Economic Systems 37 (2013) 302–329

Contents lists available at SciVerse ScienceDirect

Economic Systems

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

The exchange rate pass-through in the new EU memberstates

Ramona Jimborean

European Relations Division, 49-1487 SRE, Banque de France, 75049 Paris Cedex 1, France

A R T I C L E I N F O

Article history:

Received 16 August 2011

Received in revised form 10 August 2012

Accepted 14 August 2012

JEL classification:

C33

E31

E42

E52

F31

O52

Keywords:

Inflation and prices

Exchange rate pass-through

GMM

International topics

A B S T R A C T

This paper aims to complete our understanding of the relationship

between changes in nominal effective exchange rates and prices in

the new EU member states. I investigate the exchange rate pass-

through to import, producer and consumer prices for ten Central and

Eastern European countries with quarterly data from January 1996

to December 2011. In a first step, the pass-through estimates are

derived from a dynamic panel data model through the generalized

method of moments. A statistically significant exchange rate pass-

through to consumer, producer and import prices is found, both in

the short and long run. In a second step, I proceed to an individual

analysis, country by country, and find support for an increased

heterogeneity in the exchange rate pass-through estimates. In a

third step, I assess the drivers of the estimated exchange rate pass-

through coefficients and find support for a significant impact of

exchange rate volatility, inflation volatility, import dependence, and

the output gap, as well as the global outlook.

� 2013 Elsevier B.V. All rights reserved.

1. Introduction

In this paper, I study the exchange rate pass-through to consumer, producer and import prices forten new EU member states (NMS1).

One of the major channels through which the exchange rate traditionally affects economicperformance is through its impact on prices. The main direct effect occurs through the impact onimport prices, which further reverberates along the pricing chain to consumer prices.2 These price

E-mail addresses: rjimborean@yahoo.com, ramona.jimborean@banque-france.fr.1 Bulgaria, Czech Republic, Estonia, Hungary, Lithuania, Latvia, Poland, Romania, Slovenia and Slovakia.2 Domestic currency depreciations can increase import prices, which, in turn, may translate into domestic price inflation.

0939-3625/$ – see front matter � 2013 Elsevier B.V. All rights reserved.

http://dx.doi.org/10.1016/j.ecosys.2012.08.006

R. Jimborean / Economic Systems 37 (2013) 302–329 303

changes give rise to important indirect and second-round effects through their impact on real income,consumer spending and trade flows, with feedback effects on overall price pressures.

From a policy point of view, knowledge of the link between nominal exchange rates and inflation inthe new EU member states is important for several reasons.

First, knowledge of the exchange rate pass-through dynamics has important monetary policyimplications. For instance, in the case of inflation targeting countries, the extent and the timing of thepass-through is important both for forecasting inflation and for the monetary policy decision-makingprocess. In these countries, regular adjustments of the interest rate are needed in order to counteractdeviations from the inflation target caused by changes in the nominal exchange rate.

Second, policy discussions about exchange rate options in countries with fixed exchange rateregimes can necessitate some knowledge about the exchange rate pass-through. After the outbreak ofthe 2007 global economic and financial crisis, and once it started to affect the Central and EasternEuropean region in late 2008, there was a big debate on the pros and cons of devaluing the currency insome countries with fixed exchange rates. Quantitative knowledge of the exchange rate pass-throughis extremely important in such policy discussions since the beneficial effects of devaluation oncompetitiveness are likely to be limited if the pass-through is assessed to be high.

Third, nominal exchange rate fluctuations can have a short-term character but can equally bestructural when associated with the real appreciation trend experienced by the catching-upeconomies. The pressure of real exchange rate appreciation and the real convergence process areclosely related – this phenomenon is usually called the Balassa-Samuelson effect (Egert, 2011).Besides, during the process of real convergence, the NMS face inflationary pressures caused bothby demand factors (i.e. reduction in interest rates, credit growth, property investments, etc.) andsupply factors (i.e. inter-sectoral differences in productivity that might determine a higherinflation in the non-tradable sectors). In the catching-up countries with a nominal trendappreciation of the currency (such as Slovakia before it adopted the euro at the beginning of 2009),knowledge of the link between nominal exchange rates and inflation may shed light on the degreeto which inflation convergence vis-a-vis the euro area is sustainable, an issue that is crucial for theassessment of the convergence criteria.3 A high pass-through associated to a strong appreciationbefore joining the EU might raise some doubts on the sustainable nature of inflation. This issue waslargely debated in the 2008 Convergence Report on Slovakia, when empirical evidence of thelimited size of the Slovak pass-through played a decisive role in supporting the inflation criteria’ssustainability.4

