The effects of interest barriers on the leverage of leveraged buy-outs and on the activity of the private equity industry

(1)

MSc Business Economics: Finance track

Master Thesis

The Effects of Interest Barriers on the Leverage of Leveraged Buy-outs

and on the Activity of the Private Equity Industry

Name:

Toon Beeren

Student number: 5876117

Supervisor:

dhr. dr. J.K. Martin

(2)

2

Statement of originality

This document is written by student Toon Beeren 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

The tax deductibility of interest is considered an important incentive for excessive debt financing. In the past few years various countries have implemented interest barrier rules to limit tax deductibility in order to prevent profit shifting of interest expenses. This particularly effects leveraged buy-outs and the private equity industry, since leveraged buy-outs are deals financed largely with debt. This study aims to evaluate the impact of the implemented interest barrier rules. Specifically, which effects the rules have on the number and capital structure of leveraged buy-outs. The methodology consists out of difference-in-differences regressions performed for Germany and Sweden, while comparing the results with surrounding countries. The results suggest that the interest barrier rules have no effects on the number and leverage percentage of leveraged buy-outs. There is some evidence that suggest a lower number of leveraged buy-outs in Sweden because of the changed tax law, although this is not significant. However, the research is limited by a small number of observations for the leverage percentage of leveraged buy-outs.

(4)

4

Table of contents

1. Introduction 5

2. Literature review 6

2.1 Capital structure 6

2.2 Introduction of interest barrier rules 8

2.3 German tax reform 2008 8

2.4 Swedish interest deduction limitation rules 2009 and 2013 10 2.5 Interest barrier rules for tax deductibility in the rest of Europe 11

2.6 Evidence of thin capitalization rules 12

3. Methodology 13

4. Data and descriptive statistics 15

4.1 Data 15

4.2 Descriptive statistics Germany and Western Europe 16

4.3 Descriptive statistics Sweden and Northern Europe 19

5. Results 22

5.1 Difference-in-Differences regressions Germany 22

5.2 Difference-in-Differences regressions Sweden 25

6. Robustness checks 27

6.1 Additional Difference-in-Differences regressions Germany 27

6.2 Additional Difference-in-Differences regressions Sweden 29

7. Conclusion 31

8. References 33

(5)

5

1. Introduction

Private equity (PE) firms are known for their excessive use of leverage in acquiring other companies through leveraged buy-outs (LBO’s). Evidence by Kaplan (1989) showed that the tax benefits associated with this financing structure creates value for PE firms. The subprime mortgage crisis in 2008 reflected the dangers of using such excessive leverage. Since then, a lot has changed. Nowadays, banks are more reluctant to issue loans to companies and various countries have implemented interest barriers to reduce tax deductibility. These interest barrier rules are invented to prevent firms from shifting profits from high-tax jurisdictions to low-tax jurisdictions and to prevent firms from excessive debt financing (Cohrs, 2013). These introduced interest barriers are also called thin capitalization rules (TCR’s). Especially PE firms are affected by these TCR’s, since they impact the value of the tax shield.

Since Modigliani and Miller (1958, 1963), researchers have studied the tax benefits of debt and questioned whether they affect corporate financing decisions. The benefits of the tax shield are known. However, until this date it is still uncertain whether these tax benefits affect financing decisions. Recent research showed that the interest barrier rule, which was introduced in Germany in 2008, impacted 76.9 percent of transactions in the first year after implementation (Knauer and Sommer, 2012). Therefore it will be interesting to take a closer look at these transactions and discover whether financing decisions are affected as a result of the decreasing tax benefits. More specific, whether these decreasing tax benefits will simultaneously lead to a decrease in the number of LBO’s and/or a decrease in the leverage percentage of LBO’s.

Because of the recently implemented tax barriers in Germany and Sweden, research on this subject can be done. After Germany’s announcement of introducing the interest barrier, other European countries such as Italy, France, The Netherlands and the UK considered the possibility of implementing similar interest barrier rules. We expect that in the nearby future more countries will follow this tax legislation to limit the excessive debt financing problem. Therefore this research will be applicable to a wider range than Germany and Sweden alone.

As mentioned before, Knauer and Sommer (2012) found that the interest barrier rule impacts the value of the tax shield. However, the effects in practice are not yet researched. Knauer and Sommer (2012) state that as a result of the interest barrier rule, the number of LBO’s may be lower. This study will contribute to the existing literature by researching these effects. Specifically, which effects these interest barriers have on the leverage of LBO’s and the activity of the PE industry. Aim of the study is to find out if the (reduced) benefits of the tax shield affect the corporate financing decisions of PE firms. This research will be done by comparing the number and financing structure of LBO’s before and

(6)

6

after the implemented tax laws in Germany and Sweden. These two countries are chosen in the first place because of the recently implemented tax barriers. And in the second place because companies in both countries rely primarily on bank loans as the major source of debt financing. Also, stock ownership is concentrated in both countries and many PE firms are managed by large institutional investors or families (Knauer and Sommer, 2012). By researching the effects of the reduced tax shields in these two countries a conclusion can be drawn whether corporate financing decisions are affected by the (reduced) benefits of the tax shield.

The remainder of this paper proceeds as follows. Section 2 provides an overview of empirical papers which contains research on the capital structure of LBO’s. Moreover, the recently implemented interest barrier rules in Germany and Sweden will be explained in detail and also the effects of such rules will be discussed. Section 3 presents the methodology used. Then, the data and descriptive statistics will be presented in section 4, while the results of the research will be presented and discussed in section 5. Robustness checks will be performed in section 6. Lastly, section 7 will summarize and conclude the paper.

2. Literature review

2.1 Capital structure

Modigliani and Miller (1958) formed the basis theorem on capital structure. Their theorem states that the market value of any firm is independent of its capital structure. That is, assuming that there are no taxes and the market is efficient. Modigliani and Miller (1963) later corrected their theorem by including the tax advantages of debt financing and recognizing the difference between the naïve traditional theorem without taxes and their new model with taxes. But, they emphasized that the larger advantage of the corrected version does not necessarily mean that corporations should maximize the amount of debt in their capital structures. According to Kaplan and Strömberg (2008) there are two reasons why PE firms are financing deals with large amounts of debt. First, leverage improves corporate governance and therefore minimize agency conflicts between fund managers and investors. Because of leverage, interest and principal payments have to be made. Therefore, capital cannot be wasted and fund managers will be more monitored by the limited partners. As a result it is less likely fund managers will dissipate money. Debt can be seen as a substitute for dividends. It is a mechanism which forces managers to handle their cash efficiently instead of wasting it on empire-building projects with low or negative NPV’s (Jensen,

(7)

7 1989).

