Exchange rate pass-through and relative prices: An industry-level empirical investigation

Abstract

In this paper we explore the extent of exchange rate pass-through for the USA, UK and Japan using a post-Bretton Woods industry-level dataset. We investigate how different channels of exchange rate pass-through affect domestic and import prices. Our analysis is suggestive of two channels of transmission and we find considerable variation in the extent of pass-through across industries and countries.

Full text

Page 1

Exchange Rate Pass-Through and Relative Prices:
An Industry-Level Empirical Investigation¤
Prasad S. Bhattacharya, Cem A. Karayalcinyand Dimitrios D. Thomakos
March 4, 2004
Abstract
In this paper, we explore the extent of exchange rate pass-through for the USA, UK and Japan
using a post-Bretton Woods industry-level dataset. We then investigate how di®erent channels
of exchange rate pass-through a®ect domestic and import prices. Our analysis is suggestive of
two channels of transmission and we ¯nd considerable variation in the extent of pass-through
across industries and countries.
JEL Classi¯cations: F0, F31, F32, F41.
Keywords: Exchange Rates, Pass-Through E®ect, Expenditure-Switching E®ect, International
Competitiveness, Vector Autoregression.
¤An earlier version of this paper was presented at the 9th International Conference on Computing in Economics
and Finance in Seattle and at the Departmental Seminar Series at the Florida International University. We would
like to thank the conference and seminar participants for useful comments and suggestions. The programs and the
data used in the paper are available upon request. All remaining errors are our responsibility.
yCorresponding author, e-mail: karayalc@¯u.edu, fax: (305)348-1524. All authors are with the Department of
Economics at Florida International University, Miami, FL 33199.
1

Page 2

1 Introduction
The \exchange rate pass-through" is the percentage change in local currency import prices resulting
from a one percent change in the nominal exchange rate. As changes in the nominal exchange rate
may not be fully passed through to goods prices, consumer prices may not be very responsive
to such changes. This implies that there will be less \expenditure-switching", i.e., a change in
the exchange rate might not lead to much substitution between domestically-produced goods and
imports. Under the pricing to market (PTM, see Krugman (1987)) or local currency pricing (LCP,
see Devereux (1997)) mechanisms, ¯rms either price discriminate or set a di®erent price in foreign
currency for sales to foreign households, leading to low pass-through due to LCP and, therefore,
less expenditure-switching. This makes a ¯xed exchange rate regime preferable, because a sudden
shortage in the supply of foreign goods due to some exogenous shock will lead to a large and
undesirable currency depreciation under a °exible-rate regime. Previous empirical work report
that prices are indeed sticky in consumers' currencies.1New open economy macro models use the
low pass-through evidence from earlier studies to consumer or domestic prices in supporting the
assumption of nominal price rigidity in buyers' or consumers' currency.
However, if importables are priced in exporters' or producers' or sellers' currencies as producer
currency pricing (PCP) models assume, currency depreciation in the destination country will lead
to higher import prices in that country. With only the imported goods prices changing because of
pass-through and no change in the domestic goods prices, there is a change in relative prices. As a
result, there may be a high \expenditure-switching" e®ect of the nominal exchange rate change as
opposed to the LCP-PTM type models. In a number of in°uential new open macro papers, Obstfeld
and Rogo® (1995, 1998) show that changes in consumer prices in the short-run can be explained
with changes in nominal exchange rate assuming PCP type models. Obstfeld (2002) and Obstfeld
1The theoretical and empirical studies involving domestic and consumer prices point to a low pass-through e®ect
towards consumer prices. New theoretical models that support a ¯xed exchange rate regime based on the assumption
of domestic price rigidity are surveyed by Engel (2002). We selectively mention Mussa (1986), Rogers and Jenkins
(1995), Engel (1993, 1999, 2000), Engel and Rogers (1996, 2001) and Parsley and Wei (2001) as partial references for
this argument. A large number of works also investigate the extent of pass-through for industry-level prices. See Yang
(1997) and references therein for a review of these studies. And for pass-through studies showing evidence of PTM
in import prices, Goldberg and Knetter (1997) and Campa and Goldberg (2002) are two comprehensive references.
2

