Import response to exchange rate fluctuations:

A firm-level investigation
Yao Amber Li

Juanyi Jenny Xu

Chen Carol Zhao

Department of Economics
Hong Kong University of Science and Technology
This Version: February 2015
First Draft: May 2013

Abstract
This paper presents theory and evidence on importing responses to exchange rate fluctuations
from highly disaggregated Chinese data. The paper first develops a heterogeneous-firm trade model
and predicts firms import responses at both extensive and intensive margins: when domestic currency appreciates, more firms start importing and more products are added into the imported
inputs bundle (extensive margin effect); the import value by each firm also increases (intensive
margin effect); those import responses are found to be more profound for firms in ordinary trade
regime than those in processing trade regime. Next, the paper presents the empirical evidence of
Chinese firms import response to domestic currency appreciation in both the short run and the
medium run and confirms the predictions from the model. The results show that the extensive margin effect dominates: an appreciation of local currency significantly increases the probability of firm
entry and products adding, which counts for a major source of the increase in Chinas aggregate
import value from 2000 to 2006. We also find that the pattern is more robust for ordinary trade
than processing trade, more profound under a fixed exchange rate regime than a managed floating
regime (both expected and confirmed). Finally, we investigate the exchange rate pass-through to
import prices and find that the incomplete pass through has declined.
Keywords: exchange rate, trade imbalance, intensive margin, extensive margin, products churning, processing trade

We thank David Cook, Hiroyuki Kasahara, the participants of the World Congress of the International Economic

Association (Dead Sea, Jordan, June, 2014), the Asian Meetings of the Econometric Society (Taiwan, June 2014), the
9th Australasian Trade Workshop (Curtin University, March 2014), the Free-Trade Zone and Chinas new stage of opendoor policy workshop (Tsinghua University, December 2013), and of seminars held at HKUST for helpful discussions.
All errors are our own.

Li: Department of Economics and Faculty Associate of the Institute for Emerging Market Studies (IEMS), Hong
Kong University of Science and Technology, Clear Water Bay, Kowloon, Hong Kong SAR-PRC. Email: yaoli@ust.hk.

Xu: Department of Economics, Hong Kong University of Science and Technology, Clear Water Bay, Kowloon, Hong
Kong SAR-PRC. Email: jennyxu@ust.hk.

Zhao: Department of Economics, Hong Kong University of Science and Technology. Email: zhaochen@ust.hk.

Introduction

How do importing and exporting firms respond to exchange rate changes? This is a central question in
international macroeconomics. This question is of greater importance recently, given the existence and
persistence of substantial global imbalance over the last two decades. Several studies, such as Campa
and Goldberg (2005) and Marazzi and Sheets (2007) suggest that exchange rate pass-through to import
prices has declined in recent years in industrialized countries. For example, Dong (2012) show that
both U.S. exports and imports have become much less responsive to exchange rate movements in recent
years and firms pricing behavior and global trade pattern may help to explain this decline. Hence, it
will be informative to study importing and exporting firms response to exchange rate changes using
firm-level data, since a micro level analysis of firm import and export decision could suggest which
factors are crucial in understanding firms behavior. There are already several studies on this issue
using disaggregate level data, but they mainly focus on the export side(for example, Berman, Martin
and Mayer (2012)). In this paper, we intend to study the response of importing firms to exchange
rate changes, using a highly disaggregated Chinese firm-level data. Since China is playing a more and
more important role in the global economy, understanding the effect of exchange rate movements on
Chinese importing firms is essential for examining the source of trade imbalance and also predicting
consequences of RMB appreciations on global imbalance. This paper provides a first step towards
our understanding on this important issue. This will also shed some light on the long-lasting policy
debase on Chinese exchange rate policy.
The reason why we want to use the disaggregate firm level data to study the import response
is twofold. First, existing literatures on the Chinese import response to exchange rate fluctuations
presents ambiguous estimates of elasticity of Chinese imports with respect to exchange rate changes.
The earlier studies find that an appreciation of RMB would increase Chinese imports (e.g.Dayal-Gulati
and Cerra (1999); Dees (2001)), and the more recent ones reached a very different finding (Lau et
al, 2004; Marquez and Schindler (2007)). Cheung, Chinn and Fujii (2010) find that estimates of USChinas exchange rate elasticity of import are inconsistent with the standard model. Overall, there is
no clear consensus regarding the impacts of real appreciation of RMB on Chinas trade balance based
on earlier studies. Marquez and Schindler (2007) concludes that the estimated response of imports
is negligible and lack of precision. This inconsistency in existing literatures may be due to the fact
that they study Chinese imports using an aggregate data at either industry level or major product
level, which leads to contradiction due to the lack of firm-level information. That perhaps explains
why Garcia-Herrero and Koivu (2009) suggest that the exchange rate policy alone will not be able to
address the trade imbalance. They argue that the real appreciation of RMB reduces Chinas trade
surplus in the long run, but the effect would be limited in the short run.
Another reason to use the firm-level data is that recent research in international trade emphasizes
the importance of firms extensive margins for understanding the overall pattern, as well as the number

of goods firms import, the number of countries from which they import. Previous literature on Chinese
import response to exchange rate changes often ignores the distinction of different margins of trade.
This arose the major motivations of our investigation of import response to exchange rate fluctuation
using detailed transaction level Chinese Customs data.
To study the response of importing firms to exchange rate changes, we model importing firms
response when facing a domestic currency appreciation using a heterogeneous-firm framework. It
predicts that, during appreciation the threshold productivity for importing decreases, which means
more firms with lower productivity begin import from abroad. The increasing number of importing
firm corresponds to extensive margin (firm level) shift. Also, the fixed cost of importing implies that
marginal profit for each imported goods rises during appreciation, so the varieties of imports increase.
It corresponds to extensive margin shift at products level. If imported goods are highly substitutable,
then the expenditure-switching effect dominate pass through effect, the intensive margin will increase
as well. However, a currency appreciation may have two-way effects on processing trade profit; this
pattern could be violated when referring to processing trade.
Our empirical part utilizes the highly disaggregated transaction-level Chinese Customs data to
evaluate the impact of exchange rate fluctuations on import. A big advantage of our dataset is the
detailed information on the value and quantity of each product at HS 8-digit level that each firm
imports from each destination country. Each transaction documents the data at the firm-productcountry level. So we can use unit value as proxy for the f.o.b. (free-on-board) import price at firmproduct-destination level. Thus we could estimate both exchange rate elasticity and pass-through at
the micro level, which complements the literature that usually only adopts aggregate volume and price
index.
This transaction-level dataset enables us to investigate in detail the shift of importing behavior at
both firm and product levels as well as at both extensive and intensive margins. Extensive margin is
defined at both firm level and products level. Firm level extensive margin is referred to the number
of firms that engaging in importing; product level extensive margin is for total number of products
imported from various destination country. In the empirical investigation, we use both Probit and
linear probability regressions to test the probability change in firm entry/exit, and product or productcountry adding/dropping in response to exchange rate fluctuations.
We find that extensive margin response dominates intensive margin response, which contributes
to majority rises on aggregate import value. This is quite different from the conclusion in previous
literature regarding importers response to exchange rate changes in other small open economies, as
suggested in Lu, Mariscal and Mejia (2012). They show that the intensive margin dominates the
extensive margin when Columbia importers faces a big real appreciation. Some additional patterns
are found from trade margins when exchange rate fluctuates. When considering both a short term
(within three months) and a long term (within four quarters) response, our tests show that importers

responses are most consistent with our prediction in the long run than the short one.
Besides import volume test, we also check price pass-through effects during RMB appreciation
since previous literature finds a decline of import price pass-through in industrial countries. Using
this disaggregated transactional level data, we find some patterns for price adjustment for Chinese
importers. We find that price adjustment is less responsive than volume change because the magnitude
of price change is rather small during currency appreciation. Also, pass-through to import prices
displays a different pattern in the short run ( i.e. within three-month) and the long run( i.e. within a
year). In the short run, price level of imports from OECD countries may decrease around 10 percent
when RMB appreciates, while this reduction of import prices disappears gradually in a long run. If
we look at import from US only, price decreases in a short run, but it even increases slightly in a long
run. We further test pass-through of exchange rate fluctuation into import price at HS6 product level.
We find short run pass-through is highly incomplete. Also there is a declining trend of pass-through
to import prices, which is consistent with findings in previous literature which are based on aggregate
level date in developed countries.
According to Manova and Zhang (2012), Chinese trade has its distinctive characteristics in term
of both trade pattern and pricing behavior. A distinguished feature is that a large proportion of
processing trade in total trade volume and the potential interaction between imports and exports,
as shown in Yu (2010). Several papers explore this issue from exporters perspective to examine
how processing trade interacts with export price and quantity responses, such as Thorbecke and
Smith (2012), Liao, Shi and Zhang (2012), Koopman, Wang and Wei (2012), Xing (2011) and Ahmed
(2009). On the other hand, not only trade volume but also price behavior could be influenced by
the two-way trade (both importing and exporting). For example, Amiti, Itskhoki and Konings (2012)
find that firms varies in exchange rate pass through due to various import shares and market shares.
Another feature of Chinese trade lies in that trade performed by foreign owned firm and joint venture
firms takes a significant portion of total trade value in China. A transfer pricing issue may also affect
importers or exporters pricing decision as typical trade theories will predict. Taking all these features
of Chinese trade into consideration, we have good reasons to believe that import response and price
pass through could be rather heterogeneous in China, although little previous research has explored
that.
To explore how these features affect our result, some tests are conducted. To avoid the influence
of processing trade, we test our predictions for ordinary trade and processing trade separately and
confirm that the responses of ordinary traders are more profound than those of processing traders.
Also, to control transferring price issue associated with foreign invested firms, we conduct a robustness
test for different ownership type firms. By this way, we could compare responses of foreign invested
firms to non foreign invested firms. We find that that the ownership structure does not affect our
main conclusions about the response of Chinese import to exchange rate changes.