Second, leverage increases firm value through the tax deductibility of interest. This is proven by Kaplan (1989), where he showed that tax benefits are an important source of value creation in management buyout transactions. Kaplan (1989) estimated that U.S. PE deals generated tax benefits between 21.0 and 142.6 percent of the premium paid to shareholders during the first half of the LBO wave in the 1980s. Also, Buettner et al. (2006) found that a higher local tax rate is associated with a greater use of internal debt financing. This suggests that firms use the higher tax rate to exploit the benefits of debt financing through the tax shield.

In further research MacKie-Mason (1990) found substantial evidence of corporate financing decisions affected by taxes. His empirical results showed that firms with high tax shields are less likely to finance with debt when the marginal tax rate drops. This is in line with the results of Axelson et al. (2013). Their results suggest that buyouts become more leveraged when credit is cheap and PE firms often use as much leverage as they can. Moreover, Graham (2000) found that firms around the year 2000 use debt more aggressively than they did in the 1980s. But still, some firms seems underleveraged. This is consistent with the results of Strömberg (2010), where he found that the number of LBO

transactions increased significantly over the last few decades. However, recent LBO’s differ in their capital structure. According to Guo et al. (2011) recent LBO’s are less highly leveraged than the earlier LBO’s in the 1980s, but they still contain significant default risk. The less highly leveraged LBO’s could be an indirect result of the subprime mortgage crisis in 2008. Nowadays, national tax codes are complicated and differ significantly which creates an opportunity for tax arbitrage. These tax distortions have

probably encouraged the use of excessive leverage evident in the crisis (Keen et al., 2010). Since then, a lot has changed in the financial world. The financial crisis led to less (favorable) loans for corporations and to governmental law changes which affect the value of the tax shield. Since evidence from the past by MacKie-Mason (1990) suggested that financing decisions are affected by taxes it could be the case that the measures made as a result of the financial crisis caused PE firms to decrease their leverage. However, other research is in contradiction with this evidence. Knauer et al. (2014) suggest that the value of the tax shield does not improve the performance of PE funds and they doubt the possibility that new tax laws will result in a decline in LBO leverage. According to Knauer et al. (2014) taxes have at best an indirect effect on corporate financing decisions and the gains to PE firms are mostly because of their improvements in restructuring companies. This is consistent with the study of Jenkinson and Stucke (2011), where they concluded that government restrictions on tax benefits will not affect LBO leverage. Jenkinson and Stucke (2011) also argued that tax savings are not a source of value creation for PE

(8)

8

investors, since those benefits are already reflected in the premiums paid by PE firms to acquire companies.

2.2 Introduction of interest barrier rules

In general the return on equity, i.e. dividends, is not deductible for the distributing company under domestic tax law (Cohrs, 2013). This non-deductibility encourage firms to reinvest the distributed profits instead of distributing it to shareholders. Return on debt, i.e. interest, however is deductible for the borrowing company and taxed at the level of the lender (Cohrs, 2013). This tax law distinction between debt and equity taxation gives companies incentive to finance with debt over equity.

Moreover, according to Cohrs (2013) consolidated tax groups can transfer their profits by way of debt with the interest expenses being deductible in a high-tax jurisdiction and interest income being taxable in a low-tax jurisdiction. This significantly lowers the overall taxation for consolidated companies. As a result of the profit shifting several countries have introduced or tightened TCR’s, which limit the amount of interest payments to consolidated tax groups that is deductible from the tax base (Haufler and Runkel, 2012). These TCR’s differ between countries and have changed over time. Typically, these rules limit interest deduction for loans provided by foreign affiliates if the debt-to-equity ratio of the borrowing consolidated tax group exceeds a certain threshold. This threshold is often called a safe haven (Buetnner et al., 2012). The first country who implemented a TCR was Canada in 1971, followed by Australia (1987) and the U.S. (1989). Over time many more countries followed. By 2005, sixty percent of European countries had implemented such rules (Buettner et al., 2012).

In general, there are two types of TCR’s. The “Fixed Ratio Approach” is the traditional rule which restricts the deductibility of interest expenses to shareholders if the debt-to-equity ratio of financing exceeds a certain level set in the law (Buslei and Simmler, 2012). The new rule does not restrict the deductibility of interest expenses above a certain debt-to-equity ratio threshold, but restricts the

deduction of interest expenses to a certain share of the firm’s taxable profits (Buslei and Simmler, 2012). This new rule is also called the earnings stripping rule. However, this second type of TCR has some escape options for companies. An example of this TCR type is the recently implemented interest barrier in Germany.

2.3 German tax reform 2008

The first TCR’s in Germany were introduced as of 1 January 1994. Until the reform in 2008, interest expenses were generally deductible from total revenues as regular expenses (Buslei and Simmler, 2012).

(9)

9

These rules applied to interest payments above € 250,000 and if the debt-to-equity ratio of 1.5:1 was exceeded (Cohrs, 2013).

Huizinga et al. (2008) found that a firm’s leverage not only depends on national tax rates, but also on international tax rate differences. This relation reflects the presence of international debt shifting. According to Huizinga et al. (2008) the existence of debt shifting because of international tax rate differences understate the impact of national taxes on debt policies by about 25 percent. Mainly this study by Huizing et al. (2008) and other anecdotal evidence had convinced German politicians that profit shifting by multinationals was a major issue (Buslei and Simmler, 2012). Especially since it came to a large extent at the expense of Germany, because profits were shifted to lower taxing countries.

As a result Germany adopted the interest barrier rule as of 1 January 2008. This rule was part of a major tax reform in 2008, which aimed at increasing incentives for investments and conducting business in Germany (Cohrs, 2013). This new interest barrier is in several ways more restrictive than the old regulations. Contrary to the old rules, it takes into account interest payments from all types of creditors and applies to all types of firms. The general rule states that interest expenses that exceed the amount of incurred interest income are limited in their deductibility as a business expense up to an amount of 30 percent of taxable earnings before interest, taxes, depreciation and amortization (EBITDA) (Buslei and Simmler, 2012). Moreover, interest payments that are not deductible in a year can be carried forward indefinitely and can be deducted in later years with high levels of EBITDA. This carry-forward

arrangement was also not included in the initial regulation.

As mentioned before this new rule has some escape clauses. Knauer and Sommer (2012) found that there are three escape clauses which will exempt firms from the interest barrier rule. First, in order to prevent small firms from suffering from an additional tax burden, small and medium-sized firms with annual net interest expenses below a million euro (minimum threshold) are exempted from this rule. Second, stand-alone companies with no possibility of consolidation are also exempted (stand-alone escape). There are two variants of this escape. The first refers to single companies that do not belong to a group and do not rely on significant shareholder debt financing. The second variant refers to members of consolidated tax groups (Buslei and Simmler, 2012). However, according to the German Corporate Income Tax Act the possibility of consolidation, instead of the actual scope of consolidation is sufficient for affiliation (Knauer and Sommer, 2012). Therefore, most companies cannot take advantage of this escape clause.