Page 3

and Rogo® (2000) also use PCP, and provide evidence for the argument that a nominal exchange
rate depreciation drives down relative export prices and increases relative import prices. These
studies suggest that exporters largely invoice in home currency so that changes in the nominal
exchange rate have signi¯cant short-run e®ects on international competitiveness. In their dynamic
time-series study of German, Japanese and US automobile exports to seven industrial-country
destinations, Gagnon and Knetter (1995) ¯nd that PTM is greater in the long-run than in short-
run, which is consistent with invoicing in the exporters currency. Therefore, the pass-through to
export prices is not zero, implying that LCP towards export prices may not be true. Proponents
of PCP use the above evidence to assume price stickiness in sellers' or exporters' currency.
This debate regarding the sticky prices assumption either in terms of buyers' or sellers' currency
is important from the policy perspective as well, as domestic monetary policies may or may not
have signi¯cant e®ects depending on the extent of pass-through.2For example, a small nominal
exchange rate transmission to the import prices may mean a lower expenditure-switching in response
to changes in domestic monetary policy, with the consequence that monetary policy is more e®ective
in dealing with real shocks. As we know, the LCP-PTM argument does not rely on the expenditure-
switching e®ect of changes in the nominal exchange rate to generate price di®erences between
imported good and domestically produced goods. Therefore evidence supporting LCP-PTM would
indicate that monetary policy should be used to deal with real shocks. Evidence in support of
the PCP mechanism would have the consequence that monetary policy is ine®ective in dealing
with real shocks. Apart from the expenditure-switching role, exchange rate pass-through may also
a®ect in°ation prediction. Studies, like Ball (1999) incorporates the degree of pass-through in the
monetary policy rule to control in°ation. McCarthy (2000) addresses this issue through a VAR
model incorporating a distribution chain of pricing and ¯nds that exchange rates have a modest
e®ect on domestic in°ation while import prices have a stronger e®ect. Engel (2002), Obstfeld (2002)
and Lane (2001) call for detailed empirical investigation on the extent of pass-through e®ect for us
to be able to come to comprehensive conclusions concerning policy.
Recently, to address issues regarding price stickiness in terms of buyers' or sellers' currency,
2Lane (2001) provides an excellent survey of the growing literature on new open economy macro models involving
monetary policy.
3

Page 4

Obstfeld (2001) and Obstfeld and Rogo® (2000) built hybrid theoretical models incorporating PCP
to import prices and LCP-PTM to domestic prices from a ¯rm's perspective. These models do
generate expenditure-switching similar to the traditional Mundell-Fleming-Dornbusch models.
In this paper, we try to address the issue of exchange rate pass-through at industry-level prices.
Our analysis di®ers from the existing literature3in that we not only explore the industry-level
exchange rate pass-through, but also distinguish between various pass-through channels which may
generate changes in the relative prices of domestic and imported goods. With this relative price
change, there may be \expenditure-switching" at the industry-level decision making, as hypothe-
sized by Obstfeld (2001) and Obstfeld and Rogo® (2000).
Our analysis is carried out in two parts. In the ¯rst part of the study, following the suggestion
by Obstfeld, we empirically explore di®erent pass-through channels to prices of imports and do-
mestic goods using a monthly industry-level dataset for USA, UK and Japan. Based on markup
adjustments by domestic importers and foreign exporters, we show that there is a di®erence in
pass-through to di®erent prices in the short-run. According to Goldberg and Knetter (1997), the
markup adjustment following an exchange rate change generally occurs within a year. Thus, in the
short-run, there may be a change in relative prices that can generate expenditure-switching. We
¯nd evidence of relative price changes in industry level import and consumer prices in the short-run.
Out of 34 industries from three countries in the sample, relative prices change for forty-two percent
of the cases.4
The empirical approach in this paper is based on a \triangular" system of equations in which
a number of interesting e®ects can be clearly and adequately measured: direct short and long-run
responses of domestic (consumer and producer) and import prices to exchange rate changes; the
\carry-over" e®ect from import prices to domestic prices, as well as the \carry-over" e®ect from
3Menon (1995) provides a good review of the exchange rate pass-through literature.
4In a recent study, Shambaugh (2003) relies on Blanchard and Quah (1989)'s methodology to generate long-run
shocks in the exchange rate and prices and he ¯nds that import prices are set in the producers currency with lower
CPI pass-through re°ecting margin changes in the supply chain. This provides some support to the idea of pass-
through separation as proposed by Obstfeld (2002). However, this study does not focus on a detailed industry-level
analysis like ours. Also, it does not consider changes in export prices in response to changes in exchange rates missing
a possibly important channel.
4

