Foreign Direct Investment and Transfer Pricing

Jay Pil Choiy Taiji Furusawa z

Jota Ishikawax

June 12, 2017

Abstract

This paper analyzes an MNE’s incentives to manipulate an internal transfer price to

take advantage of tax di¤erences across countries. We …rst consider a monopoly case
and derives conditions under which FDI takes place and show that tax-induced FDI
can entail ine¢ cient internal production. With imperfect competition we show that
the internal transfer price has additional strategic e¤ects that further strengthen
incentives to in‡ate the transfer price at the expense of the rival …rm’s pro…ts. The
tax-induced FDI by the MNE has spillover e¤ects that reduce tax revenues from
other domestic …rms as well as the MNE. We also explore implications of the arm’s
length principle and import tari¤s to mitigate this problem.

JEL Codes: D4, L12, L4, L43, L51, L52

Key Words: FDI, Transfer Pricing, Arm’s Length Principle

We thank Pascalis Raimondos, Yuka Ohno, Jay Wilson and participants in various conferences and
seminars for valuable dicussions and comments. This research was initiated duirng the …rst author’s visit
to Hitotsubashi University whose hospitality is greatly appreciated.
y
Department of Economics, Michigan State University and Hitotsubashi Institute for Advanced Study,
Hitotsubashi University. E-mail: choijay@msu.edu.
z
Faculty of Economics, Hitotsubashi University, Kunitachi, Tokyo 186-8601, Japan. E-mail:
taiji.furusawa@r.hit-u.ac.jp.
x
Faculty of Economics, Hitotsubashi University, Kunitachi, Tokyo 186-8601, Japan. E-mail:
jota@econ.hit-u.ac.jp.
1 Introduction

This paper analyzes an MNE’s incentives to manipulate an internal transfer price to take
advantage of tax di¤erences across countries. We …rst consider a monopoly case and
derives conditions under which FDI takes place and show that tax-induced FDI can entail
ine¢ cient internal production. With imperfect competition we show that the internal
transfer price has additional strategic e¤ects that further strengthen incentives to in‡ate
the transfer price at the expense of the rival …rm’s pro…ts. The tax-induced FDI by the
MNE has spillover e¤ects that reduce tax revenues from other domestic …rms as well as
the MNE. We also explore implications of the arm’s length principle and import tari¤s
to mitigate this problem.
It has been well documented that MNEs engage in tax manipulation to reduce their
tax obligations by shifting their pro…ts from high tax countries to low tax jurisdictions
(see Hines and Rice, 1994 and Bauer and Langenmayr, 2013). For instance, inspections
by the Vietnamese tax authorities found that "the most common trick played by FDI
enterprises to evade taxes was hiking up prices of input materials and lowering export
prices to make losses or reduce pro…ts in books."1 In addition, Egger et al. (2010) …nd
that an average subsidiary of a multinational corporation pays about 32% less tax in a
high tax country than a similar domestically-owned …rms.
To analyze tax-induced FDI and its welfare implications, we consider a very stylized
simple set-up of two countries with di¤erent corporate tax rates. To …x the scenario, we
…rst consider a setting in which the monopolistic …nal good producer is located in Home
country (H) with a higher tax rate whereas its input can be more cheaply produced in
Foreign country (F ) with a lower tax rate. For instance, the input is labor-intensive and
country F has a lower wage. Alternatively, the necessary input is a natural resource that
is available only in country F . In this scenario, the input is needed to be procured from
F , but there are two ways to do it. It can be outsourced from outside …rms in F; or can
be produced internally with FDI. Not surprisingly, we show that FDI can be used even if
it is less e¢ cient in producing the input because it can be used as a vehicle to lessen its
tax burden with an in‡ated internal price when F has a lower tax rate.
If there is no government oversight on internal exchanges within the …rm, the MNE will
shift all pro…ts to the country with a lower tax rate via transfer price. Governments thus
impose transfer pricing rules (TPRs) to control tax manipulation. The standard practice
1
http://vietnamlawmagazine.vn/transfer-pricing-unbridled-at-fdi-enterprises-4608.html

1
is to stipulate that internal transfer prices follow the so-called "Arm’s Length Principle"
(ALP), which requires intra…rm transfer prices to meet the arm’s length standard, that
is, the transfer price should not deviate from the price two independent …rms would
trade at. Currently, the ALP is the international transfer pricing principle that OECD
member countries have agreed should be used for tax purposes by MNE groups and tax
administrations.
The basic approach of the ALP is that the members of an MNE group should be
treated "as operating as separate entities rather than as inseparable parts of a single
uni…ed business" and the controlled internal transfer price should mimic the market price
that would be obtained in comparable uncontrolled transactions at arm’s length. This
kind of “comparability analysis”, is at the heart of the application of the arm’s length
principle. For instance, the 2010 OECD Transfer Pricing Guidelines for Multinational
Enterprises and Tax Administrations 2010 states that the comparable uncontrolled price
(CUP) method

"compares the price charged for property or services transferred in a con-

trolled transaction to the price charged for property or services transferred in
a comparable uncontrolled transaction in comparable circumstances. If there
is any di¤erence between the two prices, this may indicate that the conditions
of the commercial and …nancial relations of the associated enterprises are not
arm’s length, and that the price in the uncontrolled transaction may need to
be substituted for the price in the controlled transaction." (p. 63)

As the CUP method is the most direct and reliable, it is the preferred method to
apply the ALP. In practice, however, it may be di¢ cult to …nd a transaction between
independent enterprises that is similar enough. This would be particularly so in the
monopoly context where the required input is demanded only by the monopolist and
there is no comparable input market available. In such a case, there are other methods
suggested to apply the ALP. In our theoretical set-up, we assume that the "cost plus
method" which mandates that the transfer price should re‡ect the production cost of
the input internally transacted.2 However, the true production cost is typically non-
observable to tax authorities and hard to ascertain. As a result, it can be manipulated
at certain costs.
2
Other methods suggested include the resale price method, the transactional net margin method, and
the transactional pro…t split method. See OECD (2010) for more details.

2
 To analyze incentives to engage in FDI and how the internal price is determined
when the cost plus method is used due to the absence of comparable transactions in the
market, we introduce "concealment costs." More speci…cally, when an MNE’s internal
price deviates from its true marginal cost in the presence of the ALP with the cost plus
method, there are costs to avoid such institutional constraints on the internal transfer
price. These costs can be literally concealment costs to keep two separate books or can
re‡ect expected punishment for the deviation as in Kant (1988). The MNE thus trade-
o¤s potential tax bene…ts against concealment costs in its choice of the optimal transfer
price. We show that the optimal transfer price is equivalent to the minimization of what
we call "virtual marginal cost" and this characterization provides a very simple condition
for the optimality of FDI vis-a-vis outsourcing if the concealment cost is linear in the
quantity of inputs internally transferred.
If concealment costs are convex in the quantity of inputs internally transferred, there
may be incentives for the MNE to engage in dual sourcing, that is, part of the required
input is internally produced with FDI whereas the rest is outsourced. Outsourcing,
however, creates a benchmark transaction against the internal transfer can be compared.
As a result, a dual sourcing strategy may provide the tax authority with the ability to
identify a comparable market price and adopt the CUP method as an application of
the ALP. In such a case, we demonstrate the imposition of the CUP method with dual
sourcing can have unintended consequences and detrimental e¤ects if it triggers the MNE’s
sourcing decision from dual sourcing to internal sourcing only.
We also analyze import tari¤ as a countermeasure against potential tax shifting. Even
though import tari¤ can be completely o¤set the incentives to engage in in‡ated transfer
price for tax manipulation purposes, we show that the optimal import tari¤ may be set
to allow tax manipulation to some extent. The reason is that the tax manipulation by
the MNE leads to more production in the domestic market which can alleviate allocative
ine¢ ciency due to monopoly power.
We then extend our analysis to oligopolistic market structure in the domestic market.
As the MNE has incentives to produce more for pro…t shifting motives, it can have strate-
gic e¤ects vis-a-vis its rival …rms in the …nal produce market. As a result, the rival …rms
reduce its outputs and their pro…ts su¤er. This implies that tax-induced FDI by the
MNE has spillover e¤ects that reduce tax revenues from other domestic …rms as well as
the MNE. We also consider implications of ALP when the input supplier in country F is
monopolistic. If the input purchased by the rival …rm is considered as a comparable input