Third, companies who do not finance with excessive debt compared to the entire consolidated group (equity escape). A group member can deduct all interest payments if the member’s equity rate is

(10)

10

at least higher than the equity rate in the whole group (Buslei and Simmler, 2012).

According to Knauer and Sommer (2012) the interest barrier rule in Germany in 2008 especially affected the value creating potential through the increased leverage of LBO’s. It does so in two ways. First, the lower corporate tax rate reduces the value of the tax deductibility. And second, because LBO’s cannot circumvent the interest barrier rule through one of the escape clauses mentioned above. Their results confirmed their hypotheses. They found that the interest barrier rule in 2008 in Germany reduced the value of the tax shield by 35.1 percent, which affected more than 75 percent of the LBO’s. Therefore they concluded that the implemented interest barrier in Germany led to a significant decline in tax shields in large German buyouts. It also increased the willingness of PE firms to relocate to countries with more favorable tax conditions.

2.4 Swedish interest deduction limitation rules 2009 and 2013

After Sweden discovered that the Swedish general anti-avoidance legislation was not applicable to certain tax planning schemes with interest expenses, Sweden felt the need to come up with interest deduction limitation rules to prevent this kind of tax avoidance (Cohrs, 2013). Like Germany, Sweden is a civil law country with high-tax jurisdiction. Therefore, profit shifting also was a problem in Sweden and tax was avoided at a large scale. To prevent this, Swedish government introduced interest deduction limitation rules as of 1 January 2009. However, the rules received an enormous amount of critique since they included two exceptions which made sure that non-abusive transactions would not be affected (Cohrs, 2013). Also, both the Swedish tax authorities and the Government concluded that the introduced rules still offer sufficient opportunities to circumvent the high-tax jurisdiction. As a result amendments were made as of 1 January 2013. These amendments restricted the initial rules even more. Generally, Swedish tax law allows for full deduction of interest expenses. The TCR’s however, prohibit deduction of interest payments to affiliated companies. According to Swedish law, companies are seen as affiliated if a company, directly or indirectly, has a considerable influence over another company, or if the companies are under common control (Cohrs, 2013).

However, as in the German tax reform there are a few escape clauses. First, interest expenses on intra-group debt can still be deducted if the corresponding interest income is taxed with at least 10 percent in the residence state of the beneficial owner under the condition that it was his only income. The amendments as of 1 January 2013 state that even if companies cannot fulfil this 10 percent test, they can still deduct their interest expenses when the beneficial owner is subject to Swedish yield tax or a comparable tax, under the condition that during the tax year the average interest rate of the debt has

(11)

11

not exceeded 250 percent of the average state bond interest rate of the previous year (Cohrs, 2013). Second, there is a business purpose test in case the 10 percent rule is not applicable. Intra-group interest expenses can still be deductible if these expenses are mainly based on legitimate business reasons. However, this test only applies to beneficial owners that are resident within the European Economic Area (EEA) or in a state that has a double tax treaty with Sweden covering all income by which the beneficial owner is covered (Cohrs, 2013).

2.5 Interest barrier rules for tax deductibility in the rest of Europe

Germany and Sweden are not the only two countries in Europe which adopted interest barrier rules to avoid excessive debt financing. However, most countries use different mixtures of the various TCR’s. Most of the EU countries do not accept a deduction of the interest payments attributable to the excessive debt. Examples of these countries are Belgium, Czech Republic, Denmark, France, Hungary, Lithuania, Poland, Portugal, Slovenia and the United Kingdom (Brosens, 2004). Other European countries see the interest payments as some sort of hidden profit distribution and treat the interest payments as dividends. Examples are Austria, Belgium, Germany, Ireland, Italy, Luxembourg and Spain (Brosens, 2004). However, the earnings stripping rule recently adopted in Germany is rarely used. Most European countries still make use of the fixed ratio approach. France for example, introduced a TCR in 2006 which states that interest payments on debt exceeding the debt-to-equity ratio of 1.5:1 are not deductible (Zielke, 2014). For the Netherlands this threshold is 3:1 and the amount of excess debt has to be more than €500,000. Also Denmark and Belgium have similar TCR’s. Another form of TCR is implemented by the UK. They replaced their TCR in 2004 by the transfer pricing rules of the Organisation for Economic Cooperation and Development (OECD) (Zielke, 2014). The OECD consists of 34 countries and was founded to stimulate economic progress and world trade (Bartelsman and Beetsma, 2000). This

organization created transfer pricing guidelines for multinational enterprises, since governments want to avoid profit shifting of multinational firms out of their jurisdiction. The guidelines of the OECD help countries maintain the arm’s length principle on international transactions of multinational firms (Choe and Hyde, 2004). The arm’s length principle implies that prices should be the same as they would have been if the parties of the transaction were independent of each other. Because of this principle, transfer prices between multinationals are usually close to market value. The guidelines further state that firms cannot misuse transfer pricing to benefit from low-tax jurisdictions, and firms can receive penalties when transactions are not in accordance with the arm’s length principle (Choe and Hyde, 2004).

(12)

12

they enforced a law whereby firms cannot deduct interest payments if the net interest expenses exceed 30 percent of the earnings before interest, taxes, depreciation and amortization (EBITDA) of the firm. Yet, there are still some European countries without TCR’s like Austria, Finland and Poland.

2.6 Evidence of thin capitalization rules

There has been done some research on the effects of introducing thin capitalization rules. A study of German multinationals by Buettner et al. (2012) found that TCR’s tend to reduce the tax incentive to use internal debt, which is the amount of debt owed to lenders within the country. Also, the results of the study showed that more internal debt is used in countries with high-tax jurisdiction and when there are high costs of borrowing external debt, which is debt owed to foreign creditors. Moreover, Buettner et al. (2012) found that despite their results that TCR’s encourage firms to finance with external debt, the total debt-to-equity ratio declines after a TCR is implemented. Since Germany and Sweden are both countries with a high-tax jurisdiction and rely primarily on internal debt as the major source of debt financing, these results suggest that the implemented TCR’s will reduce the amount of internal debt for German or Swedish PE firms LBO’s. As a result of the lower leverage percentage in LBO’s the number of LBO’s will possibly decrease. The same is suggested by Knauer and Sommer (2012), but is still not researched in practice. This paper will test this by using the following hypothesis:

H1: The interest barrier rule leads to a significant decline in LBO’s.

Also, Dreßler and Scheuering (2012) researched the effect of the newly introduced interest barrier rule in Germany relative to the old “Fixed Ratio Approach”. In addition to Buettner et al. (2012) their

regressions showed that the new earnings stripping rule made firms reduce their debt-to-equity ratios and their net interest payments. These results suggest that firms actually adjust their capital structure as a result of the recently implemented interest barrier. This study will test this for LBO’s in Germany and Sweden using the following hypothesis:

H2: The interest barrier rule leads to a significant decline in the leverage percentage of LBO’s.