Page 5

import prices to producer prices (after an exchange rate change has taken place). The triangularity
of the system is based on the observation that one expects no immediate short-run feedback from
domestic (consumer and producer) prices on import prices and no feedback from consumer prices
on producer prices. Using a comprehensive econometric methodology that starts by investigating
probabilistic properties of the data, we build a VAR model and suitably adjust that to get the
proposed system. We test the proposed \no-feedback" e®ect and the results support triangular
arrangement for 19 out of 34 industries in the sample. There is also considerable similarity across
industries regarding the no past-feedback. For example, in US and UK, results from industrial
prices of beverages and tobacco, chemicals and machinery and transport equipments show that
there is no lagged feedback of CPI or domestically produced goods prices to current prices of
imported goods. Similarly, results from subcategories of manufactures industry from both of these
countries support this claim.
In the second part of the study, we analyze industry-level export price movements after an unex-
pected exchange rate shock and check whether there is evidence to indicate export price stickiness
in terms of either the sellers' or buyers' currency. This is done in two ways. First, we investigate
the correlation between industry-level relative export price competitiveness5and in the nominal ex-
change rate (as in Obstfeld, 2002). The results show high correlations in most of the industry-level
export prices, providing some support to invoicing in terms of producers' or sellers' or exporters'
currency. Second, we use the changes in the logged ratio of export prices and trade-weighted
exchange rate to further corroborate the ¯nding from the ¯rst part of the export price analysis.
The results from both parts of the study point to potential di®erences in pass-through that
can generate expenditure-switching. Apart from that, the study re-a±rms the fact from earlier
empirical studies that pass through e®ect is partial. In our case, the exchange rate pass-through
e®ect for industries in USA ranges from thirty percent to ¯fty percent, which is consistent with
Goldberg and Knetter (1997) and Campa and Goldberg (2002) ¯ndings. Within manufacturing
and food industries, Japanese food, metal and textiles industries, UK's iron and steel industry and
US rubber and furniture industries support earlier results by Campa and Goldberg (2002) related
to negligible pass-through e®ect for these industrial categories. However, US food and machinery
5See the following section for precise de¯nition of relative export price competitiveness.
5

