Asymmetric and Non-linear Exchange

Rate Pass-Through:
An Empirical Analysis for Six Different Countries

Master Thesis

ERASMUS UNIVERSITY ROTTERDAM

Erasmus School of Economics
Department of Economics

Author: Supervisor:

Erna Björg Sverrisdóttir Dr. L. Pozzi

406772
[email protected]

Abstract

Due to downward price rigidities, market share objectives, binding quantity constraints
and menu costs the responses of import prices to exchange rate changes can be
asymmetric and non-linear. In this paper, regression equation augmented with interactive
dummy variables is used on quarterly data from 1980 to 2014 for six different countries,
Germany, Iceland, New Zealand, Sweden, the U.K. and the U.S., in an attempt to test
whether the direction and the size of the exchange rate change matters for pass-through
to import prices. The results indicate that asymmetry cannot be neglected. For the U.S.
and Sweden, evidence of asymmetric behavior is found but the direction of asymmetry
varies between the two countries. For no country can the restriction of linear pass-
through be rejected, unless when also taking into account the direction of the change,
suggesting that the direction effects overshadow the size effects.
Contents
1 Introduction ................................................................................................................ 2
2 Literature review ........................................................................................................ 4
2.1 Exchange rate pass-trough ................................................................................... 4
2.2 Asymmetric and non-linear exchange rate pass-through ..................................... 6
2.3 Previous empirical literature ................................................................................ 9
3 Methodology ............................................................................................................ 12
3.1 Analytical framework ........................................................................................ 12
3.2 Empirical specification ...................................................................................... 13
3.2.1 Linear specification .................................................................................... 13
3.2.2 Allowing for asymmetry and non-linearity ................................................ 15
3.4 Data ......................................................................................................................... 17
4 Results ...................................................................................................................... 19
4.1 Results from the linear specification....................................................................... 19
4.2 Robustness tests ...................................................................................................... 20
4.3 Results when allowing for asymmetry and non-linearity ....................................... 21
5 Conclusion ................................................................................................................. 24
References ........................................................................................................................ 26
Appendix .......................................................................................................................... 30

1
1 Introduction
How import prices respond to changes in the exchange rate, commonly referred to as the
exchange rate pass-through (ERPT), has long been of interest to economists. Following
the great inflation of the 1970s and the collapse of the Bretton Woods system, interest in
ERPT spiked. Due to the effects that import prices have on domestic inflation,
understanding how and to what degree exchange rate changes are passed through to
import prices is of key importance for the implementation of central bank´s monetary
policy. Furthermore, the degree of ERPT determines the impact of exchange rate changes
on the balance of payments. Other things equal, the higher degree of ERPT, the greater is
the impact of exchange rate changes.
Since the 1970s, empirical literature focusing on ERPT has grown in abundance. The
lion´s share of the literature assumes that the relationship between exchange rate and
import prices is symmetric and linear, meaning that neither the direction nor the size of
the exchange rate change matters: pass-through is the same either way. However, as
several episodes in recent years have showed, these assumptions are not realistic. For
instance, from 1999 to 2001 the euro depreciated by almost 20%, causing an increase in
euro area import prices by roughly the same magnitude. However, by 2004 the euro had
regained much of it former value but at the same time euro area import prices had only
decreased by 5% (Bussière, 2007). This example suggests that import prices do not react
in the same way to appreciations and depreciations.
There exist numerous factors that can justify asymmetric and non-linear ERPT.
Downward price rigidities, binding quantity constraints and pricing-to-market strategies
are all examples of such factors. In theory, appreciation of the importing country’s
currency (local currency) can lead to either higher or lower ERPT than depreciation. If
prices are rigid downwards, as Peltzman (2000) notes, then exporters are less keen to
decrease their export prices than increase them, implying a higher ERPT during
depreciations of the local currency than appreciations. The same results arise when
exporters are faced with binding quantity constraints. In contrast, if exporting firms
operate under a market share objective, then an appreciation of the importing country’s
currency can cause larger ERPT than depreciations. Furthermore, menu costs can
generate non-linear ERPT, as exporters only change their prices if the exchange rate
change exceeds a certain threshold. Only a handful of studies consider, and test for,
asymmetry and non-linearity. They have found mixed results, with some reporting

2
evidence of asymmetric and non-linear responses while others do not. Furthermore, there
seems to be no consistency in the direction of asymmetry.
This paper uses quarterly, aggregate data for six different countries, Germany,
Iceland, New Zealand, Sweden, the United Kingdom and the United States, in order to
answer two different questions. First, is the degree of pass-through affected by the
direction of the exchange rate change? Second, does the size of the exchange rate change
matter for pass-through?
The motivation for conducting this research stems from a price survey carried out by
the Central Bank of Iceland in 2008. Approximately two-thirds of firms raised their
prices in response to a 30% depreciation of the Icelandic krona in 2008, while only fifth
lowered their prices following a 10% appreciation in 2007 (Ólafsson, Pétursdóttir and
Vignisdóttir, 2011). In addition, studies have shown (see for example Campa and
Goldberg, 2002, and Pétursson, 2008) that ERPT is more pronounced in Iceland than in
other inflation-targeting countries, making it an interesting subject to study. New Zealand
and Sweden are chosen for their similarities with Iceland, all countries being small open
economies with their own currency, while Germany, the U.K. and the U.S. are chosen in
order to make comparison with other studies possible.
The contribution of this study to the ERPT literature is twofold. First, it provides new
up-to-date estimates of ERPT for the six countries included in the study, using a larger
dataset than has previously been done for some of them. Second, it is the first, to my
knowledge, to explore whether ERPT to import prices in Iceland, New Zealand and
Sweden is asymmetric and non-linear.
The results indicate that asymmetry and non-linearity cannot be ignored when
estimating ERPT. For two countries, the U.S. and Sweden, asymmetry cannot be
rejected. The direction of asymmetry varies between the countries, with higher ERPT
during appreciations of the US dollar but during depreciations of the Swedish krona. For
Iceland, New Zealand and the U.K. symmetric ERPT cannot be rejected. However, the
size of the difference between the estimated ERPT coefficients for appreciations and
depreciations is considerable, suggesting caution when interpreting the estimates from
the linear and symmetric model. For no country can the restriction of linear pass-through
be rejected, unless when also taking into account the direction of the change, suggesting
that the direction effects overshadow the size effects.

3
 The paper is structured as follows. Section 2 presents the theoretical microeconomic
assumptions that can generate asymmetric and non-linear ERPT and reviews the findings
of previous empirical studies. Section 3 presents the empirical framework, its underlying
foundations and describes the data. Section 4 discusses the results and several robustness
tests and finally, section 5 concludes.

2 Literature review
2.1 Exchange rate pass-trough
To explain the concept of ERPT intuitively it is convenient to use a simple example.
Suppose that there are two countries, for example the U.K. and Iceland, where the U.K.
is the importing country, Iceland the exporting country and codfish the traded good. The
export price of codfish is expressed in Icelandic kronas while the import price in British
pounds. In this simple example, the difference between import and export price only
depends on the exchange rate between the two currencies. Now, suppose that the pound
appreciates against the krona. As a result, fewer pounds are needed to buy the same
quantity of codfish as before. Therefore, the import price of codfish, in pounds,
decreases. In other words, the appreciation of the pound is “passed through” to import
prices, hence the term exchange rate pass-through.
More formally, ERPT can be defined as the impact of a one percent change in the
exchange rate between the exporting and importing countries, on import prices in local
currency. Generally speaking, ERPT is said to be “complete” if import prices respond
one-for-one to changes in exchange rates. From the literature´s standpoint1, this occurs if
exporters keep prices in their home currency stable, so the whole exchange rate change is
passed through to import prices. In contrast, ERPT is said to be zero if exporters adjust
prices in their home currency, following a change in the exchange rate, so that import
prices in local currency remain stable. However, if exporters only partially adjust prices
in their home currency, ERPT is said to be “incomplete” (Goldberg and Knetter, 1997).
For complete ERPT to be realized two conditions are required. First, mark-up of price
over marginal cost must be constant. If this condition does not hold, exporters can adjust
their mark-up in order to absorb exchange rate changes, limiting pass-through. Second,

1
The existing literature models ERPT by considering how exporters change their prices when the exchange rate
changes, see e.g. Bussière (2013).

4
marginal cost must be constant. For this condition to hold, exporters must only use
domestically produced inputs in their production process. If imported inputs are used as
well, then an appreciation of the foreign currency will reduce marginal cost, as imported
inputs will be cheaper, resulting in a decrease in export prices and thus an incomplete, or
in extreme cases zero, ERPT. These two conditions, however, are seldom satisfied in
practice (Campa, Goldberg and González-Mínguez, 2005).
Interest in ERPT research steadily increased as the number of empirical studies who
rejected the Purchasing Power Parity (PPP) grew. Unsurprisingly perhaps, as the two
concepts are closely related: if PPP holds, ERPT is complete2. Take for example the
absolute version of PPP:

𝑃 = 𝐸𝑃 ∗
where P can be interpreted as import price in local currency, P* as export price in foreign
currency and E is the exchange rate, defined as local currency per unit of foreign
currency. Suppose the exchange rate increases by 10% (more units of local currency are
needed in order to buy one unit of foreign currency) and export price remains stable. It is
clear that for PPP to hold import price must also increase by 10%, indicating complete
ERPT. Thus, complete ERPT only occurs if export price remains stable, that is, if both
mark-up and marginal cost are constants (Herzberg, Kapetanios and Price, 2003).
In line with the documented failure of PPP, researchers have in general found ERPT
to be incomplete. According to Goldberg and Knetter (1997), ERPT to U.S. import prices
is around 60%, though this number differs between industries. More recent studies on
U.S. import prices report lower pass-through, with Ihrig, Marazzi and Rothenberg (2006)
estimating a long-run pass-through as 32%, same as Bussière, Chiaie and Peltonen
(2014). It should be noted, however, that ERPT into U.S. import prices is relatively low
when compared to other advanced economies. For example Campa and Goldberg (2005)
find that average pass-through to import prices for a sample of 23 OECD countries is
46% in the short-run and 64% in the long-run.
Majority of the existing ERPT literature seeks to answer two questions, why ERPT is
incomplete, even in the long-run, and why it changes over time. Broadly speaking, the
theoretical literature can be divided into two groups: the first being from a
microeconomic perspective and the second from a macroeconomic perspective. The first
group, which could be further divided into cost theories and competition theories, focuses

