Tax Avoidance and Firm Value

(1)

Tax Avoidance and Firm Value

Name: Allard Pothaar Student number: s1971018 Degree program: Msc Fiscale Economie Supervisor: Prof. dr. I.J.J. (Irene) Burgers

Abstract:

The annual revenue loss due to Base Erosion and Profit Shifting (BEPS) is estimated to amount to USD 100 to 240 billion. With the support of the G20 leaders, the international community has taken action to increase transparency and exchange of information in tax matters, and address weaknesses of the international tax system that create opportunities for BEPS. The BEPS Actions and the Anti-Tax Avoidance Directive of the European Commission put pressure on firms to reduce tax avoidance and strategies, which are used for, base erosion and profit shifting. In this thesis I will address the

influence of tax avoidance on firm value. Specifically, I question whether (i) tax avoidance increases firm value and whether (ii) this relation is modified by a firm‟s governance. From these questions I hypothesize that tax avoidance increases firm value and that this relation is amplified by the level of a firm‟s governance. The analysis is conducted on the basis of an unbalanced panel with 7098 firm-year observations on 1227 unique firms between 2003 and 2013. The results do not indicate that tax avoidance increases firm value and that this relation is affected by the level of a firm‟s governance.

(2)

2

List of Abbreviations

BEPS Base Erosion and Profit Shifting

CAPEX Capital Expenditures

CONTROLS Control Variables

LRETR Long Run Cash Effective Tax Rate

ETR Effective Tax Rate

GOVERN Governance Proxy

OECD Organization for Economic Co-operation and Development

OLS Ordinary Least Squares

q Tobin‟s Q

R&Dexp Research and Development Expenditures

TAXVAR Tax Avoidance Proxy

US United States

(4)

4

1. Introduction

This thesis explores the effect of tax avoidance on firm value. This is interesting because the international tax landscape has changed dramatically in the recent years (OECD, 2016 (background brief)). The annual revenue loss due to Base Erosion and Profit Shifting (BEPS) is estimated to amount to USD 100 to 240 billion. With the support of the G20 leaders, the international community has taken action to increase transparency and exchange of information in tax matters, and address weaknesses of the international tax system that create opportunities for BEPS. The G20 leaders endorsed a plan developed with the OECD members to restore confidence in the international tax system and to ensure that profits are taxed where economic activities take place and value is created. The plan is composed of 15 actions in total, the BEPS Action Plan. The BEPS Actions and the Anti-Tax Avoidance Directive of the European Commission put pressure on firms to reduce tax avoidance and strategies, which are used for, base erosion and profit shifting.

It is likely that many firms will be affected by the above mentioned initiatives to reduce tax avoidance. How firm value is affected by tax avoidance is not only interesting because of the governance

initiatives to fight tax avoidance but also because not all firms are engaging in tax avoidance

strategies. Dyreng, Hanlon and Maydew (2008) report that approximately one fourth of the US firms (of their sample of 2,077 firms) are paying taxes in excess of 35 percent of their pretax income over a ten year period. Noting that the US corporate tax rate is 35 percent, these firms appear to be engaging in little or no sustainable tax avoidance. Contrary to these firms, about a quarter of the firms in their sample are able to reduce their taxes to less than 20 percent of their pre-tax earnings. Dyreng et al. (2008) question why there is so much variation in the extent to which firms pay taxes, even among firms in the same industry. The relation between tax avoidance and firm value might provide new insights to this variation.

(5)

5

Given the outcome of Desai and Dharmapala (2008) I question as well in this thesis whether the relation between tax avoidance and firm value is modified by the level of a firm‟s governance. With the analysis in this thesis I will use a different measure for tax avoidance, a different proxy for a firm‟s governance and an additional sampling criterion. The Long-Run Cash Effective Tax Rate (LRETR) developed by Dyreng, Hanlon and Maydew (2008) will be used as a measure for tax avoidance and product market competition as in Masulis, Wang and Xie (2007) will be used as a proxy for a firm‟s governance. Additionally, Porcano (1986) reports a negative association between firm size and the effective tax rate (which as described later, relates to tax avoidance) so that one might argue that larger firms are affected more by the BEPS action plan to reduce tax avoidance than small firms. For this reason and to differentiate from the sample of Desai and Dharmapala (2008) I will focus in this thesis on large firms. To connect with the initiatives to fight tax avoidance, I require firms to be large enough to qualify for country-by-country reporting as in action 13 of the Action Plan Base Erosion and Profit Shifting. Country-by-country reporting might also be of interest because Rego (2003) mentions that large firms have more opportunities to avoid income taxes. For example, large firms may be able to avoid income taxes through intercompany transactions which will be tackled by country-by-country reporting.

Action 13 of the Action Plan Base Erosion and Profit Shifting requires the development of “rules regarding transfer pricing documentation to enhance transparency for tax administration, taking into consideration the compliance costs for business. The rules to be developed will include a requirement that MNEs provide all relevant governments with needed information on their global allocation of the income, economic activity and taxes paid among countries according to a common template” (OECD, 2015). Among other approaches, country-by-country reporting has been developed in response to this requirement. A country-by-country report requires large multinational enterprises to provide tax authorities with information relating to global allocation of income, the taxes paid, and indicators of the location of economic activity among tax jurisdictions in which the multinational enterprise operates.

(6)

6

between the BEPS action plan and firm value or tax avoidance. I address the effect of tax avoidance on firm value. This relation might provide new avenues to infer on the possible consequences of the BEPS action plan.

The remainder of this thesis will be structured as follows. Section 2 discusses relevant literature and develops the hypotheses where section 3 formulates the methodology. Section 4 presents data with descriptive statistics. Section 5 provides the empirical results and section 6 concludes.

2. Literature review and hypotheses development

This section starts with discussing the understanding of tax avoidance as tax avoidance is not

unambiguously defined (Blouin, 2014). Given the understanding of tax avoidance, I will use previous literature to answer the question how firm value is affected by the level of corporate tax avoidance which results in the first hypothesis. Next it will be addressed how this relation is influenced by the level of a firm‟s governance to formulate the second hypothesis.

2.1 Addressing tax avoidance

(7)

7

In this thesis I will use Hanlon and Heitzman‟s (2010) broadly defined definition of tax avoidance. With this definition I capture the entire spectrum of tax avoidance and hence all strategies which affect firm value. Note that this would not necessarily imply that firms are engaging in anything improper. Moreover, Dyreng, Hanlon and Maydew (2008) mention that many areas of the law are unclear, particularly for complex transactions. Firms may take positions on their returns in which the ultimate tax outcome is uncertain. Dyreng et al. (2008) have established an empirical measure which captures all tax planning activities. This measure, the long-run cash effective tax rate (LRETR), is based on the ability to pay a low amount of taxes per dollar of pretax earnings over long time periods.

Furthermore, in this section of the thesis I would like to mention that firms are increasingly working to a socially responsible outcome (Hoi, Wu, and Zhang, 2013). Tax avoidance is traditionally viewed as a value maximizing activity that transfers wealth from the state to corporate shareholders (Kim, Li and Zhang, 2011). The developments towards socially responsible outcomes are also extending to paying a „fair-share‟ of tax (Happé, 2011). This fair-share of tax could mean that taxpayers are honest and willing to prepare a correct tax filing, which is consistent with the original intent of the legislation, rather than finding aggressive tax strategies to reduce their taxes within the legal scope. Given these increasing motives in working towards a socially responsible outcome it could be argued that firms are becoming less likely to favor the traditional view of tax avoidance. Given these developments it might be argued that investors do not value tax avoidance as much as assumed under the traditional view but value firms which are paying their „fair share‟ of tax.