Page 6

and transport equipment industry as well as UK tobacco and non-ferrous metal industry prices
show that there may be higher extent of pass-through. Results from Japanese wood industry and
UK pulp and metal ores industries, which can be treated as part of Non-manufacturing and Raw
materials industries show greater evidence of PCP, again supporting Campa and Goldberg (2002)
¯ndings related to rejection of LCP for these two industry categories.6
The rest of the paper is organized as follows. In section 2, we discuss a simple theoretical model
and our empirical methodology. In section 3 we present the dataset, while the results from the
empirical analysis and comments on possible long-run and short-run adjustments are discussed in
section 4. Finally, section 5 concludes. All empirical results are reported in tabular form in the
¯rst appendix. The second appendix provides certain details about the data used in this study.
2Empirical analysis from industry-level import and export prices
2.1 A simple model for import prices
To analyze di®erent channels of exchange rate pass-through (as proposed by Obstfeld (2001, 2002))
in a simple framework, we focus on the following four prices: (1) Prices of imported goods at the
point of entry denominated in local currency(denoted by imp) (2) Domestic prices of imported
goods as paid by the end-users(denoted by dig) (3) Domestic prices of import-competing goods
(denoted by ddg) (4) Domestic prices, as measured by the CPI (the ¯nal price consumers' pay at
the point of sale). The dig and ddg are the constituent parts of CPI. To investigate changes in
CPI, we look for changes in either ddg or dig or in both.
Under the LCP mechanism, neither the dig nor the CPI may change after exchange rate changes.
With PTM, there is proportionate and opposite adjustments in the markups charged by sellers
after exchange rate changes resulting in zero or low pass-through. However, with PCP both prices
respond proportionately to changes in the exchange rate. Di®erent channels of exchange rate pass-
6For instance, studies like Campa and Goldberg (2002) show that for OECD countries, Manufacturing and Food
import prices reject evidence of both PCP and LCP in the short-run. There is, however, long-run pass-through
evidence for Non-manufacturing and Raw Materials import prices in their study. Also, Goldberg and Knetter (1997)
report the median pass-through for OECD Manufacturing import prices hovers around ¯fty percent over a year period
(described as long-run).
6

Page 7

through can now be delineated in the following way.
(1) With an unexpected change in the exchange rate (say, due to unanticipated monetary shock),
there is pass-through to imp. There may be a potential markup adjustment by foreign exporters at
this point (in accordance with the PTM argument proposed by Krugman (1987)). The end result
will be either higher or lower pass-through to imp. This is the ¯rst and direct pass-through channel.
If the pass-through to imp is fast and of greater magnitude as compared to pass-through to
dig and CPI, then there would be changes in relative prices. This is important, for there to be
expenditure-switching there has to be a higher level of pass-through to imp but a low pass-through
to dig and the CPI.
With the transaction cost of the imported good added to the domestic price of the imported
good, there may be a di®erence between imp and dig. In addition, domestic importers may charge
markup on the imported good (the price of which is denoted by dig). With di®erent domestic
markup adjustment in dig vis-a-vis markup adjustment in imp (as described in (1) above), there
may be a relative price di®erence between dig and imp. Whether CPI changes or not depends on
the domestic markup adjustment re°ected in dig.
If there is domestic wage rigidity (or price rigidity, as proposed in a number of models, see the
related discussion in Bergin (2003) and the references within)7, ddg will not change very fast in
response to a domestic monetary shock that a®ects the exchange rate. The exchange rate change
will only a®ect the imp.8
(2) If there is no markup adjustment by foreign exporters at the entry point, the imp will change
proportionately with a change in the exchange rate. If the ddg responds sluggishly, there would
be a di®erence between imp and ddg. To maintain the domestic sale of imported goods, domestic
importers may adjust the domestic markup in proportion to the imp hike. This will be re°ected
in the dig. This is the mechanism emphasized by LCP or PTM. Thus the ddg and dig remain at
the level before the unanticipated shock. As a result, CPI will not change. This is the second and
indirect channel of exchange rate pass-through to dig and CPI.
7Domestic wage rigidity leads to domestic price sluggishness as there may be a wage contract signed before the
unexpected monetary shock.
8This can happen if invoicing contract is signed a period ahead.
7