3
used by the MNE, then the price set by the foreign supplier can a¤ect the internal price
of the MNE with ALP. Thus, the imposition of ALP can have implications of strategic
price setting of the monopolistic supplier of input in F .
Horst (1971) initiated the theory of multinational …rms in the presence of di¤erent
tari¤ and tax rates across countries and explores the pro…t-maximizing strategy for a
monopolistic …rm selling to two national markets, that is, how much to be produced in
each country and what would be the optimal transfer price for goods exported from parent
to subsidiary. Horst (1971) and subsequent papers (such as Batra and Hadar (1979) and
Itagaki (1979, 1981)) show that MNE’s optimum price would be either the highest or the
lowest possible allowed by the limits of government rules and regulations, depending on
tax and tari¤ schedules across countries.3 Kant (1988) shows how an interior transfer
price can be derived endogenously in the presence of so-called "concealment cost."
Bond also analyzes optimal transfer pricing when branches of a vertically integrated
enterprise are located in multiple jurisdictions with di¤erent tax rates. As in Hirshleifer
(1956), he assumes that decision-making across branches is decentralized and the transfer
prices in his model is chosen to align the production decisions of the various divisions. In
our paper, we assume that the decision making process is centralized.
Kato and Okoshi (2017) consider the optimal location of production facilities in the
presence of tax di¤erences across countries and how the ALP principle can impact the
location choice.
Samuelson (1982) is the …rst one to point out that for an MNE subject to the ALP
principle, the arm’s length reference price itself can be partially determined by the …rm’s
activities. In a similar vein, Gresik and Osmundsen (2008) consider transfer pricing
in a vertically integrated industries in the absence of transactions between independent
entities. More speci…cally, they examine the implications of arm’s length principle as a
transfer price regulation when all …rms are vertically integrated and the only source of
comparable data may be from transactions between a¢ liated …rms. In our framework
with imperfect competition in both the upstream and downstream markets, the reference
price for an MNE is determined by an outsider, but the outsider …rm recognizes the
strategic e¤ects of its price decision on its input demand via the transfer price of the
MNE. It is shown that the outsider has incentives to set a lower price compared to the
3
In contrast to Horst (1971) who assumes that output decisions are centralized, Bond (1980) considers
a situation in which decision making is decentralized. He shows that the optimal transfer prices trade
o¤ the gain from tax avoidance against the e¢ ciency losses associated from resource misallocation.

4
case of no linkage via the transfer price.
The rest of the paper is organized in the following way. Section 2 introduces the basic
set-up of the monopoly model with FDI and transfer pricing. We …rst analyze the optimal
transfer price with the concealment cost and the incentives to engage in FDI due to tax
di¤erential between the source and destination countries. We then explore implications of
such FDI for the e¢ ciency of global sourcing and identify the wedge between the e¢ cient
outcome and the market equilibrium and how such decision can be in‡uenced by the
imposition of ALP. Section 3 considers import tari¤s as a countermeasure against pro…t-
shifting. Section 4 extends the analysis to a duopoly setting to explore implications of
strategic interactions in the …nal good market. We show that pro…t-shifting strategy
of the multinational …rm has further consequences for tax revenues from other …rms due
to strategic e¤ects. We also show how the market price can be endogenized with the
imposition of ALP. Section 5 concludes the paper.

2 The Monopoly Model of FDI and Transfer Pricing

2.1 The Basic Set-up

There are two countries, Home and Foreign, with di¤erent tax rates with t and e t, respec-
tively. We assume a monopolistic …nal good producer. We assume that Home (denoted
as H) is a high tax rate country in which the headquarter that produces …nal goods is
immobile and tied to consumer markets in H while inputs are more cheaply produced in
Foreign (denoted as F ) with a lower tax rate, that is, t > e
t. The monopolist located in
H have two possible ways to procure its essential input from F . There is a competitive
open market from which the input can be procured at the price of $ (later we consider
the external source with market power and endogenize $). Alternatively, it can be an
MNE by setting up its own input production plant in F with FDI. We assume a constant
returns technology. In such a case, its input production cost is given by c. The MNE
can choose an internal transfer price ( ) when its foreign subsidiary supply its input to
the headquarter …rm that produces the …nal good. Without any tax rate di¤erential be-
tween the two countries, the MNE’s optimal internal transaction price for the input is
simply its marginal production cost of c in order to eliminate any double marginalization
problem. However, with di¤erent tax rates between H and F , the MNE can choose an
internal transfer price ( ) as a mechanism to shift pro…ts to minimize its tax burden. In

5
 Figure 1: Monopolistic MNE with Transfer Price

particular, the monopolistic MNE solves the following problem

M ax e = (1 t) (q; ) + (1 e
t) ( c)q (1)
q | {z } | {z }
Downstream Pro…ts Upstream Pro…ts

where (q; ) = [P (q) ]q and P (q) is the downward sloping inverse demand function
facing the monopolist. Figure 1 describes the basic set-up.

Note that the objective function of the monopolist can be rewritten as

e = (1 t)[P (q) ]q;

where
(1 e
t)c (t e t)
= (2)
1 t
That is, the MNE facing di¤erent tax rates across countries behaves as if its marginal
production cost were , which can be considered as the MNE’s virtual MC of production
adjusted for transfer price induced by di¤erential tax rates across countries. As the MNE’s
pro…t decreases in , the monopolist’s optimal choice of is equivalent to the choice of
that minimizes . Note that is decreasing in its internal transfer price because it
can be used as a vehicle to shift pro…t from the high tax country H to the low tax rate

6
country F: As pointed out by Horst (1971), it immediately follows that the optimal choice
is to set as high as possible potentially subject to the constraint that the downstream
headquarter pro…t cannot be negative. Otherwise, this implies that H country make up
any losses incurred by the headquarter with a subsidy up to the rate of t.4
In reality, however, there are restrictions that would prevent the choice from being
a corner solution and limit the MNE’s pro…t shifting motives. We consider two such
mechanisms and analyze how they a¤ect the MNE’s behavior and resource allocations.
The …rst one is concealment costs. We analyze concealment costs and ALP, respectively,
in subsections 2.2 and 2.3.