2
This results applies to both absolute and relative PPP.

5
on the industrial structure of the economy. For example, in a seminal paper Dornbusch
(1987) expounds how the structure of competition in an industry affects ERPT. He
argues that a more active competition prompts exporters to adjust their mark-ups in
response to changes in the exchange rate, rather than pass the change fully through to
import prices, in order to maintain their competitiveness. He summarizes that ERPT
depends on market structure, product substitutability and the number of foreign firms
relative to local firms. Froot and Klemperer (1989), using a dynamic model where
exporter´s profit tomorrow depends on market share today, show that exporter faces a
trade-off between raising current or future profits when the local currency appreciates.
The second group of literature focuses on macroeconomic environment, especially the
role of monetary policy. According to Taylor (2000), the low inflation rate in the
developed countries, due to a tightening and more credible monetary policy, has led to a
decline in the degree of ERPT. This hypothesis has been supported by several empirical
studies, for instance Gagnon and Ihrig (2004) and Bailliu and Fujii (2004). Numerous
other factors, which I will not expand on, have been put forward as determinants of
incomplete and declining ERPT3.

2.2 Asymmetric and non-linear exchange rate pass-through

The existing empirical literature on ERPT, generally assumes that the relationship
between exchange rate and import prices is (i) symmetric and (ii) linear. In other words,
it is presumed that appreciations and depreciations of the local currency have the same
impact4 on import prices and that the size of the exchange rate change does not matter for
the degree of pass-through (Bussière, 2013). These assumptions are, however, not very
realistic. In fact, there exist numerous theoretical microeconomic assumptions that can
generate asymmetric and non-linear ERPT. In this section the key assumptions will be
briefly summarized.

Market Share
Pricing-to-market (PMT), as labelled by Krugman (1986), is the most cited explanation
for incomplete ERPT. In this context, PMT refers to the pricing behavior of an exporter
when faced with local currency exchange rate changes. Suppose that an exporter has the
objective of maintaining market share in his export market. By strategic pricing the

3
See for example Mann (1986) and Gust, Leduc and Vigfusson (2010).
4
With the opposite sign.

6
exporter may adjust his mark-up downwards5 when the local currency depreciates in
order to maintain his market share but retain his mark-up and allow the import price to
fall when the local currency appreciates (Gil Pareja, 2000). Goldberg and Knetter (1997)
provide an excellent example of actual pricing-to-market strategy, implemented by
Toyota in the 1990s. In 1994 a Japanese made Toyota Celica cost $16,968 in the U.S.
One year later, the price had increased by two percent. However, over the same time
period, the yen appreciated by 34% against the dollar. Thus, in order to maintain their
market share, Toyota decreased their mark-up, letting it absorb the yen appreciation,
effectively decreasing their export price and limiting ERPT to U.S. import price.
Generally, exporter raises his export price less when the local currency appreciates than
he reduces his price when the local currency depreciates, making ERPT into import
prices greater during appreciations than depreciations, as in Marston (1990).

Binding Quantity Constraints

Binding quantity constraints arise when exporter´s ability to increase sales in his export
market when the local currency appreciates, is limited. As Knetter (1994) points out,
there are several scenarios where quantity constraints might occur, for example when
trade restrictions like quotas apply or when exporter is unable to expand his marketing
capacity. For instance, in the 1980s, U.S. imports of Japanese made automobiles were
restricted. As a result, Japanese automobiles manufacturers were faced with binding
quantity constraints: as the dollar appreciated, Japanese exporters were unable to exploit
their gain in price competitiveness by letting their dollar price fall, as the allowed
quantity of Japanese automobiles sold in the U.S. had already been reached. Since
passing the dollar appreciation to import prices was not possible, exporters increased
their mark-ups instead and charged the market clearing dollar price (Knetter, 1994).
Quantity constraints are only binding during appreciations of the local currency.
Therefore, the degree of ERPT to import prices is higher for depreciations of the local
currency than appreciations (Pollard and Coughlin, 2004).

Downward Price Rigidities

As Peltzman (2000) eloquently put, “prices rise faster than they fall”. This statement
implies an asymmetric ERPT. To elaborate, suppose that there are two countries doing
business, for example, the U.K. imports goods from Germany. Now, if the pound
appreciates against the euro, German exporters might let their mark-ups absorb the pound
5
Note that export prices are always given in the foreign currency unless otherwise stated.

7
appreciation, consequently raising their export prices in euros but keeping prices in
pounds steady. On the other hand, if the pound depreciates against the euro, absorbing
the depreciation and keeping prices in pounds steady would mean that German exporters
would have to lower their export prices in euros. As prices are generally considered rigid
downwards, one would expect German exporters to be more reluctant to decrease their
export prices than increase them. Thus, depreciation of the pound will be passed through
to U.K. import prices to a greater extent than appreciations.

Production Switching
Although not often cited, production switching can be an important reason for
asymmetric ERPT. Following Webber (2000) and Pollard and Coughlin (2004), suppose
that there are two countries trading with each other. Continuing with the example from
above, now assume that German exporters can choose whether to import production
inputs from the U.K. or use domestically produced inputs. Logically, German exporters
will only use domestically produced inputs when the pound appreciates against the euro,
as they have become relatively less expensive. Thus, as appreciation of the pound does
not affect exporters marginal cost6, it is assumed that export prices will remain stable,
implying a high degree of pass-through, ceteris paribus. On the other hand, when the
pound depreciates against the euro, German exporters will only use imported inputs.
Therefore, depreciation of the pound will decrease exporters marginal cost, resulting in
lower export prices, all other things being equal, indicating no or limited pass-through as
the decrease in export prices offsets the depreciation of the pound.

Menu and Switching Costs

Of the key assumptions mentioned in this section, menu cost is the only one which refers
to the size of the exchange rate change, not the direction of the change. In the presence of
menu costs exporter may decide to let small fluctuations in the local currency exchange
rate pass by without changing his export price, only adjusting his price if the change in
the exchange rate exceeds a certain threshold. Thus, the degree of pass-through is higher
for smaller changes than larger ones. As Pollard and Coughlin (2004) show, this result
only applies if imports are invoiced in the exporter´s currency, otherwise the opposite
holds true. To illustrate, suppose imports are invoiced in the local currency. In this
framework exporter holds the local currency price stable when the exchange rate change
is small, as changing it is costly, by letting his mark-up absorb the change. This implies
6
As the marginal cost only depends on prices of domestically produced inputs and not the exchange rate.

8
zero, or very limited, ERPT. However, if the exchange rate change is large enough the
exporter will change his price, resulting in a higher degree of pass-through. On the other
hand, if imports are invoiced in the exporter´s currency, changes in the exchange rate will
not affect the payment received by the exporter. Thus, for small changes in the exchange
rate, he will not change his export price, implying complete ERPT, other things being
equal. If the exchange rate change exceeds a certain threshold, the exporter will change
his price, limiting ERPT.
On a similar note, switching costs can also generate asymmetric pass-through as
Bussière (2013) points out. As long as the price in local currency does not exceed a
certain threshold, after which local consumer would switch to a different product,
exporter will not respond to exchange rate changes by adjusting his price.

2.3 Previous empirical literature

Despite the abundance of empirical literature on ERPT only a handful of studies
consider, and test for, asymmetry and non-linearity. Most of these studies are micro-
oriented and conducted at the industry level, while studies using aggregate data are
scarcer. Very few studies focus on ERPT to consumer prices. Previous papers have found
mixed results when testing for asymmetry and non-linearity. Furthermore, there seems to
be no consistency in the direction of asymmetry, as some studies associate appreciations
of local currency with higher degree of pass-through than depreciations, while other
report the opposite.

Industry level
While most micro-oriented studies focus on several industries at a time, Goldberg (1995)
and Kadiyali (1997) only look at one industry each. Goldberg studies ERPT in the U.S.
automobile industry. Specifically, she examines U.S. imports of Japanese and German
cars using a discrete choice model. Kadiyali, however, studies ERPT in the U.S.
photographic print film industry by focusing on pricing by Fuji Photo Film of Japan.
Both find that ERPT is greater when the dollar depreciates.
Focusing on import prices in 30 U.S. industries, Pollard and Coughlin (2004) utilize a
profit maximization model to test whether the direction or the size of the exchange rate
change matters for pass-through. They find evidence of asymmetric responses to
appreciations and depreciations of the dollar in half of the industries but the direction of
asymmetry varies. In addition, evidence of non-linearity is found in more than half of the

9
industries. In all cases, there is a positive relationship between the size of the exchange
rate change and the degree of pass-through, indicating invoicing in dollars. Finally, they
conclude that the direction effects are overshadowed by the size effects. Similarly, Yang
(2007) tests whether ERPT to US import prices is asymmetric using disaggregated U.S.
industry data. Unlike Pollard and Coughlin, Yang specifies an interactive dummy
variable which takes the value one after March 1985, when the dollar peaked, and zero
before that time. The author finds evidence of asymmetric ERPT in few industries but the
direction of asymmetry varies.
Kanas (1997) also finds evidence of asymmetric pass-through when examining export
prices of eight commodities exported from the U.K. to the U.S. On the other hand, Olivei
(2002) finds little evidence of asymmetric ERPT to U.S. import prices while Feinberg
(1989) finds no evidence at all.
Wickremasinghe and Silvapulle (2004) use asymmetric models to show that in the
long-run Japanese import prices of manufacturing respond asymmetrically to
appreciations and depreciations of the yen. Meanwhile, Ohno (1989) examines the price
setting of Japanese export manufacturers in a mark-up over cost framework. He finds
evidence of asymmetric ERPT in three machinery and equipment industries, where
exporters are keener to raise their prices when the yen depreciates than lower them when
it appreciates. In addition, Ohno tests for non-linearity in the ERPT. He obtains mixed
results: in some industries large changes in the yen exchange rate result in large price
adjustments whereas the same exchange rate change causes small adjustments in others.
Marston (1990) investigates export pricing by Japanese manufacturing firms and finds
significant evidence of PTM behavior. For five industries, out of seventeen under
observation, adjustment of mark-ups is greater when the yen appreciates. This indicates a
higher degree of pass-through when the yen depreciates. In contrast, Athukorala and
Menon (1994) reject that Japanese exporters employ PTM strategies in times of yen
appreciations. In fact, they find no evidence of asymmetric ERPT. Similarly, Knetter
(1994) seldom rejects the symmetric hypothesis for Japanese and German export prices.
Gil-Pareja (2000) tests for asymmetry in ERPT while focusing on a range of industries
across a sample of European Union countries. In most cases the author cannot reject
symmetric responses of import prices. However, analysis of coefficient estimates shows
that the direction of asymmetry varies across industries as well as countries. More
recently Campa, Mínguez and Barriel (2008) find significant evidence of non-linear

10
ERPT in EU15 countries while examining the adjustment of import prices towards their
long-run equilibrium following an exchange rate change.