2.2 Tax avoidance and firm value

The effect of tax avoidance on firm value could be addressed via the traditional view on tax avoidance. As mentioned above, this traditional view of tax avoidance states that tax avoidance is value

maximizing activity that transfers wealth from the state to corporate shareholders (Kim, Li and Zhang, 2011). Simply inferring on this traditional view implies that tax avoidance increases firm value. However, as mentioned under the introduction, Dyreng et al (2008) found a great dispersion in the level of tax avoidance. Hence, tax avoidance might not simply be a transfer of wealth from the state to corporate shareholders. A possible remark on the traditional view might be the separation of

(8)

8

traditional view of tax avoidance I will first address the effect of tax avoidance on firm value from Desai and Dharmapala (2009).

As described under the introduction, these authors addressed the influence of tax avoidance on firm value as well. With exploring this relation the authors use OLS regression with firm value as the dependent variable and tax avoidance as the primary independent variable. The results show that the effect of tax avoidance on firm value is positive but not significantly different from zero. Their

research is based on a sample of U.S. Firms with 4492 observations on 862 firms over the period 1993 – 2001. Tax avoidance was measured by inferring on the difference between income reported to capital markets and tax authorities, the book-tax gap. The book-tax gap can capture a variety of sources of which earnings management (the over reporting of financial income) and tax avoidance activity are mentioned as the main contributors. Hence, the authors adjust for earnings management with an accruals proxy to isolate the component of the gap which is attributable to tax avoidance. This component of tax avoidance is then used by the authors as the proxy for tax avoidance. The findings of these authors might be extended to this thesis since their analysis is very similar. Differences are that I use a different measure for tax avoidance, one that is based on the ability to pay a low amount of taxes per dollar of pretax earnings over longer time periods. Desai and Dharmapala (2009) use a measure of tax avoidance which is based on the annual book-tax difference and not the long-run book-tax

difference. Dyreng et al. (2008) document that annual effective tax rates are not very good predictors of long run effective tax rates and thus not accurate proxies for long run tax avoidance. It could be questioned whether this also relates to the use of the book-tax gap as an indication of tax avoidance. Also, I focus on firms that are large enough to qualify for country-by-country reporting, firms that have at least a revenue of $ 850 million. As mentioned under the introduction, Rego (2003) argues that large firms have more opportunities to avoid income taxes. These additional opportunities might have a stronger or weaker influence on firm value. Nevertheless, extending their results to this study would predict a non-significant relation between tax avoidance and firm value.

(9)

9

risk is measured by the future stock return volatility, which is, closely related to stock price crash risk. Stock prices that crash present higher tock return volatility. These authors define tax avoidance as the ability to pay a low amount of cash taxes per dollar of pretax earnings over long time periods as in Dyreng et al (2008). The authors see tax aggressiveness as if firms take tax positions that are unlikely to survive a challenge by the I.R.S. Firm risk is measured by future stock price volatility. The authors use all useful observations which are obtained from the CRSP and Compustat database for their analysis. Comparing the findings of Guenterher et al. (2013) with those of Kim et al. (2011) provides a somewhat mixed perspective. That is because it is argued Guenter et al. (2013) that stock price

volatility does not increase due to tax avoidance but Kim et al. (2011) document that future stock price crash risk increases due to tax avoidance. Actual responses of a firm‟s stock price when the

involvement of a firm in tax avoidance reaches the news is investigated by Hanlon and Slemrod (2009) and Gallemore, Maydew, and Thornock (2014).

Hanlon and Slemrod (2009) relate tax avoidance to the reputational costs that firms may be subjected to when a firm is accused of tax sheltering in the news. These authors study a sample of 108 articles which were published between 1990 and 2004 in which a firm is accused of engaging in a tax

sheltering. Using event study methodology these authors find, on average, a decline in the stock price when a firm is accused of the involvement in tax sheltering via a news article. This finding is on average because there is some cross-sectional variation in the reaction. For example, the reaction is more negative for firms in the retail sector and less negative for firms that are viewed to be generally less tax aggressive. These results suggest that the relation between tax avoidance and firm value is not as straight forward as a wealth transfer from the state to corporate shareholders. However, these authors used a three day event window with conducting their analysis. Gallemore, Maydew, Thornock (2014) used the sample of Hanlon and Slemrod (2009) and enlarged their sample with additional tax shelter revelations. These authors also find, on average, a decline in the stock price around a tax shelter revelation. However, this price decline fully reverses back to its pre-event level within 30 days. Hence, the reputational costs of being accused of using aggressive tax avoidance activities seem to be temporary. Projecting these findings on this thesis might be appropriate since a firm‟s reputation might be expressed by its share price and share price could be a determinant of firm value. Gallemore et al. (2014) mention that if these temporary reputation costs are weighed against the large benefits to after-tax earnings from after-tax avoidance it remains puzzling why not all firms engage in after-tax avoidance. Hence, reputational costs do not seem to be value reducing and therefore do not present major flaws on the traditional view on tax avoidance.

(10)

10

increases risks. This finding might be relevant for a firm‟s valuation because bank loans are an important source of financing for firms (Bharath, Sunder, and Sunder 2008). Hence, debt might be an important driver of firm values. Considering equity financing, Goh, Lee, Lim and Shevlin (2016) relate tax avoidance to a firm‟s cost of capital and find that the cost of equity is lower for tax avoiding firms. Tax avoidance is measured by inferring book-tax differences and the long run cash effective tax rate. The authors believe that this finding could be explained by investors who require a lower rate of return due to positive cash flow effects of tax avoidances. Equity and debt holders might both see tax avoidance as an as an activity which involves risk as for example the stock price crash risk. However, equity holders share the upside benefits of tax avoidance where debt holders receive a fixed return on their loans.

The above does not lead to an unambiguous prediction on the relation between tax avoidance and firm value. The traditional view on the effect of tax avoidance on firm value predicts that tax avoidance is a simple wealth transfer from state to shareholders. Hence, it is likely that tax avoidance increases firm value. However, Desai and Dharmapala (2009) did not find a statistical significant relation between tax avoidance and firm value. From Gallemore et al. (2014) it follows that reputational costs are too much of a concern which relates to tax avoidance. Hasan et al. (2014) find that firms with greater tax avoidance incur higher spreads when obtaining bank loans where Goh et al. (2016) document that tax avoiding firms face lower costs of equity capital. Despite this dispersion I follow the traditional view on tax avoidance. Namely, reputational costs do not seem to be an issue and equity financing seems to be lower for tax avoiding firms. Hence, I formulate the first hypothesis as follows:

Hypothesis 1: Tax avoidance increases firm value.

