Journal of International Economics 127 (2020) 103367

Contents lists available at ScienceDirect

Journal of International Economics

journal homepage: www.elsevier.com/locate/jie

Transfer pricing regulation and tax competition☆

Jay Pil Choi a, Taiji Furusawa b, Jota Ishikawa c,⁎
a
Department of Economics, Michigan State University, East Lansing, MI 48824, USA & School of Economics, Yonsei University, Seoul, Korea
b
Graduate School of Economics, University of Tokyo7-3-1 Hongo, Bunkyo-ku, 113-0033, Tokyo, Japan
c
Faculty of Economics, Hitotsubashi University, Kunitacchi, Tokyo 186-8601, Japan & Research Institute of Economy, Trade and Industry (RIETI), Tokyo, Japan

a r t i c l e i n f o a b s t r a c t

Article history: The paper analyzes multinational enterprises' incentives to manipulate internal transfer prices
Received 24 March 2020 to take advantage of tax differences across countries, and implications of transfer-pricing regu-
Received in revised form 5 July 2020 lations as a countermeasure against such profit shifting. We find that tax-motivated foreign di-
Accepted 14 July 2020
rect investment (FDI) may entail inefficient internal production but may benefit consumers.
Available online 24 July 2020
Thus, encouraging transfer-pricing behavior to some extent can enhance social welfare. Fur-
thermore, we consider tax competition between two countries to explore its interplay with
JEL classification: transfer-pricing regulations. We show that the FDI source country will be willing to set a
F12
higher tax rate and tolerate some profit shifting to a tax haven country if the regulation is
F23
tight enough. We also indicate a novel mechanism through which it is the larger country
H21
H26 that undertakes tax-motivated FDI, the pattern we often observe in reality.
L12 © 2020 The Author(s). Published by Elsevier B.V. This is an open access article under the CC BY
L51 license (http://creativecommons.org/licenses/by/4.0/).
Keywords:
Multinational enterprise
Corporate tax
Transfer pricing
Foreign direct investment
Arm's length principle
Tax competition

☆ This paper is an extensively revised version of a part of Choi et al. (2018). The authors wish to thank Arnaud Costinot (coeditor) and two anonymous referees for
helpful comments and suggestions. We also thank Hayato Kato, Yuka Ohno, Pascalis Raimondos, Martin Richardson, Jay Wilson and participants in various conferences
and seminars for valuable discussions and comments. This research was initiated during Choi's visit to Hitotsubashi Institute for Advanced Study whose hospitality is
greatly appreciated. Choi acknowledges financial support from the Ministry of Education of the Republic of Korea and the National Research Foundation of Korea: Grant
Number NRF-2020S1A5A2A01040865. Furusawa and Ishikawa acknowledge financial support from the Japan Society of the Promotion of Science through the Grant-in-
Aid for Scientific Research (A): Grand Numbers 19H00594 and 17H00986.
⁎ Corresponding author at: Faculty of Economics, Hitotsubashi University, Kunitacchi, Tokyo 186-8601, Japan & Research Institute of Economy, Trade and Industry
(RIETI), Tokyo, Japan.
E-mail addresses: choijay@msu.edu (J.P. Choi), furusawa@e.u-tokyo.ac.jp (T. Furusawa), jota@econ.hit-u.ac.jp (J. Ishikawa).
1
Preferential tax measures are observed in many countries. See, for example, https://www.dits.deloitte.com/#TaxGuides
2
Bernard et al. (2010) report that over 46% of U.S. imports comprised intrafirm transactions in 2000.
3
According to Tørsløv et al. (2020), more than $600 billion, which was close to 40% of multinational profits, was shifted to tax havens in 2015.

https://doi.org/10.1016/j.jinteco.2020.103367
0022-1996/© 2020 The Author(s). Published by Elsevier B.V. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
2 J.P. Choi et al. / Journal of International Economics 127 (2020) 103367

1. Introduction

Corporate tax rates substantially differ across countries, and some countries adopt preferential tax measures.1 Since multina-
tional enterprises (MNEs) actively engage in intrafirm transactions across borders,2 they have an incentive to manipulate internal
transfer prices to avoid tax payments, an activity often called “transfer pricing”. MNEs tend to shift their profits from high-tax coun-
tries to low-tax jurisdictions (see Hines and Rice, 1994; Huizinga and Laeven, 2008; Davies et al., 2018; Tørsløv et al., 2020).3 For
instance, inspections by the Vietnamese tax authorities have 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 profits in books”.4 In addition,
Egger et al. (2010) find that the average subsidiary of an MNE pays about 32% less tax than similar local firms in high-tax countries.
According to Goldman Sachs, the tax saving by U.S. MNEs amounts to $2 trillion, equivalent to four years' worth of U.S. corporate
tax revenues (Nikkei, August 31, 2016).
If governments do not regulate transfer pricing, MNEs may shift a large proportion of their profits away from their countries to
low-tax or no-tax jurisdictions, narrowing their tax bases significantly. Governments thus impose transfer-pricing rules to control
transfer-price manipulation. In particular, the Organisation for Economic Co-operation and Development (OECD) proposed that
internal transfer prices follow the so-called arm's length principle (ALP) in its guidelines for transfer pricing published in 1995
and revised in 2010.
The basic approach of the ALP is that the headquarters and affiliates of an MNE should be treated as “operating as separate
entities rather than as inseparable parts of a single unified 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 comparative anal-
ysis is at the heart of the application of the ALP. Currently, the ALP is the international transfer-pricing principle to which OECD
member countries have agreed.
Against this background, the purpose of this study is twofold. The first purpose is to closely examine MNEs' incentive to ma-
nipulate transfer prices to avoid tax payments, and governments' regulatory responses to such practices. We specifically investi-
gate the transfer pricing resulting from tax-motivated FDI and the implication of the ALP, when imperfect competition prevails in
the final-good market. To this end, we consider a stylized set-up of two countries with different corporate tax rates, where a mo-
nopolist that produces and sells its final product in the high-tax country can set up its subsidiary, which produces an intermediate
good, in the low-tax country to engage in tax-saving transfer pricing.
We show that this monopolistic MNE shifts all its profits to the low-tax country by choosing the standard monopoly price as
the transfer price, replicating the monopoly outcome in the market. To counter such extreme transfer-pricing activities, the MNE's
home country may tighten transfer-pricing regulation by lowering the transfer-price cap. In this study, we consider transfer-price
caps to be a regulatory measure.5 Tax authorities audit tax-avoidance behaviors by comparing the prices used in intrafirm trans-
actions with those of similarly uncontrolled transactions between independent parties (i.e., arm's length prices); this method of
the ALP is called the comparable uncontrolled price (CUP) method. In practice, however, it is often difficult to find a comparable
transaction of similar products between independent enterprises. In such cases, other ALP methods such as the cost plus (CP)
method, are applied.6 The price cap which we consider in this study can be thought of as a CP method, as regulatory authorities
often regard the cost plus a reasonable mark-up as a transfer price that meets the ALP.7 The CP method is also a natural ALP in the
case of monopoly, because in many monopolistic environments it is hard to find other transactions of comparable inputs.
We show that transfer pricing entails lowering what the MNE perceives as its marginal cost of production. The “perceived
marginal cost” declines as a result of transfer pricing, because the marginal tax saving that arises from an additional shipment
of the intermediate good serves as the marginal benefit of production for the MNE. Not surprisingly, therefore, FDI occurs even
if it is less efficient to produce the intermediate good internally at its foreign subsidiary, because it can be used as a vehicle to
lessen its tax burden with an inflated internal price. Interestingly, however, profit shifting with the regulation may also benefit
consumers, because transfer pricing lowers the monopolist's perceived marginal cost, thereby leading to more production
which alleviates allocative inefficiency due to market power. Thus, the MNE's home country may want to encourage its firm's
tax-saving, transfer-pricing activities.
The second purpose of the paper is to identify the economic environment in which a country is indeed willing to select a
higher tax rate than another country to encourage its monopolistic firm to engage in FDI and transfer pricing, hoping that the
resulting alleviation of allocative inefficiency enhances its social welfare. To this end, we first investigate a sequential-move,
tax-competition game (i.e., a Stackelberg tax-competition game) played by two welfare-maximizing governments within our
baseline model with transfer-pricing regulation.8 We show that the nature of tax competition can depend on the tightness of