Aggregate level
Using aggregate trade data for eight countries across the Asia-Pacific, Webber (2000)
investigates asymmetry in ERPT to import prices. Webber´s results strongly support
asymmetric pass-through: in five countries the hypothesis of asymmetry could not be
rejected. Bussière (2013) examines whether ERPT to export and import prices in the G7
countries is non-linear and asymmetric. Using a standard linear model, which is later
augmented with polynomial functions of the exchange rate, Bussière´s findings indicate
that non-linear effects are of importance and should be considered when estimating
ERPT. However, the findings show a great deal of cross-country variation as both the
direction of asymmetry and the magnitude of non-linearities are different for each
country. Stronger evidence of asymmetry and non-linearity are found in export prices
than import prices. In contrast, Herzberg et. al. (2003) find no evidence of non-linear
ERPT to U.K. import prices.
El bejaoui (2013) uses an Asymmetric Cointegrating Autoregressive Distributed Lag
(ARDL) model to investigate the possibility of asymmetric ERPT to export and import
prices in the U.S., Germany, France and Japan. He finds evidence of asymmetric
responses in the long-run, for both import and export prices.
Delatte and López-Villavicencio (2012) also use an ARDL model while focusing on
ERPT to consumer price indexes in the U.S., U.K., Germany and Japan. They find that
depreciations of the local currency are associated with higher degree of pass-through than
appreciations. They argue that their findings indicate a weak market competition.
Similarly, Przystupa and Wróbel (2011) find evidence of asymmetric responses of
consumer prices to exchange rate changes in Poland, while Correa and Minella (2006)
find higher degree of pass-through to consumer prices in Brazil when the local currency
depreciates sufficiently.

11
3 Methodology
3.1 Analytical framework
Before introducing the empirical specification it is useful to first acquaint oneself with its
underlying foundations. The most straightforward way of testing the relationship
between exchange rate and import prices is by estimating a simple reduced form
equation:

∆𝑝𝑡𝑀𝑃 = 𝛽∆𝑒𝑟𝑡 + 𝜀𝑡 (1)

where 𝑝𝑡 is the natural logarithm of import price, 𝑒𝑟𝑡 is the natural logarithm of nominal
exchange rate, defined as local currency per unit of foreign currency, 𝜀𝑡 is an error term
and 𝛽 is the ERPT coefficient. If 𝛽 = 1 then ERPT is complete. This simple specification
has not, however, established itself in the literature as it lacks economically meaningful
specifications7. Instead, researchers have adopted a micro-oriented approach, focusing on
the pricing behavior of exporters.
The model presented in this paper follows Goldberg and Knetter (1997), Bailliu and
Fujii (2004) and Bahroumi (2006), to name a few, and is common in the ERPT literature.
A representative foreign firm, which exports its product to a home country, is assumed to
enjoy market power in its exporting market. The foreign firm maximizes its profit in its
own currency (foreign currency henceforth) by solving the following profit maximization
problem:
max 𝜋 = 𝑃 ∗ 𝑄(𝑃∗ ) − 𝐶 ∗ (𝑄) (2)
𝑃∗
where 𝜋 is the profit in foreign currency, 𝑃∗ is the price of the good in foreign currency,
𝑄 is the quantity produced and 𝐶 ∗ is the cost function in foreign currency units. The first-
order condition of the profit maximization can be written as:
𝑃∗ = 𝑐 ∗ 𝜇 (3)

where 𝜇 is the mark-up over marginal cost, defined as 𝜇 = 𝜂⁄(𝜂 − 1), where 𝜂 is the
elasticity of demand8. In the home country, import price is expressed in the local
currency:
𝑃𝑀𝑃 = 𝐸𝑃∗ = 𝐸𝑐 ∗ 𝜇 (4)

7
See Campa and Goldberg (2002).
8 𝑑𝑄 𝑃
The formula for the price elasticity of demand is:
𝑑𝑃 𝑄

12
where 𝑃𝑀𝑃 is the import price expressed in the local currency and 𝐸 is the nominal
exchange rate. It is clear from equation (4) that exchange rate changes and changes in the
foreign firm´s marginal cost and mark-up can alter the local currency import price. Note
that changes in marginal cost and mark-up do not necessarily occur as results of changes
in the exchange rate. For instance, the foreign firm´s marginal cost can shift because of
changes in factor prices. As an example, wage increase in the foreign country can result
in increased marginal cost for the foreign firm. Demand conditions and demand shocks in
the home country can also shift the foreign firm´s marginal cost, as marginal cost is
increasing in quantity9. Furthermore, the firm´s mark-up depends on the elasticity of
demand. The elasticity does not only depend on pricing decisions of the foreign firm but
also on the pricing decisions of firms in the home country. If the foreign firm is engaged
in a competition with firms in the home country it will face more elastic demand, causing
the mark-up to decrease.
To finalize the model, the log-linear form of equation (4) may be expressed as
follows:

𝑝𝑡𝑀𝑃 = 𝛼 + 𝛽𝑒𝑟𝑡 + 𝛾𝑤𝑡∗ + 𝛿𝑦𝑡 + 𝜀𝑡 (5)

where 𝑤𝑡∗ denotes foreign marginal cost of production and 𝑦𝑡 is demand conditions in the
home country. Equation (5), and different variations of it, has entrenched itself in the
ERPT literature, as noted by Goldberg and Knetter (1997).

3.2 Empirical specification

3.2.1 Linear specification

In this paper, a variation of equation (5) is used to estimate ERPT. The following basic
regression equation is estimated separately for each country, in log differences and with
quarterly data:

Δ𝑝𝑡𝑀𝑃 = 𝛼 + 𝛽∆𝑒𝑟𝑡 + 𝛾Δ𝑤𝑡∗ + 𝛿𝑍𝑡 + 𝜀𝑡 (6)

(+) (+)

where 𝑝𝑡𝑀𝑃 is the natural logarithm of aggregate import price, 𝑒𝑟𝑡 is the natural logarithm
of nominal effective exchange rate, 𝑤𝑡∗ is the natural logarithm of foreign marginal cost

9
If the law of diminishing returns holds then marginal cost will eventually increase when quantity
increases.

13
of production, defined in foreign currency, and 𝑍𝑡 is a vector of other controls. In order to
estimate the long-run ERPT, dynamics are introduced to equation (6):

Δ𝑝𝑡𝑀𝑃 = 𝛼 + ∑𝑘𝑗=0 𝛽𝑗 ∆𝑒𝑟𝑡−𝑗 + ∑𝑘𝑗=0 𝛾𝑗 Δ𝑤𝑡−𝑗

∗
+ 𝛿𝑍𝑡 + 𝜀𝑡 (7)
(+) (+)
Following Campa and Goldberg (2002), short-run ERPT is given by the coefficient 𝛽0
while long-run ERPT is given by the sum of the coefficients on the contemporaneous and
lagged changes in the exchange rate. In order to keep the analysis uniform across the
observed countries, only one lag10 of each explanatory variable is included, so that k=1.
The models are estimated using Ordinary Least Squares (OLS). Furthermore, to account
for possible heteroskedasticity, robust standard errors are employed.
The expected signs for the coefficients 𝛽 and 𝛾 are given in brackets under the
equations. In this specification the exchange rate is defined as local currency per unit of
foreign currency. Therefore, an increase in 𝑒𝑟𝑡 , which in this case means a depreciation
of the local currency, should increase the aggregate import price. As measuring foreign
marginal cost of production, the primary control variable, on an aggregate level is
particularly difficult, foreign producer price index for all commodities is used as a
proxy11. It is expected that a cost increase, represented as an increase in wt∗ , should raise
import price.
The vector of control variables 𝑍𝑡 is defined as 𝑍𝑡 = [𝑃𝑃𝐼𝑡 , 𝑜𝑖𝑙𝑡 , 𝑆1, 𝑆2, 𝑆3] where
𝑃𝑃𝐼𝑡 denotes domestic producer price index, 𝑜𝑖𝑙𝑡 stands for oil prices in local currency
and 𝑆1, 𝑆2, 𝑆3 are quarterly dummy variables to capture seasonal effects, as the data is
not seasonally adjusted. Similar to Olivei (2002), the variable 𝑃𝑃𝐼𝑡 is included as a proxy
for competitor’s prices in the home country, capturing the tendencies of pricing to
market. As prices of competitive goods change, exporters will change their prices
accordingly in order to maintain their market share. Thus, one can expect a positive
relationship between import price and PPI. In addition, although not specifically
considered here, including PPI as a proxy for domestic cost can account for exporter´s
local distribution cost, as discussed by Aron, Macdonald and Muellbauer (2014). No

10
I start by including up to four lags of the explanatory variables in the regression. To determine the
appropriate value of k, I use an F-test to evaluate the joint significance of the same lag lengths of each
variable (for example the fourth lag of each variable). If they are not jointly significant, those lags are
dropped from the model. For the majority of the observed countries, only one lag of each variable is
significant. Thus, to keep the analysis uniform across countries, k=1.
11
Preferably, I would have liked to use foreign unit labor cost as a proxy for foreign marginal cost of
production, as it gives a clearer picture of the production cost firms encounter. However, due to data
limitations, this was not an option.