2.3 Corporate tax avoidance, firm value and a firm’s governance

(11)

11

measured using the percentage of ownership held by institutional investors. It is argued that institutional investors have a greater incentive and capacity to monitor managerial performance. However, a possible remark on these institutional investors to as a proxy for governance could be made from the Blaylock, 2016. Namely, managers might obscure the reporting of financial performance to make aggressive tax transactions harder for the tax authority to detect. Financial reports are used by shareholders to monitor the managers their performance. Obscuring these reports makes it more difficult for shareholders to assess and monitor managerial performance, which makes it easier for managers to use the firm‟s resources for their personal purposes without shareholder knowledge. Nevertheless, relating the findings of Desai and Dharmapala (2009) to this thesis predicts that the effect of tax avoidance on firm value is amplified with the level of a firm‟s governance.

Also favoring the agency perspective on tax avoidance is the research of Kim et al. (2011). These authors documented that tax avoidance and future stock price crash risk are positively related, as described under section 2.2. Moreover, the authors found that this relation is less pronounced at firms with stronger monitoring. Their explanation is that tax avoidance activities facilitate managerial opportunistic behavior. Complex tax shelters create tools and masks for managers to manufacture earnings and conceal negative operating outcomes for an extended period. Accordingly, negative information is likely to stockpile within the firm. Good governance institutions should reduce these opportunistic behaviors and hence negative operating outcomes. Applying this finding in the context of this thesis would imply that a firm‟s governance decreases opportunistic behavior by managers and hence increase the positive effects of tax avoidance on firm value. Wilson (2009) who developed a profile to determine the likelihood of a firm engaging in tax sheltering assessed whether tax sheltering is associated with wealth creation for shareholders. The author finds that active tax shelter firms with strong corporate governance exhibit positive abnormal returns during the period of sheltering. These results support the agency point of view stressing the importance of good corporate governance systems. Closely related as well is Wang (2010) who documented that investors place a price premium on tax avoidance but this premium decreases with corporate opacity. Tax avoidance is measured by inferring on the book tax gap and using the long term cash ETR as in Dyreng et al. (2008). The author argues that this finding is consistent with the notion that corporate transparency facilitates monitoring of managerial actions and hence reduces agency concerns. This reasoning might be extended to a firm‟s governance and hence it might be expected that an increase in a firm‟s

governance decreases the agency concerns related to tax avoidance. Hence, the level of a firm‟s governance might be expected to amplify the effect of tax avoidance on firm value.

(12)

12

aggressive if its cash effective tax rate is in the bottom decile of an industry/year. The authors argue that this finding is consistent with the negative implications which relate to tax avoidance. Namely, tax avoidance increases possibilities for managerial rent extraction which, in turn, decrease future

profitability. Hence, the authors stress that it is beneficial for firms to have good governance

mechanisms to align the managerial interests with those of shareholders. Translating these findings to this thesis would imply that the level of a firm‟s governance is relevant for the relation between tax avoidance and firm value. Reasoning further would incline that if tax avoidance increases firm value, this relation might be amplified by the level of governance since this reduces rent diverting activities of a firm‟s management.

Contrary to the above described findings favoring the agency view are the results of Blaylock (2016). This author tests empirically whether aggressive forms of tax avoidance affect shareholders due to extra facilities for managerial rent extraction. The results show no evidence between tax avoidance and rent extraction among US firms. Tax avoidance is measured using Wilson‟s (2009) model for tax sheltering, the adjusted book-tax difference and the DTAX measure from Frank, Lynch and Rego (2009). The author argues that rent extraction is most likely facilitated with the more aggressive forms of tax avoidance and uses these proxies to capture tax avoidance. Furthermore, the author finds that two of the three tax-avoidance proxies are positively associated with future performance, even among poorly governed firms. Performance is measured by the earnings before extraordinary items or cash from operations divided by lagged total assets. Moreover, the results do not indicate less optimal investment policy among poorly governed firms. Given these outcomes, the author argues that the findings are consistent with the traditional view on tax avoidance. As to the author possible

inaccuracies of the study might be that tax avoidance, rent extraction and governance are measured by proxies. Given this, the findings might also be driven by the inability of these proxies to measure their underlying constructs. Nevertheless, translating to these findings to this thesis questions the agency view.

Concluding on this section leads to the second hypotheses. The traditional view of tax avoidance states that tax avoidance results in a wealth transfer from state to shareholders. The agency view of tax avoidance where ownership and control are separated puts this traditional view into consideration. Namely, complex tax avoidance transactions can provide managers with tools, masks and

(13)

13

Given that most previous literature provides supportive evidence to the agency view on tax avoidance I formulate the second hypothesis as follows:

Hypothesis 2: The increase in firm value due to tax avoidance is amplified by the level of a firm’s governance.

3. Methodology

For the analysis of both hypotheses I will follow a similar approach as in Kim, Li and Zhang (2009) who measure the relation between tax avoidance and stock price crash risk and Desai and Dharmapala (2009) in their study on tax avoidance and firm value. To test the first hypotheses I will use ordinary least squares regression with firm value as the dependent variable and the tax avoidance as the primary independent variable. This will result in the following regression equation:

qit = α0 + α1TAXVARt-1 + αqCONTROLSt-1 + εt (1)

where q is the measure of firm value, TAXVAR is the measure of tax avoidance, CONTROLS are control variables and ε is an error term. I will modify the first equation with a governance variable to test the second hypothesis. This results in the second equation:

qit = α0 + α1TAXVARt-1 + α2GOVERNt-1 + α3(TAXVARt-1 * GOVERNt-1 ) + αqCONTROLSt-1 + εt (2)

where the variable GOVERN is a measure of the firm‟s governance. The tax avoidance, governance and control variables are all one-year lagged as in Kim et al. (2011). Recall that Hypothesis 2 predicts that the relation between tax avoidance and firm value is amplified for firms with higher levels of governance. Hence, the coefficient for TAXVARt-1 * GOVERNt-1 is expected to be positive, if a higher TAXVARt-1 and GOVERNt-1 indicate higher levels of tax avoidance and governance.

3.1 Measuring Firm Value

I will measure firm value using Tobin‟s Q (Desai and Dharmapala, 2009; Cremers and Ferrell, 2014). Tobin‟s Q will be computed as in Cremers and Ferrell (2014) as the ratio of market value of assets to book value of assets which will be calculated using the following equation:

(14)

14

where P represents the stock price at the end of the year, BTA is the book value of total assets, BE the book value of equity and ADT is the amount of deferred taxes. Given this computation, a higher value of Tobin‟s Q means that a firm is higher valued by investors. For example, a value which is higher than 1 means that the stock is more expensive than the replacement costs of its assets. Desai and Dharmapala (2009) modify this general computation of Tobin‟s Q because these authors do not include the deduction of deferred tax expenses in their computation. They mention that current tax avoidance activity may result in changes to future tax liabilities and thus create a correlation between the dependent variable and the measure of tax avoidance. However, I will use another measure of tax avoidance, the long term effective tax rate which will be a measure for tax avoidance over time. Therefore I will not have this natural correlation and include the deduction of deferred tax expense. Goh, Lee, Lim, and Shevlin (2016) mention as well that measuring tax avoidance over a longer horizon avoids annual volatility in effective tax rates, and hence, mitigates concerns about earnings management through accruals because accruals are likely to reverse in the long run.