At the same time, developments in the nominal exchange rate play an important role. If theexchange regime is fixed, as is the case for instance in a monetary union, the real appreciation passesexclusively through an inflation differential. On the contrary, in a flexible exchange regime withinflation targeting we might observe a tendency of appreciation in the nominal exchange rate, as theinflation target, if credible, makes the existence of a sensible inflation differential less probable.

My study is focused on a sample formed by the new EU member states.

3 The convergence criteria (also known as the Maastricht criteria) are the criteria for European Union member states to enter

the third stage of the European Economic and Monetary Union (EMU) and adopt the euro as their currency. These criteria are

based on Article 121(1) of the European Community Treaty and are related to: inflation rate (no more than 1.5 percentage points

higher than the average of the three best performing (lowest inflation) member states of the EU); government finance (the ratio

of the annual government deficit to gross domestic product (GDP) must not exceed 3 percent at the end of the preceding fiscal

year, the ratio of gross government debt to GDP must not exceed 60 percent at the end of the preceding fiscal year); exchange

rate (applicant countries should have joined the exchange-rate mechanism (ERM II) under the European Monetary System

(EMS) for two consecutive years and should not have devalued its currency during this period); and long-term interest rates (the

nominal long-term interest rate must not be more than 2 percentage points higher than in the three lowest inflation member

states).4 The Slovak authorities re-evaluated the central parity rate twice: in March 2007 (+8.5 percent) and in May 2008 (+17.6

percent). The size of the appreciation of the Slovak koruna was unique for an ERM II member state and caused doubt as to the

sustainability of inflation, as indicated in the 2008 ECB Convergence Report: ‘‘In recent years inflation has been dampened, in

particular, by the trend appreciation of the exchange rate of the Slovak koruna. Available assessments suggest that the

appreciation of the koruna has reduced inflation over the past year’’. The quantitative effect is nevertheless uncertain and

depends on the pass-through estimates; this makes a precise and reliable estimation of the exchange rate pass-through for the

countries entering the ERM II with a floating exchange rate (Hungary, Poland, Czech Republic and Romania) important.

R. Jimborean / Economic Systems 37 (2013) 302–329304

� T

re

he large majority of countries in the region commuted from a high inflation environment (at thebeginning of the 90s) to a relatively moderate inflation rate (by the late 90s). An analysis of the lastdecade shows a diminishing tendency in the inflation rate for the large majority of CEECs comparedto the end of the 90s. Nevertheless, inflation went up again in almost all of these countries between2003 and 2005, in particularly in Bulgaria, Estonia, Latvia and Lithuania. In Romania and Slovakiainflation continued to register a global descending trend compared to 2003, while the exchange ratewas on an appreciation tendency, even though lately this trend was reversed due to the 2007 globalfinancial crisis.

� T he NMS have maintained different exchange systems over time. These regimes included currency

boards, fixed pegs to a basket, crawling pegs, managed float and free float. Three countries arecurrently members of the ERM II5: Estonia and Lithuania with a currency board as a unilateralcommitment and Latvia with a �1 percent fluctuation band as a unilateral commitment. Bulgaria has acurrency board arrangement with the euro as the anchor currency. The Czech Republic, Hungary, Polandand Romania are under inflation targeting regimes with free floating exchange rates, while Hungary’sinflation targeting is conducted in conjunction with an exchange rate band of �15 percent against theeuro. Before joining the euro area, Slovakia practiced inflation targeting with a standard fluctuation bandof �15 percent and Slovenia practiced a managed float.

