T C G

Taxation
Andreas Oestreicher/Reinald Koch*
Taxation and Corporate Group Structure –
Evidence from a Sample of European
Multinationals**
A bstract
We empirically analyze the influence of tax considerations on the structure of investments of a parent company based in one EU member state that holds subsidiaries in
a different member state. We show that group taxation, deductibility of financing
expenses, or participation write-downs and additional taxes on intragroup dividends
may factor into the parent company’s decision on the structure of investments as tax
parameters. We find empirical evidence that a vertical structure with a pure holding
interposed is implemented more often if a domestic parent entity is required for the formation of a tax group, the semi-elasticities being 0.568 and –0.343.
JEL-Classifications: F23, H25, K34.
Keywords: Corporate Group Structure; Corporate Income Tax; Firm-Level Data; Multinationals.
1I ntroduction
One of the main issues that is frequently addressed in the empirical tax literature is the
impact of taxes on company decisions. Such analyses tend to be focused on investment,
location and financing decisions. But to date, the legal design of a given investment
program has received little attention. In this area of company decisions, only the impact
of corporation tax on the choice of legal form has been demonstrated on several occasions (see Goolsbee (1998); MacKie-Mason and Gordon (1997); Gravelle and Kotlikoff
(1989); Cragg, Harberger, and Mieszkowski (1967), following the theoretical paper by
* Andreas Oestreicher, Professor for Business Administration and Taxation, University of Göttingen, Platz der Göttinger Sieben 3, 37073 Göttingen, phone: +49 551 397308, fax: +49 551 397874, e-mail: andreas.oestreicher@
uni-goettingen.de. Reinald Koch, Senior research assistant, University of Göttingen, Platz der Göttinger Sieben 3,
37073 Göttingen, phone: +49 551 397306, fax: +49 551 397874, e-mail: [email protected].
** We are most grateful to two anonymous referees for their valuable suggestions. We also thank Alfons Weichenrieder, Michael Overesch, and Christoph Gust for their helpful advice, and the participants of the CESifo Conference on Corporate Taxes and Corporate Governance, the ZEW-Workshop “Empirische Forschung in der Betriebswirtschaftlichen Steuerlehre” and the Annual Congress of the IIPF 2010 for their useful remarks. The usual
disclaimer applies.
254
sbr 64 October 2012 254-280
Taxation
Harberger (1962)). However, the legal structure of multinational corporate groups has
received only scant analytical or empirical research on tax influences. Poppe (2007)
analyzes the overall structure of European multinational companies. Oestreicher and
Koch (2010) look at the tax motives for concluding a profit-and-loss transfer agreement
in German corporate groups. Mintz and Weichenrieder (2010), using German Bundesbank data, address the question of what determines the ownership chains of German
inbound and outbound FDI. Their focus is primarily on third country conduit entities,
compared to which the exploitation of the benefits of a country holding presents a perfect alternative. Wamser (2011) looks at the impact of different tax regimes for direct
and indirect shareholdings on foreign direct investment. Buettner, Riedel, and Runkel
(2011) analyze the consequences of group taxation on the extent of economic integration in German corporate groups.
Our objective in this paper is to analyze the tax impact on the legal structure of European local corporate subgroups, i.e., groups of companies that are located in the
same EU member state (host country) and controlled by a common parent located
in a different member state (home country). We analyze whether tax considerations
factor into the company’s decision between a horizontal structure, in which subgroup
companies in the host country are directly held by the foreign parent; a vertical structure with operative intermediary, in which subgroup companies in the host country
are indirectly held via an operative company in the host country; and a vertical structure with intermediate holding, in which subgroup companies in the host country are
indirectly held via a pure holding company in the host country. Hence, we perform
a cross-section analysis for the year 2003. Although such an analysis cannot rule out
the possibility that other determinants of group structure may also be correlated with
elements of the tax system, it makes it possible for us to compare the impact of several
tax parameters on the group structure of multinational companies across 25 EU member states. Moreover, by using cross-section analysis we can identify the relevant elements of the tax systems concerned. We can then study these influences in a longitudinal research design that makes high demands on the underlying data. We carry out
our investigation by using Update 125, February 2005, of the AMADEUS database
which includes shareholding information for one specific point in time. These restrictions on our data do not allow for application of a panel approach. Further, we doubt
whether a panel approach is appropriate for the underlying research question, since the
relevant tax provisions (e.g., group taxation regimes) have been subject to only small
changes across Europe in recent years.
Our analysis focuses exclusively on the structure of local subgroups and thus on the
choice between direct holding of the local companies by the foreign parent (horizontal
group structure) and interposing a domestic operating company or a local holding company (vertical group structure). In our view, it is important to distinguish between interposing an operative intermediary and an intermediate holding as both the economic
consequences (e.g., associated costs) and tax consequences of such interposition vary.
Second, we focus on cross-border direct investments between two EU countries, thus
enabling us to take greater account of the relevant tax provisions under the legislation of
the countries concerned. Third, we analyze the structure of local subgroups as a whole,
sbr 64 October 2012 254-280
255
A. Oestreicher/R. Koch
rather than individual ownership chains. We believe this approach is important, because
we expect the structure to depend on characteristics of the subgroup (e.g., size). To the
best of our knowledge our paper is the first to investigate empirically and in detail the
influence of taxes on the use of country holding companies in the European Union.
The research question that motivates this paper, the tax impact on the legal structure
of local subgroups, should be of interest from both a business and an economic perspective. If a group largely bases its decision regarding its structure on tax considerations, then other management decisions may also be affected, and these decisions may
have negative consequences for the group. At the very least, additional costs may be
incurred if, for tax reasons, companies deviate from what would otherwise offer the
most efficient group structure.
Management literature discusses extensively the specific advantages and disadvantages of different types of organizational structure (see, e.g., Markides and Williamson
(1996) and Hoskisson (1987)). Although internal and external structures need not correspond, the external structure may be designed to support the internal organization.
Thus, conformity of these structures may be desired for reasons of higher transparency
(see Kutschker and Schmid (2011) and Theisen (2000)). Therefore, we can assume that
the design of the corporate group is also affected by motives other than reducing taxation. Among these motives is the allocation of managers’ tasks and responsibilities,
which also depend on the legal structure under the regulations of labor and company
law. Hence, a vertical group structure can have a positive or negative impact on the development of the subordinated business if the expertise of its management is dominated
by that of the controlling company. An operative intermediary (designated an operative
holding or centralized multidivisional organizational structure (labeled “CM-form” in
the management literature)) could exercise operative control over a subordinated business, which may be desirable if the use of synergy potentials is of particular interest
(see Markides and Williamson (1996) for a detailed discussion of the usefulness of the
CM-form in this respect). In contrast, a vertical structure with an intermediate holding
focusing on strategic or financial control (“M-form”, “H-form”) should strengthen the
autonomy of a subordinated business, with the advantage that decisions can be made
more flexibly and closer to the market, and innovation is fostered. However, such a
structure is likely to be accompanied by higher monitoring costs (see, e.g., Kreikebaum,
Gilbert, and Reinhardt (2002); Scherm and Suess (2001)).
The group structure can also influence the risk situation of individual group companies. For example, for a vertical group structure the risk of insolvency faced by a
subsidiary can be affected by the risk associated with an interposed group company. If
we assume that these risks are taken into account in the decision-making process that
precedes business transactions, then the group structure can also have an impact in
this framework. But a purely horizontal group structure may raise the costs of selling
a business unit, since the constituent companies must be sold individually rather than
as holdings (Kreikebaum, Gilbert, and Reinhardt (2002). We may assume that preference for a horizontal or vertical legal group structure can also arise on operational
grounds.
256
sbr 64 October 2012 254-280
Taxation
The paper is organized as follows: In Section 2 we present a model for the tax and nontax impact that arises when a company opts for a horizontal or vertical group structure.
In Section 3 we look at some institutional details on the use of local parent companies
and give a short overview of the relevant tax legislation in the European Union. In
Section 4 we present the underlying database, describe our empirical approach used,
discuss the relevant variables, and perform an econometric identification. Section 5
concludes.
2 Tax I mpact
on
G roup Structure D ecisions : A M odel
Our analysis of tax impact on the structure of local subgroups starts by following a
simplified one-period model of a corporate group. We consider a group of companies
with a parent company P operating in the home country p and two foreign operative
subsidiaries S1 and S2 located in the host country s. Groups can opt for either a horizontal structure or a vertical architecture with either an operative intermediary or intermediate holding. We assume that the group structure affects both operating profits
and taxes. Therefore, we assume that a profit-maximizing group will choose the group
structure such that after-tax profit π* is maximized.
For a horizontal group structure, both subsidiaries are held directly by P. Here, S1,
S2, and P have pre-tax profits amounting to πS1, πS2, and πP. For simplification, we
assume in the following that tax profit equals pre-tax accounting profit π.