Page 8

After all the adjustments have taken place as above, there may still remain a di®erence between
imp and dig as well as CPI. This may trigger expenditure-switching as is proposed and shown in
Obstfeld (2001, 2002) and Obstfeld and Rogo® (2000).
As this discussion suggests, the speed and extent of markup adjustment drives all the results
of high and low pass-through to imp and dig as well as CPI.
The second and indirect channel of pass-through to CPI, as described earlier, calls for looking
at the extent of domestic markup adjustments. Since we do not have data for the domestic cost of
production at the industry-level at monthly frequency, we devise an indirect way of looking at this
e®ect. Changes in CPI show changes in either ddg or dig or both. The dig re°ects adjustment at the
second stage when the imported good enters the domestic distribution chain. So the exchange rate
change at the entry stage (re°ected in imp) can indirectly capture the change in dig because the
second stage adjustment happens only if there is a change in markups at the ¯rst stage. Therefore,
we take the exchange rate coe±cient as a proxy to the adjustment in dig based on markups. In
addition, if there is any change in the PPI or CPI due to a change in the dig (assuming that there
is tradable component in production), it is possible to separate out that e®ect. We denote this
potential e®ect as the \carry-over e®ect". For some industries, the data allows us to look for this
e®ect.
The above discussion calls for suitably disaggregated data. We do not have access to industry-
level domestic prices of the imported goods, yet the data we have include (1) consumer prices
(CPI)(which re°ect prices of ¯nal products or ¯nal consumer goods), (2) import prices (IMP) for
¯nal goods, intermediate goods as well as crude materials and (3) prices for domestic producers
(PPI), which can be categorized in terms of ¯nal goods, intermediate goods and crude materials.
With this information, we are able to distinguish in terms of end uses, and this allows us a better
understanding of the channels of pass-through under investigation.
In our setup, we take the exchange rate to be exogenously determined and focus on the e®ect of
unexpected changes in the exchange rate.9Within this framework, we build three di®erent reduced-
9Since we are only interested in the transmission of exchange rate shock and not where and how the shocks are
generated. See Adolfson (2001) for a discussion regarding exogenous exchange rate choice and omitting \controls"
for disaggregated industry-level work.
8

Page 9

form systems to capture the e®ect of changes in the exchange rate on import prices, producer prices
as well as consumer prices.
System I : This setup analyzes various channels which transmit a change in the exchange rate
to the prices of ¯nal goods as measured by the CPI. The CPI re°ects tradable (denoted by dig) and
non-tradable goods (denoted by ddg) prices. With a higher proportion of tradables, any change
in the exchange rate that a®ects their prices would lead to a signi¯cant change in CPI. Assuming
PCP, the e®ect will be full. On the other hand, under LCP-PTM mechanism, the extent of the
CPI change will be closer to zero.
In this scenario, home retailers import ¯nal goods (or tradables) from foreign countries. At the
point of entry, retailers pay the price, Pri for foreign exports, which possibly includes a markup
m¤
i¸ 0 charged by the exporters. So retailers pay the import price, Priwhere
Pri= (1 + m¤
i)EP¤
i
(1)
and where E denotes the exchange rate and P¤
iis the foreign currency price of the i-th good
imported in the home country. Depending on whether m¤
ichanges or not, a change in the exchange
rate may or may not be re°ected in Pri. As pointed out earlier, this is the ¯rst channel of short-
run exchange rate pass-through. Here foreign exporters can price discriminate between destination
markets of tradables.
In the home market, the retailers have to bear the transport and distribution costs of ¯nal
goods before selling those to the consumers. Denoting these costs for the i-th good as −ri, the total
marginal cost that the retailers face is:
MCri= Pri(1 + −ri) (2)
Letting the domestic markup charged by the retailer to be °i, we have the ¯nal price of the i-th
good as:
Pci= (1 + °i)MCri
(3)
Given (2) and (1), the ¯nal price of the i-th good is:
Pci= (1 + °i)(1 + −ri)(1 + m¤
i)EP¤
i
(4)
9

