased the risks for fines or even imprisonments as a result of incorrect financial reporting.

However, the application of real earnings management will not result in a penalty, since real earnings management does not result in improper financial reporting. The business operations are well reported in the financial statements, but it is high likely that the business decision are not optimal. This is not a valid ground to receive a penalty from any enforcement institution.

This means that on the one hand tax planning as an earnings management tool contains some benefits over real earnings management and accruals-based management. Firstly, successful practices of tax planning are beneficial for the manager and also for the shareholders, which is not the case with accruals-based earnings management, which has neutral effects for the shareholders, and real earnings management, which results in suboptimal business decisions and are not in the best interest of the shareholders. On the other hand, tax aggressiveness contains some risks that are not inherent to real earnings management, since breaking the tax code can result in penalties and reputational costs. These risks are not faced if practices of real earnings management are applied. The risks related to earnings management by taxes and accruals-based earnings management are comparable. Overall, the benefits of earnings management by being tax aggressive outweigh the benefits of accruals-based earnings management and real earnings management, but it can also be said that tax planning strategies are a risky tool to manage earnings.

2.6 Factors that influence earnings management choices by managers

The decision of managers to get engaged in earnings management practices is highly influenced by several factors. Firstly, it is important to understand that a decision to get engaged in earnings management practices is often incentivized to gain a personal benefit,

(16)

16 but also have influence on the wealth of the shareholders. This means that a manager is also deciding for others if he or she is making a decision about earnings management practices.

Taking into account personal and shareholders’ interests, earnings management by aggressive tax planning strategies is preferred above real earnings management, since successful practices of tax evasion or tax avoidance have positive consequences for both shareholders and managers. On the other hand, real earnings management can be used to gain a personal benefit for the manager, but is not in the best interest of the shareholder, since real earnings management is related to suboptimal business decisions.

However, another very important factor that has to be taken into account, while making a decision, is the manager’s willingness to tolerate risks. The cardinal utility theory states that individuals give a diminishing value to money, which means that an increase in money goes together with a degressive increase in wealth (Larrick, 1993). This theory is applied by Kahnemann and Tversky (1979) in the prospect theory about risk-taking. They state that the wealth-value functions of individuals are concave in gains and convex in losses, as shown in figure 1.

Figure 1: A hypothetical value function (Kahneman & Tversky, Prospect Theory: an Analysis of Decision under Risk, 1979).

The value function is at its steepest around a specific reference point, which is the tipping point between gains and losses. Another characteristic of such a value function is that the function is steeper before the reference point than after the reference point. This kind of function indicates that managers mostly are averse to losses. Managers are less likely to take risks if the negative result can be losses and more likely to take risks if the worst possible outcome is a decrease of profit (Kahneman & Tversky, 1992).

(17)

17

3. Hypothesis Development

Since the accounting scandals and the passing of the Sarbanes Oxley act in the beginning of this century real earnings management is serving more and more as a substitute for accrual-based earnings management. In addition, a firm can also use aggressive tax planning strategies as a tool to manage earnings.

Real earnings management on the one hand and tax aggressiveness on the other hand have different costs and benefits. Earnings management is mostly performed by managers to gain a personal benefit. The advantage of earnings management by tax aggressiveness in comparison to real earnings management is that a successful practice of tax planning is beneficial for both the shareholders and the manager, since it actually saves money. This is not the case with practices of real earnings management, since real earnings management often accompanies suboptimal business decisions regarding timing and structuring of the business operations (Zang, 2012). This means that a successful practice of earnings

management by tax planning is preferred by shareholders and managers above a practice of real earnings management, since it has benefits for both and no suboptimal business decisions are made.

However, the other side of the coin is that tax planning as an earnings management tool contains some risks that are not faced in the case that practices of real earnings

management are applied. If a firm is too aggressive in their tax planning strategy and break the tax laws, the firm is exposed to the risk that they receive a penalty from the Internal Revenue Service and have to repay the taxes that were evaded. This risk and potential costs are not present if a firm applies practices of real earnings management.

An application of the cardinal utility theory and prospect theory states that the manager’s aversion to risk is higher if a firm is not performing well, since a decrease in earnings is experienced as more undesirable in the case that a decrease means that the loss is getting bigger than in the case that the profit is getting smaller.

In sum, it is not unthinkable that a manager changes his way of managing earnings if recent firm performance is poor. It can be expected that if the recent performance of a firm is poor the manager will decide to fulfil the incentives for earnings management by applying less earnings management by tax planning, since the manager is less willing to face the risks that are inherent to tax planning. This also means that the manager should be more willing to take these risks and gain the optimal benefits from tax aggressiveness if the firm is doing well. This leads to the question whether recent firm’s performance is associated with a lesser

(18)

18 extent of tax aggressiveness in the subsequent years. The following hypothesis corresponds to this question:

H1: Recent firm performance is positively associated with earnings management by tax planning practices.