2.2 Pro…t-Shifting Transfer Pricing with Concealment Costs

As shown in the previous subsection, without any external or regulatory restriction on the
transfer price, all pro…ts would be shifted towards to a lower tax country with FDI being
used as a vehicle. However, this type of behavior can be a violation of tax laws. We thus
explore implications of institutional constraints on the internal transfer price. To this
end, we assume that a deviation of an MNE’s internal price from its true marginal cost
entails costs of ( c; q). This could be interpreted as concealment costs or can re‡ect
expected punishment for the deviation as in Kant (1988). For analytical tractability, we
assume the concealment costs are separable in the deviation of the internal price from its
true MC and the amount of inputs transferred, that is, ( c; q) = ( c) (q): We
make the following assumption about the concealment costs.
0 00 0 000
Assumption 1. > 0, > 0 for c > 0 and > 0; 0;for q > 0 with

0 0
(0) = (0) = (0) = (0) = 0

Assumption 1 states that concealment costs increases with the transfer price’s devia-
tion from its true cost and the amount of inputs transferred. In addition, concealment
costs are convex in the degree of deviations with the usual Inada conditions.
With this concealment cost constraint being into account, the monopolist solves the
following problem
4
If we consider a dynamic model, the headquarter’s loss may be used as tax o¤set against future
pro…ts. However, it cannot be used as a tax o¤set if the headquarters make losses all the time.

7
 M ax = (1 t) [P (q) ]q + (1 e
t) ( c)q ( c; q) (3)
q | {z } | {z } | {z }
Downstream Pro…ts Upstream Pro…ts Concealment Costs

= e ( c; q) (4)

The …rst order condition for this is given by

@ @ (q; ) e
= (1 t) + (1 t)( c) q( c; q) = 0 (5)
@q @q

Let qbm ( ) be the solution to (2), that is,

qbm ( ) = arg max

Note that qbm (c) = q m (c):

2.2.1 Optimal Transfer Price with Concealment Costs

If we further assume that the concealment costs are proportional to the MNE’s output,
that is, ( c; q) = ( c)q with 0 > 0; 00 > 0, and 0 (0) = 0, as in Egger and Seidel
(2013), we have a very clean characterization concerning the MNE’s optimal transfer price
and its sourcing decision. With linear concealment costs in the output, we can write down
the monopolistic MNE’s pro…t function as

= (1 t)[P (q) ]q + (1 e
t)( c)q ( c)q (6)
= (1 t)[P (q) ]q; (7)

where

e
(1
t)( c) ( c) (1 e
t)c (t e
t) + ( c)
= + =
1 t 1 t 1 t
( c)
= + ;
1 t
e e
where = (1 t)c1 (tt t) : Thus, the optimal choice of the transfer price is equivalent
to minimize the MNE’s "virtual marginal cost" of ;which is adjusted for transfer price

8
induced by di¤erential tax rates and concealment costs, and implicitly de…ned by

t e
t= 0
( c)

The optimal thus can be derived as

=c+ 0 1
(t e
t) > c

For instance, if we assume ( c) = k2 ( c)2 ; where a higher k represents better

institutional monitoring which make it more costly for the MNEs to engage in pro…t
e
shifting. Then, we have b = c+ t k t : The optimal choice of the transfer price is consistent
with empirical …ndings. For instance, Clausing (2003) shows that as the counter-party
tax rates are lower, US intra…rm import prices are higher (note that we have di¤erent
predictions without concealment costs).

2.2.2 FDI vs. Outsourcing

Let be the minimized virtual MC with the choice of optimal transfer price : Then,
the MNE’s pro…t from FDI can be written as

F DI
= (1 t)[P (q) ]q;

whereas the monopolist’s pro…t from simply outsourcing can be written as

OS
= (1 t)[P (q) $]q

Thus, the monopolist’s sourcing decision boils down a simple comparison of and $;
FDI takes place if and only if < $:

Lemma 1. <c

Proof. Note that can be written as

(t e
t)( c) ( c)
=c
1 t

9
 Figure 2: Globally E¢ cient Sourcing vs. MNE’s Sourcing Decision

Using the …rst order condition that de…nes ; we …nd that

2 3
6 ( c) 7
e 6 7
(t t)( c) + ( c) = ( c) 6 0 ( c) 7
4 ( c) 5
| {z }
>0 by the convexity of

Therefore, < c:
Lemma 1 implies that the MNE’s pro…t is global pro…t is higher due to tax manip-
ulation compared to the case where the …rm transfers its input at its marginal cost c:
The MNE’s pro…t is as if its cost were the virtual cost of which is lower than its true
marginal cost of c. This, in turn, implies that the MNE’s sourcing decision can ine¢ cient
from the global production e¢ ciency viewpoint (see Figure 2). Pro…t shifting motives
e
due to tax di¤erences across countries create a wedge of (= (t t)( 1c)t ( c) > 0), which
distorts the MNE’s sourcing decision.

Proposition 1. (Ine¢ ciency of Internal Sourcing) With tax di¤erentials across countries,
there can be excessive FDI. The global e¢ ciency requires that FDI takes place i¤ c < $
whereas FDI takes place in equilibrium i¤ < $: Thus, if c 2 ($; $ + );where =
(t e
t)( c) ( c)
1 t
> 0, there is an ine¢ cient FDI.

10
2.2.3 Parametric Example

As shown above, with ( c) = k2 ( c)2 ;we have

t e
t
=c+
k

By plugging back this into , we can easily verify that

(t e
t)2
=c
2k(1 t)

This implies that FDI takes place if and only if

(t e
t)2
c<$+
2k(1 t)

That is, unless the MNE’s internal production cost does not exceed the open market price
(t e
t)2 (t e
t)2
by 2k(1 t)
, FDI takes place. In particular, if c 2 ($; $ + 2k(1 t)
), FDI is ine¢ cient, but
still optimal from the perspective of the MNE due to tax manipulation via transfer price.

2.3 Non-Linear Concealment Costs and Dual Sourcing

With the concealment costs linear in the amount internally transferred (with an in‡ated
price) q, the MNE will procure its input only from a single source (i.e., either all from
the internal source or all from the open market). However, if the concealment costs are
convex in q;the MNE may source its inputs from the internal and external sources. To
see this, let us assume that ( c; q) = ( c) (q) with 0 and 00 > 0.

= (1 t)[P (q) ]q + (1 e t)( c)q ( c) (q)

!
(1 e t)c (t e
t) ( c)
= (1 t) [P (q) ]q (q)
1 t 1 t

Thus, given , the virtual marginal cost from internal sourcing via FDI is not constant
can be expressed as

[(1 e
t)c (t e
t) ] + ( c) 0 (q)
(q; ) =
1 t

11
This also implies that depending on the production quantity, the optimal transfer price
changes. For a given quantity q, the transfer price that minimizes the total production
cost [(1 e
t)c (t e t) ]q + ( c) (q) is given by the following …rst order condition

(t e
t)q = 0
( c) (q)

By totally di¤erentiating the condition above, we can easily verify that the optimal
internal price (q) is decreasing in q:

(t e
t)dq = 00
( c) (q)d + 0
( c) 0 (q)dq

Thus, we have
d [(t e
t) 0
( c) 0 (q)]
= 00 <0
dq ( c) (q)
because 0 ( c) 0 (q) > 0 ( c) (q)
q
= (t et) by the convexity of and the …rst order
condition for
Let qI and qO denote the amount of inputs from internal (i.e., FDI) and outside sources,
respectively. Then, the fully optimal sourcing decision can be derived from the following
optimization program

[(1 e
t)c (t e
t) ]qI + ( c) (qI )
M in + $qO
qI ;qO ; 1 t
subject to

qI + qO = q
qI ; qO 0

The Lagrangian for this problem can be written as

[(1 e
t)c (t e
t) ]qI + ( c) (qI )
$= + $qO + [q (qI + qO )];
1 t

where is the Lagrangian multiplier associated with the constraint qI + qO = q:

The …rst order conditions can be written as

12
 @$ @$
= (qI ; ) 0; qI = 0
@qI @qI
@$ @$
= $ 0; qO = 0
@qO @qO
@$ (t et)q + 0 ( c) (q)
= =0
@ 1 t

There are two types of solutions to this problem.