Another important event in Chinas exchange rate policy change is that on July 21, 2005, China
announced a move from fixed exchange rate regime to a managed floating exchange rate regime.
Afterwards, a series of appreciation of RMB against U.S. dollars took place. Hence, Chinas exchange
rate reform offers a unique laboratory to test trade response to policy change. So in this paper we also
investigate the importing firms response to real exchange rate changes before and after this regime
shift.
Interestingly, as suggested by the change of RMB-USD forward exchange rate, this exchange rate
regime reform is not a one-shot event, but a systematic process. Before 2003, a pegged to USD
exchange rate regime had been prevalent for several decades. However, since 2003 China has been
under great pressure for RMB appreciation, market expectation has changed accordingly. The shift
of expectation could be seen from the increase in forward rate between USD/RMB since the fourth
quarter of 2003. During this period, there was national wide debate and discussion about coming
reform. Uncertainty aroused for when and how a reform could be performed. This situation lasted
till July 2005, when China government announced a managed floating policy would be adopted and
gradually appreciation of RMB was foreseen then.
Intuitively, during different phases of exchange rate reform, firms may respond differently. Expecting a future uncertain appreciation, the firm may or may not want to delay its import, depending
on the uncertainty and the expected exchange rate changes. These response patterns have important
implications for exchange rate policies. Thus, a test is conducted in our paper. In this test, we segment
the overall sample periods into three; before mid 2003, 2003-2005, and after July 2005. They represent three different stages in exchange rate regime switching in China. We analyze import response
to exchange rate for each specific phase. This test suggests that firms response differently facing a
fixed exchange rate regime, an expected appreciation, and a confirmed appreciation regime. We find
that firms importing behavior is more consistent with theoretical prediction in a fixed exchange rate
regime than an expected and confirmed appreciation regime. We also examine import responses of
firms with different ownership to exchange rate changes and confirm the robustness of these results.
Our study is related to several branches of literature. The first one is the international macroeconomics literature using an aggregate level import (export) data and focusing on exchange rate elasticity
of import (export) price, e.g., Campa and Goldberg (2002);Campa and Goldberg (2005); Marazzi and
Sheets (2007), and Dong (2012). In Campa and Goldberg (2005), for example, they find a partial
pass through of exchange rate into import price in short run and a dominant effect in the long run,
especially for countries with high exchange rate volatility.
The second branch of literature uses a dis-aggregate level data set but focuses on export side. For
example, Berman, Martin and Mayer (2009) examines the heterogeneous exporters adjustments in
prices and export volume in response to exchange rate movements using French firm data. Within
this branch, there are also several papers exploring the response of Chinese exporters to exchange rate

changes, such as Yu (2009), Li et al. (2012) ,Tang and Zhang (2012) and Thorbecke and Smith (2010).
Li et al. (2012) find that exchange rate movements have a small and insignificant effect on export
quantity, and a significant and large pass-through to export price in destination currency. They also
test heterogeneous response for Chinese exporters. The most productive firms may be able adjust
their mark-up against a RMB appreciation rather than change export volume or exit from exporting
market. Our paper, instead, focus on the response of importers using Chinese custom firm-level data.
The third branch of literature uses a disaggregate level of data and focuses on importers behavior.
For example, Gopinath and Neiman (2014) explores the mechanism of trade adjustment during the
Argentine crisis 1996-2008. They find that within-firm input churning or the sub-extensive margin,
rather than firm level extensive margin, plays a significant role in import collapse during crisis in
Argentine. Also, Lu, Mariscal and Mejia (2012) use Columbia trade data and find that firm select
imported varieties and reorganize their imported inputs and production over time when exchange
rate appreciates. Our study focus on unique response of Chinese importers under a change of trade
climate. In China, in contrast, when RMB appreciates, the most responsive change comes from
extensive margin at firm level. By decomposing the change in import response to new firm entrants,
new product adding, and the net import value increase by each firm, we find the first factor contributes
most to aggregate change of import value.
Our study is also related to the studies focusing on different margins of trade. For example, Chaney
(2008), Arkolakis (2010), Eaton, Kortum and Kramarz (2011). Bernard et al. (2009) find evidence of
extensive margin accounts for a larger share of variation in import and export across countries. Also,
Hummels and Klenow (2005) argues that extensive margin at product bundle level plays an important
role in trade value.
The rest of paper is organized as follows. Section 2 builds a simple model to capture import
response to exchange rate fluctuation at both extensive and intensive margins as well as the difference
between processing and ordinary traders. Section 3 describes data and measurements and offers a short
description of changes in term of both extensive and intensive margin. Section 4 provides detailed tests
to verify the predictions from the model. Section 5 presents robustness checks for different groups of
firms by ownership and different stages in exchange rate regime reform in China. Section 6 concludes.

A Simple Model

2.1

Firms Production Problem

In this part we build a simple model to examine firms import response to exchange rate fluctuations
by extending Gopinath and Neiman (2014). We first consider an ordinary trade firm, while at the end
of this section, we compare processing trade firm with ordinary trade firm.
It is assumed that firm i draws its productivity from a uniform distribution with support (0, Amax ),
6

and follows the production function of

Yi = Ai (Ki Li1 )1 Xi

(1)

Given its productivity Ai , firm i chooses capital input Ki , labor input Li and intermediate input
bundle Xi , which is a CES aggregate of domestic intermediates Zi and imported intermediates Mi ,
with a elasticity of substitution . That is, the intermediate input bundle is defined as
1

Xi = [Zi + Mi ] , where < 1

(2)

The price of final intermediate input bundle PXi is given by equation (3)

1
PXi = (PZi
+ PM1
i )

(3)

For simplicity, we normalize domestic intermediate input price to one, and assume PM i 1.1 Then
we have

PXi = (1 + PM1
i )

(4)

Therefore, if a firm only uses domestic intermediate inputs, the input price index is one; if it uses
imported intermediate inputs, the price index is less than one.
The imported intermediate input bundle is assumed to be an aggregation of different imported
varieties j, j [1, N ]. If the price of variety j is pmj , the price index of imported intermediate bundle,
PM i , follows
PM i

Z
=[

j=1

1
pmj
]

, < 1

(5)

where is the elasticity of substitution across imported foreign varieties.

From equation (5), price index PM i is a function of the number of varieties imported, N, and
price level of imported varieties, pmj . Generally speaking, price index of imported bundle PM i is a
decreasing function of the number of varieties imported. To see this, we consider a simple case where
price of all imported input varieties are identical. Thus we can use price pm to denote the price for
all varieties. The price index for imported input bundle is the simply PM i = N

pm . Since < 1,

PM i decreases in N.
Imported input price PM i is assumed to be a function of exchange rate of domestic currency,
PM i (e), where the exchange rate e is defined as the price of domestic currency in terms of foreign
currency.2 Hence an appreciation of domestic currency implies an increase in exchange rate e. Thus,
1

This assumption ensures PXi , the price of intermediate input bundle is less than one. Intuitively, if PXi is larger
than one, i.e. the normalized domestic intermediates price, producers have no incentive to use imported intermediates.
2
In this simple model, we assume that not all export firms in the exporting country choose local currency pricing
when setting export prices. This implies that exchange rate fluctuations will always be passed through to import price
in destination countries. In other words, the price of imported intermediates will change when exchange rate changes.

we have

PM i
e

< 0, and appreciation of domestic currency will lead to decrease of import prices.

Consequently, a reduction in PM i generates a decline in PXi . Therefore, an appreciation of domestic

currency reduces the price index of intermediate inputs as long as firm i imports from abroad.
Pxi
<0
e
Overall, unit cost of goods produced by firm i is
1

Ci =

PV PXi
1
, where PV = (1 )1 r w1

1
(1 )
Ai

(6)

Since our model is a partial equilibrium one, capital price r and labor price w are both exogenously
determined. We assume they are constant, implying that the price of non-intermediate inputs PV
is identical for all firms. Under this framework, heterogeneity of cost only comes from firms own
productivity Ai and intermediate input bundle PXi. . Define =
for firm i is given by
Ci =

1
P 1 ,
(1)1 V

the cost function

PXi
Ai

(7)

Thus, within the same sector or industry, we only have to focus on heterogeneity in Ai and PXi .

2.2
2.2.1

Firms Import Decision

Demand Side

We assume a downward sloping demand function for goods produced by firm i,

Yi = oPi , where > 1 and o is a constant

(8)

where Pi is the price of good i. Profit maximization implies that, firm i sets a constant mark-up over
its unit cost Ci
Pi =
2.2.2

Ci
1

Firms Problem

The profit for firm i is denoted by i , which equals revenue minus fixed cost of production. If a firm
imports, it incurs an additional fixed cost of importing,Fimp , and a variable import cost fimp (N ),
which depends on N , the number of varieties imported. It is assumed that fimp (N ) increases with N,
and is a convex function of N.
i = Yi Pi Yi Ci Fimp fimp (N )

PV PXi 1
1 1
1
i = o (
)
Fimp fimp (N )
) (
1
(1 )1
Ai
i = (

PXi
1 1
)1 Fimp fimp (N ), where = o(
)
is a constant
Ai
1

(9)

Taking logarithm to the first term of the profit function above, we can get the following equation 10
( using lower case letter to denote logarithm).
R = (

PXi
)1
Ai

+ ( 1)(ai pxi ) where

= log()
r=
2.2.3

(10)

Extensive Margin Decision at Firm Level

In our model, a firm faces a trade-off between reducing production cost by importing more from abroad
and potential incurred cost of importing. The productivity threshold of importing can be solved from
zero profit condition. Facing a variable import cost, the more varieties a firm imports, the more cost
it must pay. The cut-off value of productivity for import, ai is given by the following equation.
ai =

log[Fimp + fimp (N )]
+ pxi
1

(11)

Since pxi decreases when domestic currency appreciates, and the cut-off productivity decreases consequently.
e , ai , i.e.,

ai
<0
e
0

This implies that the mass of importing firms shifts from (ai , amax ) to (ai , amax ), where ai < ai , and
more firms start to import after a currency appreciation.
Proposition 1. When domestic currency appreciates, more firms start to import from abroad, which
suggests an increase in extensive margin of import at firm level.
2.2.4

Extensive Margin Decision at Product Level

How do imported varieties respond to an appreciation? As shown in Section 2.1, the price index of
intermediate imported input PM i decreases with the number of imported varieties N , i.e.
Thus the price index of intermediate input bundle PXi also decreases in N ,

(N )
i.e., PXi
N

PM i (N )
N

< 0.

< 0.

Given productivity ai , firm i chooses optimal N to maximize its profit function

N = arg max[(
N

PXi
(N ) 1
)
Fimp fimp (N )]
Ai

(12)

Due to cost reduction for using imported input from abroad,3 the marginal benefit (MB) of using
3

To see this, consider a simple case: pmj = pm , PM i = N

pm (e).

PM i (N )
N

1
1
N

< 0. Therefore,

i
N

> 0.

imported intermediate input is positive.

Consequently, under domestic currency appreciation, PXi , the price level of intermediate input
bundle, decreases further due to imported varieties.4 In this way, MB curve shift upward during
currency appreciation.
Under our assumption, it incurs a variable cost fimp (N ) to import from abroad. Further more,
fimp (N ) increases with N, and is a convex function of N. That is, the marginal cost MC(N) of importing
intermediate input is positive, and MC(N) is assumed to be increasing in N.
From equation (12), the optimal number of imported variety N , becomes the intersection of
marginal cost (MC) and marginal benefit (MB) equations. Given that the slope of MC curve is
larger than the slope of MB curve,5 when the marginal benefit curve shifts upwards under a currency
appreciation, it leads the N to a larger one, which stands for an increase in the of number of imported
varieties for an importing firm.
Proposition 2. When domestic currency appreciates, a firm tends to import more varieties from
abroad, which suggests an increase in extensive margin of imports at product level.
2.2.5

Firms Intensive Margin Decision

Given its choice of importing or not and the number of importing varieties, firm i minimizes its cost by
choosing the optimal composition of domestic input Zi and imported input Mi to produce intermediate
input. In this section, we focus on the response of intensive margin to exchange rate fluctuations, so
we will treat Mi as a single variety of imported input with price PM i .

min[Pzi Zi + PM i Mi ]

(13)

Z,M

s.t.[Zi + Mi ] = 1
Solving the cost minimizing problem yields optimal input of both imported and domestic intermediate
input Mi and Zi to produce one unit of intermediate input bundle.

