FIRMS IN A GLOBAL ECONOMY
By
Kyungmin Kim

A DISSERTATION
Submitted to
Michigan State University
in partial fulﬁllment of the requirements
for the degree of
Economics – Doctor of Philosophy
2013

ABSTRACT
FIRMS IN A GLOBAL ECONOMY
By
Kyungmin Kim
Chapter 1: Multiproduct Competition in the Global Economy
This paper presents a simple model of heterogeneous multiproduct ﬁrms to examine their
strategic decisions on the product scope in the global economy. I ﬁnd that there are several
diﬀerent types of equilibria depending on technological diﬀerence and relative cost advantage
of domestic and foreign ﬁrms. When each ﬁrm has local cost advantage and technological
diﬀerence between ﬁrms is suﬃciently large, ﬁrms reallocate resources toward their more
proﬁtable products. By contrast, when there exists a small technological diﬀerence or global
cost advantage, the more productive ﬁrm may expand the product line contrary to existing
core competence literature. This is because a high-productivity ﬁrm can increase its market
power by expanding its product range while its rival’s product line expansion is limited by
cost disadvantage.

Chapter 2: Production Sharing and Exchange Rate Pass-Through
This paper proposes a theoretical background for various possibilities of exchange passthrough and expenditure-switching by investigating the eﬀect of diﬀerent channels through
which exchange rate shocks aﬀect ﬁrms’ decisions on pricing and entry and exit. I extend the
model of exchange rate pass-through with endogenous markups built by Rodriguez-Lopez
(2011) introducing the intermediate input sector: domestic and imported intermediate inputs. I show that the degrees of exchange rate pass-through and expenditure switching
depend on the shares of imported inputs in total costs. According to the relative sizes of
these shares, exchange rate movements might lead to diﬀerent cost shocks to each trading
nation. When the imported input shares are located within some range, a low but positive

rate of pass-through to aggregate import prices can be derived in the model. In addition, low
levels of exchange rate pass-through to aggregate import prices can coexist with negligible
movements in trade ﬂows unlike Rodriguez-Lopez. The results of this paper also provide a
potential explanation for the fact that the degree of pass-through varies across countries and
industries.

Chapter 3: Exchange Rate Pass-Through in Korean Manufacturing Industries
This paper examines exchange rate pass-through into Korean export prices at the industry
level using disaggreated trade data. Unlike traditional approaches, I construct a testable
model in which both the intensive and extensive margins are operative. I ﬁnd that more
import-intensive industries in Korea have higher exchange rate pass-through into their export
prices. This is because aggregate export prices are aﬀected not only by changes in ﬁrms’
marginal costs, but also by variations in the composition of exporters due to changes in
the exporting cut-oﬀ. In addition, I show that the relative value of the destination market
currency should be also considered when estimating exchange rate pass-through regardless
of a high proportion of dollar invoicing of Korean exports and imports. Finally, I ﬁnd that
pass-through is increasing both in Korea’s share in total import of the destination market
and in Korea’s comparative advantage industries.

ACKNOWLEDGMENTS

Above all, I am deeply indebted to my advisor, Professor Susan Zhu. Without her kind
guidance and persistent help, this dissertation would not have been possible. She urged
me to challenge myself by choosing emerging ﬁelds of international trade as my dissertation
topics. She has always helped me embody my vague ideas and has graciously suggested ways
to improve.
I would like to thank my committee members. Professor Steven Matusz provided the
good opportunity to teach an online class as well as his valuable suggestions for this dissertation. He taught me the basics of classroom management, and thereby I was able to
gain precious experience. In particular, I was fully impressed with his persistent eﬀorts to
develop eﬀective teaching strategies. Professor John Wilson taught me the ﬁrst graduate
course in international trade. His excellent teaching led me to take interest in the trade
literature and this served to choose international trade as my major. His deep knowledge
and useful advice greatly improved my dissertation. I also thank Professor Charles Hadlock
in the ﬁnance department. He readily accepted to be the committee member and gave me
great encouragement to ﬁnish this dissertation.
I would also like to express the deepest appreciation to Professor Jay Pil Choi. He guided
me in writing my master’s thesis in Korea and was willing to write me a recommendation
letter for a graduate school application. Thanks to his kind help, I was able to have a chance
to study at MSU.
I owe great debt to my wife, Hyejin Lee, and my two kids, Hayoung and Chanyoung for
their love and support. They have been always my staunchest supporters and have helped
me concentrate on my studies. I am also deeply indebted to my parents for their love and
devotion.
Finally, I also thank the Bank of Korea for giving the excellent opportunity to complete
my doctoral degree and providing ﬁnancial support.

iv

TABLE OF CONTENTS

LIST OF TABLES

vi

LIST OF FIGURES

vii

CHAPTER 1 Multiproduct Competition in the Global Economy
1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.2 Basic Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.2.1 Preferences and Demand . . . . . . . . . . . . . . . . . . .
1.2.2 Production Technology . . . . . . . . . . . . . . . . . . . .
1.2.3 Optimal Scale and Scope . . . . . . . . . . . . . . . . . . .
1.3 Symmetric Oligopoly . . . . . . . . . . . . . . . . . . . . . . . . .
1.3.1 Closed Economy . . . . . . . . . . . . . . . . . . . . . . .
1.3.2 Global Economy . . . . . . . . . . . . . . . . . . . . . . .
1.4 Heterogeneous Duopoly in the Global Economy . . . . . . . . . .
1.4.1 Closed Economy . . . . . . . . . . . . . . . . . . . . . . .
1.4.2 Global Economy: Local Cost Advantage . . . . . . . . . .
1.4.3 Global Economy: Global Cost Advantage . . . . . . . . . .
1.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
APPENDIX . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
REFERENCES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

1
1
5
5
6
8
10
10
13
14
14
15
21
25
27
29

Pass-Through
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

31
31
36
36
37
38
41
42
44
45
46
46
51
53
61
63
69

CHAPTER 3 Exchange Rate Pass-Through in Korean Manufacturing Industries

72

CHAPTER 2 Production Sharing and Exchange Rate
2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . .
2.2 Basic Model . . . . . . . . . . . . . . . . . . . . . . .
2.2.1 Preferences and Demand . . . . . . . . . . . .
2.2.2 Production . . . . . . . . . . . . . . . . . . .
2.2.3 Firms’ Optimization . . . . . . . . . . . . . .
2.2.4 Trade in Final Goods . . . . . . . . . . . . . .
2.2.5 Cut-oﬀ Productivity Levels . . . . . . . . . .
2.2.6 Prices, Product Varieties and Welfare . . . . .
2.2.7 Entrants, Producers and Exporters . . . . . .
2.3 The Eﬀect of Exchange Rates in Partial Equilibrium
2.3.1 The Cut-oﬀ Levels and the Exchange Rate . .
2.3.2 Welfare . . . . . . . . . . . . . . . . . . . . .
2.3.3 Exchange Rate Pass-Through . . . . . . . . .
2.4 Conclusion . . . . . . . . . . . . . . . . . . . . . . . .
APPENDIX . . . . . . . . . . . . . . . . . . . . . . . . . .
REFERENCES . . . . . . . . . . . . . . . . . . . . . . . .

v

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

3.1
3.2

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . .
Background . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.2.1 Exchange Rate and Export Price Movements . . . . .
3.2.2 Imported Inputs into Production . . . . . . . . . . .
3.3 Theoretical Framework . . . . . . . . . . . . . . . . . . . . .
3.3.1 Demand and Production . . . . . . . . . . . . . . . .
3.3.2 Cut-oﬀ Levels . . . . . . . . . . . . . . . . . . . . . .
3.3.3 Aggregate Export Prices and Exchange Rates . . . .
3.3.4 An Alternative Approach Without Selection Channel
3.4 Empirical Evidence . . . . . . . . . . . . . . . . . . . . . . .
3.4.1 Data . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.4.2 Empirical Speciﬁcation . . . . . . . . . . . . . . . . .
3.4.3 Estimation Results . . . . . . . . . . . . . . . . . . .
3.4.4 Robustness . . . . . . . . . . . . . . . . . . . . . . .
3.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . .
APPENDIX . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
REFERENCES . . . . . . . . . . . . . . . . . . . . . . . . . . . .

vi

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

. 72
. 76
. 76
. 77
. 79
. 80
. 81
. 82
. 87
. 89
. 89
. 92
. 93
. 95
. 106
. 108
. 120

LIST OF TABLES

Table 3.1 Distribution of Imported Input Shares in 2010 . . . . . . . . . . . . . .

79

Table 3.2 Correlation Coeﬃcient Among Exchange Rates . . . . . . . . . . . . .

92

Table 3.3 Baseline Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

94

Table 3.4 Robustness: Additional Controls . . . . . . . . . . . . . . . . . . . . .

98

Table 3.5 Robustness: Non-parametric Speciﬁcation . . . . . . . . . . . . . . . .

99

Table 3.6 Robustness: Real Exchange Rates . . . . . . . . . . . . . . . . . . . . 102
Table 3.7 Robustness: Alternative Samples . . . . . . . . . . . . . . . . . . . . . 103
Table 3.8 Robustness: Alternative Measure of Prices . . . . . . . . . . . . . . . . 106
Table A1 US Dollar Use in the Export and Import Invoicing . . . . . . . . . . . 109
Table A2 Share of Imported Inputs by Manufacturing Industry . . . . . . . . . . 110
Table A3 Imported Input Shares by Country . . . . . . . . . . . . . . . . . . . . 115
Table A4 Top 50 Countries in World Merchandise Trade (2002-2012) . . . . . . . 116
Table A5 Share of Imported Inputs by Manufacturing Industry Classiﬁcation of
Export Price Indexes . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119

vii

LIST OF FIGURES

Figure 1.1 Core Competency and Flexible Manufacturing

. . . . . . . . . . . . .

7

Figure 1.2 Competition Eﬀect in a Symmetric Oligopoly Model . . . . . . . . . .

12

Figure 1.3 Segmented Monopoly (β > 1) . . . . . . . . . . . . . . . . . . . . . . .

16

Figure 1.4 Segmented Monopoly and Reduction in the Product Scopes . . . . . .

17

Figure 1.5 Partial Duopoly and Segmented Monopoly (β > 1) . . . . . . . . . . .

19

Figure 1.6 Pure Monopoly (0 < β < 1) . . . . . . . . . . . . . . . . . . . . . . . .

21

Figure 1.7 Partial Duopoly (0 < β < 1) . . . . . . . . . . . . . . . . . . . . . . .

23

Figure 2.1 The Impact of Exchange Rate Movements on the Cut-oﬀ Level (k =
3, ψ = 1, τ = τ ∗ = (1.5)1/3 ) . . . . . . . . . . . . . . . . . . . . . . . .

50

Figure 2.2 The Impact of Exchange Rate Movements on Welfare (k = 3, ψ = 1, τ =
τ ∗ = (1.5)1/3 ) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

52

Figure 2.3 Exchange Rate Pass-Through and Export Quantity at the Firm Level
(k = 3, ψ = 1, τ = τ ∗ = (1.5)1/3 ) . . . . . . . . . . . . . . . . . . . . .

55

Figure 2.4 Exchange Rate Pass-Through and Export Quantity at the Aggregate
Level (k = 3, ψ = 1, τ = τ ∗ = (1.5)1/3 ) . . . . . . . . . . . . . . . . . .

58

Figure 3.1 Exchange Rate and Export Price Index . . . . . . . . . . . . . . . . .

77

Figure 3.2 Won/US Dollar Rates and Nominal Eﬀective Exchange Rates of the US
Dollar . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

85

Figure 3.3 Import Intensity and Exchange Rate Pass-Through . . . . . . . . . . .

86

Figure 3.4 Exchange Rate Pass-Through Elasticity by Import Intensity (20 Bins)

100

viii

CHAPTER 1
Multiproduct Competition in the Global
Economy
1.1

Introduction

Multiproduct ﬁrms dominate international trade in most developed countries. According
to the study on US ﬁrms by Bernard et al. (2007), 57.8 percent of exporting ﬁrms produce
multiple products, and multiproduct ﬁrms account for more than 99.6 percent of export
value in the year 2000.1 Despite this dominant role in the global economy, multiproduct
ﬁrms have received comparatively little attention in the ﬁeld of international trade so far.
This is because traditional theoretical frameworks of international trade are based on single
product ﬁrms only. However, trade economists have recently begun to take more interest in
multiproduct ﬁrms and examine their activity.2
My paper is inspired by an interesting result of recent research on multiproduct ﬁrms.
Trade liberalization induces ﬁrms to reallocate resources towards their relatively high-proﬁt
(i.e., core-competency) products. For instance, Bernard et al. (2011) model product-speciﬁc
competencies as the strength of consumers’ tastes for ﬁrm variety. The opening of trade
intensiﬁes product-market competition and so induces surviving ﬁrms to drop products with
lower consumer tastes from the domestic market. Meanwhile, Eckel and Neary (2010) as1 In this study, a product is deﬁned at the ten-digit Harmonized System (HS) code level.

Also Bernard et al. (2010) used ﬁve-digit Standard Industry Classiﬁcation (SIC) code as a
measure of product. Following their deﬁnition, multiproduct ﬁrms account for 87 percent of
total output while they represent 39 percent of total ﬁrms in 1997.
2 Besides articles introduced in the text, refer Eckel and Iacovone et al. (2010), Feenstra
and Ma (2007), Mayer et al. (2009).

1

sume that ﬁrms typically own a core competence in the production of a particular variety.
Yet, multiproduct ﬁrms are less eﬃcient in the production of their varieties which are further away from the core product. They predict that globalization encourages multiproduct
ﬁrms to focus on their core competence because greater competition hits those varieties produced at higher costs harder. In addition, Iacovone and Javorcik (2008) document empirical
evidence for core competencies from the study of Mexican ﬁrms. They found the positive
correlation between the rank of export varieties (in terms of their export values) and the
rank of expansion of export varieties. Thus, exporters tend to expand their most important
export products.
In spite of recent theoretical progress about the product scope for multiproduct ﬁrms,
there remain some unexamined questions. First of all, multiproduct ﬁrms are generally big
ﬁrms which have considerably large shares within the industry.3 Thus, we need to examine
strategic behaviors of multiproduct ﬁrms explicitly given the industry’s market structure.
In addition, multiproduct ﬁrms are likely to be heterogeneous in production technology.
For example, multiproduct ﬁrms may have diﬀerent ﬁrm-level productivity or product-level
productivity. The diﬀerent roles of multiproduct ﬁrms among countries also reﬂect heterogeneity. Iacovone and Javorcik (2008) observe that multiproduct exporters are much less
prevalent in Mexico than in the US. Putting these facts together, both market structure
and technological diﬀerences may have important roles in explaining the behaviors of multiproduct ﬁrms. In particular, globalization can cause considerable eﬀects on multiproduct
competition by increasing both market size and market competition. However, this idea has
not been explored carefully in the existing literature. For instance, Bernard et al. (2011)
extend the heterogeneous ﬁrm model of Melitz (2003) to multiproduct ﬁrms. In their model,
market structure is an ex post equilibrium outcome of monopolistic competition with free
entry and ex ante uncertainty. Eckel and Neary (2010) introduce a diﬀerentiated product
oligopoly model and so they do not consider strategic behaviors such as predatory pricing.
3 Multiproduct ﬁrms are larger than single product ﬁrms in the same industry in terms of

both shipments and employment. See Bernard et al. (2010)
2

Furthermore, their analysis excludes the possibility that the degree of competition may vary
across products produced by a multiproduct ﬁrm.
In this paper, I develop a simple multiproduct competition model in the global economy.
The basic theoretical setup is similar to Eckel and Neary (2010). They introduce both
demand and supply linkages, which make a clear distinction between multiproduct ﬁrms and
single-product ﬁrms. First, on the demand side, there exists the “cannibalization eﬀect”
when multiproduct ﬁrms produce diﬀerentiated products. Because a larger output of one
variety tends to crowd out demand for all other varieties, a multiproduct ﬁrm needs to
restrict its output of each variety. Second, on the cost side, ﬂexible manufacturing allows
ﬁrms to expand their product lines, but this expansion is limited due to diseconomies of
scope. Marginal production cost increases when ﬁrms produce varieties further from their
core competence.
I follow this framework basically, but there are some noticeable diﬀerences in my paper.
First, I formally consider the framework of a homogeneous product oligopoly model in which
ﬁrms compete directly in the same product market. When products are homogeneous, the
only source of market power is lack of competition.4 Hence, multiproduct ﬁrms have more
incentives to adjust their product scope strategically in response to changes in the market
conditions. Next, I deﬁne the diﬀerence between ﬁrms’ core products (or the gap between
ﬁrms’ feasible product lines) as technological diﬀerence. In general, ﬁrms can have their
idiosyncratic advantages in producing a speciﬁc good by the exclusive patent right or cumulative experience in production process. For example, some automakers have comparative
advantage in producing a compact car while others can produce a sports car or SUV more
eﬀectively. Finally, I assume that ﬂexible manufacturing is imperfect because the physically
feasible production lines may be diﬀerent among multiproduct ﬁrms. Sometimes it may be
impossible to enter the product market without original technology. For instance, every PC
maker cannot produce a tablet computer like the iPad and sell it at competitive prices.
4 In contrast, both lack of competition and product diﬀerentiation can create market power

under a diﬀerentiated product oligopoly model.
3

The model yields several interesting predictions about the product scope for multiproduct ﬁrms in the global economy. Relative cost advantage as well as technological diﬀerence
are main factors that aﬀect the product line selection. There are several diﬀerent types of
equilibria as a result of strategic behaviors of heterogeneous multiproduct ﬁrms. When each
ﬁrm has local cost advantage and technological diﬀerence between ﬁrms is suﬃciently large,
ﬁrms concentrate more on their core-competency products. This is similar to a traditional
Ricardian model from the standpoint of specialization due to a gap between production
technologies. Although each ﬁrm has its own segmented monopolized markets, intensiﬁed
competition among diﬀerentiated goods induces ﬁrms to focus on their more proﬁtable products. By contrast, when there exists a small technological diﬀerence or global cost advantage,
the more productive ﬁrm may expand the product line while the less productive ﬁrm reduces
its product scope in the global economy. This is because a high-productivity ﬁrm can create its monopolistic power in some varieties by expanding its product line. This result is
a new ﬁnding contrary to the existing literature.5 In reality, we can observe the product
line changes occur with frequency. Consider, for example, the history of the automobile industry. Japanese and Korean automakers began to export compact cars in the initial phase
and have gradually extended their product lines to mid-size, full-size and luxury class cars.
In contrast, General Motors reduced its product mix by dropping some unproﬁtable brands
such as Pontiac, Saturn and SAAB after its 2009 bankruptcy reorganization. Although a
dynamic model is not presented, this paper can give us some insight into adjustments in the
intra-ﬁrm extensive margin.
In conclusion, this paper suggests that we should consider various conditions in analyzing
the optimal choice of product scope for multiproduct ﬁrms. In particular, ﬁrms are likely to
have strategic incentives to adjust the number of products produced within a more concentrated industry. Thus, an empirical study on multiproduct ﬁrms needs to investigate both
5 Bernard et al. (2011) also mention that trade liberalization might have an ambiguous

eﬀect on the product scope if they allow demand heterogeneity across countries. Exporters
can add products with high consumer tastes in the foreign market.

4

competition intensity and ﬁrm heterogeneity.
The remainder of the paper is structured as follows. Section 1.2 introduces main assumptions about consumers and ﬁrms and examines an optimal choice of scale and scope
by multiproduct ﬁrms. Section 1.3 illustrates a symmetric oligopoly model to emphasize the
diﬀerence with our heterogeneous oligopoly model. In a symmetric oligopoly model, neither
a rise in market size nor an increase in the number of ﬁrms aﬀects the product line selection
of multiproduct ﬁrms. Section 1.4 analyzes a heterogeneous oligopoly model in the global
economy and shows the paper’s key results. Section 1.5 concludes the paper.

1.2

Basic Model

We introduce the behavior of consumers and multiproduct ﬁrms in a single industry. We
begin with a closed economy where L consumers exist.

1.2.1

Preferences and Demand

Following the speciﬁcation of Melitz and Ottaviano (2008) as well as Eckel and Neary
(2010), preferences are deﬁned over a continuum of diﬀerentiated products indexed by i ∈ Ω,
and a homogeneous numeraire good. All consumers share the same utility function given by

U = q0 + α

1
1
q(i)2 di − η
q(i)di − γ
2 i∈Ω
2
i∈Ω

2

q(i)di

(1.1)

i∈Ω

where q0 and q(i) denote the individual consumption levels of the numeraire good and
each variety i. The parameters α, γ and η are all positive. The parameters α and η
represent the substitution pattern between the diﬀerentiated varieties and the numeraire:
increases in α and decreases in η both raise the demand for the diﬀerentiated varieties
relative to the numeraire. The parameter γ is a measure of product diﬀerentiation between
the varieties. In the limit, when γ = 0, the goods are perfect substitutes. Since γ is positive,
diﬀerentiated products are imperfect substitutes in demand. That is, a multiproduct ﬁrm
5

produces multiple goods which are horizontally diﬀerentiated. For example, ﬁrms produce
various types of cars in the automobile industry. The degree of product diﬀerentiation
increases with γ as consumers prefer more balanced consumption across varieties.
Let Ω∗ ⊂ Ω be the subset of varieties that consumed (q(i) > 0). We assume that
consumers have positive demands for the numeraire good. Then, the inverse demand for
each variety i is given by

p(i) = α − γq(i) − ηQ, where Q ≡

q(i)di, i ∈ Ω∗

(1.2)

The market demand for a particular good i, x(i), is equal to Lq(i). This allows equation
(1.2) to be rewritten by

p(i) = α −

γ
η
x(i) − X where X ≡
L
L

x(i)di, i ∈ Ω∗

(1.3)

where X denotes the output of the entire industry.6

1.2.2

Production Technology

The technology of multiproduct ﬁrms is summarized by a core competence and ﬂexible
manufacturing. This is illustrated in Figure 1.1, where cj (i) indicates the marginal cost which
a ﬁrm j incurs to produce good i. We set a ﬁrm j’s core competence at cj with cj (cj ) = cj .7
This assumption implies two important facts. First, each ﬁrm has its own independent core
product. As a ﬁrm’s core competency product is closer to the origin, a ﬁrm can produce
6 We assume that individual outputs of diﬀerentiated goods are measured in the same

units. Thus, we can calculate the ﬁrm’s total output or industry output easily by adding up
the output of each variety. However, in general, measuring aggregate output of a multiproduct ﬁrm is more diﬃcult because of the need for diﬀerential valuation of the outputs. See
Bradley and Baron (1993) for details.
7 Eckel and Neary (2010) arrange a ﬁrm’s core competence at i = 0 with c (0) = c0 . In
j
j
their model, there is no overlap in varieties produced among ﬁrms. Therefore, we should
interpret that the origin denotes a representative ﬁrm’s core competence, not a speciﬁc ﬁrm’s
core competence.

6

¡
¡
¡ ¡
¡
cj (i)
T
¡ ¡ ¡  
 
¡ ¡ ¡
 
¡ ¡ ¡
¡ ¡ ¡  
¡ ¡ ¡  
¡ ¡ r 
¡
¡ ¡  
¡ ¡ 
¡ r 
¡
¡  
¡ 
¡
r 
 
 ◦
 45
Ei
c1
c2
c3
(a)

cj (i)
T

 
 
 
  ¨
¨
  ¨
¨¨ ¨
¨
¨
r 
  ¨¨
¨
 
¨ ¨
¨
¨¨
r  ¨¨
  ¨
 ¨¨
¨
r ¨
 
 ◦
 45
Ei
c1
c2
c3

β>1

(b)

0<β<1

Figure 1.1: Core Competency and Flexible Manufacturing

the more proﬁtable variety. Second, multiproduct ﬁrms have diﬀerent feasible product lines
because the marginal cost is lowest for core competency variety by deﬁnition. A ﬁrm can
expand its product line only towards the right-hand side from its core product. In other
words, ﬂexible manufacturing is imperfect.
Multiproduct ﬁrms must pay higher marginal production costs to produce additional
varieties, but the marginal costs of existing products remain unchanged. In this paper, we
assume that cj (i) is a linearly increasing function. Formally, the cost function is expressed
by


 c + β(i − c )
j
j
cj (i) =
 ∞


if i ≥ cj

(1.4)

otherwise

The slope β of the cost function represents the degree of ﬂexibility in expanding product
scope. The higher β implies that higher marginal costs incur in adding new products. We
need to pay regard to two interesting cases. When β > 1, each ﬁrm has local cost advantage
in producing some bounded varieties. When 0 < β < 1, the most productive ﬁrm has global
cost advantage in all varieties.

7

1.2.3

Optimal Scale and Scope

Proﬁts for a multiproduct ﬁrm j are given by

Πj =

cj +Nj
cj

pj (i) − cj (i) xj (i)di − F

(1.5)

where Nj is the scope of production of the ﬁrm j and the ﬁxed cost F is independent of
both scale and scope. So cj + Nj denotes the marginal variety which a ﬁrm j produces.
We assume that ﬁrms play a one-stage Cournot game. The ﬁrst-order condition with
respect to the scale of production of a particular good i is given by
∂Πj
γ
η
= pj (i) − cj (i) − xj (i) − Xj = 0
∂xj (i)
L
L

where Xj ≡

cj +Nj
cj

(1.6)

xj (i)di denotes the ﬁrm’s aggregate output.8

Let M (i) = {1, · · · , mi } be the set of ﬁrms which produce product i. Notice that pj (i) =
p(i) for all j ∈ M (i) because ﬁrms belonging to M (i) produce homogeneous good i.9 From
equations (1.3) and (1.6), we obtain the following equation.

α−

η
γ
η
γ
x(i) − X = cj (i) − xj (i) − Xj
L
L
L
L

(1.7)

2
8 The second-order condition can be easily shown to hold: ∂ Πj = ∂pj (i) − γ −
2

∂xj (i)
L
∂xj (i)
∂Xj
η
< 0.
L ∂xj (i)
9 Baldwin and Ottaviano (2001) categorize multiproduct competition into the following
three types. The ﬁrst, ‘full symmetry’, refers to the case where all varieties as equally good
substitutes. In other words, this type is the standard Dixit-Stiglitz model of monopolistic
competition. Eckel and Neary (2010) follow this framework. The second is, ‘ﬁrm-wise
symmetry (or market segmentation)’, where a ﬁrm’s own varieties are perceived by consumers
to be closer substitutes to each other than to those of other ﬁrms. The third is, ‘matching
product lines (or market interlacing)’, where a ﬁrm’s own varieties are less good substitutes
for each other than they are for the other ﬁrm’s varieties. The approach described in this
paper falls into the third category.

