Increasedeciency
through consolidation
and formula apportionment in the
European Union?
MichaelP. Devereuxand Simon Loretz†
4th April2008
Abstract
This paper assesses the recent reform proposals of the European Union to
introduce international loss consolidation and formula apportionment in terms
ofeciency.
For this exercise we extend theeective
tax rate methodology of
Devereuxand Grith
(1999) to includea potential loss and use a large firm
level data set to identify the distortions under the current system and after
the proposed tax reforms. While allowing international loss consolidation in
the current system would signify amove to the worse in terms ofeciency,
a formula apportionment system would increase capital export neutrality. At
the same time capital ownership neutrality would be violated because, the
widespread of statutory tax rates in the Member States would translate into
dierent
tax burdens for dierent
companies.
JEL classification: H25, H87
Keywords: Corporate Taxation; International Loss Consolidation; Appor-
tionmentRules;
Oxford University Centre for Business Taxation; Park End Street, OX11HPOxford, England,
e-mail: Michael.Devereux@sbs.ox.ac.uk
†
Oxford University Centre for Business Taxation; Park End Street, OX11HPOxford, England,
e-mail: Simon.Loretz@sbs.ox.ac.uk
1
Introduction
Reform plans in the European Union are more and more shaping towards a common
consolidated tax base. However, the process ofharmonisation has been slow and
is facing a lot of opposition from various Member States. Therefore the European
Commission adopted anew strategy of stepwise reform proposals. Along these lines
Ag´undez-Garc´ıa (2006) discusses international loss consolidation and various forms
of formula apportionment. It is widely accepted that this would significantly reduce
compliance costs, mitigate problems with transfer pricing and enhance anecient
distribution of investment within the European Union. While some empirical research
examines the impact of such tax reforms on the Member States tax revenues,
theeciency
aspects received scant attention sofar.
This paper tries to fill this gap by analysing the changes ineciency
through
the introduction of international loss consolidation and formula apportionment. For
our purpose we measureeciency
in a number of dierent
dimensions: (i) in terms
of capital export neutrality (CEN), which demands that capital should betaxed to
the same extent regardless where it is invested. (ii) capital import neutrality (CIN),
which requests that capital invested in a certain jurisdiction should face the same
tax burden, regardless of the residence of the investor and finally (iii) capital own-
ershipneutrality (CON), which stipulates that a capital owner should face the same
tax burden, regardless where she invests.1
We start with the eective
tax rates as developed by Devereuxand Grith
(1999) and extend this methodology introducinga potential loss. Further we make
use of a large micro-leveldatasetto identify the current distribution of theeective
tax burden across firms. This allows us to compare the distortions under the current
system with the potential distortions under the new system with international loss
consolidation and formula apportionment.
The introduction of international loss consolidation under the current system
would largely increase the spread of theeective
tax burdens, which would represent
amove away from both capital export neutrality and capital ownership neutrality.
Combining international loss consolidation with formula apportionment would cor-
rectmostofthese new distortions and signifya substantial improvement in terms
of capital export neutrality. However, the same move would, because ofdierences
in statutory corporate tax rates across Member States, imply reduced capital own-
ershipneutrality. Given the existing tax saving opportunities under the current
system, which would no longer exist under the formula apportionment system, this
change for the worse would only be minor.
2
Methodology
2.1 Theeective
tax rate approach
Theeective
average tax rate (henceforth EATR) literature builds on the cost of
capital approach of Jorgensen (1963) and Hall and Jorgensen (1967) which was
further developed to measure the tax burden on discrete investment choices by De-
vereuxandGrith
(1999).
2
The OECD (1991 and Yoo 2003) and the Commission
1
The concepts of capital import neutrality and capital export neutrality at least date back to
Musgrave (1969). The concept of capital ownership neutrality is introduced in Devereux (1990)
and further discussed in Desaiand Hines (2003).
2
We mainly use the notation of Devereuxand Grith
(2003) which is somewhat simplified.
See Devereuxand Grith
(1999) fora detailed description of the model.
1
of the European Union (1992 and 2001) employed and discussed this methodology
in detail, therefore we only summarize it very briefly.
The underlying model assumes that the cross-border investment increases the
capital stock in an existing foreign subsidiary for only one period. Hence an increase
in investment in periodtofdI
t
=1 impliesareduction of the investment in period
t+1 ofdI
t+1
= (1)(1+) where denote the economic depreciation and
stands for the nominal inflation rate.3 The perturbation in the capital stocks
generates an additional output in periodt+1 of (p+)(1+) where prepresents
the real economic rent. In the absence of taxation and assuming that purchasing
power parity holds the net present value of the additional income stream is
R=1+
1
1+i
{ (1+)(p+) + (1+)(1) }=
p
r
1+r
(1)
where i= (1+r)(1+) 1 denotes the nominal interest rate.4
With taxation the return is subject to the corporate tax in the host country
n
, which reduces the return in periodt+1 to (p+)(1+)(1
n
).5 DefineA
n
as the net present value of tax allowances per unit of investment, discounted with
the shareholders nominal discount rate.6 If the investment is financed through
retained earnings the shareholder needs to give up (1
A
n
) of dividends in period
t. Thedierence
between these two changes represents the additional dividend flow
to the parent which is potentially subject to taxation upon repatriation. Define
jn
as the total tax due because of repatriation to the parent. The extent of this tax
burden depends the method of the method of double taxation alleviation
jn
=
8
>
<
>
:
c
n
exemption
max
n
j
n
1
n
,c
n
o
credit with limitation
j
(1
c
n
) +c
n
deduction
(2)
wherec
n
describes the withholding taxon dividends. Hence the net present
value of the after tax income stream can be written as
R= (1
jn
) (1 A
n
) +
1
1+
((1+)(p+)(1
n
)
+ (1+)(1)(1 A
n
))
i
(3)
Other forms of finance: If the investment is financed through new equity the
cost of raising 1 unit new equity is (1
n
n
), where
n
denotes the tax depreciation
in the first period. Inturn the dividend is not reduced by (1
n
n
) shareholder.
In the second period the dividend is reduced by (1
n
n
) in order to pay the newly
raised equity of (1
n
n
) back. Abstracting from taxation at the shareholder level,
the net present value of the income stream isunaected
if the new equity is raised in
the parent country, while raising new equity in the host countryaects
the timing
of the repatriation tax
jn
.
In the case of debt financing the shareholder receivesa (1
n
n
) higher dividend
inperiodtbutin periodt+1 the debt plus the tax deductible interest needs to
3
For simplicity reasons the inflation rate is assumed to be the same for capital and output.
Further simplifying assumptions are that inflation and economic depreciation are identical across
countries and purchasing power parity holding, hence the real exchange rate is equal to unity and
we drop the subscripts for the inflation rate.
4
For the calculations we use parameter values of 0.05 for rand 0.025 for.