Mi = (1 + PM1
i )

(14)

Zi = PM1
i Mi

Since

PM i
e

< 0,

Mi
PM i

< 0 and < 1, we have

Mi
e

(15)

> 0 , implying imports from abroad Mi should

rise as exchange rate increases.

e , PM i (e) , Mi (e)
4
5

i (N )
is affected by e. From previous simple case where pmj = pm , PM i = N
Note that PM
N
The condition is explained in detail in Appendix

10

pm (e).

If we define the expenditure on imported intermediate input as Cm , we can see that it is a function
of exchange rate e.

Cm (e) = (1 + PM1
i )
Since

PM i
e

PM i (e)

(16)

< 0, imported value Cm is an increasing function of exchange rate e, given that substitution

elasticity between domestically produced and imported intermediates is high and the price of the
imported intermediates is not too low. In numerical analysis, we show that this condition hold.6 It
implies that if domestic currency appreciates, expenditure on imports increases. That is
e , PM i (e) , Cm (e) if satisfies certain condition
Proposition 3. When domestic currency appreciates, the intensive margin of import, i.e. expenditure
of each imported variety increases.
It should be noted that our model is only a simple description of import behavior in response to
exchange rate fluctuation. We did not take changes in the pricing behavior of foreign exporters or
importers after exchange rate changes into consideration. So our model con not provide predictions
about exchange rate pass-through on import prices. Nevertheless, in the empirical part, we will
investigate the import prices changes in response to exchange rate fluctuation, since this constitutes
an important channel through which exchange rate changes affect import.

2.3

Global Value Chain

Up to now, our model focuses on ordinary trade firms. In this subsection, we explore a scenario when
firms engage in global value chain and perform processing trade. In this scenario, a processing trade
firm may engage in purely assembling using imported intermediates provided by foreign partners; or it
may import intermediates inputs and produce final product to export. In both cases, facing exchange
rate changes, the marginal benefit of importing form firms will be different from those for ordinary
trade firms. In the first case, or the so-called pure assembly trade case, there is little change in
input cost since the intermediate input is directly provided by foreign firms. In this sense, there is no
incentive for Chinese assembly firms to change import value when facing exchange rate fluctuations.
However, if a firm engages in processing trade and import intermediate input itself, then the import
response to exchange rate fluctuation would be different.
We use a simple model to illustrate it. We assume the firm i use imported and domestic intermediate goods to produce export goods. The setting regarding price and quantity of domestic and

It must satisfy the following condition: (1 + PM1 )( 1) > 1. This condition is satisfied if > = arg{ =

PM

}. The LHS of equation inside the brace is increasing in , so is the RHS. In our numerical analysis, we can

1+PM

verify that, as long as PM is reasonable large within the range of [0,1], and is sufficiently large, the above condition
can be satisfied. In Gopinath and Neiman (2014), they set to be around 0.75 in calibration. Given this value of , the
above condition is satisfied.

11

imported intermediates are the same as the model above. So the cost of the export good is also
Ci =

PV1 PXi
1
,
Ai
(1)1

where PV and PXi is defined similarly as the ordinary trade model.

We also assume the processing trade firm i has a foreign partner. The foreign partner services
consumers abroad, and entails an additional outlay cost, T , for marketing purpose and managing a
distribution network. T may also include up-front sunk cost paid by firm is foreign partner. The
foreign partner and the processing-trade firm i engage in a Nash-bargaining with bargaining weights
corresponding to their contribution to the collaborated production. To be specific, we assume that
these weights reflect the share of total costs borne by each side. Under this setting, processing trade
firm would only reap a proportion of the total profit i , and the proportion depends on its input
share.
To get notation for the profit, we assume the in the international market, there is a downward
demand for goods produced by firm i.
Yi = o(Pi )
where Pi is the foreign currency price of good i. Since we have flexible price, the price faced by
foreigners is the same no matter firm uses producer currency pricing or local currency pricing.7 Suppose
firm i follows producers currency when setting export price, given that pi =

1 Ci .

So it sets exporting

price as
Pi = ePi = e

Ci
1

(17)

Hence, the profit of processing trade firm i get from production is given by
iP T = Yi Pi Yi Ci Fimp fimp (N ) T
iP T = e o(

1
)
C 1 Fimp fimp (N ) T
1
1 i

(18)
(19)

As mentioned above, the profit for processing trade firm i receives is proportional to its input share.
1
) 1 1 , profit for processing
Using the notation from the ordinary trade model, define = o( 1

trade firm receives i becomes

i = [e (

u
Ci + fimp (N ) + Fimp
Pxi
)1 Fimp fimp (N ) T ](
)
Ai
Ci + fimp (N ) + Fimp + T
Pu

If we define P T = e ( Axii )1 Fimp fimp (N ), and B =

Ci +fimp (N )+Fimp
Ci +fimp (N )+Fimp +T ,

(20)

where item B

denotes the share of inputs contributed by processing firm i. Then profit for processing-trade firm i
follows i = P T B. We denote profit of a parallel importing firm j who engages in ordinary trade
as OD . It becomes

i
P T
B P T
=
B+

e
e
e

(21)

If firm i adopts local currency pricing, the optimal price set by firm i is given by pi = e 1
Ci . So the demand follows

Yi = o(e 1
Ci ) . So the profit in turns of domestic currency is =
profit denoted in domestic currency is the same in 20.

12

Pi Yi
e

Yi Ci Fimp fimp (N ) T . Hence, the

We can show that

OD
Pu
OD
P T
= e1 ( xi )1 + e
< e
e
Ai
e
e

(22)

and
B
Ci
< 0, since
<0
e
e
Firstly, due to the presence of e in the revenue function, compared to the parallel ordinary trade
firm, the profit increase of processing firm is smaller, as show by expression (22). Intuitively, this
is because when domestic currency appreciates, export goods i becomes more expensive and abroad
demand decreases, which in turn reduces profit gain of firm brought by cost reduction after exchange
rate appreciation. Secondly, as exchange rate increases, for processing trade firm i, cost reduces and
profit rises, while bargaining power of firm i decreases as input share B decreases. So the profit increase
after a currency appreciation shrinks for the processing firm.
Hence, due to the above two effects, the response of importers to exchange rate fluctuations on
processing-trade firms should be weaker than in ordinary trade firms. Particularly, if the two adverse
effects are large enough, import responses of processing trade firm to exchange rate fluctuation may
become ambiguous. In summary, we have the following proposition.
Proposition 4. When domestic currency fluctuates, processing trade firm tend to have less or even
ambiguous import response compared to ordinary trade firms.

Data

Our empirical investigation is building upon highly disaggregated trade data on Chinese importing
firms and their imported products as well as bilateral exchange rates between China and its trading
partners.

3.1

Transaction-level trade data

The import data come from the Chinese transaction-level trade data, maintained by Chinas General
Administration of Customs. This database records monthly data of all transactions of Chinese exports
and imports between 2000 and 2006, including import and export values, quantities, quantity units, HS
8-digit product classification, firm identity information, trade destinations/origins, type of enterprises
(e.g. state owned, domestic private firms, foreign invested, and joint ventures), and customs regime
(e.g. Processing and Assembling and Processing with Imported Materials). Specifically, import
data from 29 OECD countries are included in our empirical investigation, which accounts for a majority
part of Chinas total import value (approximately 54-60 percent) from 2000 to 2006.8
8

The sample countries include Australia, Austria, Belgium, Canada, Czech Republic, Denmark, Finland, France,
Germany, Greece, Hungary, Iceland, Ireland, Italy, Japan, Luxembourg, Mexico, Netherlands, New Zealand, Norway,
Poland, Portugal, Republic of Korea, Spain, Sweden, Switzerland, Turkey, the United Kingdom, and the United States.

13

3.2

Exchange rate data

Nominal exchange rates

Nominal exchange rates are obtained from the Bloomberg Daily Exchange Rate Database. As
we only need monthly data, the average monthly exchange rates are calculated as mean of spot daily
exchange rates for that month. The Bloomberg Daily Exchange Rate Database covers all spot exchange
rates between China and her 29 OECD trading partners who use 16 currencies, including USD, AUD,
EUR, CAD, DKK, HUF, NZD, GBP, JPY, SEK, KRW, NOK, CHF,MXN, PLN, and CZK.9
Real exchange rates
We use real exchange rates instead of nominal exchange rates in the empirical tests of import
response to exchange rate fluctuations. To construct real exchange rates, monthly consumer price
index (CPI) is obtained from IMF website. CP I and CP Ic stand for consumer price index of China and
of the corresponding import partner country c, respectively. We use ec and Ec to denote real exchange
rate and nominal exchange rate between home country, China, and foreign country c, respectively. Ec
is defined as the price of domestic currency in terms of foreign currency of country c, for example, EU S
was 0.125 in 2006, i.e., one Chinese yuan was worth 0.125 USD in 2006 according to official nominal
exchange rates. Under this definition, an increase of Ec stands for an appreciation of domestic currency
against foreign currency. Then bilateral real exchange rate is given by
ec =

Ec CP I
CP Ic

To be consistent with the Customs trade data, real exchange rates in our analysis cover the period
from January 2000 to December 2006. We use a log difference to measure the change in bilateral
real exchange rate between China and country c during a certain time interval at either monthly or
quarterly frequency:

ect = log

3.3

Ec,t
Ec,t1


+ log

CP It
CP It1


log

CP Ic,t
CP Ic,t1


.

Chinas exchange rate reform and currency appreciation

Before July 2005, China followed a fixed exchange rate regime with RMB pegged to US dollar. In July
2005, China announced to adopt a managed floating exchange rate regime to replace the fixed regime.
Under the managed floating regime, based on market supply and demand, exchange rates of RMB
against USD are set with reference to a basket of foreign currencies. Figure 1 depicts fluctuations
of both nominal and real exchange rates of RMB against USD from 2000 to 2006. As illustrated in
Wo choose those countries because they are among the top trading partners of China and also because of the availability
of the bilateral exchange rates.
9
Note that some countries share a common currency, such as EU countries.

14

Figure 1, nominal exchange rates between RMB and USD did not change before 2005, and fluctuations
of nominal exchange rates only occur after the regime shift in 2005. Nevertheless, the fluctuations of
real exchange rates of RMB against USD are observed over the whole sample period. The increasing

.115

.12

.125

.13

.135

trends imply an appreciation of RMB against USD in both nominal and real exchange rates.

2000m1

2002m1

2004m1
DATE

2006m1
real

nominal

Figure 1: Nominal and Real Exchange Rate Fluctuation

Stylized facts of Chinas import

In this section, we present some stylized facts of Chinas import, including import value, import
frequency, and the changes in import margins.

4.1

Aggregate import value and its decomposition

Using transaction level Customs data, we aggregate import value by each firm-product-country triplet,
i.e., we collect all import transactions by each individual firm who imports a specific HS 6-digit product
from a certain source country during each month and aggregate to one observation as a firm-productcountry combination.10
Chinas trade volume has increased dramatically since China joined the WTO in Dec 2001: Chinas
export and import value in 2000 were approximately 266 billion and 243 billion US dollars, respectively,
while only after 5 years, these figures reached 969 billion (increased by 264%) and 791 billion (increased
10

Note that we adopt HS 6-digit, instead of 8-digit, product classifications in order to concord them consistently over
time based on the conversation table from the UN Comtrade since China adjusts her HS 8-digit product codes from time
to time.