(i) Dual Sourcing with qI > 0; qO > 0:
In this solution, (qI ; ) = $. Let qb be the unique output level such that ... This
would be the case when q > qb. Then, the amount of internal sourcing is given by qI = qb,
and the rest is outsourced, that is, qO = (q qb).
(ii) Internal Sourcing Only with qO = 0:
In this case, we have (qI ; ) = < $: This would be the case when q < qb.

(t e
t)q = 0
( c) (q)
[(1 e
t)c (t e
t) ] + ( c) 0 (q)
(q; ) = =$
1 t

Let solution to this be qb: Then, up to qb internal production and beyond which out-
sourcing. So if M R(b q ) > $, the dual sourcing. If not, then only single sourcing. That
is the MNE solves
[(1 e
t)c (t e t) ]qI + ( c) (qI )
M in
qI ; 1 t
which de…nes (q): Thus, the cost function up to qb is given by

[(1 e
t)c (t e
t) (q)]q + ( (q) c) (q)
C(q) =
1 t

@C @C @ [(1 e
t)c (t e
t) (q)] + ( (q) c) 0 (q)
C 0 (q) = + = = (q; ) < $;
@q @ @q 1 t

which is described by curve in Figure 3.

13
 Figure 3: Internal vs. Dual Sourcing

As Figure 3 illustrates, with convex concealment costs, there will be internal sourcing
only with small market demand, but as market size grows, the MNE relies on dual sourc-
ing. Note that in our model, we abstract away from …xed costs of setting up a subsidiary
by FDI. If there are any …xed costs associated with FDI, then our model would predict
that for a very small market size, the sourcing will be done by pure outsourcing, but once
the market size grows enough to justify …xed set-up costs, then the monopolist will switch
to internal sourcing, and the market size becomes su¢ ciently large, it will also use outside
sourcing. That is, the use of outsourcing is not monotonic with the market size if there
are …xed costs of FDI.

Let us work with a parametric example of ( c; q) = k ( c) (q), where (

14
c) = ( c) with > 1, and (q) = q , that is, ( c; q) = k( c) q . Thus,
0 1
( c) = k( c) : As a result, the optimal and q satis…es

(t e
t)q = k ( c) 1
q
[(1 e
t)c (t e
t) ] + k ( c) q 1
= (1 t)$
1 1
t e
t 1 1
From the …rst equation, we have (q) = c + k
q : By substituting this for in
the second equation,
we have
! 1
2 ! 1
3
t e
t
1 1
t e
t
1 1
e +k 4 5 q
1 1 1
(1 t)c (t t) q q = (1 t)$
k k

=) 1
1
(1 t)(c $) = ( )( k)
1
t e
t 1
q
1

Thus, q is given by
1
1
q = ;
(1 t)(c $)
1
where = ( )( k)
1
t e
t 1

=)

1 1
1 1 1

qb = = ( k)
1
t e
t 1

(1 t)(c $) (1 t)(c $)

To illustrate the idea, let us assume that P (q) = A q, where A represents the market
size. Then, M R = A 2q and the M R curve intersects with ! (the outsourcing M C) at
q = A 2 ! : Thus, dual sourcing takes place if and only if A 2 ! > qb, i.e.,
1
1 1

A > 2b
q+$ =2 ( k)
1
t e
t 1
+$
(1 t)(c $)

2.4 Dual Sourcing and Invocation of the CUP Method

15
In the previous section, we analyzed the MNE’s sourcing behavior in the presence of
concealment costs. The basic premise of the analysis was that for the monopoly case
we have considered the applicability of the ALP with the CUP method can be limited
if there is only one …rm that produces the product and there are no similar transactions
that can be observed and used as a benchmark. This is especially so when all input
acquisitions are done internally via FDI. Even if an alternative input is available at the
price of $, the MNE may argue that the input available in the open market is not suitable
for speci…c purposes of the MNE and the unavailability of suitable input is the reason for
FDI and internal sourcing to begin with. However, such an argument loses appeal once
the MNE engages in dual sourcing and acquires some of their input requirements from
outsourcing because it is an implicit admission that the open market input is suitable for
its …nal product. This implies that dual sourcing may entail a risk that it may induce the
government to adopt the CUP method instead of the cost plus method.
If dual sourcing invokes the use of the CUP method as an application of the ALP, the
MNE thus has two choices. One is to engage in internal sourcing only to avoid the CUP
method. The other is to do outsourcing in which case the CUP method will be imposed
and the internal price should be set at $ if internal sourcing is also used.

Proposition 2. If dual sourcing triggers the CUP method, the monopolistic …rm never
engages in (meaningful) dual sourcing.

Proof. Suppose that the monopolistic …rm engages in dual sourcing with qI > 0 and
qO > 0, where q = qI + qO : Then, its internal price should be = $. Thus, the
monopolist’s pro…t is with dual sourcing under ALP is given by

D
= (1 t)[P (qI + qO ) $](qI + qO ) + (1 e
t)($ c)qI
= (1 t)[P (q) $]q + (1 e
t)($ c)qI

We consider two cases depending on the MNE’s true marginal cost of production via FDI
exceeds the market price or not.
(i) c > $
In this case, the foreign subsidiary that produces internally makes loss due to ALP.
Thus, the pro…t from dual sourcing is less than the pro…t under outsourcing only which

16
 Figure 4: Equilibrium Change with ALP

is given by

OS
= (1 t)[P (q) $]q
> (1 t)[P (q) $]q + (1 e
t)($ c)qI = D
| {z }
( )

(ii) c < $
In this case, essentially internal sourcing only will dominate dual sourcing because D
is maximized when qI is as close as q with a minimal qO :
The imposition of ALP thus may have di¤erent e¤ects depending on the situations.
If c < $, it will have desirable e¤ects [explain]. However, if c > $; the monopolistic …rm
may engage in internal sourcing only with the imposition of ALP (see Figure 4)

3 Import Tari¤ as a Countermeasure against Pro…t-Shifting

We consider a speci…c industry in which the MNE is operating. Implicitly we assume that
the overall corporate tax rate is determined by factors beyond the speci…c industry we
consider. The overall corporate tax rate thus cannot be tailored for this particular industry

17
and is considered exogenously given. However, in face of pro…t-shifting incentives of the
MNE, the government can impose industry-speci…c ad-valorem import tari¤s to eliminate
such incentives. We analyze how import tari¤s can be adopted as a countermeasure
against pro…t-shifting.
Let m denote ad-valorem import tari¤ imposed by country H where the MNE is
located. Now the MNE’s problem with FDI can be written as

b = (1 t)[P (q) (1 + m) ]q + (1 e
t)( c)q ( c)q (8)
= (1 t)[P (q) b]q; (9)

where

(1 e
t)c (t e
t) + ( c)
b = m +
1 t
= m +

e e e
with = (1 t)c (t1 t)t + ( c) = c (t t)( 1c)t ( c) :5
In the presence of import tari¤s, the optimal choice of the transfer price b is equivalent
to minimize the MNE’s "virtual marginal cost cum tari¤s" of b, and implicitly de…ned by

t e
t m (1 t) = 0
(b c)

By totally di¤erentiating the …rst order condition above, we have

00
(1 t)d m = (b c)db

Thus, we have
@b (1 t)
= 00 <0
@ m (b c)
indicating that the incentives to in‡ate the internal price by the MNE can be mitigated
by the imposition of import tari¤s. Note that the optimal b chosen by the MNE can be
written as
b = c + 0 1( t e t m (1 t))
5
We denote all variables in the presence of import tari¤s with a hat (^)

18
 e
This implies that import tari¤s of m = m (= 1t tt ) can be used to completely o¤set
any incentives for pro…t-shifting. In addition, with m = m , b = c and the MNE will be
engaged in FDI only when its internal production is more e¢ cient than the open market.
However, consumer welfare goes down compared to the case of no import tari¤s. Thus,
the optimal import tari¤s can be lower than m , the import tari¤ that eliminates any
incentives for pro…t-shifting as shown below.