8

Summing both sides over j ∈ M (i) gives the total output of a single variety.

x(i) =

xj (i) =
j∈M (i)



1 
η
L
·
mi α − X −
γ mi + 1
L

cj (i) −
j∈M (i)

η
L

Xj 

(1.8)

j∈M (i)

Substituting (1.8) into (1.7), we can obtain a ﬁrm j’s output with respect to a single
variety.

xj (i) =



1
η
1
L
α− X +
ck (i) − cj (i)
γ mi + 1
L
mi + 1
k∈M (i)


1
η
Xk − Xj 
+ 
L mi + 1

(1.9)

k∈M (i)

The ﬁrst term shows that industry output has a negative eﬀect on the single product
output. Next, we deﬁne the second term as the relative cost advantage in producing variety
i. As the marginal costs of ﬁrm j are smaller than those of rival ﬁrms in producing the same
variety, its output increases. Finally, we deﬁne the last term as the relative cannibalization
eﬀect. A ﬁrm j’s aggregate output Xj has a more negative eﬀect on the output of a variety
when total ﬁrm output is relatively bigger than those of its rivals. That is, a ﬁrm with higher
market share is hurt more from the cannibalization eﬀect.
Next, consider the ﬁrm’s choice of product line. The ﬁrst-order condition with respect
to the scope of production is
∂Πj
= pj (i) − cj (i) xj (i)
=0
∂Nj
i=cj +Nj

(1.10)

From the FOC for scale, equation (1.6), the proﬁt margin pj (i) − cj (i)
cannot
i=cj +Nj
be zero. Therefore, multiproduct ﬁrms choose their product ranges so that the output of the
marginal variety is zero.

9

That is, xj (i)
= 0.10
i=cj +Nj

1.3
1.3.1

Symmetric Oligopoly
Closed Economy

We consider a symmetric Cournot oligopoly model as a benchmark in order to stress
the diﬀerence with our main heterogeneous model. We assume that there is an exogenously
given number of multiproduct ﬁrms m in a closed economy. Let xo (i) and Xo denote the
output of each ﬁrm in variety i and the total output of each ﬁrm, respectively. Then the
industry output X is equal to mXo . From equation (1.9), output per ﬁrm in variety i is
given by
xo (i) =

L
(m + 1)η
Xo
α − co (i) −
(m + 1)γ
L

(1.11)

In addition, from the equation of the optimal scope xo (co + No ) = 0,

co (co + No ) = α −

(m + 1)η
Xo
L

(1.12)

Thus,
xo (i) =

L
L
[co (co + No ) − co (i)] =
[β(co + No − i)]
(m + 1)γ
(m + 1)γ

(1.13)

Given co , multiproduct ﬁrms produce more of each variety the closer it is to its core
competence. Also, given the demand structure, proﬁt margins are lower for products that
10 We can show that ∂xj (i)

∂Nj

i=cj +Nj

< 0 in the equilibrium illustrated in this paper

because the marginal cost is increasing in i. Thus, second-order condition is veriﬁed:
pj (i) − cj (i)

∂xj (i)
< 0.
∂Nj i=cj +Nj

10

∂ 2 Πj
2
∂Nj

=

are further from ﬁrms’ core competence.

po (i) − co (i) =

1
[α − co (i)]
m+1

(1.14)

Now, let’s calculate the total output of each identical multiproduct ﬁrm.
co +No

Xo =

xo (i)di =
co

βL
N2
2(m + 1)γ o

(1.15)

Substituting (1.15) into (1.12), we can get the optimal scope of production.
2
βηNo + 2βγNo − 2γ(α − co ) = 0 ⇒ No =

−βγ +

(βγ)2 + 2βγη(α − co )
βη

(1.16)

Note that there exists neither L nor m in (1.16). In other words, the optimal scope
of production is independent of both market size and the number of ﬁrms in a symmetric
oligopoly equilibrium.11 This is because the change in L or m causes an equi-proportionate
change in both output per ﬁrm in variety i (xo ) and total ﬁrm output (Xo ).
Let’s examine this prediction in more detail. First, equation (1.12) gives one negative
relationship between the optimal output of each ﬁrm (Xo ) and the optimal choice of product
scope (No ). This is illustrated by the downward-sloping straight line in Figure 1.2. This
comes from the cannibalization eﬀect. A ﬁrm desires to produce less variety as its total
output increases. On the other hand, equation (1.15) shows another relationship between
Xo and No . A rise in No raises total output. The curve passing through the origin in Figure
1.2 represents this relationship. For example, suppose that m increases. Then, both loci
shift inward by the same proportion. As a result, the optimal choice of product range does
not change while the output of each ﬁrm reduces.
I now compare my results with those of Eckel and Neary (2010). The market-size eﬀect
11 This result does not depend on our linear cost function assumption. For example, if the

cost function is replaced by co (i) = co + β (i − co )2 , the optimal product scope is obtained
3
2
by solving the following cubic equation: βηNo + βγNo − γ(α − co ) = 0.

11

No
T
d
e
ed
e d
e d
e d
d
e
r
r
∗
d
e
No
d
e
d
e
d
e
d
e
d
e

←−
∗∗
Xo

∗
Xo

eq (1.15)

eq (1.12)
E Xo

Figure 1.2: Competition Eﬀect in a Symmetric Oligopoly Model

(an increase in L) is the same as that demonstrated by their model. Given a ﬁxed number
of ﬁrms, market expansion does not aﬀect ﬁrm scope in both models. On the contrary, in
the Eckel and Neary (2010) model, the competition eﬀect (an increase in m) causes a fall
in ﬁrm scope. This disparity results from diﬀerent assumptions about the degree of product
diﬀerentiation in oligopoly. In my current setup, all ﬁrms compete in the same product
market with each ﬁrm sharing an identical core competence product. Thus, an increase in
the number of ﬁrms causes an equi-proportionate increase in competition among all varieties
already produced. As a result, ﬁrms have no incentive to adjust their product scope. By
contrast, each ﬁrm produces a mutually exclusive set of products in Eckel and Neary (2010).
As a result, intensiﬁed competition hits marginal varieties harder and encourages incumbent
ﬁrms to prune their product lines.
Proposition 1 In a symmetric oligopoly model under a closed economy, the market size
eﬀect of an increase in L is an equi-proportionate increase in the output of each variety and
total ﬁrm output, but there is no change in ﬁrm scope. The competition eﬀect of an increase
in m is an equi-proportionate fall in both the output of each variety and total ﬁrm output,
but there is a rise in industry output and no change in ﬁrm scope.
Proof. All these results are clear from equations (1.13), (1.15) and (1.16). Because X =
12

m
is increasing in m, the total industry output rises in response to an increase
m+1
in the number of ﬁrms like a single product Cournot model.
mXo and

Also note the following intuitively consistent results12 :
∂No
< 0,
∂β

∂No
> 0,
∂α

∂No
> 0,
∂γ

∂No
< 0,
∂η

∂No
<0
∂co

∂Xo
< 0,
∂β

∂Xo
> 0,
∂α

∂Xo
< 0,
∂γ

∂Xo
< 0,
∂η

∂Xo
<0
∂co

Because higher β indicates less ﬂexibility in expanding product lines, multiproduct ﬁrms
will shrink their product ranges. Increases in α and decreases in η imply that consumers
much prefer the diﬀerentiated varieties to the numeraire good. Thus, more preference for the
diﬀerentiated goods raises both the number of varieties and the output of each variety. Since
γ indexes the degree of product diﬀerentiation, multiproduct ﬁrms produce more varieties
the higher γ is. Finally, multiproduct ﬁrms have narrower feasible product lines when co is
greater.

1.3.2

Global Economy

We assume that there are L consumers with identical preferences located in each of k ≥ 2
countries. In addition, we assume that the good markets of all countries are completely
integrated in a single world market as a result of free trade, so the price of a given product
is the same in every place.13
Because globalization leads to concurrent increases in both market size and the number
of ﬁrms, L = kL and m = km.14 We already know that the optimal scope of production
does not depend on L and m , so No = No .
12 See Appendix for proofs.
13 In this model, when consumers buy a speciﬁc good, they are indiﬀerent about which

ﬁrm produces it.
14 I will use prime superscript

to denote variables in the global economy.

13

Also using

k
1
1
=
>
, we obtain the following results.
km + 1
m + (1/k)
m+1

Xo =

xo (i) =

2
βL
βL
k
No =
No 2 > Xo
2(m + 1)γ
2γ (km + 1)

L
βη 2
L
k
βη 2
α − co (i) −
No =
α − co (i) −
No > xo (i)
(m + 1)γ
2γ
γ (km + 1)
2γ

We summarize these results in the following proposition.
Proposition 2 In a symmetric oligopoly model, globalization causes rises in both output of
each variety and total ﬁrm output, but there is no change in ﬁrm scope.

1.4
1.4.1

Heterogeneous Duopoly in the Global Economy
Closed Economy

In general, it is diﬃcult to analyze the equilibrium in a heterogeneous oligopoly model
because the market structure may be too complex. Instead, we consider the simplest case
as a starting point. Before globalization, there are two countries with identical consumers.
In each country, only one multiproduct ﬁrm monopolizes the industry.15 Because most
multiproduct ﬁrms are large ﬁrms, our model is not too unrealistic. In addition, despite its
simplicity, we can obtain some meaningful intuition.
From equations (1.11) and (1.15),

xj (i) =

Xj =

L
2η
α − cj (i) − Xj ,
2γ
L
cj +Nj
cj

xj (i)di =

βL 2
N ,
4γ j

j = d, f

(1.17)

j = d, f

(1.18)

where d and j denote a domestic ﬁrm and a foreign ﬁrm, respectively.
15 This is the special case that m = 1 in a symmetric model of section 1.3.

14

Also from equation (1.16), we obtain the following optimal scope of production.

2
βηNj +2βγNj −2γ(α−cj )

−βγ +
= 0 ⇒ Nj =

(βγ)2 + 2βγη(α − cj )
βη

,

j = d, f (1.19)

In the global economy, a domestic and a foreign ﬁrm compete with each other in the
fully integrated market. Thus, the market size increases from L to L = 2L. Without loss
of generality, suppose that cd < cf .
Before the analysis, we explain the equilibrium concept more clearly. A ﬁrm has its
optimal output schedules of each variety through Cournot competition given a conﬁguration
of product ranges (Nd , Nf ). These output schedules {xj (i)}j=d,f are given by equation
(1.9). Therefore, each ﬁrm selects the optimal product range to maximize its proﬁt given
the rival’s product scope and output schedules. Then, we can derive multiproduct CournotNash equilibria.16 In this paper, we sketch out some possible equilibria instead of depicting
all equilibria formally. For convenience, we analyze two diﬀerent cases (1) β > 1, and (2)
0 < β < 1 separately.

1.4.2

Global Economy: Local Cost Advantage

When β > 1, each ﬁrm has local cost advantage as depicted in Figure 1.1 (a). Above all,
we can think of the situation where the diﬀerence between cd and cf is very large. Then,
the domestic ﬁrm will not compete with the product line which the foreign ﬁrm produces
because of the prohibitively high production costs as well as the cannibalization eﬀect. This
situation is illustrated in Figure 1.3.
Let’s consider the suﬃcient condition for the segmented monopoly. Let pD (cf ) denote
the price of variety cf when both ﬁrms produce this product. If pD (cf ) − cd (cf ) < 0, the
domestic ﬁrm will not enter the product line that the foreign ﬁrm produces.
16 See Grossman (2007) for details.

15

cj (i)
T

¡
¡
¡
¡
¡
¡
¡
¡
r
¡

¡
¡
¡
¡

 
¡
 
r¡
 
¡
 
¡  
¡  
¡ 
¡
 
¡
r
 

 
 

¡
¡
¡

 
 
¡
¡  
¡  
¡ 
¡
r 
 
 ◦
 45
cd
cd + Nd

Ei

cf cf + Nf

Figure 1.3: Segmented Monopoly (β > 1)

pD (cf ) − cd (cf ) =

α−

γ
L

xd (cf ) + xf (cf ) −

η
L

1
α + cf (cf ) − 2cd (cf )
3
1
=
α − cf + 2(1 − β) cf − cd
3
α − cf
⇒ cf − cd >
2(β − 1)

(Xd + Xf ) − cd (cf )

=

<0
(1.20)

We deﬁne the diﬀerence between cd and cf as technological diﬀerence. This represents the
diﬀerence between ﬁrms’ core products and also denotes the gap between ﬁrms’ physically
feasible product ranges. The following proposition shows us the case of segmented monopoly
in the global economy.
Proposition 3 When each multiproduct ﬁrm has local cost advantage (β > 1) and technoα − cf
logical diﬀerence between ﬁrms is suﬃciently large (cf − cd >
), each ﬁrm has its
2(β − 1)
own segmented monopolized markets in the global economy. While the product scopes of both
ﬁrms decrease, total industry output increases.
16

Proof. First, let’s show that multiproduct ﬁrms shrink their product lines. The output of
each ﬁrm in variety i is given by

xd (i) =

L
2η
η
α − cd (i) − Xd − Xf
2γ
L
L

(1.21)

xf (i) =

2η
η
L
α − cf (i) − Xf − Xd
2γ
L
L

(1.22)

By using equations (1.21), (1.22) and the optimal scope equations, we can get the following
results. (See Section 1.3.1)

Xd =

2
βL
Nd ,
4γ

Xf =

2
βL
Nf
4γ

2
2
1
βηNd + 2βγNd − 2γ(α − cd ) + βηNf = 0
2
2
2
1
βηNf + 2βγNf − 2γ(α − cf ) + βηNd = 0
2

(1.23)

(1.24)

Although we cannot obtain nice analytical solutions, it is easy to show that ﬁrm scopes
decrease through globalization by comparing (1.24) with (1.19). Notice the last terms of
the simultaneous equations (1.24) are positive and the other terms are equal to those of the
equation (1.19) in a closed economy. This is illustrated in Figure 1.4. Globalization shifts
the optimal choices of product scope from A to B.

Nf T
2
2
1
→ βηNd + 2βγNd − 2γ(α − cd ) + βηNf = 0
2
rA

↓
B

r

←

2
2
1
βηNf + 2βγNf − 2γ(α − cf ) + βηNd = 0
2
E N
d

Figure 1.4: Segmented Monopoly and Reduction in the Product Scopes

17

Next, we can easily show that globalization leads to an increase in total industry output.
From the optimal scope equations of closed economy and open economy,
L
2α − (cd + cf ) − β(Nd + Nf )
2η
2L
Xd + Xf =
2α − (cd + cf ) − β(Nd + Nf )
3η
X d + Xf =

(1.25)

Since Nd > Nd and Nf > Nf , the total industry output increases from globalization.

The segmented monopoly is similar to a traditional Ricardian model. Each ﬁrm specializes in the products in which it has comparative advantage. Also, we need to note that the
foreign ﬁrm reduces its scope relatively further, as illustrated in Figure 1.4. Though the foreign ﬁrm has local cost advantage, its rival is more competitive in the entire industry. This
may be one explanation for the empirical fact that multiproduct ﬁrms are more prevalent in
advanced countries.
As another candidate of the equilibrium, we can think of the case that there exists partial
duopoly due to a relatively small diﬀerence between cd and cf . This situation is illustrated
in Figure 1.5. Assume that Nf is given. The domestic ﬁrm will not produce in the range
located at the right-hand from B on its cost function in equilibrium. It is because the
domestic ﬁrm’s production costs are higher than the foreign ﬁrm’s costs in this range. So
cf < cd + Nd ≤ cf + Nf .
Intuitively, the foreign ﬁrm may have a stronger incentive to expand its product scope
since it has relatively lower production costs within its feasible product range. It may be
possible that the foreign ﬁrm creates its monopolized markets by adding new products and
this beneﬁt outweighs the cannibalization eﬀect. Point A in Figure 1.5 represents this situation. In other words, cf < cd + Nd < cf + Nf . The following proposition veriﬁes this
equilibrium.

18

¡

cj (i)

B¡

¡
¡
r
¡
¡
 
¡
¡
 
¡
¡
 
¡
 
¡
r¡
A¡ ¡
 
r
¡
 
¡
¡
 
¡
¡
 
¡
¡  
¡
¡  
¡
¡ 
¡
¡ 
¡
¡
r
 
¡
¡  
¡  
¡ 
¡
r 
 
 ◦
 45
cd
cf
cf + Nf
T

Ei

Figure 1.5: Partial Duopoly and Segmented Monopoly (β > 1)

Proposition 4 When each multiproduct ﬁrm has local cost advantage (β > 1) and technological diﬀerence between ﬁrms is small, partial duopoly and segmented monopoly coexist in
the global economy. While the more productive ﬁrm in duopoly markets increases its product
scope, the less productive ﬁrm in duopoly markets will decrease its product scope.
Proof. Suppose that Nf is given. A domestic ﬁrm monopolizes the markets with cd ≤ i <
cf . In the region with cf ≤ i ≤ cd + Nd , there exists duopoly.17
if cd ≤ i < cf



 L

3η


α − 2cd (i) + cf (i) − Xd
3γ
L

xd (i) =


 L α − c (i) − 2η X − η X


d
 2γ
L d L f



if cf ≤ i ≤ cd + Nd

(1.26)

By integrating (1.26) and substituting the result Xd into the optimal product scope equation,
17 Note that duopoly cannot exist when the suﬃcient condition for segmented monopoly

above is satisﬁed:α − 2cd (cf ) + cf (cf ) < 0

19

we can get the following equation.
2
1
βηNd + 2βγNd − 2γ(α − cd ) + 2γ(β − 1)(cf − cd ) + βη(cf − cd )2
2
3η
+η(cf − cd ) α + (β − 2)cf + (1 − β)cd − Xf = 0
L

(1.27)

If the last term of the equation above is positive, it is suﬃcient to show Nd < Nd .18
Note that a foreign ﬁrm produces a positive output for its core product.

xf (cf ) =

L
3η
3η
L
α − 2cf (cf ) + cd (cf ) − Xf =
α + (β − 2)cf + (1 − β)cd − Xf > 0
3γ
L
3γ
L
(1.28)

Therefore, Nd < Nd .
∗
∗
Next, assume that cd + Nd = cf + Nf . Keep in mind that Nf is a possible candidate,

not a solution. Then, the foreign ﬁrm will always produce in only duopoly markets.
∗
Xf

=

∗
cf +Nf
cf

x∗ (i)di
f

=

∗
cf +Nf
cf

3η ∗
βL ∗ 2
L
N
α − 2cf (i) + cd (i) − Xf =
3γ
L
6γ f

(1.29)

∗
Substituting (1.29) into the optimal product scope equation x∗ (cf + Nf ) = 0, we can obtain
f

the following equation.
∗
∗
βηNf 2 + 2βγNf − 2γ(α − cf ) − 2γ(β − 1)(cf − cd ) = 0

(1.30)

∗
Because the last term is negative, Nf > Nf .
∗
∗
Finally, let’s check if Nf is the solution. Using Nf = Nd − cf − cd , we must get

equation (1.27) from equation (1.30). However, this is not veriﬁed.
∗
As a result, cd + Nd < cf + Nf . Therefore, Nf > Nf > Nf .

On the other hand, it is not easy to compare the total industry output before and after
18 In the equation, the ﬁrst three terms are identical to those in the equation of a closed

economy. The next two terms are always positive.
20

globalization. From the optimal scope equations,

(Xd +Xf )−(Xd +Xf ) =

L
1
3
α − (cd + cf ) − β(cf − cd ) + β(Nd + Nf ) − β(Nd + 3Nf )
3η
2
2
(1.31)

But we can guess that total industry will be more likely to increase when cf − cd is smaller
from equation (1.31). Remember the result in a symmetric oligopoly model.

1.4.3

Global Economy: Global Cost Advantage

Suppose that the domestic ﬁrm has global cost advantage (0 < β < 1). First of all, we
can consider the equilibrium where the domestic ﬁrm monopolizes the entire industry. If the
diﬀerence between cd and cf is suﬃciently large, the foreign ﬁrm may exit the industry due
to prohibitively high production costs compared with its rival. This situation is illustrated
in Figure 1.6.

cj (i)
T

 
 
 
 
 
  ¨
  ¨¨
 ¨
¨
¨
r
 
 
 

 
 
¨
 
¨¨
  ¨¨
 ¨¨
¨
r ¨
 
 ◦
 45

cd

¨
¨¨

¨
¨¨

¨
¨
r¨
¨
¨¨

cf cd + Nd

Figure 1.6: Pure Monopoly (0 < β < 1)

21

Ei

Let’s think of the suﬃcient condition for the pure monopoly. Let pD (cf ) denote the price
of variety cf when both ﬁrms produce this product. If pD (cf ) − cf (cf ) < 0, the foreign ﬁrm
will exit.
1
α + cd (cf ) − 2cf (cf )
3
1
α − cf + (β − 1) cf − cd
=
3
α − cf
⇒ cf − cd >
1−β

pD (cf ) − cf (cf ) =

<0
(1.32)

Proposition 5 When one multiproduct ﬁrm has global cost advantage (0 < β < 1) and
α − cf
technological diﬀerence between ﬁrms is suﬃciently large (cf − cd >
), the more
1−β
productive ﬁrm will monopolize the industry in the global economy. While its product scope
remains or increases, total industry output increases.
α − cf
< cf −cd ≤ Nd . Then, it is obvious that the domestic
1−β
ﬁrm (the more productive ﬁrm) does not change its product line. That is, Nd = Nd .

Proof. First, consider the case

Second, suppose that cf − cd > Nd . To deter a rival’s entry, the domestic ﬁrm should
expand its product line in the global economy. That is, Nd = cf − cd > Nd . In conclusion,
the domestic ﬁrm will not reduce its product range after globalization.
Now, let’s consider total industry output. From the optimal scope equations of closed
economy and global economy,

Xd − Xd + X f =

2
βL
βL
βL
2
2
Nd −
Nd + Nf =
4γ
4γ
4γ

2

2
2
2Nd − Nd − Nf

>0

(1.33)

Thus, total industry output increases in the global economy.

As another candidate of the equilibrium, we can think of the case that there exists partial
duopoly due to a relatively small diﬀerence between cd and cf . This situation is illustrated
in Figure 1.7. Assume that Nf is given. In equilibrium, the domestic ﬁrm will not put
22

its marginal product in the range located at the left-hand from A on its cost function. So
cf + Nf ≤ cd + Nd .
As mentioned before, the domestic ﬁrm is more likely to expand its product range. This
is because it can expect more beneﬁt by creating its market power in spite of the cannibalization eﬀect. By contrast, the foreign ﬁrm may shrink its production scope because its
competitiveness is further harmed by globalization. Point B in Figure 1.7 represents this
situation. In other words, cf < cf + Nf < cd + Nd . The following proposition proves this
intuition.

cj (i)
T

 
 
 
 
 
 
 
 

¨
¨¨
  ¨¨
r
  ¨¨
r¨
 ¨
¨¨
¨
¨
r
r¨ B
 
¨¨
 
¨ A
  ¨¨
¨
 
¨¨
¨
r 
 
 ◦
 45
cd
cf
cf + Nf
 

Ei

Figure 1.7: Partial Duopoly (0 < β < 1)

Proposition 6 When one multiproduct ﬁrm has global cost advantage (0 < β < 1) and
technological diﬀerence between ﬁrms is small, the less productive ﬁrm competes only in
partial duopoly markets in the global economy. The more productive ﬁrm monopolizes the
rest of the product markets. While the more productive ﬁrm increases its product scope, the
less productive ﬁrm will decrease its product scope.
Proof. Above all, note that the foreign ﬁrm will always produce in only duopoly markets.
23

∗
Therefore, we can obtain the following equation just by substituting Nf into Nf in equation

(1.30) of proposition 4.
2

βηNf + 2βγNf − 2γ(α − cf ) − 2γ(β − 1)(cf − cd ) = 0

(1.34)

Since 0 < β < 1, the last term is always positive. Therefore, we can conclude that Nf < Nf .
∗
∗
Next, assume that cd + Nd = cf + Nf . Keep in mind that Nd is a possible candidate,
∗
not a solution. Substituting Nd into Nd in equation (1.27) of proposition 4 and rearranging

by using x (cf + Nf ) = 0 yields the following expression.
∗
∗
βηNd 2 + 2βγNd − 2γ(α − cd ) =
∗
− βη(cf − cd )Nd

1
+ 2γ(1 − β)(cf − cd ) + βη(cf − cd )2
2

(1.35)

∗
∗
Finally, let’s check if Nd is the solution. Using Nd = Nf + cf − cd , we must obtain

equation (1.34) from equation (1.35). However, it does not work. As a result, cf + Nf <
cd + Nd .
We can obtain the following equation by considering the additional monopoly markets of
the domestic ﬁrm between cf + Nf and cd + Nd .
2

βηNd + 2βγNd − 2γ(α − cd )
2
η
Nf βηNf + βγNf − 2γ(α − cf ) − 2γ(β − 1)(cf − cd ) = 0
3γ
2
2
1
⇒ βηNd + 2βγNd − 2γ(α − cd ) − βηNf = 0
3

+

(1.36)

The ﬁnal equation is derived from the fact that the square bracket term of the ﬁrst equation
is equal to −βγNf using equation (1.34). Therefore, we can conclude that Nd > Nd .