3.2 Measuring Tax avoidance

In order to measure tax avoidance I will use the long-run cash effective tax rate (LRETR) developed by Dyreng, Hanlon, and Maydew (2008):

∑

∑ (3)

This method uses cash tax paid instead of tax expense. Tax expense is composed of the sum of current tax expense and deferred tax expense. Deferred taxes will be paid or received in the future as a result of temporary book-tax differences. Dyreng et al. (2008) argue that many tax avoidance activities involve accelerating deductions and deferring income for tax purposes. This reduces current taxes but increases deferred taxes. Because tax expense includes both current tax expense and deferred tax expenses it will not reflect these forms of tax avoidance. This problem is overcome by using cash taxes paid which represents cash payments for income taxes. This item is obtained from the statement of cash flows and includes the cash paid for income taxes applicable to both current and prior years. Special items include a variety of items. Among other items it includes restructuring charges,

(15)

15

Dyreng et al. (2008) their measure of tax avoidance is long run because it measures effective tax rates over longer time periods than a one year ETR. Dyreng et al. (2008) use in their analysis different time frames up to ten years. This is done by summing a firm‟s total cash taxes paid over a ten-year period and divide that by the sum of its total pretax income (excluding special items) over the same ten-year period. This gives an effective tax rate which tracks the firm‟s tax costs over the long run. Note that this method is different from simply averaging one year effective tax rates because this tends to over or under weigh unusually high or low tax rates.

Dyreng et al. (2008) use measurement periods of one, five and ten years to determine the cash ETR in their analysis. I will use a measurement period of five years to establish the long run cash effective tax rate (LRETR) since many prior studies use this measurement period as well (Kim, Li and Zhan, 2011; Guenther, Matsunaga and Williams, 2013; Goh, Lee, Lim and Shevlin, 2016). Dyreng et al. (2008) consider a five-year time period as a long-run measure. Given the computation of the LRETR, a lower LRETR is associated with a higher level of tax avoidance. Hence, one would expect a negative regression coefficient for the tax avoidance measure according to the first hypothesis.

3.3 Measuring Corporate Governance

Product market competition as in Masulis, Wang and Xie (2007) will be used as a proxy for a firm‟s corporate governance. It has been argued that product market competition would have a disciplinary effect on managerial behavior, aligning the manager‟s interests with those of the shareholders. Also, product market competition is suggested to be one of the most effective mechanisms to eliminate managerial inefficiency. Kim et al. (2011) mention that complex and opaque tax avoidance transactions may help managers to obscure their rent-diverting which could be seen as the unauthorized compensation and abuse of corporate funds for personal purposes. Hence, product market competition is relevant for this research because it might reduce these forms of managerial inefficiency.

Following Masulis et al. (2007) I determine product market competition with two measures. The first is the Herfindahl index, which will be calculated as the sum of squared market shares of all

(16)

16

3.4 Control variables

As in Desai and Dharmapala (2009) I will control for sales which captures firm size. Sales are used as a measure for firm size because the authors mention that assets and market value would be

mechanically correlated with Tobin‟s Q. A measure of volatility will also be included to control for changes over time in risk associated with the firm‟s stock price. Volatility will be measured by the standard deviation of the monthly closing price of the previous five years. I will also include the value of stock option grants to executives as a fraction of total compensation as a control variable. It has been found that stock-based compensation is a determinant of firm value which might be explained via incentive alignment effects (Morck, Shleifer and Vishny, 1988). Research and Development

expenditures will proxy changes in intangibles that affect Tobin‟s Q but are imperfectly measured in the book value of assets. Furthermore, Desai and Dharmapala (2009) mention that tax shields can affect the value of engaging in tax avoidance and hence I will control for changes in firms leverage by including long term debt and debt in current liabilities. As in Cremers and Ferrel (2014) I will also include the value of capital expenditures (CAPEX) as a control variable.

4. Data Selection and Descriptive Statistics

4.1 Sampling Strategy

The primary database which I will be using is Standard and Poor‟s Compustat from where I will use Compustat North America and the Execucomp database. Compustat North America will provide the data which is required to obtain information regarding firm value, tax avoidance, measures of a firm‟s governance and control variables. The Execucomp database will provide information for a specific control variable and CRSP provides stock price data. The sampling period which I will use runs from 2003 until 2013. Appendix 1 provides an overview of the data items that are extracted from these databases and used in this thesis.

The main selection criterion is that a firm is covered by Compustat North America since most

variables are obtained via this database. I do not restrict the sample to US based enterprises but include all firms that are covered by the Compustat North America database. That is because the framework for BEPS implementation should be open to all interested countries and jurisdictions, as stated by the G20 leaders (OECD, 2016). Another selection criterion is that firms are large enough to qualify for country-by-country reporting. This means for this study that the revenue of firms in the sample should exceed USD 850 million (Internal Revenue Service, 2016).

4.2 Descriptive statistics and correlations

Given the sampling method described above I obtain an unbalanced panel of 7098 firm-year

(17)

17

are used throughout this thesis. These descriptive statistics are assembled with the use of Stata13. As in Dyreng, Hanlon, and Maydew (2008) I excluded observations where the LRETRt-1 had a negative or zero denominator and winsorized at zero and one. Dyreng et al (2008) argue that the LRETRs are more interpretable when using these criteria. The other variables are winsorized at the 90% level as in Kennedy, Lakonishok, and Shaw (1992). Scaled variables are scaled by the total book value of assets. I present a LRETRt-1 based on the full sample and one which only contains United States (US) based firms. From here it could be observed whether there are distributional differences between the US and the full subsamples. Also, the US subsample could be used to verify whether the outcomes are in line with Dyreng et al. (2008) who base their analysis on US firms. Given these subsamples, one could observe that the descriptive statistics on the full sample and US only sample are very similar. Furthermore, a quarter of the sample is able to maintain a LRETRt-1 of 17 percent or less which corresponds closely to Dyreng, Hanlon, and Maydew (2008). These authors find that a quarter of their sample is able to maintain a 5 year ETR below 18 percent. It could also be observed from these subsamples that approximately 86% of the firms included in the sample are US based and 14% of the firms are based in other jurisdictions.

Furthermore, Tobins‟s Q has a mean of 1.62 and a standard deviation of 0.69 which corresponds closely to Cremers and Ferrell (2014) who obtain a q of 1.68 with a standard deviation of 0.98. Their findings are based on a sample of approximately 1200 firms which runs from 1990 until 2006. An average q of 1.62 means that a firm‟s stock is usually valued higher than the replacement costs of their assets. Although I use the Herfindahl index and Product market uniqueness as in Masulis, Wang and Xie (2007) these authors do not present descriptive statistics on these items. Nevertheless, I would like to recall from the computation of these items that a higher Herfindahl index implies a lower level of governance and higher product market uniqueness implies a higher level of governance. The average sales of firms in this sample are much higher than the sample used by Desai and Dharmapala (2009). These authors obtained firms with average sales of $ 3.58 million where I have average sales of $ 8.64 billion. This might be explained by the additional sampling criterion that firms have at least a revenue of $ 850 million.