Page 10

where, Pcidenotes the ¯nal price of the i-th good. Equation (4) shows the link between E, import
prices and Pcias mediated by the various markups clearly. An import price increase driven by an
exchange rate depreciation (assuming no change in the foreign mark-up) may or may not lead to a
proportional increase in consumer prices. This is the second channel of pass-through. The response
of domestic prices as a result of this import price hike can explained by changes in the existing
domestic markup, °iin the way we have described before.
The foreign and domestic markup adjustments can transmit the changes in exchange rate to
import prices and domestic prices. These adjustments can be a re°ection of sluggish nominal wage
adjustments in the foreign country as well as in the home country.
System II : Here, we concentrate on the case where ¯nal goods are produced domestically using
tradable and non-tradable intermediate inputs. We allow for the possibility that these intermediate
inputs themselves have some imported components. Therefore, any change in producer prices of
¯nal goods can be explained indirectly by changes in the exchange rate through changes in the
prices of domestically produced intermediate inputs as well as changes in the prices of imported
intermediate inputs.
To capture the underlying structure of System II in terms of the price adjustments, we look at
the following expression. In the case where domestic producers use imported intermediate goods,
the price (Pni) they pay at the point of entry once the markup m¤
ni¸ 0 charged by the exporters'
is taken into account, will be given by:
Pni= (1 + m¤
ni)EP¤
i
(5)
In addition to the transaction cost, −ni, associated with these imported inputs, the producer also
has to pay the wage wiper unit of output being produced domestically. Letting wibe the wage
rate and let lidenote the unit labor requirement in the production of i-th good. The marginal cost
is:
MCni= Pni(1 + −ni) + wili
(6)
Taking the domestic markup charged by the producer as ¹i, we have the producer price of the i-th
good as:
Ppi= (1 + ¹i)MCni
(7)
10

Page 11

Using (6), we get the producer price of the i-th good as:
Ppi= (1 + ¹i)[wili+ (1 + −ni)(1 + m¤
ni)EP¤
i] (8)
We have a similar interpretation in terms of markup adjustments and producer prices change
as before. The expenditure-switching is due to the change in the relative price of the imported
inputs.
System III : This is similar to System I described earlier, with the di®erence that the ¯nal
goods prices are simply replaced by producer prices. In this case, instead of ¯nished retail goods
for consumption, we have ¯nished goods for production. As before, denoting the transaction cost
for the i-th good by −pi, foreign markup by m¤
piand the domestic markup charged by the producer
by ºi, the producer price of the i-th good will be:
Ppi= (1 + ºi)(1 + −pi)(1 + m¤
pi)EP¤
i
(9)
The channels through which change in exchange rate is passed through in this system is anal-
ogous to the ones described above. What di®erentiates the three systems is that the pass-through
is towards consumer prices in system I and towards producer prices in systems II and III.
2.2 Empirical framework: import prices
The conceptual framework expressed above can be converted into an econometric framework, which
then can be used both for testing the implied triangularity of the system we propose and for
estimating short and long-run e®ects. Our analysis is conducted in growth rates. Standard unit
root and cointegration tests were performed for the levels, but we found no strong evidence of
cointegration for almost all our systems.10In what follows we describe the empirical implementation
of system I.
Consider a (3 £ 1) vector with CPI in°ation, PPI in°ation and growth rate of import prices,
say zt
= (yt;xt1;xt2)0, and rede¯ne the growth rate of the exchange rate as xt3´ wt.11We assume
that ztcan be adequately modeled by a vector autoregression with an exogenous input variable
def
10These tests are not presented here but are available on request.
11All variables are taken as deviations from their respective sample means.
11