4
http://vietnamlawmagazine.vn/transfer-pricing-unbridled-at-fdi-enterprises-4608.html
5
This assumption has often been made to investigate transfer-pricing regulations. See Raimondos-Moler and Scharf (2002), Peralta et al. (2006), Becker and Fuest
(2012), and Matsui (2012), among others.
6
Other suggested methods include the resale price method, transactional net margin method, and transactional profit split method. See OECD (2010) for more
details.
7
The other ALP methods (except for the CUP method) indirectly regulate the ceiling of the transfer price.
8
For example, Baldwin and Krugman (2004) and Stowhase (2013) also analyze tax competition in a sequential-move game.
9
We treat transfer-pricing regulation as given in our analysis, because the ALP is agreed as the transfer-pricing regulation among OECD member countries. By con-
trast, Mansori and Weichenrieder (2001) and Raimondos-Møller and Scharf (2002) focus on the strategic use of transfer-pricing rules rather than taxes when analyzing
transfer pricing in a single industry. In their analyses, governments non-cooperatively choose their optimal transfer-pricing rule, taking the tax rates as given.
 J.P. Choi et al. / Journal of International Economics 127 (2020) 103367 3

transfer-pricing regulation. In particular, the MNE's home country is willing to set a higher tax rate and tolerate some profit
shifting to an endogenously-determined, tax-haven country if the regulation is sufficiently tight. However, if regulation is too
lax, tax competition leads to a “race to the bottom” that eliminates any incentives for tax-motivated FDI. Interestingly, this implies
that a tax haven country does not always prefer lax transfer-pricing regulation. Thus, the incentives of the FDI host and source
countries can be aligned to set up global regulatory standards for transfer pricing.
Finally, we extend the sequential-move game to a simultaneous-move, tax-competition model with multiple industries, played
by two countries with different population sizes.9 This set-up allows us to endogenously determine the FDI source country that
selects a higher tax rate and investigate the characteristics of such countries. We show that a pure-strategy, subgame-perfect
equilibrium exists if countries are sufficiently different in size. The large country sources FDI, choosing a higher tax rate than
the smaller country. It is willing to create an environment in which its firm engages in FDI to save tax payments because, for
the large country, the benefit of the resulting increase in the consumer surplus is large relative to the tax revenue that it
would have earned if it hosted FDI. The smaller country, on the contrary, undercuts the large country's tax rate, because the
tax revenue collected from the subsidiary of the firm operating in the large market is large. This novel result is consistent with
the fact that most low-tax countries or tax havens are small countries.10 The result also gives some justification to use a model
with the Stackelberg tax-setting nature, because it is reasonable to presume that in reality small tax-haven countries set their
tax rates after observing large countries' tax rates. The welfare impacts of transfer pricing, which are readily obtained in the se-
quential-move game, can also extend to the setting of the simultaneous-move game.
Our study is thus at the intersection of international trade and public economics. Horst (1971) initiated the theory of MNEs in
the presence of different tariff and tax rates across countries and explored the profit-maximizing strategy of a monopolistic firm
selling to two national markets (i.e., how much it should produce in each country and what the optimal transfer price for the
good exported from the parent to the subsidiary would be). Horst (1971) and subsequent studies (e.g., Itagaki, 1979) show
that an 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 tariff schedules among countries. Samuelson (1982) is the first study to point out that
for an MNE subject to the ALP, the arm's length reference price itself can be partially determined by the firm's activities.
Lee (1998) analyzes the effects of profit taxes on firm behaviors and shows that a monopolist engaged in tax avoidance may
produce either more or less than it would in the absence of a tax.11 The focus of his analysis is on deriving the conditions under
which profit taxes are non-neutral in terms of the monopolist's output decision. It is shown that the effects of tax avoidance on
the consumer surplus depend crucially on the shape of the audit probability and penalty rate. Our model identifies a different
channel (i.e., transfer pricing to shift profits across jurisdictions with different tax rates) through which the consumer surplus
is affected when the monopolist aims to minimize the overall tax burden.
In our model, the welfare of the high-tax country can improve because the consumers there benefit from regulated transfer-
pricing activities. In particular, welfare can improve even if intermediate-good production is less efficient with FDI than without
FDI. Matsui (2012) is closely related to our study in that he considers internal transfer prices in MNEs as the channel for tax
avoidance and shows that an MNE's transfer pricing can benefit consumers through an increase in its output. In a Dixit–Stiglitz
model of monopolistic competition, he shows that tax authorities face a trade-off between consumer welfare and tax revenue.12
However, he deals with neither tax-motivated FDI nor tax competition. We also show that the relationship between the transfer-
price cap and consumer price is not monotonic.
Although tax competition has been examined in the presence of profit shifting (Stöwhase, 2005, 2013; Bucovetsky and
Haufler, 2008; Slemrod and Wilson, 2009; Johannesen, 2010; Marceau et al., 2010), only a few theoretical studies take trans-
fer-pricing regulation into consideration. Elitzur and Mintz (1996) study tax competition between the MNE's home country and
its host country, both of which try to maximize individual tax revenues. The MNE operates under rigid transfer-pricing regulation
along the lines of the CP method. In their model, however, there is no consideration of whether the firm sets up a subsidiary in
the host country. Hence, they are unable to discuss the possible inefficiency caused by tax-motivated FDI to a high-cost country.
Moreover, in their model, transfer pricing is used not only for profit shifting but also for an incentive scheme to improve the
subsidiary's efficiency. Peralta et al. (2006) identify an interesting role of transfer-pricing regulation in the context of bidding
for a firm: loosening the control of profit shifting can mitigate tax competition so that winning tax competition becomes more
beneficial in the presence of indigenous firms that also pay corporate taxes.13 In some cases, therefore, the winning country
does not strictly enforce the regulation even though it entails tax revenue leakage. Hosting a firm (or the headquarters of a
firm in our context) is different from getting hold of its profits. Because of its reduced-form structure of the model, however,
they are unable to examine the effect of transfer pricing on the firm's output and its welfare consequences through the effect
on the consumer surplus. In our study, this channel is crucial to identifying the importance of market size in the endogenous de-
termination of the FDI source country and a tax haven that hosts FDI.
Moreover, tax competition between countries of different size has been explored extensively (Bucovetsky, 1991; Wilson, 1991;
Haufler and Wooton, 1999; Stöwhase, 2005, 2013). The literature also shows that the smaller country is likely to set a lower tax
rate in various settings. Stöwhase (2005, 2013) is related to our analysis in the sense that both difference in country size and
profit shifting are explicitly taken into account in a tax-setting game between two tax-revenue-maximizing governments. In

10
These countries include Ireland, Luxembourg, Singapore, Switzerland and the Netherlands, among others.
11
We thank an anonymous referee for directing our attention to this study.
12
Matsui (2012) focuses on the long-run equilibrium in which the MNE's post-tax profits are zero.
13
Becker and Fuest (2012) introduce a model that does not require the presence of indigenous firms and obtain qualitatively the same result as Peralta et al. (2006).
4 J.P. Choi et al. / Journal of International Economics 127 (2020) 103367

our analysis, however, the governments maximize social welfare, caring about not only tax revenue but also profits and consumer
welfare. Stöwhase (2005, 2013) does not deal with transfer-pricing regulation either. We identify an aforementioned novel logic
of an equilibrium outcome that the larger country is willing to set a higher corporate tax rate than the smaller country in the
presence of transfer-pricing regulation.
The rest of the paper is organized as follows. Section 2 introduces the basic set-up of a monopoly model with transfer pricing.
In the model in which the monopolist has already set up a subsidiary in the foreign country, we examine this monopolistic MNE's
transfer-pricing behavior when its home tax rate is exogenously given at a higher rate than the foreign rate. In particular, we
show that the MNE increases its output from the monopolistic production level when its home country imposes a binding cap
on the transfer price. In Section 3, we model a sequential-move, tax-competition game, played by two countries, still assuming
that only one firm, in the first-mover country, sources inputs from the foreign country. We show that the home country selects
a higher tax rate than the foreign country if and only if transfer-pricing regulation is sufficiently tight. Section 4 extends the tax-
competition game to a simultaneous-move game with a monopoly in each country operating in each of the two industries. More-
over, we allow the two countries to differ in size. Then, we show that there is a pure-strategy equilibrium in which the larger
country sets a higher tax rate than the smaller one to become the source country of tax-motivated FDI. Section 5 concludes.

2. Monopoly model with FDI: Transfer pricing and regulations

2.1. Baseline model

There are two countries, Country 1 and Country 2, with possibly different corporate tax rates, t1 and t2. In Country 1, there is a
monopolistic final-good producer, which we often call the MNE. The MNE's headquarters that produces the final good is immobile
and tied to Country 1, while its intermediate input can be procured either from the open market at a price of ω or from its sub-
sidiary in Country 2.14 We assume that one unit of the input is converted into one unit of the final good without incurring any
additional costs and that input production is subject to constant-returns-to-scale technology. We also assume away the costs as-
sociated with FDI.15 The subsidiary's unit cost of input production (i.e., the marginal production cost) is given by c. The final good
is consumed only in Country 1.
The MNE can choose an internal transfer price of γ, when its foreign subsidiary supplies its input to the headquarters. Without
any tax rate differential between the two countries, it does not matter for the MNE to set γ at any level between between c and
the price for the final good. However, as the two countries have different tax rates, the MNE can engage in transfer pricing such
that it chooses γ to optimally allocate its profits between its headquarters and subsidiary and thus minimize its tax burden.
In this section, we assume that the two countries' corporate tax rates are exogenously given such as t1 > t2, meaning that the
MNE in Country 1 has an incentive to manipulate its internal transfer price to avoid tax payments. We also assume throughout
this section that the MNE has already set up its subsidiary in Country 2 to procure the intermediate good, thereby focusing the
analysis on the MNE's transfer-pricing strategy and its home country's regulatory response.