Ignoring taxes on distributions from S1 and S2 to P, we can determine the total aftertax profits of the horizontal group as
πh∗ = (πS1 + πS2) ∙ (1 − τS) + πP ∙ (1 − τP) + G ∙ gtss;fp ∙ τs + [(ES1 + ES2) ∙ expp;f + (WS1 + WS2) ∙ wdp;f ] ∙ τ p.
(1)
We define tax payments of S1 and S2 as pre-tax profits πS1 + πS2 minus the tax base
reduction G resulting from a possible application of group taxation in the host country
s. We multiply the result by the corporate tax rate τS. For a horizontal group structure,
application of group taxation in s requires that the tax law in s offers a group taxation regime for sister companies that are controlled by a common foreign parent (gtss;fp = 1).
We start to define tax payments of the parent company as pre-tax profit πP reduced by
potential write-downs on the book values of S1 and S2 (WS1 + WS2) and expenses
for financing S1 and S2 (ES1 + ES2). We multiply the result by the corporation tax
rate τ p. We take into account the respective income reductions if a write-down of foreign corporate investments (wdp;f = 1) and the deduction of expenses for financing
foreign subsidiaries (expp;f = 1) is permitted by the law applicable in the home country p. Because we disregard loss-carry forwards in our model, we assume that P earns
sufficient income to ensure full deductibility of write-downs and financing expenses
(πP > WS1 + WS2 + ES1 +ES2).
sbr 64 October 2012 254-280
257
A. Oestreicher/R. Koch
As one alternative to a horizontal structure, S1 and S2 can be held vertically, meaning that P controls S2 indirectly via S1. Structuring the group in a way such that S1
functions as an intermediary may bring with it in particular the advantage of enhancing the transparency and flexibility of the group structure (Scheffler (2004)). Greater
flexibility is achieved, since for a vertical group structure only the shares of S1 have to
be transferred for a possible sale of the subgroup. The transparency of a group structure
is enhanced if ownership rights and management rights are distributed accordingly.
Therefore, interposing S1 as an operative intermediary is especially useful if S1 exercises strategic and operating control in S2. Since there is the chance to better exploit
synergy advantages, we can expect to see this particular distribution of management
tasks more frequently if both companies belong to the same industry (see Kutschker
and Schmid (2011); Kreikebaum, Gilbert, and Reinhardt (2002)). This also holds true
if S1 dominates S2 in size and market power (see Scheffler (2004)). In contrast, a horizontal group structure appears to be more appropriate if the managements of S1 and
S2 operate independently. Apart from this argument, interposing an intermediary can
be useful as a way to structure a domestic subgroup, e.g., according to different lines of
business, which may be beneficial especially for a large local subgroup. However, a vertical group structure can be disadvantageous in the context of the risk situation of the
subgroup, because a subsidiary’s risk of insolvency can be affected by the risk associated
with an interposed group company (see Scheffler (2004)).
The operative advantages and disadvantages of a vertical group structure result in additional costs or income, and therefore affect the operating income of the group. Thus,
we assume that the operating profit of a vertically structured group differs from that of
a horizontally structured group in the (positive or negative) amount ∆πv(o). For simplification we define ∆πv(o) as net of tax and therefore ignore this amount when determining tax payments.
From a tax point of view, the consequence of using a vertical group structure is that
S1 and S2 can also be taxed as a group if the tax law in the host country requires a
domestic parent (gtss;dp = 1), i.e., group taxation is not applicable for sister companies
of a common foreign parent. Furthermore, S1 can deduct expenses for financing S2 if
debt is raised by the intermediary company (α = 1) and deduction is permitted by law
(exps;d = 1). Also, a write-down of the investment in S2 reduces taxable income in the
host country if deductible (wds;d = 1). Assuming full distribution of profits by S2, additional taxes may arise on dividend payments between S2 and S1 (divtaxS2). Hence,
divtaxS2 = max(πS2 − tS2; 0) ∙ divtaxs
(2)
where divtaxs denotes the tax rate for domestic intragroup dividend payments in s.
The starting point for the determination of tax payable by P is the pre-tax profit πP.
We then generally determine the tax payments as in the previous case. However, writedown is limited to the investment in S1 and expenses for financing S2 can be deducted only if debt is raised at the parent company level (α = 0). Again, we ignore
258
sbr 64 October 2012 254-280
Taxation
taxes on distributions by S1 to P. We also assume sufficient income to ensure full
deductibility of write-downs and financing expenses (we note that this assumption is
more restrictive in this case than it is for the horizontal structure, since it requires both
S1 and P to be profitable). Further, we assume that the corporate group chooses α in
a way that the after-tax profits are maximized. Thus, the total after-tax profits of the
vertical group amount to
π∗v(o) = ∆πv(o) + (πS1 + πS2) · (1−τs) + πP · (1 − τ p)−divtaxS2 + G · gtss;dp · τS + WS2 · wds;d · τS + [ES1 · expp;f
exps;d · τs
​
​
+ WS1 · wdp;f ] · τ p + ES2 · max exp
p;f · τ p
(3)
[
A profit-maximizing group will then opt for a vertical group structure if this structure
yields a higher after-tax profit compared to a horizontal group structure, i.e., π∗v(o)− πh∗
> 0. When we compare the total after-tax profits arising for a vertical group structure
using an operative intermediary company with those of a horizontal group structure
(Equation (3) – Equation (1)) we find that
π∗v(o) – πh∗ = ∆πv(o) – divtaxS2 + G ∙ (gtss;dp – gtss;fp) ∙ τs + W2 ∙ (wds;d ∙ τs – wdp;f ∙ τ p)
​ + ES2 ∙ max _______________
[​ (4)
exps;d ∙ τs – expp;f ∙ τ p
0
Based on Equation (4) we derive the following propositions on the choice between a
vertical group structure with operative intermediary and a horizontal structure.
Proposition 1: Ceteris paribus, we expect a profit-maximizing group to opt for a vertical
group structure with operative intermediary more frequently if:
(A) A domestic parent is required for implementing a group tax regime in the host country
(gtss;dp – gtss;fp = 1). However, this is the case only if S2 suffers losses.
(B) Investment write-downs and/or financing expenses can be deducted only in the host
country, or such deductions become effective in the host country at a higher rate (wds;d ∙
τs – wdp;f ∙ τ p and/or exps;d ∙ τs – expp;f ∙ τ p > 0). However, the first of these tax
advantages applies only if S2 suffers losses.
(C) Domestic dividends are not subject to additional taxation in the host country s (divtaxS2 = 0).
(D) This structure is beneficial from a business point of view (∆πv(o) > 0), particularly if
both operative companies are engaged in the same line of business, the operating companies
differ in size, and the subgroup in the host country is large.
According to the assumption formulated above we assume S1 is profitable.
sbr 64 October 2012 254-280
259
A. Oestreicher/R. Koch
As a second alternative, an additional domestic company (H) that is functioning as
an intermediate holding can be interposed between P and S1/S2. Interposing a pure
holding company is also assumed to affect both the operating profits and the tax payments of the group companies. From a business point of view, interposing a pure holding company might be used to combine the advantages of greater transparency and
flexibility without centralizing operative management. The latter could cause negative
incentive effects for the subsidiaries’ management. This outcome could be expected
especially if S1 and S2 are equal in both size and market power (Scheffler (2004)).
Instead, corporate management often uses pure holding companies to centralize the
management of investments and financing activities or to centralize strategic management. However, interposing a pure holding company might induce additional implementation or administration costs as well as costs for an extended intragroup reporting system (see Scheffler (2004)). Therefore, we can expect to see pure holding
companies more frequently in large subgroups. Because of these advantages and disadvantages of an intermediate holding company, we assume that the operating income
of a group with an intermediate holding differs from that of a horizontally structured
group in the (positive or negative) amount ∆πv(h). Again, we assume that ∆πv(h) is
defined net of tax.
From a tax perspective, interposing a pure holding company has the effect that financial expenses and investment write-downs can become tax effective in the host country
only if a group tax regime is available (gtss;dp = 1). This is due to the fact that in
functioning as a pure holding company, the latter generates no own taxable income.
Furthermore, the corporation can deduct financing expenses and investment writedowns in the host country s now with expenses referring to both operative subsidiaries.