In the next section will be discussed in detail how these hypotheses will be tested and which models will be used to derive the results.

(13)

13

3. Methodology

To analyze the causal effects of the implemented interest barrier rule we want to compare the number of LBO’s before and after the law change in Germany and Sweden. We will do this by relying on a difference-in-differences (DID) approach. In a DID estimation the effect of a reform is measured by comparing the outcome for a treatment group (which is affected by the reform) with a control group both before and after a reform. In this case, the law change. According to Bertrand et al. (2004) standard errors in the DID model often understate the standard deviation of the treatment effects, which lead to overestimated t-statistics and significance levels. By collapsing the data into pre- and post- periods consistent standard errors are produced. Furthermore, two requirements have to be fulfilled for a valid DID estimation. Firstly, the treatment and the control group should have a common trend in the absence of the treatment. We control for this by including surrounding countries of Germany and Sweden as the control group in the DID model. As a result, the change in number of LBO’s will be controlled for omitted variables such as macroeconomic conditions of the country. Any bias caused by variables common to Germany/Sweden and the surrounding countries are implicitly controlled for, even when these variables are unobserved. Secondly, a valid DID analysis states that the treatment has to be exogenous. This means that there is no selection possible into the treatment and the control group. We believe that this condition is fulfilled by the formation of the treatment and control group.

As mentioned above, a valid DID regression requires two groups: a treatment group and a control group. Since this analysis will be done both for Germany and Sweden we have two treatment- and two control groups. The control group for Germany will consist of Western European countries. These countries are chosen because they have a parallel trend with Germany before the law change and they do not use an earnings stripping rule. These countries do use various forms of TCR’s which are explained in the previous section. However, according to Dreßler and Scheuering (2012) there is a significant change between these old rules and the newly introduced earnings stripping rule. By using these countries as the control group we can research if this new rule changes the number and leverage percentage of LBO’s.

For Sweden the control group will consist of Northern European Countries. Finland has no TCR’s, while Norway just implemented a TCR in 2014. Denmark uses a fixed ratio approach as explained in the previous section, which differs from the introduced interest barriers in Sweden. These countries have a common trend with Sweden before the law change. Also, we believe that the difference in TCR’s should lead to less noise and therefore give credible results. The macroeconomic conditions in the Southern- and Eastern European countries are less comparable to these conditions in Germany and Sweden.

(14)

14

Including these countries will lead to more noise, therefore they are excluded from the regression. By including Western- and Northern European countries as the control groups with a common trend, we assume there are no other changes over time.

We will test the hypotheses created in the previous section, by making use of the DID analyzes. The first hypothesis will be tested by the following DID model:

(1) Number of LBO’si,t = β0 + β1Xit + β2Gi + β3Dt + εit

In this model the number of LBO’s is a count variable. Furthermore, β0 is a constant, Gi is a dummy

indicating 1 for the treatment group (Germany or Sweden), Dt is a dummy indicating 1 for years after the

law change and Xit is the treatment effect Gi x Dt. Also, I is an index across countries and t is an index

across years. The variable of interest β1 is the difference-in-differences estimator. The DID estimator is

measured by the following formula:

(2) β1 = (Ȳtreatment, after - Ȳtreatment, before) – (Ȳcontrol, after - Ȳcontrol, before) = ΔȲtreatment - ΔȲcontrol

The DID estimator measures the difference in the number of LBO’s changes between the treatment group (Germany or Sweden) and the control group (Western Europe or Northern Europe) as a result of the implemented interest barrier. To avoid biased results due to firm specific effects we estimate equation (1) in differences between the period after the law change and before. The equation we estimate is given by (3). The dependent variable is now the change in the average number of LBO’s between the years before and after the law change. Because the regression is estimated in differences we automatically control for time-invariant unobserved factors between the treatment and control group.

(3) ΔNumber of LBO’si = β0 + β1ΔXi + β2ΔGi + β3ΔD + Δεi

This equation (3) will be estimated twice for both Germany and Sweden as the dependent variable. Also, we want to research the change in leverage in LBO’s by comparing the leverage ratio in LBO’s before and after the law change in Germany and Sweden. The leverage percentage is defined by the amount of debt per LBO divided by the amount of the total deal value. With the leverage percentage as the dependent variable we will test the second hypothesis. We will do so using the following DID model:

(15)

15

(4) Leverage Percentagei,t = β0 + β1Xit + β2Gi + β3Dt + Country fixed effectsi + Time fixed effectst + εit

Here, the dependent variable is the leverage percentage of the LBO for which t = 1 if it contains an LBO before the law change and t = 2 if it contains an LBO after the law change. The rest of the variables are interpreted the same as in the first regression. Again the DID estimator β1, described in the second

formula, is the variable of interest.

Also for Sweden we avoid biased results due to firm specific effects by estimating equation (3) in differences. This equation is given by (5). The dependent variable is now the change in the average leverage percentage between the years before and after the law change. Time-invariant unobserved factors are automatically controlled for.

(5) ΔLeverage Percentagei = β0 + β1ΔXi + β2ΔGi + β3ΔD + Country fixed effectsi + Δεi

4. Data and descriptive statistics

4.1 Data

Data for this thesis is obtained from the Thomson One database. Since this thesis concentrates on the PE industry the dataset contains data on all LBO’s and secondary buyouts done in Western- and Northern Europe in the period 2002-2014. A secondary buyout is a type of LBO in which the PE firm sells its investment in a company to another PE firm, thereby ending its involvement with the company. This is usually done when the selling PE firm already realized significant profit from the investment or when the buying PE firm can add more value to the company than the selling PE firm. The dataset contains

information on the value of the deal and the amount of equity used in the transaction. The amount of debt is derived by deducting the amount of equity from the value of the deal. The leverage percentage is calculated by dividing the debt value of the LBO by the deal value.

The period 2002-2014 is chosen to compare the 6 years after the changed tax law in Germany as of 1 January 2008 with the 6 years prior. Therefore, we drop the year in which the law change took place (2008) when estimating the equations. Otherwise we compare 7 years after with 6 years prior, which causes problems when performing the regressions. Also, the sample period captures the third PE boom (2003-2007) and the financial crisis and its consequences. The dataset contains a total of 5,130 LBO’s and

(16)

16

secondary buyouts. 973 of the LBO’s were performed in Germany and are subject to German tax

legislation. The other 4,157 LBO’s are financed in the rest of Western Europe. It should be noted that not all details on the LBO’s are known. Some LBO’s in the dataset have not disclosed the equity- and debt value of the deal, which leads to less observations for the regressions on the leverage percentage. Still, we believe that the average leverage percentages should give a good estimate of the true value. The sample period 2002-2014 is also used for Sweden. However, in Sweden they introduced interest deduction limitation rules as of 1 January 2009, which were restricted even more since 1 January 2013 (Cohrs, 2013). Therefore, the years 2009-2014 will be compared with the years 2003-2008. The year 2002 will be dropped from this regression when estimating the equations. The dataset contains a total of 1,037 LBO’s and secondary buy-outs. 437 of these LBO’s were performed in Sweden and are subject to Swedish tax legislation. The other 600 LBO’s are financed in the rest of Northern Europe. Also in this dataset not all equity- and debt values of the LBO’s are disclosed. It is possible this will cause less credible results for the DID estimation with the leverage percentage as the dependent variable.