5
Note that throughout the paper ndenotesthe host country while jdenotesthe parent country.
6
Abstracting from taxation at the shareholder level equals to the nominal interest rate i.
2
be paid back, which amounts to (1
i (1
j
)). If the subsidiary borrows at the
parent company the interest payments are deductible there, however potential the
additional tax burden because of withholding taxes and double taxation must be
taken into account. It is therefore useful to define!
jn
as the total taxon interest
payments from the subsidiary to the parent, again depending on the method of
double taxation alleviation.
!
jn
=
8
<
:
!
n
n
exemption
max{
j
, !
n
}
n
credit with limitation
j
(1
!
n
) +!
n
n
deduction
(4)
where!
n
denotes the withholding taxon interest payments between subsidiaries
and parent companies. Denoting the change in new equity as dNandthechange
indebtasdBfor additional costs of financing in the parent or host country can
therefore be written as
F=
(1
n
n
)
1+
[i (1
j
)]dB
j
+
jn
1+
(1
n
n
) dN
n
+
(1
n
n
)
1+
{
jn
[1+i (1
n
) (1+)]!
jn
i}dB
n
(5)
To extent which extent the investment is tax depreciable and therefore the size
ofA
n
depends on the type of assets. Machinery is tax depreciated faster as build-
ingsand inventories cannot be depreciated at all. However, if the inventories are
valuated according to the FIFO method, the increase in value because of inflation
is due to taxation and therefore RRE needs to be adjusted to
R
INV
=R
1
(1+)
n
(1
n
)(1+)
(6)
Solving the after tax income stream Rforthenecessary economic rent to break
even yields the cost of capital
ep=
(r+)(1 A
n
)
1
n
+
F (1+r)
(1
jn
)(1
n
)
(7)
where Fdescribes the additional costs of other forms of finance as defined in (5).
In the case of a FIFO valuation of the inventories the adjustment term in equation
(6) needs to be accounted for. The EMTR is then defined as
EMTR=
e
p
r
e
p
(8)
The EATRisdefined as thedierence
between the NPVof the income stream
in the absence of taxes and NPV of the income stream in the presence of taxes
in relation to the NPV of thepre-tax total income stream p/ (1+r). Using the
equations (1) and (3) the EATR is given through
EATR=
R
R
p/ (1+r)
=
pr
1+r
+ (1
jn
)
h
1
1+r
((p+)(1
n
) (1 A
n
)(r+))
i
p/ (1+r)
(9)
Equation (9) describes the EATR fora not further defined investment financed
through retained earnings. Analogously the EATR for other forms of finance can
3
be calculated if the relevant additional costs of finance as defined in equation (5) are
included. This allows to calculate EATR sforspecific investments and for specific
forms of finance. We adopt this convention throughout the rest of the paper.
2.2 Eective
tax rates fora new subsidiary with potential
losses
To fully capture the potential changes due to the introduction of aformulaappor-
tionmentsystemwe need to extend the standardeective
tax rate framework. We
depart from the original model insofar as we assume that the subsidiary is new, i.e.
has neither earnings to be retained as a possible form of finance nor there are any
existing profits where any tax depreciation in periodtcanbe claimed. Therefore
they present a taxable loss, that can be carried forward into the next period.
More importantly we also introducea potential loss to the model to account for
thedierences
in the outcome under variousdierent
systems of group taxation.
With a possibility ofqthenew investment yields a positive outcome ofg>0. In
the other cases the firm suers
areal economic loss
b<0. In order to compare it
to the existing model we assume that the expected value of the outcome equals the
assumed pretax rate of return, i.e. thatp=qg+ (1
q) bholds.7
Good outcome: With a good outcome the subsidiary is profitable, i.e. (g+
)(1+) >1, and therefore liable to taxation in periodt+1. The taxable loss of
the first period (the taxable loss of period tequalsto the depreciation allowances
n
) can be oset
against the profit in period 1. Therefore we need to adjust the
net present value of the depreciation allowances A
n
for this delay.
ˆ
A
n
=A
n
i
nn
1+
(10)
Apart from this delayed depreciation the good outcome is exactly the same as
described above, therefore equation (3) can be rewritten using equation (10) and
replacing pwithg.
R
GOOD
= (1
jn
) (1
ˆ
A
n
) +
1
1+
((1+)(g+)(1
n
)
+ (1+)(1)(1 A
n
))
i
(11)
Concerning the additional cost because ofdierent
forms of finance equation (5)
needs to be adjusted insofar, as the costs for one additional unit are now 1 instead
of (1
n
n
).
F
GOOD
=
1
1+
[i (1
j
)]dB
j
+
jn
1+
dN
n
+
1
1+
{
jn
[1+i (1
n
) (1+)]!
jn
i}dB
n
(12)
Bad outcome: If the subsidiary has a bad outcome the return on investment in
periodt+1 is (b+)(1+), wherebcanbe negative implying areal loss. Identical
to the case of the good outcome the taxable loss brought forward from periodt
7
In particular we assumeg=0.3 andb=0.2 with a probability of the good outcome of
q=0.8. This impliesavalueof 0.2 forpwhichis usually used in the calculation of EATRs.
4
equals to the first year depreciation allowance
n
. However, given that the project
yields no taxable profits, these losses cannot be used. Further we assume that
after learning about the bad nature of the outcome the project is abandoned and
the investment is sold. Therefore thedierence
between the real value and the tax
depreciated value must be added as a balancing charge. Accounting also for the tax
deductibility of interest payments the taxable income is negative if the following
condition is met.8
b<
idB
n
1+
(13)
Under the assumptions we use (b=0.2 and
n
=0.025) this condition is met
even for no debt finance at all, i.e. iB
n0
=0. Further if this condition holds, the
bad outcome is a real economic loss. Therefore there is no dividend tax on the repayment
of the equity in periodt+1. Consequently the NPV of the income stream
with the bad outcome simplifies to
R
BAD
=1+
(b+)(1+) + (1)(1+)
1+
=
b
r
1+r
(14)
The additional costs for the other forms of finance simplify as well, as no dividend
taxis due and as the interest payments are only tax deductible at the parent.
F
BAD
=
i (1
j
)
1+
dB
j
+
!
jn
i
1+
dB
n
(15)
The new measure of the EATR is then simply the probability weighted NPVs
of the good and the bad outcome as in equation (11) respectively (14) in relation to
the income in the absence of taxation. For the other forms of finance the additional
cost as described in equation (12) respectively (15) need to be added.
EATR
n.c.
=
R
[qR
GOOD
+ (1q) R
BAD
]
p/ (1+r)
(16)
To calculate the cost of capital we solve the expected net present value of the
income stream for the necessary good outcome to breakeven. In doing so, we hold
the bad outcome b and the probability of a good outcome q fixed.9 Using equations
(11)(12)(14) and (15) the necessary return in the case of the good outcome without
loss consolidation can be written as
eg
n.c.