But taxes on dividend distributions in s now arise for the distributions from both S1
and S2 to H (see Equation (5)).
divtaxS1 + divtaxS2 = [max(πS1 − tS1; 0)
(5)
+ max(πS2 − tS2;0)] ∙ divtaxs
Thus, the total after-tax profit for a vertical group structure with domestic holding
company computes as
π∗v(h) = ∆πv(h) + (πS1 + πS2) ∙ (1 − τs) + πP ∙ (1 − τ p) − divtaxS1 (6)
− divtaxS2 + G ∙ gtss;dp ∙ τS + (WS1 + WS2) ∙ wds;dp
[​ exps;d ∙ gtss;dp ∙ τs
∙ gtss;dp ∙ τs + (ES1 + ES2) ∙ max ____________
​
exp ∙ τ p;f
p
To compare the after-tax profit of this alternative with that of a horizontal structure
and a vertical structure without holding, we first consider the case in which the host
country does not provide for a group tax regime (gtss;dp = 0). In this case, the difference in after-tax profits between this structure and a horizontal structure (π∗v(h) – πh∗)
computes as
260
sbr 64 October 2012 254-280
Taxation
π∗v(h) – πh∗ = ∆πv(h) − (WS1 + WS2) ∙ wdp;f ∙ τ p − divtaxS1 − divtaxS2. (7)
Based on Equation (7) we derive the following propositions on the advantages of interposing a holding company in comparison to a horizontal structure, if the host country
does not provide for a group tax regime.
Proposition 2a: If the host country does not provide for a group tax regime, then interpos-
ing a holding company is beneficial compared to using a horizontal group structure if:
(A) It does not prevent a potential write-down of the investments in S1 and S2 at the level
of P (wdp;f = 0).
(B) Dividend distributions in the host country do not result in an additional tax burden
(divtaxS1 = divtaxS2 = 0).
(C)It has advantages from a business point of view (∆πv(h)). Such advantages are expected,
in particular, if S1 and S2 do not differ in size and the subgroup in the host country is
large.
When we compare the after-tax profits for the two alternatives of a vertical group structure under the assumption that no group tax regime is applicable in the host country,
we find
(π∗v(h) – π∗v(o)) = (∆πv(h) – ∆πv(o)) – WS2 ∙ wds;d ∙ τs [​ (8)
ES2 ∙ (exps;d ∙ τs – expp;f ∙ τ p)
– WS1 ∙ wdp;f ∙ τ p – divtaxS1 – max _________________________
​
0
Proposition 3a: If the host country does not provide for a group tax regime, then interpos-
ing a holding company is beneficial in comparison to interposing an operative intermediary
if
(A) Investment write-downs are deductible neither in the home country nor in the host
country (wdp;f = 0 and wds;d = 0).
(B) A deduction of financing expenses in the host country is not beneficial compared to a
deduction in the home country (exps;d ∙ τs – expp;f ∙ τ p ≤ 0).
(C) Dividend distributions in the host country do not result in an additional tax burden
(divtaxS1 = divtaxS2 = 0).
(D)An intermediate holding is beneficial from a business point of view (∆πv(h) − ∆πv(o)
> 0). Such advantages are expected particularly where S1 and S2 do not differ in size,
S1 and S2 operate in different industries and the subgroup in the host country is large.
Different tax consequences arise for a vertical structure with intermediate holding if
the host country s provides for a group tax regime in its domestic tax system (at least in
the case of a domestic parent; gtss;dp = 1). Under this assumption, comparison of the
after-tax profits for the vertical structure using an intermediate holding with those that
arise for a horizontal structure (π∗v(h) – πh∗) or for a vertical structure with operative
intermediary (π∗v(h) – π∗v(o)) results in (9) and (10).
sbr 64 October 2012 254-280
261
A. Oestreicher/R. Koch
(π∗v(h) – πh∗) = ∆πv(h) + G ∙ (gtss;dp – gtss;fp) ∙ τs (9)
+ (WS1 + WS2) ∙ (wds;d ∙ τs – wdp;f ∙ τ p) – divtaxS1 – divtaxS2
[​ exp
∙ τ − exp
∙τ
s;d
s
p;f
p
+ (ES1 + ES2) ∙ max _______________
​
0 π∗v(h) – π∗v(o) = ∆πv(h) – ∆πv(o) + divtaxS1 + WS1 ∙ (wds;d ∙ τs [​ (10)
exps;d ∙ τs − expp;f ∙ τ p
− wdp;f ∙ τ p) + ES1 ∙ max _______________
​
0 Under the assumption that the host country does allow for group taxation, tax payments arising for the vertical structure with an intermediate holding are very similar
to those that arise for the vertical structure with operative intermediary. Small differences may stem from higher dividend tax in s, as both the profits of S1 and S2 are
distributed to a domestic parent. Besides, also ES1 and WS1 can be deducted in the
host country (instead of p) if permitted by law. Based on these results we derive the following propositions:
Proposition 2b: If the host country provides for a group tax regime, interposing a holding
company is beneficial in comparison to a horizontal structure, under the conditions mentioned in Propositions 1(A) to 1(C) (tax factors) and 2a (C) (non-tax factors).
Proposition 3b: If the host country provides for a group tax regime, interposing a holding
company is beneficial in comparison to interposing an operative intermediary if:
(A) Dividend distributions in the host country do not result in an additional tax burden
(divtaxS1 = 0).
(B) Investment write-downs and/or financing expenses can be deducted only in the host
country, or such deductions become effective in the host country at a higher rate (wds;d ∙
τs – wdp;f ∙ τ p and/or exps;d ∙ τs – expp;f ∙ τ p > 0).
(C) An intermediate holding is beneficial from a business point of view (∆πv(h) − ∆πv(o)
> 0). For possible explanations see Proposition 3a(C).
3I nstitutional D etails
3.1 C orporation Tax R ates
D ividend Payments
and Taxation of D omestic and
C ross - border Corporate profits are subject to corporation tax and – if applicable – surcharges and/or
local profit taxes. Taking these three components into account, the following tax rates
on corporate profits apply in the EU member states.
262
Tax law provisions presented in this section refer to the year 2003, i.e., the year of our empirical group structure
information.
sbr 64 October 2012 254-280
Taxation
Table 1: Tax rate on corporate profits for 2003 (τ)
Austria
Great Britain
30
Malta
Greece
35
Netherlands
15
Hungary
18
Poland
Czech Rep.
31
Ireland
Denmark
30
Italy
Finland
29
Latvia
Belgium
Cyprus
34
33.99
France
34.33
Lithuania
Germany
39.58
Luxembourg
35
34.5
27
12.5
Portugal
33
38.25
Slovakia
25
19
Slovenia
25
15
Spain
35
Sweden
28
30.38
Source: KPMG (2006).
Profit distributions between domestic corporations are subject to an additional tax burden (divtaxs) if there is neither a full tax credit for the corporation tax underlying the
dividend nor a full exemption of the dividend income. According to Table 2, this is the
case in Belgium and France, where only 95 percent of domestic dividends are exempted.
Table 2: Double taxation relief for domestic inter-company dividends
Exemption 100 per cent (ε = 1)
Austria, Cyprus, Czech Republic, Denmark, Finland, Germany, Greece,
Ireland, Italy, Hungary, Latvia, Lithuania, Luxembourg, Netherlands,
Poland, Portugal, Slovakia, Slovenia, Sweden, United Kingdom
Exemption 95 per cent (ε = 0.95)
Belgium, France1
Credit method (ε = 1)
Malta, Spain
1
: In France, non-deductibility is limited to expenses actually paid.
Source: Endres et al. (2007); Kesti (2003).
An additional tax burden arises if domestic corporate dividends are subject to withholding tax. This also holds true if the withholding tax can be credited against the corporation tax of the receiving company. In this case, tax payments have to be financed for
an interim period of up to 12 months. As summarized in Table 3, in 2003 five member
states collected withholding taxes on dividend payments to domestic corporations.
Table 3: Withholding taxes on domestic income distributions (τ wh;s)
Withholding taxes, fully creditable (φ = 1)
Spain (15 percent), Germany (20 percent), Poland
(15 percent)
Withholding taxes, partially creditable (φ < 1) Czech Republic (15 percent, half creditable)
Withholding taxes, non- creditable (φ = 0)
Slovakia
No withholding taxes on domestic income Austria, Belgium, Cyprus, Denmark, Finland, France, Greece,
distributions
Hungary, Ireland, Italy, Luxembourg, Lithuania, Latvia, Malta,
Netherlands, Portugal, Sweden, Slovenia, United Kingdom
Source: Kesti (2003).
sbr 64 October 2012 254-280
263
A. Oestreicher/R. Koch
As we have to take into account both the additional tax at the level of the dividend
receiving company, due to only incomplete exemption (ε < 1), and the timing or final
effect of a withholding tax, and the fact that creditable withholding taxes (φ ∙ τ wh;s)
have to be financed for a period of nine months at a model rate of six percent, we determine the tax burden on dividends paid to a domestic corporation thus
divtaxs = (1 − ε) ∙ τS + (1 − φ) ∙ τ wh;s + φ ∙ τ wh;s ∙ 0,06 ∙ 9/12.
(11)
In Table 4 we summarize the resulting tax burdens on dividends paid to domestic corporations in accordance with (11).
Table 4: Tax burden on dividends paid to domestic corporations (divtaxS )
Austria
0
Belgium
1.6995
Cyprus
Czech Rep.