A rigorous quantitative analysis of the exchange rate pass-through in these countries might answerboth political economy questions (sustainable convergence) and theoretical questions (the impact ofthe exchange rate regime on the magnitude of pass-through). I thus estimate the average exchangerate pass-through, both in the short and the long run, with the use of the generalized method ofmoments estimator for dynamic panel data models, while considering the exchange rate pass-throughto consumer, producer and import prices. In a second step, I proceed to an individual country-specificanalysis that is prolonged in a third step by the assessment of the macroeconomic drivers of theexchange rate pass-through.

I find evidence of a positive exchange rate pass-through to consumer, producer and import pricesthat seems to increase slightly in the long run. In other words, on average, the nominal exchange ratedepreciation is followed by an increase in consumer, producer and import prices.

The paper is organized as follows. Section 2 presents the theoretical framework, while enumeratingthe factors that influence the exchange rate pass-through; in the same section, I present themethodology that can be employed and what has been done in previous studies. In Section 3 I presentthe data and focus on the econometric specification. Section 4 concludes.

2. Theoretical framework

Both nominal and real exchange rates play an important role in the monetary policy transmission.There are two stages of the transmission mechanism: the first stage shows the way monetary policyimpacts exchange rates, while the second stage consists in the pass-through from exchange rates toimport and domestic prices, followed by adjustments in real variables (imports, exports, investment).

Nominal exchange rate movements caused by monetary policy actions can be translated intodomestic inflation through changes in the prices of imported final goods, as well as through theupward or downward pressure of imported intermediate goods prices on domestic inflation via theprice of domestically manufactured tradable and non-tradable goods.

The pricing behavior of importing firms is an important contribution to the way the exchange rateaffects domestic prices through imported prices. If the prices of imported goods are set in theimporter’s currency (producer currency pricing), any change in the exchange rate will beautomatically transmitted to the prices of the destination country and thus the pass-through iscomplete. On the other hand, if the prices of imported goods are set in local currency (local orconsumer currency pricing), exchange rate movements are not reflected in domestic prices and thepass-through is zero.

5 Estonia joined the euro area on the 1st of January 2011 and is no longer an ERM II member. Nevertheless, since the analysis is

alized over the period 1996–2010, we can still talk about the existence of three members of the ERM II.

R. Jimborean / Economic Systems 37 (2013) 302–329 305

Two general opinions emerge in the empirical literature related to the exchange rate pass-through.First, the exchange rate pass-through is generally higher in developing and emerging countriescompared to developed economies (Campa and Goldberg, 2002); it declines over time with thecatching-up in terms of living standards, both in the industrialized and the developing countries(Coricelli et al., 2006; Bitans, 2004). These evolutions are related mainly to the role of macroeconomicvariables, especially inflation (Taylor, 2000; Devereux and Yetman, 2003; Choudhri and Hakura,2006), as well as to the shift in imports from goods with higher pass-through elasticities to goods withlower ones (Frankel et al., 2005). Second, the pass-through is highest for imported goods, lower forproducer goods and lowest for consumer prices. Among the potential explanations, we mention therole of distribution costs for final imported goods6 (Burnstein et al., 2002) and the role of intermediateimported goods (Engel, 2002).

2.1. Factors influencing the size of the exchange rate pass-through

The pass-through of the exchange rate to prices is only partly due to monetary policy actions.Several other factors, both macroeconomic and microeconomic, might cause changes in exchangerates. Among these factors we mention the inflation rate, the exchange rate regime, the output gap,openness and the composition of trade and imports, as well as expectations.

A first strand of literature underlines the importance of the rate of inflation (Taylor, 2000). Thehigher the inflation rate, the larger the exchange rate pass-through, since prices are adjusted morefrequently in an inflationary environment. Moreover, as noted by Corsetti et al. (2007), a more stableinflation environment reduces the incentive of producers to price discriminate across countries(implying lower pass-through). The need to account for the observed inertial behavior of inflation(inflation persistence) has been emphasized by the literature on inflation dynamics (Gali and Gertler,1999). The more persistent inflation is, the less exchange rate movements are perceived to betransitory and the more firms might respond via price adjustments. A reduction of inflationpersistence would cause a decline in the long-run effect that exchange rate fluctuations might have oninflation (Takhtamanova, 2008). According to this theory, the pass-through is positively related to theaverage inflation rate and negatively related to inflation persistence (Bitans, 2004).