15

(a) Aggregate Import Value

(b) Number of Product-Country

(c) Number of Importing Firms

Figure 2: Time trend of import value and decomposition

by 226%) US dollars in 2006. Figure 2(a) presents monthly total import value from Jan 2000 to Dec
2006. The aggregate import value shows a significantly increasing trend over time, especially after
the accession to the WTO in Dec 2001, though fluctuations are also observed from time to time. To
decompose the aggregate import value, we also plot the number of importing firms and the number
of product-origin country combinations are imported (see Figures 2(b) and 2(c)). Figure 2 shows
that both number of importing firms and the number of product-country bundles present a steady
increasing trend over time, jointly contributing to the phenomenon that aggregate import value has
soared after the WTO accession.

4.2

The margins of Chinas imports

It is important to understand the contribution of different margins to trade (e.g., Chaney, 2008;
Bernard et al., 2009, among others). We thus follow the method in Bernard et al. (2009) to decompose
aggregate import value into different margins and to examine both cross-sectional and time-series
variations in Chinas imports.
Cross-sectional decompositions.Chinas aggregate import with partner country c (denoted
by xc ) is decomposed into the number of firms who import from that country (fc ), the number of
all products imported from the country (Nc ), and the average value of imports per firm-product,
xc /(fc Nc ). However, it should be noted that the term xc /(fc Nc ) is not the intensive margin, because
the full firm-product bundle fc Nc may not all be imported from abroad. Therefore, to account for
density of actual trade (Bernard et al., 2009), we shall introduce an additional term dimp which is
defined as the fraction of all possible firm-product combinations that is non zero for a given sample
period, i.e., dimp =

oimp
fimp pimp ,

where oimp is the total number of firm-product observations that have

non-zero import value. Then intensive margin x

can be calculated as

16

ximp
oimp .

Firm only imports a small

fraction of the total product bundles, so trade density dimp is usually negative related with the total
number of importing firms fimp or the number of imported products pimp . We can now decompose
import value according to the following equation:
ximp = fimp pimp x
dimp ,

(23)

where fimp is the firm-level extensive margin; pimp represents the product-level extensive margin; x
is
the intensive margin; dimp is the trade density. By taking logarithm on both sides of Equation (23),
the total import value is divided into different margins as well as import density.
We then apply a similar approach as in Bernard et al. (2009) to compute the relative contribution of
each margin to total import. We regress the logarithm of each trade margin (see the terms at the righthand side of Equation (23)) on the logarithm of total import value at the left-hand side of Equation
(23). The coefficients sum to unity, with each coefficient representing the share of the overall variation
in total import value explained by each margin. For example, the coefficient on fimp represents the
contribution share of firm-level extensive margin in explaining aggregate value of imports. We find
that the shares of aggregate import value being explained by average firm-level extensive margin,
product-level extensive margin, and intensive margin are 0.42, 0.37, and 0.53, respectively while trade
density takes the weight of the rest -0.35. If we include two types of extensive margins together, their
total contributions outweigh intensive margin in explaining the variations in aggregate imports. This
is consistent with the finding in Bernard et al. (2009) using US data.11
Time-series decompositions.We also follow Bernard et al. (2009) to decompose the timeseries variations in Chinese imports between periods. We again decompose total imports to two types
of extensive margins (firm entry and exist and continuing firms adding and dropping of countryproducts) and one intensive margin (continuing firm-product-countrys growth and decline). Table
1 reports the results and shows that the changes in Chinas imports due to two types of extensive
margins also dominate the changes from intensive margin.
[INSERT TABLE 1]

4.3

Import Frequency Analysis

At this subsection we look at frequency of import adjustment, such as how long (in terms of the
number of months) a firm continually imports from abroad, imports the same product from abroad,
and how many times it changes its major import product. If most firms are continuing importers,
we are more confident later when we use monthly data to analyze the short-run import response to
exchange rate movement.
11
In Bernard et al. (2009), they find that 0.58 of US aggregate import value comes from extensive firm margin, 0.54
from product churning, and 0.318 for intensive margin.

17

Table 2 presents the summary statistics of import frequency. Columns 1 and 2 shows that within
one year how many months firms import continually. We list percentage of firms by the duration in
their continuous importing. We find that most firms import very frequently over a year: more than
71% of firms import in all twelve months within a year. In other words, seasonal importing firms
who only import for a few months in a specific season only constitute a small percentage among all
importers.
[INSERT TABLE 2]
In columns 3 and 4 of Table 2, we report how many months firms consistently import the same
HS4 product within one year. The results show that firms also tend to import the same HS4 product
frequently. This will also alleviate the concern for the short-run analysis of product-level extensive
margin later. The mean interval of importing the same product is over 8.86 months, suggesting that
most Chinese importers are importing the same HS4 product for at least three out four quarters within
one year. We also investigate how many times importers change their major product in columns 5 and
6, where major product is defined as the HS4 product with the largest import value within a given
month. The results show that about 20 percent of firms do not change their major product within
one year, and the majority of importing firms switch major product less than three times in one year.

Import response to exchange rate fluctuations

In this section, we investigate firm import response to exchange rate changes and report results on
both extensive and intensive margins based on micro-level data. In extensive margin regressions, we
further unfold extensive margin at both firm and product levels. In intensive margin, we test both
import value and price changes for a specific firm-product-country bundle. Our approach is similar to
Tang and Zhang (2012) in their test for adjustment of export margins.
In reality, importing firms response in our data set and exchange rate fluctuations may not occur
at the same pace. For example, there is lag between firms decision and import reporting to the
Customs. Also firms may adjust import between months to keep the inventory at a constant level.
Thus, we test our model predictions for both short run (monthly interval) and long run (quarterly
interval). Later in the robustness checks we will also report results using yearly interval as long run.
In our model, an appreciation of domestic currency implies a reduction of import cost. Consequently, firms tend to enter importing market and to import more varieties than before. If substitution
of imported inputs is high, the increase in import quantity may offset the price reduction, which leads
to an increase of total import value.
If our prediction is correct, we expect a significant positive coefficient of exchange rate for both
extensive margin and intensive margin, and a negative coefficient of import price at intensive margin.
18

Also, this pattern could be more stable in a longer horizon than a short term, due to a time lag in
firms responses.

5.1
5.1.1

Import responses at extensive margins

Firm entry/exit

We first test extensive margin at firm level by examining the effect of exchange rate fluctuations on
firms probability of entry or exit using a Probit regression. To be specific, we set entry and exit
dummy as the dependent variable in our regressions. We define that entry equals one if a firm imports
in time t but not imports in t 1; entry is set to be zero if a firm imports in both time t and t 1,
where time t could be either monthly or quarterly.
As this test is for firm entry/exit probability, exchange rate is accordingly calculated at firm level,
defined as the sum of weighted exchange rate among all the trade partners of the firm. To address
the potential endogeneity issue using firm-level exchange rates, we adopt two alternative ways to
construct firm level exchange rates. First, we set country weight to be the share of total import value
from one country in this firms aggregate import value from all its trade partners over the whole
sample period. This could avoid the changing weight issue due to the changes in trading partners over
time. In an alternative way to construct firm-level exchange rates, we use real exchange rates between
USD and RMB as firm-level exchange rates because the majority of Chinese trade transactions are
denominated in US dollar. In both ways in computing the firm exchange rates, the weights for different
trade partners are constant over our sample period. Therefore, the constructed firm-level exchange
rates alleviate the concerns that an importing firm adjusts its trade partners according to the exchange
rate fluctuations.
The regression equation of firm entry is given by
Pr(Entry = 1)i,(tn,t)

3
X
= (
k1 ei,(tkn,tk) + z1 Zit + g1 gt + Fi + Ft ),
k=0

where eit is firm level exchange rate fluctuations, Zit is an export dummy to indicate whether firm i
engages in two-way trade (i.e., export and import at the same time interval) and gt is GDP growth
to control for demand changes in domestic market. F are a set of fixed effect terms, including firm
fixed effects Fi and time fixed effect Ft . In Probit regression we only include time fixed effects while
in linear probability regressions we include both firm fixed effects and time fixed effects.
To take into account potentially sluggish import responses to exchange rate shocks, as postulated
by the standard arguments for the J-curve response, we exploit the high frequency nature of the
P
data and design the test in the following way. Elasticity of exchange rate 3k=0 k1 is a sum of both
contemporaneous coefficient of exchange rate fluctuations (monthly or quarterly) and three lagged
coefficients of exchange rate fluctuations. This approach is often used to test price adjustments to
19

exchange rate fluctuations, e.g. Campa and Goldberg (2005). In our test, we distinguish long run
and short run responses. We define long run as quarterly changes and short run as monthly changes.
In the long-run test, variables (including entry dummy) capture quarterly adjustment, i.e. n=3. In
this case , exchange rate fluctuation etk3,k is quarterly change covering three months. While in the
short-run regression, variables capture monthly adjustment and exchange rate fluctuations are also
defined at monthly level, i e., n=1.
Similarly, we test firm exit decision from importing markets during the period t according to the
following equation:
3
X
Pr(Exit = 1)i,(tn,t) = (
k2 ei,(tkn,tk) + z2 Zit + g2 gt + Fi + Ft ),
k=0

where Exit is set to be one if firm i imports in time t 1 but not in time t; it equals zero if firm i
continues to import in both time t and t 1.
Table 3 shows the baseline regression results for firm entry probability using the weighted firm-level
exchange rates. For OECD countries, the coefficients on exchange rate fluctuations are significantly
positive across all specifications in both Probit and linear probability estimations. The positive coefficient suggests that it is more likely for firms to overcome fixed costs and to import from abroad
when domestic currency appreciates. The net coefficient for long run is obtained by adding coefficients
over four smooth-moving regressions, similarly short run coefficient is the sum of monthly coefficients
over four continuing monthly regressions. In contrast, the significantly negative coefficients for firm
exit suggest that it is unlikely for firms to exit from importing market when facing domestic currency
appreciation. The results here are consistent with Proposition 1 in our model. We also calculate
marginal effect of exchange rate fluctuations on import decisions of firm entry/exit. In the long run,
a 10 percent RMB appreciation improves probability of firm entry by 0.02 percent, and reduces probability of exit by 0.02 percent for OECD countries. In the short run, probability of entry increases by
0.007 percent, and probability of firm exit reduces by 0.04 percent for imports from OECD countries.
[INSERT TABLE 3]
In Table 4, we report the results using the alternative USD/RMB approach in calculating firmlevel exchange rates. In the long run, the marginal effect of exchange rate fluctuations show that a
10 percent RMB appreciation improves the probability of firm entry by 0.05 percent, and reduces
probability of exit by more than 0.33 percent for imports from OECD countries. In the short run,
the effects are less significant, but still show the predicted signs as in Proposition 1. Comparing with
previous results in Table 3, two approaches in the calculation of firm-level exchange rates yield similar
results, and the only difference is that coefficients are larger in the second test than the one using
weighted firm-level exchange rates. Generally speaking, the firm-level extensive margin responses
follow our predictions (see Proposition 1) in both long run and short run, which contributes to the
20

aggregate increase of import value after domestic currency appreciation.