2.2. The benchmark case: No regulation

As a benchmark case, we first consider the choice of transfer price by the MNE without any transfer-pricing regulation. In this
case, the MNE selects its transfer price, γ, and output, q, to maximize the following global profits after tax:

Π ¼ ð1−t 1 Þ ½P ðqÞ−γ q þð1−t 2 Þ ðγ−cÞq ; ð1Þ

|fflfflfflfflfflfflffl{zfflfflfflfflfflfflffl} |fflfflfflffl{zfflfflfflffl}
Downstream Profits Upstream Profits

where P(q) is the downward-sloping inverse demand function faced by the monopolist (with the corresponding demand function
q = D(p)). Note that this expression for the profits is only valid when both downstream and upstream profits are non-negative,
given that the MNE cannot pay negative taxes (i.e., receipt of a transfer from the government). Even so, the expression in (1) is
the relevant profit function in our analysis, because the MNE would never select γ such that profits in either country become neg-
ative; the MNE's tax payment would increase if it increases γ beyond P(q), for example, because it would increase tax payment to
Country 2 while the payment to Country 1 remains zero.16
Decision making is assumed to be centralized. That is, with the firm's objective function above, the headquarters producing the
final good makes an output decision that maximizes the overall profit of the firm, not only the profit of the downstream division.
We can rewrite the objective function of the monopolist, expressed by (1), as

Π ¼ ð1−t 1 Þ½P ðqÞ−ξq; ð2Þ

14
Alternatively, we can assume that ω is the headquarters' constant marginal production cost of the intermediate input.
15
If a fixed cost of setting up the subsidiary exists, FDI becomes less likely. However, the essence of our results would not change.
16
In our (static) model, the MNE has no incentive to make the downstream profit negative. In reality, however, MNEs may have such an incentive if they can carry
forward loss.
 J.P. Choi et al. / Journal of International Economics 127 (2020) 103367 5

where

ð1−t 2 Þc−ðt 1 −t 2 Þγ
ξ≡ : ð3Þ
1−t 1

That is, the MNE facing different tax rates across countries behaves as if its marginal production cost were ξ, which can be con-
sidered as its perceived marginal cost of production. Since the MNE's profit decreases as ξ increases, the monopolist's optimal
choice of γ is to minimize ξ regardless of its output level. Note that ξ is decreasing in its transfer price γ because it can be
used as a vehicle to shift profit from the high-tax country (Country 1) to the low-tax country (Country 2).
As pointed out by Horst (1971), it immediately follows that the optimal choice is to set γ at the highest level that ensures non-
negative downstream profits. This would imply that all profits from a high-tax country are shifted toward to a low-tax country in
the absence of any regulation. This simple model illustrates the need for regulations to counter such MNE's profit-shifting motives
to reduce the tax burden.
To show this rigorously, we first derive q∗(γ), which we define as the MNE's optimal output when its transfer price is set at γ.
Let qm(ξ) denote the optimal monopoly quantity that maximizes the monopolist's profits when its perceived marginal cost is ξ.
Since qm(ξ(γ)) is increasing in γ and hence P(qm(ξ(γ))) is decreasing in γ, there is a unique γ = γ0 such that P(qm(ξ(γ0))) =
γ0. In other words, if γ < γ0 so that P(qm(ξ(γ))) > γ, the optimal output by the MNE is given by q∗(γ) = qm(ξ(γ)) with the down-
stream division in the high-tax Country 1 making a positive profit. However, if γ > γ0, the non-negative profit condition for the
headquarters is binding because P(qm(γ)) < γ. Therefore, the MNE's optimal output q∗(γ) cannot be qm(ξ(γ)). The maximum out-
put that can be used to transfer profit is capped by P−1(γ). That is, if γ > γ0, we have q∗(γ) = P−1(γ) = D(γ), which is decreasing
in γ. We can conclude that
(
m 0
 q ðξðγ ÞÞ if γ < γ
q ðγ Þ ¼ −1 0
P ðγ Þ ¼ Dðγ Þ if γ ≥ γ :

The following lemma helps us determine the MNE's optimal choice of γ.

Lemma 1. Without any transfer-pricing regulation, the MNE's optimal choice of γ cannot be less than γ0.

Proof. Suppose, on the contrary, that γ < γ0. Then, the optimal output by the MNE is given by q∗(γ) = qm(ξ(γ)). With an in-
finitesimal increase in γ, the new transfer price is still less than γ0 while reducing ξ. As a result, the MNE's profit increases,
resulting in a contradiction. □
The above lemma implies that all the downstream profits are transferred to the subsidiary located in the country with a lower
tax rate, that is, P(q∗(γ)) = γ. Given Lemma 1, we can rewrite the optimal choice of γ by the MNE as

max ð1−t 2 Þðγ−cÞDðγ Þ: ð4Þ

γ

This is mathematically equivalent to the standard monopoly pricing problem where γ can be considered as the monopolist's price.
Thus, the MNE chooses a monopoly price equal to P(qm(c)). Proposition 1 summarizes our discussion in the case of no regulation
and calls for regulations to counter the MNE's profit-shifting motives to reduce the tax burden.

Proposition 1. (Full Profit Shifting with the Monopoly Outcome under No Regulation) In the absence of any transfer-pricing regulation,
the MNE shifts all its profits to the country with the lower tax rate by choosing the transfer price at the standard monopoly price, rep-
licating the monopoly outcome in the market.

2.3. Transfer-pricing regulation and the MNE's decisions

In reality, regulations would prevent the choice from being a corner solution and limit the MNE's profit-shifting motives. The
most-widely adopted and agreed-upon standard practice is the ALP, which requires intrafirm transfer prices to mimic the market
prices that would be obtained in comparable uncontrolled transactions at arm's length, as discussed in the Introduction. Although
this principle is conceptually sound and straightforward, its implementation as a regulatory policy may be difficult and subject to
different interpretations. For instance, in the monopoly context, such a comparative analysis is unlikely to be feasible, simply be-
cause there is no comparable transactions as it is the only firm that produces the final good; no other firms acquire similar inputs.
Even if similar inputs are transacted in the market by other firms for different purposes, a firm may argue that the available inputs
are not suitable to meet its specifications, which explains why it is engaged in its production in the first place. In other words,
what constitutes a similar input may not be clear-cut and could be subject to disputes unless comparable inputs are identical.
In our analysis, we simply model Country 1's transfer-pricing regulation as a price cap for the transfer price to limit the extent
to which the MNE can shift its profits to the low-tax country. Country 1 caps the transfer price at γ.17 Under this regulation, the
MNE can only select γ that is less than or equal to γ.

17
As will be discussed further in section 3, the cap on the transfer price can be endogenously derived as the outcome of the MNE's optimal decision if we consider the
government's monitoring and the MNE's concealment costs with the CP method.
6 J.P. Choi et al. / Journal of International Economics 127 (2020) 103367

Fig. 1. MNE's optimal output level.

We derive the MNE's optimal choice of γ under the restriction and its optimal output level as a function of γ. As shown in the
previous subsection, the MNE raises γ as much as it can, all the way up to the monopoly price P(qm(c)) if possible. If γ < γ0, for
example, the MNE can raise its profit by increasing γ, thereby decreasing the perceived marginal cost ξ. If γ ≥ γ0, it can also raise
its profit by increasing γ because γ here serves as the price of the product and increasing γ is equivalent to bringing the price
closer to the monopoly price P(qm(c)), as shown in (4). Therefore, the MNE chooses the largest possible γ under the price cap
of γ, as long as this regulation is binding, that is, as long as γ ≤ Pðqm ðcÞÞ. If γ > Pðqm ðcÞÞ, the MNE simply chooses γ at the mo-
nopoly price and produces qm(c) units of the good in such a case. Fig. 1 illustrates the MNE's optimal output level as a function
of the price cap γ (in the case of linear demand, as given by (5)−(7) below). An increase in γ increases the optimal output if c
≤ γ ≤ γ0 but decreases it if γ 0 < γ ≤ Pðqm ðcÞÞ.

Lemma 2. With the given transfer-price cap γ, the MNE chooses γ ¼ γ if γ ≤ Pðqm ðcÞÞ, and γ = P(qm(c)) if γ > Pðqm ðcÞÞ. The optimal
output level is given by
8 m
>
< q ðξðγÞÞ if c ≤ γ ≤ γ ;
0
  m 
q ðγ Þ ¼ P −1 ðγ Þ 0
if γ < γ ≤ P q ðcÞ ;
>
: m  m 
q ðcÞ if γ > P q ðcÞ :

As an illustration and for later purposes, we explicitly derive some of the above key functions and variables in the case of
linear demand, such that P(q) = 1 − bq:

m 1−c ðt 1 −t 2 Þðγ−cÞ
q ðξðγ ÞÞ ¼ þ ; ð5Þ
2b 2bð1−t 1 Þ

−1 1−γ
P ðγ Þ ¼ ; ð6Þ
b

0 ð1−t 1 Þ þ ð1−t 2 Þc
γ ¼ : ð7Þ
ð1−t 1 Þ þ ð1−t 2 Þ

2.4. Optimal taxation and transfer-pricing regulation

Fig. 1 shows the possibility that Country 1, the home country of the MNE, can mitigate the monopoly distortion by relaxing
transfer-pricing regulation and hence inducing the MNE to produce more. Here, we investigate whether Country 1 can indeed en-
hance its social welfare by optimally adjusting the transfer-price cap. In addition, we derive its optimal policy when both the
transfer price cap and the corporate tax rate can be freely chosen, for the given tax rate in Country 2 (i.e., the country that
hosts the MNE's subsidiary).
 J.P. Choi et al. / Journal of International Economics 127 (2020) 103367 7

Fig. 2. The effect of an increase in the source country's tax rate.