Another key macroeconomic variable is the exchange rate regime. As noted by Bussiere andPeltonen (2008), a more stable exchange rate regime is likely to induce more pricing-to-market fromforeign exporters and to decrease the pass-through to import prices. On the contrary, if the exchangerate is not used as an intermediate target, expectations are not strongly associated with the exchangerate, resulting in lower pass-through.7 In a country with a crawling peg regime, the preannounceddevaluation offers a nominal anchor for inflationary expectations. Any changes in the exchange ratewill be rapidly incorporated into expectations and thus into prices, both for tradable and non-tradablegoods. This implies a high and relatively homogenous pass-through to consumer prices. The movetoward a more flexible exchange regime, combined with inflation targeting, can break the linkbetween the exchange rate and prices by disconnecting the level of the non-tradable goods price fromthe exchange rate.

Egert and MacDonald (2009) noted little interest in the literature in the role played by the exchangerate regime, despite the fact that the pass-through was thought to be higher for countries where theexchange rate served as a nominal anchor to inflationary expectations.8

As outlined in previous studies, it may be argued that the adoption of a fixed exchange rate regimewas appropriate in the environment of high inflation rates observed in many East European countries(see Bitans, 2004). According to Bitans (2004), the high exchange rate pass-through enhanced theeffectiveness of exchange-rate-based stabilization policies; it also implied that the level at which theexchange rate was fixed had little impact on the real economy. However, when inflation rates in these

6 As noted by Egert and MacDonald (2009), imported goods reach consumers through wholesaling and retailing networks, so

that their prices have a substantial local input which serves as a buffer to soften the impact of exchange rate changes.7 In a floating regime, exchange rate changes have little influence on non-tradable prices.8 In these countries, any change in the exchange rate will be rapidly incorporated into expectations and prices (both tradable

and non-tradable).

R. Jimborean / Economic Systems 37 (2013) 302–329306

countries declined over time, the stabilizing role of the exchange rate diminished along with thedecreasing pass-through effect. At the same time, the exchange rate’s role in the real economygradually increased as shocks to the nominal effective exchange rate produced increasingly persistentdeviations of the real exchange rate from the equilibrium level. In these circumstances, severalcountries opted for the abandonment of their fixed exchange rate regimes; the central banks in thesecountries became less concerned about the impact of exchange rate fluctuations on inflation and,instead, allowed their currencies to respond to various economic shocks. The move toward moreflexible exchange rate regimes tended to reduce the degree of exchange rate pass-through evenfurther. For countries which retained rigid exchange rate regimes either explicitly or implicitly, thedecline in pass-through was relatively less pronounced. As a result, exchange rate fluctuations arerelatively more important in determining inflation and relatively less important for the real exchangerate in these countries.

The business cycle is another determining factor. One measure of the country-specific businesscycle stage is the output gap; it consists of the deviation of the actual real GDP from the ‘‘potential’’ realGDP. A positive gap signals that the economy is running above potential, therefore domestic demand isexpanding and a lower exchange rate pass-through may be observed if export companies try to gainmarket share by absorbing the exchange rate fluctuations in their profit margins in order to quotecompetitive prices. This way, importing countries with growing output gaps may be perceived byexporting firms as an opportunistic incentive to reduce pass-through in terms of scale expansion.

Openness, measured as imports to GDP ratio, is equally an important determining factor. Accordingto Dornbush (1987), higher import penetration should be associated to higher pass-through to importprices.

The importance of the composition of trade and imports (Campa and Goldberg, 2002) has to beconsidered. The pass-through is large for homogenous goods and reduced for differentiated goods,where the pricing-to-market practices are more frequently used. As a consequence, poor countries(with imports formed mainly by homogenous goods) face a higher pass-through compared to richcountries (where the share of manufactured goods in total imports is higher). The changingcomposition of imports (toward more differentiated goods) through the process of economicdevelopment might equally explain the decrease in pass-through during the catching-upprocess.

Expectations represent another important factor that influences the pass-through. When exchangerate changes are perceived as permanent, the pass-through is generally larger compared to situationswhen exchange rate changes are temporary. Nevertheless, in transition economies, exchange ratesrepresent nominal anchors for inflation expectations. If the exchange rate is a credible anchor in acrawling peg or a fixed-type regime, exchange rate changes are rapidly incorporated into expectationsand, consequently, prices. On the contrary, the inflation rate should not be highly connected with theexchange rate in an inflation targeting environment, where the expectations are anchored by theannounced inflation target and credible rules.