(19)

19

4. Methodology

4.1 Sample

This study is performed by using panel data in an empirical archival research. Since the data required for this research consists of financial data, the Compustat-database is sufficient to gather the necessary data. After the gathering of data Stata is used to perform the regression and statistical tests.

The sample consists of firms headquartered in the United States over the years 2003 until 2016. This time period is chosen, since Keung, Lin and Shih (2010) have shown that the attitude of investors and analysts towards earnings management has changed in the period after the passing of the Sarbanes-Oxley legislation in 2002 in comparison to the period before the passing of the Sarbanes-Oxley legislation. As a consequence of the numerous accounting scandals, investors and analysts are more sceptical towards practices of earnings

management. The introduction of the Sarbanes-Oxley-act in 2002 was a consequence of those accounting scandals. In order to use as much valuable data as available after the introduction of the Sarbanes-Oxley act, 2016 is chosen as the last year of the time period, since this is the last year that is completely covered in the database.

As shown in table 1, the initial sample consists of 180,085 observations from 19,581 firms. After generating the necessary variables and dropping observations with missing variables, 29,860 firm-year observations of 6,136 unique US firms remain.

TABLE 1

Sample Selection OBS FIRMS

All US firm-years with Compustat data coverage for the period

2003-2016 180,085 19,581

Less: firm-years with lack of necessary data to compute the dependent variable or main variable of interest

-48,192 -3,259

131,893 16,322

Less: firm-years with lack of important control variables -102,033 -10,186

29,860 6,136

(20)

20 4.2 Empirical design

In order to test whether earnings management by aggressive tax planning strategies is associated with recent firm performance, I used a model based upon the models of Stickney and McGee (1982) and Gupta and Newberry (1997). The aim of those models is to examine which factors influence the effective tax rate of a firm. I extended those models by adding my main variable of interest, a measure of firm performance, and important control variables for the level of accruals-based earnings management and real earnings management. This resulted in the following model.

𝐸𝑇𝑅_𝑖,𝑡 = 𝛽₀+ 𝛽₁∗ 𝑃𝐸𝑅𝐹_{𝑖,𝑡−1}+ 𝛽₂∗ 𝑆𝐼𝑍𝐸 + 𝛽₃∗ 𝐿𝐸𝑉_𝑖,𝑡+ 𝛽₄∗ 𝐶𝐴𝑃𝐼𝑁𝑇_𝑖,𝑡+ 𝛽₅∗ 𝐴𝐶𝐶_𝑖,𝑡+

𝛽₆𝑅𝐴𝑀_𝑖,𝑡+ 𝛽₇𝐶𝐹𝐿_𝑖,𝑡 (1)

In this model the effective tax rate, ETR, serves as the dependent variable. The recent firm performance, PERF is the main variable of interest. The other variables serve as control variables that are selected based on prior literature. In addition to this model, time-dummies and industry-dummies are added to control for fixed effects that are the consequence of time-specific and industry-time-specific characteristics.

In the following of this section a description of the variables included in the model is given. In addition, the appendices contain three tables, appendices A,B and C, that show the way how the variables are computed and what the meaning of those variables are.

4.2.1 Variables used 4.2.1.1 Dependent variable

A key concept for getting a better idea about tax aggressiveness is the effective tax rate of a firm. To explain this concept it is necessary to describe the concept of dual accounting.

Dual accounting means that in a firm a distinction is made between tax reporting and financial reporting, which is often called book reporting (Plesko, 1999). These firms have two administrations. First of all, the firm has the regular financial administration. This

administration forms the backbone of the financial statements. Financial reporting practices aim for giving a relevant and reliable insight in the practices of the firm. Financial reporting has to deal with the valid financial reporting standards that are applied to a firm. In addition

(21)

21 there is tax reporting. This type of administration serves to make a yearly fiscal profit

calculation. Tax laws and regulations serve to give firms guidance for this profit calculation. It is an important notion that financial managers can face two different incentives within the two types of administration. As seen before, financial managers tend to present the results in the financial statements as positive as possible. This can be incentivized by, for instance, requirements in compensation contracts, debt covenants or government regulations (Healy & Wahlen, 1999; Fields, Lys, & Vincent, 2005)

The incentive within the tax bookkeeping is to keep the profit as low as possible. The consequence of a low profit is that the tax base is low, what results in a low amount of taxes that have to be paid (Tang & Firth, 2011).