4.2 Descriptive statistics Germany and Western Europe

In the previous section we described the treatment- and control groups. The control group for Germany consists of Western European countries. According to the Thomson One database these countries are Austria, Belgium, Channel Islands, France, Gibraltar, Greenland, Ireland, Liechtenstein, Luxemburg, Monaco, the Netherlands, Switzerland and the UK. Germany is excluded from this group, since it is already in the treatment group. Tables 1a and 1b give an overview of the descriptive statistics for the treatment- and control group before and after the changed law.

Table 1a shows that in the full sample there were approximately 244 LBO’s per year in the period 2002-2007. For Germany about 82 LBO’s per year and approximately 406 LBO’s per year in the rest of Western Europe. The average yearly deal value of a LBO is for Germany about €29M lower than for the rest of Western Europe in the period before the law change. Also, the LBO’s in Germany are financed with relatively less debt in both periods. On average 87.47 percent of the LBO’s before the law change are financed with debt in Germany. For the control group this is 84.43 percent. In the post-treatment period these numbers are 89.50 percent and 81.43 percent respectively. This means that German LBO’s are more highly levered since the law change, while other Western European LBO’s decreased in leverage. Furthermore, we plotted the number of the LBO’s of the control group and treatment group next to each other to discover whether there is a common trend in the period before the law change. As stated before, without a common trend DID estimations are not valid. The graph is presented in figure 1.

(17)

17

As expected, the number of LBO’s is increasing for both groups in the PE boom from 2003-2007 and declining since the financial crisis started around 2007. Whether this decline is because of the

implemented interest barrier rule cannot be derived from this figure. This will be analyzed in the results section. It is clear that the treatment- and control group are moving coherently in the pre-treatment period, despite that the numbers are absolute. Obviously, there are more LBO’s in the rest of Western Europe than in Germany alone. Still, the treatment and the control group have the same pattern. Increasing up to 2007, declining from 2007-2009 and then slowly recovering. Therefore, the common trend condition is satisfied. More details on the yearly number of LBO’s and leverage percentages of the treatment- and control group can be found in table A1, presented in the appendix.

Table 1a: Pre-treatment descriptive statistics of Germany and Western Europe

Summary statistics for the full sample (Germany and Western Europe), treatment group (Germany) and the control group (Western Europe) for the years 2002-2007. Minimum, 25th and 75th percentile, median, maximum, mean and standard deviation. The leverage% is calculated by dividing the average debt value by the average deal value. With the exception of the number of LBO's all numbers are yearly averages. Deal value, equity value, and debt value are in millions of euro's.

Full sample

Variable Min P25 Median P75 Max Mean S.D.

Number of LBO's 38 85 152 427 549 244 190 Deal Value 39.291 56.809 68.995 125.337 182.900 89.957 50.297 Equity Value 2.228 5.883 7.890 14.050 24.890 10.647 7.553 Debt Value 33.138 43.804 62.604 110.892 168.460 79.310 49.823 Leverage% 61.48% 78.94% 90.76% 94.64% 96.08% 85.95% 11.34% Treatment group

Number of LBO's 38 53 85 103 126 82 33 Deal Value 39.291 50.598 65.547 74.939 156.584 75.418 41.738 Equity Value 2.228 2.938 5.883 6.800 24.586 8.053 8.302 Debt Value 33.138 39.238 55.014 72.000 149.784 67.365 42.826 Leverage% 61.48% 84.34% 93.63% 95.66% 96.08% 87.47% 13.49% Control group

Number of LBO's 178 394 427 463 549 406 124

Deal Value 47.140 63.020 82.405 169.110 182.900 104.497 57.587 Equity Value 7.170 8.610 12.168 14.440 24.890 13.241 6.359 Debt Value 33.480 54.410 66.375 158.434 168.460 91.256 57.312 Leverage% 71.02% 73.55% 88.10% 92.10% 93.69% 84.43% 9.76%

(18)

18

Table 1b: Post-treatment descriptive statistics of Germany and Western Europe

Summary statistics for the full sample (Germany and Western Europe), treatment group (Germany) and the control group (Western Europe) for the years 2008-2014. Minimum, 25th and 75th percentile, median, maximum, mean and standard deviation. The leverage% is calculated by dividing the average debt value by the average deal value. With the exception of the number of LBO's all numbers are yearly averages. Deal value, equity value, and debt value are in millions of euro's.

Full sample

Number of LBO's 155 223 247 293 303 246 49

(19)

19

Figure 1: Evolution of the number of LBO’s for Germany and Western Europe

4.3 Descriptive statistics Sweden and Northern Europe

The same DID estimations in the methodology section will be performed for Sweden and Northern Europe. Sweden is the treatment group, since they changed the law as of 1 January 2009 and made amendments as of 1 January 2013. The control group in this case is Northern Europe. According to Thomson One the Northern European countries are Denmark, Faroe Islands, Finland, Iceland and Norway. Sweden is excluded from this group, since it is included in the treatment group. Tables 2a and 2b give an overview of the descriptive statistics for Sweden and the rest of Northern Europe. On average, there were about 44 LBO’s per year for the whole group in the pre-treatment period. 38 LBO’s for

Sweden and 50 for the rest of Northern Europe. Furthermore, the average deal value is approximately €7.5M higher for Sweden. The average percentage of leverage that is used to finance the LBO is around 67.4 percent for the whole group. The leverage percentage used in the treatment group is on average around 3 percentage points higher than the percentage used in the control group. After the law change this difference increased to almost 23 percent. This increase is not conform expectations, but is

(20)

20

consistent with the findings of German LBO’s. The standard deviations for the leverage percentage variable are however quite high in both periods.

Table 2a: Pre-treatment descriptive statistics of Sweden and Northern Europe

Summary statistics for the full sample (Sweden and Northern Europe), treatment group (Germany) and the control group (Northern Europe) for the years 2002-2008. Minimum, 25th and 75th percentile, median, maximum, mean and standard deviation. The

leverage% is calculated by dividing the average debt value by the average deal value. With the exception of the number of LBO's all numbers are yearly averages. Deal value, equity value, and debt value are in millions of euro's.

Full sample

Number of LBO's 8 33 52 70 83 50 24

Source: Thomson One database, years 2002-2014, own calculations.