As expected, it is not simple to compare the total industry output before and after globalization. From the optimal scope equations,

24

(Xd + Xf ) − (Xd + Xf ) =
L
1
3
α + (7cf − 9cd ) − 2(β − 1)(cf − cd ) + β(Nd + Nf ) + 2β(Nf − 3Nd )
3η
2
2

1.5

(1.37)

Conclusion

Firm heterogeneity under monopolistic competition has become an essential part in modern trade theory. Although recently developed theories like Melitz (2003) explain some empirical ﬁndings well, there remain insuﬃciently examined areas. First of all, we often observe
that some industries are dominated by a small number of big ﬁrms. Therefore, we cannot
neglect ﬁrms’ strategic behaviors within a given market structure. On the other hand, we
should develop a new trade model for multiproduct ﬁrms in view of their huge roles in the
global economy. The growing evidence tells us that many ﬁrms, and especially most large
exporters, are multiproduct ﬁrms. This paper is inspired by these studies.
In the paper, I have developed a simple model of multiproduct ﬁrms, which highlights
the strategic decision regarding their product scope. The model yields several interesting
predictions about the product scope for multiproduct ﬁrms in the global economy. I ﬁnd
that there are several diﬀerent types of equilibria according to technological diﬀerence and
relative cost advantage. When each ﬁrm has local cost advantage and technological diﬀerence
between ﬁrms is suﬃciently large, ﬁrms have their own segmented monopolized markets after
globalization. And each ﬁrm shrinks its product scope and concentrates on more proﬁtable
products. This result is consistent with existing core competence literature. By contrast, I
show that the more productive ﬁrm can expand its product scope after globalization when
there exists a small technological diﬀerence or global cost advantage. This is because it
can obtain market power that oﬀsets the cannibalization eﬀect when its rival’s product line
expansion is limited by cost disadvantage. Therefore, this paper suggests that we should be
more careful with the eﬀects of globalization in more concentrated industries.

25

There are some areas in which my framework might be improved. First, we need to
measure explicitly changes in product diversity and overall industry productivity in order
to analyze gains from globalization. Second, multiproduct ﬁrms add and drop products
with surprising intensity and frequency as indicated by Bernard et al (2010).19 Thus, a
dynamic optimization model may be more useful in analyzing multi-stage competition among
multiproduct ﬁrms. Third, it will be important work to ﬁnd empirical evidence that supports
main predictions of this paper.

19 On average, 54 percent of US manufacturing ﬁrms change their mix of ﬁve-digit (SIC)

products every ﬁve years.
26

APPENDIX

27

Appendix
Proofs omitted in the text: page 13

For convenience, let A ≡ (βγ)2 + 2βγη(α − co ).
1
∂No
γ
= − A− 2 (α − co ) < 0
∂β
β
1
∂No
= γA− 2 > 0
∂α
1

1
−1 + A− 2 [βγ + η(α − co )]
∂No
=
>0
βγ + η(α − co ) > A 2
∂γ
η
1
∂No
γ
= 2 [1 − A− 2 {βγ + η(α − co )}] < 0
∂η
η
1
∂No
= −γA− 2 < 0
∂co

Use Xo =

1
βL
L
2
[βγ + η(α − co ) − A 2 ].
No =
2
2(m + 1)γ
(m + 1)η

∂Xo
∂β
∂Xo
∂α
∂Xo
∂γ
∂Xo
∂η
∂Xo
∂co

1
γL
[1 − A− 2 {βγ + η(α − co )}] < 0
(m + 1)η 2
1
1
L
βγ < A 2
=
[1 − A− 2 βγ] > 0
(m + 1)η
1
βL
=
[1 − A− 2 {βγ + η(α − co )}] < 0
(m + 1)η 2
βL
∂No
=
<0
(m + 1)γ ∂η
1
L
=−
[1 − A− 2 βγ] < 0
(m + 1)η

=

28

REFERENCES

29

REFERENCES

[1] Allanson, P., Montagna, C., 2005. Multiproduct ﬁrms and market structure: An explorative application to the product life cycle. International Journal of Industrial Organization 23, 587-597.
[2] Baldwin, R.E., Ottaviano, G.I.P., 2001. Multiproduct multinationals and reciprocal FDI
dumping. Journal of International Economics 54, 429-448.
[3] Bernard, A.B., Jensen, J.B., Redding, S.J., Schott, P.K., 2007. Firms in international
trade. Journal of Economic Perspectives 21(3), 105-130.
[4] Bernard, A.B., Redding, S.J., Schott, P.K., 2010. Multi-product ﬁrms and product
switching. American Economic Review 100(1), 70-97.
[5] Bernard, A.B., Redding, S.J., Schott, P.K., 2011. Multi-product ﬁrms and trade liberalization. The Quarterly Journal of Economics 126, 1271-1318.
[6] Bradley M.D., Baron, D.M., 1993. Measuring performance in a multiproduct ﬁrm: An
application to the U.S. postal service. Operations Research 41(3), 450-458.
[7] Eckel, C., Iacovone, L., Javorcik, B., Neary, J.P., 2010. Multi-product ﬁrms at home and
away: Cost- versus quality- based competence. mimeo.
[8] Eckel, C., Neary, J.P., 2010. Multi-product ﬁrms and ﬂexible manufacturing in the global
economy. The Review of Economic Studies 77, 188-217.
[9] Feenstra, R., Ma, H., 2007. Optimal choice of product scope for multiproduct ﬁrms under
monopolistic competition. NBER Working Paper No. 13703.
[10] Grossman, V., 2007. Firm size and diversiﬁcation: Multiproduct ﬁrms in asymmetric
oligopoly. International Journal of Industrial Organization 25, 51-67.
[11] Iacovone, L., Javorcik, B.S., 2008. Multi-product exporters: Diversiﬁcation and microlevel dynamics. World Bank Working Paper No. 4723.
[12] Mayer, T., Melitz, M.J., Ottaviano, G.I.P., 2009. Market size, competition, and the
product mix of exporters. NBER Working Paper No. 16959.
[13] Melitz, M.J., 2003. The impact of trade on intra-industry reallocations and aggregate
industry productivity. Econometrica 71(6), 1695-1725.
[14] Melitz, M.J., Ottaviano, G.I.P., 2008. Market size, trade, and productivity. Review of
Economic Studies 75, 295-316.

30

CHAPTER 2
Production Sharing and Exchange Rate
Pass-Through
2.1

Introduction

Price changes caused by exchange rate ﬂuctuations can have a considerable impact on
the proﬁtability of domestic ﬁrms. The most traditional view is that exchange rate shocks
change the price competitiveness of domestic ﬁrms against foreign ﬁrms. For example, an
appreciation in the domestic currency makes the foreign export price of goods produced
by domestic ﬁrms more expensive. However, some countries have continued to outperform
their export competitors despite the massive appreciation (for instance, see Athukorala and
Menon (1994) for a study on Japanese exporters). As a result, many economists have been
motivated to analyze the relationship between exchange rates and prices of traded goods
more clearly.
One remarkable outcome is now popularly known as the exchange rate pass-through
(ERPT) relationship. The rate of exchange rate pass-through can be deﬁned as the percent
change in import prices in the importing nation’s currency due to 1% change in the exchange
rate between the two trading partners. A rate of 1 (less than 1) indicates a complete (incomplete) exchange rate pass-through. Empirical studies for industrialized countries consistently
show a low pass-through of nominal exchange rate changes to import prices.1 An implication
of this empirical ﬁnding is that the “expenditure-switching” eﬀect of exchange rate changes
might be very small. If exchange rate movements have little eﬀect on prices, there is lit1 See Yang (1997), Engel (2002), Campa and Goldberg (2005).

31

tle change in the quantities traded. However, this interpretation has no micro foundation
regarding ﬁrms’ adjustment after exchange rate shocks as Rodriguez-Lopez (2011) points
out.
There are various possible explanations about sources of incomplete pass-through. In
this paper, I focus on two factors which Athukorala and Menon (1994) point out.2 The ﬁrst
factor is “pricing to market (PTM)” behavior called by Krugman (1986). Exporting ﬁrms
may take strategic pricing behavior which aims to protect market share during currency
appreciation or to augment proﬁt margins during currency depreciation. Let’s introduce a
∗
stylized framework of exchange rate pass-through and pricing to market. Let PH be the
∗
Home-currency import price of a Foreign good and PF denote the Foreign-currency export

price of a Foreign good. Also E indicates the bilateral exchange rate deﬁned as the Home
currency per unit of the Foreign currency. Assuming the law of one price holds, then
∗
∗
d ln PF
d ln PH
∗
∗
=
+1
PH = PF · E =⇒
d ln E
d ln E
ERP T

(2.1)

PTM

∗
d ln PF
= 0, there is no PTM by Foreign exporters into the Foreign-currency prices.
d ln E
∗
d ln PF
Consequently, this implies full pass-through to Home import prices. By contrast, if
=
d ln E
−1, this implies full PTM by Foreign exporters and zero pass-through into Home import

If

prices. Markup adjustment to exchange rate changes has strong empirical support. For
example, Goldberg and Knetter (1997) conclude that destination-speciﬁc changes in markups
are a very signiﬁcant factor in incomplete exchange rate pass-through. Therefore, we need
to allow endogenous markups in analyzing pass-through to ﬁrm-level prices.3
Secondly, incomplete pass-through may result from changes in the marginal cost due
to changes in imported input costs. Recently, the role of imported intermediate inputs
2 Current research points to other sources of incomplete pass-through: the importance of

local nontraded costs in the total costs, the costs of nominal price adjustment and so on.
3 Exogenous markups imply full exchange rate pass-through to ﬁrm-level import prices.

32

has received a lot of attention in explaining the eﬀect of exchange rate shocks. One of
the recent noticeable trends in world trade involves the increasing interconnectedness of
production processes in a vertical trading chain across many countries (Hummels et. al
(2001)4 ). With the growth of vertical specialization5 , we need to pay more attention to
changes in input prices due to exchange rate movements. In an economy with intermediate
goods trade, exchange rate changes will aﬀect not only the relative price of ﬁnal goods, but
also the relative price of local inputs to imported inputs. Exchange rate appreciations have
a negative eﬀect on domestic ﬁrms through the traditional revenue channel, but a positive
eﬀect through the cost channel by lowering prices of imported inputs.
These diﬀerent channels of exchange rate shocks have been examined more vigorously in
the empirical area by the practical necessity of policy-makers. Athukorala and Menon (1994)
conclude that the cost lowering eﬀect of exchange rate changes seems to have provided
Japanese exporters with considerable leverage in enduring the yen appreciation. Ahmed
(2009) shows there is a signiﬁcant negative eﬀect on exports of Chinese Renminbi (RMB)
appreciation against the currencies of China’s advanced-economy trading partners, and there
is a positive eﬀect on processing exports of Chinese RMB appreciation against other emerging
Asian currencies.6 Greenway et al. (2010) ﬁnd evidence of both a negative eﬀect from
appreciation and an oﬀsetting eﬀect through imported intermediate inputs for a sample of
UK manufacturing ﬁrms between 1988 and 2004.
With production sharing the role of imported inputs in aﬀecting pass-through becomes
important. Existing literature suggests that pass-through is muted when ﬁrms use imported
inputs. Goldberg and Knetter (1997) state that if changes in the value of the home cur4 For their 14-country sample, the vertical specialization share grew by about 30% between

1970 and 1990, and growth in vertical specialization exports accounted for 30% of the growth
in the overall export /GDP ratio.
5 There exist various expressions with the same meaning in the trade literature:“slicing
up the value chain”, “disintegration of production”, “international fragmentation” etc.
6 China’s exports include a substantial amount of intermediate inputs imported from
other East Asian economies. Koopman, Wang, and Wei (2008) illustrate that this East
Asian supply chain is particularly dominant in electronic products.

33

rency against the dollar inﬂuence exporters’ marginal costs by inducing imported input
price changes, pass-through to the U.S. could be muted relative to other markets. Yang
(1997, 1998) shows that there is a negative relationship between the elasticities of marginal
cost with respect to exchange rate and pass-through. Cost shocks oﬀset the eﬀect of the
exchange rate change on the Foreign ﬁrm’s export price in the importing nation’s currency.
The aim of this paper is to investigate the eﬀect of diﬀerent channels through which
exchange rate shocks aﬀect ﬁrms’ decisions on pricing and entry and exit. I extend the
model of exchange rate pass-through with endogenous markups built by Rodriguez-Lopez
(2011) introducing the intermediate input sector. In his paper, he assumes that exchange
rate movements are exogenous and that wages are ﬁxed.7 Exchange rate movements generate reallocation of ﬁrms because ﬁrms are heterogeneous with respect to their levels of
productivity. As a result, exchange rate movements aﬀect the extensive margin of trade–the
number of goods traded–by altering the cut-oﬀ productivity levels. In addition, I introduce two factors, domestic and imported intermediate inputs. Therefore, ﬁrms’ exposure
to exchange rates includes two aspects: exports of ﬁnal goods and imports of intermediate
inputs.
Rodriguez-Lopez shows that low levels of exchange rate pass-through to ﬁrm- and aggregatelevel import prices coexist with large movements in trade ﬂows. However, these results may
not be reconciled with some empirical observations. First, the pass-through to the aggregate
import price is always negative in his original model.8 This consequence results mainly from
a composition bias. In response to a Home currency depreciation, the least productive Foreign exporters leave the Home market as their competitiveness weakens. Under intensiﬁed
competition with Home ﬁrms, the surviving Foreign exporters must decrease their markups.
In the end, aggregate import prices are computed using only the survivors who are the most
7 There exists strong evidence of nominal wage rigidities from many countries. See Akerlof

(2007) for detail.
8 To solve this problem, he proposes an extended model considering product quality. His
quality model predicts that the pass-through to the aggregate unit import price can be
positive.

34

productive.9 Second, his model predicts large and unilateral expenditure-switching eﬀects
of exchange rate ﬂuctuations. But some empirical studies show us that imports and exports
are much less responsive to exchange rate movements.10
This paper proposes a theoretical background for various possibilities of exchange passthrough and expenditure-switching. The shares of imported inputs in total costs are important determinants.11 According to the relative sizes of these shares, exchange rate movements might lead to diﬀerent cost shocks to each trading nation. When the imported input
shares are located within some range, a low but positive rate of pass-through to aggregate
import prices can be derived in the model. In addition, both the size and the direction of
expenditure-switching eﬀects are ambiguous because they also depend on the imported input
shares. Therefore, low levels of exchange rate pass-through to aggregate import prices can
coexist with negligible movements in trade ﬂows unlike Rodriguez-Lopez. Although I use a
partial equilibrium model with ﬁxed wages, we can investigate welfare implications of exchange rate changes brieﬂy. Under the current model, the eﬀect of exchange rate movements
on welfare in each trading country is ambiguous. This is because a change in the overall
competitive environment by exchange rate shocks may vary with the impact of cost shocks.
In this paper, the overall competitive environment is captured by the number of competitors
(or the number of product varieties) in each market.

9 Rodriguez-Lopez indicates that aggregate prices can diﬀer drastically from the average

ﬁrm-level pass-through rate due to a sample selection problem related to changes in the
extensive margin of trade.
10 For example, Dong (2012) indicates that both U.S. imports and exports have become
much less responsive to exchange rate movements in recent years.
11 According to Campa and Goldberg (2006), the ratio of imported inputs to total costs
diﬀers signiﬁcantly by country and industry. In general, larger countries have a lower share
of imported inputs into production while smaller countries have a higher share. Within
the manufacturing sector, chemicals has the largest share of imported inputs, 67 percent of
total costs, followed by electrical machinery and medical and precision instruments, both
with imported input shares above 50 percent. The industries within manufacturing with the
lowest imported input shares are forestry and metal ores.
35

2.2
2.2.1

Basic Model
Preferences and Demand

We begin by considering an economy which imports intermediate inputs from a trading
partner, but does not trade diﬀerentiated ﬁnal goods. There are two countries, Home,
Foreign. We use a star (*) to denote Foreign variables.
L identical consumers share the same utility function given by
1
1
z(ω)2 dω − η
U = z0 + α
z(ω)dω − γ
2 ω∈Ω
2
ω∈Ω

2

z(ω)dω

(2.2)

ω∈Ω

where z0 and z(ω) are, respectively, the individual consumption of the numeraire good and
quantity of variety ω in the diﬀerentiated sector. The demand parameters α, γ and η are
all positive. The parameter α and η denote the substitution pattern between the numeraire
and the diﬀerentiated varieties: both increases in α and decreases in η raise the demand for
the diﬀerentiated varieties relative to the numeraire. The parameter γ represents the degree
of product diﬀerentiation among the varieties.
Total expenditure of the representative consumer is given by

I = z0 +

p(ω)z(ω)dω

(2.3)

ω∈Ω

where p(ω) is the price that consumers pay for a speciﬁc variety ω.
We assume that consumers have positive demands for the numeraire good (z0 > 0). The
inverse demand for each variety ω is then given by

p(ω) = α − γz(ω) − ηZ

where Z ≡

z(ω)dω denotes the consumption level over all varieties.
ω∈Ω

36

(2.4)

Let Ω∗ ⊂ Ω be the subset of varieties that are consumed (z(ω) > 0). We can derive the
following linear market demand system for these varieties.

y(ω) ≡ Lz(ω) =

L
γα
ηN
− p(ω) +
p ,
¯
γ ηN + γ
ηN + γ

where N is the measure of consumed varieties and p =
¯
price. The set Ω∗ is the largest subset of Ω that satisﬁes

p(ω) ≤

∀ω ∈ Ω∗

(2.5)

1
p(ω)dω is their average
N ω∈Ω∗

1
(γα + ηN p) ≡ p
¯
ˆ
ηN + γ

(2.6)

where p is the price-ceiling for all varieties, above which the demand for an individual variety
ˆ
will be zero.
Let’s substitute (2.6) into (2.5). The quantity demanded for each variety is

y(ω) =

2.2.2

L
[ˆ − p(ω)]
p
γ

(2.7)

Production

Labor is the only factor of production and is assumed to be inelastically supplied in
a competitive market. There are three sectors in each country. The ﬁnal-good sector is
monopolistically competitive while the intermediate-input sector and the numeraire good
market are competitive.
Production of the ﬁnal good requires two types of intermediate inputs: domestic (Id )
and imported (Im ). We assume that producing one unit of a domestic input requires one
unit of labor and, given the nominal wage rate W , entails cost W . It will turn out that in
equilibrium the price of each intermediate input equals the marginal cost of producing the
input. The numeraire good is produced under constant to returns at unit cost. That is, one
1
unit of the numeraire good is produced using
units of labor. A parallel assumption holds
W

37

for Foreign country. In the Foreign country, the marginal cost of producing the input is W ∗
1
and one unit of the numeraire good is produced using ∗ units of labor.
W
We adopt a CES production function that transforms intermediate inputs to ﬁnal goods
like Bas (2009).
σ−1
σ−1 σ
1
y = [(Id ) σ + (φIm ) σ ] σ−1
c

(2.8)

1
indexes ﬁrm productivity and is assumed to be diﬀerent among heterogeneous ﬁrms.
c
Each ﬁrm learns about c only after paying the entry cost.
where

σ is the elasticity of substitution between two inputs. We assume that domestic and
imported inputs are imperfect substitutes: 1 < σ < ∞. The parameter φ measures the
imported inputs requirements in the production process. Higher φ indicates higher eﬃciency
of imported inputs.

2.2.3

Firms’ Optimization

First, we can consider the following cost minimization problem conditional y.
min

Id ,Im

C(y) = W Id + EW ∗ Im

σ−1
σ−1 σ
1
s.t. y = [(Id ) σ + (φIm ) σ ] σ−1
c

(2.9)

where E is the nominal exchange rate, measured as the Home currency per unit of the
Foreign currency. Actually, E can be interpreted as the real exchange rate in this paper.
This is because W and W ∗ denote labor productivity in Home and Foreign, respectively.12
Equation (2.9) yields the following relationship between factor demands and their relative
W∗

∗
∗
∗
E
12 Q = EP = M P L∗ = E W M P L = E W W = E where Q is the real exchange
W
P
W M P L∗
W W∗
MPL
rate, P and P ∗ are the price levels of Home and Foreign country. M P L and M P L∗ are
marginal products of labor in the non-tradable numeraire sector of the two countries, respectively.

38

prices.
Im
=
Id

σ
W
φσ−1
∗
EW

(2.10)

From equation (2.10), exchange rate appreciations (a decrease in E) or higher imported
input requirements raise the relative demand for imported intermediate inputs.
Substituting (2.10) into (2.8),
σ

1
y=
c
=

1+

σ−1
σ−1
W
Id
φσ−1
EW ∗

(2.11)

Aσ

I
c d
1

σ−1
σ−1
W
where A ≡ 1 +
φσ−1
EW ∗
Now let’s calculate the eﬀective marginal cost of producing one unit of ﬁnal good, which

is identical to the average cost.13
C(y)
W Id + EW ∗ Im
W Aσ−1 Id
Wc
=
=
=
σ
A I
y
y
A
c d

(2.12)

W
and c. For our
A
better understanding, suppose that we can split production process of a ﬁnal good into two
Therefore, the eﬀective marginal cost is the product of two terms,

stages. At the ﬁrst stage, ﬁrms produce virtual intermediate inputs IV using domestic and
σ−1

σ−1

σ

+ (φIm ) σ ] σ−1 . In the second stage, a ﬁnal good is
I
produced using these virtual inputs, that is y = V . Then c can be interpreted as unit
c
W
virtual input requirement for a ﬁnal good. The price of a virtual input is given by
. In
A
W
other words,
represents the eﬀective input prices of Home ﬁrms. If Home ﬁrms do not
A
use imported inputs at all (φ = 0), the eﬀective input price would be W . Access to foreign
imported inputs. IV = [(Id ) σ

inputs lowers the eﬀective input price from the existence of an additional term A > 1 in
13 As we see in equation (2.12), the marginal cost is not given directly and is computed

indirectly. Thus, I use the term “eﬀective marginal cost”.

39

the denominator. Therefore, we can interpret A as the coeﬃcient of productivity growth
due to imported inputs. Note that exchange rate movements cause marginal cost shocks by
changing A.
Given the technology, the Home ﬁrms’ proﬁt function is
π(c) =p(c)y(c) − W Id − EW ∗ Im
Wc
y(c)
A
Wc L
= p(c) −
[ˆ − p(c)]
p
A γ

= p(c) −

(2.13)

The pricing rule is derived by maximizing proﬁts with respect to the price. Given the
continuum of competitors, each individual ﬁrm regards itself as quite small relative to the
market as a whole and treats p and N as unaﬀected by its own choices. Therefore, a ﬁrm
¯
takes p as given.
ˆ
1
2

p(c) =

p+
ˆ

Wc
A

(2.14)

Let µ(c) and π(c) denote the (absolute) mark-up and proﬁt.

µ(c) =

1
2

p−
ˆ

Wc
A

(2.15)

y(c) =

L
2γ

p−
ˆ

Wc
A

(2.16)

π(c) = µ(c)y(c) =

L
4γ

p−
ˆ

Wc 2
A

(2.17)

As expected, lower cost ﬁrms (high productivity ﬁrms) set lower prices and higher markups and earn higher proﬁts.14
14 In this paper, c is basically a physical unit which is not measured in terms of money

value. Because we assume ﬁxed wages, c is expressed as cost draws by abuse of notation.

40

2.2.4

Trade in Final Goods

Home and Foreign countries have respectively L and L∗ identical consumers who share
the same preferences. The ﬁnal goods markets in the two counties are segmented, but ﬁrms
can produce in one market and sell in the other, incurring a per-unit trade cost. Let τ > 1
be the iceberg cost for Home ﬁrms. In the same way, τ ∗ accounts for the iceberg cost for
Foreign ﬁrms.
Let pD (c) and pX (c) denote the nominal domestic and export prices of a Home ﬁrm with
cost c. These prices are set in the currency of the destination country.
Following equation (2.14), we can write the pricing equations for a Home ﬁrm with cost
c as
pD (c) =

1
2

p+
ˆ

Wc
A

,

pX (c) =

1
2

p∗ +
ˆ

τ Wc
·
E A

1
2

p−
ˆ

Wc
A

,

µX (c) =

1
2

p∗ −
ˆ

τ Wc
·
E A

Mark-ups are given by

µD (c) =

Also we can get the following proﬁt functions.
L
πD (c) =
4γ

Wc 2
p−
ˆ
,
A

L∗
πX (c) =
4γ

p∗
ˆ

τ Wc 2
− ·
E A

Let A∗ be the the coeﬃcient of productivity growth due to imported inputs by Foreign
ﬁrms.

1

A∗ =

1+

EW ∗ σ−1 ∗ σ−1 σ−1
(φ )
W

Analogously, prices, mark-ups and proﬁts by a Foreign ﬁrm are given by
p∗ (c) =
D

1
2

p∗ +
ˆ

W ∗c
A∗

,

p∗ (c) =
X

41

1
2

p + τ ∗E ·
ˆ

W ∗c
A∗

µ∗ (c) =
D
∗
πD (c)

2.2.5

1
2

L∗
=
4γ

p∗ −
ˆ
p∗
ˆ

W ∗c
A∗

µ∗ (c) =
X

,

W ∗c 2
− ∗
,
A

∗
πX (c)

1
2

L
=
4γ

p − τ ∗E ·
ˆ
p − τ ∗E
ˆ

W ∗c
A∗
W ∗c 2
· ∗
A

Cut-oﬀ Productivity Levels

∗
We deﬁne the cut-oﬀ rules as cr = sup{c : πr (c) > 0} and c∗ = sup{c : πr (c) > 0} for
r

r ∈ {D, X}.
cD =
c∗ =
D

Aˆ
p
,
W

A∗ p∗
ˆ
,
W∗

cX =

AE p∗
ˆ
τW

A∗ p
ˆ
c∗ = ∗
X
τ EW ∗

Then, we derive the following relationships between cut-oﬀ levels.
c∗ =
X

cX =

A∗ W
Aτ ∗ EW ∗
AEW ∗
A∗ τ W

cD

c∗
D

(2.18)

(2.19)

where the term in parentheses in each equation is the relative variable cost.
We assume that cost draws c are Pareto distributed in the interval [0, cM ] in both counc k
, where k > 1
tries. The cumulative distribution function of costs is given by G(c) =
cM
indexes the dispersion of cost draws. As k increases, the ratio of high-cost ﬁrms increases
and the cost distribution is more clustered near the upper bound.
Entry is unrestricted in both countries. Firms will enter in each country as long as their
∗
expected proﬁts are no less than the sunk entry cost. Let fE and fE be the entry cost in

units of eﬀective labor of the Home and Foreign country, respectively. Consequently, the
∗
sunk cost in each country is represented by W fE and W ∗ fE in nominal terms.

42

Free entry conditions for both countries are given by
cD
0
c∗
D
0

cX

πD (c)dG(c) + E

0

πX (c)dG(c) = W fE

(2.20)

∗

∗
πD (c)dG(c) +

1 cX ∗
∗
πX (c)dG(c) = W ∗ fE
E 0

(2.21)

Using the Pareto parametrization for the cost draws, the free entry conditions can be
re-written as follows.