Page 12

(VARX) as:
zt
def
=
p
X
i=1
¦izt¡i+
q
X
j=1
¯jwt¡j+ ut
(10)
where f¦igp
error vector utis assumed to be multivariate white noise with variance-covariance matrix §. The
i=1are (3 £ 3) parameter matrices and f¯gq
j=1are (3 £ 1) parameter vectors. The
model in the above equation will be our broadest, unrestricted model (U-model).12The implied
triangularity of the conceptual model of the previous section can now be tested using this model.
Consider the restrictions implied by the following null hypothesis and corresponding to our ¯rst
restricted model (R1-model):
H0:
©¼i
ab= 0 j for a > b and a;b = 1;2;3ª
8i
(11)
where ¼i
abis the (a;b) coe±cient of ¦i. These restrictions imply absence of feedback from CPI
in°ation to PPI in°ation and from CPI and PPI in°ation to growth of import prices; they are
immediately testable using a Wald-type test applied to the U-model.
If the above null hypothesis is rejected, we proceed by eliminating the insigni¯cant coe±cients
from the U-model and by re-estimating the remaining parameters by seemingly unrelated regression
(SUR). This is our second restricted model (R2-model), which we then compare to the U-model
using a likelihood ratio (LR) test. If the R2-model is rejected in favor of the U-model we use the
estimates from the U-model to compute the long-run e®ects; if the R2-model is not rejected we use
its estimates to calculate long-run e®ects. Similarly, if the null hypothesis of triangularity is not
rejected, we proceed by eliminating the insigni¯cant coe±cients from the R1-model and re-estimate
the remaining parameters using SUR. This constitutes our third restricted model (R3-model), which
we then compare to the R1-model using a LR test. Depending on whether the R3-model is rejected
or not we use the estimates from either the R1-model or the R3-model to calculate the long-run
e®ects.
To illustrate the computation of the long-run e®ects, consider the U-model and re-write it using
lag operator notation as:
¦(L)zt= ¯(L)wt+ ut
(12)
12The U-model was estimated using conditional least squares with the orders chosen by the Schwarz (BIC) criterion.
12

Page 13

where ¦(L)
def
= I3¡
p
X
i=1
¦iLiand ¯(L)
def
=
q
X
j=1
¯jLj. When the system is in long-run equilibrium
we expect that the variables do not deviate substantially from some ¯xed values, say z¤, w¤and
u¤def
= E[ut] = 0. Therefore, we have the representation:
¦(1)z¤= ¯(1)w¤
(13)
from which all long-run e®ects can be easily computed by summing the estimates of the ¦i's and the
¯i's. For example, the long-run e®ects of the exchange rate growth on CPI in°ation, PPI in°ation
and growth of import prices are given by the estimate of the vector @z¤=@w¤
def
= [¦(1)]¡1¯(1).
Standard errors for the long-run e®ects were obtained using the Delta method. In the ¯rst appendix
we report the total short and long-run e®ects, as well as the carry-over e®ect.
2.3 Empirical framework: export prices
Let ptdenote the price of export of the home country, p¤
tdenote the export price of the foreign
country and etdenote the nominal exchange rate (in units of home's currency per unit of foreign
currency). For the ¯rst part of the analysis, we construct home's relative export prices (denoted
by REP) vis-a-vis foreign's export prices in the following way:
REP = et¢p¤
t
pt
(14)
We then look at the correlation of monthly changes in logarithm of relative export prices and the
nominal exchange rate. With export prices remaining ¯xed in the originating countries currencies,
we expect that the relative prices in the above equation will not vary. Thus, relative export prices
and exchange rates °uctuations will be highly correlated. This would support the notion of invoicing
in terms of sellers currencies, which provides evidence against the LCP mechanism.
For the second part of the analysis, the dependent variable is the monthly growth in the ratio of
export prices ¢yt= (1¡L)yt. The explanatory variable is the monthly growth of the trade-weighted
exchange rate xt3. The model we consider is a regression in monthly growth rates, namely:
¼(L)¢yt= c + ±(L)xt3+ zt(#) (15)
13