Also, like the graph of German and Western European LBO’s, we plotted the number of LBO’s of the control group and treatment group next to each other to discover whether there is a common trend in the pre-treatment period. This graph is presented in figure 2. The common trend between Sweden and Northern Europe is clear. The pattern however changes a little from Germany and Western Europe in the post-treatment period. The number of LBO’s are increasing up to 2007-2008 and decreasing in the financial crisis. This is similar to Germany and Western Europe. However, the recovery after the financial

(21)

21

crisis is different. The number of LBO’s is upward trending from 2009-2011, but declines thereafter. This also shows why it is better to have two different control groups for both countries. The difference between Germany and Sweden from 2011 onwards means that both treatment groups do not have the same trend in the post-treatment period. This is probably caused by the difference in implemented tax deductibility laws in 2008 and 2009. The control groups of both countries do have the same common trend with the treatment group and therefore we can continue performing the regressions. Using the control group for Germany also for Sweden would be comparing apples and oranges. The same is true for using Sweden’s control group for Germany. More details on the yearly number of LBO’s and leverage percentages for both groups are presented in table A2 in the appendix.

Table 2b: Post-treatment descriptive statistics of Sweden and Northern Europe

Summary statistics for the full sample (Sweden and Northern Europe), treatment group (Germany) and the control group (Northern Europe) for the years 2009-2014. Minimum, 25th and 75th percentile, median, maximum, mean and standard deviation. The

leverage% is calculated by dividing the average debt value by the average deal value. With the exception of the number of LBO's all numbers are yearly averages. Deal value, equity value, and debt value are in millions of euro's.

Full sample

Number of LBO's 24 28 45 55 56 42 14

(22)

22

Figure 2: Evolution of the number of LBO’s for Sweden and Northern Europe

5. Results

This section will describe the results of the DID regressions. Section 5.1 will present the regressions for Germany. Then the same regressions will be performed for Sweden in section 5.2. The regressions are generated using Stata.

5.1 Difference-in-Differences regressions Germany

First, hypothesis H1 is tested for Germany. It examines whether the recently implemented earnings

stripping rule will lead to a decline in the number of LBO’s. The results of the regression are presented in table 3. Then the second hypothesis is tested for Germany. H2 examines if the earnings stripping rule

leads to a decrease in the leverage percentage of LBO’s. Results of this regression are displayed in table 4.

(23)

23

Table 3: Average number of LBO's per country before and after the changed tax law.

LBO's by country Difference, WE GER GER-WE

Variable (i) (ii) (iii)

1. Avg. number of LBO's before, 406*** 82*** -325***

all available observations (51) (13) (53)

2. Avg. number of LBO's after, 236*** 66*** -171***

3. Change in mean LBO's -170*** -16 154***

(33) (11) (35)

Notes: Heteroscedasticity-robust standard errors are shown in parentheses. The sample consists of all

LBO’s in the period 2002-2014, with the exception of the year 2008. This year is omitted to compare the 6 years after the law change with the 6 years prior. Treatment group Germany is indicated as GER. Control group Western Europe is indicated as WE. Germany is excluded from the control group. Stars (***/**/*) indicate significance at the 1%/5%/10% levels.

Table 4: Average leverage percentage per country before and after the changed tax law.

Leverage percentage by country

Difference,

WE GER GER-WE

1. Leverage percentage before, 84.43%*** 87.47%*** 3.04%

all available observations (3.98%) (5.51%) (6.80%)

2. Leverage percentage after, 80.12%*** 89.33%*** 9.20%

3. Change in mean leverage -4.30% 1.86% 6.16%

percentage (4.54%) (7.46%) (8.73%)

LBO’s in the period 2002-2014, with the exception of the year 2008. This year is omitted to compare the 6 years after the law change with the 6 years prior. Treatment group Germany is indicated as GER. Control group Western Europe is indicated as WE. Germany is excluded from the control group. Stars (***/**/*) indicate significance at the 1%/5%/10% levels.

(24)

24

Table 3, row 1 shows that Germany has an average of approximately 82 LBO’s per year in the period before the earnings stripping rule. The rest of Western Europe has about 406 LBO’s per year in the same period. As expected both numbers are lower in the period after the law change, since the law change took place in the middle of the financial crisis. Also, the period before the law change contains the third private equity boom period. The post-treatment period in row 2 shows that Germany now has an average of approximately 66 LBO’s per year, against 236 LBO’s for the control group. All averages are significant at 1 percent. Row 3 presents the change in the average number of LBO’s per year between the period before and after the law change. For the control group, a given year in the post-treatment period has on average 170 less LBO’s than a given year in the pre-treatment period. This number is significant at the 1 percent level. Germany has a decrease of approximately 16 LBO’s on average between both periods. However, this number is not significant. Column (iii) in row 3 gives the DID estimator and measures the difference in the average number of LBO’s between Germany and the rest of Western Europe. The difference is about 154 LBO’s, which is significant at 1 percent. However, the numbers are absolute so in this regression the DID estimator is not very

meaningful. Therefore, we convert the numbers to percentages. The average decrease of LBO’s for Germany is then -16/82 = 19.51 percent, against -170/406 = 41.87 percent for the control group. So between both periods the average number of LBO’s for the control group decreases stronger than for the treatment group. As said earlier, this difference is significant at 1 percent. This shows that there is no evidence that German LBO’s are significantly decreasing because of the implemented earnings stripping rule. Therefore H1 is rejected for Germany. The results

even suggest that German LBO’s are decreasing less relative to the rest of Western Europe. However, the difference for the treatment group is not significant, whereas the difference for the control group is significant. These findings mean that the law change has no effect on the treatment group, but does have an effect on the control group. This evidence is not convincing and therefore it cannot be concluded that the changed law in Germany results in less LBO’s for the control group.

Table 4 presents the results of the second regression. The first row shows that in the pre-treatment period German LBO’s were financed with 87.47 percent debt on average. Other Western European LBO’s were financed with slightly less debt, namely 84.43 percent. After 2008 this percentage decreased for the control group to 80.12 percent debt. German PE companies, in contradiction with H2, increased the leverage percentage of LBO’s in this period

(25)

25

to 89.33 percent. This is shown in row 2 of table 4. These numbers are significant at 1 percent. This means that the control group lowered the leverage percentage with 4.30 percent, while Germany increased the leverage ratio with 1.86 percent. However, these differences shown in row 3 are not significant. Also, the DID estimator which equals 6.16 percent is not significant. We expected this number to be negative and significant. Despite the suggestion of Dreβler and Scheuering (2012) that the new earnings stripping rule let firms reduce their debt-to-equity ratios, the results do not support this hypothesis. Therefore, also the second hypothesis is rejected for Germany.