0
7.8375
Great Britain
0
Malta
0
Greece
0
Netherlands
0
Hungary
0
Poland
Ireland
0
Portugal
0.675
0
Denmark
0
Italy
0
Slovakia
15
Finland
0
Latvia
0
Slovenia
0
France
1.7165
Lithuania
0
Spain
Germany
0.9495
Luxembourg
0
Sweden
0.675
0
Source: Own calculations.
Table 5: Deductibility of expenses relating to domestic and foreign shareholdings
Expenses deductible
domestic shareholdings (exps;d = 1)
Belgium, Cyprus, Denmark, Finland, France, Great Britain, Hungary,
Ireland, Italy, Latvia, Lithuania, Malta, Netherlands, Poland,
Portugal, Slovenia, Spain, Sweden
foreign shareholdings (expp;f = 1)
Belgium, Cyprus, Denmark, Finland, Germany, Great Britain,
Greece, Hungary, Ireland, Italy, Latvia, Lithuania, Luxembourg ,
Malta, Netherlands, Poland, Portugal, Slovakia, Spain, Sweden
Expenses non-deductible
domestic shareholdings (exps;d = 0)
Austria, Czech Republic, Germany, Greece, Slovakia
foreign shareholdings (expp;f = 0)
Austria, Czech Republic, Slovenia
Expenses deductible only under
certain conditions
domestic shareholdings (exps;d = 1)
Luxembourg
foreign shareholdings (expp;f = 1)
France
Source: Endres et al. (2007); Kesti (2003).
264
sbr 64 October 2012 254-280
Taxation
3.2 Deductibility of Related Expenses and Extraordinary Write-down of Investments
According to the Parent-Subsidiary Directive, EU member states that apply the exemption
method for taxing domestic and foreign dividends may disallow shareholders the possibility of deducting expenses associated with financing their investments. Whereas two member states regard five percent of domestic dividends as non-deductible expense but accept
the deduction of actual expenses (see Table 2), in five (three) member states financing costs
relating to domestic (foreign) shareholdings are non-deductible (see Table 5). Two member
states allow deduction of these expenses only subject to certain conditions.
Table 6: Extraordinary write-down of long-term securities and investments
Extraordinary write-down is mandatory
(wds;d = 1 and/or wdp; f = 1)
Austria (domestic investments)*, Czech Republic, Hungary,
Spain***
Extraordinary write-down is prohibited or
not tax effective(wds;d = 0 and/or wdp;f = 0)
Austria (foreign investments), Belgium, Cyprus, Denmark,
Finland, France**, Germany, Greece, Ireland, Italy, Latvia,
Lithuania, Luxemburg, Malta, Netherlands, Poland, Portugal,
Slovakia, Slovenia, Sweden, United Kingdom
* Only if the decrease in value seems to be permanent.
** For investments in participating shares, write-down is not tax-effective but treated as long-term capital
loss which might be offset against long-term capital gains.
*** Limitations for unquoted securities.
Source: Endres et al. (2007).
Only three member states, Spain, Hungary, and the Czech Republic, recognize writedown of long-term domestic and foreign investments. Austria allows resident corporations limited extraordinary write-down of long-term domestic securities and investments (see Table 6).
3.3 Availabilit y
and S cope of
G roup Taxation R egimes
Of the 24 EU member states considered in this paper 17 countries provide group taxation regimes (see Table 7 ). For the remaining seven member states, Belgium, Greece,
Lithuania, Slovakia, Czech Republic, Hungary and Italy, there was no group taxation
regime available in 2003.
Table 7: Group taxation regimes in the European Union
Group taxation regime available
Austria, Cyprus, Denmark, Finland, France, Germany,
Ireland, Latvia, Luxembourg, Malta, Netherlands, Poland,
Portugal, Sweden, Slovenia, Spain, United Kingdom
No group taxation regime available
Belgium, Czech Republic, Greece, Hungary, Italy, Lithuania,
Slovakia
Source: Endres et al. (2007).
sbr 64 October 2012 254-280
265
A. Oestreicher/R. Koch
Establishing a tax group has its merits for the companies involved, since it allows
offset of profits and losses thereby reducing the tax liabilities of the entire group. It
can also be advantageous with regard to potential double taxation of dividend income as well as withholding taxes on distributed profits between members of a tax
group.
Table 8: Territorial scope
Foreign permanent establishments of foreign
parent companies acceptable (group consists
only of resident subsidiaries)
gtss;fp = 1
gtss;dp = 1
Cyprus, Ireland*, Latvia**, Malta, Sweden,
United Kingdom
Domestic entity required (group consists of
a domestic parent or the resident permanent
establishment of the foreign parent and
resident subsidiaries)
gtss;fp = 0
gtss;dp = 1
Austria, Denmark, Finland, France,
Germany, Luxembourg, Netherlands,
Spain
Domestic holding required (group consists of
a domestic parent and resident subsidiaries)
gtss;fp = 0
gtss;dp = 1
Poland, Portugal, Slovenia
* Only with respect to companies resident in the EU or a contracting party to the EEA agreement.
** Only with respect to companies resident in the EU or in a country with which a tax treaty is in effect.
Source: Endres et al. (2007).
The territorial scope of a group taxation regime differs widely across the European
Union. For the analysis conducted in this paper, it is particularly relevant, whether
a domestic parent entity (or intermediary) is required for establishing a tax group, or
whether profits and losses can also be consolidated across subsidiaries of a common
foreign parent. If foreign companies abroad are not eligible to act as head company
of a national tax group, a domestic entity is inevitable. This approach is found in six
EU member states (see Table 8). All other group taxation regimes require that a domestic entity (Poland, Portugal, Slovenia) or at least a domestic permanent establishment of a foreign company heads the domestic tax group (Austria, Denmark, Finland,
France, Ger­many, Luxembourg, Netherlands, Spain). However, as we cannot observe
permanent establishments in our data, we assume for our empirical analysis that group
taxation is not available in these member states in case of horizontal domestic subgroup
structures (i.e., gtss;fp = 0).
4E mpirical A nalysis
4.1 Dataset
The empirical data was collected from the AMADEUS database of the provider Bureau
van Dijk. The version used (update 125) includes about 6.15 m companies in the Eu-
266
sbr 64 October 2012 254-280
Taxation
ropean Union. Of these, 2.25 m companies provided information on their controlling
entities. Based on this shareholding information, corporate groups and subgroups are
re-created using the following definition: A corporate group consists only of Europeanbased incorporated enterprises. These must be directly or indirectly majority owned by
another European company. This of course does not apply for the ultimate owner entity. Once the ownership chain leaves the European Union it is severed meaning that a
European corporate group placing a conduit entity for investments in a non-European
country is treated here as two separate groups.
Domestic subgroups are defined as a group of two or more companies located in the
same EU member state and controlled by the same foreign company (which does
not necessarily have to be the ultimate parent). For simplification purposes, direct
and indirect shareholdings are not combined and we disregard indirect shares held
through a foreign company. We also ignore subgroups in which less than two of
its companies are operative, i.e., have a NACE code other than 7415 (“management
activities of holding companies”). In the course of our regression analysis, group
structure variables based on this shareholding information (see subsection 4.2 for details) are matched with balance sheet data and other company data for the subgroups
(also taken from AMADEUS) as well as country controls and tax data. Shareholding information in AMADEUS refers to the last available point in time (which is in
our case the annual accounts for 2003 for most companies). We therefore employ
company and tax data referring to this year. We exclude subgroups for which the
provided balance sheet and company data does not allow computation of the non-tax
variables.
According to the above definition we identified 3,336 domestic subgroups with a nonresident parent company . The geographic structures of these subgroups and their (immediate) foreign parent companies are summarized in Table 9, and the size of these
subgroups is given in Table 10.
The ownership chains are cut due to the fact that the AMADEUS database covers Europe only.
Subgroups with only one operative company could not be structured vertically with operative intermediary and
are therefore excluded.
Country data (GDP per capita, GDP) is taken from Eurostat. For the sources of the tax data see Section 3.
For deduction of financing expenses and participation write-downs we had to rely on information from Endres
et al. (2007) referring mainly to the year 2005. As far as possible we used the European Tax Handbook 2003 in
order to check for differences regarding 2003; this was not possible for all countries, however. As these provisions
regarding the deductibility of financing expenses and participation write-down are usually subject to only minor
changes over time, we do not believe that this should bias our results.