14
direct demand terms, like gross domestic product (GDP) or output gap, are included in
the specification, even though theory would suggest otherwise, mainly because in
practice they often prove to be insignificant12. The reasoning behind this is that PPI
already captures the shift in domestic demand, as prices tend to rise as demand increases.
Following Marazzi et al. (2005), I include oil prices in local currency as a measure of
commodity prices. By doing so, I control for the direct effects that oil prices have on
import prices. As Bussière (2013) points out, the pass-through to oil prices, and
commodity prices in general, is usually very high. Therefore, failing to take them into
account could result in overestimating the ERPT. In addition, it is advisable to account
for the volatile nature of oil prices by including them as an explanatory variable when
dealing with aggregate import price.
Finally, to test for the presence of unit root in the series, an Augmented Dicky Fuller
(ADF) test is conducted. The test results are reported in Table A1 in the Appendix. The
results indicate that all variables are stationary in first difference, or integrated of order
one, I(1). Given the non-stationarity of the series in log levels and integration of the same
order, I perform an Engle-Granger (1987) two-step method to determine whether the
three key variables are cointegrated. That is, I test whether a linear combination of import
price, exchange rate and foreign marginal cost of production produces a stationary
process. The results, also reported in Table A1, do not support the cointegration
hypothesis, as the residuals are non-stationary for all countries except New Zealand.
Consequently, I estimate the models in log differences, instead of applying an error
correction model.

3.2.2 Allowing for asymmetry and non-linearity

In equations (6) and (7) ERPT is assumed to be symmetric and linear, meaning that
neither the direction nor the size of the exchange rate change matters, the degree of pass-
through is the same either way. However, as previously noted, these assumptions are not
very realistic. Thus, the possibility of asymmetric and non-linear responses has to be
taken into account. Following Pollard and Coughlin (2004), two dummy variables are
created13, one representing appreciations of the local currency and the other

12
This is formally tested below using real GDP as a proxy for demand condition in the home country.
13
It is also possible to estimate ERPT for appreciations and depreciations by creating only one dummy
variable, generating the same results. However, as using two dummy variables generates results that are
easily interpreted, and is in fact more common in the literature, I choose that method.

15
depreciations, in order to determine whether the direction of the exchange rate change
matters for the degree of ERPT. More specifically:
1 𝑖𝑓 Δ𝑒𝑟𝑡 < 0 1 𝑖𝑓 Δ𝑒𝑟𝑡 > 0
𝐴𝑡 = { 𝑎𝑛𝑑 𝐷𝑡 = {
0 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒 0 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒
These dummy variables are then interacted with the exchange rate and plugged into
equation (6), generating an equation that provides two different estimates of the ERPT,
one for appreciations and the other for depreciations of the local currency:

Δ𝑝𝑡𝑀𝑃 = 𝛼 + 𝛽𝐴 (𝐴𝑡 Δ𝑒𝑟𝑡 ) + 𝛽𝐷 (𝐷𝑡 Δ𝑒𝑟𝑡 ) + 𝛾Δ𝑤𝑡∗ + 𝛿𝑍𝑡 + 𝜀𝑡 (8)

The inclusion of two dummy variables allows for formally testing whether the degree of
ERPT is significantly different for appreciations and depreciations. This is done by
testing the coefficient restriction 𝛽𝐴 = 𝛽𝐷 . Rejection of this restriction indicates that
asymmetry is present.
To determine whether the size of the exchange rate change matters for the degree of
ERPT, two dummy variables are created, one representing large exchange rate changes
and the other small exchange rate changes. When deciding what constitutes as a large
exchange rate change I follow Bussière (2013) and define the threshold value for each
country14 as being equal to one standard deviation of the quarterly exchange rate change.
More specifically:
1 𝑖𝑓 |Δ𝑒𝑟𝑡 | ≥ 𝑠𝑡. 𝑑𝑒𝑣 1 𝑖𝑓 |Δ𝑒𝑟𝑡 | < 𝑠𝑡. 𝑑𝑒𝑣
𝐿𝑡 = { 𝑎𝑛𝑑 𝑆𝑡 = {
0 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒 0 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒
Like before, these dummy variables are interacted with the exchange rate and plugged
into equation (6), generating an equation that provides two different estimates of ERPT,
one for large exchange rate changes and the other for small exchange rate changes:

Δ𝑝𝑡𝑀𝑃 = 𝛼 + 𝛽𝐿 (𝐿𝑡 Δ𝑒𝑟𝑡 ) + 𝛽𝑆 (𝑆𝑡 Δ𝑒𝑟𝑡 ) + 𝛾Δ𝑤𝑡∗ + 𝛿𝑍𝑡 + 𝜀𝑡 (9)

Rejection of the restriction 𝛽𝐿 = 𝛽𝑆 indicates that the size of the exchange rate change
matters for the degree of ERPT. As discussed in Pollard and Coughlin (2004), it depends
on the currency of invoice whether or not one would expect the degree of ERPT and the
size of the exchange rate to be positively correlated. For example, as imports of larger
economies are usually invoiced in their own currencies, one can assume that the degree
of pass-through is higher for larger exchange rate changes. The opposite might be

14
Pollard and Coughlin (2004) choose the same threshold value for all sectors in their study. However,
doing so in this study could generate misleading results, as the exchange rate changes in the countries
under observation are vastly different. What constitutes as a large change in Germany for example would
be unremarkable in Iceland.

16
assumed to hold for smaller economies, where the home market is small and the
exchange rate often more volatile.
The final specification in this section combines the former two, i.e. it takes into
account both the direction and size of the exchange rate change. Thus, four new dummy
variables are created:
1 𝑤ℎ𝑒𝑛 𝐿𝑡 = 1 𝑎𝑛𝑑 𝐴𝑡 = 1 1 𝑤ℎ𝑒𝑛 𝑆𝑡 = 1 𝑎𝑛𝑑 𝐴𝑡 = 1
𝐿𝐴𝑡 = { 𝑎𝑛𝑑 𝑆𝐴𝑡 = {
0 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒 0 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒
1 𝑤ℎ𝑒𝑛 𝐿𝑡 = 1 𝑎𝑛𝑑 𝐷𝑡 = 1 1 𝑤ℎ𝑒𝑛 𝑆𝑡 = 1 𝑎𝑛𝑑 𝐷𝑡 = 1
𝐿𝐷𝑡 = { 𝑎𝑛𝑑 𝑆𝐷𝑡 = {
0 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒 0 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒
Interacting the dummy variables with the exchange rate and substituting into equation (6)
gives an equation that provides different estimates of ERPT, for large appreciations, for
small appreciations, for large depreciations and the final for small depreciations:

Δ𝑝𝑡𝑀𝑃 = 𝛼 + 𝛽𝐿𝐴 (𝐿𝐴𝑡 Δ𝑒𝑟𝑡 ) + 𝛽𝑆𝐴 (𝑆𝐴𝑡 Δ𝑒𝑟𝑡 ) + 𝛽𝐿𝐷 (𝐿𝐷𝑡 Δ𝑒𝑟𝑡 ) + 𝛽𝑆𝐷 (𝑆𝐷𝑡 Δ𝑒𝑟𝑡 ) +
𝛾Δ𝑤𝑡∗ + 𝛿𝑍𝑡 + 𝜀𝑡 (10)

As before, the threshold is defined as being equal to one standard deviation of the
quarterly exchange rate change.

3.4 Data
The dataset is composed of quarterly, aggregate data for six different countries:
Germany, Iceland, New Zealand, Sweden, the United Kingdom and the United States.
The sample period is from 1980:Q1 through 2014:Q4 for each country. All series are in
index form, with 2010:Q3 = 100, and are not seasonally adjusted. For most variables,
data is collected from the International Monetary Fund´s (IMF) International Financial
Statistics (IFS) database.
The quarterly series for import prices come from the IFS, for all countries except
Iceland. From the IFS, I use import price for all commodities as the dependent variable.
The IFS does not report import price for Iceland so instead I use the import price deflator
calculated by the Central Bank of Iceland as dependent variable. Series for nominal
effective exchange rates (NEER) for all countries under observation come from the IFS.
NEER is a trade weighted series, calculated by using consumer price index (CPI) of the
home country and of its trading partners. As in my specification the exchange rate is
defined as domestic currency per unit of foreign currency, I divide one with the NEER
series (1/NEER).