Country 1's social welfare is defined as the sum of the consumer surplus (CS), producer surplus (PS), and tax revenue (TR). The
producer surplus here is the MNE's total profits derived from both its headquarters and subsidiary, which can be written as (2)
with the use of the perceived marginal cost ξ. Then, it follows from
Z q ðγÞ    
CS ¼ P ðqÞdq−P q ðγ Þ q ðγ Þ;
0      
PS ¼ Π1 ¼ ð1−t 1 Þ P q ðγ Þ −ξðγ Þ q ðγ Þ;
     
TR ¼ t 1 P q ðγ Þ −γ q ðγÞ

that Country 1's social welfare can be written as

Z q ðγ Þ 
s 
W1 ¼ P ðqÞ−ξ ðγÞ dq; ð8Þ
0

where the social marginal cost ξs is defined and can be written with the use of (3) as

s
ξ ðγ Þ ¼ ð1−t 1 ÞξðγÞ þ t 1 γ
ð9Þ
¼ c þ t 2 ðγ−cÞ:

s
Obviously, W1 increases with the MNE's output level, q ðγÞ, and decreases with the social marginal cost, ξ ðγÞ.

Lemma 3. The optimal transfer price cap for Country 1 is not greater than γ0. It equals γ0 if t2 is small, while it is less than γ0 if t2 is
sufficiently large to be close to t1.
s
Proof. As Fig. 1 shows, q ðγÞ increases by lowering γ when γ > γ0 . Because a decrease in γ also decreases ξ ðγÞ as (9) shows,
W1 increases by lowering γ in this range of γ. Thus, the optimal γ is not greater than γ0.
s
In the range with c ≤ γ ≤ γ 0 , a reduction in γ involves a trade-off between the beneficial effect of lowering ξ ðγÞ and adverse

effect of decreasing q ðγÞ. If t2 is small, the latter adverse effect outweighs the former effect. Indeed, if t2 = 0 at the extreme, (9)
s s
shows that ξ ðγÞ ¼ c, and hence decreasing γ from γ0 only reduces q ðγÞ leaving ξ ðγÞ as it is. Hence, it is optimal for Country 1 to
s
set γ ¼ γ . However, if t2 is not so small, the beneficial effect of lowering ξ ðγÞ outweighs the adverse effect at γ ¼ γ 0, making it
0

optimal for Country 1 to set γ below γ0. At the extreme where t2 is so large to be equal to t1, we know from (3) that ξðγÞ ¼ c and
s
thus q ðγÞ ¼ qm ðcÞ. Hence, the optimal γ is as small as c to minimize ξ ðγÞ. □
What if Country 1 can adjust t1, for the given t2, as well as choose an arbitrary γ? Redefining ξ as a function of t1 as well as γ
(i.e., ξðγ; t 1 Þ), we obtain from (3) that ∂ξ=∂t 1 ¼ −ðγ−cÞð1−t 2 Þ=ð1−t 1 Þ2 < 0. Since qm is decreasing in ξ, this means that qm in-
creases as t1 rises, except at γ ¼ c. This relationship is easily verified from (5) in the linear demand case, as illustrated by the
counterclockwise rotation of the qm line in Fig. 2. As this figure shows, an increase in t1 coupled with a decrease in γ (from γ0
s
to γ0 ) enables Country 1 to increase q ðγÞ and decrease ξ ðγÞ at the same time.
'

Proposition 2. The MNE's home country can enhance its social welfare by increasing its corporate tax rate and tightening transfer-pric-
ing regulation simultaneously. The monopoly distortion is completely eliminated, meaning social welfare is maximized at the limit where
the tax rate approaches 1, whereas the transfer-price cap approaches the marginal cost of the intermediate good.

Proof. As t1 increases, so does qm ðξðγ; t 1 ÞÞ for any γ. It also follows from γ0 = P(qm(γ0, t1)), the definition of γ0, that γ0 de-
creases accordingly, as Fig. 2 illustrates. Thus, Country 1 can induce the MNE to produce more and lower the social marginal
8 J.P. Choi et al. / Journal of International Economics 127 (2020) 103367

cost, as seen in (9), by raising t1 and simultaneously lowering γ to keep γ ¼ γ 0 . That is, Country 1 can increase its social welfare
by raising t1 and lowering γ, while keeping P −1 ðγÞ ¼ qm ðξðγ; t 1 ÞÞ. As shown in (3), at the limit where t1 approaches 1, γ ap-
s
proaches c, meaning that ξ remains finite to satisfy the above equality. Noting that ξ ðγÞ also approaches c, we conclude that
Country 1 can maximize its social welfare by raising t1 and lowering γ, thereby attaining a market outcome equivalent to that
under perfect competition at the limit where t1 = 1 and γ ¼ c. □
In the current scenario in which the MNE's home country has two policy tools, namely the transfer-price cap and corporate tax
rate, it can manage to increase the MNE's output and minimize tax leakage at the same time. Even though the country lowers γ,
which dampens the MNE's incentive to produce more, it can increase t1 sufficiently large to more than offset the incentive; hence
the MNE is induced to increase its output because of the increased incentive to save tax payments. This benefits the country be-
cause it can increase the MNE's output while reducing tax leakage. Indeed, this proposition indicates that the country can lower
the price of the good all the way to the MNE's actual marginal cost of production by tailoring its tax rate and price-cap.

3. Tax competition between the MNE's home country and the host country

We have thus far analyzed how the MNE reacts to transfer-pricing regulation and corporate taxation in its home country, and
discussed the optimal policy for the home county. Proposition 2 reveals that the home country can completely eliminate the mo-
nopoly distortion through extreme transfer-pricing regulation and taxation. In reality, however, the transfer price is regulated by
the ALP, which suggests that it should be set at the level that provides reasonable markups to the marginal production cost of the
intermediate good. On the contrary, sovereign countries have more freedom to set their individual corporate tax rates. In this sec-
tion, therefore, we fix γ, and discuss how the tax rates are determined between the MNE's home government and host govern-
ment. We can motivate this approach by linking the cap on transfer prices to the costs of monitoring transfer pricing behavior of
the MNE. More specifically, consider a scenario in which the CP method is adopted as an application of the ALP due to the lack of
comparable inputs transacted in the market by uncontrolled parties. The CP method mandates that the transfer price should re-
flect the production cost of the input internally transacted. This implies that a deviation of MNE's internal price from its true MC
entails “concealment costs” or expected punishment for the deviation as in Allingham and Sandmo (1972) and Kant (1988). These
costs will be higher with better institutional monitoring. We thus can interpret the cap on the transfer price as the outcome of the
MNE's optimum behavior because more efficient monitoring will lower the maximum transfer price that will be optimal for the
MNE.18
We consider a sequential-move game in which Country 1 (the MNE's home country) first sets its tax rate, and then Country 2
tries to undercut Country 1's tax rate to attract FDI, given the transfer-pricing regulation represented by the price cap of γ. We
show that the nature of tax competition can depend on the tightness of transfer-pricing regulation. In particular, if γ is not too
large, Country 1 selects a relatively high tax rate to allow Country 2 to undercut its tax rate in a unique subgame perfect equilib-
rium. The monopolist in Country 1 sets up a subsidiary in Country 2 and engages in transfer pricing to save tax payments. Country
1 is willing to let its own firm do so, because it leads to an expansion of production and helps increase its social welfare. If reg-
ulation is too lax (i.e., γ is sufficiently large), however, tax competition leads to the “race to the bottom” and eliminates any in-
centive for tax-motivated FDI.
The next section considers a simultaneous-move game between the two countries, where the direction of FDI is endogenously
determined. Although the model setting in the next section may be considered as more general than the one here, it entails mul-
tiple equilibria in some cases and no pure-strategy equilibrium in other cases. Thus, it is meaningful to consider a sequential-move
game that pins down a unique equilibrium, to gain insights into the tax competition when a monopolistic firm may engage in
transfer pricing. Moreover, the timing assumption seems rather realistic, especially when Country 2 merely acts as a tax haven.
Here, we relax the assumption that the MNE has already set up its subsidiary in Country 2. That is, we now explicitly incor-
porate the firm's choice of FDI for intermediate-good procurement, while maintaining the assumption that the firm's headquarters
and its production facility for the final good is located in Country 1.
Social welfare for either country depends on whether the firm engages in FDI. In the case where the firm engages in FDI, Coun-
try 1's social welfare can be written as

Z q ðt 1 ;t 2 Þ 
s 
W1 ¼ P ðqÞ−ξ ðt 2 Þ dq; ð10Þ
0

s
where ξ ðt 2 Þ ¼ c þ t 2 ðγ−cÞ, and the MNE's output is now written as a function of t1 and t2. When the firm does not engage in FDI,
then it loses the opportunity of tax avoidance while its marginal cost is ω. Country 1's social welfare in this case is given by

Z q m ðω Þ 3ð1−ωÞ2
W1 ¼ ½P ðqÞ−ωdq ¼ :
0 8

18
See Choi et al. (2018) for an analysis that explicitly accounts for concealment costs in the derivation of the optimal transfer price.
 J.P. Choi et al. / Journal of International Economics 127 (2020) 103367 9

Country 2's social welfare, by contrast, only consists of the tax revenue from the MNE's subsidiary, and can be written as

W 2 ¼ t 2 ðγ−cÞq ðt 1 ; t 2 Þ; ð11Þ

if the firm engages in FDI, whereas it equals 0 otherwise.

With sequential moves, Country 1 maximizes W1 subject to

t 2 ∈BR2 ðt 1 Þ ≡ arg max t 2 ðγ−cÞq ðt 1 ; t 2 Þ;
t2

taking into account that Country 2 will undercut its tax rate to obtain tax revenue. We solve this problem by backward induction
as usual.
For the sake of concreteness, we henceforth assume that the demand for the final good is linear such that P(q) = 1 − bq.