5.2 Difference-in-Differences regressions Sweden

This section examines the same hypotheses as tested in section 5.1. However, this time the treatment group is Sweden and the control group consists of other Northern European countries. Sweden did not implement an earnings stripping rule like Germany but introduced more standard thin capitalization rules which prohibit the tax deduction of interest payments to affiliated companies. Results of the DID regressions are presented in table 5 and table 6.

Table 5: Average number of LBO's per country before and after the changed tax law.

LBO's by country Difference, NE SWE SWE-NE

1. Avg. number of LBO's before, 57*** 42*** -15

2. Avg. number of LBO's after, 42*** 29*** -14*

3. Change in mean LBO's -14 -13* 1

(10) (7) (12)

LBO’s in the period 2002-2014, with the exception of the year 2002. This year is omitted to compare the 6 years after the law change with the 6 years prior. Treatment group Sweden is indicated as SWE. Control group Northern Europe is indicated as NE. Sweden is excluded from the control group. Stars (***/**/*) indicate significance at the 1%/5%/10% levels.

(26)

26

Table 6: Average leverage percentage per country before and after the changed tax law.

Leverage percentage by country

Difference,

NE SWE SWE-NE

1. Leverage percentage before, 68.50%*** 80.42%*** 11.92%

2. Leverage percentage after, 57.43%*** 88.16%*** 30.73%*

3. Change in mean leverage -11.07% 7.75% 18.82%

percentage (29.65%) (13.31%) (32.50%)

LBO’s in the period 2002-2014, with the exception of the year 2002. This year is omitted to compare the 6 years after the law change with the 6 years prior. Also the years 2013 and 2014 are omitted, since details on the LBO’s of the treatment group are not available. Therefore, we now compare 4 years after the law change with the 4 years prior. Treatment group Sweden is indicated as SWE. Control group Northern Europe is indicated as NE. Sweden is excluded from the control group. Stars (***/**/*) indicate significance at the 1%/5%/10% levels.

The first row in table 5 shows that Sweden has approximately 42 LBO’s in a given year in the pre-treatment period. For the control group, this number is 15 LBO’s higher. The second row gives the average number of LBO’s in the post-treatment period. A given year in this period contains about 29 LBO’s for Sweden. The control group has in this period on average 42 LBO’s per year. All averages in the pre- and post-treatment period are significant at 1 percent. As expected, the number of LBO’s are decreasing for both the treatment- and the control group. The decrease of 13 LBO’s for Sweden is significant at 10 percent. The control group decrease of 14 LBO’s is not significant. Converting this absolute numbers to percentages gives a -14/57 = 24.56 percent decrease for the control group and a -13/42 = 30.95 percent decrease for the treatment group. This is consistent with the first hypothesis. Also, the significance for the difference in the treatment group and the non-significance for the

difference in the control group means that the treatment has an effect on the treated group and not on the control group. Although the DID estimator (1) is not significant, this empirical evidence suggests that the implemented law change in Sweden leads to significantly less LBO’s relative to countries where there are no interest barrier rules. Therefore, H1 is not entirely rejected for Sweden. It seems that the interest

deduction limitation rules in Sweden do have an effect on the number of LBO’s, whereas the earnings stripping rule in Germany has not.

(27)

27

Table 6 shows the leverage percentages for Swedish and other Northern Europe LBO’s in both periods. Since the details of Swedish LBO’s for the years 2013 and 2014 are not disclosed, we had to omit these years from the regression. Unfortunately this is exactly the period in which Sweden made

amendments which restricted the interest barrier rules even more. This could lead to less credible results. The leverage percentage for Swedish LBO’s in the pre-treatment period was 80.42 percent. For the control group this was significantly less, namely 68.50 percent. In the period after the law change this leverage percentage gives 88.16 percent for the treatment group and 57.43 percent for the control group. All leverage percentages are significant at the 1 percent level. In contradiction with the second hypothesis, Sweden finances LBO’s with 7.75 percent more debt in the period after the law change. However, this percentage is not significant. The control group finances LBO’s with 11.07 percent less debt in the post-treatment period. Also this number is not significant. Moreover, the DID estimator (18.82 percent) is positive and not significant, which leads to the rejection of the second hypothesis for Sweden. Surprisingly, the leverage percentages for Swedish and German LBO’s are both increasing between the two time periods, while the leverage percentages for both control groups LBO’s are decreasing. Although these findings are not significant, it will be interesting to research this further.

6. Robustness checks

The tax laws in Germany and Sweden were changed at the same time as the financial crisis was active, which could lead to biased results in section 5. In this section will be controlled for the financial crisis by performing additional DID regressions on the leverage percentage of LBO’s for Germany and Sweden. The years of the financial crisis will thereby be added as a control variable. Moreover, DID regressions will be performed with the average number of LBO’s per year and the average deal value per LBO as the dependent variable.

6.1 Additional Difference-in-Differences regressions Germany

The DID regression for Germany will be performed with the leverage percentage as the dependent variable. Thereby the following DID regression is used:

(6) Leverage percentagei,t = β0 + β1LawChangei,t + β2Germanyi + β3LawChange*Germanyi,t

(28)

28

The same DID regressions are performed with the number of LBO’s and the average deal value per LBO as dependent variables. For all three regressions the same data for Western European countries, as in the DID regressions for Germany in section 5.1, is used. Results of the DID regressions are displayed in table 7.

On average the leverage percentage for the whole sample is 84.43 percent for Western European LBO’s excluding Germany. The results suggest that LBO’s done after the implemented earnings stripping rule in 2008 are financed with approximately 3 percent less debt relative to LBO’s pre-2008, although this number is not significant. German LBO’s are about 3 percent more levered than other Western European countries in the whole sample period. After the changed tax law this number even increased. Although it contradicts with the second hypothesis, this is consistent with the results in section 5.1. Furthermore, it seems that the financial crisis did not have much impact on the leverage percentage. However, as in the DID regression results presented in table 4 in section 5.1, all explanatory variables for the first DID regression are not significant. Also, the German leverage percentage of LBO’s is rising after the changed law, which does not confirm the second hypothesis. Therefore, H2 is rejected.

The second regression shows that the law change led to a an average decline of approximately 162 LBO’s per year for the whole sample. This is significant at 1 percent. Surprisingly, this decline is not directly caused by the financial crisis, since in the years 2007-2009 there were on average 11 LBO’s more than in other years. However, this number is not significant. Also, the year 2007 was at the end of the third private equity boom and table A1 in the appendix shows that the number of LBO’s in this year was relatively high. The results of the second regression in table 7 further suggest that Germany had on average 325 less LBO’s then the rest of Western Europe, which is obvious since numbers are absolute. However, after the law change German LBO’s decreased less than other Western European LBO’s. This is significant at 1 percent and in contradiction with the first hypothesis. It seems that the decline in LBO’s is an indirect result of the financial crisis. Table A1 shows that the number of LBO’s started to decline from the end of 2008 onwards. Since the results do not suggest that this decline is attributable to the changed law, and even suggest that German LBO’s are decreasing significantly less because of the treatment, the first hypothesis is rejected.