17
 Measuring foreign marginal cost of production is hard, as it is not directly observable.
In the literature it is common, see for example Campa and Goldberg (2002), to construct
a proxy for foreign marginal cost of production by computing 𝑈𝐿𝐶𝑡∗ = 𝑁𝐸𝐸𝑅𝑡 𝑈𝐿𝐶𝑡 /
𝑅𝐸𝐸𝑅 where 𝑈𝐿𝐶𝑡∗ is foreign unit labor cost, 𝑈𝐿𝐶𝑡 is the local unit labor cost and NEER
and REER are the nominal and real effective exchange rate, respectively, calculated by
using unit labor cost. This gives a proxy for foreign marginal cost of production, where
the cost of each trading partner is weighted by its importance in the home country´s
trade. Because of insufficient data15, I choose a slightly different method. Instead of
using foreign ULC as a proxy for marginal cost of production I use foreign PPI. For each
country under observation I calculate 𝑤𝑡∗ = ∑5𝑗=1 𝜔𝑗,𝑡 𝑃𝑃𝐼𝑗,𝑡
∗
where 𝜔𝑗,𝑡 is a weight,
assigned to each of the top five trading partners of the home country, depending on their
importance in the home country´s trade. The weights are changed every five years, to
account for changes in trading patterns during the observed time period16. The necessary
data to calculate the weights come from statistical offices of each country while the PPI
series come from the IFS.
Data for domestic PPI comes from the IFS for all countries, except Iceland. As
Statistics Iceland has only compiled data on Icelandic producer prices from 2006, no
older data is available. Thus, for Iceland, I include real GDP instead of domestic PPI. Oil
prices are from the U.S. Energy Information Administration (EIA), where spot prices per
barrel of WTI are used for the U.S. and New Zealand and spot prices per barrel of Brent
for the European countries. The spot prices are converted into domestic currencies.
Finally, I exploit that the IFS reports both NEER and REER, calculated from CPI, to
create a variable for foreign prices by computing 𝐶𝑃𝐼𝑡∗ = 𝑁𝐸𝐸𝑅𝑡 𝐶𝑃𝐼𝑡 /𝑅𝐸𝐸𝑅 where 𝐶𝑃𝐼𝑡
is the price level in the home country while 𝐶𝑃𝐼𝑡∗ is the trade-weighted price level of
home´s trading partners.

15
Time series for ULC were too short so including them meant shortening my sample by ten years. In
addition, REER is not calculated for Iceland by using ULC.
16
For example, after 2000, imports from China increased in all observed countries.

18
4 Results
4.1 Results from the linear specification
The results from the linear specification are summarized in Table 1. I start by estimating
the simple reduced form equation ∆𝑝𝑡𝑀𝑃 = 𝛽∆𝑒𝑟𝑡 + 𝜀𝑡 . As there are no variables
included to control for certain factors and to isolate the effects of the exchange rate on
import price, the pass-through coefficient reported in Panel A of Table 1 can be
interpreted as a combination of direct and indirect effects of the exchange rate on import
price. Because of the inclusion of indirect responses, the pass-through coefficients
reported in Panel A should be considerably higher than the ones in Panels B and C, as is
the case for most countries.
The estimated pass-through coefficients reported in Panel B are obtained by
estimating equation (6), which in addition to the exchange rate includes a set of
explanatory variables. By controlling for indirect effects of the exchange rate on import
price, operating through changes in different factors, the pass-through coefficients should
only reflect direct effects of the exchange rate. Finally, I estimate equation (7). As the
equation includes dynamics I can calculate both the short- and long-run pass-through.
The results are reported in Panel C. The full results of equations (6) and (7) are reported
in Tables A2 and A3 in the Appendix. Overall, the models perform are satisfactorily:
most key variables are statistically significant with the expected signs, the goodness-of-
fit seems to be high and the residuals show low or no serial correlation.

Table 1: Summary results for the linear specification

DE IS NZ SE UK US
Panel A: Import price on exchange rate
ERPT 0.688 *** 0.965*** 0.678*** 0.391*** 0.331*** 0.401***

Panel B: Import price on exchange rate and controls, static

ERPT 0.498*** 0.962*** 0.586*** 0.208*** 0.351*** 0.164***

Panel C: Import price on exchange rate and controls, dynamic

Short-run ERPT 0.531*** 0.909*** 0.562*** 0.215*** 0.312*** 0.120***

Long-run ERPT 0.556*** 0.940*** 0.719*** 0.275*** 0.482*** 0.195***

Note: The full results are reported in Tables A2 and A3 in the Appendix. A positive coefficient indicates
that a depreciation of the home currency results in an increase in import price. ***, ** and * indicates
significance at 1%, 5% and 10% level respectively.

On the whole, the results are in line with the literature. The pass-through coefficients
reported in Panels B and C can be compared with existing results for most of the

19
countries included in this study. Starting with the U.S., the pass-through coefficients
reported here are slightly lower compared to Campa and Goldberg (2005, henceforth
CG), Choudhri and Hakura (2012, henceforth CH) and Bussière (2013, henceforth MB)
who estimate short-run ERPT coefficient as 23%, 38% and 23%, respectively. The
difference between the long-run estimates is slightly greater, perhaps unsurprisingly
considering I only use one lagged term of the exchange rate to calculate the long-run
pass-through while other studies use up to three or four lagged terms. For the U.K., the
present estimates are close to CG (36% and 46%)17 and MB (39% and 48%) but slightly
lower than Ihrig et al. (2006, henceforth IMR), who estimate a long-run pass-through
coefficient as 59%. For Germany, the estimates presented in Panel C are similar to CG
(55% and 80%) but higher than MB (33% and 36%). The present estimates for Sweden
are lower than CG (48% and 38%) and CH (39% in the short-run). For New Zealand the
estimates reported here are closer to CH (65% in the short-run) than CG (22% and 22%).
Estimated ERPT coefficients for Iceland stand out as they are much higher than those
for other countries. The results, however, are consistent with a recent report from the
Central Bank of Iceland (2011) who claims that ERPT is more prominent in Iceland than
in most other developed countries. As far as I know, only Campa and Goldberg (2002)
have estimated ERPT to aggregate import prices in Iceland. They report a short-run pass-
through coefficient of 118%, which seems unrealistic, and a long-run pass-through
coefficient of 76%. Pétursson (2008) estimates that ERPT to consumer prices in Iceland
is 43%, which is considerably higher than estimates for most other countries in his study.

4.2 Robustness tests

In order to check the robustness of the results four tests are conducted. First, I estimate
equation (6) using instrumental variable techniques in order to account for the possibility
of domestic prices being endogenous. If domestic firms are engaged in a competition
with foreign firms, they take import prices into account when pricing their goods.
Therefore it might be inaccurate to treat producer prices, which are used as a proxy for
domestic competitive prices, as an exogenous regressor. However, the results obtained by
using instrumental variable techniques, with lagged producer prices as instrument, are
similar to the OLS estimators, see Table A4 in the Appendix.

17
The first number represents short-run ERPT while the second number represents long-run ERPT. This
applies for all numbers in parenthesis in this section.

20
 Second, I add real GDP as an additional explanatory variable in order to test the
chosen specification, which does not include direct demand terms even though theory
suggests otherwise. Real GDP is chosen as a proxy to capture shifts in domestic demand,
primarily because of data availability. The demand term is insignificant for all
countries,18 see results in Table A5 in the Appendix, indicating that shifts in domestic
demand are already captured through producer prices.
Third, different proxy is used to capture domestic competitive prices. The literature is
divided when it comes to choosing appropriate proxy: some use ULC, others PPI and a
few CPI. As mentioned above, data for ULC is not available for all countries included in
the study and not for the whole sample period. Therefore I only replace PPI with CPI in
order to see if using different proxy yields different findings. The results, see Table A6 in
the Appendix, are by and large not that different from the benchmark case. However,
they indicate that using PPI is more appropriate, as the CPI coefficient is not significant
for all countries and noticeably lower than the PPI coefficient in most cases. In addition,
the goodness-of-fit falls slightly.
Finally, one may question the validity of the proxy for foreign marginal cost of
production and its construction. Therefore, a different proxy is tested, specifically foreign
consumer prices, whose construction is perhaps more solid considering that the CPI of
each trading partner is weighted by its importance in the home country´s trade instead of
just the top five trading partners. However, based on the estimation results, see Table A7
in the Appendix, CPI does not work as well as a proxy for foreign marginal cost of
production as PPI: the coefficient is only statistically significant for four countries, and of
those four coefficients, only two have the expected sign.

4.3 Results when allowing for asymmetry and non-linearity

Estimated ERPT coefficients, when allowing for asymmetry and non-linearity, are
presented in Table 2. Estimation results in Panel A correspond to the coefficients of
interest in equation (8), Panel B corresponds to equation (9) and Panel C to equation
(10). Full results can be found in the Appendix.

18
Iceland is not included in this robustness test.

21
Table 2: Summary results when allowing for asymmetry
DE IS NZ SE UK US
Panel A
Appreciations 0.503 *** 1.173*** 0.517*** 0.099* 0.267** 0.169***
Depreciations 0.492 *** 0.925*** 0.628*** 0.245*** 0.407*** 0.152
Wald test statistic 0.076 1.343 -0.878 -2.377** -1.085 0.149
Panel B
Large changes 0.501*** 0.961*** 0.563*** 0.219*** 0.345*** 0.159***
Small changes 0.486*** 0.970*** 0.722*** 0.153*** 0.374*** 0.181***
Wald test statistic 0.154 -0.121 -1.158 1.447 -0.252 -0.364
Panel C
Large 0.494*** 1.441*** 0.487*** 0.115** 0.260* 0.176***
appreciations
Small 0.412** 0.976*** 0.761*** -0.038 0.429*** 0.246**
appreciations
Large 0.503*** 0.934*** 0.603*** 0.247*** 0.391*** 0.129
depreciations
Small 0.546*** 0.962*** 0.668*** 0.330*** 0.303** 0.120
depreciations
Note: The full results are reported in Tables A8, A9 and A10 in the Appendix. ***, ** and * indicates
significance at 1%, 5% and 10% level respectively. Significant Wald test statistic, indicated by ***, ** and
* at 1%, 5% and 10% respectively, indicates that the coefficients are statistically different.