3.1. Country 2's best response

Let us first consider the best response for Country 2, whose social welfare is given by (11). Now, we see from (5) that qm de-
creases with t2. Therefore, if t2 is smaller than the threshold given by (7), the firm's output is constant at P −1 ðγÞ. Hence, Country 2
is better off by increasing t2 up to the threshold. If t2 is sufficiently large that qm ðξðt 1 ; t 2 ÞÞ ≤ P −1 ðγÞ, we have q∗(t1, t2) = qm
(ξ(t1,t2)). Therefore, we can use (5) to write (11) as
 
1−c ðt 1 −t 2 Þðγ−cÞ
W 2 ¼ t 2 ðγ−cÞ þ :
2b 2bð1−t 1 Þ

It is readily verified that the t2 that maximizes W2 in this case is given by

1−c−ð1−γÞt 1
t2 ¼ : ð12Þ
2ðγ−cÞ

This t2 is a valid best response only if the resulting γ0 is greater than or equal to γ. To see the condition for this requirement, we
substitute (12) into (7) and find that γ 0 ≥γ if and only if

3ð1−γÞ−ðγ−cÞ
t1 < t ≡ :
3ð1−γÞ

If this condition is violated, t2 is chosen to satisfy γ0 ¼ γ, i.e.,

1−γ
t 2 ¼ 1− ð1−t 1 Þ:
γ−c

Fig. 3. Country 2's reaction curve.

10 J.P. Choi et al. / Journal of International Economics 127 (2020) 103367

Another condition that is required for the t2 in (12) to be the best response is t2 ≤ t1, which reduces to

1−c
t 1 ≥t ≡ :
2ðγ−cÞ þ ð1−γÞ

Consequently, Country 2's best response function is given by

8
>
> t1 if t 1 ≤ t;
>
>
>
>
>
< 1−c−ð1−γÞt 1 if t < t ≤ t;
1
B2 ðt 1 Þ ¼ 2ðγ−cÞ ð13Þ
>
>
>
>
>
> 1−γ
>
: 1− ð1−t 1 Þ if t < t 1 < 1;
γ−c

where it is readily verified that t < t. Fig. 3 depicts Country 2's best response function.
The intuition for this result is as follows. When t1 is low (more precisely, when t 1 ≤ t ), it is optimal for Country 2 to just
undercut Country 1's tax rate, as reducing the tax rate further and inducing more output does not raise tax revenue. Thus, in
this range, Country 2's tax rate is a strategic complement to Country 1's tax rate. However, if t1 becomes moderately large (i.e.,
t < t 1 ≤ t), then it can be optimal for Country 2 to lower the tax rate to induce more output and hence tax revenue in response
to an increase in Country 1's tax rate.19 In this case, Country 2's tax rate is a strategic substitute to Country 1's tax rate. Finally, if
the tax rate in Country 1 becomes sufficiently large (i.e., t < t 1 < 1), the incentive to shift profit by raising production becomes too
large, while the zero-profit condition for the MNE's headquarters is binding. As the output is capped and cannot increase any
more by lowering t2, Country 2 follows suit as Country 1 increases its tax rate. Once again, Country 2's tax rate becomes a strategic
complement to Country 1's tax rate.

3.2. Country 1's optimal tax rate

Now, we analyze Country 1's optimal tax rate when it considers the undercutting threat by Country 2. To avoid unnecessary
complication in the exposition, we assume henceforth (1 − c)2/2b > 3(1 − ω)2/8b, which means that the highest social welfare
Country 1 can achieve through regulation and tax policies in the presence of FDI, which is the same as welfare under perfect com-
petition as Proposition 2 shows, is greater than its social welfare without FDI. If this condition is violated, Country 1 will never set
a higher tax rate than Country 2. In the subgame perfect equilibrium, both countries thus set their tax rates to 0, whereas the
monopolist engages in FDI if and only if c < ω. Here, we make the following assumption

2 3 2
ð1−cÞ > ð1−ωÞ ; ð14Þ
4

which is equivalent to (1 − c)2/2b > 3(1 − ω)2/8b to focus on the interesting and meaningful case.
As Country 2 will always choose a point on its reaction curve, Country 1's optimal choice can be considered as the choice of a
point on Country 2's reaction curve, as shown in Fig. 3. Then, we have the following proposition.

Proposition 3. The subgame perfect equilibrium of tax competition can be summarized as follows. There is a critical level of γ; denoted
by γ  ð< ð1 þ cÞ=2Þ, such that (i) the equilibrium tax profile is ðt 1 ; t 2 Þ ¼ ðt; 2=3Þ if γ < γ  and (ii) (t1, t2) = (0, 0) otherwise.

Proof. We consider three segments on Country 2's reaction curve: the segments between A = (0, 0) and B ¼ ðt; tÞ, B and
C ¼ ðt; 2=3Þ, and C and D = (1, 1) (Fig. 3). First, we argue that Country 1 prefers point A to any other points in the segment
AB because along this segment, ξ = [(1 − t2)c − (t1 − t2)γ]/(1 − t1) = c. Thus, the monopolist's output is fixed at q∗
s
(t1, t2) = qm(c) = (1 + c)/2. By contrast, ξ ¼ c þ t 2 ðγ−cÞ is increasing as we move from A to B as t2 increases with t1. Thus,
point A is preferred to any other points in the segment AB. Moreover, point C is preferred to any other points in the segment
BC. Along the BC segment, output is increasing from qm(c) to P −1 ðγÞ: In addition, Country 2 responds by decreasing its tax rate
as t1 increases. Thus, on both accounts, Country 1 is better off, and hence it prefers point C to any other points in the segment
BC. Finally, in the segment CD, output is once again fixed at the level of P −1 ðγÞ: Thus, as in the segment AB, ξ ¼ c þ t 2 ðγ−cÞ in-
s

creases as we move from C to D as t2 increases with t1.

19
An increase in t2 directly contributes to Country 2's tax revenue, while it hurts Country 2 due to the resulting reduction of the MNE's output. It can be shown that in
this region, both effects increase as t1 rises, but the latter output effect grows more than the former direct effect. Consequently, Country 2 lowers t2 in response to an
increase in t1.
 J.P. Choi et al. / Journal of International Economics 127 (2020) 103367 11

This leaves us two candidate equilibrium points, A and C. When point A is chosen, Country 1's welfare is given by
( )
3ð1−ωÞ2 3ð1−cÞ2
W 1 ðAÞ ¼ W 1 ð0; 0Þ ¼ max ; :
8b 8b

In this case, the monopolist's FDI decision is always efficient from the viewpoint of global production efficiency. When point C is
chosen by the monopolist, it can be easily verified that

  ð3−2c−γÞð1−γ Þ
W 1 ðC Þ ¼ W 1 t; 2=3 ¼ : ð15Þ
6b

Observe that W1(C) is decreasing in γ, while W 1 ðCÞjγ ¼ c ¼ ð1−cÞ2 =2b > W 1 ðAÞ under the assumption expressed by (14), and
W 1 ðCÞj 1þc ¼ 5ð1−cÞ2 =24b < W 1 ðAÞ: Therefore, there is a unique γ  between c and P(qm(c)) = (1 + c)/2 such that
γ¼
2
W 1 ðCÞjγ ¼ γ ¼ W 1 ðAÞ: This implies that in the subgame perfect equilibrium, (i) ðt 1 ; t 2 Þ ¼ ðt; 2=3Þ if γ < γ and (ii) (t1, t2) =
(0,0) otherwise. □
Proposition 3 reveals that Country 2 will undercut Country 1's tax rate so that the firm in Country 1 establishes a foreign subsidiary to
engage in transfer pricing if and only if transfer-pricing regulation is tight. Otherwise, the two countries race to the bottom in tax com-
petition, and the firm undertakes FDI if and only if c < ω (i.e., no tax-motivated FDI occurs). It is rather counter-intuitive that loosening
regulation deters transfer pricing. If the price cap is high, however, so is the resulting price for the final good. Then, Country 1 would ob-
tain an insufficient benefit from allowing its firm to engage in transfer pricing. Consequently, it selects zero tax rate to prevent Country 2
from undercutting the tax rate; then the firm has no incentive to engage in FDI to avoid tax.

3.3. Welfare impacts

Next, we assess the welfare impacts of opening the possibility of the firm's FDI. We examine the impact on world welfare,
s
W1 + W2, as well as on each country. It follows from (10) and (11) as well as t 2 ðγ−cÞ−ξ ¼ −c that world welfare can be writ-
ten as
Z q ðt 1 ;t 2 Þ
W1 þ W2 ¼ ½P ðqÞ−cdq; ð16Þ
0

when the firm engages in FDI. As the expression in (16) indicates, there is a trade-off between enhancing allocative efficiency,
namely, the efficiency attained by increasing quantity from the monopolistic production level, and sacrificing production efficiency,
namely, the cost-minimizing production location. As we have seen, Country 1 can induce the firm to produce more, only by en-
couraging it to engage in FDI in Country 2, which may have a higher cost of production. It is readily shown, on the contrary, that
W2 = 0 and hence world welfare equals W1 = 3(1 − ω)2/8 if the firm does not undertake FDI.
Let us first consider the case in which γ < γ . In this case, the firm engages in FDI if and only if ω > ξ, as illustrated in Fig. 4. Since
Country 2 enjoys a positive welfare level if and only if the firm sets up its subsidiary there, Country 2's welfare improves if and only if
ω > ξ. As Fig. 4 illustrates, Country 1's welfare increases if and only if ω > ω∗1, and world welfare enhances if and only if ω > ω∗.
Country 1's welfare increases if and only if W1(C) > 3(1 − ω)2/8, which is readily shown to be equivalent to ω > ω∗1,
where

 2 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ω1 ≡ 1− ð1−γ Þð3−γ−2cÞ:
3

The threshold ω∗1 lies between ξ and c. We know from the definition of γ that if ω = c, then W1(C) > 3(1 − ω)2/8. That is, under
tight transfer-pricing regulation, Country 1 prefers an equilibrium with FDI because of the increase in allocative efficiency, and this
is particularly so in the absence of production inefficiency when ω ≥ c. As ω falls and hence c − ω rises, the gain in allocative
efficiency decreases, while the loss in production efficiency increases. The open-market marginal cost ω reaches the threshold
before it reaches ξ, where the gain in allocative efficiency disappears.