The third regression shows the development of the average LBO deal value in millions of euro’s. Over the whole sample the average LBO was €70.7M for Germany and €99.8M for other Western European countries, although the difference is not significant. The average deal value declined after the changed law with €31.3M, which effected especially non-German LBO’s. However, also these numbers

(29)

29

are not significant. Moreover, LBO’s in the financial crisis were on average €28.2M higher than in other years, which is remarkable although not significant at the 10 percent level.

Table 7: Effect of earnings stripping rule on German leverage percentage, number of LBO's and deal value

Results of the DID regressions for Germany. The first dependent variable is the percentage of leverage used in the capital structure of the LBO's. The second dependent variable is the number of LBOs per year. The third

dependent variable is the average deal value per LBO in millions of euro's. Law change is a dummy variable indicating one for LBO's that took place in the period after the law change and zero for the period before the law change. Germany is a dummy variable equaling one if it contains a German LBO and zero for the control group. Law change*Germany contains German LBO's in the post-treatment period. Financial crisis is a dummy variable indicating one for the years 2007-2009 and zero otherwise. The numbers contain yearly averages.

Heteroscedasticity-robust standard errors are shown in parentheses. Stars (***/**/*) indicate significance at the 1%/5%/10% levels.

(1) (2) (3)

Dependent variable: Leverage % No. Of LBO's Deal value

Law change -0.0299 -162*** -31.27 (0.0614) (53) (24.09) Germany 0.0304 -325*** -29.08 (0.0691 (53) (25.15) Law change*Germany 0.0503 148*** 15.10 (0.0916) (57) (31.75) Financial crisis -0.0003 11 28.23 (0.0619) (25) (22.31) Constant 0.8443*** 404*** 99.79*** (0.0416) (53) (22.14) Observations 26 26 26 R² 0.084 0.829 0.222

6.2 Additional Difference-in-Differences regressions Sweden

The same DID regression as presented in equation 6 will be performed for Sweden and its control group. Only now the variable ‘Amendments’ is added, which is a dummy variable equaling one for the years after the amendments were made (2013 and 2014). Also here the same regression will be conducted with the number of LBO’s and the average deal value as the dependent variable. The same dataset as in the DID regressions for Sweden in section 5.2 will be used. Results are presented in table 8.

(30)

30

in the sample period with the exclusion of Sweden is on average 61.94 percent. Swedish LBO’s are 2.86 percent more levered than the rest of Northern Europe. The introduction of the TCR led to an average decline of 8.05 percent in the leverage ratio of all Northern European LBO’s. However, Swedish LBO’s after the law change were financed with relatively more debt than other Northern European LBO’s. This contradicts the second hypothesis, but is in line with the results of Germany. Also remarkable is that LBO’s in the financial crisis are more levered than LBO’s in other years. The results are consistent with the results of the DID regression in table 6. However, all explanatory variables are not significant. The second hypothesis is rejected for Sweden, since the results suggest the opposite of H2 and all

explanatory variables are not significant.

The average number of LBO’s per year are displayed in the second column. Sweden has over the whole period about 12 LBO’s per year less than the rest of Northern Europe. The law change in 2009 led to a decrease of approximately 3 LBO’s per year on average for the whole sample. For Swedish LBO’s, this number is even higher. This in line with our first hypothesis and consistent with the results in table 5. However, this difference is not significant. In years of the financial crisis the number of LBO’s is higher relative to other years. Again, this is possibly attributable to the higher number of LBO’s in the year 2007 and the beginning of 2008, which meant the end of the third PE boom. This is shown in table A2 in the appendix. Although the lower number of Swedish LBO’s relative to other Northern European LBO’s after the law change is in line with H1, the first hypothesis is rejected since this difference is not significant.

The amendments as of 1 January 2013 led to even less LBO’s for both groups. But also here, the difference is not significant.

The third column shows the development of the average deal value of LBO’s in millions of euro’s. Swedish LBO’s are on average €7.6M higher in value than LBO’s in the rest of Northern Europe. After the changed tax law the average deal value increased for Swedish LBO’s, but declined for other Northern European LBO’s. However, all these differences are not significant. The financial crisis seems to have caused LBO’s of lower value in comparison with other years, but also this variable is not significant. The deal value of the years 2013 and 2014 is significantly higher than the years prior. This difference is not caused by the amendments, since the interaction variable Amendments*Sweden is not significant. It does indicate that the private equity industry is recovering from the financial crisis, since the LBO’s in 2013 and 2014 have higher deal values. The leverage percentage of these years cannot be derived, since the details of the deals after the amendments are not available.

(31)

31

Table 8: Effect of TCR on Swedish leverage percentage, number of LBO's and deal value

Results of the DID regressions for Sweden. The first dependent variable is the percentage of leverage used in the capital structure of the LBO's. The second dependent variable is the number of LBO's per year. The third

dependent variable is the average deal value per LBO in millions of euro's. Law change is a dummy variable indicating one for LBO's that took place in the period after the law change and zero for the period before the law change. Sweden is a dummy variable equaling one if it contains a Swedish LBO and zero for the control group. Law change*Sweden contain Swedish LBO's in the post-treatment period. Financial crisis is a dummy variable indicating one for the years 2007-2009 and zero otherwise. Amendments is a dummy variable equaling one for the years 2013-2014. Amendments*Sweden contain Swedish LBO's after amendments were made. The numbers contain yearly averages. In the first regression the years 2013-2014 are omitted, since details on Swedish LBO's are not available. Heteroscedasticity-robust standard errors are shown in parentheses. Stars (***/**/*) indicate significance at the 1%/5%/10% levels.

(1) (2) (3)

Dependent variable: Leverage % No. of LBO's Deal value

Law change -0.0805 -3 -27.13 (0.1957) (13) (26.35) Sweden 0.0286 -12 7.58 (0.1881) (12) (31.62) Law change*Sweden 0.2788 0 18.18 (0.2403) (16) (35.13) Financial crisis 0.1413 7 -0.68 (0.1043) (11) (17.81) Amendments - -11 113.28* (10) (62.34) Amendments*Sweden - -6 -75.32 (12) (91.97) Constant 0.6194*** 48*** 55.46** (0.1568) (9) (27.95) Observations 22 26 26 R² 0.154 0.277 0.239

7. Conclusion

In order to prevent profit shifting several countries have implemented thin capitalization rules to limit the amount of interest payments to consolidated tax groups. This especially effects leveraged buy-outs, since these deals are largely financed with debt. The aim of this thesis is to research which effects the introduction of interest barriers have on the capital structure of LBO’s and if the TCR’s lead to a decline in LBO’s. Difference-in-differences regressions are used to test the capital structure and the number of