3.1 Optimal Import Tari¤s

Let us analyze the government’s optimal choice of import tari¤s given (t; e
t) when it max-
imizes domestic social welfare, which is de…ned as

W = Producer Surplus + Consumer Surplus + Tax Revenue.

We consider import tari¤s as a second-best policy when the transfer price and output
choices are left to the …rm. Let b be the minimized virtual MC with the choice of
optimal transfer price b , that is,

(1 e
t)c (t e
t)b + (b c)
b = mb +
1 t

Let the corresponding output level be q(b ): Then, social welfare with FDI can be written
as
Z q
W = (1 t)[P (q) b ]q + P (x)dx P (q)q + [t [P (q) (1 + m ) ]q + m q]
0
= (1 t)[P (q) (1 + m) ]q + (1 e
t)( c)q ( c)q
Z q
+ P (x)dx P (q)q + [t [P (q) (1 + m) ]q + m q]
0
Z q(b ) h i
= P (x) bSP dx
0

SP
where b = c + e t(b c) + (b c) and represents the marginal cost of FDI production
from the perspective of the social planner of H country. It consists of the physical
production cost of c, tax transfer to the host country, and any concealment costs incurred
by the MNE. Note that the production level by the MNE is determined by its perceived
SP
MC of b , not the social planner’s b . This implies that the choice of m that minimizes

19
bSP is not necessarily the optimal import tari¤.

To be more precise, the marginal e¤ect of import tari¤ on social welfare can be written
as follows:
"Z b #
q( (b ; m ))
dW d SP
= P (x) b (b )dx
d m d m 0
2 3
h SP
i dq 6 @b @ b @b 7 Z q(b (b )) dbSP (b) @b
= P (q(b (b ; m ))) b (b )) 6 + 7 dx
db 4 @ m |@b{z @ m5
} 0 db @ m
=0
h i b @b
SP

= P (q(b (b ; m )))
bSP (b )) dq @ q(b (b ))
( ) @b
= 0;
db @ m @b @ m

where

@b (t e
t) 0
(b c)
= m + = 0 by the envelope theorem
@b 1 t 1 t
SP
@b (b) dq @b @b (1 t)
and = e
t+ 0
(b c) > 0; < 0; =b ; = 00 <0
@b db @ m @ m (b c)

In a more compact form, we have

dW h i dq( ) db
SP

= P (q( m ))
bSP ( m )
m
q( m )
d m d m d
| {z } | {zm}
( ) (+)

The …rst term on the RHS represents the negative e¤ect on consumer welfare as the
imposition of import tari¤s increases the MNE’s virtual MC which induces the …rm to
reduce outputs in the domestic market. The second term on the RHS is the positive
e¤ect of reducing tax shifting to country F and concealment costs.
t e
t
Lemma 2. The optimal import tari¤ is never higher than m = 1 t
:

Proof. See the Appendix.

We thus consider only import tari¤ m m : Let us de…ne the wedge between the
SP
social planner’s perceived MC and the MNE’s perceived MC as = b( ( m )) b ( ( m )):

20
Then, we have

b( ( bSP ( ( t e
m )) m )) = m + ( c) (1 t)( c) ;
1 t

where b = c + 0 1 ( t e
t m (1 t)):
t e
t
Since = c when m = 1 t
; we have

t e
t
j t te = b( ) bSP ( )j t te = c>0
m= 1 t m= 1 t 1 t

We can also easily verify that

bSP ( )j t
j m =0 = b( ) m =0 = ( c) (1 e
t)( c) j m =0 <0
1 t

To see this, note that at satis…es the …rst order condition t e

t = 0
( c): Since
is convex, we have
( c) 0 e
< ( c) = t t <1 t
c
Thus, (1 e
t)( c) > ( c): We also know that
h i
d b( ( m ))
bSP ( ( ))
m
e 0 @b
= t+ ( c) >0
d m @ m
0 @b
because ( c) > 0 and @ m
< 0: Thus, we have the following lemma.
e
Lemma 3. There is a unique o
m 2 (0; 1t tt ) such that
8
>
> b bSP ( ) if o
< ( )> m < m
b( ) = bSP ( ) if m = o
>
>
m
: b( ) < bSP ( ) if > o
m m

A su¢ cient condition for the optimal import tari¤ to be less than m is
"Z #
q(b( ))
dW d b SP
j m= m = P (x) ( )dx j t te
m= 1 t
<0
d m d m 0

The following proposition provides a su¢ cient condition for this.

21
 0 dq
Proposition 3. Let =P d
denote the cost-price pass-through rate for the monopolist.
SP
db
d m
dW
d m
j m= m < 0 if > d j t te
m= 1 t
:
d m

Proof. We have

dW h SP
i dq( ) db
SP

= P (q( m ))
b ( m)
m
q( m )
d m d m d
| {z } | {zm}
( ) (+)

h i dq( db
SP
b + (b bSP ( m)
= P (q( m )) m )) q( m )
d m d m

By Lemma 3, we know b bSP > 0 at m = m: In addition, we know that

P (q( m ))
b = 0
P q

dq( m)
by the …rst order condition for the MNE’s pro…t maximization and d m
< 0: As a
result, we have
SP
dW dq( m )
0 db
j m= m < P q q( m ) j t te
d m d m d m m= 1 t
" SP #
0 dq d db
= q P + j m = t te
d d m d m 1 t

Therefore,

" SP # db
SP

dW 0 dq d db d m
j m= m < q P + j t te
m= 1 t
< 0 if > d
j t te
m= 1 t
d m d d m d m d m

e
Thus, we can conclude that the optimal import tari¤ m < m = 1t tt , that is, the
optimal import tari¤s mitigates incentives to engage in tax manipulation via transfer price,
SP
db
d m
but does not completely eliminate it, if > d j m= m . This is because the transfer
d m
price induces the MNE produces more, which can enhance consumer welfare. For instance,
bSP
this condition is satis…ed if e
t is su¢ ciently small because dd m =et 00(1(b t) c) ' 0 and
m= m