L

L∗

W 2
(cD )k+2 + EL∗
A

τW 2
(cX )k+2 = ϕW fE
AE

W ∗ 2 ∗ k+2 1
cD
+ L
A∗
E

τ ∗ EW ∗ 2 ∗ k+2
∗
cX
= ϕW ∗ fE
A∗

(2.22)

(2.23)

where ϕ ≡ 2γ(k + 1)(k + 2) (cM )k
We now solve for the equilibrium cut-oﬀ levels.
1

A2 ϕW fE (τ ∗ )k {τ k − ψ} k+2
·
cD =
LW 2
(τ τ ∗ )k − 1

A2 EϕW fE

cX =


c∗ = 
D

L∗ (τ W )2

∗
(A∗ )2 ϕW ∗ fE

L∗ (W ∗ )2

(2.24)

1

·

(τ ∗ )k ψ − 1 k+2
(τ τ ∗ )k − 1

τk
·

 1
k+2
1
−
ψ 
(τ τ ∗ )k − 1

(2.25)

(τ ∗ )k

 1
k+2
τk − 1
∗ 2
∗ ∗
ψ
 (A ) ϕW fE

c∗ = 
·

X
LE (τ ∗ W ∗ )2 (τ τ ∗ )k − 1

(2.26)



(2.27)

∗
EW ∗ fE EW ∗ /A∗ k
is a combined measure of the entry cost and the variable
W fE
W/A
cost of Foreign country relative to Home country. Therefore, ψ can be deﬁned as the index

where ψ ≡

43

describing relative competitiveness of Home ﬁrms.
1
We see that ψ must range between
and τ k in order to obtain positive equilibrium
∗ )k
(τ
15 Because we treat all the variables except exchange rates as ﬁxed, this
cut-oﬀ levels.
condition consequently limits exchange rate ﬂuctuations. Throughout this paper, we will
assume this necessary condition holds.

2.2.6

Prices, Product Varieties and Welfare

Let g(c|c ≤ cr ) denote the probability density function for costs of Home ﬁrms that
actually sell in market r ∈ {D, X}. Given the Pareto distribution assumption, we obtain

 g(c)
kck−1


=
if c ≤ cr
G(cr )
(cr )k
g(c|c ≤ cr ) =



0
otherwise
A parallel conditional distribution holds for Foreign ﬁrms.
Let pr and p∗ denote the average price of Home goods and Foreign goods available in
¯
¯r
market r, for r ∈ {D, X}. Also, let p and p∗ represent the average price of all goods at
¯
¯
Home and at Foreign, respectively. We obtain the following proposition.
Proposition 7
The average prices of domestic and imported goods are equal:

p = p D = p∗ =
¯ ¯
¯X

1
k+2
k+1

W
c and p∗ = p∗ = pX =
¯
¯D ¯
A D

k+1
2
k+1

W∗ ∗
c
A∗ D

The equivalence of average prices of domestic and imported goods results from the identical price distribution faced by all ﬁrms competing in the same market. From pricing
equations, (2.18) and (2.19), we can check this fact easily.
15 To ensure an interior solution, we also need the equilibrium cut-oﬀ levels to be smaller

than or equal to cM .

44

From equation (2.6) and Proposition 7, the number of ﬁrms selling in each country is
given by equation (2.28).16
W
A
2(k + 1)γ α − A cD
2(k + 1)γ
·
N=
α−1
=
Wc
η
η
W cD
D
A
W ∗ c∗
2(k + 1)γ
2(k + 1)γ α − A∗ D
A∗
=
N∗ =
·
α−1
η
η
W ∗ c∗
W ∗ c∗
D
A∗ D

(2.28)

With Meltiz-Ottaviano (2008) utility function, the overall competitive environment is
characterized by the number of competing varieties and average prices. When the domestic
cut-oﬀs decrease, average prices decline and product variety increases. That is, the competition environment becomes tougher. The following corollary shows us that welfare increases
with decreases in the domestic cut-oﬀs.
Corollary 1 Welfare can be evaluated using the following indirect utility functions.
W
k+1W
1
α − cD
α−
c
2η
A
k+2 A D
1
W∗
k + 1 W∗ ∗
c
U∗ = W ∗ +
α − ∗ c∗
α−
2η
A D
k + 2 A∗ D
U = W+

2.2.7

Entrants, Producers and Exporters

∗
Let Ne (Ne ) denote the mass of entrants at Home (Foreign). Also let ND and NX be the

mass of Home ﬁrms selling at Home and at Foreign, respectively. With similar expressions
for Foreign ﬁrms, for r ∈ {D, X} we have

Nr = G(cr )Ne =

cr k
Ne
cM

(2.29)

16 In Rodriguez-Lopez (2011), N and N ∗ are constant due to the use of translog preferences

instead of a quadratic utility. However, the composition of domestic and imported goods in
each market may change in response to exchange rate ﬂuctuations.

45

∗
Nr

=

∗
G(c∗ )Ne
r

=

c∗ k ∗
r
Ne
cM

(2.30)

∗
∗
Because N = ND + NX and N ∗ = ND + NX , we can solve the number of entrants in each

country.

Ne =

∗
Ne

=

2(k + 1)γ (cM )k

(τ τ ∗ )k

η (τ τ ∗ )k − 1

(cD )k

2(k + 1)γ (cM )k

(τ τ ∗ )k

η (τ τ ∗ )k − 1

c∗
D

k

A
1
α−1 −
W cD
(cX )k
A∗
1
∗ c∗ α − 1 − ∗ k
W D
cX

A E
α−1
W cX τ
A∗
1
∗ c∗ Eτ ∗ α − 1
W X

(2.31)

(2.32)

∗
Both Ne and Ne must be non-negative. The following corollary shows that if trade costs

are suﬃciently high, exporters always are more productive than non-exporters. Empirical
studies ﬁnd that exporting ﬁrms tend, on average, to be more productive.17 Therefore, I
assume that Corollary 2 holds throughout this paper.18
1
Corollary 2 If ∗ ≤ E ≤ τ , cD ≥ cX and c∗ ≥ c∗ .
D
X
τ

2.3

The Eﬀect of Exchange Rates in Partial Equilibrium

2.3.1

The Cut-oﬀ Levels and the Exchange Rate

To understand how exchange rate shocks are reﬂected in prices, we begin by analyzing
their impact on the cut-oﬀ levels.
17 Bernard and Jensen (1999), Clerides, Lach and Tybout (1998) and Aw, Chung and

Roberts (2000) are frequently cited empirical papers which support a positive correlation
between exporting and productivity.
18 Purchasing power parity states that the real exchange rate is equal to 1. Therefore,
assuming this parity holds, the suﬃcient condition of Corollary 2 is also satisﬁed.

46

Taking the natural logarithm of equation (2.24), we obtain
1
ln cD =
ln νD + 2 ln A + ln(τ k − ψ)
k+2

ϕfE
(τ ∗ )k
where νD ≡
·
LW (τ τ ∗ )k − 1

Let c ,E be the elasticity of cD with respect to the exchange rate.
D

cD ,E

=

∂ ln cD
1
∂ ln A
ψ
∂ ln ψ
=
2·
− k
·
∂ ln E
k+2
∂ ln E τ − ψ ∂ ln E

(2.33)

Thus, c ,E is comprised with combination of two diﬀerent eﬀects. From the bracket in
D
equation (2.33), the ﬁrst term indicates a negative cost shock due to higher prices of imported
inputs. With a depreciation in the Home currency, Home ﬁrms experience a rise in imported
input prices because the coeﬃcient A decreases. All other things being equal, Home ﬁrms’
cut-oﬀs must decrease to compensate for higher input prices.
Next, the second term indicates the competitiveness eﬀect. As deﬁned, an increase in ψ
implies that Home ﬁrms become more competitive. A depreciation of the Home currency
aﬀects relative competitiveness in two opposite directions. To begin with, a Home currency
depreciation directly raises both the variable cost and entry cost of Foreign ﬁrms expressed
in the Home currency. On the other hand, with a depreciation in the Home currency, Home
ﬁrms experience a negative cost shock while Foreign ﬁrms face a positive cost shock. This
indirect channel makes Home ﬁrms less competitive contrary to the former direct channel. As
∂ ln ψ
, is unclear. Suppose ﬁrst
a result, the net eﬀect of these two forces, namely the sign of
∂ ln E
∂ ln ψ
that a depreciation of the Home currency makes Home ﬁrms more competitive
>0
∂ ln E
in the Foreign market. Then, an increase in the ex ante expected proﬁt from exporting
gives rise to downward pressure on the domestic cut-oﬀ for Home ﬁrms taking into account
the free-entry condition. Reversely, the weakened competitiveness of Home ﬁrms due to a
∂ ln ψ
depreciated domestic currency
< 0 tends to put upward pressure on cD .
∂ ln E

47

From the deﬁnition of A and A∗ ,

∂ ln A
=−
∂ ln E

∂ ln A∗
∂ ln E

σ−1

W
EW ∗

W
EW ∗

1+

EW ∗
W

=
1+

φσ−1

σ−1

σ−1

(2.34)

= A∗

(2.35)

(φ∗ )σ−1

σ−1

EW ∗
W

=− A
φσ−1

(φ∗ )σ−1

These elasticities (in absolute value) range between 0 and 1. We can interpret the economic meaning of A and A∗ easily. From equation (2.10),
EW ∗ Im
W Id

=

W
EW ∗

σ−1

φσ−1

=⇒ A =

W
EW ∗

σ−1

W
1 + EW ∗

φσ−1

σ−1

=
φσ−1

EW ∗ Im
(2.36)
W Id + EW ∗ Im

Therefore, A represents the share of imported inputs in total variable costs of Home ﬁrms
and A∗ denotes the corresponding imported input share of Foreign ﬁrms.
From the deﬁnition of ψ,
∂ ln ψ
= (k + 1) − k ( A + A∗ )
∂ ln E

(2.37)

Substituting (2.34) and (2.37) into (2.33), we obtain the following expression.

cD ,E

=

1
(k + 2)(τ k − ψ)

kψ − 2(τ k − ψ)

A

+ kψ A∗ − ψ(k + 1)

(2.38)

Therefore, we see that the impact of exchange rate shocks on the cut-oﬀ level depends on
some parameters.

48

Proposition 8
1. (a) If τ k ≥

k+1
2

ψ, c ,E < 0.
D

2(τ k − ψ)
k+1
ψ, c ,E ≥ 0 only when A∗ ≥
−1
D
2
kψ
Otherwise, c ,E < 0.
D

(b) If ψ < τ k <

A +1+

1
.
k

k+1 1
, ∗
> 0.
2
ψ cD ,E
k A − (k + 1)
1
k+1 1
(b) If < (τ ∗ )k <
, c ∗ ,E ≤ 0 only when A∗ ≥
.
D
ψ
2
ψ
2[(τ ∗ )k ψ − 1] − k
Otherwise, c∗ ,E > 0.

2. (a) If (τ ∗ )k ≥

D

Consider ﬁrst the impact of a depreciation of the Home currency on cD , the cut-oﬀ level
for domestic producers at Home. Equation (2.33) tells us that the competitiveness eﬀect is
very slight when trade costs are suﬃciently large. In this case, the cut-oﬀ for Home ﬁrms
selling in their domestic market always decreases to make up for higher prices of imported
inputs as shown in Proposition 8-1 (a). Unless trade costs are large enough to incapacitate
the competitiveness eﬀect, the sign of c ,E is determined by the relative size of a directlyD
induced negative cost shock and the competitiveness eﬀect. Figure 2.1 (a) illustrates how
the sign of c ,E is determined within A A∗ range. Note that c ,E < 0 holds in most
D
D
areas.
For c ,E to be positive, a currency devaluation must weaken the competitiveness of
D
Home ﬁrms considerably. Then, pre-entry expected proﬁt from exporting decreases for Home
producers. In order to satisfy the free-entry condition, the pre-entry expected proﬁt from
domestic sales increases for Home ﬁrms. This occurs if and only if cD increases. To obtain
this result, A∗ needs to be suﬃciently large and so Foreign ﬁrms enjoy the huge beneﬁt from
cheaper prices of imported inputs after a Home currency depreciation. On the contrary, A
may aﬀect c ,E through both a directly-induced cost shock eﬀect and the competitiveness
D
eﬀect in the opposite direction.
Analogously, we can obtain the cut-oﬀ rule for Foreign ﬁrms selling domestically as the
second part of Proposition 8. From Figure 2.1, a depreciation of Home currency is likely to
49

A∗

A∗

4/3
cD ,E

1

1

cD ,E

c∗ ,E
D

>0

c∗ ,E
D

<0

0

1

>0

0

A

<0

1

4/3 A

(b) c∗ ,E
D

(a) c ,E
D

Figure 2.1: The Impact of Exchange Rate Movements on the Cut-oﬀ Level (k = 3, ψ =
1, τ = τ ∗ = (1.5)1/3 )
For interpretation of the references to color in this and all other ﬁgures, the reader is referred
to the electronic version of this dissertation.

lead a decline in the cut-oﬀ for Home domestic producers and a rise in the cut-oﬀ for Foreign
∂ ln ψ
domestic producers. This is because Home ﬁrms’ competitiveness generally improves (
∂ ln E
∂ ln ψ
is positive) or weakens mildly (
is negative but relatively small in absolute value) due to
∂ ln E
a depreciation of Home currency. For example, suppose that a Home-currency depreciation
leads to the improvement of competitiveness of Home ﬁrms. Then, the entry at Home
increases while the entry at Foreign declines. Assuming the trade cost is suﬃciently large,
increased competition among local ﬁrms dominates in the Home market. cD decreases by
the intensiﬁed competition at Home combined with higher imported input costs.
Given the relationships established in equations (2.18) and (2.19),

c∗ ,E
X

= c ,E + A + A∗ − 1
D

50

(2.39)

cX ,E

= c∗ ,E + 1 − A − A∗
D

(2.40)

From Proposition 8, if c ,E ≥ 0, c∗ ,E ≥ 0. Also, if c∗ ,E ≤ 0, c ,E ≤ 0. But with
D
X
X
D
cD ,E < 0 ( c∗ ,E > 0), we cannot determine the sign of c∗ ,E ( cX ,E ).
X
D
In conclusion, the eﬀect of exchange rate movements on the cut-oﬀ levels is not unilateral
when we include a tradable intermediate input sector. This property yields distinguishing
features of this model.

2.3.2

Welfare

According to the results in Section 2.2.6, we must analyze the change in the eﬀective
marginal cost cut-oﬀ to investigate the impact of exchange rate ﬂuctuations on average
prices, product varieties and welfare. For simplicity of notation, we deﬁne the eﬀective
marginal cost cut-oﬀ as the product of the eﬀective input price and cost cut-oﬀ. That is,
W
W∗
λr =
cr and λ∗ = ∗ c∗ for r ∈ {D, X}. And the elasticities of these cut-oﬀs with respect
r
A
A r
to the exchange rate are deﬁned as λr ,E and λ∗ ,E for r ∈ {D, X}. Then, the following
r

equations are readily obtained.
= A + c ,E
D

(2.41)

= − A∗ + c∗ ,E
D

(2.42)

λD ,E
λ∗ ,E
D

Therefore, λ ,E < 0 ( λ∗ ,E < 0) is the suﬃcient condition for Home welfare (Foreign
D
D
welfare) to increase due to a Home currency depreciation. Proposition 9 tells us that the
impact of exchange rate shocks on each country’s welfare is ambiguous.
Proposition 9
τk
1
1. If A∗ < −
A + 1 + , a depreciation of the Home currency raises Home welfare.
ψ
k
Otherwise, Home welfare decreases or remains unchanged due to a depreciation of the
Home currency.

51

2. If A∗ > −

1

+

k+1

, a depreciation of the Home currency raises Foreign
kψ (τ ∗ )k
welfare. Otherwise, Foreign welfare decreases or remains unchanged due to a depreciA
(τ ∗ )k ψ

ation of the Home currency.

A∗

A∗

4/3
1

1

∂U ∗
>0
∂E

∂U
<0
∂E

∂U ∗
<0
∂E

∂U
>0
∂E

0

1

0

A

1

4/3 A

(b) λ∗ ,E
D

(a) λ ,E
D

Figure 2.2: The Impact of Exchange Rate Movements on Welfare (k = 3, ψ = 1, τ = τ ∗ =
(1.5)1/3 )
Figure 2.2 illustrates the impact of a Home currency depreciation on welfare in each
country. In comparison with Figure 2.1, the welfare implication of exchange rate changes
is more ambiguous. Again, the eﬀective marginal cost cut-oﬀ is the product of the eﬀective
input price and cost cut-oﬀ. In the case of λD , the cost cut-oﬀ cD is likely to decrease in
response to a Home currency depreciation as shown in Figure 2.1 (a). However, there exits
W
a negative cost shock represented by an increase in
, which is a factor for the rise in λD .
A
W
W∗
We need to note that a rise in
causes a decrease in Ne and a decline in ∗ leads to an
A
A
∗ from equation (2.31) and (2.32) when the cost cut-oﬀs remain unchanged. In
increase in Ne
other words, a negative cost shock due to a Home-currency depreciation reduces the number
of entrants at Home and a positive cost shock increases the number of entrants at Foreign.
We have to consider the eﬀects of cost shocks on entry decision together with the changes in

52

cut-oﬀ levels to investigate the changes in the overall competitive environment determining
welfare changes.
However, we must keep in mind that proposition 9 is derived from a restricted model.
Consequently, there are deﬁnite limitations of this welfare analysis. First, most economists
believe a signiﬁcant depreciation of a country’s currency always reduces its consumption and
welfare. This is because a depreciation of the country causes a deterioration in its terms
of trade. With our current partial equilibrium model, we cannot investigate the eﬀect of
exchange rate shocks on terms of trade eﬀectively. Secondly, exchange rate changes may
lead to substantial valuation eﬀects in the form of capital gains or losses. Given a country’s
international investment position, asset channel can oﬀset some or all of the eﬀects of the
trade channel.19

2.3.3

Exchange Rate Pass-Through

Firm-level Prices and Trade Flows
The Home-currency import price of a good produced by a Foreign ﬁrm with cost c is
given by
W∗
1
p∗ (c) = τ ∗ E ∗ c∗ + c
X
X
2
A

(2.43)

Therefore, the pass-through rate deﬁned as the elasticity of p∗ (c) with respect to the
X
exchange rate is given by
∗
ξX (c) =

∂ ln p∗ (c)
c∗
X
= 1 − A∗ + ∗ X · c∗ ,E
∂ ln E
cX + c
X

for c ≤ c∗
X

(2.44)

c∗
where note that ∗ X ∈ [1/2, 1).
cX + c
From equation (2.44), we can sort out three forces when the exchange rate changes: (1)
the direct eﬀect on the ﬁrm’s variable cost (1 − A∗ ), (2) the change in the competitive
19 See Tilte (2005) for valuation eﬀect of exchange rate movements.

53

environment of Foreign exporters ( c∗ ,E ), (3) ﬁrm-speciﬁc eﬀect reﬂecting a Foreign ﬁrm’s
X
c∗
relative position with respect to the cut-oﬀ level ( ∗ X ). The following proposition states
cX + c
exchange rate pass-through to ﬁrm-level import prices.
Proposition 10
∗
1. If c∗ ,E > 0, 1− A∗ < ξX (c) < λ ,E . More productive ﬁrms have higher pass-through
D
X

rates.
∗
2. If c∗ ,E ≤ 0, λ ,E ≤ ξX (c) ≤ 1− A∗ . More productive ﬁrms have lower pass-through
D
X
rates.

First, suppose that c∗ ,E > 0. This implies that Foreign exporters face a weaker comX
petitive environment at Home due to a depreciation of the Home currency. We can also show
that λ ,E > 0 whenever c∗ ,E > 0. So N decreases. Because the number of competitors
D
X
in the Home market declines, Foreign exporters can increase their mark-ups. That is, they
raise their product prices by more than the percentage increase in variable costs (1 − A∗ ).
From the point of view of more productive Foreign ﬁrms, they have a bigger incentive to
raise prices because less competitors exist and they are weaker than before. The area (a) of
Figure 2.3 represents this situation.
By contrast, if c∗ ,E < 0, a depreciation of the Home currency gives Foreign exporters
X
tougher competition. As a result, more productive Foreign ﬁrms absorb a higher proportion
of an exchange rate shock to remain competitive in the Home market. The rate of pass∗
∗
through ξX (c) is incomplete (less than 1) and less than 1 − A∗ . When λ ,E > 0, ξX (c)
D
is always positive. The area (b) in Figure 2.3 represents this situation. When λ ,E < 0
D

(and so N increases),

∗
ξX (c)

can be negative. Foreign exporters must compete with more and

stronger competitors. Thus, some high productivity Foreign ﬁrms may lower their product
prices. This corresponds to the area (c) in Figure 2.3.
There are some interesting implications related to Proposition 10. First, when we introduce a tradable intermediate-input sector, the Foreign exporter’s variable cost increases
54

A∗

4/3
13/11
1
a

c

b

0

1

A

Figure 2.3: Exchange Rate Pass-Through and Export Quantity at the Firm Level (k =
3, ψ = 1, τ = τ ∗ = (1.5)1/3 )

by 1 − A∗ percent with 1 percent depreciation of the Home currency. This fact is consistent with the existing literature which suggests that pass-through is muted using imported
inputs in production.20 Second, we cannot guarantee that exchange rate pass-through to
ﬁrm-level import prices is incomplete contrary to Rodriguez-Lopez (2011). In his model,
Foreign exporters are always in tougher competition due to a Home currency depreciation.
On the contrary, Foreign exporters may face a diﬀerent competitive environment after an
exchange rate shock in the current model. For instance, it is possible that Foreign exporters
are in weaker competition at Home in spite of a Home currency depreciation shown as (a) in
Figure 2.3. This case happens because the opposite cost eﬀects between the two countries are
considerably large. Next, Rodriguez-Lopez shows that the ﬁrm productivity associated with
exchange rate pass-through is sensitive to the choice of the utility function.21 Furthermore,
the current model tells us that the change in the competitive environment driven by c∗ ,E is
X
20 See Ghosh (2009) for details.
21 He shows that more productive ﬁrms have higher pass-through rates with the translog

expenditure function. But he obtains the opposite result with the quasilinear-quadratic
utility function: lower pass-through rates for high productivity ﬁrms.

55

an important factor in determining the relationship between ﬁrm-speciﬁc productivity and
exchange pass-through.
The quantity that a Foreign ﬁrm with c ≤ c∗ sells in the Home market is given by
X
∗
yX (c) =

L ∗ W∗ ∗
τ E ∗ cX − c
2γ
A

(2.45)

The export sales in terms of the Home currency are given by
∗
p∗ (c)yX (c) =
X

L
W∗ 2
τ ∗E ∗
4γ
A

c∗
X

2

− c2

(2.46)

The impact of exchange rates on ﬁrm-level traded quantities may be relatively substantial compared with the exchange pass through eﬀect, but it does not always hold unlike
Rodriguez-Lopez (2011). The following proposition presents the results regarding exchange
rates and ﬁrm-level quantity adjustments.
Proposition 11
1. If c∗ ,E ≥ 0, the export quantity and export sales are increasing with respect to the
X
∗
∂ ln yX (c)
∗
exchange rate:
> ξX (c) > 0.
∂ ln E
2. If c∗ ,E < 0 and λ ,E > 0, some high productivity Foreign ﬁrms may increase their
D
X
export quantities and export sales while the other Foreign ﬁrms decrease their export
∗
∂ ln yX (c)
∗
quantities and export sales:
< ξX (c).
∂ ln E
3. If c∗ ,E < 0 and λ ,E ≤ 0, the export quantity and export sales are decreasing with
D
X
∗
∗
∂ ln yX (c)
∗ (c) and ∂ ln yX (c) < 0.
respect to the exchange rate:
< ξX
∂ ln E
∂ ln E
Each item 1, 2, 3 of Proposition 11 corresponds to each area (a), (b), (c) in Figure 2.3: (a)
When Foreign exporters face a weaker competitive environment at Home, they increase their
export quantities. (b) With less and stronger competitors, some more productive Foreign
exporters increase their export quantities, but the other Foreign ﬁrms decrease their export
56

quantities. (c) With more and stronger competitors, Foreign ﬁrms decrease their export
quantities.
Therefore, at the intensive margin, quantities adjust in a variety of ways after an exchange
rate shock. Roughly speaking, when the role of imported inputs in production is not large,
low levels of exchange rate pass-through coexist with relatively larger movements in trade
ﬂows at the ﬁrm level like Rodriguez-Lopez (2011). On the contrary, if the shares of imported
inputs in total costs are suﬃciently large, the impact of exchange rates on ﬁrm-level trade
ﬂows occurs diﬀerently at both the extensive and intensive margin.22

Aggregate Prices and Trade Flows
The aggregate import price is a weighted average of prices of imported goods. The market
share of a Foreign ﬁrm with c is given by

s∗ (c)
X

∗
yX (c)

y ∗ (c)
= X∗ =
YX
0

c∗
X ∗
yX (c)dc

=

L ∗ W∗ ∗
2γ τ E A∗ cX − c
L ∗ W∗ ∗ 2
4γ τ E A∗ cX

=

2(c∗ − c)
X
c∗
X

2

(2.47)

∗
where YX denotes the quantity of Foreign exports. Therefore, the aggregate import price is

∗
PX

=
0

c∗
X ∗
pX (c)s∗ (c)dc
X

W∗
2W
2
c
= τ ∗ E ∗ c∗ =
X
3
A
3A D

(2.48)

The following proposition presents the result regarding exchange rate pass-through to
the aggregate import price.
Proposition 12
The pass-through rate of exchange rate to the aggregate import price is
∗
∗ = ∂ ln PX =
ΞX
λD ,E .
∂ ln E
22 At the extensive margin, the cut-oﬀ levels and mass of entrants change. At the intensive

margin, the original or surviving ﬁrms adjust their export quantities.