For Estonia we identified a total of 54 subgroups. However, investments in Estonia were disregarded since the Estonian tax system differs from that of all other EU member states and does not fit the assumptions of the model
presented in section 2.
sbr 64 October 2012 254-280
267
268
1
1
40
113
13
37
19
3
ES
4
3
1
8
2
18
FI
IT
1
37
132
14
21
63
98
3
37
11
GB
1
5
4
2
1
8
1
GR
8
1
7
2
2
2
16
1
16
HU
IE
3
1
1
2
1
12
53
2
4
5
21
12
IT
2
2
LU
1
4
3
LV
6
24
11
2
1
6
14
8
1
NL
1
5
15
6
2
11
34
2
5
3
PL
27
385
1
27
175
132
53
346
111
66
8
1
6
718
33
152
687
63
1
184
3
20
33
1
10
57
1
1
5
161
1
4
44
10
2
85
10
1
1
132
14
1
26
5
39
1
6
1
3
8
13
7
PT
204
33
1
12
21
30
1
72
11
22
1
SE
69
3,336
2
285
9
4
829
21
1
127
42
1
3
321
513
82
93
237
418
2
3
274
All figures in one column refer to the same location of the domestic subgroup, whereas all numbers in one row refer to the same (immediate) foreign
parent country.
SK
10
7
18
4
79
SE
1
1
1
31
2
49
7
1
7
23
PT
PL
1
183
5
LU
5
NL
LT
57
2
132
11
38
23
101
153
10
FR
1
1
18
5
7
12
2
1
DK
1
6
24
38
1
5
11
9
19
DE
IE
2
2
6
1
8
CZ
HU
2
33
GB
1
98
FR
GR
3
3
1
FI
ES
29
8
17
2
BE
DK
DE
CZ
CY
BE
AT
AT
A. Oestreicher/R. Koch
Table 9: Number of domestic subgroups
sbr 64 October 2012 254-280
Taxation
Table 10: Size of domestic subgroups
No. of companies
(local sub-group)
2
3
4
5
6-10
11-20
21-50
>50
Total
Frequency
1,287
656
364
243
473
198
81
34
3,336
4.2 A pproach
4.2.1 D efinition
of
G roup Structure Variables
In order to analyze tax impact on the structure of local subgroups operating in the
host country s, we distinguish horizontal structures (subsidiaries in the host country
are held directly by the parent resident in the home country) from vertical structures
(subsidiaries in the host country are controlled indirectly through an intermediary in
the host country). In the latter case, subsidiaries held through an operative intermediary are distinguished from those held through an intermediate holding company.
Analyzing the structure of these subgroups as a whole involves the problem that the
number of possible structures increases exponentially with the number of companies included. This makes it necessary to distil the group structure into a set of variables that
may function as dependent variables in the course of the following regression analysis.
As a starting point to comparing vertical and horizontal subgroup structures we define
the variable VERTICALITYrate as the share of the group companies held by a domestic intermediary company. As one domestic company is necessarily held directly by a
foreign company, VERTICALITYrate is determined as follows:
Number of group companies in state s
held
by a domestic intermediary company
VERTICALITYrate = ​ ___________________________
​.
Number of group companies in state s – 1
​ _______________________________
​
(12)
For illustration purposes an example is given in Figure 1. In this example five out of the
nine subsidiaries in the host country are held domestically. VERTICALITYrate therefore amounts to 5/8 = 0.625.
Although VERTICALITYrate does not account for the entire group structure, high
values for this index indicate a more vertical layout and can therefore function as a
measure for the decision to choose a horizontal or a vertical group structure. Furthermore, this measure takes into account the fact that the tax consequences of different
subgroup structures are due to the existence of an intermediary company rather than
to the depth of the group structure. However, it does not distinguish between a verti
The deduction of one in the denominator ensures that VERTICALITYrate can take any value in a range between
zero and one.
sbr 64 October 2012 254-280
269
A. Oestreicher/R. Koch
cal arrangement with an operative intermediary company and a vertical one with an
intermediate holding company, as would be necessary to test the hypotheses outlined
in Section 2. Therefore, OPERATIVErate and HOLDINGrate are determined similarly to VERTICALITYrate, but taking into account only those parts of the subgroup
that are controlled (a) directly by P (and do not function as intermediate holding) or
through an operative intermediary company (OPERATIVErate: S2 – S4, S7 and S9 )
or (b) directly by P (and do not function as an operative intermediary) or through an
intermediate holding company (HOLDINGrate: S1 – S2, S4 – S6 and S810 ). In the
example given in Figure 1, OPERATIVErate takes the value of 0.5 (= 2/4), but HOLDINGrate amounts to 0.6 (= 3/5). For the purpose of identifying holding companies we
apply the NACE industry code in AMADEUS. Thereby, companies with an industry
code of 7415 are regarded as holding companies.
Figure 1: Determination of VERTICALITYrate
P
S 2(h)
S 1(h)
S5
S 3(o)
S 4(o)
S7
S6
S9
S8
To picture the decision between an operative intermediary and an intermediate holding company, we compute OP_vs_HOLDrate. This variable is defined as the share of
indirectly held subsidiaries in the host country that are controlled by an operative intermediary as compared to the total number of indirectly held subsidiaries in the host
country (in the example: 2/5 = 0.4). High values indicate structures in which operative
companies can be observed more frequently, whereas low values stand for group structures in which holding companies dominate.
Holdings without subsidiaries are regarded as part of a horizontal group structure and therefore included in this
definition.
10 Operative companies without subsidiaries are regarded as part of a horizontal group structure and therefore included in this definition.
270
sbr 64 October 2012 254-280
Taxation
4.2.2 R egression M odels
In the following, we apply a set of regression models in order to analyze each of the
three different decisions regarding the structure of local subgroups as discussed theoretically in Section 2. In particular this concerns the choices between
(a)A horizontal structure and a vertical structure with operative intermediary (Specifications (1) to (3) in Table 11 and Specification (1) in Table 12 testing Proposition 1).
(b)A horizontal structure and a vertical structure with intermediate holding (Specifications
(4) to (6) in Table 11 and Specification (2) in Table 12 testing Propositions 2a and 2b).
(c)A vertical structure with operative intermediary and a vertical structure with intermediate holding (Specifications (7) to (9) in Table 11 and Specification (3) in Table 12
testing Propositions 3a and 3b).
Each decision is analyzed on the basis of four regression models. Three of these models relate to the subgroup structure as a whole and are based on the group structure
variables as described in Section 4.2.1. We consider a count data model to be more
appropriate (preferred specification) as this type of model can take account of the nonnegative nature of the dependent variables (Specifications (1), (4) and (7) in Table 11).
Since we cannot rely directly on the rate variables defined in Section 4.2.1, we use the
number of indirectly held subsidiaries in a subgroup (nominator of the rate variables) as
dependent variables (OPERATIVEcount: number of companies held indirectly through
a domestic operative intermediary; HOLDINGcount and OP_vs_HOLDcount: number
of companies held indirectly through a domestic intermediate holding11). The denominator of the rate variables enters the regression models as exposure variable12. In order
to control for unobserved heterogeneity at group level we include group fixed effects.
To establish the appropriate type of count data model we test for overdispersion as proposed by Cameron and Trivedi (2010, 575-576) and compare the number of actual
and predicted zero counts in our sample, in order to observe potential zero-inflation13.
Due to the existence of overdispersion or a better fit in terms of predicted zeros and the
non-existence of zero-inflation we choose a negative binomial model14 for the dependent variables HOLDINGcount and OP_vs_HOLDcount. A poisson model is applied to
OPERATIVEcount.
In addition, for each of the three decisions we use a linear model with group fixed effects to
our rate-variables, and a count data model without consideration of group fixed effects: In
both cases this procedure serves as a robustness check. Specifications (2), (5) and (8) refer to
the linear model. Specifications (3), (6) and (9) apply the count data model without group
11 For OP_vs_HOLDcount we count, similarly to OPERATIVEcount, the number of subsidiaries held by an operative intermediary. Here, however, the total number of indirectly held subsidiaries serves as exposure variable.
12 This ensures that the regression model effectively also looks at the share of indirectly held subsidiaries but using
a different distribution.
13 The results of these tests will be provided to interested readers upon request.
14 To be more precise, we use the NB2-model described by Cameron and Trivedi (2010), 577-579.
sbr 64 October 2012 254-280
271
A. Oestreicher/R. Koch
fixed effects. The latter specification is designed to investigate the consequences of the substantially reduced sample that results when group fixed effects are taken into account15.
None of the specifications considered so far enables us to look at the determinants of
intragroup differences in the subgroup structure (why are some subsidiaries held directly and others not?). In order to model certain non-tax parameters company specifically, we therefore estimate, as a fourth specification, logit models with group fixed
effects including each subsidiary as a separate observation. This approach has the advantage that non-tax impacts may be modeled more accurately. However, no variance
in the tax parameters occurs for observations belonging to the same subgroup. The dependent variable takes the value of one if the respective subsidiary is held indirectly via
a domestic operative (OPERATIVEdummy, OP_vs_HOLDdummy) or holding company
(HOLDINGdummy)16. The results of the logit models are reported in Table 12.