22
 d
d m
j m= m= c:
In particular, if we assume quadratic concealment costs and linear demand of P =
A q;the condition can be written as
So far, our discussion of the optimal ad valorem import tari¤ was on the premise that
the MNE is engaged in FDI. However, if the optimal m derived above is su¢ ciently
large, we may have a situation in which b( m = 0) < $, but b( m = m ) > $(1 + m ): In
this case, there is a unique import tari¤ level m 2 (0; m ) such that b( m ) = $(1 + m ):
In this case, the optimal tari¤ is m : The government in country H then sets an import
tari¤ at the rate of min[ m ; m ] because H country prefers outsourcing at m = m , that
is, b( m = m ) = c(1 + m ) > $(1 + m ):
To see this, note that at m = m ; the MNE produces the same quantity and thus CS
is the same. With outsourcing, welfare can be written as
Z q
c OS =
W [P (x) $] dx
0

23
What we need to show is b( m) > $:
h SP i
(1 + m ) b ( m) $ = (1 + m) c+e
t(b c) + (b c)
" #
(t e
t) (b c) + (b c)
mb +c+
1 t
> (1 + m) c+e
t(b c) + (b c)
" #
(t e
t) (b c) + (t e
t) (b c)
mb +c+
1 t
1 e
= m (b c) + (1 + m) t(b c) + (b c)
1 t
t
+ (b c)
"1 t #
(t e
t)
= m (b c) + (1 + e
m )t(b c)
1 t
1
+ (1 + m) (b c)
1 t
" #
(t e
t) e
> m (b c) + (1 + m )t(b c)
1 t
1 e
+ (1 + m) (t t) (b c)
1 t
t e
t t
= (1 + m )t m > 0 because m < <
1 t 1 t

3.1.1 Parametric Example

As shown above, with ( c) = k2 ( c)2 ;we have

t e
t m (1 t)
b =c+
k

By plugging back this into b, we can easily verify that

24
 " # e e (t et) m (1 t) (t et) m (1 t) 2
e (1 t)c (t t) c + k
+ k2 ( k
)
b t t m (1 t)
= m c+ +
k 1 t
2
t e
t m (1 t)
= c(1 + m)
2k(1 t)

From outsourcing, the MNE’s cost becomes $(1+ m ): So given m; the MNE engages
in FDI if and only if
2
t e
t m (1 t)
c(1 + m) < $(1 + m)
2k(1 t)

or
c < $ + b;
2 2
[(t et) t)]
m (1 (t et)
where b = 2k(1 t)(1+ m )
: Note that b is decreasing in m and bj m =0 = = 2k(1 t)
and
bj m = m = 0:

4 The Duopoly Model with Strategic Interactions

In this section we consider a duopoly model in which an MNE competes with another
…rm in the domestic market in order to explore implications of strategic e¤ects. The
set-up is otherwise the same as in the monopoly model. More speci…cally, two …nal good
producers, …rm 1 and …rm 2, compete in H. Firm 2 is a domestic …rm and simply
procures its input from F with an exogenously given market price e (later we extend the
model to endogenize e). Firm 1 has two choices as before. It can procure its input from
F like …rm 2. Or it can be an MNE by setting up its own input production plant in F
with FDI. In such a case, its input production cost is given by c. In this case, the MNE
can choose an internal transfer price ( ) when its foreign subsidiary supply its input to the
headquarter …rm that produces the …nal good. Figure 4 describes the duopoly model.
In the monopoly case, we assumed that the CUP method is not applicable because
there is no comparable downstream …rm and the input market simply does not exist in
the case of FDI (unless the MNE is engaged in dual sourcing that also relies on outside
suppliers). As a result, the ALP was based on the cost plus method and the MNE

25
 Figure 5: Duopoly Model with Strategic Interactions

was assumed to operate with concealment costs when its transfer price deviates from its
marginal cost. In the case of duopoly, the applicability of the CUP method depends
on whether the transactions between the rival downstream …rm and its input suppliers
can be regarded as "externally comparable" to internal transactions of the MNE (OECD
2010, p. 71). We present two sets of results depending on the comparability of the
external transactions. First, we consider a scenario in which the external transactions
are not considered as comparable. This would be the case if the two downstream …rms
produce di¤erentiated products and use very di¤erent types of inputs. Then, the ALP
should be based on the cost-plus method and the MNE operates with concealment costs.
In contrast, if the external transactions are considered as comparable, then the MNE is
constrained to use the comparable market price as the internal transfer price.

4.1 Pro…t Shifting in Duopoly Model with Concealment Costs

We …rst analyze the case in which the transactions between the rival downstream …rm
and its input suppliers are not comparable to the internal transactions of the MNE. In
this case, the MNE’s behavior can be described with the presence of concealment costs for
transfer price that deviates from its true marginal cost, as in the previous section. The
case of comparable external transactions is analyzed in section 4.2.
To focus on implications of strategic interactions for the MNE’s behavior, we assume
concealment costs that is linear in output, that is, ( c; q) = ( c)q with 0 >

26
 00
0; > 0 with 0 (0) = 0:
Firm 1 solves the following problem

M ax 1 = (1 t) (q1 ; q2 ; ) + (1 e
t) ( c)q ( c)q1 (10)
q1 |1 {z } | {z }1 | {z }
Downstream Pro…ts Upstream Pro…ts Concealment Costs

where 1 (q1 ; q2 ; ) = [P (q1 ; q2 ) ]q1 : Once again, by collecting terms with q1 , we can
rewrite it as
1 = (1 t)[P (q1 ; q2 ) ]q1 ; (11)

where
(1 e e
t)c
t) + ( (t c)
=
1 t
The …rst order condition for …rm 1 is given by

1 @ 1 @ 1 (q1 ; q2 ; )
= =0 (12)
1 t @q1 @q1

Firm 2 similarly makes its decision on q2 to solve the following problem:

M ax 2 (q1 ; q2 ; $) = [P (q1 ; q2 ) $]q2 (13)

q2

)
@ 2 (q1 ; q2 ; $)
=0 (14)
@q2
Eq (12) and (14) implicitly de…ne reaction functions for …rm 1 and …rm 2 respectively.
The equilibrium quantities for each …rm, q1 ( ) and q2 ( ), given the transfer price are
at the intersection of these two reaction functions.
2 2
Assume @@q2i > @q@ i @qij ; where i = 1; 2 and j 6= i:
i

Lemma 4. Let be the unique that minimizes the virtual cost of such that (t e t) =
0 dq1
( c): Then we have the following comparative statics results: (i) if < ; d > 0
dq2 dq dq dq dq
and d < 0, (ii) if > ; d 1 < 0 and d 2 > 0, and (iii) if = b; d 1 = d 2 = 0.

Proof. See the Appendix.