(18)

18

Table 1 descriptive statistics

This sample below is an unbalanced panel consisting of 7098 firm years with 1227 unique firms. The sample runs from 2003 until 2013 and includes all firms that are covered by Standard and Poor‟s Compustat North America with an annual revenue of at least $ 850 million. Q proxies firm value and is computed as the ratio of market value of assets to book value of assets. LRETR refers to the Long Run Cash Effective Tax rate which is computed by summing a firm‟s total cash taxes paid over a five year period which is divided by the sum of its total pretax income less special items over the same five year period. A firm‟s governance is determined on the basis of the product market competition where product market competition is measured as in Masulis et al. 2007 using the Herfindahl index and product market uniqueness (Unique). The Herfindahl index is calculated as the sum of squared market shares of all COMPUSTAT firms in each Fama French (1997) industry. Unique is measured by each industry‟s median ratio of selling expense to sales. Sales are the sales in millions of dollars. Total assets represents the book value of assets in millions of dollars. R&Dexp are the research and development expenditures. Debt long is the long term debt. Debt Short represents short term debt. CAPEX are the capital expenditures. Stock_compensation is the percentage of stock option grants to executives compared to the total compensation. Volatility represents stock price volatility measured by the monthly closing price of the preceding five years. Variables which are stated to be (scaled) are scaled by the total book value of assets.

Quintiles

Variable n Mean S.D. Min 0.25% Mdn 0.75% Max

Q 7098 1.62 0.69 0.9 1.09 1.4 1.94 3.42

LRETR full sample t-1 7098 0.25 0.13 0 0.17 0.25 0.32 1

LRETR USfirms t-1 6118 0.25 0.13 0 0.18 0.25 0.32 1

Herfindhal index t-1 7098 7.53 5.11 2.09 4.06 5.80 9.49 64.78

Unique t-1 7098 0.23 0.13 0.04 0.13 0.2 0.31 0.69

Sales t-1 (millions of $) 7098 8645.56 10640.03 1194.94 2110.00 4097.68 10057.51 41862

Total assets t-1 (millions of $) 7098 14906.20 22753.11 824.99 2107.60 5074.79 16005.00 92389

R&Dexp t-1 (scaled) 3526 0.03 0.03 0 0 0.01 0.04 0.12

Debt Long t-1 (scaled) 7098 0.2 0.14 0 0.08 0.19 0.29 0.48

Debt Short t-1 (scaled) 7098 0.03 0.04 0 0 0.01 0.05 0.14

CAPEX t-1 (scaled) 7098 0.05 0.04 0 0.02 0.04 0.06 0.14

Stock_compensation t-1 5704 0.2 0.19 0 0 0.17 0.33 0.61

Volatility t-1 7098 10.13 6.72 2.53 5.12 8.21 13.14 27.57

Table 2 provides correlations between the variables which will be used in the regression analyses. It could be observed that the LRETRt-1 and Tobin‟s Q are negatively correlated. Although correlation is not causation, this negative correlation supports hypotheses 1 which states that tax avoidance increases firm value. However, this correlation is not statistically significant as it has a p-value of 0.804.

Furthermore, the correlation of -0.200 between the two measures of a firm‟s governance (Herfindahl index and Product market uniqueness) is statistically significant. This negative correlation is expected because a higher Herfindahl index corresponds to a lower level of a firm‟s governance where a higher value of uniqueness corresponds to a higher level of a firm‟s governance. Nonetheless, it is worth mentioning that a correlation of -0.200 is only a weak correlation and might have been expected to be stronger because both items proxy for a firm‟s level of governance. Both measures of a firm‟s

(19)

19

Table 2 descriptive statistics

The correlations below are based on an unbalanced panel consisting of 7098 firm years with 1227 unique firms. The sample runs from 2003 until 2013 and includes all firms that are covered by Standard and Poor‟s Compustat North America America with an annual revenue of at least $ 850 million. The variables and their description are similar to those presented in table 1.The value between parentheses represents the p-value. Scaled variables are scaled by the total book value of assets.

A B C D E F G H I J K Q A 1 LRETR full t-1 B -0.002 1 (0.909) Herfindhal index t-1 C -0.098 -0.044 1 (0.000) (0.000) Unique t-1 D 0.266 0.008 -0.200 1 (0.000) (0.506) (0.000) Sales t-1 (millions of $) E -0.043 0.039 0.025 0.043 1 (0.000) (0.001) (0.033) (0.000) R&Dexp t-1 (scaled) F 0.318 -0.181 -0.024 0.535 0.052 1 (0.000) (0.000) (0.160) (0.000) (0.002)

Debt Long t-1 (scaled) G -0.146 -0.098 0.135 -0.163 -0.085 -0.258 1

(0.000) (0.000) (0.000) (0.000) (0.000) (0.000)

Debt Short t-1 (scaled) H -0.037 0.027 0.025 0.037 0.148 -0.024 0.102 1

(20)

20

5. Empirical Results

5.1 Regression analysis on hypothesis 1

Table 3 presents the results of the regression analyses which are conducted to test the first hypothesis. All regressions are conducted with the use of Stata13. All columns have Tobin‟s Q (q) as the

dependent variable where the first column has the Long Run Cash Effective Tax Rate (LRETRt-1) as the primary independent variable. Column 2 presents the results from the same analysis but with an indicator variable as its primary independent variable. I included this indicator variable as in Guenther, Matsunga, and Williams (2013) which marks if a firm is tax aggressive. The variable takes on a value of one if the firm-year observation is in the lowest LRETR quintile and zero otherwise. This may identify whether the relation between tax avoidance and firm is more evident for a higher level of tax avoidance. I included the third column for completeness, this column uses the same variables as the first column but restricts the observations included to US based firms.

As in Desai and Dharmapala (2009) I conducted all three regressions with both time and firm fixed effects. It could be observed from table 7 in the second appendix that the average qis for example much lower in 2008 than in 2013 which confirms the inclusion of time fixed effects. Desai and Dharmapala (2009) argue that the inclusion of firm fixed effects controls for many of the sources of cross-sectional variation in q across firms that have been discussed in the literature. Also, the

Huasman tests included in Appendix 3 indicates that fixed effects are preferred over a random effects model. Furthermore, the regressions are conducted using robust standard errors to correct for

heteroskedasitcity. Normality tests indicate that the residuals are not normally distributed, which might lead to wrong inferences about the coefficient estimates. However, the sample sizes are probably large enough so that I could be less concerned about non normality of the residuals (Brooks, C. 2007). According to the central limit theorem, the test statistics will asymptotically follow an appropriate distribution. Appendix 3 contains the output of the normality test, heteroskedasticity test as well as the Hausman test.

(21)

21

However, as with the LRETRt-1 this coefficient on the indicator variable is not statistically significant. Column 3 which presents the results of the US based firms is consistent with the first column and hence does not provide any contrasting evidence. In conclusion, these results do not support the traditional view of tax avoidance and hence do not support the first hypothesis.