Fig. 4. Welfare impacts.

12 J.P. Choi et al. / Journal of International Economics 127 (2020) 103367

Similarly, world welfare increases if and only if W1(C) + W2(C) > 3(1 − ω)2/8, where W1(C) + W2(C) is calculated from (16)
with C ¼ ðt; 2=3Þ. Deriving the threshold

 2 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ω ≡ 1− pffiffiffi ð1−γ Þð1 þ γ−2cÞ
3

from

ð1−γ Þð1 þ γ−2cÞ 3ð1−ω Þ2

W 1 ðC Þ þ W 2 ðC Þ ≡ ¼ ;
2b 8b

it is readily shown that ω∗ < ω∗1, and W1(C) + W2(C) > 3(1 − ω)2/8 if and only if ω > ω∗. Now, both Country 1 and Country 2
enjoy gains from allocative efficiency. Consequently, the threshold for world welfare is less than that for Country 1's welfare.
Under tight transfer-pricing regulation, both countries benefit from the firm's FDI and transfer pricing, even though it involves
a loss of production efficiency. However, these gains diminish as foreign production costs c rise above home production costs ω.
In the case where γ > γ , (t1, t2) = (0,0) in the equilibrium, and the firm engages in FDI if and only if c < ω. Production ef-
ficiency is guaranteed with FDI. The countries, however, do not enjoy gains from allocative efficiency with FDI because the firm
sets γ = c and the standard monopoly equilibrium is realized.
The above analysis is summarized in the following proposition.

Proposition 4. (i) Suppose γ < γ. Then, there are three welfare level thresholds: ξ < ω∗ < ω∗1. FDI arises and Country 2's welfare im-
proves if and only if ω > ξ, Country 1's welfare increases if and only if ω > ω∗1, and the world welfare increases if and only if ω > ω∗.
(ii) Suppose γ > γ  . Then, FDI arises if and only if ω > c. FDI improves Country 1's welfare but does not affect Country 2's welfare.

Before closing this section, we should mention an interesting, counter-intuitive effect of transfer-pricing regulation on Country 2.
Since Country 2, as a tax haven, can generate tax revenue from the MNE only when γ < γ  ; it does not always prefer laxer transfer-
pricing regulation, once we take into consideration that Country 1's willingness to allow transfer pricing depends on the tightness
of such regulation. Thus, there is some room for harmonizing and setting up global regulatory standards in transfer pricing. As Fig. 4
indicates, tighter transfer-pricing regulation benefits both countries as long as the loss in production efficiency is small.

4. Endogenous selection of the FDI source country

We have thus far assumed that the MNE's headquarters is located in Country 1 and thus Country 1 always sources FDI. What
characteristics does the FDI source country have? Does it always have an incentive to select a higher tax rate to become an FDI
source country? To answer these questions, we consider here a simultaneous-move, tax-competition game, played by two coun-
tries, either of which can be an FDI source country. Each country has its own industry: Industry 1 in Country 1 and Industry 2 in
Country 2.20 In Country 1, a monopolist, Firm 1, sells its product only in Country 1. Similarly, in Country 2, Firm 2 in Industry 2
produces a good and sells it only in Country 2. Each firm, however, can set up its subsidiary in the other country to produce an
intermediate good, which is shipped to its headquarters for final-good production. Demand for Firm 1's final good is characterized
by the inverse demand function P1(q1) = 1 − q1/L1, whereas that for Firm 2's good is given by P2(q2) = 1 − q2/L2, where L1 and
L2 represent the two countries' populations, respectively.21
We suppose L1 ≥ L2, and show that otherwise symmetric countries have different tax strategies in the equilibrium when L1 is
sufficiently larger than L2: the larger country, Country 1, will select a tax rate higher than Country 2's to deliberately allow its
own firm to engage in transfer pricing to save tax payments. Thus, this section provides a theoretical foundation for the commonly
observed phenomenon that tax-haven countries tend to be small and that firms in large countries set up subsidiaries there to en-
gage in transfer pricing.
Let us first derive Country 1's best response function. If Country 1 chooses a higher tax rate than Country 2, it becomes an FDI
source country that maximizes its social welfare, given by (10), for the given t2. We find immediately that t1 can affect W1 only
through q∗(t1,t2). As shown in Fig. 1, if t1 is sufficiently large that γ 0 ≤ γ, Firm 1 selects q ¼ P −1 ðγÞ, which is independent of t1. On
the contrary, if t1 is so small that γ0 > γ, Firm 1 chooses qm ðξðγÞÞ, which is smaller than P −1 ðγÞ and hence less favorable for Coun-
try 1. Thus, Country 1 chooses any t1 that yields γ0 ≤ γ, so that the price for the final good attains γ, the minimum of all the prices
that the MNE possibly chooses under the transfer-pricing regulation. It follows from (7) that the threshold t1 that yields γ0 ¼ γ
should satisfy

γ−c
1−t 1 ¼ ð1−t 2 Þ; ð17Þ
1−γ

20
We thank Arnaud Costinot (the coeditor of this journal) for suggesting this extended model.
21
The specified linear demand functions can be derived if all consumers have a common demand function of q = 1 − p, for example.
 J.P. Choi et al. / Journal of International Economics 127 (2020) 103367 13

so that Country 1's best response function as an FDI source country is a correspondence given by

s γ−c
B1 ðt 2 Þ ¼ t 1 1− ð1−t 2 Þ ≤ t 1 < 1 : ð18Þ
1−γ

Any effective price cap must satisfy γ < ð1 þ cÞ=2 ¼ Pðqm ðcÞÞ, which directly means that ðγ−cÞ=ð1−γÞ < 1. Thus, we have
Bs1(t2) > t2 for any t2 < 1.
If Country 1 chooses a lower tax rate than Country 2, on the contrary, it becomes an FDI host country and its best response is
given by
8
>
> t2 if t 2 ≤ t;
>
>
>
> γ−c þ ð1−γ Þð 1−t Þ
< 2
if t < t 2 ≤ t;
h
B1 ðt 2 Þ ¼ 2ðγ−cÞ ð19Þ
>
>
>
> 1−γ
>
> ð1−t 2 Þ if t < t 2 < 1;
: 1−
γ−c

as it was derived as a best response function for Country 2, shown in (13), in the last section.
The following lemma shows the two countries' individual best response functions.

Lemma 4. For a given L1/L2, there is a threshold ~t 2 such that Country 1's best response function is
(
if t 2 < ~t 2 ;
s
B1 ðt 2 Þ
B1 ðt 2 Þ ¼ h ð20Þ
B1 ðt 2 Þ otherwise;

and ~t 2 is decreasing in γ and increasing in L1/L2. Country 2's best response function is symmetric to Country 1's. The only substantial
difference is that the threshold ~t 1 decreases with L1/L2.

Proof. See the Appendix. □

Fig. 5 illustrates Country 1's reaction curve, which is characterized by (20). The threshold ~t 2 increases if γ decreases; that is,
Country 1 is more willing to select a higher tax rate than Country 2 to become the FDI source country when the transfer-pricing
regulation is tight. As we have seen in the derivation of the FDI source country's best response function, the source country always
induces its MNE to select the price for its good at γ. The lower the γ, the greater the positive effect of transfer-pricing in mitigat-
ing the monopoly distortion. Therefore, Country 1 is more willing to become the FDI source country under a tighter transfer-
pricing regulation.
Given the above lemma, we are ready to derive the Nash equilibrium of the simultaneous-move, tax-competition game played
by the two countries. We consider two cases. In the first case, γ is sufficiently small that ~t 1 ¼ ~t 2 ≥2=3 holds when L1 = L2. In the
second case, γ is large, meaning that ~t 1 ¼ ~t 2 < 2=3 when L1 = L2. The threshold of 2/3 here is derived from Bhi ðtÞ ¼ 2=3 for i = 1,
2 (recall Fig. 3).
As a preliminary investigation, consider the knife-edge case where ~t 1 ¼ ~t 2 ¼ 2=3 holds when L1 = L2. The two countries are
perfectly symmetric when L1 = L2. If t 2 ¼ ~t 2 ¼ 2=3, one of Country 1's best responses to t2 = 2/3 is to select t 1 ¼ t to become
the FDI source country, as Fig. 6 shows, while Country 2's best response to t 1 ¼ t is to select t2 = 2/3. Similarly, if

Fig. 5. Country 1's reaction curve.