[INSERT TABLE 4]
5.1.2

Product adding/dropping

According to the model prediction in Proposition 2, firms add more product varieties into the import
set or import from more foreign countries while drop less product varieties or stop importing from
less countries when facing a currency appreciation. We thus test product adding and dropping using
Probit and linear probability regressions similar to the previous tests for firm entry/exit. We classify
product variety at HS 6-digit level, and define that a HS-6 product from different origin countries as
different varieties. In other words, according to HS-6 product and country origin, we construct a new
product-country bundle to test product churning effect. We use the following regression equations:
3
X
k1 ec,(tkn,tk) + z1 Zit + g1 gt + Fihc + Ft )
Pr(Add = 1)ihc(tn,t) = (

(24)

k=0

Pr(Drop = 1)ihc,(tn,t)

3
X
= (
k2 ec,(tkn,tk) + z2 Zit + g2 gt + Fihc + Ft )

(25)

k=0

where i, h, c, t represent firm, HS6 product, country and time (month or quarter), respectively. ect
stands for exchange rate changes between country c and domestic country at time t. We use dummy
variables Add or Drop to capture firm is adding/dropping a specific product h from country c at time
t. To be specific, Add equals one if a product appears in period t but not in previous period t 1,
and zero otherwise; Drop equals one if a product appears in period t 1 but not in period t. For
control variables, we include two-way trade dummy Zit and GDP growth rate gt to control for the
domestic demand. Firm-product-country fixed effects and year fixed effects are also included in the
linear probability regression.
[INSERT TABLE 5]
The results at both quarterly and monthly intervals are reported in Table 5. In columns 1 and 2
of Table 5, we notice a rise in exchange rates (domestic currency appreciation) has a positive impact
on product adding. In columns 3 and 4, there are significantly negative coefficients on exchange rates
in product dropping regressions. The result suggests that, parallel to a firm-level extensive margin
test, an appreciation of local currency leads to an increase in probability of adding imported products
and a decrease in probability of dropping imported products. Columns 5-8 report the results using
linear probability estimation. The results support the model predictions on product adding/dropping
as stated in Proposition 2.

21

5.2

Import responses at intensive margin

To test the impact of exchange rate fluctuations on intensive margin, we estimate the following two
specifications:
xihc(tn,t) =

3
X

k1 ec,(tkn,tk) + z1 Zit + g1 gt + Fihc + Ft + ihct

(26)

k2 ec,(tkn,tk) + z2 Zit + g2 gt + Fihc + Ft + ihct

(27)

k=0

pihc(tn,t) =

3
X
k=0

where i, h, c, t represent firm, HS6 product, country and time (month or quarter), respectively. xihct
stands for the logarithm of import value of product h by firm i from country c at time t, and e is the
logarithm of real exchange rate between RMB and importing country cs currency at time t. Zi is a
specific dummy for whether firm i engages in two-way trade (export and import at the same time).
gt is GDP growth rate at time t to control for demand shift in domestic market. F are set of fixed
effect terms. We add fixed effects at firm-country-product level Fihc , and year Ft level. ihct denotes
unobserved shocks.
In the second specification (see equation (27)), we focus on exchange rate pass-through effect on
domestic import price changes. The dependent variable becomes pihct , the price change of product
h imported from country c by firm i during time (t n, t). The independent variables are the same
as the regression for import value.12
The baseline regression results for intensive margin are reported in Table 6. Columns 1-4 report
import value (see columns 1 and 3) and import price (see columns 2 and 4) regression for imports
from all OECD countries. Columns 5 and 6 report results for import value from the US in both long
run and short run.
[INSERT TABLE 6]
Table 6 shows that the coefficients for net effect within a long run is significantly positive for
imports from OECD countries (see column 1). It suggests from a longer time perspective, the average
import value from OECD for a current firm-product-country triplet increases by 5.3 percent under
a 10 percent of RMB appreciation. The effect is robust after controlling for fixed firm-product-country
and time fixed effects. This is consistent with Proposition 3 that the intensive margin of import, i.e.
expenditure of each imported variety increases when domestic currency appreciates. For import price
changes, if local currency appreciates, the representative price of imports in terms of local currency
12

Since we are using a monthly panel data, we perform a DickeyFuller test for stationary of RMB fluctuation with
trading partners currency. The p value suggests that we reject the null hypothesis of a unit root for both monthly or
quarterly change of exchange rate at all common significance levels in our sample. Thus, we do not use VEC estimation
as in a typical time-series analysis.

22

Figure 3: Fluctuations in margins of import over time

should be reduced. In our results, in the long run import price for imports from OECD countries
reduces by 0.344 under a 10 percent of RMB appreciation (see column 2).
In a short run, when we look at the net coefficients, the results at intensive margin turn to be less
robust. To be specific, import value even decreased slightly by 0.6 percent in the short run under a 10
percent appreciation which contradicts our prediction. The price of imports reduces by 1.052 percent
in a short run. By comparing the long-run response to the short-term one, we find that a short-run
response is less robust than the long run. For import values from the US (see columns 5 and 6), we
observe that intensive margin rises by 7.8 percent in a short run and 4.5 percent in the long run under
a 10 percent appreciation of local currency.

5.3

Decomposition of changes in import margins with exchange rate fluctuations

So far, we have tested each type of import margin separately using Chinese data that support Proposition 1-3. Now, we present fluctuations of different margins and aggregate import value together in
Figure 3. Intuitively, all fluctuations follow the same trend and aggregate import value fluctuates
dominates. It suggests that both changes in extensive and intensive margins contribute to aggregate
import value fluctuation. In order to quantitatively test response of different import margins to exchange rate fluctuations, we use a simple regression to estimate the response of the change in each
margin with respect to exchange rate changes. We first difference import value as well as different
margins of import to obtain the changes in (1) total import value, (2) firm margin, (3) product-country
margin, and (4) intensive margin. To be specific, the change in firm-level extensive margin stands
for net entry out of exit in a given month; similarly, the change in product-level extensive margin
represents the net number of product adding out of dropping. The change in intensive margin is the

23

adjustment of import value of each product-country bundle of existing import firms in our sample.
Table 7 reports the results of regressing the first difference (monthly) of firm number, products
and intensive margins on exchange rate changes, respectively. Column 1 of Table 7 shows coefficient
on exchange rate fluctuations for the changes in aggregate import value, columns 2 and 3 report
changes in the number of firms and the number of products, and the last column is for changes in
the intensive margin fluctuations. After controlling for country fixed effects, we find that exchange
rate fluctuations play a significant role in determining firm entry and product churning, which leads
to an overall increase in import value. A one percent appreciation of currency significantly increases
the probability of entry by 0.23 percent and product adding for 0.24 percent. But exchange rate
fluctuations have an insignificant positive coefficient on intensive margin when facing appreciation.
The insignificant coefficient for intensive margin suggests an insignificant increase of import value by
continuing importers. To conclude, although all margins of import contributes to aggregate increase
of import value, the extensive margins rather than intensive margin are the major contributors to
aggregate import response to exchange rate fluctuations.
[INSERT TABLE 7]

Robustness

6.1

A longer run (yearly) response investigation

Previously, we focus on firms responses, e.g. entry/exit and product churning, mainly at monthly
and quarterly basis. According to an import frequency analysis in Table 2 in Section 4.3, the majority
of importers contentiously import during most of months within a year, and they import the same
product(or major product) consistently for many months within one year. However, one might still
concern about some seasonal importers. Because those seasonal production activities may incur a
continuing cease of importing for successive months within a year, for which case it may be treated
as firm exit or product dropping in our previous tests. To exclude this concern, we conduct a yearly
regression to capture firms responses at a longer horizon.
In this yearly investigation, we examine yearly import value changes rather than quarterly or
monthly changes in the intensive margin test. Also, at extensive margin levels, if a firm imports at
least once within a year we treat it as a non-exit importer; if a product-country bundle appears at
least once within a year, we treat it as a non-dropping product bundle. Exchange rates between origin
country and China are also computed as yearly average changes in exchange rates.
The results of yearly regressions are presented in Table 8. Extensive margin at firm level is reported
in columns 1-4; extensive margin at product-country level is in columns 5-8; and intensive margin is
displayed in columns 9-11. Following a similar approach as in previous main results, we use both

24

Probit and linear probability regressions to test extensive margin adjustment and linear regressions
for intensive margin test. We find that all responses, e.g. entry/exit, adding/dropping and import
value changes, follow the same pattern as the previous tests. Firms increase entry or product adding
probability, and also increase import value under a domestic currency appreciation, while reduce exit
or product dropping probability at the same time. Our model predictions are also supported by yearly
investigation at an even longer horizon.
[INSERT TABLE 8]

6.2

Subsample: Ordinary Trade vs. Processing Trade

One of the distinctive features of Chinese trade is that importers often engage in global value chain.
Those importers use imported intermediate inputs for production and then sell products to foreign
partners. A possible trade pattern is that importers may import raw and auxiliary materials, parts
and components, accessories or materials from abroad, and re-exporting the finished products after
processing or assembly by Chinese enterprises. Hence there are differences between ordinary and
processing trade in the sense that processing trade yields a special linkage between importing and
exporting behaviors. Hence, exchange rates may exert different effects on imports for processing trade
and for ordinary trade. In this section, we separate all transactions into two groups: processing trade
and ordinary trade. Here we do not distinguish pure assembling and processing trade, and label
both types as processing trade. In the test, we re-check responses of intensive and extensive margins
to exchange rate fluctuations for both ordinary trade and process trade transactions.
The results for product-level extensive margin are reported in Table 9. Columns 1-4 present
results for ordinary trade, while columns 5-8 display results for processing trade. The most significant
difference between ordinary trade and processing trade exists in the response of extensive margin, i.e.,
the effect of exchange rate appreciation on product-country churning. In Table 9, for the long run, the
probability of adding imported products decreases for processing trade under RMB appreciation, which
is opposite with a positive coefficient for ordinary trade presented in the left panel. Intuitively, for
processing trade importers, a decreasing marginal benefit of exporting offsets an increase in marginal
benefit of import due to a reduction in importing input costs. In this way, a predicted positive
stimulating effect of currency appreciation for ordinary trade importers is not guaranteed for those
who engage in global value chain. Since one firm may involve both processing and non-processing
transactions, extensive margin at firm level is less meaningful for processing trade. So we mainly focus
on extensive margin changes at product-country bundle level. However, we still check extensive
margin at firm level in our test which are presented in Table 10 where we use a weighted firm-level
exchange rate as in the previous tests to evaluate firm entry/exit probability with exchange rate
fluctuations.