=
(r+)(1 A
n
)
1
n
+
i
nn
(1+)(1
n
)
1+r
(1
jn
)(1
n
)
F
GOOD
+ (R
BAD
+F
BAD
)
1
q
q
(17)
Weighting the necessary good outcome and the fixed bad outcome with the
probability yields the cost of capital
ep=qeg
n.c.
+ (1
q) b
(18)
The EMTRisthen calculated using the standard equation as defined in (8).
8
Detailed derivations can be found in a technical appendix, available from the authors upon
request.
9
There are two other ways to calculatea measure of the cost of capital. Instead of fixing the
badoutcomeb, the good outcome gcouldbe hold constant to calculate the worst outcome still
breaking even (
e
b). Alternatively, one can hold the good and the good and bad outcome constant
and calculate the necessary probability (eq) of a good outcome.
5
2.3 Eective
tax rates with international loss consolidation
Good Outcome: If the investment project is successful the first year depreciation
allowances that are claimed as group relief in period tneedtobe deducted from
theNPVofthe depreciation allowances. Define A
n
as the net present value of the
depreciation allowances in the host country if immediate group relief is claimed.
A
n
=A
n
nn
(19)
The immediate group relief can be tax deducted at the parent and therefore
leads to a direct tax saving of
jn
. Note that the group relief is not subject to
dividend taxation. Therefore the posttax rate of return with international loss
consolidation can be written as
Rr
GOOD
=
j
n
+ (1
jn
) (1 A
n
) +
1
1+
((1+)(g+)(1
n
)
+ (1+)(1)(1 A
n
))
i
(20)
The additional costs of the other forms of finance are similar to the one for the
original model with the onlydierence
that the first year depreciation
n
allowance
is now tax deducted at the parent country tax rate
j
.
Fr
GOOD
=
(1
jn
)
1+
[i (1
j
)]dB
j
+
jn
1+
(1
j
n
) dN
n
+
(1
jn
)
1+
{
jn
[1+i (1
n
) (1+)]!
jn
i}dB
n
(21)
Bad Outcome: If the project is unsuccessful and sold after periodt+1 the
calculation of theeective
tax rates with international loss consolidation is slightly
more complicated. It is now possible that the group relief claimed in period tis
large enough to imply a balancing charge in periodt+1 that turn the negative
taxable income into a taxable profit. In contrast to condition stated in equation
(13) there is no loss brought forward into periodt+1 and therefore the balancing
charge does not cancel out. Hence the taxable income in the host country T
n
is
T
n
=b+b++
n
idB
n
(22)
which is negative if the following condition holds.
b<
idB
n
n
1+
(23)
If the condition in equation (23) is met, the investment project is not subject to
tax in the host country. However, the taxable loss can beoset
against tax profit
at the parent company. In contrast if the inequality in equation (23) does not hold,
the balancing charge is large enough to create a tax liability at the subsidiary and
no taxable loss exists to beoset
against the parent profits.
The rest of the NPV of the income stream with the bad outcome is identical to
the case described in equation (14). Hence the after-tax NPV of the income stream
with the bad outcome and international loss consolidation is
Rr
BAD
=
(
j
h
n
T
n
1+
i
+R
BAD
ifT
n
0
j
n
n
T
n
1+
+R
BAD
ifT
n
>0
(24)
6
In line with the distinction above the additional costs of other finance forms
depend now also on whether the condition in equation (23) is met. Given that the
taxable income in the subsidiary is positive the interest payments for debt raised at
the subsidiary can be deducted at the subsidiary level against
n
. If in contrast the
taxable income is negative (because of the interest deduction) the deduction of the
interest increases the loss that can beoset
against the tax profits at the parent.
Fr
BAD
=
(1
jn
)
1+
[i (1
j
)]dB
j
+
(1
jn
)
1+
{ID!
jn
i}dB
n
(25)
where
ID=
(
min
h
i, i
b+b++
n
dB
n
i
j
ifT
n
<0
i
n
ifT
n
0
(26)
denotes the interest deductible depending on whether the condition in equation
(23) holds. The EATR is again the probability weighted NPVsofthegoodand the
bad outcome following the same logic as in equation (16).
EATR
l.c.
=
R
[qRr
GOOD
+ (1q) Rr
BAD
]
p/ (1+r)
(27)
Following the same logic as in equation (17) the necessary return in the good
outcome for the loss consolidation case can be calculated.
eg
l.c.
=
(r+)(1 A
n
)
1
n
+
nn
(1+r)
(1
n
)
1+r
(1
jn
)(1
n
)
F
GOOD
+
nj
+ (R
BAD
+F
BAD
)
1
q
q
(28)
Using equation (28) in equations (18) and (8) then yields the EMTR for the
loss consolidation case.
The decision whether to participate: Given that a international group relief
is sofar only allowed in few countries, most notably in Austria or Denmark,
and even in these countries international loss consolidation is an option it seems
plausible to model the participation as voluntary. Therefore we assume that the
firm can choose between the immediate group relief against the taxable profits of
the parent company or carry the loss forward to relief it against potential future
profits in the subsidiary. However, it is not possible to claim relief for the same
loss at both, the parent and the subsidiary level. Hence the extent of the relief will
depend on the choice of the firm.
Leaving the loss in the subsidiary and carry it forward to periodt+1 the firm
can claim relief against the taxable profit at the subsidiary in periodt+1. Further
the reduced tax burden at the subsidiaryaects
the dividend repatriated to the
parent, which inturn is subject to
jn
. However, the probability of a good outcome
isonlyq<1. Hence with probability (1 q) >0 the firm cannot claim a tax
relief in the subsidiary and can only claim a group relief with the taxable profit at
parent company.10 Therefore the expected value of the tax relief if no group relief
is claimed is:
10
We assume that a corporate group can again choose to participate in periodt+1 if it did not
participate in periodt. Therefore it will participate incase of a bad outcome.
7
E
no
=
n
1+
{q[
n
(1
jn
)]+ (1
q)
j
}
(29)
If immediate group relief is claimed the value of the immediate tax relief is certain
and amounts to
jn
. Inperiodt+1 the additional balancing charge in the
bad outcome can lead to a positive tax in the subsidiary if the condition in equation
(23) does not hold. Therefore the positive taxable income is now subject to
taxation in the host country rather than tax deductible as a loss in the home country.
E
cons
=
jn
+
1
1+
{ (1q) max[T
n
, 0](
n
j
) }
(30)
Comparing these two outcomes, the firm will choose to immediately claim group
relief if the expected tax relief in the future is lower than the certain immediate tax
relief, taking into account that the balancing charge might lead to a positive taxable
income in the bad outcome. Define the choice of the firm as which takes the value
1 if the firm claims immediate group relief and 0 otherwise.