57

A∗

A∗

4/3
21/17

Ξ∗ > 1
X
1

1
0 < Ξ∗ < 1
X

a
b

Ξ∗ < 0
X
d
0

1
(a)

0

A

Ξ∗
X

c

∗
∂ ln YX
(b)

1

A

∂ ln E

Figure 2.4: Exchange Rate Pass-Through and Export Quantity at the Aggregate Level
(k = 3, ψ = 1, τ = τ ∗ = (1.5)1/3 )

Figure 2.4 (a) illustrates how Ξ∗ is determined within A A∗ range. Let’s examine
X
the diﬀerence between Rodriguez-Lopez (2011) and the current model. In a basic model of
Rodriguez-Lopez, the exchange rate pass-through to the aggregate import price is always
negative. That is, a depreciation of the Home currency always lowers the aggregate import
price. This surprising result is the consequence of a composition bias due to changes in the
extensive margin of trade. With a Home-currency depreciation, the least productive Foreign
ﬁrms exit the Home market. The new aggregate import price is computed taking into account
only the surviving Foreign exporters, who are the most productive and so charged lower prices
before the exchange rate shock. Furthermore, these surviving Foreign ﬁrms must reduce
their markups in response to intensiﬁed competition at Home. In contrast, there are two
important diﬀerences in this paper. First, the change in competitive environment (related
to the extensive margin) due to an exchange rate shock is not unilateral. For example,
the least productive Foreign exporters may enter the Home market after a Home currency
depreciation due to relative cost advantage caused by lower imported input prices. Second,

58

as we see in section 2.3.2, the number of competing varieties in the Home market (N ) may
decrease in response to the depreciation of the Home currency. Then, original or surviving
Foreign exporters can raise their mark-ups or reduce their mark-ups by a smaller margin.
Taken together, various adjustments occur on not only the extensive but also the intensive
margin in my model.
The empirical evidence supports small but positive pass-through rates to aggregate import prices. To reconcile the theoretical model with the empirical evidence, Rodriguez-Lopez
(2011) develops an extended model that includes product quality.23 On the other hand, this
model tells us that exchange rate pass-through rates depend on shares of imported inputs
in total costs of the two countries. Based on the current model, the empirical evidence thus
suggests that a pair of shares of imported inputs is likely to be located within some middle
range.
Now let us look into the expenditure-switching eﬀect of exchange rates. We can calculate
the response of an aggregate export quantity to an exchange rate movement.
∗
∂ ln YX
= 1 − A∗ + 2 c∗ ,E = 1 + λ∗ ,E + c∗ ,E = λ ,E + c∗ ,E = Ξ∗ + c∗ ,E (2.49)
X
D
∂ ln E
X
X
X
X
X

In addition, the value of Foreign exports in terms of the Home currency is given by
∗
VX

=
0

c∗
X ∗
∗
pX (c)yX (c)dc

c∗
X

=
0

L
W∗ 2
τ ∗E ∗
4γ
A

c∗
X

2

− c2 dc =

L
W∗ 2 ∗ 3
cX
τ ∗E ∗
6γ
A
(2.50)

Therefore,
∗
∗
∂ ln VX
∂ ln YX
= 2 (1 − A∗ ) + 3 c∗ ,E = 2 λ ,E + c∗ ,E = Ξ∗ +
X
D
∂ ln E
∂ ln E
X
X

(2.51)

23 He assumes that higher productivity (or capability) is related to higher quality.

If
product quality is suﬃciently positively related to ﬁrm productivity, the unit price increases
with productivity while quality-adjusted price decreases with productivity. As a result, a
positive pass-through rate to the aggregate unit import price can be obtained.

59

The following proposition states our result regarding the expenditure switching eﬀect of
an exchange rate movement.
Proposition 13
1. If c∗ ,E ≥ 0, the quantity and value of Foreign exports are increasing with respect
X
to the exchange rate and their elasticities are greater than exchange rate pass-through
∗
∗
∂ ln VX
∂ ln YX
rate;
>
> Ξ∗ > 0
X
∂ ln E
∂ ln E
2. If − λ ,E < c∗ ,E < 0, the quantity and value of Foreign exports are increasing with
D
X
respect to the exchange rate. The exchange rate pass-through rate is between the two
∗
∗
∂ ln YX
∂ ln VX
exchange rate elasticities; 0 <
< Ξ∗ <
X
∂ ln E
∂ ln E
3. If c∗ ,E < − λ ,E < 0, the quantity of Foreign exports is decreasing with respect to
D
X
the exchange rate. The exchange rate elasticity of Foreign export value is positive or
∗
∂ ln YX
<
negative but smaller in absolute value than that of Foreign export quantity;
∂ ln E
∗
∗
∂ ln YX
∂ ln VX
0 < Ξ∗ and
<
X
∂ ln E
∂ ln E
4. If c∗ ,E < 0 and λ ,E ≤ 0, the quantity and value of Foreign exports are decreasing
D
X
with respect to the exchange rate and their elasticities are greater in absolute values
∗
∗
∂ ln VX
∂ ln YX
than exchange rate pass-through rate;
<
< Ξ∗ < 0
X
∂ ln E
∂ ln E
Figure 2.4 (b) presents a graphical illustration about the response of Foreign’s export
volume to an exchange rate movement. Each area (a), (b), (c), (d) corresponds to each item
1, 2, 3, 4 of Proposition 13. The last part of Proposition 13 is similar to the prediction
of Rodriguez-Lopez model, namely large and unambiguous expenditure-switching eﬀect of
exchange rate ﬂuctuations. However, the current model suggests there might be various
cases regarding the expenditure-switching eﬀect. For example, the trade ﬂows move in the
unexpected direction as described in the ﬁrst two parts of Proposition 13. If a Home-currency
depreciation causes a less competitive environment for Foreign exporters in the Home market,
both the quantity and value of Foreign exports can increase. Another possibility is that the
60

expenditure-switching eﬀect may be trivial or ambiguous like the third case of Proposition
13. One important implication of Proposition 13 is that a depreciation of the domestic
currency may lead to a negligible eﬀect or even a negative impact on the current account
considering global production sharing.

2.4

Conclusion

The issue of exchange rate pass-through has been extensively studied in international
economics, but relatively less attention has been paid to imported input prices in the literature. To examine this issue, I present a partial equilibrium model of production sharing
and exchange pass-through with monopolistic competition among heterogeneous ﬁrms, endogenous markups and sticky wages. Exchange rate ﬂuctuations has two distinct eﬀects.
First, a Home-currency depreciation improves directly the Home ﬁrms’ competitive position
through a rise in the production costs of Foreign ﬁnal goods, in terms of the Home currency.
Second, a depreciation causes an indirect eﬀect by increasing imported input prices, which
leads to a rise in the production costs of Home ﬁnal goods. Consequently, the domestic and
foreign ﬁrms’ competitive positions depend on the relative importance of these two eﬀects.
In the end, the degrees of exchange rate pass-through and expenditure switching depend on
adjustments in ﬁrms’ prices and quantities due to altered competitive positions caused by
an exchange rate shock.
With production sharing involving two trading partners, the model derives some interesting results. Both exchange rate pass-through and expenditure-switching eﬀect are ambiguous
unlike the prediction of a standard model built by Rodriguez-Lopez (2011). While the passthrough to the aggregate import price is always negative in his original model, a low but
positive rate of pass-through can be derived assuming the shares of imported inputs in total
costs are located within some range. In addition, a depreciation of the domestic currency
may lead to a negligible eﬀect or even a negative impact on the current account contrary to
popular belief.
61

The results of this paper also provide a potential explanation for the fact that the degree
of pass-through varies across countries and industries. This is because the ratio of imported
inputs to total costs diﬀers signiﬁcantly by country and industry. Future research might
explore the implications of these ﬁndings for exchange rate pass-through patterns across
countries with diﬀerent industry mixes and thus verify roles of imported inputs in production.

62

APPENDIX

63

Appendix
Proof of Proposition 7.

pD =
¯

cD

1W
kck−1
(cD + c)
dc
2A
(cD )k

0

(A.1)

cD
1W
cD ck−1 + ck dc
k (cD )−k
2A
0
1W
2k + 1
=
k (cD )−k
(c )k+1
2A
k(k + 1) D

=

k+1
2
k+1

=

p∗
¯X

c∗
X

=
0

W
c
A D

1 ∗ W∗ ∗
kck−1
τ E ∗ cX + c
dc
k
2
A
c∗
X

=

k+1
2
k+1

τ ∗E

W∗ ∗
c
A∗ X

=

k+1
2
k+1

τ ∗E

W∗
A∗ W
·c
A∗ Aτ ∗ EW ∗ D

=

k+1
2
k+1

(A.2)

W
c
A D

Analogously, p∗ = p∗ = pX =
¯
¯D ¯

1
k+2
k+1

W∗ ∗
c .
A∗ D

Proof of Corollary 1.
From equation (2.5), z(c) =

1
γ
(α − p) − (p(c) − p) for c ≤ cD ,
¯
¯
γ ηN + γ
cD
0

cD
0

z(c)2 dc =

z(c)dc =

N
(α − p)
¯
ηN + γ

2
cD
N
1
(α − p)2 + 2
¯
(p(c) − p)2 dc
¯
ηN + γ
γ 0

64

(A.3)

(A.4)

cD

p(c)z(c)dc =

0

1 cD
N
(α − p)¯ −
¯p
(p(c) − p)2 dc
¯
ηN + γ
γ 0

(A.5)

Substituting (2.3), (A.3), (A.4) and (A.5) into (2.2),
1 N
1 N 1 cD
(α − p)2 +
¯
(p(c) − p)2 dc
¯
2 ηN + γ
2γ N 0


Wc
1 α − A D 
1
k
= W+
(α − p)2 +
¯
2η
α−p
¯
2η 2(k + 1)(k + 2)

U = W+

= W+

1
2η

W
c
A D

α−

(A.6)
α−

W
c
A D

W
c
A D

k+1W
c
k+2 A D

α−

cD
1 cD
(p(c) − p)2 dc =
¯
(p(c) − p)2 g(c|c ≤ cD )dc.
¯
N 0
0
Analogously, we can derive the indirect utility function of Foreign country.

where we use

Proof of Corollary 2.
Using (2.18) and (2.19), we can re-write equation (2.31) in a diﬀerent way

Ne =

2(k + 1)γ (cM )k
η (τ τ ∗ )k − 1

c∗ k
D
cX

1
c∗
X

Ne ≥ 0 ⇐⇒

k

1
A∗
1
α−1 −
k
W ∗ c∗ τ ∗ E
c∗
X
D

c∗ k
D
·
c∗
X

A∗
1 α−1
W ∗ c∗ τ ∗ E
X
A∗ α − 1
W ∗ c∗
D

A∗
α−1
W ∗ c∗
D
(A.7)

≥1

1
which is incompatible with c∗ < c∗ since ∗ ≤ 1. Therefore, c∗ ≥ c∗ .
D
X
D
X
τ E
We also re-write equation (2.32) as the following expression

∗
Ne

=

2(k + 1)γ (cM )k
η

(τ τ ∗ )k

−1

1
cD k
∗
cX
(cX )k

∗
Ne

≥ 0 ⇐⇒

A E
1
α−1 −
W cX τ
(cD )k

A E
cD k W cX τ α − 1
·
≥1
A α−1
cX
Wc
D

65

A
α−1
W cD

(A.8)

which is incompatible with cD < cX since

E
≤ 1. Therefore, cD ≥ cX .
τ

Proof of Proposition 8.
From equation (2.38), for c ,E to be negative
D

A∗ <

2(τ k − ψ)
−1
kψ

A

+1+

1
k

(A.9)

1
2(τ k − ψ)
−1 ≥ − , this condition is satisﬁed for ∀ A ∈ (0, 1) and ∀ A∗ ∈ (0, 1).
kψ
k
k − ψ)
2(τ
1
k+1
We can check that
− 1 ≥ − ⇐⇒ τ k ≥
ψ.
kψ
k
2
Now let’s prove part 2. From equation (2.26),

However, if

c∗ ,E
D

1

=
(k

+ 2){(τ ∗ )k ψ

− 1}

{2[(τ ∗ )k ψ − 1] − k} A∗ − k A + k + 1

(A.10)

For c∗ ,E to be positive
D
{2[(τ ∗ )k ψ − 1] − k} A∗ − k A + k + 1 > 0
First, if 2[(τ ∗ )k ψ − 1] − k ≥ 0 ⇐⇒ (τ ∗ )k ≥

k+2
2

(A.11)

1
, the condition above is obviously
ψ

satisﬁed.
Now suppose that 2[(τ ∗ )k ψ − 1] − k < 0.
Because A ∈ (0, 1) and A∗ ∈ (0, 1), {2[(τ ∗ )k ψ −1]−k} A∗ −k A +k +1 > 2 (τ ∗ )k ψ −k −1.
k+1 1
Therefore, if 2 (τ ∗ )k ψ − k − 1 ≥ 0 ⇐⇒ (τ ∗ )k ≥
, (A.11) is always satisﬁed.
2
ψ
Proof of Proposition 9.
If we derive λ ,E and λ∗ ,E , the result then follows by determining their signs.
D
D

λD ,E

=

1
kτ k A + kψ A∗ − ψ(k + 1)
(k + 2)(τ k − ψ)

66

(A.12)

λ∗ ,E
D

1

=
(k

+ 2){(τ ∗ )k ψ

− 1}

−kψ (τ ∗ )k A∗ − k A + k + 1

(A.13)

Proof of Proposition 10.
c∗
Suppose c∗ ,E > 0. ∗ X · c∗ ,E is a positive value and approaches to c∗ ,E as c → 0.
cX + c
X
X
X
∗ (c) < 1−
Hence, 1− A∗ < ξX
A∗ + c∗ ,E . In addition, 1− A∗ + c∗ ,E = 1+ λ∗ ,E = λ ,E
X

X

by deﬁnition. Therefore, 1 − A∗ <

∗
ξX (c)

X

D

< λ ,E . Actually, c∗ ,E > 0 ⇐⇒ λ ,E >
D
D
X

1 − A∗ .
To show that more productive Foreign ﬁrms have higher pass-through rates,
∗
∂ξX (c)
c∗
X
=−
· c∗ ,E < 0
∗ +c 2
∂c
X
cX

(A.14)

Analogously, we can prove the second part when we assume c∗ ,E ≤ 0
X
Proof of Proposition 11.
From equation (2.45),
∗
∂ ln yX (c)
c∗
= 1 − A∗ + ∗ X · c∗ ,E
∂ ln E
cX − c
X

c∗
Note ∗ X ∈ (1, ∞).
cX − c

(A.15)

∗
∂ ln yX (c)
First, if c∗ ,E ≥ 0,
≥ 1 − A∗ + c∗ ,E = λ ,E > 0.
D
∂ ln E
X
X
∗
∂ ln yX (c)
< λ ,E . Hence, if we add the condition
Next, suppose that c∗ ,E < 0. Then,
D
∂ ln E
X
∗ (c)
∂ ln yX
is always negative. With λ ,E > 0, some high productivity Foreign
λD ,E ≤ 0,
D
∂ ln E
∗
∂ ln yX (c)
ﬁrms may increase their export quantities since
is decreasing in c and its upper
∂ ln E
bound is positive.

67

Now let’s examine the relationship between exchange rates and ﬁrm-level export sales.
2

∗
∂ ln[p∗ yX (c)]
c∗
X
X
= 2 1 − A∗ +
· c∗ ,E
∗ 2 − c2
∂ ln E
X
cX

Therefore, we can apply the same argument to the value of exports.

Proof of Proposition 12.
∗
∂ ln PX
2
∗
From equation (2.48), PX = λD . Hence, it is obvious that Ξ∗ =
= λ ,E .
X
D
3
∂ ln E

Proof of Proposition 13.
Given the equations (2.49) and (2.51), all the results are obtained easily.

68

(A.16)

REFERENCES

69

REFERENCES

[1] Ahmed, S., 2009. Are Chinese exports sensitive to changes in the exchange rate?. International Finance Discussion Papers No. 987, Board of Governors of the Federal Reserve
System
[2] Akerlof, G.A., 2007. The missing motivation in macroeconomics. American Economic
Review 97(1), 5-36.
[3] Athukorala, P., Menon, J., 1994. Pricing to market behaviour and exchange rate passthrough in Japanese exports. Economic Journal 104(423), 271-281.
[4] Aw, B.Y., Chung, S., Roberts, M.J., 2000. Productivity and turnover in the export
market: Micro-level evidence from the Republic of Korea and Taiwan (China). World
Bank Economic Review 14(1), 65-90.
[5] Bas, M., 2009. Trade, foreign inputs and ﬁrms’ decisions: Theory and evidence. CEPII
Working Paper No. 2009-35.
[6] Bernard, A.B., Jensen, J.B., 1999. Exceptional exporter performance: Cause, eﬀect or
both?. Journal of International Economics 47, 1-25.
[7] Berman, N., Martin, P., Mayer, T., 2012. How do diﬀerent exporters react to exchange
rate changes?. The Quarterly Journal of Economics 127, 437-492.
[8] Campa, J.M., Goldberg, L.S., 2005. Exchange rate pass-through into import prices. The
Review of Economics and Statistics 87(4), 679-690.
[9] Campa, J.M., Goldberg, L.S., 2006. Distribution margins, imported inputs, and the
sensitivity of the CPI to exchange rates. NBER Working Paper No. 12121.
[10] Clerides, S.K., Lach, S., Tybout, J.R., 1998. Is learning by exporting important? Microdynamic evidence from Colombia, Mexico, and Morocco. The Quarterly Journal of Economics 113(3), 903-947.
[11] Dong, W., 2012. The role of expenditure switching in the global imbalance adjustment.
Journal of International Economics 86, 237-251.
[12] Engel, C., 2002. Expenditure switching and exchange rate policy. NBER Working Paper
No. 9016.
[13] Ghosh, A., 2009. Implications of production sharing on exchange rate pass-through.
International Journal of Finance and Economics 14, 334-345.
[14] Goldberg, P.K., Knetter, M.M., 1997. Good prices and exchange rates: What have we
learned?. Journal of Economic Literature 35, 1243-1272.

70

[15] Greenaway, D., Kneller, R., Zhang, X., 2010. The eﬀect of exchange rates on ﬁrm
exports: The role of imported intermediate inputs. The World Economy 33(8), 961-986.
[16] Hallak, J.C., 2006. Product quality and the direction of trade. Journal of International
Economics 68, 238-265.
[17] Hummels, D., Ishii, J., Yi, K., 2001. The nature and growth of vertical specialization
in world trade. Journal of International Economics 54, 75-96.
[18] Hummels, D., Klenow, P.J., 2005. The variety and quality of a nation’s exports. American Economic Review 95(3), 704-723.
[19] Krugman, P., 1986. Pricing to market when the exchange rate changes. NBER Working
Paper No. 1926.
[20] Koopman, R., Wang, Z., Wei, S., 2008. How much of Chinese exports is really made
in China? Assessing domestic value-added when processing trade is pervasive. NBER
Working Paper No. 14109.
[21] Kugler, M., Verhoogen, E., 2012. Prices, plant size, and product quality. Review of
Economic Studies 79, 307-339.
[22] Liao, W., Shi, K., Zhang, Z., 2010. Vertical trade and China’s export dynamics. Hong
Kong Institute for Monetary Research Working Paper No. 10/2010.
[23] Melitz, M.J., 2003. The impact of trade on intra-industry reallocations and aggregate
industry productivity. Econometrica 71(6), 1695-1725.
[24] Melitz, M.J., Ottaviano, G.I.P., 2008. Market size, trade, and productivity. Review of
Economic Studies 75, 295-316.
[25] Nucci, F., Pozzolo, A.F., 2001. Investment and the exchange rate: An analysis with
ﬁrm-level panel data. European Economic Review 45, 259-283.
[26] Rodriguez-Lopez, J.A., 2011. Prices and exchange rates: A theory of disconnect. Review
of Economic Studies 78, 1135-1177.
[27] Tille, C., 2005. Financial integration and the wealth eﬀect of exchange rate ﬂuctuations.
Staﬀ Report No. 226, Federal Reserve Bank of New York.
[28] Verhoogen, E.A., 2008. Trade, quality upgrading, and wage inequality in the Mexican
manufacturing sector. The Quarterly Journal of Economics 123(2), 489-530.
[29] Wauchy, X., 1996. Quality choice in models of vertical diﬀerentiation. The Journal of
Industrial Economics 44(3), 345-353.
[30] Yang, J., 1997. Exchange rate pass-through in U.S. manufacturing industries. The Review of Economics and Statistics 79(1), 95-104.
[31] Yang, J., 1998. Pricing-to-market in U.S. imports and exports: A time series and crosssectional study. The Quarterly Review of Economics and Finance 38(4), 843-861.

71

CHAPTER 3
Exchange Rate Pass-Through in Korean
Manufacturing Industries
3.1

Introduction

The exchange rate pass-through has substantial economic eﬀects. From a macroeconomic
perspective, the extent of exchange rate pass-through has an important implication for monetary policy as it aﬀects domestic inﬂation and foreign transactions. From a microeconomic
perspective, the degree of pass-through helps us examine the response of ﬁrms’ export and
price setting decisions to exchange rate shocks. Therefore, the empirical literature on the
exchange rate pass-through is extensive.
Current trade literature has highlighted the role of both the intensive margin (markup
adjustments) and extensive margin (entry and exit decision) in explaining ﬁrm export behaviour. However, most empirical works on exchange rate pass-through have focused mainly
on the intensive margin and have paid little attention to the extensive margin.1 That is, most
articles have concentrated on the heterogeneity in the responses to exchange rate shocks by
exporters and hence they do not model entry and exit into exporting (Berman et al. (2012),
Amiti et al. (2012)2 , Fauceglia et al. (2012)).
1 Many ﬁrm-level studies condition their analysis on exporting ﬁrms’ pricing strategies

related to exchange rate shocks (Amiti et al. (2012), Berman et al. (2012), Li et al. (2012)).
On the other hand, the extensive margin has been vigorously examined in investigating the
relationship between exchange rates and trade ﬂows. See Berman et al. (2012), Colacelli
(2010), Tang and Zhang (2012) for example.
2 Instead, they rather focus on the import decisions of the ﬁrms. In equilibrium, the more
productive ﬁrms choose to source a greater share of their inputs internationally with ﬁxed
costs of importing.

72

In addition, a single bilateral exchange rate does not move around separately as Hummels
et al. (2010) point out. Other things being equal, a depreciation of the Korean won against
the US dollar would make dollar-denominated Korean export products more attractive in
the importing country. But if other currencies are also depreciating against the US dollar,
then the net eﬀect on prices of Korean goods sold abroad will also depend on the extent of
exchange rate pass-through by competing countries. Therefore, we need to pay attention to
shocks on multiple exchange rates in estimating pass-through.3
In this paper, I attempt to empirically analyze the role that imported inputs and adjustments in the extensive margin play in exchange rate pass-through at the industry level
using disaggregated Korean data. I develop further a quadratic utility model introduced in
the second chapter to guide the empirical analysis. I ﬁnd that more import-intensive industries in South Korea (hereafter simply “Korea”) have higher exchange rate pass-through
into their export prices. This result cannot be explained by previous ﬁrm-level studies, but
it is consistent with my theoretical model including the extensive margin of trade. This is
because aggregate export prices are aﬀected not only by changes in ﬁrms’ marginal costs, but
also by variations in the composition of exporters due to changes in the exporting cut-oﬀ.
For example, suppose that the exporting cost cut-oﬀ declines owing to a competitive disadvantage caused by higher imported input prices following a depreciation of the Korean won.
The new aggregate export price is then computed taking into account only the surviving
Korean ﬁrms who are the most productive and had lower prices before the exchange rate
shock. Next, I show that the relative value of the destination market currency should be
also considered when estimating exchange rate pass-through regardless of an astonishingly
high proportion of dollar invoicing of Korean exports and imports. Finally, I ﬁnd that passthrough is increasing both in Korea’s share in total import of the destination market and in
3 Hummels et al. (2010) document that a depreciation of currency of the competing

country shifts the residual demand curve inward and reduces the exchange rate pass-through
elasticity. They also suggest that cross-currency exchange rate shocks become more crucial
when an exporting country is relatively small.

73

Korea’s comparative advantage industries.
This paper is related to two strands of recent literature. First, it relates to the existing
literature emphasizing the importance of ﬁrm heterogeneity and the extensive margin of
trade. Bernard et al. (2009) ﬁnd that short-run changes in US exports are largely accounted
for by the intensive margin. By contrast, the extensive margins, which are decomposed into
ﬁrm entry and exit and continuing ﬁrms’ adding and dropping of country-products, play an
important role in explaining long-run changes in US exports.4 Using a bilateral trade sample
of 136 countries, Colacelli (2010) shows that the extensive margin of trade plays a signiﬁcant
role in export adjustments response to real exchange rate ﬂuctuations at the yearly frequency.
Berman et al. (2012) ﬁnd that a 10% real depreciation increases the exporting probability
by around 1.8 percentage point from a French ﬁrm-level data set. Tang and Zhang (2012)
ﬁnd that a 10% real appreciation of the renminbi is associated with a 1 percentage point
decline in the probability of entry, and a 0.2 percentage point increase in the probability of
exit using Chinese ﬁrm-level data.
Second, this paper is related to the recent empirical work on the role of imported inputs in
exchange rate adjustments of exports. The rationale for studying this area is that a currency
depreciation not just lowers the foreign currency prices of exports, but also raises the prices
of imported inputs. Campa and Goldberg (2010) document that integrated production
(through cost variation from imported input use in goods production) has become large
enough to dominate the direct consumption of imported goods as the channel for transmission
of exchange rates into the CPI. Berman et al. (2012) show that ﬁrms that are more dependent
on imports increase their export price expressed in home currency more than others (that
is, lower exchange rate pass-through) because they see their input costs rise when the euro
depreciates. Using detailed Belgium micro data, Amiti et al. (2012) ﬁnd that more importintensive exporters have signiﬁcantly lower exchange rate pass-through into their export
4 This evidence is consistent with interpretation by Eaton et al. (2008). Conditional on

survival, entering exporters and recently added product-countries grow more rapidly than
incumbent exporters and product-countries.