4.2.3 Parameters
To analyze the different group-structure decisions under consideration, a set of tax and
non-tax parameters enters the regression equation as right-hand variables. According to
Proposition 1, we expect vertical group structures to appear less frequently if domestic
dividends are subject to an additional tax burden (divtaxs). But implementing a domestic intermediary company can be beneficial if a domestic parent is required for (a) use
of a group tax system and/or (b) a write-down on the investments or the deduction of
financial expenses related to these investments. gtss reflects the influence of the group
taxation requirements on the structure of the subgroups. Such influence is deemed to
exist if the application of the group taxation regime is restricted to domestic parent entities (gtss;dp − gtss;fp = 1). In this case, gtss takes the value of one, otherwise zero
is assigned. Advantages with regard to the deduction of participation write-downs and
­f inancing expenses arise from a vertical group structure if these deductions are offered
by tax law in the host country (wds;d = 1 and exps;d = 1) and are not allowed by tax
law in the home country (wdp;f = 0 and expp;f = 0). Apart from this, a vertical structure can also be beneficial from a tax perspective if these opportunities are offered by
tax law in both the host and home country, but a deduction in the host country would
become effective at a higher rate. We account for both effects by employing the independent variables WDs;d, WDp;f and exp.WDs;d (WDp;f) is determined by multiplying the dummy variable reflecting the deductibility of participation write-down wds;d
(wdp;f) with the relevant tax rate τs (τ p). For the deductibility of financing expenses we
determine exps;d ∙ τs and expp;f ∙ τP accordingly, but combine them into a single variable exp which is determined as the positive difference of both terms17.
15 The inclusion of group fixed effects brings with it the disadvantage that groups with only one observation (one local
subgroup) are disregarded. This reduces our sample to 505 groups with 1,723 observations for OP_vs_HOLDrate,
430 groups with 1,417 observations for OPERATIVErate, and 134 groups with 457 observations for HOLDINGrate.
16 Regressions employing OPERATIVEdummy and OP_vs_HOLDdummy differ with regard to the underlying observations. For OPERATIVEdummy we disregard all subsidiaries held by a pure holding company in the host country. Moreover, for OP_vs_HOLDdummy we exclude subsidiaries directly held by a foreign parent.
17 WDs;d and WDp;f are not combined accordingly since it is not the difference between these two terms that
should factor into the group structure.
272
sbr 64 October 2012 254-280
Taxation
In the case of a vertical group structure with interposed holding company, deductions
for financing expenses and participation write-downs in the host country become tax
effective only if the host country offers a group tax regime (see Hypothesis 2a/2b and
3a/3b). We therefore include the respective variables WDs;d, WDp;f and exp as plain
variables and in interaction with (1 − gtss;dp) in this case18. The nominal tax rate τs
serves as explanatory variable, as we expect tax considerations to be more influential if
the subgroup in the host country faces a higher tax burden.
We assume the size of the effects arising from applying a group taxation regime and
write-down of participations to be dependent on the profit situation of the domestic
subsidiaries. LOSSsubgroup therefore measures the share of companies included in the
subgroup that reported an overall loss in the five previous years (1999 and 2003).
Control variables are employed in order to capture non-tax impacts on the group structure choice. To this end, the number of companies forming part of the domestic subgroup
(SIZEsubgroup) and the overall group (SIZEgroup) are included (both in terms of their natural logarithms), whereby the first variable may function as an indicator of the organizational necessity of a vertical group structure. To control for other legal and economic
influences, the GDP per capita is included both for the country of the ultimate group parent (GDPpcu) and the subgroup (GDPpcs). It may be assumed, for example, that multinationals resident or subgroups operating in highly developed EU member states (high
values for GDP per capita) make use of tax planning strategies to a greater extent than
others. In order to capture further influences related to the size of the economy we also
control for the GDP of the host country in terms of its natural logarithm (ln(GDPs)).
We include variables for differences in the size of the subgroup companies (Varsize) and
their industry (Varindustry). In accordance with Hypothesis 1 we expect to observe vertical structures more frequently if the subsidiaries are engaged in the same line of business (small values for Varindustry). As we assume that in case of a vertical structure using
an operative intermediary small subsidiaries are usually controlled by larger ones, we
expect this structure to be applied more frequently if subsidiaries differ in size (high
values for Varsize, see Proposition 1). Contrastingly, horizontal structures and vertical
structures with intermediate holding company are rather expected if subsidiaries are of
equal size (see Proposition 2a-3b).19
Concerning the logit models, intra-subgroup heterogeneity is modeled by employing subsidiary-specific definitions of LOSSsubgroup (= LOSSsubsidiary) and Varindustry
(= Sameindustry). Also company size is defined with reference to the respective subsidiary (SIZEsubsidiary) instead of the variance in the size of the subgroup companies (Varsize). SIZE subsidiary takes the value of one if operating revenue of the subsidiary exceeds
the average value of the subgroup.
18 We use 1 − gtss;dp instead of gtss;dp for the purpose of interaction in order to avoid high correlation with the
plain variables.
19 Descriptive statistics will be provided to interested readers upon request.
sbr 64 October 2012 254-280
273
274
–
+
+
+
–
+
+
+
–
Linear1
Yes
–.010
(.014)
.052*
(.028)
–.002
(.001)
Poisson2
Yes
.043
(.030)
.091
(.061)
–.005**
(.002)
.079**
(.033)
.120*
(.067)
–.410***
(.111)
1,417
–.009
(.026)
.072**
(.034)
–.018**
(.009)
.009***
(.002)
.001
(.007)
.083***
(.024)
.099**
(.039)
–.156***
(.059)
2,963
.078
–.060***
(.018)
.087***
(.022)
–.002
(.004)
.003***
(.001)
–.006
(.004)
.004***
(.001)
OPERATIVErate
OPERATIVEcount
.008***
(.002)
(2)
(1)
–.022
(.022)
.008***
(.003)
–.003***
(.001)
.016
(.039)
.076***
(.020)
–.052***
(.010)
.035
(.028)
.003
(.057)
–.371***
(.111)
2,958
–.000
(.001)
.006***
(.002)
Poisson3
No
.047
(.063)
.160
(.142)
.003***
(.001)
OPERATIVEcount
(3)
+
+
–
–
–
+
–
+
+
–
–.074
(.124)
.572**
(.230)
–.223
(.370)
457
.319**
(.133)
.226*
(.123)
NBinom. 2
Yes
–.066
(.259)
.568*
(.297)
–.006
(.009)
–.091
(.738)
.011
(.008)
–.692***
(.171)
–.013
(.340)
–.383***
(.077)
–.061
(.051)
.012
(.010)
HOLDINGcount
(4)
–.023
(.031)
.117**
(.056)
–.054
(.068)
1,557
.205
.069***
(.025)
.177***
(.029)
Linear1
Yes
–.001
(.019)
.127***
(.038)
–.004*
(.002)
–.002
(.002)
.004**
(.002)
.001
(.005)
–.010
(.007)
–.006
(.008)
–.013***
(.005)
.003***
(.001)
HOLDINGrate
(5)
NBinom. 3
No
–.037
(.369)
.604
(.439)
.001
(.003)
–.063*
(.033)
.002
(.004)
–.045
(.053)
.003
(.004)
–.458***
(.011)
–.027
(.079)
.021**
(.010)
–.002
(.002)
.181
(.173)
.316***
(.086)
.051*
(.027)
–.101***
(.034)
.100*
(.054)
–.416*
(.240)
1,553
HOLDINGcount
(6)
–
+
–
+
+
+
–
+
–
–
+
.025
(.097)
.216*
(.130)
.149
(.208)
1,193
–.120**
(.053)
–.522***
(.167)
NBinom. 2
Yes
.023
(.049)
–.343***
(.111)
.012**
(.006)
–.005
(.006)
–.000
(.003)
–.012
(.013)
–.005
(.019)
.033
(.184)
.033*
(.017)
–.004*
(.002)
OP_vs_
HOLDcount
(7)
.046*
(.026)
.062
(.044)
–.102
(.068)
2,433
.087
–.065***
(.020)
–.104***
(.026)
Linear1
Yes
.009
(.018)
–.164***
(.035)
.005***
(.002)
–.001
(.002)
.001
(.001)
–.006*
(.004)
.000
(.002)
.011**
(.005)
.020***
(.006)
–.002***
(.001)
OP_vs_
HOLDrate
(8)
NBinom. 3
No
.001
(.058)
–.228***
(.083)
.005***
(.001)
–.001
(.002)
.005***
(.002)
–.010
(.014)
–.000
(.003)
–.000
(.003)
.007
(.017)
–.001
(.003)
–.002*
(.001)
–.098*
(.052)
–.145***
(.028)
–.059***
(.014)
.124***
(.037)
–.019
(.057)
–.106*
(.057)
2,433
OP_vs_
HOLDcount
(9)
*, **, and *** denote significance levels of 10, 5, and 1 percent respectively; we report standard errors in parentheses; 1 Standard errors are robust and clustered on group level;
2
Standard errors are bootstrapped (400 iterations); 3 Standard errors are robust and clustered on country level; we include but not report a constant in our model; in Specification 3, 6, and 9 we include but not report industry dummies; we indicate expected coefficients according to our hypotheses as + or –.