=
1
ifE
no
<E
cons
0
ifE
no
>E
cons
(31)
If the firm chooses to participate the cost of capital and averageeective
tax
rate are the same as described in equations (27) and (28). The cost of capital and
eective
average tax rate fora non-participating company closely resemble the case
of no consolidation as described in equations (10) to (18). The onlydierence
is, in
the bad case the taxable loss of the subsidiary in periodt+1 can beoset
against
the taxable profits at the parent level.11 Therefore the EATR and the necessary
return in the good outcome with voluntary consolidation can be written as.
EATR
v.c.
=
(
EATR
l.c.
if=1
R
[
qR
GOOD
+(1q)(R
BAD
j
b+b+
1+
) ]
p/ (1+r)
if=0
(32)
eg
v.c.
=
eg
l.c.
if=1
eg
n.c.
+
1
q
q
j
b+b+
1+
if=0
(33)
To obtain a measure of the cost of capital and the EMTR equation (33) needs
again be inserted in equation (18) and equation (8).
Consolidation with an already existing subsidiary: In the subsequent
analysis we will use a large firm-leveldataset which also allows us to identify whether
a corporate groups is already operating a potential host country. For these compa-
niesitis reasonable to assume that they are conducting their investment through
the existing companies. Assuming that the existing business is profitable enough,
the initial depreciation and any potential losses areoset
against the subsidiary
profits. Hence the EATR and the cost of capital can be simplified to the case as
described in equations (7) and (9).12
2.4 Eective
tax rates under formula apportionment
The calculation ofeective
tax rates under a formula apportionment quite closely
resembles the standard unilateraleective
tax rates. This is due to the fact that first
year depreciation allowances are nowoset
against the consolidated group profit,
11
The underlying assumption is, that if the firm did chose not to participate in periodt, it chose
again whether to participate in periodt+1.
12
The exact calculations are available from the authors upon request.
8
regardless the profit situation at the new subsidiary. The main and most important
dierence
is the applicable corporate tax rate, which is now not only depending
on the country where the investment takes place. In the extreme case where the
investment is small enough to not significantly alter the distribution of the factors
used to determine the apportionment, the applicable tax rate independent from the
tax rate in the host country. Define 0
1 as the share of the new investment
in terms of the existing investment the applicable tax rate can be written as
FA
n
=
n
+ (1
)
l
X
m=1
m
µX
m
(34)
where mnowdenotes all the countries the corporate group operates in andµX
m
the share of the factor Xemployedin countrym.13
µX
m
=
X
m
P
l
m=1
X
m
(35)
In line with Ag´undez-Garc´ıa (2006) we consider the following apportionment
factorsasX: number of employees, cost of employees, turnover, total assets anda
measure of value added. Further we also follow Ag´undez-Garc´ıa (2006) in assuming
that each Member State is still allowed to define its own tax base and tax rate. This
implies that the net present value of the tax allowances is now not only host country
specific, as it is calculated withFA
n
. Further, as all the income is consolidated across
countries and then apportioned, there is no longer any dividend tax. Using the
applicable tax rate as defined in (34), the income stream from (3) can be rewritten
as
RFA
GOOD
=
(1
AFA
n
) +
1
1+
(1+)(g+)(1FA
n
)
+ (1+)(1)(1 AFA
n
)
(36)
The tax eects
of the other forms of finance are now significantly reduced, as the
lending from the parent to the subsidiary cancels out because of the consolidation.
The only remaining other form of finance that influences the income stream is
the outside debt raised at the parent level.14 Note that these additional costs are
identical for the good and the bad outcome.
FFA=
(1
FA
n
n
)
1+
i (1
FA
n
) dB
n
(37)
In the bad outcome the first year depreciation allowance
n
can be oset
against
the consolidated group profit at the applicable tax rateFA
n
. However this claimed
depreciation reduces the taxable loss (increases the taxable profit) in periodt+1
because of the balancing charge. In contrast to the separate accounting system,
the taxable income is now subject to the factor weighted tax rateFA
n
regardless
whether the income is positive or negative. Therefore the net present value of the
income stream in the bad outcome can be written as
RFA
BAD
=FA
n
n
+R
BAD
T
n
FA
n
(38)
Regardless the state of the outcome, there is now an additional feedbackeect
of the investment, as it may alter the applicable tax rate for the existing operations.
13
For simplicity we assume that the new investment is proportional in the apportionment factor(s).
Otherwise the becomes location specific and cannot be moved in front of the summation.
14
Assuming that interest rates are equal across countries, it is in fact irrelevant where the debt
is raised.
9
Define
R
as the sum of the changes in the income streams in all existing locations
due to the new investment with a relative size of.
R
=
l
X
m=1
R
m
(FA|
>
0) R
m
(FA|=0)
(39)
This requires an assumption about the income stream of the existing operations.
However, as the whole concept of this sort ofeective
tax rate calculations is based
on a hypothetical investment project in the future, we assume that the existing
operations takeasimilarform. We therefore use a assume that the existing opera-
tionsprovidea certain income stream withp=0.2. Hence the R
m
are given through
equation (3).
Theeective
average tax rate under a formula apportionment system is then,
similar to the to the calculation in equation (16), given through the probability
weighted average of (36) and (38) plus the additional feedback term as defined in
(39)
EATR
FA
=
R
qRFA
GOOD
+ (1q) RFA
BAD
+1
R
p/ (1+r)
(40)
The necessary return in the good outcome is again obtained in solving the ex-
pectedrateof return forgandcanbe written as
eg
FA
=
(r+)(1 AFA
n
)
1
FA
n
1+r
1
FA
n
FFA+
1
q
q
(RFA
BAD
+FFA) +
1
q
R
(41)
The EMTRforformula apportionment is again calculated using (41) in equations
(18) and (8).
2.5 Firm-specificeective
tax rates
Usuallyeective
tax rates are calculated for countries, or at most country pairs.
However, these hypothetical scenarios fall short to capture all the potentialeects
of a loss consolidation and a potential formula apportionment system. Therefore
we follow the concept of Egger et al. (2008) and use firm-specific information about
asset, finance and ownership structure from the ORBIS database.