Table 3 OLS regression

This table contains the output of the ordinary least squares (OLS) regressions which are based on an unbalanced panel consisting of 7098 firm years with 1227 unique firms. The sample runs from 2003 until 2013 and includes all firms that are covered by Standard and Poor‟s Compustat North America America with an annual revenue of at least $ 850 million. The dependent variable is Tobin‟s Q which measures firm value. The primary independent variable is the LRETRt-1 in column (1). Column (2) uses an indicator variable as its primary

independent variable. This indicator variable takes on a value of one if the observation is in the upper quintile of the LRETRt-1 variable and

zero otherwise. Column (3) is restricted to US based firms and uses the LRETRt-1 as its primary independent variable. The other variables

correspond to those in table 1. Furthermore, the regressions are conducted using robust standard errors to correct for heteroskedasticity. The *, ** and *** denote significance levels at the 10%, 5% and 1% level respectively. The value between brackets in the table represents the t-statistic. (1) (2) (3) Tax avoidance LRETRt-1 -0.06 (-0.45) Indicator LRETRt-1 -0.01 (-0.37) LRETR USt-1 -0.06 (-0.42) Control variables Salest-1 (millions of $) -1.96*10^-5*** (-4.46) -1.96*10^-5*** (-4.49) -1.78*10^-5*** (-4.07) RD expt-1 (scaled) 4.77*** (2.98) 4.75*** (2.99) 4.50*** (2.68) Debt Longt-1 (scaled) -0.97***

(-6.15)

-0.98*** (-6.17)

-0.98*** (-6.04) Debt Shortt-1 (scaled) -0.47*

(-1.90) -0.47* (-1.87) -0.52** (-2.06) CAPEXt-1 (scaled) -0.09 (0.15) -0.10 (-0.15) -0.11 (0.17) Stock compt-1 0.05 (0.80) 0.05 (0.80) 0.05 (0.76) Volatilityt-1 -0.02*** (6.73) -0.02*** (6.70) -0.02*** (6.71) Intercept 2.02*** (23.39) 2.01*** (24.23) 2.01*** (22.30)

Firm and year fixed effects YES YES YES

# of observations 2909 2909 2802

R2 _0.31 _0.31 _0.31

Adjusted R2 _0.31 _0.31 _0.30

5.2 Regression analysis on hypothesis 2

Table 4 presents the results of the regression analyses which are conducted to test the second

hypothesis. As for the first hypotheses I conducted three regressions where all three columns use q as the dependent variable. The first column uses the LRETRt-1 as a measure for tax avoidance, the second uses the same indicator variable as in section 5.1 which marks if a firm is tax aggressive. The variable takes on a value of one if the firm-year observation is in the lowest LRETR quintile and zero

(22)

22

The second hypothesis states that the relation between tax avoidance and firm value is amplified by the level of a firm‟s governance. Hence, the interaction term between tax avoidance and the

governance proxy is of interest in this regression equation. Recall that a lower value of LRETRt-1 corresponds to a higher level of tax avoidance, a higher Herfindahl index corresponds to a lower level of governance and a higher Unique measure corresponds to a higher level of governance. Thus, the coefficient on TAXVARt-1*Herfindahlt-1 is expected to be negative where the coefficient on TAXVAR*Uniquet-1 is positive when TAXVARt-1 is captured by the LRETRt-1. The coefficient on TAXVAR*Herfindahlt-1 is expected to be positive where the coefficient on TAXVARt-1*Uniquet-1 is expected negative when TAXVARt-1 is expressed by the indicator variable. It can be observed from Table 4 that for all three regression equations there are no significant coefficients on these terms indicating that a firm‟s governance does not modify the relation between tax avoidance and firm value.

Table 4 OLS regression

This table contains the output of the ordinary least squares (OLS) regressions which are based on an unbalanced panel consisting of 7098 firm years with 1227 unique firms. The sample runs from 2003 until 2013 and includes all firms that are covered by Standard and Poor‟s Compustat North America America with an annual revenue of at least $ 850 million. The dependent variable is Tobin‟s Q which measures firm value. The primary independent variables are the LRETR and its interaction terms with the governance proxies in column (1). Column (2) uses an indicater variable as its primary independent variable and this variable also interacts with the governance proxies. This indicator variable takes on an value of one if the observation is in the upper quintile of tax aggressiveness and zero otherwise. Column (3) is restricted to US based firms and uses the LRETR and interaction terms as its primary independent variables. The variable TAXVARt-1 captures the

proxy for tax avoidance. The other variables correspond to those in table 1. Furthermore, the regressions are conducted using robust standard errors to correct for heteroskedasticity. The *, ** and *** denote significance levels at the 10%, 5% and 1% level respectively. The value between brackets in the table represents the t-statistic.

(1) (2) (3)

LRETR Indicator US based LRETR

Tax avoidance measure

TAXVARt-1 -0.37 (-1.14) 0.01 (0.15) -0.34 (-1.04) Herfindahlt-1 -0.00 (-0.26) 0.00 (0.74) -0.00 (0.66) TAXVAR * Herfindahlt-1 0.02 (0.77) -0.00 (-0.10) 0.02 (0.80) Uniquet-1 -1.89*** (-3.30) -1.75*** (-3.43) -1.58*** (-2.75) TAXVAR* Uniquet-1 0.66 (0.65) -0.07 (-0.26) 0.51 (0.49) Control variables Salest-1 (millions of $) -2.04*10^-5*** (-4.71) -1.99*10^-5*** (-4.67) -1.87*10^-5*** (-2.75) RD expt-1 (scaled) 4.99*** (3.20) 5.07*** (3.28) 4.80*** (2.88) Debt Longt-1 (scaled) -0.94***

(-5.98)

-0.93*** (-6.18)

-0.95*** (-5.93) Debt Shortt-1 (scaled) -0.47*

(-1.88) -0.46* (-1.85) -0.51** (-2.02) CAPEXt-1 (scaled) -0.12 (-0.19) -0.12 (-0.19) -0.07 (-0.10) Stock compt-1 0.04 (0.68) 0.05 (0.72) 0.05 (0.67) Volatilityt-1 0.02*** (6.94) -0.02*** (6.93) 0.02 (6.84) Intercept 2.57*** (14.59) 2.48*** (15.57) 2.47 (13.89)

Firm and year fixed effects YES YES YES

# of observations 2909 2909 2802

R2 _0.32 _0.32 _0.31

(23)

23

5.3 Additional analysis

Table 5 presents results on additional regression analyses. As in Desai and Dharmapala (2009) I conduct two separate regressions where one is based on a sample of firm year observations which have a high level of governance and one that is low. The high and low samples are determined based on the sample median of the governance proxies. Consistent with the second hypotheses, the effect of tax avoidance on firm value should be significant and stronger for well-governed firm year observations. Recall that a high Herfindahl index implies a low level of governance where a high product market uniqueness variable implies a high level of governance. Hence, columns 2 and 3 contain a sample of firm year observations which are high on governance and columns 1 and 4 contain observations which are low on governance. I conducted these regressions for the full sample with q as the dependent variable and the LRETRt-1 as the primary independent variable. I only computed these regressions with the LRETRt-1 as the independent variable because this variable is the primary variable of interest. Again, all regressions presented in table 5 are conducted with both time and firm fixed effects and with robust standard errors to correct for heteroskedasticity.

(24)

24

Table 5 OLS regression:

This table contains the output of the ordinary least squares (OLS) regressions which are based on an unbalanced panel consisting of 7098 firm years with 1227 unique firms. The sample runs from 2003 until 2013 and includes all firms that are covered by Standard and Poor‟s Compustat North America America with an annual revenue of at least $ 850 million. The dependent variable is Tobin‟s Q which measures firm value. The primary independent variable is the LRETRt-1 which proxies for tax avoidance. The regressions in column 1 and 4 are based

on subsamples which are low on governance and columns 2 and 3 are based on subsamples which are high on governance. These subsamples are established according to the sample median of their corresponding governance proxies. The other variables correspond to those in table 1. Furthermore, the regressions are conducted using robust standard errors to correct for heteroskedasticity. The *, ** and *** denote

significance levels at the 10%, 5% and 1% level respectively. The value between brackets in the table represents the t-statistic.