Observations
Adjusted R²
Varindustry
LOSSsubgroup
Varsize
SIZEgroup
SIZEsubgroup
ln(GDPs )
GDPpcu
GDPpcs
τs
(1 − gtss;dp) ∙ WDp;f
WDp;f
(1 − gtss;dp) ∙ WDs;d
WDs;d
(1 − gtss;dp) ∙ exp
Exp
gtss
divtaxs
Model type
Group fixed effects
Dependent
variable
A. Oestreicher/R. Koch
Table 11: Regression results (1)
sbr 64 October 2012 254-280
Taxation
4.3 R egression R esults
Table 11 displays the regression results for the count data and linear models for all
three group-structure decisions under scrutiny. Specifications (1) to (3) refer to the decision between a horizontal structure and a vertical structure with operative intermediary (testing Proposition 1), and specifications (4) to (6) refer to the decision between a
horizontal structure and a vertical structure with intermediate holding (testing Propositions 2a and 2b). Specifications (7) to (9) address the choice between the two types of
a vertical subgroup structure (testing Propositions 3a and 3b). In each case, the first
specification (count data model with group fixed effects) is ‘preferred’ (specifications
(1), (4) and (7)), and the specifications serving as robustness check (linear model with
group fixed effects and count data model without group fixed effects) follow (specifications (2) and (3), (5) and (6), and (8) and (9)).
As expected, the results presented in Table 11 show a significant and substantial impact
of the applicable tax provisions on the corporate group structure. Firstly, the requirement of a domestic parent entity for application of group taxation (gtss) factors particularly strongly into the decision for or against interposing a domestic holding company. In line with Propositions 2b and 3b we estimate positive coefficients for this covariate in Specifications (4) to (6) and negative coefficients in Specifications (7) to (9),
all meaning that subsidiaries in the host country are more frequently held through a
domestic intermediary holding company if a parent entity is required for application of
a group tax regime. This effect is significant in both of the preferred specifications ((4)
and (7)) and in three out of four robustness checks. In order to compare the size of the
effects estimated in different specifications we translate the coefficients estimated with
the linear models into semi-elasticities20. As a result, semi-elasticities estimated for gtss
in Specifications (4) to (6) vary in a range between .464 and .604 (preferred specification: .568), meaning that the share of subsidiaries held by an intermediate holding company in the host country (as compared to subsidiaries held by a company resident in the home country) is some 46 to 60 percent higher in European member states
where a domestic parent is required for application of group taxation. In comparison to
subsidiaries held by an operative intermediary in the host country (Specifications (7) to
(9)), the respective semi-elasticities vary between –.201 and –.343 (preferred specification: –.343). For the choice between a horizontal structure and a vertical structure with
operative intermediary (Specifications (1) to (3)) positive coefficients are estimated for
gtss, as expected. However, this effect is considerably reduced in both significance and
size. Here, we estimate semi-elasticities varying between .088 and .160 (preferred specification: .091), being significant at the ten percent level only for the linear model.
A second significant tax impact stems from provisions regarding the write-down of participation values. According to our Proposition 1 interposing an operative intermediary
is beneficial in this respect if such deduction is allowed only in the host country or would
become tax effective in the host country at a higher rate. In line with this proposition we
20 In contrast to the linear model, coefficients estimated with the count data models can be interpreted directly as
semi-elasticities.
sbr 64 October 2012 254-280
275
A. Oestreicher/R. Koch
estimate significantly positive coefficients for WDs;d for all specifications (1) to (3), but
the coefficients for WDp;f are insignificant. Semi-elasticities for WDs;d vary in a range
between .006 and .008 (preferred specification: .008). Consequently, the number (share)
of subsidiaries held through an operative intermediary is increased by some six to eight
percent if participation write-downs can be deducted in the host country at a ten percent
higher rate. In the case of a pure holding being interposed, these advantages (WDs;d >
0) would become effective only if the host country offers a group tax regime (see Propositions 2a and 3a). However, in the absence of a group tax regime implementing a vertical
structure with intermediary holding would be disadvantageous if the tax law in the home
country permits a participation write-down (WDp;f >0) (see Propositions 2b and 3b). In
line with this argumentation, we estimate in Specifications (4) to (6) positive coefficients
for WDs;d and by majority negative coefficients for (1 − gtss;dp) ∙ WDs;d and (1 −
gtss;dp) ∙ WDp;f. These effects are significant at the one-percent level ((1 − gtss;dp) ∙
WDs;d and (1 − gtss;dp) ∙ WDp;f) and at the 15-percent level (WDs;d) in the preferred
Specification (4). These findings are supported by significant coefficients in one of the two
robustness checks in the case of WDs;d and (1 − gtss;dp) ∙ WDp;f21. Contrastingly, these
provisions exercise no significant impact on the choice between a horizontal subgroup
structure and a vertical structure with intermediary holding (Specifications (7) to (9)).
The advantages of a vertical group structure concerning intragroup loss offset and deduction of participation write-downs become effective in particular if subgroup companies suffer losses. In line with this expectation we estimate significantly positive coefficients in most of the Specifications (1) to (6).
The results for the other tax parameters included in our regression model show no clear
impact. According to Propositions 1, 2a and 2b, coefficients for divtaxs are expected
to be negative in the case of Specifications (1) to (6), meaning that an additional tax
burden on domestic dividend payments between companies hinders corporate groups
from interposing an intermediary company. Contrastingly, Propositions 3a and 3b predict positive coefficients for Specifications (7) to (9) as such tax burden on dividends
should discriminate against intermediary holdings as compared to operative intermediaries22. Although estimated coefficients predominantly show the expected sign, in
particular as far as Specifications (4) to (9) are concerned, this effect is significant in
none of the cases concerned. This result should come as no surprise since the extent of
this additional tax burden – if any – is small (with only few exceptions).
For exp an influence similar to WDs;d/WDp;f was expected according to our propositions.
However, estimated coefficients are insignificant (Specifications (4) to (6)), show significantly
differing plus or minus signs in the best-method specification and in robustness checks (Specifications (1) to (3)) or point in a direction opposite to theoretical expectations (Specifications
(7) to (9). Concerning the nominal tax rate in the host country (τs) we find that in the pres21 (1 − gtss;dp) ∙ WDp;f and (1 − gtss;dp) ∙ WDs;d take values different from zero only for a very small number of observations. This may serve as an explanation for the large size of the effects estimated for these interaction terms.
22 In contrast to a vertical structure with operative intermediary, distributions of all operative subsidiaries are subject to dividend taxation in the host country if a pure holding company is interposed.
276
sbr 64 October 2012 254-280
Taxation
ence of high tax rates horizontal structures (vertical structures with operative intermediary)
are preferred to vertical structures (vertical structures with holding intermediary).
Furthermore, the results presented in Table 11 imply that also non-tax considerations
factor significantly into the choice between a horizontal and a vertical subgroup structure. On this note, we find that a vertical group structure is applied significantly more
frequently the more companies belong to the subgroup (significantly positive coefficient for SIZEsubgroup in Specifications (1)-(6)). The size of the companies also plays
a significant role. The more the size of the companies included in the local subgroup
differs (high values for Varsize), the more frequently vertical structures with operative
intermediary can be observed. This outcome was expected and should be due to the
fact that negative management incentives of such a structure are expected to be smaller
or of minor importance if a large operative company controls one or only a few smaller
operative companies. As far as the choice between the two types of a vertical subgroup
structure is concerned, we observe holding structures more frequently in large subgroups (negative coefficient for SIZEsubgroup) and subgroups with companies that differ
only to a minor extent in size (positive coefficient for Varsize).
Two of the non-tax parameters may be considered to be of endogenous nature. With
regard to the number of companies in the subgroup (SIZEsubgroup) one may argue that
this variable depends on the group structure if multinationals avoid splitting up their
activities in several companies per member state if no group tax regime applies23. We
can take the same argument for LOSSsubgroup assuming that companies make use of
different strategies to avoid the occurrence of loss situations if no group tax regime applies. We therefore estimated the count data models in Table 11 without considering
these two variables24. However, concerning the tax influences this did not affect our
findings to any significant extent.
In addition to the regression results reported in Table 11, we estimate fixed effects logit
models in order to capture non-tax parameters subsidiary-specifically and therefore
more accurately. The results displayed in Table 12 confirm most of our previous findings relating to the tax variables. However, we find here a significantly negative coefficient for divtaxs in Specification (2), meaning that an additional tax burden on dividends in the host country hinders groups from interposing a pure holding company.