25

[INSERT TABLES 9 and 10]

For intensive margin test (see Table 11), the coefficients of exchange rate are quite similar for both
ordinary and processing trade in the long run. However, in the short run, opposite to ordinary trade,
exchange rate coefficient of processing trade decreases under a currency appreciation. For import
price, we find for both types of trade, import price reduces when RMB appreciates in short run. In
the long run, the magnitude of price decrease for processing trade is larger than that of ordinary trade.
[INSERT TABLE 11]

Further discussions

7.1

Different stages in exchange rate regime switching

Although Chinas exchange rate regime reforms and RMB appreciation take place since July 2005,
the reforms can be divided into several phrases. Before early 2003, China adopted a firmly pegged US
dollar exchange rate policy, and there was no shift in foreseeing a reform in Chinese foreign currency
policy.
However, as early as in February 2003, in the G7 meeting, Japan proposed a reform towards Chinas
exchange rate policy to be taken. Since then there had been a lot of debate and discussions about the
necessity and feasibility of exchange rate reform in China. Chinese government hence faced more and
more pressure from western society to reform the foreign currency policy. For example, in September
2003, during the visit of Secretary of the Treasury of the US in China, a less government intervened
exchange rate policy was required toward a free floating exchange rate regime. In the G7 meeting in
2004, more countries and global institutions including IMF started to demand Chinas adopting of
floating exchange rate policy to replace the previously fixed one. Western countries believed that RMB
had been undervalued severely which led to a huge trade surplus for China. Starting from 2003, the
foreign currency market also anticipated an appreciation of RMB. The forward exchange rate between
USD and RMB well reflected markets expectation that forward rate started to appreciate since late
2003. The president of China Central Bank in Boao Forum for Asia in May 2005 also announced that
an exchange rate reform would be listed on the agenda.
It was believed that a reform would certainly to come. But there was still uncertainty about in what
form this reform is to be taken: is it a steady increase or an abrupt adjustment? This discussion lasted
until July 2005 when China government announced that the reform would be a steady appreciation
of RMB against USD in exchange rate policy instead of previous pegged USD. Since then RMB had
been gradually appreciated.
Although a reform in exchange rate regime was announced in July 2005, the change is not a oneshot shock. With the changes in expectations of subsequent appreciation, firms responded differently
26

for each phase of the reform. Thus, we explore the difference in firms import responses under each
phase. We construct there phase dummies to indicate the three phases of regime switching during
Chinas exchange rate reform between 2000 and 2006. The first phase is from 2000 to late 2003. During
that period there was neither change of exchange rate policy nor expectation of a change in exchange
rate policy. The second phase lasts from the forth quarter of 2003 to 2005 July, when the debate of an
exchange rate reform was heated and market had expected an appreciation of RMB/USD in the near
future. The last phase begins from July 2005, when the exchange rate reform toward RMB/USD was
announced officially.
We test the following specification by including the three phase dummies to capture three different
phrases in Chinas exchange rate reform. Ri is the set of regimes dummies for {R1 , R2 , R3 }, corresponding to three different phases in our sample. xihct is import value response for firm i importing
product h from country c at time t. Then we interact exchange rate fluctuations with those three
dummies to investigate their respective effects on import responses:
yihct = et R1 + et R2 + et R3 + Fihc + Ft + ihct

(28)

We test response at intensive margin as well as extensive margins to exchange rate fluctuation
under different regimes. Table 12 shows extensive margins at product-country level, and Table 13
lists results for intensive margin results, respectively.13
Table 12 displays extensive margin response to exchange rate under different regimes. As in
previous tests, we use both Probit in columns 1-4 and linear probability regressions in columns 5-8 to
test product adding/dropping probability. We find that firms tend to adjust product-country bundle
mostly under regime 2 and 3, i.e., the expected and confirmed exchange rate regimes. This pattern
is very obvious in quarterly response test, but monthly response is more noisy. Those results suggest
that under appreciation stages, either expected or confirmed, favorable exchange rate fluctuations
encourage firms to import more varieties or to import from more countries. This extensive margin
adjustment is through adding and dropping products-country bundles within firm.
Table 13 shows intensive margin response to exchange rate fluctuations under different regimes.
Columns 1-3 show quarterly response while columns 4-6 show monthly response. The columns 1, 3,
4 and 6 are for import value and columns 2 and 5 refer to import price. We present results for both
OECD countries and the US in testing intensive margin regarding import value. We find that under
regime 1, i.e., the fixed exchange rate regime, the import value responses are most obvious among all
three regimes in both long run and short run. Import value increases significantly less in regime 2 and
even decreases in regime 3 in the long run. For import price response, there is little difference among
regimes in the short run. It indicates that in the short run, regime shifts do not influence import
13

Since firms entry/exit decisions cover longer time interval (usually lasts for years), our regime regression for extensive
margin at firm level may not capture the accurate different responses under various phrases. Thus we mainly focus on
intensive margin test and extensive margin at product level.

27

price adjustment much. In the long run, import price reduces most during the first and third regimes,
which are either confirmed fixed or confirmed appreciation stages. For imports from the US, import
value also responds to exchange rate most in regime 1. The above finding suggests that firms are more
willing to adjust import value according to real exchange rate fluctuations under a fixed exchange rate
regime. Under such a fixed exchange rate regime without uncertainty, importers adjust import value
of the imported varieties to respond to exchange rate fluctuations in a more predictable way.
[INSERT TABLES 12 and 13]
Now we summarize the patterns related with the above results. First, in the short run, firms
responses to exchange rate changes show a lack of variations between different stages during the
reform. Second, in the long run, firms display significantly different responses at extensive margin
during different stages of the reform. Continuing importers tend to adjust product-country under
both expected and confirmed appreciation stages, which lasts from late 2003 to 2006. Third, also
in the long run, firms display various responses in intensive margin under different stages of reform.
Under a fixed exchange rate regime (2000-early 2003), importers behave most responsively to exchange
rate fluctuations. But in phase 2 and 3, this responsiveness diminished a lot. In terms of import price,
we find little difference under different stages in the short run; but in the long run, import price is
more likely to adjust to respond to exchange rate fluctuations under a fixed exchange rate regime.

7.2

Import Price Pass-through

Import price elasticity to exchange rate, which is also known as pass-through of exchange rate, is one
of the most important issues in the literature. In this part, we test pass-through of nominal exchange
rates between China and OECD countries to product price by using this highly disaggregate level
data.14 Our product price, computed as unit value using total value divided by total quantity, is
calculated at HS 6-digit level. We test both short run (within 3 months) and long run (within 12
months) pass-through of exchange rate fluctuations.
Our results are presented in Figure 4(a) and 4(b) for both short-run and long-run pass-through.
In the short run, we notice that the elasticity within three months is quite small as to around 0.10 to
0.15.15 This suggests that, in a short run (quarterly or monthly), this pass-through effect is highly
incomplete. In the long run (yearly), we find that the elasticity value rises to around 0.4 to 0.5. This
suggests that the pass-through of exchange rates on import price has an accumulative effect, and the
elasticity grows gradually towards a larger value though one-year elasticity is still incomplete in our
test.
14
Since before 2005, China had adopted a pegged to USD exchange rate policy, we drop observations between US and
China before July 2005 for a constant nominal exchange rate between USD/RMB.
15
We also test import price pass-through within one month, the elasticity is around 0.02-0.03.

28

(a) Pass Through of Import Price in Short Run: 3-month

(b) Pass Through of Import Price in Long Run: 1-year

If we further investigate elasticity value over a longer time interval, we find that both short-run
and long-run pass-through presents a declining pattern over time. This pattern is more obvious in the
long run than in the short run. The pass-through elasticity is reduced from 0.55 in 2000 to around
0.4 in 2006. Although pass-through in the short run is volatile, the value still goes down towards a
smaller value.
Both the incomplete and the declining patterns of pass-through are also found among developed
countries, which are documented in Campa and Goldberg (2005) and Marazzi and Sheets (2007). The
former finds incomplete pass-through to import price for major developed countries, while the latter
use a reduced form analysis and find that pass-through elasticity is declining over time for US. Here
we find those patterns also exist for Chinese imports.

Conclusion

In this paper, we use micro-level data to investigate firms import response to real exchange rate
fluctuations covering the RMB exchange rate reform. After decomposing the changes of import, the
extensive margin, classified both as firm entry and products adding, contributes to a magnificent part
of the overall changes in the aggregate import. It suggests that more firms participating in importing
activities, and once starting import, they tend to import more varieties and more value when domestic
currency appreciates. The drastically rise in the number of importers or the imported product varieties
are the main force to drive up Chinas aggregate import value during 2000-2006. The predicted pattern
is more significant for ordinary trade than for processing trade. Those patterns are explained by our
heterogeneous-firm trade model and the empirical investigations support the model predictions. Also,
there is incomplete pass-through effect on import price and a declining pattern of the elasticity of
pass through over time. Firms responses to exchange rate fluctuations present heterogeneity when

29

switching between different exchange rate regimes, i.e., from a fixed exchange rate regime, then to an
expected appreciation and finally to a confirmed appreciation regime.
A potential direction of future research would be heterogeneous response of firms to exchange rate
fluctuations, including firm productivity, financial status and ownerships. Previous study of a similar
topic only focuses on export side. For example, Berman, Martin and Mayer (2012) and Li et al.
(2012) find that high productivity firms has a lower pass-through and more price-to-market behavior
in exporting. Besides, other heterogeneous firm characteristics (e.g., financial status, ownerships) may
be also attributed to heterogeneity in firms import behavior or pass-through. It also has important
policy implication to explore how and which groups of importers respond to exchange rate fluctuations.

30

References
Ahmed, Shaghil. 2009. Are Chinese exports sensitive to changes in the exchange rate?
Amiti, Mary, Oleg Itskhoki, and Jozef Konings. 2012. Importers, Exporters, and Exchange
Rate Disconnect. National Bureau of Economic Research, Inc NBER Working Papers 18615.
Arkolakis, Costas. 2010. Market Penetration Costs and the New Consumers Margin in International Trade. Journal of Political Economy, 118(6): 1151 1199.
Berman, Nicolas, Philippe Martin, and Thierry Mayer. 2009. How do different exporters
react to exchange rate changes? Theory, empirics and aggregate implications. , (7493).
Berman, Nicolas, Philippe Martin, and Thierry Mayer. 2012. How do Different Exporters
React to Exchange Rate Changes? The Quarterly Journal of Economics, 127(1): 437492.
Bernard, Andrew B., J. Bradford Jensen, Stephen J. Redding, and Peter K. Schott. 2009.
The Margins of US Trade. American Economic Review: Papers and Proceedings, 99(2): 48793.
Campa, Jose Manuel, and Linda S. Goldberg. 2002. Exchange rate pass-through into import
prices: a macro or micro phenomenon?
Campa, Jos Manuel, and Linda S. Goldberg. 2005. Exchange Rate Pass-Through into Import
Prices. The Review of Economics and Statistics, 87(4): 679690.
Chaney, Thomas. 2008. Distorted Gravity: The Intensive and Extensive Margins of International
Trade. American Economic Review, 98(4): 170721.
Cheung, Yin-Wong, Menzie D. Chinn, and Eiji Fujii. 2010. Chinas Current Account and
Exchange Rate. In Chinas Growing Role in World Trade. NBER Chapters, 231271. National
Bureau of Economic Research, Inc.
Dayal-Gulati, Anuradha, and Valerie Cerra. 1999. Chinas Trade Flows-Changing Price Sensitivies and the Reform Process. International Monetary Fund IMF Working Papers 99/1.
Dees, Stephane. 2001. The opening policy in China: Simulations of a macroeconometric model.
Journal of Policy Modeling, 23(4): 397410.
Dong, Wei. 2012. The role of expenditure switching in the global imbalance adjustment. Journal
of International Economics, 86(2): 237251.
Eaton, Jonathan, Samuel Kortum, and Francis Kramarz. 2011. An Anatomy of International
Trade: Evidence From French Firms. Econometrica, 79(5): 14531498.
Garcia-Herrero, Alicia, and Tuuli Koivu. 2009. Chinas exchange rate policy and Asian trade.
BIS Working Papers, , (282).
31