Definet
i
,i
i
ands
i
as the firm specific share of tangible fixed assets TFA
i
,
intangible fixed assets IFA
i
and stocks STO
i
over the sum of them.
t
i
=
TFA
i
IFA
i
+TFA
i
+STO
i
i
i
=
IFA
i
IFA
i
+TFA
i
+STO
i
s
i
=
STO
i
IFA
i
+TFA
i
+STO
i
(42)
In order to exploit more information from the national tax laws we need to
further distinguish between various forms of tangible fixed assets. Therefore we use
10
information about the industry and size specific structure of capital assets froma
Canadian study by McKenzie, Mansourand Brule (1998).15 Denote thekindustry
and size specific weights withb
k
for buildings,m
k
for machineryandl
k
for land
the firm specific share of tangible fixed assetst
i
can be decomposed into
b
i
=
t
i
b
k
m
i
=
t
i
m
k
l
i
=
t
i
l
k
(43)
Note, that the way we calculate these shares ensures that they add up to 1, i.e.
thatb
i
+m
i
+l
i
+i
i
+s
i
=1 holds. The weights can now be used to calculate
the firm specific tax depreciation for the first year
i
and the firm specificNPVof
the depreciation allowances A
i
i
=b
n
b
i
+m
n
m
i
+l
n
l
i
+i
n
i
i
+s
n
s
i
(44)
A
i
=Ab
n
b
i
+Am
n
m
i
+Al
n
l
i
+Ai
n
i
i
+As
n
s
i
(45)
Accordingly the parameter for the economic depreciation
i
needs to be weighted
firm specifically.
To calculate the firm specific finance structure we define the share of debt finance
dB
n
at the subsidiary level as the sum of current liabilities CL
i
and non-current
liabilitiesNL
i
over total assets TA
i
dB
n
=dB
i
=
CL
i
+NL
i
TA
i
(46)
It is now possible to calculate all the above introduced measures of the cost of
capital and EATRs at the individual firm level by replacing A
n
,
n
, anddB
n
with their firm-specific counterparts A
i
,
i
,
i
anddB
i
.
3
Data
The firm level data is from largest available set of firm level data Orbis, provided
by the Bureau van Dijk. We start with 930,588 companies which report total assets
higher than 2 million Euros in two consequent years 2001 to 2005.16 As we
use information at the firm level only for weighting and are not interested infirm
behaviour, we average the data across the years and use only the cross-sectional
information. This sample, including non-European companies is then used to iden-
tifythegroup structures. A company is treated as part of a group if the database
reports a majority shareholder (more than 50%director indirectshareholding)
that is within our sample. Further a company is considered to be part of a group if
the database reportsaglobal owner which itself hasa BvD identification number.
15
See Eggeretal. (2008) for the matching of the industry codes and for the used weights.
16
The datasetisvery similar to the one in Devereuxand Loretz (2007), we therefore only present
it very briefly here.
11
Table
1:
Descriptive
Statistics:
Country
coverage
and
average
weights
Observations
Tax
rate
buildings
machinery
land
inventories
intangibles
leverage
Country
all
firms
parents
(
n
)
(
FA
i
)
(
b
i
)
(
m
i
)
(
l
i
)
(
s
i
)
(
i
i
)
(
dB
i
)
Austria
1,461
128
25.0%
27.2%
0.220
0.279
0.074
0.374
0.052
0.683
Belgium
10,761
371
34.0%
32.7%
0.239
0.257
0.095
0.364
0.046
0.686
Bulgaria
1,101
0
10.0%
n.a.
0.239
0.305
0.090
0.337
0.028
0.631
Cyprus
92
4
10.0%
15.7%
0.363
0.198
0.128
0.199
0.113
0.419
Czech
Republic
6,459
12
24.0%
24.4%
0.239
0.319
0.116
0.305
0.021
0.524
Germany
14,054
669
36.4%
33.3%
0.247
0.241
0.099
0.368
0.044
0.676
Denmark
5,073
254
25.0%
26.5%
0.339
0.241
0.151
0.213
0.056
0.590
Spain
62,650
305
33.0%
32.6%
0.231
0.214
0.101
0.360
0.093
0.610
Estonia
1,025
15
22.0%
20.8%
0.314
0.273
0.135
0.264
0.014
0.543
Finland
5,242
135
26.0%
26.5%
0.240
0.254
0.094
0.335
0.076
0.539
France
48,199
621
33.3%
32.9%
0.194
0.179
0.081
0.402
0.144
0.630
United
Kingdom
22,353
425
30.0%
30.0%
0.283
0.248
0.116
0.309
0.044
0.636
Greece
7,940
32
25.0%
24.1%
0.212
0.268
0.079
0.393
0.049
0.624
Hungary
4,118
9
16.0%
18.3%
0.221
0.315
0.099
0.332
0.032
0.566
Ireland
1,126
44
12.5%
24.9%
0.286
0.231
0.112
0.338
0.032
0.601
Italy
95,158
568
37.3%
35.6%
0.177
0.228
0.072
0.440
0.083
0.773
Lithuania
866
5
18.0%
21.0%
0.206
0.342
0.076
0.366
0.009
0.529
Luxembourg
493
42
29.6%
30.8%
0.235
0.235
0.087
0.365
0.078
0.656
Latvia
724
2
15.0%
16.7%
0.235
0.329
0.091
0.333
0.012
0.592
Malta
107
0
35.0%
n.a.
0.316
0.180
0.127
0.371
0.006
0.555
Netherlands
3,949
486
25.5%
30.3%
0.275
0.220
0.111
0.334
0.061
0.669
Poland
7,710
14
19.0%
20.8%
0.264
0.329
0.101
0.277
0.028
0.512
Portugal
6,060
48
26.5%
27.6%
0.198
0.300
0.075
0.397
0.030
0.695
Romania
2,552
0
16.0%
n.a.
0.221
0.374
0.082
0.303
0.019
0.622
Slovak
Republic
1,641
4
19.0%
20.6%
0.270
0.307
0.149
0.258
0.016
0.498
Slovenia
1,786
7
23.0%
23.5%
0.234
0.386
0.082
0.271
0.028
0.556
Sweden
10,742
367
28.0%
28.0%
0.270
0.236
0.113
0.333
0.046
0.615
Europe
323,442
4,567
24
.
2%
26
.
0%
0
.
218
0
.
233
0
.
091
0
.
379
0
.
079
0
.
664
Notes:
n
denotes
the
statutory
corporate
tax
rates
including
local
profit
taxes.
FA
i
denotes
the
applicable
tax
rates
under
a
formula
apportionment
system.
Only
multinational
companies
included.
12
We include all 27 EU Member States, which leaves us with a sample of 410,222
companies for which all the necessary data is reported.17 We then average the
114,853 companies within corporate groups across the 28,703 groups to end up with
323,442 observations. Each of these observations is then attributed to the country
of the headquarter company, unless the corporate owner is outside Europe. In these
cases we treat the national groups within this multinational group as individual
companies. Table 1 summarises the country coverage and the relevant variables
that are used for the weighting of the firm specific EATRs.
In total our sample includes 4,567 corporate groups that operate in more than
one European country. For these corporate groups we calculate the applicable
tax rate under a formula apportionment system. For this purpose we weight the
corporate tax rate in the countries the group operates with the shares of the apportionment
factors employed there. Like in Devereuxand Loretz (2007) we use an
composite apportionment factor with on sixth number of employees, on sixth cost
of employees, one third turnover and one third total assets.