14 J.P. Choi et al. / Journal of International Economics 127 (2020) 103367

Fig. 6. Nash equilibrium when L1 = L2 and ~t 1 ¼ ~t 2 ¼ 23.

t 1 ¼ ~t 1 ¼ 2=3, one of Country 2's best response is to select t 2 ¼ t, while Country 1's best response to t 2 ¼ t is to select t1 = 2/3.
Consequently, as Fig. 6 illustrates, there are two Nash equilibria ðt 1 ; t 2 Þ ¼ ðt; 2=3Þ; ð2=3; tÞ.
When γ is so small that ~t 1 ¼ ~t 2 > 2=3 holds with L1 = L2, there is a continuum of equilibria, in each of the two cases where
Country 1 becomes the source country and where it becomes the host. As illustrated in Fig. 7, the Nash equilibrium set is
given by

1−γ 1−γ
ðt 1 ; t 2 Þj1−t 2 ¼ ð1−t 1 Þ; t 1 ≥ ~t 1 ; t 2 ≤ ~t 2 ∪ ðt 1 ; t 2 Þj1−t 1 ¼ ð1−t 2 Þ; t 2 ≥ ~t 2 ; t 1 ≤ ~t 1 : ð21Þ
γ−c γ−c

In the first equilibrium set, Country 1 plays the role of the source country, whereas Country 2 plays the role of the host. The
roles are switched in the second equilibrium set. In this case, each country is willing to be the source because γ is so small that
transfer-pricing by its own firm entails substantial elimination of monopoly distortion. Each country also has an incentive to be
the host country because the source country's tax rate is so high that it can maintain a relatively high tax rate to obtain sufficient
tax revenue, while undercutting the other country's.
As L1/L2 increases, however, the equilibrium set under which Country 1 is the host shrinks and eventually disappears. As Fig. 8
depicts, ~t 1 decreases and ~t 2 increases as L1/L2 rises. Consequently, the first equilibrium set in (21) expands, whereas the second set
shrinks. The second equilibrium set eventually disappears when ~t 1 becomes smaller than 2/3. In this asymmetric case, the

Fig. 7. Nash equilibrium when L1 = L2 and ~t 1 ¼ ~t 2 > 23.

 J.P. Choi et al. / Journal of International Economics 127 (2020) 103367 15

Fig. 8. Nash equilibrium when L1 > L2.

equilibrium set becomes

1−γ
ðt 1 ; t 2 Þj1−t 2 ¼ ð1−t 1 Þ; t 1 ≥t 1 ; t 2 ≤ ~t 2 : ð22Þ
γ−c

In the second case where ~t 1 ¼ ~t 2 < 2=3 holds with L1 = L2, there is no pure-strategy Nash equilibrium as long as the two
countries are completely symmetric. As L1/L2 rises, however, the Nash equilibrium set, as described by (22) and graphically rep-
resented in Fig. 8, will appear.
We summarize the above results as a proposition.

Proposition 5. If Country 1 is sufficiently large relative to Country 2, there is a Nash equilibrium set of the tax-competition game, in
which Country 1 sets a higher tax rate than Country 2 so that Firm 1 sets up a subsidiary in Country 2 and engages in transfer pricing.
If the two countries are of a similar size and regulation is so tight that the transfer-price cap is low, then in addition to this Nash equi-
librium set, there is a Nash equilibrium set in which Country 2, the smaller country, selects a higher tax rate to become the FDI source
country.

On the one hand, an increase in tax revenue is large for the FDI host country if the FDI source country is large so that its firm's
output level is high. On the other hand, the source country's benefit from mitigating monopoly distortion through its own firm's
transfer-pricing activity is large if the source country itself is large. A small country tends to be the host, because the former tax-
revenue effect is likely to outweigh the latter effect on consumer surplus, whereas a large country tends to be the source exactly
for the opposite reason. As in reality, the host country will not “just” undercut the source country's tax rate in equilibrium,
because the source country selects a sufficiently high tax rate in order to induce its own firm to lower the price while the host
country optimally selects a tax rate that maximizes tax revenue instead of “just” undercutting the source country's tax rate.

5. Concluding remarks

We have analyzed MNE's incentives to manipulate an internal transfer price to take advantage of tax differences across coun-
tries and discussed implications of transfer-pricing regulation as a countermeasure against such profit shifting. We found that tax-
motivated FDI may entail inefficient internal production but could benefit consumers. Thus, encouraging transfer-pricing behavior
to some extent can enhance social welfare.
We have also considered tax competition between (exogenously determined) source and host countries to explore the inter-
play between tax competition and transfer-pricing regulation. In tax competition, each government non-cooperatively sets the tax
rate to maximize its social welfare. We showed that the nature of tax competition can depend on the tightness of transfer-pricing
regulation. In particular, the source country is willing to set a higher tax rate and tolerate profit shifting to a tax-haven country
under sufficiently tight regulation. However, if regulation is too lax, tax competition leads to a “race to the bottom” and eliminates
any incentives for tax-motivated FDI. This finding implies that a tax-haven country does not always prefer lax transfer-pricing reg-
ulation. Thus, the incentives of the host and FDI source country can be aligned to set up global regulatory standards for transfer
pricing.
Finally, we have extended our tax competition model to endogenously determine the identity of the source country in a set-up
with multiple industries. This extended set-up allows us to rationalize our basic model by deriving an equilibrium outcome that
the larger country is willing to set a higher corporate tax rate than the smaller country in the presence of transfer-pricing
16 J.P. Choi et al. / Journal of International Economics 127 (2020) 103367

regulation. We often observe in reality that firms in large countries establish subsidiaries in small tax-haven countries and engage
in transfer pricing. The result that the large country sources FDI also provides some justification for using a model with the
Stackelberg tax-setting nature because it is reasonable to presume that in reality small tax-haven countries set their tax rates
after observing large countries' tax rates. The welfare impacts of transfer pricing obtained in the sequential-move game can
also be extended to the setting of the simultaneous-move game.
Our study has mainly focused on a simple monopoly setting. With oligopolistic market competition, additional issues could
arise, however. For instance, with oligopolistic competition in the final-good market of the FDI source country, the internal trans-
fer price has additional strategic effects that further strengthen the incentive to inflate the transfer price at the expense of rivals'
profits. Tax-motivated FDI by the MNE has spillover effects that reduce tax revenue from other final-good producers as well as the
MNE. Moreover, with the presence of competitors that use similar inputs, the CUP method may be adopted as an application of
the ALP. In such a case, we can also uncover a novel mechanism for input foreclosure when the input market is also imperfectly
competitive. The MNE may have an incentive for input foreclosure even if it is a more efficient input producer. The new mecha-
nism stems from the dependence of the transfer price on the market price of a “comparable” input, which is endogenously deter-
mined. Some of these issues are analyzed by Choi et al. (2018). In addition, with oligopolistic competition, each firm's FDI decision
may depend on other firms' FDI decisions. These issues represent potential areas for future research.

Appendix A. Proof of Lemma 4

To show that Country 1's best response is given by (20), we derive Country 1's social welfare when it sources and hosts FDI,
respectively.
When County 1 sources FDI, it chooses t1 so that the price equals γ regardless of t2 as we have seen. Country 1's social welfare
s
in this case consists only of the total surplus in Industry 1. Thus, it follows from q ¼ L1 ð1−γÞ and ξ ¼ c þ ðγ−cÞt 2 that its social
welfare as a source country is given by

s L1 2  s
W1 ¼ ð1−γÞ þ γ−ξ L1 ð1−γÞ
2
L1 2
¼ ð1−γ Þ þ L1 ðγ−cÞð1−γÞð1−t 2 Þ;
2

for any t2. It is readily verified that Ws1 is decreasing in γ, meaning that Ws1 takes the largest value of L1(1 − c)2/2 when γ ¼ c.
Country 1 compares this welfare Ws1 with that when it selects a lower tax rate than Country 2 and hence hosts a subsidiary of
Firm 2. Social welfare as an FDI host country consists of the total surplus in Industry 1 and that in Industry 2, and it thus takes a
different form depending on the level of t2 as (19) suggests.
When t 2 ≤ t, the best response of Country 1 is to slightly undercut t2, or t1 is set equal to t2 while Country 1 hosts FDI. In this
case, Firm 2's perceived marginal cost becomes ξ2 = c and hence q∗2 = L2(1 − c)/2, meaning that we have

h 3L1 ð1−ωÞ2 L ð1−cÞ

W1 ¼ þ t 2 ðγ−cÞ 2 :
8 2

Consequently, we have Ws1 ≥ Wh1 if and only if

" #
2 2
L1 ð1−γ Þ 3ð1−ωÞ t ðγ−cÞð1−cÞ
þ ðγ−cÞð1−γÞð1−t 2 Þ− ≥ 2 : ð23Þ
L2 2 8 2
It is readily verified that the left-hand side of this inequality is decreasing in γ. It follows also from the assumption expressed by
(14) that as γ decreases to c, the left-hand side of this inequality approaches to a positive number, while the right-hand side con-
verges to 0. Thus, this inequality is satisfied for any t2 if γ is sufficiently small. If γ is not that small, there exists a ~t 2 such that
Ws1 ≥ Wh1 if and only if t 2 ≤ ~t 2 . It is easy to see that the threshold ~t 2 decreases with γ and increases with L1/L2.
We obtain a similar result when t < t 2 < t. In this case, we have t 1 ¼ ½1−c þ ð1−γÞð1−t 2 Þ=½2ðγ−cÞ as indicated in (19).
Then, we have