Gopinath, Gita, and Brent Neiman. 2014. Trade Adjustment and Productivity in Large Crises.
American Economic Review, 104(3): 793831.
Hummels, David, and Peter J. Klenow. 2005. The Variety and Quality of a Nations Exports.
American Economic Review, 95(3): 704723.
Koopman, Robert, Zhi Wang, and Shang-Jin Wei. 2012. Estimating domestic content in
exports when processing trade is pervasive. Journal of Development Economics, 99(1): 178189.
Liao, Wei, Kang Shi, and Zhiwei Zhang. 2012. Vertical trade and Chinas export dynamics.
China Economic Review, 23(4): 763775.
Li, Hongbin, Hong Ma, Yuan Xu, and Yanyan Xiong. 2012. How do exchange rate movements
aect Chinese exports? A firm-level investigation.
Lu, Dan, Asier Mariscal, and Fernando Mejia. 2012. Imports Switching and the Impact of
Large Devaluation.
Manova, Kalina, and Zhiwei Zhang. 2012. Export Prices Across Firms and Destinations. The
Quarterly Journal of Economics, 127(1): 379436.
Marazzi, Mario, and Nathan Sheets. 2007. Declining exchange rate pass-through to U.S. import prices: The potential role of global factors. Journal of International Money and Finance,
26(6): 924947.
Marquez, Jaime, and John Schindler. 2007. Exchange-rate Effects on Chinas Trade. Review
of International Economics, 15(5): 837853.
Tang, Heiwai, and Yifan Zhang. 2012. Exchange Rates and the Margins of Trade: Evidence from
Chinese Exporters. CESifo Economic Studies, 58(4): 671702.
Thorbecke, Willem, and Gordan Smith. 2012. ARE CHINESE IMPORTS SENSITIVE TO
EXCHANGE RATE CHANGES? China Economic Policy Review, 01(02): 1250012.
Thorbecke, Willem, and Gordon Smith. 2010. How Would an Appreciation of the Renminbi
and Other East Asian Currencies Affect Chinas Exports? Review of International Economics,
18(1): 95108.
Xing, Yuqing. 2011. Processing Trade, Exchange Rates and Chinas bilateral Trade Balances.
National Graduate Institute for Policy Studies GRIPS Discussion Papers 10-30.
Yu, Miaojie. 2009. Revaluation of the Chinese Yuan and triad trade: A gravity assessment. Journal
of Asian Economics, 20(6): 655668.

32

Yu, Miaojie. 2010. Processing Trade, Firms Productivity, and Tariff Reductions : Evidence from
Chinese Products. East Asian Bureau of Economic Research Macroeconomics Working Papers
22799.

33

34

4.17

(3) Net Firm Entry

54
72
-26.30

% Net Firm Entry (rows (1+2)/8)

% Net Product-Country Add (rows (4+5)/8)

% Net Intensive Margin (rows 7/8)

Percent of Annual
Growth Due to

-2.02
7.68

5.53

(6) Net Adding

(7) Net Intensive Change

(8) Total Import Growth

- 88.79

(5) Dropping

94.32

-2.81

(2) Firm Exit

(4) Adding

6.98

(1) Firm Entry

2000-2001

Intensive Margin

Adding/Dropping

Product-Country

Firm Entry/Exit

Time Period

1.25

43
56

0.16
12.82

7.18

-83.46

90.64

5.48

- 3.23

8.71

2001-2002

1.63

36
63

0.23
14.11

8.89

-82.7

91.59

4.99

-3.03

8.02

2002-2003

Table 1: Time-series Decompositions of Changes in Chinas Imports Over Time

1.71

37
61

0.25
14.64

8.93

-82.19

91.12

5.46

-2.67

8.13

2003-2004

1.36

48
51

0.11
8.06

4.11

-88.58

92.69

3.84

-3.08

6.92

2004-2005

37.65

45
16

2.4
6.38

1.02

-90.24

91.26

2.82

- 3.16

5.98

2005-2006

35

(3)
(4)
Continuing Product
continuing months percentage
1
4.2112
2
4.1413
3
4.1868
4
4.4367
5
4.5312
6
4.8222
7
5.1017
8
5.5228
9
6.2647
10
7.4469
11
10.5867
12
38.7479

(5)
(6)
Continuing Major Product
times of churning percentage
0
20.82
1
20.37
2
17.08
3
13.26
4
9.65
5
7.05
6
4.97
7
3.24
8
1.99
9
1.04
10
0.41
11
0.13

Notes: In this table we use 2005 as sample year. We also experimented with other years and obtained similar results.

(1)
(2)
Continuing Firm
continuing months percentage
1
1.16
2
1.26
3
1.38
4
1.61
5
1.76
6
2.02
7
2.26
8
2.64
9
3.27
10
4.34
11
7.17
12
71.14

Table 2: Summary Statistics: Firm Entry and Product Churning within One year

36

Notes: , , and
constant term.

N
ad-R

Year fixed effects

Firm fixed effects

Expdummy

Growth

1794694

38.18***
(3.2746)
-0.245***
(0.0038)
yes

0.00737***
0.002359

1892676

19.00***
(2.7226)
-0.206***
(0.0035)
yes

(3)
exit

1794694

-55.71***
(3.2822)
-0.223***
(0.0038)
yes

-0.00618***
-0.00193

Probit

0.002292***
0.000724

(2)
entry

1892676

-39.63***
(2.7227)
-0.191***
(0.0035)
yes

-0.01342***
-0.00417

(4)
exit

1794694
0.374

14.42***
(0.8470)
-0.0188***
(0.0013)
yes
yes

0.00172 ***

(5)
entry

1892676
0.372

6.949***
(0.7018)
-0.0150***
(0.0011)
yes
yes

0.00173***

1794694
0.387

-28.87***
(0.8288)
-0.0109***
(0.0013)
yes
yes

-0.00443***

(6)
(7)
entry
exit
Linear Probability

1892676
0.384

-20.04***
(0.6871)
-0.0112***
(0.0011)
yes
yes

-0.004378

(8)
exit

indicate significance at the 1%, 5%, and 10% level. Robust standard errors are in parentheses. All regressions include a

Monthly Fluctuation
marginal effect

Quarterly Fluctuation
marginal effect

(1)
entry

Table 3: Baseline Regression: Firm Entry/Exit Probability

37
1189055

1157833

-0.240***
(0.0109)
yes

-5.996***
(0.0296)

1189055

-0.182***
(0.008)
yes

-6.183***
(0.0312)

-0.08228
-0.02736

(4)
exit

-0.0135***
(0.0029)
yes
yes
1157833

0.291***
(0.0079)

0.00258

(5)
entry

-0.0149***
(0.0024)
yes
yes
1189055

-34.66***
(4.2232)

0.017074

(6)
entry

0.0180***
(0.0030)
yes
yes
1157833

-2.283***
(0.0083)

-0.05872

(7)
exit
Linear

-0.00628***
(0.0023)
yes
yes
1189055

-10.53**
(4.1119)

-0.04007

(8)
exit

indicate significance at the 1%, 5%, and 10% level. Robust standard errors are in parentheses. All regressions include a

1157833

Notes: , , and
constant term.

-0.210***
(0.008)
yes

-0.310***
(0.0109)
yes

Expdum

Year fixed effects

Firm fixed effects
N

1.899***
(0.272)

0.244***
(0.0299)

Growth

(3)
exit

-0.0982***
-0.03263

Probit

0.074
0.023567

0.01718***
0.005416

(2)
entry

Monthly Fluctuation
marginal effect

Quarterly Fluctuation
marginal effect

(1)
entry

Table 4: Alternative Firm Entry/Exit Regression with USD/RMB EXR

38

Notes:
term.

N
ad-R

, ,

and

1050585

936434

1050585

0.248***
(0.0058)
yes

936434

0.285***
(0.006)
yes

-1.033 ***
-0.39914

(4)
drop

1050585
0.235

0.0276***
(0.022)
yes
yes

0.6362 ***

(5)
add

936434
0.246

(7)
drop

1050585
0.175

0.0514***
(0.023)
yes
yes

-0.9815***

Linear

0.0694***
(0.022)
yes
yes

0.0533

(6)
add

936434
0.181

0.0490***
(0.023)
yes
yes

0.086 ***

(8)
drop

indicate significance at the 1%, 5%, and 10% level. Robust standard errors in parentheses. All regressions include a constant

Year fixed effects

Firm-product-country fixed effects

0.361***
(0.006)
yes

0.214***
(0.0058)
yes

Expdum

(3)
drop

-2.026***
-0.78673

Probit

-0.261 ***
-0.06589

1.78 ***
0.700536

(2)
add

Monthly Fluctuation
marginal effect

Quarterly Fluctuation
marginal effect

(1)
add

Table 5: Baseline Regression: Extensive Margin at Product-Country Level

39

Notes: , , and
constant term.

-0.00632***
(0.0012)
yes
yes
14525409
0.914
3695.2

-0.0349***
(0.0061)

-0.1052 ***

(4)
OECD
price

0.148***
(0.0118)
yes
yes
340465
0.109
79.9

0.651***
(0.0629)

0.451 ***

(5)
US
value

0.00971*
(0.0052)
yes
yes
2658169
0.715
401.2

-0.292***
(0.0296)

0.781***

(6)
US
value

indicate significance at the 1%, 5%, and 10% level. Robust standard errors in parentheses. All regressions include a

Year fixed effect

Firm-product-country fixed effect
N
adj. R-sq
F

0.0410***
(0.0019)
yes
yes
14525409
0.736
1813.6

0.171***
(0.0033)
yes
yes
3115504
0.11
959.8

Expdum

-0.00327*
(0.0018)
yes
yes
3115504
0.112
186.4

-0.276***
(0.0093)

0.0787***
(0.0115)

0.811***
(0.0207)

Growth

-0.03448***

(3)
OECD
value

-0.0622***

0.53***

(2)
OECD
price

Monthly Fluctuation

Quarterly Fluctuation

(1)
OECD
value

Table 6: Baseline Regression: Intensive Margin

40
2205
0.0110

adj. R-sq

0.0090

2205

yes

0.235***
(0.0909)

(2)
4 Firm Number

0.0070

2205

yes

0.243***
(0.0739)

Margins of Imports
(3)
4Products-Country Number

Notes: , , and indicate significance at the 1%, 5%, and 10% level. Robust standard errors are
in parentheses. All regressions include a constant term.

yes

0.2360
(0.1566)

Country fixed effect

4EX

(1)
4 Total Value

0.0120

2205

yes

0.1950
(0.3531)

(4)
4Intensive Margin

Table 7: General Regression: Change of Margins w.r.t.Exchange Rate

41

Notes: , , and
constant term.