Starting with the results in Panel A, the estimated ERPT coefficients are statistically
significant at conventional levels, both for appreciations and depreciations, in all
countries except the U.S., where the coefficient for depreciation is not significant. Of
those countries, Sweden is the only one where the restriction 𝛽𝐴 = 𝛽𝐷 , meaning
symmetric pass-through, can be rejected. In the case of Sweden, ERPT is significantly
higher for depreciations of the Swedish krona than appreciations. These findings could
indicate downward price rigidities in Sweden´s import prices. In contrast, the estimated
ERPT coefficients for Germany are nearly identical, irrespective of the direction of the
exchange rate change, suggesting that the ERPT estimates in Table 1 could be considered
accurate. For Iceland, New Zealand and the U.K. however, the size of the difference
between the estimated ERPT coefficients for appreciations and depreciations is
considerable, even though the difference is not statistically significant. Therefore, I
conclude that for those three countries the symmetric estimates in Table 1 should be
taken with a grain of salt.
Ignoring for the time being that symmetric pass-through cannot be rejected, the
estimates for New Zealand and the U.K could indicate that import prices are rigid
downwards. On the other hand, for Iceland, appreciations appear to be passed through to
import prices to a greater extent than depreciations. However, the estimated ERPT

22
coefficient for appreciations is unrealistically high, suggesting caution when interpreting
the results. Pass-through to U.S. import prices is only significant when the US dollar is
appreciating, consistent with the market share theory. Thus, for the U.S., ERPT estimates
in Table 1 provide a misleading picture.
Turning to the results in Panel B, estimated ERPT coefficients are statistically
significant for all countries, irrespective of the size of the exchange rate change. For no
country can the restriction of 𝛽𝐿 = 𝛽𝑆 , or linear pass-through, be rejected. For Germany,
Iceland, U.K. and the U.S. the estimated ERPT coefficients for large and small exchange
rate changes are extremely similar. In contrast, for Sweden and New Zealand the size of
the difference between ERPT coefficients is larger, though not statistically significant. In
the case of New Zealand, where the difference is most noticeable, ERPT is greater when
the exchange rate changes are small. This is consistent with the menu cost assumption
when imports are priced in the exporter´s currency. Even though I am unable to
discriminate statistically between the coefficient estimates for large and small exchange
rate changes, the size of the difference suggests that menu costs should not be ignored.
There are no guidelines or rules that dictate how large and small exchange rate
changes should be defined. Thus, I apply alternative threshold values to test the
robustness of the results in Panel B. First, I define the value of the threshold for each
country as being equal to one and a half standard deviation of the quarterly exchange rate
change and second, as being equal to two standard deviations of the quarterly exchange
rate change. Overall, the basic results hold when the threshold value is increased, see
Table A11 in the Appendix. Estimated ERPT coefficients are fairly similar to the
estimates reported in Panel B and as before, linearity cannot be rejected.
Finally, looking at the results in Panel C, the estimated ERPT coefficients are
statistically significant at conventional levels for large appreciations, small appreciations,
large depreciations and small depreciations in Germany, Iceland, New Zealand and the
U.K. For those countries, four different restrictions are tested, 𝛽𝐿𝐴 = 𝛽𝐿𝐷 , 𝛽𝑆𝐴 = 𝛽𝑆𝐷 ,
𝛽𝐿𝐴 = 𝛽𝑆𝐴 and 𝛽𝐿𝐷 = 𝛽𝑆𝐷 . Iceland is the only country where one or more of those
restrictions can be rejected. To elaborate, 𝛽𝐿𝐴 = 𝛽𝐿𝐷 and 𝛽𝐿𝐴 = 𝛽𝑆𝐴 are rejected at five
percent significance level. However, the estimated ERTP coefficient for large
appreciations is unrealistically high, questioning the validity of this result. For New
Zealand and the U.K. there is a considerable variation among the estimated pass-through
coefficients, albeit not significant. For both countries small appreciations appear to be

23
passed through to import prices to a greater extent than other exchange rate changes,
which is interesting seeing as a higher degree of ERPT is associated with depreciations,
see Panel A.
For Sweden, the estimated ERPT coefficient for small appreciations is not statistically
significant. The restriction of 𝛽𝐿𝐴 = 𝛽𝐿𝐷 is rejected at five percent significance level
while the restriction 𝛽𝐿𝐷 = 𝛽𝑆𝐷 is not. Thus, one might tentatively conclude that in
Sweden the direction of the exchange rate change matters more for ERPT than the size of
the change. Finally, for the U.S., only estimated ERPT coefficients for appreciations of
the US dollar are statistically significant, both for large and small changes. However the
restriction of 𝛽𝐿𝐴 = 𝛽𝑆𝐴 cannot be rejected. Therefore, as for Sweden, one might
cautiously interpret the results as such that the direction of the exchange rate is the key
determinant for ERPT.

5 Conclusion
This paper investigates the possibility of asymmetric and non-linear responses of import
prices to exchange rate changes, using quarterly data from 1980 to 2014. The study is
conducted with aggregate import prices for six countries, Germany, Iceland, New
Zealand, Sweden, the United Kingdom and the United States and by estimating a simple
regression equation separately for each country, in log differences, which is later
augmented with interactive dummy variables to account for possible asymmetries and
non-linearities.
The results indicate that asymmetry and non-linearity cannot be neglected when
estimating ERPT. For the U.S. and Sweden, evidence of asymmetric behavior is found
but the direction of asymmetry varies between the two countries. Pass-through to U.S.
import prices is only significant during appreciations of the dollar, consistent with the
market share theory, while pass-through to Swedish import prices is significantly higher
during depreciations of the Swedish krona, indicating downward price rigidities in
Sweden’s import prices. For Iceland, New Zealand and the U.K. symmetric ERPT cannot
be rejected. However, the size of the difference between the estimated ERPT coefficients
for appreciations and depreciations is considerable, suggesting caution when interpreting
the estimates from the linear and symmetric model.

24
 For no country can the restriction of linear pass-through be rejected. Furthermore, in
most cases the difference between estimated pass-through for large and small exchange
rate changes is miniscule. However, in some cases, for instance New Zealand, the
difference in pass-through is quite substantial, indicating that non-linearity cannot be
ignored. In the case of New Zealand, ERPT to import prices is greater when the
exchange rate changes are small, consistent with the menu cost assumption when imports
are priced in the exporter´s currency. When combining the direction and the size of the
exchange rate change, the results indicate that the direction effects overshadow the size
effects.
While the aim of this paper is to investigate the possibility of asymmetric and non-
linear responses of import prices to exchange rate changes, future research might
examine the characteristics that explain cross-country differences. Another interesting
area is to test for asymmetries and non-linearities using disaggregate, industry level
import prices. Furthermore, focusing on asymmetries and non-linearities in the second-
stage ERPT, or the pass-through of exchange rate changes to consumer prices, is of key
importance for policy makers with the objective of price stability.
There remains a lot of work to do when it comes to the issue of asymmetries and non-
linearities in ERPT. With the literature still relatively scarce, and mostly restricted to a
small group of countries, it seems like the surface has barely been scratched.

25
References
Aron, J., R. Macdonald and J. Muellbauer. (2014). Exchange Rate Pass-through in
Developing and Emerging Markets: A Survey of Conceptual, Methodological and
Policy Issues, and Selected Empirical Findings, The Journal of Development
Studies 50(1): 101-143.

Athukorala, P. and J. Menon. (1994). Pricing to Market Behavior and Exchange Rate
Pass-Through in Japanese Exports, The Economic Journal 104: 271-281.

Barhoumi, K. (2006). Differences in long run Exchange Rate Pass-through into Import
Prices in Developing Countries: An Empirical Investigation, Economic Modelling
26(6): 926-951.

Bailliu, J. and E. Fujii. (2004). Exchange Rate Pass-through and the Inflation
Environment in Industrialized Countries: an Empirical Investigation. Bank of
Canada Working Paper 2004-21.

Bussière, M. (2007). Exchange Rate Pass-Through to Trade Prices: The Role of

Nonlinearities and Asymmetries. ECB Working Paper No. 822.

Bussière, M. (2013). Exchange Rate Pass-through to Trade Prices: The Role of

Nonlinearities and Asymmetries, Oxford Bulletin of Economics and Statistics
75(5): 731-738.

Bussière, M., S. D. Chiaie and T. A. Peltonen. (2014). Exchange Rate Pass-Through in

the Global Economy: The Role of Emerging Market Economies, IMF Economic
Review 62: 146-178.

Campa, J. M. and L. S. Goldberg. (2002). Exchange Rate Pass-Through into Import

Prices: A Macro or Micro Phenomenon? NBER Working Paper No. 8934.

Campa, J. M. and L. S. Goldberg. (2005). Exchange Rate Pass-Through into Import

Prices, Review of Economics and Statistics 87(4): 679-690.

Campa, J. M., L. S. Goldberg and J. M. González-Mínguez. (2005). Exchange Rate Pass-

Through to Import Prices in the Euro Area. Banco de España Working Paper No.
0538.

Campa, J. M., J. M. G. Mínguez and M. S. Barriel. (2008). Non-linear Adjustment of

Import Prices in the European Union. Bank of England Working Paper No. 347.

Central Bank of Iceland. (2011). Exchange Rate Pass-through, Monetary Bulletin 44(2):
44-46.

26
Choudri, E. U. and D. S. Hakura. (2012). The Exchange Rate Pass-Through to Import
and Export Prices: The Role of Nominal Rigidities and Currency Choice. IMF
Working Paper No. 12/226.

Correa, A. S. and A. Minella. (2006). Nonlinear Mechanism of the Exchange Rate Pass-
Through: A Phillips Curve Model with Threshold for Brazil. Banco Central do
Brasil Working Paper No. 122.

Delatte, A. L. and A. López-Villavicencio. (2012). Asymmetric Exchange Rate Pass-

through: Evidence from Major Countries, Journal of Macroeconomics 34(3): 833-
844.

Dornbusch, R. (1987). Exchange Rates and Prices, American Economic Review 77(1):
93-106.

El bejaoui, H. J. (2013). Asymmetric Effects of Exchange Rate Variations: An Emprical

Analysis for four Advanced Countries, International Economics 135-136: 29-46.

Engle, R. F. and C. W. J. Granger. (1987). Co-Integration and Error Correction.

Representation, Estimation, and Testing, Econometrica 55(3): 251-276.

Feinberg, R. M. (1989). The Effects of Foreign Exchange Movements on U.S. Domestic

Prices, Review of Economics and Statistics 71(3): 505-511.

Froot, K. A. and P. Klemperer. (1989). Exchange Rate Pass-Through when Market Share
Matters, American Economic Review 79(4): 637-654.

Gagnon, J. E. and J. Ihrig. (2004). Monetary Policy and Exchange Rate Pass-through.
Board of Governors of the Federal Reserve System, International Finance
Discussion Paper No. 704.