Page 14

where ¼(L) = 1 ¡Pr
and where the term zt(µ) captures the regression error dynamics of the equation and depends
i=1¼iLiand ±(L) = ±0+Ps
i=1±iLiare polynomials in the lag operator
on the auxiliary parameter vector #. For example, if the equation includes the ¯rst lag of ¢yt
and xt3 and the regression error follows a seasonal autoregression of orders one and twelve then
¼(L) = 1¡¼1L, ±(L) = ±0+±1L and zt(#) = ut, with (1¡Á1L)(1¡Á12L12)ut= "t, "t» iid(0;¾2
and #
= (Á1;Á12;¾2
")
def
")0. If no lagged dynamics of ¢ytand xt3are explicitly included then ±¤= ±0
gives us the long-run \equilibrium" e®ect of a change in monthly relative export prices from a
change in the monthly trade-weighted exchange rate. If lagged terms are present, then the long-
run e®ect is computed as ±¤= ±(1)=¼(1), i.e. as the equilibrium solution of the dynamic part of
the model. As above, standard errors for the long-run e®ects are obtained using the Delta method.
3 Data
3.1Import prices
Dornbusch (1987) uses disaggregated industry-level prices to investigate the strategic interactions
among domestic producers and producers of competing goods' imports to explain low pass-through
to import prices, in accordance with the PTM argument of Krugman (1987). In our analysis, the
data choice follows that of Dornbusch.
The data for USA is obtained from the Bureau of Labor Statistics's web site. Import prices
are taken from SITC classi¯cation with the corresponding PPIs are taken both from industry-level
and commodity-level classi¯cations. We use the CPI for urban consumers as our sample CPI series
in the study.13For import prices from Food and Beverages, Mineral Fuels and Lubricants and
Textiles industries, we use the corresponding CPI for these industries. In every case of our data for
USA, the end point is December, 2002. However, the starting point varies across industries.14For
example, the sample of import prices for Food and Beverages begins from January, 1989, while that
for Textiles, Organic chemicals, Furnitures and Metalworking machinery starts from January, 1994
and for the rest of the industries, the starting date is January, 1993. We look for industry-level
13We use CPI as a measure of retail prices for these industries in order to match the corresponding import prices
data for these industries.
14This is due to the monthly industry-level data availability.
14

Page 15

import prices and use trade-weighted exchange rates for all the countries in the sample. The data
for trade-weighted exchange rate for USA comes from the Federal Reserve Bank of St. Louise web
site. For import prices, the base year is 2000 and for consumer and producer prices the base year
is between 1982-84. We change all the series base to an uniform base of average of 1995 prices.
The analysis is carried out for fourteen industries in US, including broad as well as subcategories
of SITC level industries.15
In the case of UK, the data for SITC level import prices and PPIs come from the National
Statistics Online. The trade-weighted exchange rate data is available from Bank of England's web
site. The price sample ranges from January, 1991 to December, 2002. We look at eleven industries
in our analysis.
For Japan, the wholesale price index (WPI)16, domestic producer price index (PPI), import
prices (IMP) and trade-weighted exchange rates are obtained from the Bank of Japan's web site.
Our sample ranges from January, 1971 to December, 2002 and we have seven broad SITC industries
for our study.
3.2Export prices
In case of USA, SITC level data for export prices are obtained from the web site of the Bureau of
Labor Statistics. In all, we investigate pricing behavior in sixteen industries including subcategories
of broad SITC level industries. Most of the monthly data start from January, 1993 and we have
taken observations till December, 2002 in our analysis. The reported base for all the prices' in the
sample is 2000 = 100. We have changed the base to average of 1995 for the present analysis.
For UK, sixteen SITC level industries' export prices including prices for some subcategories of
industries are taken for the present analysis. The reported base in this case is 1995 = 100. The
data sample period is from January, 1983 to December, 2002.
Japanese export prices are taken from Bank of Japan's web site. The data correspond to six
industries with the time period starting from January, 1983 and ending at December, 2002. We
15Appendix 2 provides the detailed import data.
16We use this as a measure of retail prices as described in our System I earlier. Now this index is replaced by the
corporate goods index in the source web site.
15