Year fixed
Firm fixed
Prod-cty fixed
N
adj. R-sq

GDP growth

Marginal effect
Expdummy

EXR fluctuation

973573

1378112
0.564

1378112
0.366

Product Extensive
(6)
(7)
(8)
OECD
OECD
OECD
drop
add
drop
Probit
Linear
0.00222*
-0.00333***
0.000796
-0.00141**
(0.0012)
(0.0012)
(0.0006)
(0.0006)
0.000815
-0.00122
0.0410***
0.0111*
0.00163
-0.00677*
(0.0059)
(0.0058)
(0.0036)
(0.0036)
0.124***
-0.136***
0.0373*** -0.0274***
(0.0156)
(0.0007)
(0.0004)
(0.0004)
yes
yes
yes
yes
yes
yes
yes
yes
1113268
1113268
1113268
1113268
0.054
0.051
(5)
OECD
add

(9)
OECD
value

0.124***
(0.0118)
0.0237***
(0.0019)
yes
yes
yes
1526498
0.418

0.00957
(0.0061)

indicate significance at the 1%, 5%, and 10% level. Robust standard errors in parentheses. All regressions include a

1094824

Firm Extensive
(1)
(2)
(3)
(4)
OECD
OECD
OECD
OECD
entry
exit
entry
exit
Probit
Linear
0.216***
-0.347***
0.0784***
-0.101***
(0.0053)
(0.0048)
(0.0012)
(0.0012)
0.081019
-0.11526
-0.414***
0.126***
-0.0198***
0.0227***
(0.0062)
(0.0061)
(0.0017)
(0.0017)
50.59***
-2.154***
84.54***
-123.54***
(1.5318)
(0.1343)
(0.4975)
(0.481)
yes
yes
yes
yes
yes
yes

Table 8: Robustness Check: Yearly Regressions

-0.0178**
(0.0087)
-0.00111
(0.0014)
yes
yes
yes
1504661
0.737

Intensive
(10)
OECD
price
Linear
-0.0255***
(0.0049)

0.275***
(0.0031)
0.0293***
(0.0004)
yes
yes
yes
482682
0.928

1.604***
(0.0243)

(11)
US
value

42
-0.0649***
(0.0077)
yes
yes
yes
568673

Expdum
-0.0625***
(0.0072)
yes
yes
yes
578005

0.00596
(0.0394)

-0.0967***

-0.0152**
(0.0077)
yes
yes
yes
568673

-3.854***
(0.0589)

-1.774 ***

(2)
(3)
Ordinary Trade
add
drop

-0.0613***
(0.0072)
yes
yes
yes
578005

-2.065***
(0.04)

-0.248***

drop

(4)

-0.0804***
(0.0117)
yes
yes
yes
241378

-0.219***
(0.0818)

-0.064***

add

(5)

Notes: , , and indicate significance at the 1%, 5%, and 10% level. Robust standard errors are
in parentheses. All regressions include a constant term.

Time fixed
Firm fixed
prod-cty fixed
N

-0.119**
(0.0566)

0.02***

Growth

Monthly Fluctuation

Quarterly Fluctuation

add

(1)

-0.0630***
(0.0093)
yes
yes
yes
296803

-0.0391
(0.0517)

0.165 ***

-0.0588***
(0.0117)
yes
yes
yes
241378

-4.871***
(0.0848)

-3.1 ***

(6)
(7)
Processing Trade
add
drop

(8)

-0.0773***
(0.0093)
yes
yes
yes
296803

-2.160***
(0.0524)

-0.953 ***

drop

Table 9: Robustness: Ordinary Trade V.S. Processing Trade (Extensive Margin at Product-Country Level )

43

yes
1604398

Time fixed
N
yes
1646494

30.15***
(3.5148)

0.003325***
0.00107

yes
1604398

-62.34***
(4.056)

-0.00584***
-0.00184

(2)
(3)
Ordinary Trade
entry
exit
Probit

yes
1646494

-47.95***
(3.5262)

-0.01463***
-0.00461

exit

(4)

yes
190296

13.27**
(5.7773)

0.004918***
0.001329

entry

(5)

Notes: , , and indicate significance at the 1%, 5%, and 10% level. Robust standard errors are
in parentheses. All regressions include a constant term.

41.16***
(4.0338)

0.00834***
0.002687

Growth

Monthly Fluctuation
Marginal effect

Quarterly Fluctuation
Marginal effect

entry

(1)

yes
246182

-6.146
(4.4327)

0.001566***
0.000435

yes
190296

-55.00***
(5.7532)

2.60E-05***
7.20E-06

(6)
(7)
Processing Trade
entry
exit
Linear

Table 10: Robustness: Ordinary Trade V.S. Processing Trade (Extensive Margin at Firm Level )

yes
246182

-28.19***
(4.4192)

0.005468***
0.001529

exit

(8)

44

Notes: , , and
a constant term.

-0.00271
(0.0028)
yes
yes
yes
1597503
0.114
112.2

0.0652***
(0.0179)

-0.01***

0.0127***
(0.0024)
yes
yes
yes
9078587
0.744
1709.2

-0.369***
(0.013)

0.064***

(2)
(3)
Ordinary Trade
price
value

-0.0154***
(0.0017)
yes
yes
yes
9078587
0.908
1892.4

-0.0818***
(0.0092)

-0.1124***

price

(4)

0.232***
(0.0048)
yes
yes
yes
1518001
0.11
639

1.075***
(0.0288)

0.5994***

value

(5)

-0.00385
(0.0024)
yes
yes
yes
1518001
0.111
52.23

0.0795***
(0.0148)

-0.072 ***

0.103***
(0.0032)
yes
yes
yes
5446822
0.745
782.9

-0.128***
(0.0132)

-0.166 ***

(6)
(7)
Processing Trade
price
value

0.00954***
(0.0018)
yes
yes
yes
5446822
0.903
2035

0.00777
(0.0075)

-0.10018***

price

(8)

indicate significance at the 1%, 5%, and 10% level. Robust standard errors are in parentheses. All regressions include

0.112***
(0.0048)
yes
yes
yes
1597503
0.12
430.5

Expdum

Time fixed
Firm fixed
prod-cty fixed
N
ad-R
F

0.553***
(0.0303)

0.612 ***

Growth

Monthly Fluctuation

Quarterly Fluctuation

value

(1)

Table 11: Robustness: Ordinary Trade V.S. Processing Trade (Intensive Margin)

45

Notes: , , and
a constant term.

-0.0720***
(0.0134)
0.558***
(0.0142)
-0.681***
(0.0198)
0.310***
(0.0053)
yes
1106580

0.265***
(0.0051)
yes
1269543

-0.352***
(0.0160)
0.963***
(0.0116)
0.401***
(0.0141)

(3)
drop

1269543

0.278***
(0.0051)
yes

-0.0496***
(0.0159)
-0.723***
(0.0113)
-0.461***
(0.0139)

(2)
add
Probit

1106580

-0.141***
(0.0133)
0.319***
(0.0142)
-0.782***
(0.0193)
0.287***
(0.0053)
yes

(4)
drop

0.0441***
(0.0019)
yes
yes
1269543
0.230

-0.122***
(0.0056)
0.351***
(0.0040)
0.176***
(0.0048)

(5)
add

-0.0177***
(0.0048)
0.215***
(0.0050)
-0.158***
(0.0067)
0.0634***
(0.0020)
yes
yes
1106580
0.211

(7)
drop

0.0573***
(0.0020)
yes
yes
1269543
0.185

0.0262***
(0.0058)
-0.359***
(0.0041)
-0.155***
(0.0050)

(6)
add
Linear

(8)
drop

-0.0243***
(0.0049)
0.0950***
(0.0051)
-0.175***
(0.0068)
0.0521***
(0.0021)
yes
yes
1106580
0.192

indicate significance at the 1%, 5%, and 10% level. Robust standard errors are in parentheses. All regressions include

Year
Firm-product-country fiexed
N
ad-R

expdummy

Regime3 Monthly

Regime2 Monthly

Regime1 Monthly

Regime3 Quarterly

Regime2 Quarterly

Regime1 Quarterly

(1)
add

Table 12: Response in Different Exchange Rate Regime with Extensive Margin at Product Level

46
0.177***
(0.0034)
2775039
0.109

Expdum
-0.000110
(0.0019)
2775039
0.113

-0.0470***
(0.0065)
-0.00914***
(0.0035)
-0.0220***
(0.0031)
0.155***
(0.0117)
340465
0.108

0.228***
(0.0382)
0.0489**
(0.0196)
0.0628***
(0.0166)
0.0354***
(0.0020)
11867240
0.741

0.00853***
(0.0015)
0.00635***
(0.0015)
0.00510***
(0.0015)
-0.00663***
(0.0013)
11867240
0.915

-0.0341***
(0.0010)
-0.0354***
(0.0010)
-0.0355***
(0.0010)

0.0119**
(0.0052)
2658169
0.715

0.197***
(0.0249)
0.196***
(0.0250)
0.195***
(0.0253)

(4)
(5)
(6)
OECD
OECD
US
value
price
value
Monthly Changes

Notes: , , and indicate significance at the 1%, 5%, and 10% level. Robust standard errors are
in parentheses. All regressions include a constant term.

N
adj. R-sq

Regime3

Regime2

0.541***
(0.0120)
0.0446***
(0.0065)
-0.0132**
(0.0057)

(2)
(3)
OECD
US
price
value
Quarterly Change

Regime1

(1)
OECD
value

Table 13: Response in Different Exchange Rate Regime (Intensive Margin)

Appendices
In order to get Proposition 2, we need the following condition: the slope of marginal benefit (MB) of
increasing imported variety N is smaller than the slope of marginal cost (MC) of increasing imported
variety N. This condition is shown in Graph 4, and it assures that when MB curve shifts upwards
during currency appreciation, the crossing point N of MB and MC become larger.
To derive this condition, we first get the marginal benefit function MB, which is given by the
following equation
M B(N ) = (1 )A1
Px+1
i
Since

PM
N

PX PM
>0
PM N

< 0, and > 1, the slope of MB curve becomes

PX PM 2
PX
2 PM
M B(N )
[( + 1)Px+2
(
= (1 )A1
) + Px+1
]
i
N
PM
PM N
N 2
Also, we have

PM
1 1
=
N PM
N

1 1 1 1
2 PM
=
( )N PM
N 2

By inserting

PM
N

and

2 PM
N 2

into

M B(N )
N

equation, we have the following

M B(N )
1 2 2 2
PX PX
1 1 1 1
= (1)A1
[
(+1)Px+2 (
) N PM +Px+1
( )N PM ]
i
N
PM PM

On the other hand, we assume that fixed cost for importing fimp (N ) is a function of variety N .
For example, a specific form of fimp (N ) function could be
fimp (N ) = f N , where > 1
Similarly, the MC(N) curve and marginal MC(N) of N become
M C(N ) =

2 (N )
fimp
fimp (N )
M C(N )
> 0, and
=
N
N
N 2

A sufficient condition for Proposition 2 is that

M B(N )
N

<

M C(N )
.
N

In order to get it, the slope

of MC curve must be higher than MB curve, which is shown in Graph 4 Inserting items into this
condition, it becomes
(1 )A1
i

2 (N )
fimp
PX +1 PM PX
PM
1 1
Px
[
( + 1)Px1 (
)
]<
PM
N PM
N
N
N 2

47

Figure 4: Condition of Proposition 2

It can be proven that, given are sufficiently large, this condition is likely to be satisfied as N become
larger.

48