74

prices. They also provide a second channel that limits the eﬀect of exchange rate shocks
on export prices. This channel is based on the fact that import intensive ﬁrms have higher
market shares and hence actively move their markups in response to changes in marginal
costs. On the other hand, Fauceglia et al. (2012) document that Swiss exporters optimally
choose to absorb changes of the imported input prices in their markups and hence imported
input price changes do not signiﬁcantly change exchange rate pass-through behavior.
Korea is a small open economy that depends highly on foreign trade5 and has a relatively high share of imported inputs intro production.6 Thus, estimating the size of exchange
rate pass-through has received considerable attention in empirical studies for Korea. Before
turning to my study, I brieﬂy review the empirical literature. Athukorala (1991) ﬁnds an
average pass-through into foreign prices for the nominal eﬀective exchange rate to be around
28 per cent. Yang and Hwang (1994) estimate pass-through from real sectoral exchange
rate shocks into Korean export prices in six manufacturing sectors. They have documented
that Korean exporters absorb 70% of a given exchange rate change in their margin and pass
through the remaining 30%. Lee (1995) estimates the response of Korean manufacturedexport prices to nominal eﬀective exchange rate changes for 16 industries. He shows that
pricing-to-market behaviour is prevalent in Korean export industries and explains this result
by the market power asymmetry between home and export markets. Lee (1997) estimates
exchange rate pass-through of industry speciﬁc real exchange rate changes into Korean import prices. The average pass-through elasticity for all manufacturing imports was 0.62 and
market concentration systematically reduced the pass-through. Choi and Kim (2001)7 and
Kim and Lee (2009)8 show that Korean export prices have become less responsive to the
5 The proportion of Korea’s trade to its gross national income (GNI) stood at 112.7 percent

in 2012.
6 See Section 3.2.2 for details.
7 The pass-through elasticities before the ﬁnancial crisis (1981:1Q ∼ 1997:3Q) and in the
whole sample period (1981:1Q ∼ 2000:4Q) were 0.61 and 0.39, respectively.
8 The long run elasticities of pass-through before and after the ﬁnancial crisis were 0.61
and 0.24, respectively.

75

won/dollar exchange rate since the ﬁnancial crisis in 1997. Decline in prices of major export
products and increased competition with China have been suggested as the main reasons for
the decline in exchange rate pass-through. Cho (2010) estimates exchange rate pass-through
into export prices for 13 manufacturing industries. He suggests that industries with smaller
destination market share or more competitive in the foreign market have lower exchange rate
pass-through.

3.2
3.2.1

Background
Exchange Rate and Export Price Movements

As Pollard and Coughlin (2006) pointed out, pass-through estimates are sensitive to the
exchange rate index. Therefore, it is important to choose a proper exchange rate index to
examine the relationship between exchange rate and export price. Empirical pass-through
studies on Korea can be divided into two groups. The ﬁrst group usually uses bilateral
exchange rate, most notably the Korean won/US dollar exchange rate (Yang and Hwang
(1994), Choi and Kim (2001), Kim and Lee (2009), Cho (2010)). The second group of
studies uses a composite exchange rate which are constructed according to trade weights
(Athukorala (1991), Lee (1995, 1997)). Because Korea uses the US dollar for more than 80%
of trade transactions9 , the nominal won/dollar exchange rate is thought be the preferred
index in examining exchange rate pass-through.
Figure 3.1 depicts the evolution of the nominal won/dollar exchange rate and the export
price10 in dollar terms over the period 1990-2012. With a few exceptions, there has been
a negative correlation between the won/dollar exchange rate and the export price index.
9 Appendix Table A1 shows the dollar share in the invoicing of exports and imports.

As shown in Table A1, the extent to which the US dollar is used in trade invoicing varies
substantially across countries.
10 The Export Price Index is based, as a principle, on the FOB (free on board) price, that
is the price at the time of shipping oﬀ the export goods from Korea. The index is computed
on the won basis, the dollar basis and the contract currency basis.
76

Between 1996 and 1998, the Korean won depreciated greatly against the US dollar and the
export price index (in dollar terms) declined. Over the period 2002-2005, the export price
index increased gradually along with the appreciation of the Korean won.

Data Source: The Bank of Korea
Figure 3.1: Exchange Rate and Export Price Index

3.2.2

Imported Inputs into Production

The shares of imported inputs are calculated from the Input-Output Table. Deﬁne a
vector of share of import content in ﬁnal demand for domestically produced products by
IM .11
IM = uAM [I − AD ]−1

(3.1)

where u is a 1 × n vector of 1’s, AM is the n × n imported coeﬃcient matrix, I is the identity
matrix, AD is the n × n domestic coeﬃcient matrix and n is the number of sectors.
For industry s, IMs is the column sum of the coeﬃcient matrix for total intermediate
11 See Koopman et al. (2012) for the derivation of equation (3.1).

77

import requirement. We can know that the concept of vertical specialization by Hummels
et al. (2001) and that of import content in total exports are identical.12
Since 2000, the Korean Input-Output tables have been compiled with 404, 168, 77 and
28 industrial sectors in basic, small, medium and large classiﬁcations, respectively. I use 168
sector classiﬁcation as a benchmark to measure shares of imported inputs into production.
This classiﬁcation is not only appropriately disaggregated but also consistent with sectoral
categories of other independent variables such as producer price index. Again, I focus on
100 manufacturing industries out of 168 sectors.
I have calculated the share of imported inputs by manufacturing industry using the
Korean Input-Output tables for 1990, 1995, 2000, 2005, 2010. These shares are reported in
Appendix Table A2. As shown in Table A2, imported input shares of given the industry
tend to increase but are relatively stable over time while there is considerable variation across
industries at a speciﬁc time.13
We see the distribution of imported input shares in production among 100 Korean manufacturing industries in Table 3.1 (as of 2010). For the majority of industries, the share of
imported inputs in production ranges between 30% and 60%. At the same time, nearly 80%
of (manufacturing) export sales are generated by the industries belonging to that interval.
The export-value-weighted median of imported input shares is 41%. The industry with the
highest share is naphtha (88%) while cereal husking has the lowest share (11%).
In addition, we need to notice that the dispersion of imported inputs into production also
diﬀers signiﬁcantly by country (Campa and Goldberg (2010)). Appendix Table A3 reports
the import content of exports for manufacturing industries across countries. In general,
larger countries (Brazil, Russia, United States) have a lower share of imported inputs while
smaller countries (Belgium, Hungary, Ireland, Singapore, Taiwan) have a higher share. Ko12 Vertical specialization share of total exports for country k is calculated by V SX =
k

uAM [I − AD ]−1 X/Xk , where X is an n × 1 vector of exports and Xk is total country
export. The result is also called the import content of exports.
13 Over the period 1990-2010, between standard deviation is 0.14 while within standard
deviation is 0.06.

78

rea has a relatively high share of 42% and is expected to be more sensitive to cost shocks
due to exchange rate movements.

Table 3.1: Distribution of Imported Input Shares in 2010
# Industries

3.3

Fraction of Export Value

≤ 0.3
≤ 0.4
≤ 0.5
≤ 0.6
≤ 0.7
> 0.7

10
38
16
20
7
9

10.0%
38.0%
16.0%
20.0%
7.0%
9.0%

0.9%
30.4%
31.5%
17.5%
13.4%
6.3%

Total

0 < IMs
0.3 < IMs
0.4 < IMs
0.5 < IMs
0.6 < IMs
IMs

Fraction of Industries

100

100.0%

100.0%

Theoretical Framework

In this section, I construct a theoretical framework linking an industry’s exchange rate
pass-through to its import intensity. In order to do that, I extend the two-country model
introduced in the second chapter (Production Sharing and Exchange Rate Pass-Through)
to three-country model which would be more appropriate for empirical analysis. That is, I
divide the global economy into three categories; Korea (K), an export destination country
(D) and other competing countries (R).
International trade transactions can be invoiced in the producer currency, in the destination currency, or in a third currency, that is, vehicle currency. Both the US dollar and
the euro are widely used for invoicing and settling international trade around the world. In
particular, the U.S. dollar has been the dominant vehicle currency in Korea. Consequently,
most empirical research articles focus on the Korean won-U.S. dollar nominal exchange rates
in analyzing exchange rate pass-through into export price. For simplicity, I suppose that
the US dollar is the only currency in the invoicing and payment of international trade. In
particular, this assumption plays a crucial role in examining the impact of imported inputs
79

on marginal costs in response to exchange rate shocks. It seems to be an oversimpliﬁcation
of the real world but would be a benchmark which provides necessary insights to help us
understand the mechanism. More will be said on this assumption later.

3.3.1

Demand and Production

Each economy has identical consumers denoted by LK , LD and LR , respectively. These
consumers are assumed to share the quasilinear-quadratic utility function proposed by Melitz
and Ottaviano (2008). Let the source country be indexed by i and the destination country by
j for i, j ∈ {K, D, R}. Also let EiU (EjU ) be the exchange rate deﬁned as units of country
i (j) currency per unit of the US dollar for i, j ∈ {K, D, R}. The quantity demanded for a
variety imported from country i and consumed by country j is

yij (ω) =

Lj
pj − EjU · pU (ω)
ij
γ

(3.2)

where pj denotes the price-ceiling for all varieties sold in country j and pU represents the
ij
US dollar-denominated export price of country i to country j. The parameter γ stands for
the degree of product diﬀerentiation among the varieties.
I adopt the same assumptions for production process as those used in the second chapter.
1
W
Then, i c represents the eﬀective marginal cost for a ﬁrm with productivity located in
Ai
c
country i, where Wi is the nominal wage rate in country i and Ai denotes the coeﬃcient of
productivity growth due to imported inputs.
To be more speciﬁc, Ai is given by

Ai =





1+

Wi
U
EiU · W−i

σi −1

σ −1

φi i

 1
 σi −1

(3.3)



U
where W−i indexes the average US dollar-denominated wage rate of the rest of the world

except country i, σi indexes the elasticity of substitution between domestic and imported

80

inputs in country i and φi measures the relative eﬃciency of imported inputs in country i.
For simplicity, assume that per-unit trade costs are identical across countries (τK = τD =
τR = τ ). Then, we can obtain the following price and proﬁt functions in US dollar terms for
i = j.
pU (c) =
ij
U
πij (c)

3.3.2

1
2

Lj
=
E
4γ jU

pj
τ
W
+
· ic
EjU
EiU Ai

(3.4)

pj
τ
Wi 2
c
−
·
EjU
EiU Ai

(3.5)

Cut-oﬀ Levels

We deﬁne the cut-oﬀ rules as cij = sup{c : πij (c) > 0} for i, j ∈ {K, D, R}.14 For
W
notational convenience, let wi = i . Then, we derive the following relationships between
Ai
cut-oﬀ levels.
τ wK EDU
wK EDU
wK EDU
·
·
·
cKD , cDK =
cKK , cDR =
wD EKU
τ wD EKU
wD EKU
wK ERU
wK ERU
τ wK ERU
·
·
·
cRD =
cKD , cRK =
cKK , cRR =
wR EKU
τ wR EKU
wR EKU

cDD =

cKR
cKR
(3.6)

In addition, I assume that the sunk entry costs are identical across countries (fK = fD =
fR = f ) for simplicity. Free entry conditions for all countries are given by
cKK
0

cKD

πKK (c)dG(c) + EKU

0

U
πKD (c)dG(c) + EKU

cKR
0

U
πKR (c)dG(c) = WK f

(3.7)
cDD
0
cRR
0

πDD (c)dG(c)+EDU
πRR (c)dG(c) + ERU

cDK
0
cRK
0

U
πDK (c)dG(c)+EDU
U
πRK (c)dG(c) + ERU

14 Note that π (c) = E · π U (c)
ij
iU
ij

81

cDR
0
cRD
0

U
πDR (c)dG(c) = WD f (3.8)
U
πRD (c)dG(c) = WR f (3.9)

Using the Pareto parametrization for the cost draws (c)15 and equations (3.6)-(3.9), we
can solve the equilibrium cut-oﬀ levels for Korean ﬁrms.

cKK =

ϕWK
LK (wK )

2

·

τ k {(τ k

+ 1) − ψD − ψR }
+ 2)(τ k − 1)

1
k+2

(3.10)

(τ k

1

E
(τ k + 1)ψD − ψR − 1 k+2
· KU ·
cKD =
(τ k + 2)(τ k − 1)
LD (τ wK )2 EDU
ϕWK

cKR =

ϕWK
2

LR (τ wK )

·

(τ k

+ 1)ψR − ψD − 1
(τ k + 2)(τ k − 1)

EKU
·
ERU

(3.11)

1
k+2

(3.12)

EKU k+1
and
EDU
WR wR k EKU k+1
E
E
ψR =
. Because KU = EKD and KU = EKR , ψD and ψR
WK wK
ERU
EDU
ERU
measure relative competitiveness of Korean ﬁrms against the destination country D and the
where ϕ = 2γf (k + 1)(k + 2)ck ,
M

ψD =

WD
WK

wD k
wK

rest of the world R, respectively.16

3.3.3

Aggregate Export Prices and Exchange Rates

The aggregate export price set in the US dollar from Korea to the destination D is
U
PKD =

0

cKD
pU (c)sKD (c)dc
KD

=

2 τ
w c
3 EKD K KD

15 The cumulative distribution function of costs is given by G(c) =

(3.13)
c

k

, where k > 1
cM
indexes the dispersion of cost draws. As k increases, the ratio of high-cost ﬁrms increases
and the cost distribution is more clustered near the upper bound.
16 To obtain positive cut-oﬀs, both ψ and ψ must lie within some bounds. Graphical
D
R
illustration shows us the condition is stricter than that of two-country model. That is,
1
< ψD (ψR ) < τ k cannot guarantee the existence of positive equilibrium cut-oﬀs.
k
τ

82

where sKD (c) denotes the market share of a Korean ﬁrm with cost c in total exports from
Korea to country D.17
From equation (3.3), d ln Ai = − i · d ln EiU , where i is the imported input share in total
costs for ﬁrms located in country i. Taking logarithms and diﬀerentiating totally of both
sides of equation (3.13) yield the following relationship.
U
d ln PKD = −d ln EKU − d ln AK + d ln cKD = −(1 − K )d ln EKU + d ln cKD

(3.14)

As shown in equation (3.11), cKD is the function of three bilateral exchange rates EKU ,
EDU and ERU . As a result, the aggregate Korean export price is aﬀected by these various
exchange rate shocks.
Let’s obtain d ln cKD from equation (3.11) and substitute it into equation (3.14), then
U
d ln PKD = −(1 − K )d ln EKU

1
 +
k


 (k + 2){(τ k + 1)ψ − ψ − 1)} {(τ + 1)ψD − ψR − 1)}×

D
R



 {(1 − 2 )d ln E
K
KU − d ln EDU }
d ln cKD −→


 +(τ k + 1)ψD {(k + 1 − k K ) · d ln EKU − (k + 1 − k D ) · d ln EDU }





 −ψ {(k + 1 − k ) · d ln E
R
K
KU − (k + 1 − k R ) · d ln ERU }

=

1
(k + 1 − k K ) · d ln EKU .
(k + 2){(τ k + 1)ψD − ψR − 1}
− {(τ k + 1)ψD − ψR − 1 + (k + 1 − k D )} · d ln EDU
− (k + 1 − k R ) · d ln ERU

(3.15)

Equation (3.15) carries several implications. First, we need to be cautious about empirical
work on exchange rate pass-through because export prices in principle can be aﬀected by
movements of all exchange rates. The equation can be rewritten in the following simple
17 For derivation, see the subsection ‘Aggregate Prices and Trade Flows’ of the second

chapter.

83

form.
U
d ln PKD
d ln ERU
d ln EDU
= βK + βD ·
+ βR ·
d ln EKU
(+)
(−) d ln EKU
(−) d ln EKU

(3.16)

Consider ﬁrst two other bilateral exchange rates have negative correlation with the won/dollar
d ln EDU
d ln ERU
exchange rate
< 0 and
< 0 . For instance, suppose that the Korean
d ln EKU
d ln EKU
won depreciates against the US dollar in spite of global weakness of the US dollar. Then,
Korea’s export price at the aggregate level would rise because new Korean entrants have
lower productivity and charge higher prices than existing Korean exporters. If so, traditional estimation (using a single bilateral exchange rate) may ﬁnd a pass-through estimate
whose sign is opposite to that based on existing theoretical frameworks.
To take an extreme example, consider two other exchange rates have perfect positive
d ln ERU
d ln EDU
=
= 1 . It is reasonable
correlation with the won/dollar exchange rate
d ln EKU
d ln EKU
to assume D > R when we take two economies’ sizes into consideration. Then, we can see
that the aggregate Korean export price in dollar terms declines in response to a depreciation
of the Korean won against the US dollar.18
Figure 3.2 depicts movements in the Korean won/US dollar exchange rates and nominal
eﬀective exchange rates of the US dollar.19 Two series display considerable co-movement and
the correlation coeﬃcient is 0.692, which is signiﬁcant at the 1% level. From equation (3.16),
if exchange rates of the US dollar against the Korean won and against other currencies have a
strong positive correlation, we can expect the same sign of pass-through as that of estimates
obtained from traditional methods. That is, Korea’s dollar-nominated export prices decrease
along with a depreciation of the won against the US dollar.

18
U
d ln PKD
d ln EKU

= −

(k

+ 2){(τ k

1
{(τ k + 1)ψD − ψR − 1}
+ 1)ψD − ψR − 1}

+ k{(1 − K ) + ( D − R )} + 1 < 0
19 By deﬁnition, an increase in both series indicates an appreciation of the US dollar.

84

Data Source: BIS, The Bank of Korea
Figure 3.2: Won/US Dollar Rates and Nominal Eﬀective Exchange Rates of the US Dollar
Another implication of equation (3.15) is that industries with higher import intensity
have higher pass-through rates: the coeﬃcient on the interaction term ( K · d ln EKU ) is
negative. This result depends basically on the assumption of the given model. In this paper,
the import intensity is the same for all ﬁrms in an industry regardless of how productive they
are. As shown in the second chapter, the relative demand for imported inputs to domestic
inputs is determined by an corresponding exchange rate, imported input requirement (φ)
and the elasticity of substitution between these two inputs (σ). On the contrary, Amiti et
al. (2012) suggest that the more productive ﬁrms source a greater share of their inputs from
abroad, which in turn lead to a further increase in their productivity using the endogenous
choice of importing at the ﬁrm level. They also ﬁnd that import intensive exporters are
more productive and greater market shares from detailed Belgium data. Given that both
import intensity and market share distributions are skewed toward the largest exporters,
they conclude that their ﬁndings help explain the observed low rate of pass-through at the
aggregate level.
85

But there is an inherent limitation in the use of ﬁrm-level import intensity to analyze
exchange rate pass-through at the industry level. This is because ﬁrms producing multiple products are prevalent in manufacturing (Bernard et al. (2010))20 while all ﬁrms are
assumed to be basically single-product ﬁrms in Amiti et al (2012). For instance, if ﬁrms produce several products across diﬀerent industries, import intensity data constructed at the
individual ﬁrm level would be inappropriate to examine industry-level pass-through. Furthermore, if import-intensive exporters have much larger market share, heterogeneity within
exporting ﬁrms plays relatively little role in examining import intensity at the industry level.
This is because a few larger exporters may determine industry import intensity.

Exchange Rate
Shocks

Direct Channel

Imported Input
Prices

Change in marginal
cost

Selection Channel

Change in composition of exporters
Aggregate Export
Prices

Figure 3.3: Import Intensity and Exchange Rate Pass-Through
Coming back to equation (3.15), the imported input share ( K ) aﬀects Korea’s export
price through two channels following exchange rate changes as shown in Figure 3.3. To begin
with, exchange rate shocks have a direct eﬀect on the marginal cost by making changes in
imported input prices. Korean ﬁrms may thus absorb some of exchange rate variations
through imported inputs, which act as a natural means for hedging exchange rate risks.
20 Bernard et al. (2010) ﬁnd that ﬁrms that produce multiple products represent 87 percent

of output in 1997, while ﬁrms present in multiple industries and sectors are responsible for
81 percent and 66 percent of output, respectively.
86

Next, aggregate export prices are aﬀected by variations in the composition of exporters due
to changes in the exporting cut-oﬀ (cKD ). Suppose that the cut-oﬀ declines owing to a
competitive disadvantage caused by higher imported input prices following a depreciation
of the won. The new aggregate export price is then computed taking into account only
the surviving Korean ﬁrms who are the most productive and had lower prices before the
exchange rate shock. Equation (3.15) shows that the direct channel is dominated by the
selection channel.

3.3.4

An Alternative Approach Without Selection Channel

Now I introduce an alternative model which includes only the intensive margin of trade
in order to emphasize the outstanding feature of my model. That is, the approach here
involves the subset of exporting ﬁrms, and hence does not model entry and exit decision.
Because we now condition our analysis on exporters, there is no selection channel unlike my
theoretical model.
Following Yun (2002), I investigate the extent of exchange rate pass-through when imported intermediate inputs are settled in the US dollar. Consider a Korean ﬁrm which exports
its product to n foreign markets. The US dollar is assumed to be the only settlement and
invoicing currency in trade.
The ﬁrm’s proﬁt maximization problem can then be written as follows.
n

max

pU j=1
j

n

EKU · pU yj
j

pU
j

· EjU − C(EKU )

yj

(3.17)

j=1

where pU denotes price of exports to country j in US dollar terms and yj is demand for
j
import from the Korean ﬁrm to country j. Let EKU be the exchange rate deﬁned as Korean
won per unit of US dollar and EjU be the exchange rate deﬁned as units of country j currency
per unit of US dollar. C denotes the marginal cost set in the Korean won. For simplicity, we
assume that marginal costs are constant and increase due to higher imported input prices

87

when the Korean won depreciates.21
Notice that pj destination-currency price of exports equals to pU · EjU . Even if a Korean
j
ﬁrm sets its export price in US dollar terms, import demand depends upon the price of
imports in destination currency. As a result, we can expect that export prices in Korea will
be aﬀected not only by the won-dollar exchange rate but also by the destination currencydollar exchange rate.
From the ﬁrst order condition,

pU =
j

ηj
C(EKU )
·
EKU
ηj − 1

(3.18)

∂yj pj
is the price elasticity of demand in country j.
∂pj yj
Take natural logarithms of both sides in equation (3.18) and diﬀerentiate, then

where ηj = −

dpU
j
U
pj
Let θj =

dηj
EjU dpU + pU dEjU
j
j
dpj
dEKU
C (EKU )
EKU
−
=− 1−
C(EKU )
EKU
ηj (ηj − 1)

(3.19)

1 d ln ηj
. Then, equation (3.19) can be re-written as
ηj − 1 d ln pj
dpU
j

dpU
dEjU
dEKU
C (EKU )
j
=− 1−
EKU
− θj
+
C(EKU )
EKU
EjU
pU
pU
j
j

(3.20)

C (EKU )
Notice that K =
E
is the share of imported inputs in total costs of the
C(EKU ) KU
Korean ﬁrm.
dpU
j
pU
j

=−

θj dEjU
1
dEKU
(1 − K )
−
1 + θj
EKU
1 + θj EjU

(3.21)

Equation (3.21) has the following implications. First, ﬁrms with higher import shares have
21 This implies a relatively low elasticity of substitution between domestic and foreign

intermediate inputs.
88

lower exchange rate pass-through as they face oﬀsetting exchange rate eﬀects on their
marginal costs. This prediction is consistent with the ﬁrm level studies like Amiti et al.
(2012) and Berman et al. (2012). Korean ﬁrms which are more dependent on imported
inputs in production will experience their input costs rise when the Korean won depreciates
against the US dollar and therefore lower their export prices in dollar terms less than others.
This conﬂicting result is based on the fact that there is no selection channel caused by variations in the composition of exporters. Consequently, the alternative model suggests that
prediction of ﬁrm-level may be valid at the industry level. That is, industries with higher
import intensity also exhibit lower exchange rate pass-through. Second, a depreciation of
the destination currency against the US dollar increases prices of Korean products sold in
the importing country. Hence, Korean exporters need to reduce their export prices in US
dollar terms. Third, the size of exchange rate pass-through depends on the price elasticity
of demand.

3.4
3.4.1

Empirical Evidence
Data

The annual data in this analysis covers the period from 2002 to 2012. I use data on export
unit values from the World Integrated Trade Solution (WITS), classiﬁed up to 6 digits by
the Harmonized System (HS) Combined.22 . Export prices in US dollar terms are proxied
by product-level unit values computed as the ratio of export values to export quantities.
Because export unit value is built using a free-on-board (FOB) price, we need to note that
it is diﬀerent from a ﬁnal retail price of a traded good. In particular, it does not include
transport or distribution costs.
Unit values may suﬀer from measurement errors due to compositional changes in quan22 HS Combined combines all revisions of HS (HS88/92, HS96, HS2002 and HS2012).

For more details on the nomenclature confer http://wits.worldbank.org/WITS/WITS/
WITSHELP/WITSHelp.htm
89

tities and quality mix or to errors in measuring quantities.23 Even if the use of highly
disaggregated data limits this problem, we need to implement some additional measures in
order to improve the reliability of unit values as proxies for prices. First, I drop annual
changes in unit values (expressed in absolute value) that are more than 200% to remove unusual price changes.24 Second, I adopt ﬁxed eﬀects to control for unobserved measurement
errors.25
The shares of imported inputs in production at the industry level are obtained using
the method introduced in section 3.2. In the regression, I use the time-averaged shares
of imported inputs IMs (2000, 2005 and 2010) by industry. Basically, we cannot obtain
observable annual data on imported input shares using Input-Output tables. However, the
average shares over the period might be justiﬁed by the fact that within-industry variation is
quite small relative to between-industry variation and they can help to avoid the endogeneity
and causation problems.26 These average shares by industry are listed in Appendix Table
A2. To link these shares to export unit values at the product level, I map each HS 6-digit
production code to the corresponding Korean sector classiﬁcation. To begin with, I employ
the concordance between 10-digit HS codes and the 5-digit Standard Industrial Classiﬁcation
(SIC) product classes used to classify US manufacturing production.27 I take the ﬁrst 6digits of the 10-digits HS code, and I include only the corresponding SIC code when it is a
unique mapping. Some HS 6-digit codes map to several SIC codes, so that I exclude these
23 See Silver (2007) for the concerns about unit value indices.
24 Although improved deletion routines are certainly advocated by the IMF Manual, there

is the arbitrary nature of the cut-oﬀ values often used in practice for deletion. See for
example Amiti et al. (2012) and Gaulier et al. (2008).
25 In particular, Gaulier et al. (2008) document that the importer-speciﬁc eﬀect should
control for trends in the evolution of the demand for quality. This is because country growth
is often accompanied by an improvement in the quality of imports.
26 Firms may adjust their imported input shares in response to exchange rate movements.
As a result, the shares at a speciﬁc time cannot be regarded as exogenous variables.
27 See Pierce and Schott (2009) for more detailed information. The concordance data are
downloadable from Schott’s website (http://faculty.som.yale.edu/peterschott/sub_
international.htm).