Dual accounting can be more clarified by the definition of the effective tax rate. The effective tax rate can be defined as the ratio of paid taxes to income (Gupta & Newberry, 1997). The GAAP effective tax rate is defined as follows:

𝐺𝐴𝐴𝑃 𝐸𝑓𝑓𝑒𝑐𝑡𝑖𝑣𝑒 𝑇𝑎𝑥 𝑅𝑎𝑡𝑒 = 𝑇𝑜𝑡𝑎𝑙 𝑖𝑛𝑐𝑜𝑚𝑒 𝑡𝑎𝑥𝑒𝑠

𝑃𝑟𝑒𝑡𝑎𝑥 𝑝𝑟𝑜𝑓𝑖𝑡 𝑎𝑐𝑐𝑜𝑟𝑑𝑖𝑛𝑔 𝑡𝑜 𝐺𝐴𝐴𝑃× 100%

(2)

If firms do not apply an aggressive tax planning strategy, the reported income in the tax and financial books is approximately equal to each other, resulting in an effective tax rate equal to the statutory rate. However, if a manager succeeds in fulfilling the incentives to present high profits in the financial statement and a low profit in the fiscal profit calculation, the result is a low effective rate. Such a low effective tax rate is a sign of an aggressive tax planning

strategy. In other words, a low effective tax rate is an indication of tax avoidance and tax evasion (Gupta & Newberry, 1997).

Since the ETR is a ratio between two variables, it is necessary to make some

adjustments to the ETR to make the ETR more useful (Gupta & Newberry, 1997). First of all, if a firm gets a tax refund, ETR is set at 0, since the effective tax rate cannot be negative. Further, the ETR is limited to a maximum value of 100 percent to decrease the influence of observations where the net income is really small and the amount of taxes paid is higher.

(22)

22 4.2.1.2 Independent variable: Main variable of interest

To test whether recent firm performance is associated with earnings management by tax aggressiveness in the subsequent year, it is important to define a firm as poor-performing or well-performing in the previous year. The measure that is used is the adjusted industry return on assets. This number is calculated as follows:

𝑅𝑂𝐴_𝑎𝑑𝑗 = 𝑠𝑖𝑔𝑛∆𝑅𝑂𝐴 ∗ √(|∆𝑅𝑂𝐴|)

(3) In this formula, ∆ROA is calculated as the difference between the ROA of the firm and the ROA of the specific industry. The ROA of a specific industry is calculated as the median of the ROA in all the industries, based on the two-digit SIC-levels. The use of the industry adjusted ROA to determine how a firm is performing is preferred above the normal ROA, since the industry-adjusted ROA controls for industry-specific and general economic factors (Eisenberg, Sundgren, & Wells, 1998). This means that the industry-adjusted ROA gives is a better measure to compare firm performances of firms that are active in different industry than the regular ROA.

Subsequently, I used the dummy variable, PERF, to classify the observations as good-performing or bad-good-performing. The dummy is coded 1 if the industry adjusted ROA for a specific observation exceeds the median of the industry adjusted ROA’s. Otherwise the firm

is classified as bad-performing with the consequence that the dummy is coded 0. This dummy variable, PERF, serves as the main variable of interest that distinguishes good recent firm performance and poor recent firm performance.

4.2.1.3 Independent variable: Control variables

In order to determine whether an association exists between recent firm performance and the extent of earnings management by tax aggressiveness I have included some control variables in my empirical model to isolate the effect of recent firm performance on the extent of tax aggressiveness.

First of all, I have included firm size as a control variable in my empirical model. Prior literature is not unambiguous about the effect of firm size on the effective tax rate. Some literature states that there is positive association between firm size and the effective tax rate (Zimmerman, 1983). An argument for this statement is that big firms are more

(23)

23 relationship is found between the size of a firm and the effective tax rate (Graham, Hanlon, Shevlin, & Shroff, 2014). Other studies found that a negative relationship exists between firm size and the effective tax rate. The underlying reasoning for this argument is that big firms have more resources available to spend on tax planning with the consequence that they are better able to lower their effective tax rate (Gupta & Newberry, 1997). Furthermore, extant literature show that firm size can also be linked to the main variable of interest, firm

performance. Hall and Weiss (1967) state that big firms tend to have higher profit rates as a consequence of the advantages of the economies of scale.

Secondly, I have included the proportion of long-term debt within the firm as a control variable. This variable is included, since there is evidence that the capital structure of a firm has an effect on the effective tax rate (Stickney & McGee, 1982; Mills, Erickson, & Maydew, 1998). The underlying reasoning is that many national tax codes contain provisions that allow that the costs for debt, interest, are deductible, while the costs for equity,

dividends, are not. This means that there are more legal opportunities to lower the effective tax rate for highly-leveraged firms. Moreover, prior literature shows a negative relationship between profitability and a firm’s proportion of debt (Wald, 1999). This relationship can be explained with the pecking order theory that states that firms prefer internal financing above debt-financing. Since highly profitable firms have more options to use internal financing, they are less likely to need external financing in the form of debt (Myers & Majluf, 1984).