(1) (2) (3) (4)

High Herfindahlt-1 Low Herfindahlt-1 High Uniquet-1 Low Uniquet-1

Tax avoidance variable

LRETRt-1 0.07 (0.41) -0.13 (-0.72) 0.02 (0.13) -0.24 (-1.47) Control variables Salest-1 (millions of $) -2.25*10^-5*** (-2.97) -1.8*10^-5*** (-3.44) -1.75*10^-5*** (-3.27) -2.75*10^-5*** (-3.45) RD expt-1 (scaled) 7.77*** (2.73) 3.37* (1.82) 4.86*** (2.77) 4.42 (1.34) Debt Longt-1 (scaled) -1.26***

(-5.10) -0.93*** (-4.63) -0.92*** (-5.02) -0.83*** (-2.74) Debt Shortt-1 (scaled) -0.68*

(-1.77) -0.50 (-1.64) -0.67** (-2.13) 0.01 (0.03) CAPEXt-1 (scaled) -0.92 (-0.77) 0.27 (0.35) 0.45 (0.56) -0.76 (-0.73) Stock compt-1 0.17 (1.48) -0.00 (-0.01) 0.07 (0.74) -0.05 (-0.59) Volatilityt-1 0.01* (1.87) 0.02*** (6.54) 0.02*** (6.16) 0.01 (1.49) Intercept 2.09*** (13.89) 2.09*** (18.82) 2.04*** (18.10) 2.06*** (14.26) Firm and year fixed

effects

YES YES YES YES

# of observations 1232 1677 1775 1134

R2 _0.32 _0.34 _0.34 _0.30

Adjusted R2 0.31 0.33 0.34 0.29

6. Conclusion

This thesis investigated whether (i) tax avoidance increases firm value and whether (ii) this relation is modified by a firm‟s governance. Both questions have been investigated using Ordinary Least Squares (OLS) regression with both time and firm fixed effects. The first question is addressed by using firm value as the dependent variable and tax avoidance as the primary independent variable. Firm value is measured by Tobin‟s Q and tax avoidance is measured by the long run cash effective tax rate

(LRETR).The results do not show a significant coefficient on the tax avoidance variable suggesting a non-significant relation between tax avoidance and firm value. The second question is investigated by augmenting the first regression with an interaction term between tax avoidance and the level of a firm‟s governance. This augmented regression equation uses the same proxies for tax avoidance and firm value and measures the level of a firm‟s governance by inferring on the product market

(25)

25

The first hypothesis states that tax avoidance increases firm value. This prediction is primarily driven by the traditional view on tax avoidance as the wealth transfer from the state to corporate shareholders. Given the results, I can not confirm this hypothesis and hence do confirm the findings of Desai and Dharmapala (2009) who found a positive but insignificant relation between tax avoidance and firm value. The second hypothesis predicts that the relation between tax avoidance and the level of a firm‟s governance is modified by the level of a firm‟s governance. Again, I can not provide evidence which supports this hypothesis. A possible explanation for these deviations from the traditional view on tax avoidance could be provided by the recent change in attitude of companies in respect to paying taxes from minimizing tax expenses in order to maximize shareholder value to paying their „fair share‟ of tax. As mentioned under section 2.1, firms are increasingly working towards a socially responsible outcome which extents as well towards paying a „fair share‟ of tax. This development might as well be demanded by investors resulting in firms that are paying their „fair share‟ of tax are valued higher by these investors. Another possible explanation might be that I base my analysis on large firms, firms which are large enough to qualify for country-by-country reporting. Compared to small firms, these large firms might be more likely to be published about in the news when involved in aggressive tax avoidance strategies and hence be scrutinized by the public.

A potential limitation of this thesis might be the choice of proxies for tax avoidance and governance. Blaylock (2016) mentioned as well that the chosen proxies might not be able to measure their

(26)

26

Sources

Bankman, J., 1999. The new market in corporate tax shelters. Tax Notes, June 21, p. 1775; 83

Bharath, S., Sunder, J., Sunder, S., 2008. Accounting quality and debt contracting. The Accounting Review 83, 1 – 28

Blaylock, B.S., 2016. Is Tax Avoidance Associated with Economically Significant Rent Extraction among U.S. Firms? Contemporary Accounting Research, Vol. 33, No. 3, pp. 1013 – 1043

Blouin, J., 2014. Defining and Measuring Tax Planning Aggressiveness. National Tax Journal, December 2014, 67 (4), 875 – 900

Brooks, C., 2008. Introductory econometrics for finance. Cambridge University Press, Cambridge.

Cremers, M., Ferrel, A., 2014. Thirty Years of Shareholder Rights and Firm Value. The Journal of Finance. Vol. LXIX, No. 3, June 2014

Desai, M.A., Dharmapala, D., 2009. Corporate tax avoidance and firm value. The Review of Economics and Statistics, August 2009, 91(3): 537 – 546

Dyreng, S.D., Hanlon, M., Maydew, E.L., 2008. Long-Run Corporate Tax Avoidance. The Accounting Review. Vol. 83, NO. 1, 61 – 82

Frank, M., Lynch, L., Rego, S., 2009. Tax reporting aggressiveness and its relation to aggressive financial reporting. The Accounting Review 84 (2): 467 – 96

Gallemore, J., Maydew, E.L., Thornock, J.R., 2014. The Reputational Cost of Tax Avoidance. Contemporary Accounting Research, Vol. 31 No. 4 pp. 1103 – 1133

Goh, B.W., Lee, J., Lim, C.Y., Shevlin, T., 2016. The Effect of Corporate Tax Avoidance on the Cost of Equity. The Accounting Review, Vol. 91, No. 6, November 2016, pp. 1647 – 1670

(27)

27

Guenter, D.A., Matsunaga, S.R., Williams, B.M., 2013. Tax Avoidance, Tax Aggressiveness, Tax Risk and Firm Risk. Lundquist College of Business, University of Oregon, Eugene, OR 97403 USA, August 2013

Hanlon, M., Heitzman, S., 2010. A review of tax research. Journal of Accounting and Economics, 50 (2010), 127 – 178

Hanlon, M, Slemrod, J., (2009). What does tax aggressiveness signal? Evidence from stock price reactions to news about tax shelter involvement. Journal of Public Economics, 93 (2009), 126 – 141

Happé, R.H., 2011. Belastingrecht en de geest van de wet, Een pleidooi voor een beginsel-benadering in de wetgeving. Afscheidscollege op 20 mei 2011. Available via:

https://www.tilburguniversity.edu/upload/de20e45d-a089-4ffc-8bff-fb99387b34ec_afscheidsrede.pdf

(assessed at: 4-11-2016)

Hasan, I., Hoi, C.K., Wu, Q., Zhang, H., 2014. Beauty is in the eye of the beholder: The effect of corporate tax avoidance on the cost of bank loans. Journal of Financial Economics, 113 (2014), 109 - 130

Hoi, C.K., Wu, Q., Zhang, H., 2013. Is Corporate Social Responsibility (CSR) Associated with Tax Avoidance? Evidence from Irresponsible CSR Activities. The Accounting Review, Vol. 88, No. 6, pp. 2025 - 2059