90

codes. Next, I match 5-digit SIC codes to 100 manufacturing sectors in small classiﬁcation
of the Korean Input-Output table.
Due to the nature of the model, each country in the sample should be a Korea’s major
competitor in the world market. Thus, I limit the set of destination countries according to
their shares in global trade. Another evidence supporting the use of a set of major countries
is that the non-major currencies have little help in explaining import price changes (Pollard
and Coughlin (2006)28 ). Speciﬁcally, I rank countries based on their averaged merchandise trade volumes (exports plus imports) for the period 2002 to 2012.29 And I focus on
25 countries, whose shares in world trade are more than 1%, except Korea itself and two
oil-exporting countries (Saudi Arabia, United Arab Emirates): United States, China, Germany, Japan, France, United Kingdom, Netherlands, Italy, Belgium, Canada, Hong Kong,
Spain, Singapore, Mexico, Russia, Taiwan, India, Switzerland, Australia, Malaysia, Thailand, Brazil, Austria, Sweden, Poland, accounting for 78% of global trade. I will also report
a robust test with other sets of countries later.
Data on bilateral exchange rates come from the International Financial Statistics of the
International Monetary Fund. To compute the US dollar exchange rate against the rest of
the world except Korea and country d, I consider product-speciﬁc exchange rates using the
following formula.
ln ERRUi,d,t =

βi,j,t ln ERDUj,t

(3.22)

j∈Jd

where Jd denotes all countries in the sample except Korea and country d, and βi,j,t is the
trade-based weight of country j for product i in the basket.30 There are two logical rea28 They examined exchange rate pass-through into US import prices for 29 manufacturing

industries using eight exchange rate indexes. They showed that major currency indexes
perform better than their broad currency counterparts.
29 Appendix Table A4 lists the top 50 countries with their averages trade volumes and the
corresponding shares in world trade.
j
j
Xi,t
Mi,t
j
j
30 To be more concrete, β
+ 0.5 ×
where Xi,t and Mi,t
i,j,t = 0.5 ×
j
j
j∈Jd Xi,t
j∈Jd Mi,t
denote total export and total import of product i by country j, respectively.

91

sons for me to adopt product-speciﬁc exchange rates instead of aggregate trade-weighted
exchange rates. First, the importance of each country as a competitor within a product can
diﬀer substantially from its importance in the aggregated world trade. As a consequence,
product-speciﬁc exchange rates may be more eﬀective than aggregate trade-weighted exchange rates in capturing changes in the competitive environment at the product level caused
by exchange rate movements.31 Second, product-speciﬁc exchange rates can help to reduce
multi-collinearity problem among multiple exchange rates. Indeed, we can observe reductions in correlation with other exchange rates when using product-speciﬁc exchange rates as
shown in Table 3.2.

Table 3.2: Correlation Coeﬃcient Among Exchange Rates
∆ ln ERKUt
∆ ln ERRUd,t

∆ ln ERDUd,t

0.086∗∗∗
(0.002)

0.356∗∗∗
(0.002)

0.066∗∗∗
0.209∗∗∗
(0.002)
(0.002)
Note: Asymptotic standard errors are reported in parentheses with ∗∗∗
denoting signiﬁcance at 1%.
∆ ln ERRUi,d,t

3.4.2

Empirical Speciﬁcation

According to the discussion in the previous theoretical framework, I estimate the following
speciﬁcation, where products are indexed by i, industries (or sectors) by s, destinations by
d and ∆ is the ﬁrst diﬀerence operator.
∆ ln U Vi,s,d,t = α1 ∆ ln ERKUt + α2 ∆ ln ERKUt · IMs + α3 ∆ ln ERDUd,t

(3.23)

+ α4 ∆ ln ERRUi,d,t + λt + µs,d + υi,s,d,t
31 From a similar perspective, Pollard and Coughlin (2006) conclude that industry-speciﬁc

exchange rate indexes are preferred to aggregate indexes.

92

where U Vi,s,d,t denotes the unit value of exports set in US dollars, used as a proxy for export
prices, ERKUt is the Korean won/US dollar rate, ERDUd,t is the exchange rate deﬁned as
units of country d currency per unit of the US dollar and ERRUi,d,t is the US dollar exchange
rate against the rest of the world except Korea and country d for product i. λt denotes year
dummies and µs,d represents industry-destination ﬁxed eﬀects to capture the time-invariant
characteristics that vary by industry, by destination or by industry-destination.
From equation (3.15), we expect a positive sign on α1 and a negative sign on α2 , α3
and α4 . In particular, the second coeﬃcient α2 captures the heterogeneity of pass-through
rates, that is, the fact that high import intensity increases industry-level pass-through rate
following a depreciation of the Korean won. In contrast, it is worth noting that the alternative
model in Section 3.3.4 predicts a positive sign on the coeﬃcient on the interaction term (α2 ).

3.4.3

Estimation Results

To examine the relationship between pass-through, multiple exchange rate movements
and import intensity, I start with a simple speciﬁcation and build up to the benchmark
empirical speciﬁcation in equation (3.23). Table 3.3 reports the results. First, column 1
shows that the unweighted average exchange rate pass-through elasticity in the sample is
0.61 when we use only the won/dollar exchange rate. In column 2, I include an interaction
between the won/dollar exchange rates and an industry’s import intensity. We see that
import intensity has a crucial characteristic to explain diﬀerent pass-through rates between
industries. Industries with a high share of imported inputs exhibit higher pass-through: a
1 percentage point higher import intensity leads to a 1.1 percentage higher pass-through.
For example, tobacco industry with a 9% import content has a pass-through of 30%, while
semiconductor industry with a 49% import content has a pass-through of 75%.
In column 3, I include a bilateral exchange rate between the currency of destination d and
the US dollar. All coeﬃcients corresponding to exchange rate shocks are of the expected sign
and these are also highly signiﬁcant except for the coeﬃcient on won/dollar exchange rate.

93

Table 3.3: Baseline Results
(1)
∆ ln ERKUt
∆ ln ERKUt × IMs

−0.610∗∗∗
(0.134)

(2)

(3)

(4)

−0.199
(0.163)

0.008
(0.176)

0.010
(0.186)

−1.142∗∗∗
(0.257)

−1.136∗∗∗
(0.256)

−0.136∗∗∗
(0.039)

−0.140∗∗∗
(0.040)

−1.131∗∗∗
(0.256)

∆ ln ERDUd,t
∆ ln ERRUi,d,t

0.001
(0.015)

Observations
283,722
283,722
283,722
278,252
Note: All regressions include country-industry ﬁxed eﬀects and yearly dummies.
Standard errors are clustered at the country-year level, reported in parenthesis with ∗ ,
∗∗ and ∗∗∗ denoting signiﬁcance at 10%, 5% and 1%, respectively.

We see that Korean exporters lower their export prices in US dollar terms in response to a
depreciation of the destination currency against the US dollar. A depreciation of the destination market’s currency against the US dollar not only raises the prices of Korean products
sold in the importing country directly, but also lowers the exporting cut-oﬀ of Korean ﬁrms
due to intensiﬁed competition and so causes a decrease in aggregate export price. In column
4, I estimate the main empirical speciﬁcation in equation (3.23) by adding product-speciﬁc
exchange rates between the other competitors’ currencies and US dollar. We see that the
coeﬃcients both on the import intensity interaction and on the destination currency/US
dollar exchange rates remain almost unchanged and strongly signiﬁcant. However, contrary
to expectations, the estimated coeﬃcient on the trade-weighted exchange rates (ERRU ),
which I will refer to as third-country exchange rates, has the negligible positive sign and is
not statistically signiﬁcant.
There are several possible explanations for why third-country exchange rates do not have
a signiﬁcant impact on export prices. First, we need to re-examine the assumption about the
invoice currency of international trade. In this paper, I assume that every country invoices

94

its trade in the US dollar. The US dollar is still the primary invoice currency in international
transactions, but the extent to which the dollar is used diﬀers substantially across countries
as shown in Appendix Table A1. As a result, my model might not fully capture the eﬀect of
exchange rate shocks on ﬁrms’ pricing behavior. Furthermore, there exists some empirical
evidence to support that exchange rate pass-through varies considerably with the choice of
invoicing currency.32 Another possible explanation is that Korean exporters in practice do
not care too much about the other competitors’ exchange rates in setting their prices to each
destination. If we assume that prices are set conditional on the available information before
the realization of exchange rates, it would be diﬃcult or costly for ﬁrms to gather suﬃcient
information to forecast multiple exchange rates.

3.4.4

Robustness

In this section, I provide diﬀerent sets of robustness checks. First, I control for some
other industry characteristics that could generate heterogeneity in the pass-through rates.
Second, I test how robust my results are to alternative non-parametric speciﬁcation. Third,
I check that my results are robust to the use of real exchange rates. Fourth, I further check
the robustness of my results within alternative samples of the dataset. Finally, I use the
export price index as an alternative price indicator instead of unit value.
Additional Controls Although I focus on import intensity, the empirical work undertaken so far has been related the degree of exchange rate pass-through to other industry
characteristics. Dornbusch (1987) suggests that the extent of the pass-through depends on
the degree of product substitution, the relative number of domestic and foreign ﬁrms, and
market structure. Yang (1997) documents that the degree of pass-through is found to be
positively correlated to product diﬀerentiation and negatively to the elasticity of marginal
cost. In addition, Choi and Kim (2001) note the share of Korean exporters in the foreign
32 For instance, Gopinath et al. (2010) ﬁnd that there is a large diﬀerence in the exchange

rate pass-through of the average US import good priced in dollars (25 percent) versus nondollars (95 percent).

95

market. Unfortunately, most variables mentioned above are not directly observable, but I
employ some observable variables available to control product or industry heterogeneity.
To begin with, I deﬁne the product-speciﬁc Korea’s share in total import of country d in
product i33 as follows.

M Si,d,t =

d
Mi,K,t
d
Mi,t

(3.24)

d
d
where Mi,K,t is total import of product i from Korea to country d in period t and Mi,t is

total import of product i by country d in period t.
The elasticity of the exporter price also depends on the degree of competition in the sector.
As a measure of the extent of product diﬀerentiation, I use the elasticity of substitution (σs,d )
estimated by Broda and Weinstein (2006). They provide their estimates at the HS 3-digit
level. Thus, I match them with product codes in the HS 6-digit classiﬁcation.
Lastly, the revealed comparative advantage index (RCAIi,t ) is widely used to measure
comparative advantages of nations in international trade. This index measures a country’s
relative export performance in a speciﬁc product category compared to its overall export
performance in the following manner.

RCAIi,t =

K
W
Xi,t /Xi,t
K
W
Xt /Xt

(3.25)

K
W
Here, Xi,t is total export of product i by Korea in period t and Xi,t is total export of product
K
W
i by world in period t. Xt is total export of Korea in period t and Xt is total export of

world in period t. If RCAIi,t > 1, it is assumed that Korea has comparative advantage in
product i in period t.
Table 3.4 re-estimates the main empirical speciﬁcation in column 4 of Table 3.3 with
33 As Feenstra et al. (1996), it would be more appropriate to employ Korea’s share in total

destination sales which also covers competition with destination ﬁrms. But constructing
product-speciﬁc market shares in this way could be very time consuming. Instead, I select
an alternative measure which easily be built by customs data.

96

additional controls. We ﬁnd that pass-through is increasing in Korea’s share in total import
of the destination market in column 1. It contrasts with the previous ﬁrm-level studies such
as Berman et al. (2012) and Amiti et al. (2012), which document that large ﬁrms absorb
more exchange rate movements in their markups. However, this result is similar to that
reported by Auer et al. (2012) who ﬁnd that the rate of pass-through into import prices
following trade-partner currency movements is increasing in the trade partner’s sector-speciﬁc
market share.34 In column 2, the coeﬃcient on the elasticity of substitution interaction is
insigniﬁcant.35 One possibility is that there might be a considerable amount of heterogeneity
among more detailed products (HS 6-digit) within a broad product classiﬁcation (HS 3-digit).
Next, column 3 suggests that exporters pass exchange rate shocks through their prices more
in comparative advantage industries. It is likely that revealed comparative advantage indexes
are systematically associated with market shares in the destination market. In that regard,
this result is compliant with the outcome in column 1. Indeed, we observe that the coeﬃcients
on both the market share and the revealed comparative advantage index interaction terms
drop slightly in size in column 4 when we include both variables.
Finally, I include all the variables in column 4. Once again, variations in import intensity,
market shares and comparative advantage are the eﬀective tools for explaining diﬀerences in
pass-through across industries. The inclusion of other control variables does not modify substantially the size and statistical signiﬁcance of the coeﬃcients both on the import intensity
interaction and on the destination currency/US dollar exchange rates.
Nonparametric Speciﬁcation I interact the won/dollar exchange rates with diﬀerent
bins built from percentiles of industries’ import intensity to examine the robustness of alternative non-parametric speciﬁcations. First, I construct dummy variables for industries
34 Meanwhile, Feenstra et al. (1996) develop a Bertrand diﬀerentiated products model and

show that the relationship between pass-through and market share is signiﬁcantly nonlinear
in the global automobile industry.
35 Broda and Weinstein (2006) do not provide their estimates regarding four countries;
Belgium, Russia, Taiwan and Singapore. Hence, these countries were excluded from regression. I re-estimated column 2 without country-industry ﬁxed eﬀects, but the result remain
unchanged.
97

Table 3.4: Robustness: Additional Controls
(1)
∆ ln ERKUt

(2)

(3)

(4)

0.127
(0.192)

−0.017
(0.220)

0.133
(0.185)

0.203
(0.221)

∆ ln ERKUt × IMs

−1.143∗∗∗
(0.249)

−1.111∗∗∗
(0.292)

−1.072∗∗∗
(0.250)

−1.094∗∗∗
(0.288)

∆ ln ERDUd,t

−0.126∗∗∗
(0.040)

−0.150∗∗∗
(0.041)

−0.130∗∗∗
(0.039)

−0.132∗∗∗
(0.039)

∆ ln ERRUi,d,t

−0.012
(0.015)

−0.004
(0.018)

−0.007
(0.015)

−0.025
(0.017)

∆ ln ERKUt × M Si,d,t

−0.875∗∗
(0.344)

M Si,d,t

−0.779∗∗
(0.381)

0.561∗∗∗
(0.034)

0.353∗∗∗
(0.040)

∆ ln ERKUt × σs,d

−0.001
(0.001)

0.000
(0.001)

σs,d

−0.000
(0.000)

0.000
(0.000)
−0.033∗∗∗
(0.008)

∆ ln ERKUt × RCAIi,t

0.024∗∗∗
(0.001)

RCAIi,t

−0.021∗∗
(0.008)
0.018∗∗∗
(0.001)

Observations
246,362
203,525
278,252
180,162
Note: All regressions include country-industry ﬁxed eﬀects and yearly dummies.
Standard errors are clustered at the country-year level, reported in parenthesis with ∗ ,
∗∗ and ∗∗∗ denoting signiﬁcance at 10%, 5% and 1%, respectively.

pertaining to each category, based on the median, quintiles and deciles of import intensity.
And I replace the import intensity with those dummy variables and also interact them with
the won/dollar exchange rates. In table 3.5, I replicate column 4 of Table 3.3 and column 4
of Table 3.4.
The interaction terms are always negative and statistically signiﬁcant. The diﬀerence
in the pass-through between industries with the highest import intensity and the other

98

Table 3.5: Robustness: Non-parametric Speciﬁcation
(1)
∆ ln ERKUt

(2)

(3)

(4)

−0.339∗∗ −0.362∗∗ −0.392∗∗ −0.121
(0.163)
(0.162)
(0.162)
(0.191)

∆ ln ERKUt × IMT op50% −0.155∗∗∗
(0.042)

(6)

−0.150
(0.189)

−0.164
(0.189)

−0.158∗∗∗
(0.045)
−0.404∗∗∗
(0.079)

∆ ln ERKUt × IMT op20%

(5)

−0.421∗∗∗
(0.087)
−0.407∗∗∗
(0.122)

∆ ln ERKUt × IMT op10%

−0.510∗∗∗
(0.156)

∆ ln ERDUd,t

−0.138∗∗∗ −0.140∗∗∗ −0.139∗∗∗ −0.131∗∗∗ −0.132∗∗∗ −0.131∗∗∗
(0.040)
(0.040)
(0.040)
(0.039)
(0.039)
(0.039)

∆ ln ERRUi,d,t

−0.000
(0.015)

−0.000
(0.015)

−0.001
(0.015)

−0.026
(0.017)

−0.025
(0.017)

−0.027
(0.017)

−0.932∗∗ −0.858∗∗ −0.942∗∗
(0.402)
(0.389)
(0.395)

∆ ln ERKUt × M Si,d,t
M Si,d,t

0.350∗∗∗ 0.352∗∗∗ 0.351∗∗∗
(0.040)
(0.040)
(0.040)

∆ ln ERKUt × σs,d

0.000
(0.001)

0.000
(0.001)

0.000
(0.001)

σs,d

0.000
(0.000)

0.000
(0.000)

0.000
(0.000)

−0.025∗∗∗ −0.021∗∗ −0.032∗∗∗
(0.009)
(0.008)
(0.009)

∆ ln ERKUt × RCAIi,t

0.018∗∗∗ 0.018∗∗∗ 0.018∗∗∗
(0.001)
(0.001)
(0.001)

RCAIi,t

Observations
278,252 278,252 278,252 180,162 180,162 180,162
Note: All regressions include country-industry ﬁxed eﬀects and yearly dummies.
Standard errors are clustered at the country-year level, reported in parenthesis with ∗ ,
∗∗ and ∗∗∗ denoting signiﬁcance at 10%, 5% and 1%, respectively.

industries is strongly signiﬁcant: For example, as shown in column 1, export industries below
the median decrease their prices by 0.3% following a 1 % depreciation of the won against
the US dollar while export industries above the median decrease their prices by 0.5%. The

99

pass-through rates of industries with much higher import intensity are even higher. Column
3 shows us that export industries below the top decile decrease their prices by 0.4% following
a 1 % depreciation of the won against the US dollar while export industries above the top
decile decrease their prices by 0.8%.
I now present graphically the fragmented set of non-parametric interaction terms using
20 bins (20 percentiles) classiﬁed by industry’s import intensity in Figure 3.4, together with
90% conﬁdence intervals and a lowess smoother36 . On the whole, we can see that the passthrough elasticity (in absolute value) increases with import intensity. However, Figure 3.4
also suggests the possibility of nonlinear relationships between exchange rate pass-through
rates and imported input shares. In particular, for industries with smaller imported input
shares, the negative trend is not clear or a modest positive trend is detected.

Figure 3.4: Exchange Rate Pass-Through Elasticity by Import Intensity (20 Bins)
Real Exchange Rates In table 3.6, I replicate the last columns of Table 3.3 and Table
36 The default bandwidth of 0.8 in STATA is used for smoothing. This means that 80% of

the data are used for calculating smoothed values for each point in the data except for the
end points. The greater the bandwidth, the greater the smoothing.

100

3.4 by replacing nominal exchange rates with real exchange rates.37 Real exchange rates
are computed from the Penn World Table.38 Due to data limitations, the analysis period
is curtailed to 2002-2010. One interesting feature is the coeﬃcient on the won/dollar real
exchange rates is positive as expected from the theoretical model. We also observe that the
key result related to import intensity is strengthened. On the other hand, the coeﬃcient on
third-country exchange rates have the opposite sign or statistically insigniﬁcant.
Alternative Samples I further investigate the robustness of the baseline results using
alternative samples of the dataset. Table 3.7 provides the results of the main speciﬁcation
from column 4 of table 3.3 in eight alternative samples.
First, I add 28 countries including oil-exporting countries to the baseline sample, whose
shares in world trade (2002-2012) are more than 0.3%.39 A total of 53 countries account
for 93% of global trade. Column 1-5 of Table 3.7 report the results for ﬁve alternative sets
of destination countries based on the extended sample– all sample countries, the US only,
all sample countries excluding the US, all Euro countries, non-Euro countries.40 Column
1 reveals the same patterns we ﬁnd in the baseline sample. It is remarkable that Korean
exporters pass won/dollar exchange rate shocks through their prices to the US less than to
other countries as shown in column 2 and 3. This is consistent with previous studies on low
pass-through into the US. Interestingly, we see a huge diﬀerence between the coeﬃcients on
the bilateral exchange rates between destination currency and US dollar in column 4 and
5. Although both coeﬃcients are not signiﬁcant, the (absolute) pass-through elasticity for
Euro countries is about four times larger than that for non-Euro countries. It suggests that
37 In the second chapter, I mention that exchange rates can be interpreted as real exchange

rates because each country’s nominal wage reﬂects productivity by assumption.
38 See Rodrick (2008) for more detailed calculation.
39 The list of countries to be added is as follows: Saudi Arabia, United Arab Emirates,
Turkey, Indonesia, Czech, Norway, Ireland, Denmark, Hungary, South Africa, Finland, Iran,
Vietnam, Portugal, Israel, Slovakia, Chile, Venezuela, Philippines, Nigeria, Ukraine, Argentina, Romania, Greece, Algeria, Kuwait, Kazakhstan, Qatar. Iraq was ruled out due to
data unavailability regardless of its share (0.3%) in global trade.
40 Slovakia joined the Euro Zone in 2009. Considering its late joining, I exclude Slovakia
from EU countries.
101

Table 3.6: Robustness: Real Exchange Rates
(1)

(2)

0.728∗∗∗
(0.233)

0.315
(0.278)

∆ ln RERKUt × IMs

−1.170∗∗∗
(0.283)

−1.173∗∗∗
(0.320)

∆ ln RERDUd,t

−0.161∗∗∗
(0.046)

−0.170∗∗∗
(0.042)

∆ ln RERRUi,d,t

0.130∗∗
(0.063)

∆ ln RERKUt

−0.010
(0.074)
−0.866∗∗
(0.426)

∆ ln RERKUt × M Si,d,t
M Si,d,t

0.374∗∗∗
(0.042)

∆ ln RERKUt × σs,d

0.000
(0.001)

σs,d

−0.000
(0.000)

∆ ln RERKUt × RCAIi,t

−0.017∗
(0.009)
0.018∗∗∗
(0.001)

RCAIi,t

Observations
218,955
141,477
All regressions include country-industry ﬁxed eﬀects and yearly dummies.
Standard errors are clustered at the country-year level, reported in
parenthesis with ∗ , ∗∗ and ∗∗∗ denoting signiﬁcance at 10%, 5% and 1%,
respectively.

diﬀerences in pass-through might be related to the currency of invoicing.

102

Table 3.7: Robustness: Alternative Samples
Countries (Extended Sample)
All
(1)
∆ ln ERKUt

−0.120
(0.156)

∆ ln ERKUt × IM

−1.107∗∗∗
(0.215)

∆ ln ERDUd,t

0.001
(0.001)

W/O US
(3)

Only Euro W/O Euro
(4)
(5)

0.699∗∗∗
(0.107)

−0.202
(0.161)

−0.073
(0.418)

−0.117
(0.168)

−1.117∗∗∗
(0.227)

−1.076∗∗∗
(0.377)

−0.066∗∗
(0.029)
0.002
(0.008)

−0.082∗∗∗
(0.029)

∆ ln ERRUi,d,t

Only US
(2)

−0.865∗∗
(0.344)

−0.014
(0.055)

Products (Deletion of Outlier)
∆P rice ≤ |∆P rice| ≤
All
100%
Products
50%
(7)
(6)
(8)
0.075
(0.241)

0.040
(0.183)

0.212∗∗
(0.090)

−1.102∗∗∗
(0.244)

−0.785∗∗
(0.316)

−1.180∗∗∗
(0.251)

−1.233∗∗∗
(0.171)

−0.136
(0.102)

−0.036
(0.030)

−0.103∗∗
(0.052)

−0.138∗∗∗
(0.040)

−0.101∗∗∗
(0.013)

−0.001
(0.019)

0.001
(0.009)

0.036∗∗
(0.018)

−0.006
(0.016)

−0.003
(0.005)

Observations
383,145
20,584
362,561
68,156
314,989
302,740
262,639
206,032
Note: All regressions include country-industry ﬁxed eﬀects and yearly dummies. Standard errors are clustered
at the country-year level, reported in parenthesis with ∗ , ∗∗ and ∗∗∗ denoting signiﬁcance at 10%, 5% and 1%,
respectively.

103

Next, I check the robustness against outliers. As IMF (2010) points out, unit value
indices rely to a large extent on outlier detection and deletion. Therefore, I provide the
results for three alternative sets in column 6-8 within the previous baseline country sample–
all products (no outlier exclusion), dropping yearly unit value changes of over 100 percent
and dropping changes of plus or minus over 50 percent. In column 6-8, we see that the
results are insensitive to outliers.
Alternative Measure of Prices Unit value indices are widely used as measures of
price changes of traded goods for economic analysis due to the relatively low cost of such
data. However, they are exposed to well-recognized bias as mentioned in Section 3.4.1.41
Therefore, I examine whether my results are robust to replacing unit values with export
price indexes.
The quarterly data in this analysis covers the period from 1994 to 2012.42 Export
prices used here are export price indexes on the US dollar basis obtained from the Bank of
Korea. I select 42 industries that encompass a broad spectrum of manufacturing industries
using mapping between product classiﬁcation of export prices and sector classiﬁcation of the
Korean Input-Output table.43 In the regression, I use the time-averaged shares of imported
inputs IMs (1990, 1995, 2000, 2005 and 2010) by industry. These average shares by industry
are listed in Appendix Table A5.
41 United Nations (1981) provided an international guideline on choosing price measure-

ment in external trade. Well-endowed countries were advised to conduct a comprehensive
price survey in order to complement unit value indices. Countries under tight budgetary
conditions were advised to use unit value indices but deﬁne each item in the narrowest sense
possible.
42 When selecting the analysis period, two main factors are considered. First, the existing empirical studies have concluded that the pass-through into export prices has declined
since the ﬁnancial crisis in 1997 (Kim and Choi (2001), Kim and Lee (2009)). Thus, I
estimate pass-through coeﬃcients using recent time series. Second, the nominal eﬀective exchange rates for the US dollar are obtained from BIS broad indices comprising 61 economies
(2010=100) with data from 1994. With drastic changes in global trade over recent decades,
the broad indices are more appropriate than the narrow indices comprising 27 economies
with data from 1967.
43 42 industries are comprised of 1 large (Food), 3 medium (Apparel, Pulp & Paper, Other
Manufacturing) and 38 small industries.
104

Following Kim and Choi (2001), I estimate the following speciﬁcation.
∆ ln XPs,t = α1 ∆ ln ERKUt + α2 ∆ ln ERKUt · IMs + α3 ∆ ln EERU SDt

(3.26)

+ α4 ∆ ln P Ps,t + α5 ∆ ln M Os,t−1 + s,t
where XPs,t is the export price index for industry s (set in the US dollar), ERKUt is
the exchange rate for the Korean won per US dollar, EERU SDt is the US dollar eﬀective
exchange rate obtained from BIS broad indices, P Ps,t is the producer price index for industry
s reﬂecting changes in production costs, M Os,t−1 is the manufacturing operation ratio index
for industry s measuring demand-side pressures. IMs is the share of imported inputs in
domestic production for industry s as deﬁned earlier.
Because export price indexes are not computed on the destination basis, I use the US
dollar eﬀective exchange rate to capture changes in the value of the dollar versus other
currencies.44 An increase in EERU SDt indicates an appreciation of the US dollar.
I estimate equation (3.26) with quarterly dummies in order to correct for seasonal factors.
The literature generally assumes pass-through to occur over time after the initial exchange
rate shock. So I also estimate the long-run pass through, which are the sum of the coeﬃcients
on the contemporaneous exchange rates and two lags of exchange rates. The number of lag
terms is chosen using the standard “general-to-speciﬁc” modeling.
Table 3.8 reports the results of the estimations of short-run and long-run exchange rate
pass-through. All the coeﬃcients associated with exchange rates have the expected sign and
are strongly signiﬁcant. Industries with a high share of imported inputs exhibit higher passthrough into US dollar export prices. Korean ﬁrms lower their dollar-denominated export
prices in response to the worldwide US dollar appreciation. This result is consistent with
44 Similarly, Kim and Lee (2009) added the dollar’s eﬀective exchange rate provided by

FRB to their regression equation. They did not derive a speciﬁc relationship among exchange
rates, but considered that other competitors’ export prices might be aﬀected by changes in
the exchange rates between their currencies and US dollar. Meanwhile, Kim and Choi (2001)
estimated the exchange rate pass-through by including the bilateral exchange rates between
the Japanese yen and the US dollar.