Lemma 2 implies that with concealments costs, the strategic e¤ects of an increase in the
transfer price depends on the relative magnitudes of the tax di¤erential and the deviation
of the transfer price from the MNE’s true marginal costs. Note that dq d
1
= dq
d
2
= 0 if

27
t=e t: That is, the transfer price is irrelevant for downstream competition if there is no
tax di¤erential between H and F. As the MNE behaves as if its MC were which is
the weighted average of its true marginal cost c and the internal transfer price, with the
weights determined by the tax rates. The higher the transfer price, the MNE behaves
as if its MC were lower and compete more aggressively in the downstream market and
this e¤ect is larger with the tax rate di¤erential (t e t): Note that = c if t = e t or
= c. This implies that both di¤erential tax rates and in‡ated internal transfer price
are required for strategic e¤ects on the downstream market.
Now let us analyze the optimal choice of the MNE’s transfer price : To this end, we
analyze the e¤ects of on the overall net pro…ts of the MNE.

d 1 d 1
e @ 1 dq2 0
= [(1 t) + (1 t)q1 ] + (1 t) ( c)q1
d d @q d | {z }
| {z } | {z 2 } Concealment Cost E¤ect
Pro…t Shifting E¤ect Strategic E¤ect

e 0 @ 1 dq2
= [(t t) (
c)]q1 + (1 t) =0
| {z } @q2 d
Direct Tax E¤ect | {z }
Strategic E¤ect

By lemma 2, we know that the direct tax e¤ect and the strategic e¤ect move in the
same direction. That is, 8
> > 0 if <
d 1 <
= = 0 if =
d >
:
< 0 if >
Thus, the optimal transfer price is given by as in the monopoly case.
As expected, the MNE’s tax manipulation strategy reduces its tax obligation in H
(high tax rate country). In fact, in our model, its internal transfer price is arti…cially
jacked up to the extent to shift all its pro…t at the downstream stage to the upstream
subsidiary which resides in F (low tax rate country). As a result, the tax revenue from
the MNE is zero because all pro…ts are shifted to a subsidiary that is located in a lower
tax country. In addition, we uncover additional tax revenue loss from other …rms in the
presence of imperfect competition due to strategic e¤ects. However, it is not the end of
the story; there is a collateral damage due to spillover e¤ects. The aggressive behavior of
the MNE with the tax-induced transfer price also reduces the rival …rm’s pro…ts. Thus,
the tax revenue from the other …rm that is not engaged in tax manipulation also decreases

28
(even though consumer surplus increases).
It is also worthwhile to point out the crucial di¤erence between the strategic e¤ects
driven by tax di¤erences in our model and strategic transfer pricing in the IO and man-
agement literature (see Alles and Datar (1998)). The basic premise of strategic transfer
pricing in oligopoly models is to assume decentralized decision making and each division
maximizes its own pro…ts, rather than the overall pro…ts of the …rm. Otherwise, the
optimal decisions will be based on true marginal costs and the transfer prices would not
matter and does not generate any strategic e¤ects because internal transfer prices are just
transfers among divisions within the …rm and cancel out each other from the perspective
of …rm’s overall pro…ts. Only when the decision of each division is driven by its own
pro…ts, transfer price can have any meaningful e¤ects. In contrast, our model assumes
centralized decision making. If the decision is not centralized, when the transfer price is
in‡ated to reduce the tax burden, the strategic e¤ects will go the other way around.

4.2 Arm’s Length Principle with the CUP Method

We now consider a scenario in which the transactions between the rival downstream …rm
and its input suppliers can be considered comparable. In this case, the ALP can be
applied as a requirement that the transfer price be equal to similar input price in the
market, which is the input price paid by …rm 2.6 If the input market for …rm 2 is
perfectly competitive with the price of ! and the same input can be used for …rm 1, the
analysis is trivial. As its transfer price is constrained to be at $ with the CUP method,
it will engage in FDI if and only if FDI is e¢ cient from the global e¢ ciency point, that
is, c < $. In this case, pro…t-shifting will take place to some extent, but it is limited by
the competitive market price $: If c > $; there is no ine¢ cient FDI for pro…t shifting
purpose. However, if we assume that the input available in the open market is supplied
by a …rm with market power, we can restore ine¢ cient FDI for tax manipulation.
To see this, we now endogenize e by assuming that F input supplier has market power.
More speci…cally, let us assume that F input supplier is a monopolist and sets the input
6
Gresik and Osmundsen (2008) analyzes the arm’s length principle when all …rms are vertically in-
tegrated and comparable but independent transactions on which the application of the arm’s length
principle can be based is not available. Such issue does not arise in our model because the rival down-
stream …rm acquires its input from an independent source.

29
price e given its marginal cost of e
c:

M ax(e c)q2CU P (e)

e
e

Let q(x; y) denote the equilibrium output level for a downstream …rm when its input
cost is x while the rival …rm’s cost is given by y: Let us assume that …rm 1 engages in
FDI. Then,
q2AL (e) = q(e; CU P );

where
CU P (1 e
t)c (t e
t)e
=
1 t
The …rst order condition on e is given by

CU P @q @q @ CU P
q(e; ) + (e e
c) + =0
@x @y @e

Thus, the optimal e is implicitly de…ned by

CU P
q(e ; )
e =e
c h AL
i >e
c
@q @q @
@x
+ @y @e

With the CUP applied as the ALP, upstream …rm F’s input price e in‡uences the
MNE’s transfer price and indirectly a¤ects MNE’s competitive behavior in the down-
AL e
stream market via its e¤ect on CU P . Since @ @e = (t1 tt)e < 0; a higher input price to
…rm 2 reduces …rm 1’s virtual cost AL and indirectly reduces …rm 2’s output via strategic
e¤ects. Thus, under the CUP, upstream …rm F charges a lower input price compared to
the case of its absence if …rm 1 produces internally with FDI.
We now consider …rm 1’s FDI decision. Let us assume that the monopolistic input
supplier cannot price discriminate between two downstream …rms and charge an input
price of e: Then, …rm 1 engages in FDI if and only if

CU P (1 e
t)c (t e
t)e
= e;
1 t

which is equivalent to c e: This implies that if the monopolist charges an input price
higher than c, its demand is q(e; AL ) whereas if it charges an input price less than c, its
demand is 2q(e; e): It is immediate that …rm 1 always engages in FDI if its production

30
cost is lower than the input supplier, i.e., c ec . We thus forcus on the case where c > ec;
that is, …rm 1’s internal production is less e¢ cient than the input supplier. In this case,
the input monopolist’s pro…t can be written as

m
m
2 = (e c) q(e; CU P ) if e > c
e
= m
b = (e ec) 2q(e; e) if e c

Note that CU P = c when e = c: This implies that the monopolist’s pro…t jumps down
discretely (more precisely, halves down) as its price is increased from c by when e c < c:
AL
For analytical simplicity, let us assume that both m 2 = (e e
c) q(e; ) and mb =
(e e c) 2q(e; e) are concave in e: To reduce the number of cases to consider, we assume
that the gap between e c and c is not too large. More speci…cally, we assume

@ m
b @q(c; c) @q(c; c)
= 2 q(c; c) + (c e
c) + >0
@e e=c @x @y

that is,

q(c; c)
(c e
c) <
@q(c;c) @q(c;c)
@x
+ @y

It can also be easily veri…ed that this condition also implies that e (e
c) > c. Otherwise,
the outside monopolistic input supplier’s cost is drastically lower than that of the MNE
and FDI would never be a viable strategy. In such a case, the monopolistic input supplier
can either to set the priced at em = e (ec) and sell only to …rm 2 or set the price at em = c
and sell to both …rms. That is,

m CU P
= max[(c e
c) 2q(c; c); (e e
c) q(e ; )]

It is clear that selling only to …rm 2 at the price of e (e c) is a better option than selling
at the price of c if e
c is very close to c: However, if e
c is su¢ cently lower than c, it may be
optimal to set the input price at c; as the next lemma shows.