The third control variable that is added to the model is the proportion of the value of the assets that consist of the value of the property, plant and equipment. This is added as a control variable, since an inverse relationship exists between the extent of capital intensity within a firm and the effective tax rate (Gupta & Newberry, 1997; Stickney & McGee, 1982; Mills, Erickson, & Maydew, 1998). The underlying reasoning is tax codes contain many provisions for firms that are capital intensive. An example of such a provision is the accelerated depreciation facility. In addition, the capital structure of a firm can also affect recent firm performance. Firms that are active in a market that requires a lot of capital tend to be more profitable, since those markets are less competitive. This lower level of competition is caused by the high entry barriers in the form of high capital requirements for new potential competitors (Hall & Weiss, 1967).

The fourth and fifth control variables serve to control for the extent of accrual-based earnings management respectively real earnings management within a firm. The variable for the level of accruals-based management and real earnings management, ACC and RAM, are determined by the Jones-model (1991) respectively the model of Rowchowdhury (2006).

(24)

24 Both models consist of a regression in which the error term is considered to be the extent of earnings management within firms.

The Jones model can be described as follows:

𝑇𝐴_𝑖,𝑡 𝐴𝑖,𝑡−1 = 𝛼0+ 𝛼1∗ ( 1 𝐴𝑖,𝑡−1 ) + 𝛼2∗ ( ∆𝑠𝑎𝑙𝑒𝑠 − ∆𝑟𝑒𝑐𝑒𝑖𝑣𝑎𝑏𝑙𝑒𝑠 𝐴𝑖,𝑡−1 ) + 𝛼3 ∗ ( 𝐺𝑃𝑃𝐸_{𝑖,𝑡−1} 𝐴𝑖,𝑡−1 ) + 𝜖𝑖,𝑡 (4) The Rowchowdhury-model is described as follows:

𝐶𝐹𝑂𝑖,𝑡 𝐴_{𝑖,𝑡−1}

= 𝛼

0

+ 𝛼

1

∗ (

1 𝐴_{𝑖,𝑡−1}

) + 𝛼

2

∗ (

𝑆𝑎𝑙𝑒𝑠𝑖,𝑡 𝐴_{𝑖,𝑡−1}

) + 𝛼

3

∗ (

∆𝑠𝑎𝑙𝑒𝑠 𝐴_{𝑖,𝑡−1}

) + 𝜀

𝑡 (5)

These variables are necessary to test whether firms will be using tax aggressiveness as a tool to manage earnings to a greater or lesser extent in comparison to the total level of earnings management that is measured within a firm. Furthermore, the relationship between recent firm performance and the effective tax rate can be influenced by the total level of earnings management, since it is not unlikely that good-performing firms are more likely to apply more practices of earnings management in general. These variables are computed with the help of the earlier described Jones-model and Rowchowdhury-model.

The sixth control variable is the amount of carry forward rights that a firm has in comparison to the total value of its assets. Those rights are the consequence of recent poor firm performance and firms can use those rights to lower its tax base in the current year. The consequence is that the taxes payable will be also lower, resulting in a lower effective tax rate. In previous literature on tax aggressiveness firms with carry forward rights are mostly left out of the sample (Gupta & Newberry, 1997; Wilkie & Limberg, 1993). Since those firms may also have incentives to be tax aggressive, valuable data will be lost if these firms are left out of the sample. To control for the mechanical effect that carry forward rights may have on the effective tax rate this variable is added to the model.

(25)

25 Lastly, I control for the fixed effects that are a consequence of the industry a firm is active in and the year of the observation. This is done by adding a dummy for each year that is in the sample, respectively 2003 until 2016. The industry a firm is active in is determined with the help of the Standard Industrial Classification Code. I used the first two numbers of this code to classify an observation into a specific industry group. A dummy variable is added for each of the industry subgroups. This resulted in 58 different industry-specific dummies that represent 58 different industry groups.

4.3 Hypothesis testing

To determine whether the association exist between recent firm performance and the extent of earnings management by tax aggressiveness, I run an ordinary least squares regression. To check whether the regression is valid I check whether the residuals of my regression are normally distributed, linear and independent. To prevent that the results are influenced by heteroscedasticity errors the ordinary least squares regression is run in a robust form. Further, the regression is run as a clustered regression to control for intra-firm correlations.

To check whether multicollinearity exists I use a correlation matrix after winsorizing the variables at the first and 99th percentile to exclude outliers from the sample. Moreover, I check the VIF-values to check whether multicollinearity is present.