The applicable tax rates under the formula apportionment are also depending
on the already existing investment inaditionto location of the new investment.
The latter is only true, if the new investment is large enough to alter the overall
distribution for the relevant apportionment factors. As abase case we assume that
this is not the case, i.e. we assume that is zero. This assumption we be changed
in the later on in the paper. Regardless the relative size of the new investment
the applicable tax rate under a formula apportionment must be a weighted average
of the statutory corporate tax rates in the individual Member States. Therefore
they are bound with the highest and the lowest rate in the EU and the formula
apportionment rates can only be equal or higher in the lowest tax country Cyprus.
Equally the rate under a formula apportionment system must be no higher than
the statutory rate in Italy, which has the highest statutory tax rate. We therefore
expect that companies headquartered in a high tax country would benefit froma
lower tax rate under a formula apportionment system, while the firms located ina
low tax country would face an increased tax rate. This holds true, as companies
in all the high tax countries, Germany, Italy, Spain, France, Belgium would benefit
from a lower tax rate. In contrast the tax rates for the firms located in low tax
countries as Cyprus, Ireland, Latvia, Lithuania and Hungary would increase. It is
noteworthy that the unweighted average is higher under formula apportionment,
which reflects that the economic activity within the corporate groups tends to be
more within high tax countries.18
In comparison to the statutory corporate tax rates, the overall distribution of
the tax rates is necessarily less dispersed, but the same countries emerge as high tax
countries. Despite only looking at multinational companies, the tax rate applicable
under a formula apportionment system is very similar to the statutory tax rate in
the headquarter country. This mirrors the fact that the domestic activities ofa
multinational dominate, even more so in large countries as France, Germany, Italy
or the United Kingdom.
17
In addition to all observations that report missing values for the relevant variables we also
exclude corporate groups that report more than 100%debtor report zero in all three asset variables,
i.e. stocks, tangibles and intangibles.
18
This is in line with previous research, e.g. Fuestetal. (2007) and Devereuxand Loretz
(2007). However compared to this studies we do not relate the apportionment factors to profits
and therefore we cannot infer from this that profit shifting takes place.
13
4
Results
We are mainly interested in the dispersion of the tax burden under the current
and the proposed tax systems and less so in the bilateral tax burden fora specific
country combination. Therefore we only present very summarised results. The next
section includes some numerical results for the EATR sandagraphical presentation
of the results. The results for the cost of capitals are subsequently presented only in
graphical form.19 Finally we also discuss the importance of the existing investment
under a formula apportionment system.
4.1 Eective
Average TaxRates
To geta first impression of the impact of thedierent
tax systems it is useful to
look at the overall dispersion of the tax burden across all firms. We do so for three
dierent
scenarios; for the current system, i.e. without the possibility of interna-
tionallossoset,
fora system of voluntary international loss consolidation without
formula apportionment and fora system with international loss consolidation and
formula apportionment.
Figure 1shows histograms of the EATR sforthesethree scenarios. Each his-
togramshowsthe dispersion of more than 8.7 million tax rates as we calculate the
potential tax burden for 323,442 companies in all 27 European countries. Like in
all the figures, the graphs are arranged as follows. The upperpart of the figure
displays the results for the current system, the middle part for the voluntary loss
oset
and the lower part for the formula apportionment system. Moving from the
top downwards two main changes can be observed. First the distribution shifts left,
because allowing loss consolidation reduces theeective
tax burden. Further the
middle part of Figure 1 show that the introduction of loss consolidation without
formula apportionment would significantly increase the dispersion of the tax burdens.
Note, that because of the fact that the loss consolidation would be voluntary,
the distribution only widens at the lower tail. The lower part of the figure indicates
that the overall tax burden and its dispersion would be significantly reduced under
a formula apportionment system.20
19
Moredetailled results are available from the authors on request.
20
The few tax rates above 40%are due to the particular system in Estonia, not allowing tax
depreciation.
14
0
2
4
6
8
Density
Existing system (EATR
n.c.
)
0
2
4
6
8
Density
-.5
0
Voluntary consolidation (EATR
v.c.
)
0
2
4
6
8
Density
-.5
0
.5
Formula Apportionment (EATR
FA
)
Figure1: Histograms of EATR
15
Table
2:
Summary
of
Results:
Domestic
and
Bilateral
EATR
s
for
domestic,
inbound
and
outbound
investment
domestic
investment
outward
FDI
inward
FDI
current
formula
current
voluntary
formula
current
voluntary
formula
country
system
apportionment
system
consolidation
apportionment
system
consolidation
apportionment
Austria
18.8%
19.3%
27.6%
21.0%
19.6%
30.0%
21.5%
24.17%
Belgium
24.2%
24.1%
30.2%
19.1%
25.9%
38.0%
30.0%
23.90%
Bulgaria
10.1%
7.6%
27.0%
25.0%
7.8%
22.2%
9.3%
23.12%
Cyprus
8.8%
9.4%
28.0%
26.2%
9.1%
17.7%
7.3%
26.12%
Czech
Republic
20.8%
20.3%
31.2%
25.0%
20.0%
29.6%
21.1%
25.04%
Germany
27.3%
26.8%
30.4%
18.1%
27.5%
40.9%
34.0%
23.71%
Denmark
21.3%
20.0%
27.7%
21.3%
20.0%
31.4%
18.0%
25.33%
Spain
27.3%
26.1%
28.4%
18.1%
26.1%
38.1%
30.6%
24.85%
Estonia
22.2%
21.4%
27.8%
22.5%
17.9%
30.3%
23.8%
29.22%
Finland
21.9%
22.0%
27.8%
21.0%
21.3%
32.2%
22.6%
25.78%
France
26.0%
25.9%
29.9%
19.1%
26.2%
38.4%
30.5%
24.17%
United
Kingdom
24.1%
24.1%
35.6%
26.4%
23.4%
36.2%
27.5%
25.87%
Greece
21.7%
19.0%
31.9%
25.2%
19.9%
29.5%
20.6%
23.46%
Hungary
13.4%
13.7%
27.7%
24.3%
13.2%
22.9%
13.0%
25.72%
Ireland
10.0%
12.6%
27.1%
24.6%
12.4%
19.6%
9.1%
25.05%
Italy
25.8%
25.7%
31.2%
18.0%
27.3%
40.5%
34.6%
22.62%
Lithuania
14.0%
14.7%
27.8%
23.9%
15.1%
24.6%
3.8%
24.29%
Luxembourg
21.8%
22.3%
28.1%
19.4%
23.2%
34.1%
25.3%
23.62%
Latvia
11.6%
11.8%
27.6%
24.5%
12.2%
21.7%
4.7%
24.06%
Malta
28.3%
28.3%
40.6%
30.0%
28.7%
39.8%
32.6%
24.12%
Netherlands
21.6%
20.4%
27.5%
20.9%
20.8%
30.6%
20.9%
24.89%
Poland
16.0%
16.2%
29.0%
24.7%
15.9%
25.0%
15.6%
24.66%
Portugal
19.8%
19.4%
28.6%
21.3%
20.3%
31.3%
21.5%
23.68%
Romania
12.5%
12.7%
27.8%
24.4%
12.7%
30.3%
10.2%
25.68%
Slovak
Republic
15.7%
15.8%
28.1%
23.8%
16.0%
24.9%
14.2%
24.41%
Slovenia
18.1%
18.4%
28.0%
22.1%
18.7%
28.2%
18.8%
24.28%
Sweden
22.5%
22.5%
28.1%
20.4%
22.1%
34.0%
24.3%
25.47%
Europe
24
.