Gil-Pareja, S. (2000). Exchange Rates and European Countries´ Export Prices: An

Empirical Test for Asymmetries in Pricing to Market Behavior,
Weltwirtschaftliches Archiv 136(1): 1-23.

Goldberg, P. K. (1995). Product Differentiation and Oligopoly in International Markets:

The Case of the U.S. Automobile Industry, Econometrica 63(4): 891-951.

Goldberg, P. K., and M. M. Knetter. (1997). Goods Prices and Exchange Rates: What
Have We Learned?, Journal of Economic Literature 35(3): 1243-1272.

Gust, C., S. Leduc and R. Vigfusson. (2010). Trade Integration, Competition and the
Decline in Exchange Rate Pass-through, Journal of Monetary Economics 57(3):
309-324.

27
Herzberg, V., G. Kapetanios and S. Price. (2003). Import Prices and Exchange Rate Pass-
through: Theory and Evidence from the United Kingdom. Bank of England
Working Paper No. 182.

Ihrig, J. E., M. Marazzi and A. Rothenberg. (2006). Exchange Rate Pass-through in the
G7 Economies. Board of Governors of the Federal Reserve System, International
Finance Discussion Paper No. 851.

Kadiyali, V. (1997). Exchange Rate Pass-through for Strategic Pricing and Advertising:
An Empirical Analysis of the U.S. Photographic Film Industry, Journal of
International Economics 43(3/4): 437-461.

Kanas, A. (1997). Is Economic Exposure Asymmetric Between Long-Run Depreciations

and Appreciations? Testing Using Cointegration Analysis, Journal of Multinational
Financial Management 7(1): 27-42.

Knetter, M. M. (1994). Is Export Price Adjustment Asymmetric? Evaluating the Market

Share and Marketing Bottlenecks Hypotheses, Journal of International Money and
Finance 13(1): 55-70.

Krugman, P. (1986). Pricing to Market when the Exchange Rate Changes. NBER
Working Paper No. 1926.

MacKinnon, J. G. (2010). Critical Values for Cointegration Test. Queen´s Economics

Department Working Paper No. 1227.

Mann, C. L. (1986). Prices, Profit Margins and Exchange Rates, Federal Reserve
Bulletin 72: 366-279.

Marazzi, M., N. Sheets, and R. Vigfusson, and J. Faust, J. Gagnon, J. Marquez, R.

Martin, T. Reeve and J. Rogers. (2005). Exchange Rate Pass-through to U.S.
Import prices: Some New Evidence. Board of Governors of the Federal Reserve
System, International Finance Discussion Paper No. 832.

Marston, R. C. (1990). Pricing to Market in Japanese Manufacturing, Journal of

International Economics 29(3-4): 217-236.

Ohno, K. (1989). Export Pricing Behavior of Manufacturing: A U.S. – Japan

Comparison, IMF Staff Papers, 36(3): 550-579.

Olivei, G. P. (2002). Exchange Rates and the Prices of Manufacturing Products Imported
into the United States, New England Economic Review First Quarter: 3-18.

28
Ólafsson, T. T., A. Pétursdóttir and K. Á. Vignisdóttir. (2011). Price Setting in Turbulent
Times: Survey Evidence from Icelandic Firms. Central Bank of Iceland Working
Paper No. 54.

Peltzman S. (2000). Prices Rise Faster than They Fall, Journal of Political Economy
108(3): 466-502.

Pétursson, Þ. G. (2008). How hard can it be? Inflation Control around the World. Central
Bank of Iceland Working Paper No. 40.

Pollard, P. S. and C. C. Coughlin. (2004). Size Matters: Asymmetric Exchange Rate

Pass-through at the Industry Level. Federal Reserve Bank of St. Louis Working
Paper 2003-029C.

Przystupa, J. and E. Wróbel. (2011). Asymmetry of the Exchange Rate Pass-through: An

exercise on the Polish data, Eastern European Economics 49(1): 30-51.

Taylor, J. B. (2000). Low Inflation, Pass-through and the Pricing Power of Firms,
European Economic Review 44(7): 1389-1408.

Webber, A. (2000). Newton´s Gravity Law and Import Prices in the Asia Pacific, Japan
and the World Economy 12(1): 71-87.

Wickremasinghe, G. B. and P. Silvapulle. (2004). Exchange Rate Pass-Through to

Manufactured Import Prices: The Case of Japan. International Trade 2004:
0406006.

Yang, J. (2007). Is Exchange Rate Pass-through Symmetric? Evidence from U.S.

Imports, Applied Economics 39(2): 169-178.

29
Appendix
Table A1: Stationarity test for key series
DE IS NZ SE UK US
A. Variables in first differences (ADF test statistic):
𝑝𝑀𝑃 -6.577*** -8.212*** -9.433*** -8.388*** -7.811*** -8.657***
𝑒𝑟 (𝑁𝐸𝐸𝑅) -8.988*** -4.564*** -9.524*** -9.049*** -9.389*** -7.871***
∗
𝑤 -6.126*** -9.852*** -7.082*** -8.080*** -6.541*** -7.029***
𝑃𝑃𝐼 -5.683*** NA -6.532*** -8.091*** -6.239*** -8.408***
𝑂𝑖𝑙 𝑝𝑟𝑖𝑐𝑒𝑠 -10.47*** -9.969*** -10.637*** -10.330*** -12.372*** -10.493***
B. Cointegration test (ADF test statistic):
-2.548 -2.474 -4.199** -1.924 -3.406 -2.102
Notes: ***, ** and * denotes rejection of the null hypothesis of a unit root at the 1% level, 5% level and
10% level respectively. Different critical values apply for section A and B of this table. Section B relies
upon critical values provided by MacKinnon (2010).

Table A2: Full model, linear specification equation (6)

DE IS NZ SE UK US
∆𝑒𝑟 0.498*** 0.962*** 0.586*** 0.208*** 0.351*** 0.164***
(0.060) (0.021) (0.044) (0.024) (0.054) (0.032)
∗
∆𝑤 0.927*** 0.269 0.645*** 0.141** 0.448*** 0.228
(0.163) (0.198) (0.142) (0.056) (0.170) (0.260)
∆𝑃𝑃𝐼 0.536*** 0.034 0.735*** 1.089*** 1.392*** 1.064***
(0.179) (0.044) (0.086) (0.082) (0.415) (0.112)
∆𝑂𝑖𝑙 𝑝𝑟𝑖𝑐𝑒𝑠 0.016*** 0.002 -0.013 0.018*** 0.016*** 0.013***
(0.005) (0.009) (0.009) (0.003) (0.005) (0.005)
R-squared 0.831 0.881 0.731 0.915 0.681 0.837
Durbin-Watson stat 1.586 2.051 2.349 1.581 1.558 1.885
Notes: Standard errors in parenthesis. ***, ** and * indicates significance at 1%, 5% and 10% level
respectively. All regressions include quarterly seasonal dummies. Because of data limitations I use GDP
instead of PPI in the regression for Iceland.

Table A3: Full model, linear specification equation (7)

DE IS NZ SE UK US
∆𝑒𝑟 0.556*** 0.940*** 0.719*** 0.275*** 0.482*** 0.195***
(0.068) (0.038) (0.067) (0.041) (0.042) (0.042)
∗
∆𝑤 0.478*** 0.462*** 0.792*** 0.143*** 0.628*** 0.365
(0.136) (0.178) (0.217) (0.065) (0.177) (0.258)
∆𝑃𝑃𝐼 0.662*** -0.090 0.446*** 0.863*** 1.074*** 0.769***
(0.137) (0.045) (0.140) (0.072) (0.301) (0.121)
∆𝑂𝑖𝑙 𝑝𝑟𝑖𝑐𝑒𝑠 0.044*** 0.047*** 0.022*** 0.042*** 0.018 0.048***
(0.005) (0.020) (0.011) (0.017) (0.013) (0.008)
R-squared 0.886 0.911 0.773 0.946 0.738 0.862
Durbin-Watson stat 1.742 2.081 2.486 1.813 1.571 1.863
Notes: Standard errors in parenthesis. ***, ** and * indicates significance at 1%, 5% and 10% leve
respectively. All regressions include quarterly seasonal dummies. Because of data limitations I use GDP
instead of PPI in the regression for Iceland.

30
Table A4: Robustness test, instrumental variable estimator
DE IS NZ SE UK US
∆𝑒𝑟 0.528*** 0.976*** 0.590*** 0.276*** 0.368*** 0.197***
(0.056) (0.025) (0.059) (0.043) (0.049) (0.068)
∆𝑤 ∗ 1.166*** 0.281 0.661*** 0.301*** 0.702*** 0.507
(0.178) (0.194) (0.242) (0.091) (0.261) (0.589)
∆𝑃𝑃𝐼 0.149 0.153** 0.714*** 0.797*** 0.947*** 0.821*
(0.214) (0.065) (0.216) (0.164) (0.254) (0.450)
∆𝑂𝑖𝑙 𝑝𝑟𝑖𝑐𝑒𝑠 0.014*** -0.002 -0.013 0.022*** 0.015*** 0.017*
(0.005) (0.010) (0.009) (0.004) (0.005) (0.010)
R-squared 0.818 0.874 0.731 0.895 0.665 0.826
Durbin-Watson stat 1.568 2.097 2.357 1.965 1.526 1.943
Notes: Standard errors in parenthesis. ***, ** and * indicates significance at 1%, 5% and 10% level
respectively. All regressions include quarterly seasonal dummies. Because of data limitations I use GDP
instead of PPI in the regression for Iceland.