The value of the β1-coefficient in model 1 results shows whether an association exists between recent firm performance and the effective tax rate. If this β1-coefficient is negative, a negative relationship between recent firm performance and the effective tax rate is found, which implies that a positive relationship exists recent firm performance and the extent of earnings management by applying aggressive tax planning strategies. This would be in line with the hypothesis that poor-performing firms are more risk-averse and less likely to use tax aggressiveness as a tool to manage earnings, while well-performing firm are more likely to tolerate the risks that are related to practices of earnings by applying aggressive tax planning strategies.

(26)

26

5. Results

5.1 Descriptive statistics

Table 2A and 2B show the descriptive statistics on the variables that are used in this research. Table 2A shows the descriptive variables on the variables before transformations are applied to solve the problem of skewed date, while table 2B shows the statistics of the variables in the format as they are used in the regression.

The mean of the effective tax rate for US-firms over the time period 2003-2016 is 29.70 percent, which is below the statutory tariff of 35 percent. This statistic suggests that tax planning practices have been applied within the US-firms during the period beginning in 2003 and ending in 2016. This finding is in line with the findings in other studies that have investigated the effective tax rate. All these studies have found effective tax rates of approximately 30 percent (Dyreng, Hanlon, & Maydew, 2008; Gupta & Newberry, 1997). The sample firms have an average asset value of $8,042 million (after log-transformation: 5.95). On average, these assets are financed with 27.5 (-2.61) percent long-term debt, as can be derived from the variable LEV. On average 54.1 (-2.23) percent of the assets consist of plant, property and equipment. The extent of accruals based earnings management and real earnings management are on average 0.03 respectively -0.01, which represent the level of discretionary accruals and the abnormal cash flows from operations. The last variable that is used is the amount of carryforward rights that has a mean of 17.355 (-2.17). To safeguard the normality of the variables the variables are winsorized and a logarithm-transformation is applied on the variables SIZE, LEV, CAPINT and CFL.

Table 2A: Descriptive statistics of variables before transformations

Variable Obs Mean Std. Dev. Min Max

ETR 29860 29.70881 31.93765 0 100 PERF 29860 .5 .5000084 0 1 SIZE 29860 8.041.798 73412.15 .001 3270108 LEV 29860 .2746902 5.565938 0 8.217.272 CAPINT 29860 .5413681 1.410155 0 113.8 ACC 29860 .0325401 .1500266 -.309738 .3582939 RAM 29860 .0290244 .1584302 -.3510058 .3389435 CFL 29860 17.35526 471.5424 0 41431

(27)

27 Table 2B: Descriptive statistics of variables after transformations

Variable Obs Mean Median Std. Dev. Min Max

ETR 29860 29.70881 5.246413 31.93765 0 100 PERF 29860 .5 0.5 .5000084 0 1 SIZE 29860 5.714759 5.950159 2.771851 -2.312635 11.62975 LEV 29860 -2.613424 -2.228667 1.609133 -4.60517 .352014 CAPINT 29860 -1.216829 -0.9768769 1.250486 -4.60517 .8507923 AAC 29860 .0325401 0.0127663 .1500266 -.309738 .3582939 RAM 29860 .0290244 -0.0103149 .1584302 -.3510058 .3389435 CFL 29860 -2.16873 2.347986 -2.566509 -4.60517 4.535606 The sample has 29,860 unique firm-year observations. The ETR is the effective tax rate that is determined by dividing the total income taxes to the pretax income. PERF is a dummy variable that has value 1 if the observation has an industry adjusted return on assets in the previous year that is above the median. The value for the dummy variable is 0 if the industry adjusted return on assets is below the median. In order to deal with skewness of the above data, the following transformations are applied. SIZE is defined as the logarithm of the lagged assets. LEV is calculated as the logarithm of the ratio between the lagged long-term debt and the lagged assets. CAPINT is the logarithm of the ratio between the lagged value of the gross value of plant, property and equity. AAC is the error term of the regression on the Jones model. RAM is the error term of the Rowchowdhury model. CFL is defined as the logarithm of the ratio between the lagged value of the carry forward rights and the lagged assets. SIZE, LEV, CAPINT and CFL are winsorized for the bottom and top 1 percent. ACC and RAM are winsorized for the bottom and top 5 percent.

Table 3 provides the descriptive statistics for two different subsamples. The left side of the table shows the descriptive statistics of the observations where the recent firm performance is classified as below median and the right side of the table shows the descriptive statistics of the observations for what the recent firm performance is classified as above median. Both groups consist of 14,930 observations.