6%
24
.
2%
30
.
2%
19
.
9%
24
.
7%
30
.
2%
19
.
9%
24
.
57%
16
To assess the impact on the EATR sin the individual Member States Table 2
compares the average tax burdens for a domestic investment, for inbound and outbound
foreign direct investment. We thereby assume that the domestic investment
is done through the existing profitable parent company, which allows the losses to
beoset.
Further we assume that if the international investment takes place in an
existing subsidiary that the losses can beoset
there. In contrast if the company
has no subsidiaries in the country sofar, we assume that no loss consolidation is
possible under the current system.21 For the domestic investment the current sys-
temandthe voluntary consolidation lead to the same outcome as it is possible to
consolidate domestic losses under the current system. Further, it is always ben-
eficialtouse losses immediately in a domestic subsidiary because the loss carry
forward could only be used against tax same tax rate in the future. Even under formula
apportionment the tax burden for domestic investment changes little, which is
partly due to the fact the majority of our sample are domestic firms, for which the
applicable tax rate remains unchanged. As a result the dierences
in the domestic
EATR acrosscountries persist and are only reduced insignificantly.
Comparing the domestic EATR under the current system with the tax burden
for either outbound or inbound investment it is evident that the lack of international
loss consolidation increases the EATR for international investment. This is at odds
with both capital export neutrality and capital import neutrality because domestic
investment receives more favourable tax treatment. While the tax burden for do-
mesticinvestment are between 8.8%for Cyprus and 28.3%in Malta, the country
averages of EATRs for outbound investment range between 27.0%in Bulgaria to
40.6%in Malta. Similarly the country averages of EATR sforinbound investment
vary from 17.7%in Cyprus to 40.9%in Germany.
Moving to a system of voluntary loss consolidation without formula apportion-
mentwould overcorrect the distortion between domestic and foreign investment,
as foreign investment would receive on average a more favourable tax treatment.
Further, the overall reduction in the tax burden of five percentage points is very
unevenly distributed. In fact outbound investment from high tax countries would
facea significantly reduced tax burden, while outbound investment from low tax
countries would still face a high tax burden. Overall, the spread of the average
EATR foroutbound investment is comparable to the current system, with values
between 18.0%for Italy and 30.0%for Malta. The other side of this phenomenon is
that the attractiveness of low tax countries for inbound investment is amplified. For
countries withacombination of generous depreciation allowances and low statutory
tax rates, for example Lithuania, the average tax burden for inbound investment
would be very low. This leads to an extremely largedierential
between country
averages of EATR for inbound investment, with values as low as 3.8%for Lithua-
niaontheonehand and tax burdens as high as 34.6%in Italy.
A switch to a formula apportionment system would almost eliminate the dier-
encesintheEATR for domestic investment and outbound investment. While the
taxdierential
between domestic and outbound investment is reduced, theeective
tax burden still do largely vary across thedierent
member states. Country averages
range from as little as 7.8%for outbound investment from Bulgaria to 28.7
%for investment from Malta. However, for the inbound investment the dispersion
amongst host countries is significantly reduced, with a lowest average EATR of 22.6
%inItalyanda highest average of 29.2%in Estonia.
21
For simplicity reasons and to allow abetter comparison we also do not allow loss consolidation
in countries like Austria and Denmark, which do allow some form loss consolidation.
17
Capital export neutrality: It is, however, not possible to evaluate thee-
ciencyonlyfrom country averages of EATRs. We therefore fully exploit the firm
level information and examine capital export neutrality and capital ownership neu-
tralityforthe individual companies. From a company perspective capital export
neutrality is given if the investment faces the same tax treatment regardless where it
takes place. Technically, this can be measured as the variation between the EATRs
fora firm in each country, including its home country. The lower this variation is,
the better capital export neutrality is fulfilled. The first three columns of Table
3 summarise the standard deviation between the EATR sforeachofthe 323,442
firms under the various tax systems on a country by country basis. Figure 2 further
shows histograms of these standard deviations for all firms.
18
0
20
40
60
80
100
Density
0
Existing system (sd[EATR
n.c.
])
0
20
40
60
80
100
Density
0
.05
.1
.15
.2
Voluntary consolidation (sd[EATR
v.c.