Table A5: Robustness test, additional explanatory variables

DE IS NZ SE UK US
∆𝑒𝑟 0.499*** NA 0.585*** 0.208*** 0.350*** 0.163***
(0.062) (0.046) (0.024) (0.053) (0.032)
∗
∆𝑤 0.932*** NA 0.643*** 0.141** 0.436*** 0.224
(0.171) (0.142) (0.056) (0.155) (0.261)
∆𝑅𝑒𝑎𝑙 𝐺𝐷𝑃 -0.028 NA -0.022 0.001 0.055 -0.027
(0.052) (0.104) (0.023) (0.171) (0.110)
∆𝑃𝑃𝐼 0.537*** NA 0.735*** 1.089*** 1.412*** 1.067***
(0.180) (0.086) (0.084) (0.400) (0.117)
∆𝑂𝑖𝑙 𝑝𝑟𝑖𝑐𝑒𝑠 0.016*** NA -0.013 0.018*** 0.016*** 0.013***
(0.006) (0.010) (0.003) (0.005) (0.005)
R-squared 0.831 NA 0.731 0.915 0.682 0.838
Durbin-Watson stat 1.563 NA 2.345 1.581 1.546 1.885
Notes: Standard errors in parenthesis. ***, ** and * indicates significance at 1%, 5% and 10% level
respectively. All regressions include quarterly seasonal dummies.
Iceland is not included in this robustness test as real GDP is already used in the estimated equation
instead of PPI.

Table A6: Robustness test, alternative proxy for competitor´s prices

DE IS NZ SE UK US
∆𝑒𝑟 0.540*** 0.887*** 0.686*** 0.441*** 0.421*** 0.340***
(0.058) (0.044) (0.047) (0.038) (0.046) (0.049)
∗
∆𝑤 1.266*** 0.226 1.136*** 0.639*** 0.862*** 0.777***
(0.127) (0.192) (0.137) (0.159) (0.214) (0.159)
∆𝐶𝑃𝐼 0.150 0.157** 0.299*** 0.343*** 0.738 1.570***
(0.113) (0.063) (0.104) (0.106) (0.453) (0.396)
∆𝑂𝑖𝑙 𝑝𝑟𝑖𝑐𝑒𝑠 0.013*** 0.002 -0.015 0.034*** 0.019*** 0.025***
(0.005) (0.009) (0.009) (0.007) (0.007) (0.007)
R-squared 0.808 0.886 0.685 0.655 0.567 0.722
Durbin-Watson stat 1.517 2.051 2.361 1.981 1.440 1.922
Notes: Standard errors in parenthesis. ***, ** and * indicates significance at 1%, 5% and 10% level
respectively. All regressions include quarterly seasonal dummies.

31
Table A7: Robustness test, alternative proxy for foreign marginal cost of production
DE IS NZ SE UK US
∆𝑒𝑟 0.477*** 0.931*** 0.524*** 0.170*** 0.372*** 0.192***
(0.095) (0.026) (0.048) (0.030) (0.029) (0.034)
∆𝐶𝑃𝐼 ∗ 0.220 1.007*** -0.185 -0.244* -0.581*** 0.150**
(0.152) (0.324) (0.228) (0.134) (0.033) (0.064)
∆𝑃𝑃𝐼 1.230*** 0.059 1.084*** 1.253*** 1.557*** 1.080***
(0.140) (0.046) (0.122) (0.114) (0.091) (0.058)
∆𝑂𝑖𝑙 𝑝𝑟𝑖𝑐𝑒𝑠 0.029*** 0.002 -0.008 0.017*** 0.021*** 0.012***
(0.009) (0.009) (0.009) (0.003) (0.005) (0.004)
R-squared 0.732 0.884 0.705 0.914 0.826 0.847
Durbin-Watson stat 1.658 2.043 2.179 1.567 1.796 2.112
Notes: Standard errors in parenthesis. ***, ** and * indicates significance at 1%, 5% and 10% level
respectively. All regressions include quarterly seasonal dummies. Because of data limitations I use GDP
instead of PPI in the regression for Iceland.

Table A8: Full model, allowing for asymmetry, equation (8)

DE IS NZ SE UK US
𝐴∆𝑒𝑟 0.503 *** 1.173*** 0.517*** 0.099* 0.267** 0.169***
(0.093) (0.155) (0.093) (0.057) (0.118) (0.046)
𝐷∆𝑒𝑟 0.492 *** 0.925*** 0.628*** 0.245*** 0.407*** 0.152
(0.087) (0.037) (0.061) (0.022) (0.046) (0.094)
∆𝑤 ∗ 0.927*** 0.255 0.659*** 0.139*** 0.504*** 0.227
(0.164) (0.192) (0.144) (0.053) (0.148) (0.261)
∆𝑃𝑃𝐼 0.537*** 0.034 0.739*** 1.103*** 1.334*** 1.065***
(0.179) (0.045) (0.088) (0.080) (0.375) (0.113)
∆𝑂𝑖𝑙 𝑝𝑟𝑖𝑐𝑒𝑠 0.016*** 0.002 -0.013 0.017*** 0.015** 0.012***
(0.005) (0.010) (0.009) (0.003) (0.006) (0.005)
R squared 0.831 0.884 0.732 0.919 0.684 0.837
Durbin Watson stat 1.589 2.075 2.307 1.622 1.544 1.883
Notes: Standard errors in parenthesis. ***, ** and * indicates significance at 1%, 5% and 10%
respectively. All regressions include quarterly seasonal dummies. A denotes appreciations of the home
currency while D denotes depreciations. Because of data limitations I use GDP instead of PPI in the
regression for Iceland.

Table A9: Full model, allowing for non-linearity, equation (9)

DE IS NZ SE UK US
𝐿∆𝑒𝑟 0.501*** 0.961*** 0.563*** 0.219*** 0.345*** 0.159***
(0.066) (0.024) (0.051) (0.023) (0.073) (0.035)
𝑆∆𝑒𝑟 0.486*** 0.970*** 0.722*** 0.153*** 0.374*** 0.181***
(0.093) (0.072) (0.120) (0.048) (0.069) (0.056)
∆𝑤 ∗ 0.926*** 0.269 0.631*** 0.146*** 0.442** 0.227
(0.172) (0.200) (0.136) (0.055) (0.181) (0.263)
∆𝑃𝑃𝐼 0.539*** 0.034 0.722*** 1.088*** 1.394*** 1.065***
(0.189) (0.044) (0.086) (0.081) (0.421) (0.112)
∆𝑂𝑖𝑙 𝑝𝑟𝑖𝑐𝑒𝑠 0.016*** 0.002 -0.013 0.017*** 0.016*** 0.013***
(0.005) (0.010) (0.010) (0.003) (0.005) (0.005)
R squared 0.831 0.881 0.735 0.916 0.682 0.838
Durbin Watson stat 1.587 2.049 2.333 1.575 1.550 1.882

32
Notes: Standard errors in parenthesis. ***, ** and * indicates significance at 1%, 5% and 10% level
respectively. All regressions include quarterly seasonal dummies. L denotes large exchange rate changes
while S denotes small exchange rate changes. Because of data limitations I use GDP instead of PPI in the
regression for Iceland.

Table A10: Allowing for asymmetry, equation (10)

DE IS NZ SE UK US
𝐿𝐴∆𝑒𝑟 0.494*** 1.441*** 0.487*** 0.115** 0.260* 0.176***
(0.098) (0.232) (0.092) (0.057) (0.147) (0.052)
𝑆𝐴∆𝑒𝑟 0.412* 0.976*** 0.761*** -0.038 0.429*** 0.246**
(0.216) (0.118) (0.260) (0.089) (0.149) (0.121)
𝐿𝐷∆𝑒𝑟 0.503*** 0.934*** 0.603*** 0.247*** 0.391*** 0.129
(0.101) (0.035) (0.074) (0.022) (0.058) (0.111)
𝑆𝐷∆𝑒𝑟 0.546*** 0.962*** 0.668*** 0.330*** 0.303** 0.120
(0.172) (0.125) (0.212) (0.070) (0.141) (0.157)
∆𝑤 ∗ 0.926*** 0.233 0.656*** 0.130** 0.513*** 0.233
(0.173) (0.177) (0.143) (0.052) (0.151) (0.270)
∆𝑃𝑃𝐼 0.537*** 0.029 0.717*** 1.116*** 1.327*** 1.063***
(0.192) (0.043) (0.088) (0.080) (0.371) (0.113)
∆𝑂𝑖𝑙 𝑝𝑟𝑖𝑐𝑒𝑠 0.016*** 0.003 -0.013 0.017*** 0.015** 0.012**
(0.005) (0.009) (0.010) (0.003) (0.006) (0.005)
R squared 0.831 0.888 0.737 0.921 0.687 0.838
Durbin Watson stat 1.586 2.121 2.271 1.682 1.515 1.877
Notes: Standard errors in parenthesis. ***, ** and * indicates significance at 1%, 5% and 10% level
respectively. All regressions include quarterly seasonal dummies. LA denotes large appreciations, SA
small appreciations, LD large depreciations and SD small depreciations. Because of data limitations I use
GDP instead of PPI in the regression for Iceland.

Table A11: Different thresholds

DE IS NZ SE UK US
Panel A: Threshold equal to one and a half standard deviation
Large changes 0.560 *** 0.936*** 0.574*** 0.224*** 0.329*** 0.186***
(0.060) (0.034) (0.056) (0.023) (0.086) (0.040)
Small changes 0.435 *** 1.038*** 0.610*** 0.174*** 0.384*** 0.139***
(0.088) (0.065) (0.083) (0.033) (0.037) (0.051)
Wald test statistic 1.312 -1.170 -0.341 1.670* -0.597 0.698
Panel B: Threshold equal to two standard deviations
Large changes 0.588*** 0.964*** 0.571*** 0.221*** 0.421*** 0.235***
(0.093) (0.030) (0.071) (0.028) (0.053) (0.056)
Small changes 0.467*** 0.958*** 0.599*** 0.189*** 0.290*** 0.123***
(0.064) (0.039) (0.062) (0.028) (0.083) (0.043)
Wald test statistic 1.271 0.107 -0.287 1.047 1.330 1.469
Notes: Standard errors in parenthesis. ***, ** and * indicates significance at 1%, 5% and 10% level
respectively. All regressions include quarterly seasonal dummies.

33