Lemma 5. Suppose that e c < c: Then, there is a critical level of e

c (which is strictly lower
than c) such that the optimal price for the monopolistic input supplier em is given by

c if e
c<e c
em =
e (e
c) if e
c e c

31
There is ine¢ cient FDI if e
c 2 (e
c ; c):
CU P CU P
Proof. Let (e c) = (c e
c) 2q(c; c) (e e
c) q(e ; ): We know (e
c = c) = (e e
c) q(e ; )<
0: In addition,
@ (e
c) CU P AL
= 2q(c; c) + (e ; )
@e
c
Since e > c and AL < c, we have q(e ; CU P ) < q(c; c): Thus, @ @ec(ec) < 2q(c; c) +
q(c; c) < 0: Therefore, there can be a critical value of e
c such that the statement in the
Lemma is true. If (e c = 0) < 0, the input monopolist always sell only to downstream
…rm 2 and we can take e c = 0:

Lemma 6. Let be the optimal price for the input monopolist if the MNE supplies its
input at its marginal cost c without in‡ated transfer price. Under ALP with the CUP
2 q(x;y)
method, the monopolistic input supplier charges a price e < if @ @x@y 0

Proof. By de…nition, satis…es the following …rst order condition

@q( ; c)
q( ; c) + ( e
c) =0
@x

If we evaluate the …rst order conditon under the ALP at e = ; we have

2 3
6 @q( ; AL ) @q @ CU P 7 @q( ; c)
CU P 6 7
q( ; )+( e
c) 6 + 7 < q( ; c) + ( e
c) =0
4 @x @y @e 5 @x
| {z }
( )

@q( ; AL ) @q( ;c)

because CU P < c, which implies that q( ; CU P
) < q( ; c) and @x
< @x
< 0: As
a result, we have e < :
Linear Demand Example:
We can easily verify that

AL (1 e
t)c (t e
t)e
a 2e + a 2e + 1 t
q2CU P (e) = =
3 3

The optimal choice of F input supplier and the subsequent internal transfer price due
to the arm’s length principle is given by

a + (1 + )c + (2 + )e
c
eCU P =
4+

32
 e
where = (t1 tt) :
In contrast, if the MNE transfers its input at its marginal cost of c, we have

a + c + 2e
c
= > eAL
4

It can also be easily veri…ed that this gap between and eAL increases in .

5 Concluding Remarks

We have analyzed an MNE’s incentives to manipulate an internal transfer price to take

advantage of tax di¤erences across countries. Our analysis of the monopoly case de-
rives conditions under which FDI takes place and shows that tax-induced FDI can entail
ine¢ cient internal production. We also analyzed implications of ALP and it can have
the opposite e¤ect to the one intended if it induces the MNE’s sourcing decisions from
dual sourcing to internal sourcing only to avoid the application of ALP. With imperfect
competition we show that the internal transfer price has additional strategic e¤ects that
further strengthen incentives to in‡ate the transfer price at the expense of the rival …rm’s
pro…ts. The tax-induced FDI by the MNE has spillover e¤ects that reduce tax revenues
from other domestic …rms as well as the MNE.
We have analyzed FDI decision of only one …rm in isolation in the oligopoly case
assuming that other …rms engage in outsourcing. However, each …rm’s FDI decision may
also depend on other …rm’s FDI decision. This is an area of future research.

33
References

[1] Alles, M. and Datar, S. (1998). Strategic Transfer Pricing. Management Science (44),
451-461

[2] Bauer, C. J. and Langenmayr, D. (2013). Sorting into outsourcing: Are pro…ts taxed
at a gorilla’s arm’s length? Journal of International Economics, 90(2):326-336.

[3] Bond, E. W. (1980). Optimal transfer pricing when tax rates di¤er. Southern Eco-
nomic Journal 47, 191-200.

[4] Clausing, Kimberly A. (2003). Tax-motivated transfer pricing and US intra…rm trade
prices. Journal of Public Economics, 87(9): 22-7-2223.

[5] Copithorne, L.W. (1971). Internal corporate transfer prices and government policy.
Canadian Journal of Economics, pp. 324-341.

[6] Davies, R., Martin, J., Parenti, M., Toubal, F. (2016). Knocking on tax haven’s door:
multinational …rms and transfer pricing, CEPR Discussion Paper No. 10844.

[7] Egger, P., Eggert, W., Winner, H., 2010b. Saving taxes through foreign plant own-
ership. Journal of International Economics 81, 99–108.

[8] Egger, P., Seidel, T., 2013. Corporate Taxes and Intra-Firm Trade, European Eco-
nomic Review 63, 225-242.

[9] Gresik, T., Osmundsen, P., 2008. Transfer pricing in vertically integrated industries.
International Tax and Public Finance 15, 231–255.

[10] Hau‡er, A., Schjelderup,G., 2000. Corporate tax systems and cross country pro…t
shifting. Oxford Economic Papers 52, 306–325.

[11] Hirshleifer, J., 1956. On the economics of transfer pricing. Journal of Business, 172-
184.

[12] Horst,. T. 1971. The theory of the multinational …rm: Optimal behavior under dif-
ferent tari¤ and tax rates. Journal of Political Economy 79, 1059-1072.

[13] Kant, C. (1988), ‘Endogenous transfer pricing and the e¤ects of uncertain regulation’,
Journal of International Economics 24(1-2), 147–157.

34
[14] Kato, H., Okoshi, H., 2017. Production location of multinational …rms under transfer
pricing: The impact of Arm’s Length Principle, unpublished manuscript.

[15] OECD (2010), OECD Transfer Pricing Guidelines for Multinational En-
terprises and Tax Administrations 2010, OECD Publishing, Paris. DOI:
http://dx.doi.org/10.1787/tpg-2010-en

[16] Samuelson, Larry (1982). The multinational …rm with arm’s length transfer price
limits. Journal of International Economics Volume 13, Issues 3–4, November 1982,
Pages 365-374.

35
 Appendix
Proof of Lemma 2:
By totally di¤erentiating these two equilibrium conditions, we have

@2 1 (q1 ; q2 ; ) @2 1 (q1 ; q2 ; ) @2 1 (q1 ; q2 ; )d

dq1 + dq2 + d = 0
@q12 @q1 @q2 @q1 @ d
2 2
@ 2 (q1 ; q2 ) @ 2 (q1 ; q)
dq1 + dq2 + 0d = 0
@q1 @q2 @q22

To conduct a comparative static analysis on , we can rewrite the equations above as

" @2 @2
# " #
1 (q1 ;q2 ; ) 1 (q1 ;q2 ; ) dq1 @2 1 (q1 ;q2 ; )d
@q12 @q1 @q2 d @q1 @ d
@ 2 2 (q1 ;q2 ) @ 2 2 (q1 ;q) dq2
=
@q1 @q2 @q22 d
0

@2
Note that @ 1 (q@1 ;q2 ; ) = q1 by the envelope theorem. This implies that 1 (q1 ;q2 ; )
@q1 @
= 1:
By applying the Cramer’s rule, we have

d @2 1 (q1 ;q2 ; )
d @q1 @q2
@ 2 2 (q1 ;q)
dq1 0 @q22
=
d D
d @ 2 2 (q1 ;q)
d @q22
=
D
@2 1 (q1 ;q2 ; ) d
@q12 d
@ 2 2 (q1 ;q2 )
dq2 @q1 @q2
0
=
d D
d @ 2 2 (q1 ;q2 )
d @q1 @q2
=
D
h i
@2 1 (q1 ;q2 ; ) @2 2 (q1 ;q) @2 1 (q1 ;q2 ; ) @2 2 (q1 ;q2 ) dq1
:where D = @q12 @q22 @q1 @q2 @q1 @q2
> 0: Thus, the sign of d
is the
d dq2 d
opposite of the sign of whereas the sign of
d d
is the same as the sign of d
: We have
the desired result because we have
8
>
< < 0 if <
d
= = 0 if =
d >
:
> 0 if >

36