])
0
20
40
60
80
100
Density
0
.05
.1
.15
.2
.25
Formula apportionment (sd[EATR
FA
])
Figure2: Histograms of standard deviation of EATR
19
Table
3:
Summary
of
Results:
Standard
deviation,
average
and
minimum
EATR
s
for
the
individual
corporations
Standard
deviation
average
EATR
minimum
EATR
current
voluntary
formula
current
voluntary
formula
current
voluntary
formula
country
system
consolidation
apportionment
system
consolidation
apportionment
system
consolidation
apportionment
Austria
0.081
0.098
0.012
27.3%
20.9%
19.6%
13.0%
3.3%
18.1%
Belgium
0.072
0.100
0.016
29.9%
19.3%
25.8%
16.2%
-0.5%
23.8%
Bulgaria
0.080
0.078
0.005
26.3%
24.4%
7.8%
9.9%
9.1%
7.2%
Cyprus
0.083
0.082
0.005
27.3%
25.6%
9.1%
8.8%
8.8%
8.3%
Czech
Republic
0.049
0.062
0.011
30.8%
24.9%
20.0%
20.8%
14.2%
18.6%
Germany
0.069
0.099
0.017
30.2%
18.5%
27.5%
16.5%
-2.1%
25.4%
Denmark
0.077
0.094
0.012
27.5%
21.3%
20.0%
12.7%
3.8%
18.3%
Spain
0.075
0.103
0.016
28.4%
18.4%
26.1%
13.7%
-1.9%
24.1%
Estonia
0.078
0.089
0.010
27.6%
22.5%
18.0%
12.3%
5.5%
16.6%
Finland
0.078
0.098
0.013
27.6%
21.0%
21.4%
12.6%
3.0%
19.7%
France
0.073
0.103
0.017
29.8%
19.4%
26.1%
15.6%
-1.7%
24.1%
United
Kingdom
0.030
0.049
0.014
35.2%
26.3%
23.4%
24.1%
14.1%
21.5%
Greece
0.044
0.056
0.011
31.6%
25.1%
19.8%
21.7%
14.9%
18.4%
Hungary
0.080
0.082
0.007
27.2%
23.9%
13.2%
11.7%
7.7%
12.3%
Ireland
0.078
0.077
0.007
26.5%
24.1%
12.4%
10.0%
9.9%
11.5%
Italy
0.067
0.096
0.019
31.0%
18.3%
27.2%
17.7%
-2.4%
25.1%
Lithuania
0.085
0.089
0.008
27.3%
23.6%
15.1%
11.4%
5.9%
14.1%
Luxembourg
0.075
0.100
0.015
27.9%
19.5%
23.2%
13.7%
0.9%
21.3%
Latvia
0.080
0.081
0.007
27.0%
24.1%
12.2%
10.8%
8.0%
11.3%
Malta
0.026
0.036
0.015
40.1%
29.9%
28.7%
28.3%
18.0%
26.6%
Netherlands
0.076
0.094
0.013
27.3%
20.9%
20.7%
13.3%
4.3%
19.1%
Poland
0.066
0.072
0.009
28.5%
24.3%
15.9%
16.0%
13.3%
14.7%
Portugal
0.076
0.092
0.012
28.3%
21.2%
20.2%
13.8%
3.8%
18.7%
Romania
0.074
0.076
0.008
27.2%
24.0%
12.7%
12.5%
12.1%
11.7%
Slovak
Republic
0.081
0.087
0.008
27.6%
23.5%
16.0%
11.8%
6.6%
14.9%
Slovenia
0.081
0.097
0.011
27.6%
21.9%
18.7%
12.5%
4.6%
17.3%
Sweden
0.075
0.095
0.013
27.9%
20.5%
22.1%
13.2%
2.6%
20.3%
Europe
0
.
068
0
.
093
0
.
016
30
.
0%
20
.
0%
24
.
7%
16
.
3%
1
.
2%
22
.
8%
20
In the top part of Figure 2, showing the standard deviation under the current
system, a very peculiar distribution of the standard deviation can be observed. In
fact, there are three distributions within this histogram. Starting from the left,
the first peak at a standard deviation of around 0.03 there are firms with their
headquarter in countries with a credit system and a relatively high tax rate, like
the United Kingdom, Greece or Malta. The smaller second peak at a standard
deviation of approximately 0.05 represents firms in a country with a credit system
and a moderate tax rate, e.g. the Czech Republic or Poland. The large bulk of
companies is located in either a country with an exemption system, or in a country
with a relative low corporate tax rate, which quasi exempts foreign income for most
outbound investment. These countries have a standard deviation of their EATRs
between0.06 and 0.1. Therefore, under the current system, capital export neu-
tralityismostly fulfilled for high tax credit countries, but less so for exemption
countries.
The middle part of Figure 2 displays the standard deviation for each firm under
a voluntary loss consolidation. Compared to the current system the large increase
in the spread of tax rates for each country demonstrates that such a tax reform
would represent amove away from capital export neutrality. This large variation of
tax rates stems from the fact that low tax countries will not gain significantly from
loss consolidation, while the high tax countries will benefit most. Therefore, the
firms that face a low domestic tax burden, will face relatively high tax burdens for
outbound investment. At the same time firms with a relatively high domestic tax
burden will have increased low tax opportunities. This results in aaveragestan-
darddeviationof almost 0.1, with values up to more than 0.2 for some firms. The
three distinct peaks for thedierent
combinations of double taxation and tax rate
combinations are no longer apparent. Nevertheless, in Table 3 the these countries
can still be distinguished having a lower average standard deviation.
The lower part of Figure 2 presents the standard deviations under a formula
apportionment system. The standard deviations are now significantly reduced and
less dispersed. The peak is now at approximately 0.02 which signals the improve-
mentintermsof capital export neutrality because of this tax reform. In the third
column in Table 3 it can be now seen that firms located in the countries with the
lowest corporate tax rate have the smallest variation in their EATRs. This shows
that the eective
tax burden is largely dominated by the statutory tax rate of the
headquarter country.
Capital Ownership Neutrality: To assess the ability of the tax reforms to
improve capital ownership neutrality we follow two approaches. First we followa
more conservative approach and stipulate that the requirements for capital ownership
neutrality are met if all potential competitors face the on average the same
tax burden. This is measured in comparing the average EATR for all 27 invest-
mentlocations-the home country and all the other 26 European countries-for
all firms. The results are presented graphically in the upper row of Figure 3 and
on a more detailled country by country basis in the middle three columns of Table 3.
21
0
5
10
15
20
Density
-.6
-.4
-.2
0
.2
.4
0
5
10
15
20
Density
-.6
-.4
-.2
0
.2
.4
10
15
20
Existing
system
(
av
[
EATR
n.c.
])
Voluntary
consolidation
(
av
[
EATR
v.c.
])
Formula
Apportionment
(
av
[
EATR
FA
])
Averages
of
EATR
s
0
5
10
15
20
Density
-.6
-.4
-.2
0
.2
.4
0
5
10
15
20
Density
-.6
-.4
-.2
0
.2
.4
10
15
20
Existing
system
(
min
[
EATR
n.c.
])
Voluntary
consolidation
(
min
[
EATR
v.c.
])
Formula
Apportionment
(
min
[
EATR
FA
])
Minimums
of
EATR
s
Figure
3:
Histograms
of
average
and
minimum
EATR
22
Additionally we also examine a stronger requirement for capital ownership neutrality.
We assume that multinational companies can engage in tax planning or
profit shifting. This implies that each corporation would setup the subsidiary in
the country with the lowest EATR.22 Therefore capital ownership neutrality is only
achieved if the tax optimal location decisions lead to an identical tax treatment for
all potential competitors. In consequence we measure the degree of capital own-
ershipneutrality through comparing the lowest possible EATR for each company.
Again, the results are presented both graphically in the lower row of Figure 3 and
on a country per country basis in the last three columns of Table 3.
The upper left of Figure 3 presents the distribution of the averages of the EATR
in all potential location decisions for each firm under the current system without
loss consolidation. Most of the firms face an average EATR between 25%and
35%andboththeupper and lower tail are relatively short. This implies that the
current system performs reasonably well in terms of capital ownership neutrality.
The country averages in Table 3 strengthen this impression, as the only moderately
vary between 26.3%in Bulgaria and 40.1%in Malta.
In comparison, the lower left part of Figure 3 depicts the distribution of the low-
estpossibleEATR under the current system for each firm. Relative to the average
EATR intheupper row the distribution is shifted to the left and more dispersed.
This is also reflected in the country averages in Table 3 where thedierences
across
cou