With regard to the non-tax parameters we find that subsidiaries held by an operative
intermediary are usually smaller than average (negative coefficient for SIZE subsidiary
in Specification (1)). The opposite applies to subsidiaries that are held by intermediate holdings. Furthermore, our findings show that subsidiaries are held indirectly in
a larger number of cases if at least one company in the subgroup belongs to the same
industry (positive coefficients for sameindustry in Specifications (1) and (2)). Interestingly, in contrast to the definition referring to the subgroup as a whole (LOSSsubgroup
included in the specifications reported in Table 11) the company-specific definition of
LOSS (LOSSsubsidiary) turns out to be insignificant. This finding could indicate that it
23 In an extreme case, subgroups may even be selected out of the sample if only one subsidiary is used.
24 Results of these additional models will be provided to interested readers upon request.
sbr 64 October 2012 254-280
277
A. Oestreicher/R. Koch
is the profitability of the subgroup as a whole that factors into the decision for a specific
subgroup structure rather than that of an individual subsidiary.
Table 12: Regression results (2)
Dependent variable
(1)
(2)
(3)
OPERATIVEdummy
HOLDINGdummy
OP_vs_ HOLDdummy
Logit
Logit
Logit
Yes
Yes
Yes
–.101
–.527**
–.190
Model type
Group fixed effects
divtaxs
gtss
Exp
(.064)
(.267)
.222
1.285***
(.163)
(.469)
–.018**
–.028
(.009)
(.032)
∗
(1 − gtss;dp) exp
.031***
WDs;d
(.007)
τs
.011
–.001
(.018)
(.020)
–.064
(.072)
(.354)
ln (GDPs)
Sameindustry
.350**
(.057)
(.120)
.282***
(.060)
(.089)
.018
–.036*
(.005)
(.013)
(.019)
–.416***
.397*
–.875***
(.113)
LOSSsubsidiary
(.117)
–.387***
–.106*
(.103)
SIZEsubsidiary
–.374***
.009
1.360***
SIZEsubgroup
.618***
(.219)
(.024)
.019***
GDPpcs
(.215)
(.271)
1.624***
–.541**
(.205)
(.264)
–.430***
.419***
–.359***
(.083)
(.115)
(.098)
.082
.173
.059
(.095)
(.163)
(.128)
.440***
.395**
.103
(.100)
(.175)
(.163)
4,606
2,485
3,112
–1556.40
–604.79
Observations
Log likelihood
.046
(.089)
–.023
(1 − gtss;dp) ∙ WDp;f
.074**
(.038)
–.447
(.110)
WDp;f
(.540)
(1.701)
–.724***
(1 − gtss;dp) ∙ WDs;d
(.301)
–1.923***
–994.47
*, **, and *** denote significance levels of 10, 5, and 1 percent respectively; we report standard errors in parentheses; standard errors are bootstrapped (400 iterations); we include but not report a constant in our model.
278
sbr 64 October 2012 254-280
Taxation
5 C onclusion
The point of departure for this paper was the assumption that businesses face additional costs if they adopt a legal group structure deviating from the pre-tax optimal
setting for tax reasons. Apart from planning costs and costs for establishing and maintaining a certain group structure, undesired consequences may result for the risk situation or the allocation of managers’ tasks and responsibilities. Although the tax impact
on the choice of legal form has already been demonstrated empirically, only scant literature exists with regard to group structure decisions. This paper therefore aims at analyzing the influence of tax considerations on the structure of local corporate subgroups,
i.e., the structure of investments of a parent company resident in one EU member state
(home country) holding subsidiaries resident in a different member (host country). To
this end, we distinguish horizontal and vertical group structures. Vertical group structures can take the form of a group with an operative holding or with a pure holding
as its domestic parent company. We compare the tax consequences of these different
structures on the basis of a simple one-period model. Our results show that a vertical group structure is beneficial from a tax perspective if a domestic parent is required
for application of a group tax regime and – at least in case of an operative intermediary – brings with it advantages concerning the deduction of participation expenses and
participation write-downs. Additional tax payments on intragroup dividends in the
host country, however, make the implementation of a vertical structure, especially by
interposing a domestic holding company, a less attractive alternative.
To test this model, we analyzed a sample of 3,336 domestic subgroups with regard to
their structure. The findings support most of the propositions derived from our model.
In particular, we find that a vertical structure with a pure holding interposed is implemented significantly more frequently if in the host country a domestic parent entity is
required for the formation of a tax group. This result emerges in comparison to both a
horizontal subgroup structure and a vertical subgroup structure with operative intermediary, with the size of the effect being substantial in both cases (semi-elasticities of
.568 and –.343 in the preferred specifications). Furthermore, our analysis shows that
the choice between a horizontal structure on the one hand and a vertical structure with
an operative intermediary or with an intermediate holding, on the other hand, is affected by the deductibility of a participation write-down in the host country. Other tax
parameters, such as an additional tax burden on dividend distributions or the deductibility of financing expenses in the host country, show no significant impact. Among
the non-tax parameters, in particular the size of the subgroup, the relative size of the
subgroup companies and their profitability exercise significant influence.
All in all, it emerges that corporate groups frequently base their choice of group structure on the prospect of achieving tax advantages to a significant degree. As this rationale can lead to groups abandoning the pre-tax optimal structure, the European member states would be well advised to amend their regimes in such a way as to facilitate
group taxation in general. Multinational companies could then combine their legal
group structure with the pre-tax optimal arrangement and, consequently, avoid possible inefficiencies or social waste.
sbr 64 October 2012 254-280
279
A. Oestreicher/R. Koch
R eferences
Buettner, Thiess, Nadine Riedel, and Marco Runkel (2011), N., Strategic consolidation under formula apportionment, National Tax Journal 64, 225-254.
Cragg, John G., Arnold C. Harberger, and Peter Mieszkowski (1967), Empirical Evidence on the Incidence of the
Corporation Tax, The Journal of Political Economy 75, 811-821.
Endres, Dieter, Andreas Oestreicher, and Wolfram Scheffler et al. (2006), The Determination of Corporate Taxable
Income in the EU Member States, Alphen aan den Rijn: Kluwer Law International.
Goolsbee, Austan (1998), Taxes, organizational form, and the deadweight loss of the corporate income tax, Journal
of Public Economics 69, 143-152.
Gravelle, Jane and Laurence J. Kotlikoff (1989), The Incidence and Efficiency Costs of Corporate Taxation When
Corporate and Noncorporate Firm Produce the Same Good, The Journal of Political Economy 97, 749-780.
Harberger, Arnold C. (1962), The Incidence of the Corporation Income Tax, The Journal of Political Economy 70,
215-240.
Hoskisson, Robert E. (1987), The Contingency of Diversification Strategy, The Academy of Management Journal 30,
625-644.
Kesti, Juhani (2003), European Tax Handbook, Amsterdam: IBFD Publications.
KPMG (2006), Corporate Tax Rate Survey: An international analysis of corporate tax rates from 1993 to 2006,
downloadable at http://www.kpmg.com.
Kreikebaum, Hartmut, Dirk Ulrich Gilbert, and Glenn O. Reinhardt (2002), Organisationsmanagement internatio­
naler Unternehmen, 2nd ed., Wiesbaden: Gabler.
Kutschker, Michael and Stefan Schmid (2011), Internationales Management, 7th ed., München: Oldenbourg
MacKie-Mason, Jeffrey K., and Roger H. Gordon (1997), How Much Do Taxes Discourage Incorporation?, Journal
of Finance 52, 477-505.
Markides, Constantinos C. and Peter J. Williamson (1996), Corporate Diversification and Organizational Structure:
A Resource-Based View, The Academy of Management Journal 39, 340-367.
Mintz, Jack M. and Alfons Weichenrieder (2010), Holding Companies and Ownership Chains, in Mintz, Jack and
Alfons Weichenrieder (eds.), The Indirect Side of Direct Investment, Cambridge/London: The MIT Press.
Oestreicher, Andreas and Reinald Koch (2010), The determinants of opting for the German group taxation regime
with regard to taxes on corporate profits, Review of Managerial Sciences 4, 119-147.
Poppe, Andreas (2007), Auswirkungen der Einführung einer konsolidierten Körperschaftsteuer-Bemessungsgrundlage in
der Europäischen Union, Frankfurt et al.: Peter Lang.
Scheffler, Eberhard (2004), Vor- und Nachteile der Holding, in Lutter, M. (ed.), Holding Handbuch, Köln: Dr. Otto
Schmidt Verlag, 30-42.
Scherm, Ewald and Stefan Suess (2001), Internationales Management: eine funktionale Perspektive, München:
Vahlen.
Theisen, Manuel R. (2000), Der Konzern, 2nd ed., Stuttgart: Schaeffer-Poeschel.
Wamser, Georg (2011), Foreign (in)direct investment and corporate taxation, Canadian Journal of Economics 44,
1497-1524.
280
sbr 64 October 2012 254-280