It can be seen that the average ETR is significantly lower (p=0.0000) for good-performing firms in comparison to bad-good-performing firms. The difference is almost one percent, taking into account that bad-performing firms have an ETR of 30.2 percent and good-performing firms of 29.2 percent. Table 3 also shows that good-performing firms tend

(28)

28 to be smaller in terms of asset value than bad-performing firms (p=0.0000). In addition, the average of level of leverage and capital intensity is significantly lower (p=0.0000 and p=0.0000) for the subsample of good-performing firms than for the subsample of bad-performing firms. Furthermore, the levels of accruals-based earnings management and real earnings management are significantly different in the two subsamples. Those levels are both significantly higher in the subsample of good-performing firms, since the p-value is in both cases 0.0000. The last variable that is significantly different between the two different subsamples is the amount of carryforward rights that is owned by a firm. Table 3 shows that the ratio between the value of the carryforward rights and asset value is more than 12 percent (1.90) is significantly higher (p=0.0032) for firms in the subsample of bad-performing firms. The logical explanation for this statistic is that rights for carryforward arises if a firm makes losses. This means that firms that have performed poorly in the recent history are more likely to have rights for carryforwards on their balance sheet.

(29)

29

TABLE 3: Descriptive statistics split up for PERF=0 and PERF=1

Subsample of bad-performing firms Subsample of good-performing firms

PERF=0 (N=14930) PERF=1 (N=14930) T-tests

Variable Obs Mean Std. Dev. Min Max _{Variable Obs} Mean Std. Dev. Min Max T-test p-value

ETR 14930 30.20556 38.71623 0 _{100 ETR 14930} 29.21206 23.25199 0 ₁₀₀ -5,2208 0,0000 PERF 14930 0 0 0 _{0 PERF} 14930 1 0 1 ₁ 77,0412 0,0000 SIZE 14930 4.984342 2.988595 -2.312635 _{11.62978 SIZE} 14930 6.445176 2.316902 -2.312635 _11.62978 -13,7139 0,0000 LEV 14930 -2.58879 1.68584 _{-4.60517 0.3520111 LEV} 14930 -2.638057 .1.528241 _{-4.60517 0.3520111} -4.000 0,0000 CAPINT 14930 -1.27754 1.347848 _{-4.60517 0.8507923 CAPINT 14930 .-1.156118 1.141695 -4.60517 0.8507923} -5,7869 0,0000 ACC 14930 .014211 .1674787 -.309738 _{.3582939 ACC} 14930 .0508693 .1276574 -.309738 _.3582939 35,0878 0,0000 RAM 14930 -.0229942 .1652595 -.3510058 _{.3389435 RAM} 14930 .0810431 .1322083 -.3510058 _.3389435 96,1524 0,0000 CFL 14930 -1.220719 2.495454 -4.60517 4.535606 CFL 14930 -3.116749 1.732505 -4.60517 4.535606 -2,9528 0,0032 The sample has 29860 unique firm-year observations. PERF serves to make a distinction between good-performing and bad-performing firms. The value is 0 if the firm has an industry-adjusted ROA under the median for that specific year. The value of the dummy is 1 is the firm has an industry adjusted roa that is above the median. The following transformations are applied on the above-mentioned variables to deal with the skewness of the variables. The ETR is the effective tax rate that is determined by dividing the total income taxes to the pretax income. SIZE is defined is the logarithm of the lagged assets. LEV is calculated as the logarithm of the ratio between the lagged long-term debt and the lagged assets. CAPINT is the logarithm of the ratio between the lagged value of the gross value of plant, property and equity. ACC is the error term of the regression on the Jones model. RAM is the error term of the

Rowchowdhury model. CFL is defined as the logarithm of the ratio between the lagged value of the carry forward rights and the lagged assets. SIZE, LEV, CAPINT and CFL are winsorized for the bottom and top 1 percent. ACC and RAM are winsorized for the bottom and top 5 percent. The tests report the t-statistic and p-value (two-tailed) of a difference between the means of the variables for firms that are qualified as bad-performing respectively good-performing.

(30)

Table 4 provides a Pearson correlation matrix containing the dependent, main variable of interest and the control variables of the regression. Overall, the correlation coefficients among the dependent variables are not very high, which mitigates the concerns that the regression results are influenced by multicollinearity. The significantly negative coefficient for ETR and PERF may indicate that good-performing firms are more likely to apply tax aggressiveness as a practice of earnings management. This relationship is in line with my hypothesis that firms that recently performed well are more likely to apply tax planning practices as a tool to manage earnings, while poor-performing firms are less likely to do so.