Internal Revenue Service, 2016. Country-by-Country Reporting. Federal Register, Vol. 81, No. 126 / Thursday, June 30, 2016 / Rules and Regulations

Katz, S.P., Khan, U., Schmidt, A., 2013. Tax Avoidance and Future Profitability. Columbia Business School Research Paper No. 13-10. Available at SSRN: https://ssrn.com/abstract=2227149 (assessed at: 17-11-2016)

Kim, J., Li, Y., Zhang, L., 2011. Corporate tax avoidance and stock price crash risk: Firm-level analysis. Journal of Financial Economics 100 (2011) 639 – 662

(28)

28

Lamberts, P.B.W.L., 2016. Fair Taxation: Truth is in The Eye of the Beholder. Weekblad Fiscaal Recht, 2016/123

Masulis, R.W., Wang, C., Xie, F., 2007. Corporate Governance and Acquirer Returns. The Journal of Finance. Vol. LXII, No. 4, August 2007

Morck, R., Shleifer, A., Vishny, R.W., 1988. Management ownership and market valuation: an empirical analysis. Journal of Financial Economics 20 (1988) 293 – 315

OECD, 2015. Transfer Pricing Documentation and Country-by-Country Reporting, Action 13 - 2015 Final Report, OECD/G20 Base Erosion and Profit Shifting Project, OECD Publishing, Paris.

http://dx.doi.org/10.1787/9789264241480-en

OECD, 2016. Background brief. Inclusive framework for BEPS implementation (march, 2016). Available via: https://www.oecd.org/tax/background-brief-inclusive-framework-for-beps-implementation.pdf (assessed at: 21-09-2016)

Porcano, T.M., 1986. Corporate Tax Rates: Progressive, Proportional, or Regressive. Journal of the American Taxation Association; Spring 86, Vol. 7 Issue 2, p17

Rego, S.O., 2003. Tax-Avoidance Activities of U.S. Multinational Corporations. Contemporary Accounting Research Vol. 20 No. 4 (Winter 2003) pp. 805 - 33

Wang, X., 2010. Tax Avoidance, Corporate Transparency, and Firm Value. Unpublished dissertation. The University of Texas at Austin. https://repositories.lib.utexas.edu/bitstream/handle/2152/ETD-UT-2010-12-2219/WANG-DISSERTATION.pdf?sequence=1 (assessed at 04-11-2016)

(29)

29

Appendix 1

This appendix contains table 6 in which an overview of the source of the data items used in this thesis is presented.

Table 6

Data source of the variables used in this thesis.

Panel A: Variables from Compustat

Data Item Used for Data code

Price Close – Annual (PRCC) Firm value Item #199 Common Shares Outstanding (CSHO) Firm value Item #25

Assets – Total (AT) Firm value Item #6

Common/Ordinary Equity – Total (CEQ) Firm value Item #60 Deferred Taxes Balance Sheet (TXDB) Firm Value Item #74

Income Taxes Paid (TXPD) LRETR Item #317

Pretax Income (PI) LRETR Item #170

Special Items (SPI) LRETR Item #17

Sales/Turnover Net (SALE) Control variable Item #12 Research and Development Expense (XRD) Control variable Item #46 Long-Term Debt – Total (DLTT) Control variable Item #9 Debt in Current Liabilities – Total (DLC) Control variable Item #34 Capital Expenditures (CAPX) Control variable Item #128 Sales/Turnover Net (SALE) Herfindahl index / product uniqueness Item #12 Selling, General and Administrative

Expense (XSGA)

Product uniqueness Item #132

Panel B: Variables from Execucomp

Data Item Used For Data Code

Total Compensation Control variable TDC1

Grant Date Fair Value of Options ($ - as valued by company)

Control variable OPTION_AWARDS_FV

Option Granted ($ - Compustat Black-Scholes value)

Control variable OPTION_AWARDS_BLK_VALUE

Panel C: Variable from CRSP

Data Item Used for Data Code

(30)

30

Appendix 2

This appendix contains table 7 in which the yearly distributions of the average value of q and the LRETRt-1 used in this thesis is presented.

Table 7

Sample distribution of Q and Long Run Effective Tax Rate by year. This distribution is based on an unbalanced panel consisting of 7098 firm years with 1227 unique firms. The sample runs from 2003 until 2013 and includes all firms that are covered by Standard and Poor‟s Compustat North America with an annual revenue of at least $ 850 million. q proxies firm value and is computed as the ratio of market value of assets to book value of assets. LRETRt-1 refers to the Long Run Cash Effective Tax rate which is computed by summing a firm‟s

total cash taxes paid over a five year period which is divided by the sum of its total pretax income less special items over the same five year period.

Year # of observations Average q Average LRETRt-1

(31)

31

Appendix 3

This appendix contains the test results on the regression analysis conducted under section 5.1. I have only presented the test results on the first column of table 3 because the relation between q and LRETRt-1 is of primary interest and the test results on the other columns did not differ tremendously. To test for heteroskedasticity I used the xttest3 command in Stata13 to conduct a modified Wald test which tests for groupwise heteroskedasticity in a fixed effects model. The output of this

heteroskedasticity test is presented in table 8. Table 9 presents characteristics and a formal normality test on the residuals. Table 10 presents a Hausman tests which is conducted to verify whether a fixed effects model is preferable over a random effects model.

Table 8 heteroskedasticity test

This table contains the output of the heteroskedasticity test on the regression which is presented in the first column of table 3. The regression heteroskedasticity test is based on the modified Wald test which tests for groupwise heteroskedasticity.

Chi2 (503) 4.5*10^36

P-value chi2 0.00

Table 9 Normality test on the residuals

This table contains the output of the normality test on the residuals of the OLS regression presented in the first column of table 3.

Skewness 0.19

Kurtosis 4.72

Shapiro Wilk test p-value 0.00

Table 10 Hausman test

This table presents the output of the hausman test on the regression in the firmst column of table 3.

Chi2 (7) 183.21

(32)

32

Appendix 4

This appendix contains the test results on the regression analysis which is conducted under section 5.2. Again, I have only presented the test results on the first column of table 5 because the relation between q and LRETRt-1 is of primary interest and the test results on the other columns did not differ

tremendously. To test for heteroskedasticity I used the xttest3 command in Stata13 to conduct a modified Wald test which tests for groupwise heteroskedasticity in a fixed effects model. The output of this heteroskedasticity test is presented in table 11. Table 12 presents characteristics and a formal normality test on the residuals. Table 13 presents a Hausman tests which is conducted to verify whether a fixed effects model is preferable over a random effects model.

Table 11 heteroskedasticity test

This table contains the output of the heteroskedasticity test on the regression which is presented in the first column of table 3. The regression heteroskedasticity test is based on the modified Wald test which tests for groupwise heteroskedasticity.

Chi2 (503) 2.7*10^33

P-value chi2 0.00

Table 12 Normality test on the residuals

This table contains the output of the normality test on the residuals of the OLS regression presented in the first column of table 3.

Skewness 0.18

Kurtosis 4.68

Shapiro Wilk test p-value 0.00

Table 13 Hausman test

This table presents the output of the hausman test on the regression in the firmst column of table 3.

Chi2 (7) 193.5

Tax Avoidance and Firm Value