105

the predicted negative sign of the coeﬃcients on ∆ ln ERDUd,t and ∆ ln ERRUi,d,t in the
baseline model. Moreover, the coeﬃcients on the control variables have the expected positive sign: Both higher production costs and increased global demand put upward pressure
on export price.

Table 3.8: Robustness: Alternative Measure of Prices
Short-run

Long-run

0.510∗∗∗
(0.089)

0.424∗∗∗
(0.125)

∆ ln ERKUt × IMs

−1.803∗∗∗
(0.246)

−1.977∗∗∗
(0.306)

∆ ln EERU SDt

−0.550∗∗∗
(0.076)

−0.621∗∗∗
(0.153)

∆ ln P Ps,t

0.761∗∗∗
(0.055)

0.577∗∗∗
(0.051)

∆ ln M Os,t−1

0.035∗∗∗
(0.011)

0.077∗∗∗
(0.025)

∆ ln ERKUt

Observations
3,150
3,066
Note: Both regressions include industry ﬁxed eﬀects and quarterly dummies.
Standard errors are clustered at the industry level, reported in parenthesis
with ∗∗∗ denoting signiﬁcance at 1%.

3.5

Conclusion

I ﬁnd that Korean manufacturing industries with higher import intensity have higher
exchange rate pass-through elasticities. This result is consistent with predictions of the
theoretical model developed in this paper. My work suggests that the extensive margin
of trade may play an important role in determining the degree of aggregate exchange rate
pass-through. Higher imported input prices following a depreciation not only have a direct
eﬀect on the marginal cost, but also aﬀect the exporting cut-oﬀ due to changes in the
competitive environment. Consequently, aggregate export prices are aﬀected by variations
106

in the composition of exporters due to changes in the exporting cut-oﬀ.
This paper relates industries’ import intensities to pass-through rates, but does not mention about the relationship between ﬁrm-speciﬁc import intensity and pass-through. In
future research I want to analyze more the puzzling diﬀerence in the current industry-level
study and previous ﬁrm-level studies. For instance, Amiti et al. (2012) ﬁnd that ﬁrms
with high import shares have low exchange rate pass-through and import intensity is heavily
skewed toward the largest exporters. Therefore, they suggest that their ﬁndings help explain
low aggregate pass-through elasticities. In my opinion, there are two limitations in linking
their results to aggregate pass-through. First, they do not consider the extensive margin of
trade at all. Second, ﬁrm-level data may suﬀer from serious measurement errors due to the
prevalence of multiproduct ﬁrms.
Finally, we need to develop a more realistic model to examine in detail the role of imported
inputs in exchange rate pass-through. I assume full exchange rate pass-through into imported
input prices like most researchers, but it seems to be a rather strong assumption given the
empirical evidence of partial exchange rate pass-through into import prices. In addition, the
invoice currency in trade is very important in capturing precisely changes in imported input
prices in response to exchange rate shocks.

107

APPENDIX

108

Appendix

Table A1: US Dollar Use in the Export and Import Invoicing
Observation2 US Dollar Share in Export
United States

US Dollar Share in Import

2003

99.8

92.8

Japan

2001

52.4

70.7

Korea

2012

85.1

83.9

Thailand

1996

83.9

83.9

Australia

2002

67.9

50.1

France1

2002

34.2

43.2

Germany1

2002

32.3

37.9

Italy

2002

20.5

30.8

United Kingdom

2002

26.0

37.0

Bulgaria

2002

44.5

37.1

Czech

2002

14.7

19.5

Poland

2002

29.9

28.6

Slovenia

2002

9.6

13.3

Asia

EU

EU accession

1
2

Invoicing data refer only to the invoicing of “extra euro-area” trade.
Latest Observations are annual except for: Japan-January 2001, Germany-2002 Q3,
United States-2003 Q1. Thailand is for overall trade and is not broken down by imports.
Source: Goldberg and Tille (2008) except for Korea – Korea Customs Service (January
2013)
109

Table A2: Share of Imported Inputs by Manufacturing Industry
1990

1995

2000

2005

2010

Average1

Naphtha

0.93

0.77

0.94

0.87

0.88

0.89

Coal

0.39

0.64

0.67

0.78

0.86

0.77

Non-ferrous Metal Ingot

0.60

0.58

0.68

0.72

0.78

0.72

Basic Organic Chemical

0.67

0.48

0.68

0.70

0.78

0.72

Leather and Fur

0.48

0.54

0.60

0.67

0.80

0.69

Sugar

0.65

0.64

0.57

0.57

0.81

0.65

Cereal Milling

0.73

0.63

0.60

0.57

0.76

0.65

Other Petroleum

0.48

0.48

0.49

0.71

0.73

0.64

Fuel

0.78

0.54

0.58

0.60

0.68

0.62

Computer

0.45

0.54

0.55

0.61

0.69

0.62

Pig Iron and Ferrous-Alloy

0.45

0.54

0.60

0.53

0.71

0.61

Non-ferrous Metal

0.57

0.55

0.53

0.59

0.62

0.58

Synthetic Rubber

0.48

0.31

0.45

0.62

0.63

0.57

Man-made Fiber

0.51

0.51

0.52

0.55

0.62

0.56

Synthetic Resin

0.52

0.41

0.51

0.55

0.62

0.56

Aircraft

0.48

0.47

0.57

0.53

0.58

0.56

Crude Steel

0.45

0.48

0.54

0.54

0.59

0.56

Watch

0.44

0.37

0.46

0.55

0.63

0.55

..

0.57

0.55

0.48

0.55

0.53

Animal Feed

0.47

0.49

0.46

0.54

0.57

0.53

Fertilizer and Agricultural Chemical

0.48

0.44

0.47

0.53

0.54

0.51

..

0.38

0.48

0.51

0.54

0.51

0.34

0.36

0.49

0.49

0.54

0.50

Leather and Fur Garment

Electronic Video and Audio
Telecommunication and Broadcasting

Continued on next page

110

Table A2 (cont’d)
1990

1995

2000

2005

2010

Average1

Cold Rolled Iron

0.37

0.39

0.40

0.52

0.58

0.50

Fiber

0.48

0.56

0.42

0.48

0.57

0.49

Wood

0.63

0.50

0.47

0.49

0.51

0.49

Semiconductor

0.61

0.33

0.50

0.48

0.48

0.49

Vegetable and Animal Oil

0.31

0.38

0.46

0.49

0.51

0.48

Other Basic Organic Chemical

0.42

0.35

0.40

0.49

0.54

0.48

Other Chemical

0.40

0.41

0.40

0.48

0.54

0.48

Electronic Display

0.27

0.30

0.46

0.44

0.47

0.45

Paper

0.46

0.44

0.42

0.43

0.51

0.45

Dye and Paint

0.42

0.41

0.41

0.43

0.51

0.45

Starche and Glucose

0.42

0.39

0.44

0.40

0.51

0.45

Railway Locomotive

0.29

0.41

0.47

0.39

0.46

0.44

Cast and Forged Steel

0.27

0.31

0.36

0.37

0.57

0.44

Pulp

0.44

0.38

0.39

0.49

0.44

0.44

Inorganic Chemical

0.32

0.24

0.34

0.43

0.51

0.43

Hot Rolled Iron

0.36

0.36

0.40

0.40

0.48

0.43

Oﬃce Machinery

0.34

0.31

0.37

0.42

0.47

0.42

Other Transport Equipment

0.28

0.27

0.33

0.34

0.51

0.39

Tire and Tube

0.38

0.35

0.33

0.39

0.46

0.39

Texile

0.38

0.37

0.34

0.37

0.45

0.39

Electronic Component

0.31

0.30

0.35

0.35

0.43

0.38

Ship

0.33

0.33

0.32

0.40

0.41

0.37

Plastic

0.35

0.30

0.33

0.35

0.44

0.37

Optical Instrument

0.30

0.17

0.34

0.36

0.42

0.37

Continued on next page
111

Table A2 (cont’d)
1990

1995

2000

2005

2010

Average1

..

0.31

0.31

0.36

0.43

0.37

Other Electrical Equipment

0.36

0.31

0.34

0.35

0.41

0.37

Wood Product

0.33

0.35

0.37

0.34

0.39

0.37

Cement

0.26

0.19

0.21

0.32

0.56

0.36

Footwear

0.31

0.33

0.35

0.32

0.42

0.36

Other Steel

0.39

0.29

0.26

0.39

0.44

0.36

Engine and Turbine

0.27

0.41

0.36

0.34

0.34

0.35

Textile Dyeing

0.28

0.23

0.34

0.31

0.36

0.34

..

0.34

0.34

0.32

0.36

0.34

0.34

0.24

0.28

0.33

0.40

0.34

Trailer and Container

..

0.37

0.41

0.29

0.29

0.33

Metal Container

..

0.30

0.29

0.31

0.38

0.33

Paper Product

0.32

0.30

0.32

0.29

0.36

0.33

Other Manufacturing

0.27

0.28

0.33

0.30

0.34

0.32

Medical and Precision Instrument

0.28

0.23

0.31

0.31

0.35

0.32

Other Rubber

0.33

0.29

0.27

0.31

0.38

0.32

Glass

0.27

0.26

0.24

0.32

0.38

0.32

Other Texile

0.30

0.27

0.29

0.29

0.37

0.31

..

0.28

0.26

0.29

0.40

0.31

Motor Vehicle Component

0.25

0.25

0.27

0.31

0.36

0.31

Electric Motor and Generator

0.31

0.28

0.29

0.30

0.34

0.31

Structural Metal

0.27

0.25

0.30

0.29

0.34

0.31

Other Special Purpose Machinery

0.30

0.26

0.28

0.28

0.36

0.31

Motor Vehicle

0.23

0.25

0.27

0.31

0.34

0.31

Other Leather

Agricultural and Construction Machinery
Domestic Electric Appliance

Knitted Apparel

Continued on next page
112

Table A2 (cont’d)
1990

1995

2000

2005

2010

Average1

Furniture

0.36

0.28

0.28

0.29

0.34

0.30

Toy and Athletic Good

0.24

0.22

0.27

0.28

0.35

0.30

Other Non-metallic Mineral

0.20

0.17

0.24

0.29

0.36

0.30

Clay

0.20

0.22

0.27

0.29

0.33

0.30

..

0.25

0.26

0.29

0.33

0.29

Condiment

0.26

0.28

0.27

0.28

0.34

0.29

Bakery and Noodle

0.33

0.28

0.27

0.26

0.35

0.29

Other General Machinery

0.31

0.25

0.28

0.27

0.33

0.29

..

0.29

0.29

0.25

0.33

0.29

0.22

0.26

0.29

0.28

0.30

0.29

..

0.26

0.25

0.26

0.35

0.29

Pharmaceutical

0.20

0.24

0.26

0.27

0.33

0.29

Meat

0.24

0.25

0.26

0.26

0.34

0.29

Cosmetic and Soap

0.29

0.24

0.26

0.25

0.33

0.28

General Machinery Component

0.21

0.22

0.24

0.27

0.33

0.28

Luggage and Handbag

..

0.23

0.27

0.25

0.30

0.27

Fabric Apparel

..

0.32

0.32

0.23

0.28

0.27

Machine Tool

0.25

0.26

0.26

0.24

0.31

0.27

Other Metal

0.29

0.23

0.25

0.25

0.31

0.27

Other Food

0.21

0.25

0.23

0.24

0.31

0.26

Non-alcoholic Beverage

0.19

0.17

0.22

0.24

0.30

0.25

Dairy

0.24

0.23

0.24

0.22

0.29

0.25

Ceramic

0.20

0.17

0.23

0.21

0.29

0.24

Printing and Reproduction

0.23

0.21

0.23

0.19

0.25

0.22

Refrigerator and Air Conditioning

Industrial Conveying
Fish
Hand Tool and Metal Wire

Continued on next page
113

Table A2 (cont’d)
1990

1995

2000

2005

2010

Average1

Concrete

0.15

0.13

0.16

0.21

0.30

0.22

Fruit and Vegetable

0.12

0.10

0.14

0.17

0.20

0.17

Alcoholic Beverage

0.12

0.12

0.12

0.11

0.15

0.13

Tobacco

0.06

0.06

0.07

0.09

0.12

0.09

Cereal Husking

0.05

0.04

0.05

0.09

0.11

0.08

1

The time-averaged shares during 2000-2010 used in the main empirical analysis.
Descriptive statistics are as follows: Min=0.08, Max=0.89, Mean=0.40, Median=0.36,
Standard Deviation=0.15.

114

Table A3: Imported Input Shares by Country

Country

IO Year

Import Content

Country

IO Year

of Exports
Australia

Import Content
of Exports

2004/05

0.202

Japan

2005

0.176

Belgium

2005

0.540

Korea

2005

0.417

Brazil

2005

0.177

Mexico

2003

0.475

Canada

2005

0.377

Netherlands

2005

0.478

China

2005

0.304

Norway

2005

0.318

Denmark

2005

0.370

Poland

2005

0.386

Finland

2005

0.409

Russia

2000

0.152

France

2005

0.333

Singapore

2000

0.698

Germany

2005

0.309

Spain

2005

0.426

Greece

2005

0.351

Sweden

2005

0.393

Hungary

2005

0.630

Switzerland

2005

0.291

2003/04

0.267

Taiwan

2006

0.545

Indonesia

2005

0.232

Turkey

2002

0.292

Ireland

2005

0.582

United Kingdom

2005

0.318

Israel

2005

0.507

United States

2005

0.164

Italy

2004

0.341

Vietnam

2000

0.463

India

Source: OECD, STAN Database for Structural Analysis (www.oecd.org/sti/stan)

115

Table A4: Top 50 Countries in World Merchandise Trade (2002-2012)
Trading Volume1
Rank

Country

Share
($US Billion)

1

United States

2,910

11.2%

2

China

2,110

8.1%

3

Germany

2,050

7.9%

4

Japan

1,260

4.9%

5

France

1,060

4.1%

6

United Kingdom

961

3.7%

7

Netherlands

924

3.6%

8

Italy

852

3.3%

9

Belgium

727

2.8%

10

Canada

727

2.8%

11

Hong Kong

698

2.7%

12

Korea

696

2.7%

13

Spain

534

2.1%

14

Singapore

528

2.0%

15

Mexico

521

2.0%

16

Russian Federation

518

2.0%

17

Taiwan

429

1.7%

18

India

400

1.5%

19

Switzerland

314

1.2%

20

Australia

313

1.2%

21

Saudi Arabia

307

1.2%

Continued on next page

116

Table A4 (cont’d)
Trading Volume1
Rank

Country

Share
($US Billion)

22

United Arab Emirates

305

1.2%

23

Malaysia

300

1.2%

24

Thailand

292

1.1%

25

Brazil

286

1.1%

26

Austria

283

1.1%

27

Sweden

273

1.1%

28

Poland

264

1.0%

29

Turkey

245

0.9%

30

Indonesia

220

0.8%

31

Czech

207

0.8%

32

Norway

187

0.7%

33

Ireland

178

0.7%

34

Denmark

173

0.7%

35

Hungary

157

0.6%

36

South Africa

143

0.6%

37

Finland

136

0.5%

38

Iran

122

0.5%

39

Viet Nam

115

0.4%

40

Portugal

114

0.4%

41

Israel

104

0.4%

42

Slovak

101

0.4%

Continued on next page

117

Table A4 (cont’d)
Trading Volume1
Rank

Country

Share
($US Billion)

43

99

0.4%

44

Venezuela

98

0.4%

45

Philippines

98

0.4%

46

Nigeria

97

0.4%

47

Ukraine

96

0.4%

48

Argentina

95

0.4%

49

Romania

91

0.4%

50
1

Chile

Greece

84

0.3%

All volumes are the annual average during 2002-2012 and valued at
current price.
Source: World Trade Organization

118

Table A5: Share of Imported Inputs by Manufacturing Industry Classiﬁcation of Export Price
Indexes
Industry

Import Shares1 Industry

Import Shares1

Naphta

0.875

Textile

0.382

Non-ferrous Metal

0.672

Paper

0.370

Organic Chemicals

0.659

Other Electrical Equipment

0.355

Fuel

0.635

Plastic

0.355

Leather

0.617

Other Basic Iron

0.351

Non-ferrous Metal Product

0.571

Electronic Component

0.350

Computer

0.566

Apparel

0.320

Man-Made Fiber

0.541

Domestic Electric Appliance

0.318

Synthetic Resin

0.523

Spectacle

0.317

Fiber

0.502

Cement

0.308

Synthetic Rubber

0.497

Electric Motor

0.306

Fertilizer

0.493

Other Manufacturing

0.306

Semiconductor

0.482

Special Purpose Machinery

0.296

Electronic Video and Audio

0.478

Industrial Conveying

0.290

Cold Rolled Iron

0.452

Engine

0.288

Telecommunication

0.443

Industrial Refrigerator

0.284

Other Chemicals

0.438

Hand Tool

0.283

Hot Rolled Iron

0.399

Motor Vehicle

0.281

Electronic Display

0.386

Machine Tool

0.262

Oﬃce Machinery

0.383

General Purpose Machinery

0.253

Tire and Tube

0.383

Food

0.215

1

The time-averaged shares during 1995-2010.

119

REFERENCES

120

REFERENCES

[1] Amiti, M., Itskhoki, O., Konings, J., 2012. Importers, exporters and exchange rate disconnect. NBER Working Paper No. 18615.
[2] Athukorala, P., 1991. Exchange rate pass-through: The case of Korean exports of manufactures. Economics Letters 35, 79-84.
[3] Auboin, M., 2012. Use of currencies in international trade: Any changes in the picture?.
WTO Staﬀ Working Paper ERSD-2012-10.
[4] Auer, R.A., Schoenle, R.S., 2012. Market structure and exchange rate pass-through.
Federal Reserve Bank of Dallas Globalization and Monetary Policy Institute Working
Paper No. 130.
[5] Bailliu, J., Dong, W., Murray, J., 2010. Has exchange rate pass-through really declined?
Some recent insights from the literature. Bank of Canada Review Autumn 2010, 1-8.
[6] Balassa, B., 1965. Trade liberalization and revealed comparative advantage. The Manchester School 33, 99-123.
[7] Berman, N., Martin, P., Mayer, T., 2012. How do diﬀerent exporters react to exchange
rate changes?. The Quarterly Journal of Economics 127, 437-492.
[8] Bernard, A.B., Jensen, J.B., Redding, S.J., Schott, P.K., 2009. The margins of US trade.
The American Economic Review 99(2), 487-493.
[9] Bernard, A.B., Redding, S.J., Schott, P.K., 2010. Multi-product ﬁrms and product
switching. American Economic Review 100(1), 70-97.
[10] Berner, E., 2011. Exchange rate pass-through: New evidence from German micro data.
Economics Working Papers from Christian-Albrechts-University of Kiel No. 2011-01.
[11] Burstein, A., Gopinath, G., 2013. International prices and exchange rates. NBER Working Paper No. 18829.
[12] Campa, J.M., Goldberg, L.S., 2010. The sensitivity of the CPI to exchange rates: Distribution margins, imported inputs and trade exposure. The Review of Economics and
Statistics 92(2), 392-407.
[13] Campa, J.M., Gonzalez-Minguez J.M., 2006. Diﬀerences in exchange rate pass-through
in the euro area. European Economic Review 50, 121-145.
[14] Campos, C.F.S., 1998. Incomplete exchange rate pass-through and extensive margin of
adjustment. mimeo.
[15] Cho, T., 2010. An estimation of exchange rate pass-through to export prices by industry
(in Korean). Korean Economic Studies 28(9), 117-147.
121

[16] Choi, Y., Kim, C., 2001. Analysis of exchange rate pass-through (in Korean). The Bank
of Korea, Economic Analysis 7(3), 63-103.
[17] Colacelli, M., 2010. Intensive and extensive margins of exports and real exchange rates.
mimeo.
[18] Dornbusch, R., 1987. Exchange rates and prices. American Economic Review 77(1),
93-106.
[19] Eaton, J., Eslava, M., Kugler, M., Tybout, J., 2008. The margins of entry into export
markets: Evidence from Colombia. In the Organization of Firms in a Global Economy,
ed. Helpman, E., Marin, D., Verdier, T., 231-272. Cambridge, MA: Harvard University
Press.
[20] Fauceglia, D., Shingal, A., Wermelinger, M., 2012. Natural hedging of exchange rate
risk: The role of imported input prices. MPRA Paper No. 39438.
[21] Feenstra, R.C., Gagnon, J.E., Knetter, M.M., 1996. Market share and exchange rate
pass-through in world automobile trade. Journal of International Economics 40, 187-207.
[22] Gaulier, G., Lahreche-Revil, A., Mejean, I., 2008. Exchange-rate pass-through at the
product level. Canadian Journal of Economics 41(2), 425-449.
[23] Goldberg, L.S., Tille, C., 2008. Vehicle currency use in international trade. Journal of
International Economics 76, 177-192.
[24] Gopinath, G., Itskhoki, O., Rigobon, R., 2010. Currency choice and exchange rate
pass-through. American Economic Review 100(1), 304-336.
[25] Halpern, L., Koren, M., Szeidl, A., 2011. Imported inputs and productivity. mimeo.
[26] Hummels, D., Ishii, J., Yi K., 2001. The nature and growth of vertical specialization in
world trade. Journal of International Economics 54, 75-96.
[27] Hummels, D., Kejriwalz, M., Naknoix, K., 2010. Exchange rate pass-through and market
structure in multi-country world. mimeo.
[28] Hummels, D., Klenow, P.J., 2005. The variety and quality of a nation’s exports. American Economic Review 95(3), 704-723.
[29] IMF, 2010. Developing a revised manual for the export and import price indices http:
//www.imf.org/external/np/sta/tegeipi/
[30] Kim, H., Lee, H., 2009. Declines in exchange rate pass-through to export prices in Korea
(in Korean). Korea Development Institute, Korea Development Review 31, 235-266.
[31] Klau, M., Fung, S., 2006. The new BIS eﬀective exchange rate indices. BIS Quarterly
Review March 2006, 51-65.
[32] Knetter, M.M., 1993. International comparisons of price-to-market behavior. American
Economic Review 83(3), 473-486.
122

[33] Koopman, R., Wang Z., Wei, S., 2012. Estimating domestic content in exports when
processing trade is pervasive. Journal of Development Economics 99(1), 178-189.
[34] Lee, J., 1995. Pricing-to-market in Korean manufacturing exports. International Economic Journal 9(4), 1-12.
[35] Lee, J., 1997. The response of exchange rate pass-through to market concentration in
a small economy: The evidence from Korea. The Review of Economics and Statistics
79(1), 142-145.
[36] Li, H., Ma, H., Xu, Y., Xiong, Y., 2012. How do exchange rate movements aﬀect Chinese
exports?: A ﬁrm-level investigation. mimeo.
[37] Melitz, M.J., Ottaviano, G.I.P., 2008. Market size, trade, and productivity. Review of
Economic Studies 75, 295-316.
[38] Pollard, P.S., Coughlin, C.C., 2006. Passthrough estimates and the choice of an exchange
rate index. Review of International Economics 14(4), 535-553.
[39] Pierce, J., Schott, P., 2009. A concordance between ten-digit U.S. Harmonized System
Codes and SIC/NAICS product classes and industries. Center for Economic Studies
Paper No. CES 09-41.
[40] Rodrik, D., 2008. The real exchange rate and economic growth. Brookings Papers on
Economic Activity, Fall 2008, 365-412.
[41] Silver, M., 2007. Do unit value export, import, and terms of trade indices represent or
misrepresent price indices?. IMF Working Paper, WP/07/121.
[42] Tang, H., Zhang, Y., 2012. Exchange rates and the margins of trade: Evidence from
Chinese exporters. CESifo Economic Studies 58, 671-702.
[43] United Nations, 1981. Strategies for price and quantity measurement in external trade.
Department of International Economic and Social Aﬀairs, Statistical Papers, Series M,
No. 69.
[44] Yang, J., 1997. Exchange rate pass-through into U.S. manufacturing industries. The
Review of Economics and Statistics 79(1), 95-104.
[45] Yang Y.Y., Hwang, M., 1994. Price behavior in Korean manufacturing. The Review of
Economics and Statistics 76(3), 461-470.
[46] Yun. S., 2002. Problems of estimating exchange rate pass-through and theoretical explanation on currency selection (in Korean). The Bank of Korea, Economic Analysis 8(3),
76-105.

123