ð1−t 1 Þc−ðt 2 −t 1 Þγ γ−c−ð1−γÞð1−t 2 Þ

ξ2 ¼ ¼ γ− ;
1−t 2 2ð1−t 2 Þ
and hence

 L2 ð1−ξ2 Þ L2 ½γ−c þ ð1−γ Þð1−t 2 Þ

q2 ¼ ¼ :
2 4ð1−t 2 Þ
 J.P. Choi et al. / Journal of International Economics 127 (2020) 103367 17

Country 1 obtains a tax revenue of

2
 L2 ½γ−c þ ð1−γÞð1−t 2 Þ
t 1 ðγ−cÞq2 ¼ ;
8ð1−t 2 Þ

meaning that social welfare as an FDI host country is given by

h 3L1 ð1−ωÞ2 L2 ½γ−c þ ð1−γ Þð1−t 2 Þ2

W1 ¼ þ :
8 8ð1−t 2 Þ

Thus, we have Ws1 ≥ Wh1 if and only if

" #
L1 ð1−γ Þ2 3ð1−ωÞ2 L ½γ−c þ ð1−γ Þð1−t 2 Þ2
þ ðγ−cÞð1−γÞð1−t 2 Þ− ≥ 2 : ð24Þ
L2 2 8 8ð1−t 2 Þ

As γ falls to c, the right-hand side of this inequality decreases and converges to L2(1 − c)2(1 − t2)/8. However, we have

2 2
L2 ð1−cÞ ð1−t 2 Þ L2 ð1−cÞ ðγ−cÞ
<
8 8½2ðγ−cÞ þ ð1−γ Þ

under t 2 > t, while the right-hand side of this inequality converges to 0 as γ falls to c. This implies that Ws1 > Wh1 for any t2 if γ is
small enough. In addition, the right-hand side of (24) is increasing in t2 if and only if 1−t 2 < ðγ−cÞ=ð1−γÞ, which is true when
t 2 > t since 1−t 2 < ðγ−cÞ=½2ðγ−cÞ þ ð1−γÞ < ðγ−cÞ=ð1−γÞ, and increasing in γ. This implies that when the threshold ~t 2 falls in
this range, ~t 2 falls with γ and rises with L1/L2.
We also obtain a similar result in the final case where t 2 ≥t. In this case, we have t 1 ¼ ½γ−c−ð1−γÞð1−t 2 Þ=ðγ−cÞ and q2 ¼ L2
ð1−γÞ. Thus, we have

2
h 3L1 ð1−ωÞ
W1 ¼ þ t 1 ðγ−cÞL2 ð1−γ Þ
8
2
3L1 ð1−ωÞ
¼ þ L2 ð1−γ Þ½γ−c−ð1−γÞð1−t 2 Þ:
8

We have Ws1 ≥ Wh1 if and only if

" #
L1 ð1−γ Þ2 3ð1−ωÞ2
þ ðγ−cÞð1−γÞð1−t 2 Þ− ≥ ð1−γÞ½γ−c−ð1−γ Þð1−t 2 Þ:
L2 2 8

It is readily verified that this inequality is satisfied if γ is sufficiently small, and that if ~t 2 falls in this region it decreases with γ and
increases with L1/L2.
Before turning to Country 2's best response, we show the relationship between γ and ~t 2 more rigorously. We restrict γ to the
range (c, (1 + c)/2), that is, the transfer price lies between the marginal cost and monopoly price. As we have shown, if γ is suf-
ficiently close to c, we have Ws1 > Wh1 under the assumption (14), meaning that ~t 2 ¼ 1. As γ increases, Ws1 decreases while Wh1
increases. At a certain level of γ, these two payoffs become equal, and ~t 2 starts to decrease as γ further increases. Tentatively as-
suming that ~t 2 becomes smaller than t when γ reaches its upper bound, the equation Ws1 = Wh1, which is characterized by (23)
with equality when t 2 ¼ ~t 2 and γ ¼ ð1 þ cÞ=2, can be written as
" #
L1 3ð1−cÞ2 3ð1−ωÞ2 ~t 2 ð1−cÞ2 ~t ð1−cÞ2
− − ¼ 2 :
L2 8 8 4 8

If c < ω, the ~t 2 that satisfies this equation is positive and thus the threshold ~t 2 is positive even at the limit. This is because FDI
itself is beneficial for the source country, because of the gains in production efficiency. We may rather focus on the more inter-
esting case where c ≥ ω, while satisfying the assumption (14). In that case, ~t 2 reaches 0 before γ reaches its upper bound: the
assumption that ~t 2 < t is thus satisfied in this case.
Country 2's best response is similarly derived. Recall that Bs1 and Bh1 do not depend on L1/L2, as (18) and (19) show, and that L1/
L2 only affects Country 1's best response function B1 through ~t 2 , as (20) shows. Country 2's best response function is therefore
symmetric to Country 1's, including the property that the threshold ~t 1 decreases with γ. The only substantial difference is that
the threshold ~t 1 decreases with L1/L2, whereas ~t 2 increases.
18 J.P. Choi et al. / Journal of International Economics 127 (2020) 103367

References

Allingham, M.G., Sandmo, A., 1972. Income tax evasion: a theoretical analysis. J. Public Econ. 323–338.
Baldwin R., Krugman, P., 2004, Agglomeration,integration and tax harmonisation, Eur. Econ. Rev. 48, 1–23.
Becker, J., Fuest, C., 2012. Transfer pricing policy and the intensity of tax rate competition. Econ. Lett. 117, 146–148.
Bernard, A.B., Jensen, J.B., Redding, S.J., Schott, P.K., 2010. Intrafirm trade and product contractibility. Am. Econ. Rev. 100, 444–448.
Bucovetsky, S., 1991. Asymmetric tax competition. J. Urban Econ. 30, 167–181.
Bucovetsky, S., Haufler, A., 2008. Tax competition when firms choose their organizational form: should tax loopholes for multinationals be closed? J. Int. Econ. 74,
188–201.
Choi, J.P., Furusawa, T., Ishikawa, J., 2018. Transfer Pricing and the arms's Length Principle under Imperfect Competition, HIAS Discussion Paper E-73.
Davies, R., Martin, J., Parenti, M., Toubal, F., 2018. Knocking on tax haven's door: multinational firms and transfer pricing. Rev. Econ. Stat. 100, 120–134.
Egger, P., Eggert, W., Winner, H., 2010. Saving taxes through foreign plant ownership. J. Int. Econ. 81, 99–108.
Elitzur, R., Mintz, J., 1996. Transfer pricing rules and corporate tax competition. J. Public Econ. 60, 401–422.
Haufler, A., Wooton, I., 1999. Country size and tax competition for foreign direct investment. J. Public Econ. 71, 121–139.
Hines, J., Rice, E., 1994. Fiscal paradise: foreign tax havens and American business. Q. J. Econ. 109, 149–181.
Horst, T., 1971. The theory of the multinational firm: optimal behavior under different tariff and tax rates. J. Polit. Econ. 79, 1059–1072.
Huizinga, H., Laeven, L., 2008. International profit shifting within multinationals: a multi-country perspective. J. Public Econ. 92, 1164–1182.
Itagaki, T., 1979. Theory of the multinational firm: an analysis of effect of government policies. Int. Econ. Rev. 10, 437–448.
Johannesen, N., 2010. Imperfect tax competition for profits, asymmetric equilibrium and beneficial tax havens. J. Int. Econ. 81, 253–264.
Kant, C., 1988. Endogenous transfer pricing and the effects of uncertain regulation. J. Int. Econ. 24, 147–157.
Lee, H., 1998. Tax evasion, monopoly, and nonneutral profit taxes. Natl. Tax J. 51, 333–338.
Mansori, K.S., Weichenrieder, A.J., 2001. Tax competition and transfer pricing disputes. Finanz Arch. 58, 1–11.
Marceau, N., Mongrain, S., Wilson, J.D., 2010. Why do most countries set high tax rates on capital? J. Int. Econ. 80, 249–259.
Matsui, K., 2012. Auditing internal transfer prices in multinationals under monopolistic competition. Int. Tax Public Financ. 19, 800–818.
OECD, 2010. OECD transfer pricing guidelines for multinational enterprises and tax administrations 2010. OECD Publish. Paris https://doi.org/10.1787/tpg-2010-en.
Peralta, S., Wauthy, X., van Ypersele, T., 2006. Should countries control international profit shifting? J. Int. Econ. 68, 24–37.
Raimondos-Møller, P., Scharf, K., 2002. Transfer pricing rules and competing governments. Oxf. Econ. Pap. 54, 230–246.
Samuelson, L., 1982. The multinational firm with arm's length transfer price limits. J. Int. Econ. 13, 365–374.
Slemrod, J., Wilson, J.D., 2009. Tax competition with parasitic tax havens. J. Public Econ. 93, 1261–1270.
Stöwhase, S., 2005. Asymmetric capital tax competition with profit shifting. J. Econ. 85, 175–196.
Stöwhase, S., 2013. How profit shifting may increase the tax burden of multinationals: a simple model with discrete investment choices. J. Public Econ. Theory 15,
185–207.
Tørsløv, T.R., Wier, L.S., Zucman, G., 2020. The missing profits of nations. NBER Working Paper No. 24701.
Wilson, J.D., 1991. Tax competition with interregional differences in factor endowments. Reg. Sci. Urban Econ. 21, 423–451.