Table 4: Pearson Correlation matrix

ETR PERF SIZE LEV CAPINT AAC RAM CFL

ETR 1.0000 PERF -0.0156* 1.0000 SIZE 0.1210* 0.2635* 1.0000 LEV 0.0270* -0.0153* 0.3845* 1.0000 CAPINT 0.0484* 0.0486* 0.0900* 0.2370* 1.0000 AAC -0.0737* 0.1222* 0.0055 0.0473* 0.0162* 1.0000 RAM 0.0084 0.3283 0.2056* -0.0035 0.0244* 0.1130* 1.0000 CFL -0.0815* -0.4038* -0.5727* -0.1436 -0.0590* -0.0315* -.02606* 1.0000 Please see table 2 and 3 for an explanations of the variables. The sample contain of 29860 unique firm-year observations.

*) p<0.01

5.2 Main results: Hypothesis test

In this section I test whether an association exists between the effective tax rate and recent firm performance within US firms in the period from 2003, after the introduction of the Sarbanes-Oxley Act until 2016. The results are shown in Table 5.

The regression results show that the hypothesis is confirmed, since the value of coefficient β1 for the main variable of interest, PERF, is -4.144. This high negative

coefficient is highly significant (p<0.01). The meaning of this coefficient is that the effective tax rate is lower in the case that firms performed well in the recent history, while the effective tax rate is higher if firms performed poor in the recent history. Since a low effective tax rate is considered to be a consequence of an aggressive tax planning strategy, the highly

significant negative coefficient is a confirmation of the hypothesis that recent firm performance and the extent of earnings management by tax aggressiveness are positively

(31)

31 associated. This may indeed indicate that the cardinal utility theory and the prospect theory of Kahnemann and Tversky (1979) hold in the setting of tax aggressiveness as a tool to manage earnings. Apparently, managers of firms are less willing to take the risks that are incorporated in earnings management by using taxation if a firm recently performed not so well.

Table 5 also shows that the ETR is positively associated with the firm size. This positive coefficient is highly significant and indicates that bigger firms have higher effective tax rates, which possibly means that less tax planning practices are applied in big firms (β=-1.452| p<0.01). This coefficient is in line with the claim of Zimmerman that bigger firms have higher effective tax rates (Zimmerman, 1983), while other literature often state that no significant relationship was found between the firm size and the effective tax rate (Graham, Hanlon, Shevlin, & Shroff, 2014). A potential explanation for this relationship is that big firms are heavily investigated by the Internal Size Revenue in comparison to smaller firms, which would mean that big firms face higher risks if they apply an aggressive tax planning strategy.

Furthermore, the negative coefficient for the variable LEV shows that

highly-leveraged firms make use of some facilities in the tax code that can only be used for the costs of debt, as is suggested by extant literature (Mills, Erickson, & Maydew, 1998; Stickney & McGee, 1982). The coefficient related to this relationship is significant at a 95 percent level (β=-0.401| p<0.05).

The coefficient that contradicts with the extant literature is the 1.223 (p<0.01) for the variable CAPINT (Stickney & McGee, 1982). This finding suggests that a firm is paying more taxes if the firm has a lot of plant, property and equipment on its balance sheet. A potential explanation for the contradiction with previous literature is that my sample is from a different time period. In the time period I have used firms were probably less likely to make use of facilities, like accelerated depreciation and investment credit, since the investment climate was unfavorable during the crisis years, which form an important part of the time frame of my sample.

Another interesting finding is that no significant relationship can be found between the level of real earnings management and the effective tax rate. The absence of a significant relationship means that real earnings management practices and practices of earnings

management by applying aggressive tax planning strategies do not act as substitutes or complements. This is different for accruals-based earnings management and earnings management by the means of taxation, since the results show that a significantly negative relationship is found between the level of accruals-based earnings management and the

(32)

32 effective tax rate (β=-15.862| p<0.01), which implies that a positive relationship exists

between the level of accruals-based earnings management and the level of earnings

management by taxation. An explanation for this relationship can be found in the fact that the same kind of risks are faced by applying accrual-based earnings management and earnings management by taxation. This suggests that if a firms’ manager is willing to take risks related to accruals-based earnings management, he is also not scared by the risks of earnings

management by applying aggressive tax planning strategies.

Another coefficient that is determined in the OLS regression represents the

relationship between the amount of carryforward rights a firm owns and the effective tax rate. As expected, a highly significant negative relationship exists between the value of the

carryforward rights and the ETR (β=-0.826| p<0.01).

Furthermore, table 5 shows that the adjusted r-squared is 0.053. This indicates that 5.3 percent of the variance of the effective tax rate is explained by the model.

TABLE 5: RESULTS OF OLS CONTROLLED FOR FIXED TIME AND INDUSTRY EFFECTS

DV:ETR OLS controlled for

fixed time and industry effects

Robust t-statistics in parentheses

*** p<0.01, ** p<0.05, * p<0.1 PERF -4.144*** (-8.169) SIZE 1.452*** (-10.951) LEV -0.401** (-2.081) CAPINT 1.223*** (-4.164) ACC -15.862*** (-10.530) RAM -1.972 (-1.131) CFL -0.826*** (-5.231) CONSTANT 9.844*** (-6.000) OBSERVATIONS 29860

(33)

33 5.3 Robustness tests

In order to test the construct validity of the empirical test in this study, the same regression is run with an alternative measure for recent firm performance. This test provides evidence whether the positive relationship between recent firm performance and earnings management by using aggressive tax planning strategies is biased by an incorrect operationalization of the theoretical concept, recent firm performance.

Model 1 uses a comparison between the performance in the industry and a firm’s performance to classify a firm-year observation as good-performing or bad-performing. In this regression to test the construct validity of the model the change of earnings is used as an alternative measure for recent firm performance. If the earnings of a firm are increased in the previous year, the recent firm performance is classified as good, while no recent earnings increase has the consequence that recent firm performance is qualified as poor. The main difference between the original and alternative measure is that model 1 uses an external benchmark and the alternative measure an internal benchmark to classify the recent firm performance (Eccles, 1991).

The same sample as for the main regression is used to test whether an association exists between recent firm performance measured by earnings growth and the effective tax rate. 13,020 of these observations are classified as bad-performing, while the other 16,840 observations are classified as good-performing. The results in table 6 show a significantly negative relationship between recent firm performance and the effective tax rate (β=-1.422| p<0.01), which is in line with the main results. These results provide evidence that another operationalization of the concept ‘recent firm performance’ gives comparable results. This means that multiple measures of recent firm performance provide support to confirm the hypothesis.

TABLE 6: RESULTS OF THE REGRESSION WITH ANOTHER MEASURE OF RECENT FIRM PERFORMANCE

DV:ETR OLS Robust t-statistics in parentheses *** p<0.01, ** p<0.05, * p<0.1 PERF -1.295*** (-3.485) SIZE _1.401*** (-10.632) LEV _-0.301 (-1.566) CAPINT _1.137***

(34)

34 (-3.871) AAC -15.363*** (-10.560) RAM -2.555 (-1.530) CFL _-0.462*** (-3.005) CONSTANT _8.554*** (-5.204) OBSERVATIONS 29,860 ADJ. R-SQUARED _0.050 5.4 Additional tests

In this section the results of some additional analyses are presented. These results serve as an addition and clarification to the results of the main hypothesis test.

5.4.1 The driver of the relationship between recent firm performance and the effective tax rate

The results in this study show that a significantly negative relationship exists between recent firm performance and the effective tax rate. It can be questioned if these results are driven by poor-performing firms that apply less practices of tax aggressiveness or by good-performing firms that apply more practices of tax aggressiveness.

To investigate this, I ran an additional regression with the dummy variable, PERF, reverse-coded and an extra dummy, CF, added to the model. The dummy is coded 0 if the firm does not own any rights to carryforward losses and valued 1 if the firm has those rights. In addition to this variable, I added a variable, CFxPERF, which is the multiplication of the dummy indicating recent firm performance and the dummy for the carryforward of losses. Since the presence of rights to carryforward losses is a consequence of poor recent firm performance, the new variable indicates whether the relationship in the main results is driven by bad-performing firms that apply less practices of earnings management by using

aggressive tax planning strategies.

The results of this regression in table 7 show that the coefficient belonging to PERF is not significant anymore after adding the dummy, CF, and the interaction term. Furthermore the variable, CF, and the interaction term have both a significant highly positive relationship

(35)

35 with the ETR (β=1.248| p<0.10 & β=-6.870| p<0.01). These results show that for both bad-performing and good-bad-performing firms, the firms that own carryforward rights have higher effective tax rates. This suggests the main relationship hold for both good-performing and bad-performing firms. However, the coefficient is bigger for the interaction term, which implies that the main relationship is stronger for poor-performing-firms.

TABLE 7: RESULTS OF REGRESSION TO TEST WHETHER THE RELATIONSHIP IS DRIVEN BY BAD-PERFORMING FIRMS

DV:ETR OLS Robust t-statistics in parentheses

*** p<0.01, ** p<0.05, * p<0.1 PERF _-0.986 (-1.202) CF _8.013*** (-7.573) CFXPERF _6.765*** (-6.870) SIZE _1.262*** (-8.966) LEV _-0.371* (-1.939) CAPINT _1.226*** (-4.199) AAC _-15.712*** (-10.476) RAM _-2.011 (-1.163) CFL _-1.420*** (-7.290) CONSTANT _4.719*** (-2.759) OBSERVATIONS _29,860 ADJ. R-SQUARED _0.056

5.4.2 Saving money by tax evasion and tax avoidance

The underlying explanation of the results found in this study is that managers apply

Recent firm performance and the extent of tax aggressiveness as a practice to manage earnings