2.2. Previous work

Two approaches are generally used in the empirical literature for estimating the exchange ratepass-through. These are the structural VAR (Vector Auto Regressive) models introduced by McCarthy(1999) and the unique econometrical equation including variables in difference (Campa and Goldberg,2002).

In this paper I develop a three step analysis. In a first step I estimate the average exchange rate pass-through in the new EU member states, using the generalized method of moments estimator fordynamic panel-data models. I apply the standard specification used in the pass-through literature forthree price indices: the consumer, the producer and the import price index.9 In a second step, I proceedto a country-by-country analysis, which is completed by an examination of the drivers of theestimated exchange rate pass-through in a third step. Before presenting the empirical work, I brieflyreview previous studies on this issue.

9 Due to poor data availability, the analysis on import prices could not be carried out for Romania and Slovenia.

Table 1Existing studies on exchange rate pass-through in Central and Eastern Europe.

Author Methodology Sample of

countries

Period of

analysis

Results pass-through (PT)

Vector Auto Regressive (VAR) models

Kuijs (2002) Separate estimation

of long-run

relationships

used as error

correction terms

Slovakia 1993–2000 Short-term PT: 40%,

declining gradually

with the appreciation

Gueorguiev (2003) First-difference VAR Romania 1997–2002 PT to CPI: 30–40%, most

of the impact within

12 months

Coricelli et al. (2004) Co-integrated VAR Hungary, Czech

Rep., Poland,

Slovenia

1993–2002 Full PT: Slovenia, Hungary

80%: Poland,

50%: Czech Rep.

Bitans (2004) Recursive VAR in

first differences

13 East

European

countriesa

1993–2003 Cross-country variation

50% PT decline over time

Dynamic panel data

Mihaljek and

Klau (2001)

Single equation

estimation

technique

13 emerging

economiesb

1993–2000 6%: Czech Rep.

45%: Poland54%:

Hungary

Darvas (2001) Time-varying

parameters

framework,

accounting for

regime shifts

during the 1990s

Hungary

Czech Rep.

Poland Slovenia

1993–2000 Long-run PT: 15%:

Czech Rep.

20%: Poland

40%: Hungary, Slovenia

Short run PT: 0 to 10%

a Bulgaria, Cyprus, Czech Republic, Croatia, Estonia, Hungary, Latvia, Lithuania, Macedonia (FYR), Poland, Romania, Slovak Republic,

Slovenia.b Brazil, Chile, Mexico, Peru, Czech Republic, Hungary, Poland, South Africa, Korea, Malaysia, Philippines, Thailand, Turkey.

R. Jimborean / Economic Systems 37 (2013) 302–329 307

The number of empirical studies on the exchange rate pass-through in East European countriesis rather limited. The existing studies consider individual countries or narrow groups of countriesand present as a main drawback the fact that they do not allow for systematic cross-countrycomparisons. In addition, the degree of pass-through depends considerably on the appliedestimation techniques.

In Table 1 I briefly resume the existing studies. The main conclusion that emerges is theexistence of a large heterogeneity across countries regarding the pass-through to the consumerprice index (CPI), which is found to be higher in less developed countries. At the same time, thepass-through to producer prices (PPI) is found to be higher than the pass-through to consumerprices (CPI).

3. Empirical framework and data

3.1. Data description

The sample consists of quarterly data and covers the period January 1996 to December 2011. I usethe Eurostat data for the consumer price index, the unit labor cost (ULC), the ULC-based real effectiveexchange rate, the nominal effective exchange rate, the GDP volume and imports (as share of GDP).The International Financial Statistics (IFS, IMF) database is used for the producer price index. Data onimport prices come from national sources. Information on the exchange rate regimes was taken fromIlzetzki et al. (2008). Data on the Global Manufacturing Purchasing Managers Index (PMI) are from J.P.Morgan and Markit Economics. Tables 9 and 10 in Appendix A provide a description of the data andfurther information on data availability.

R. Jimborean / Economic Systems 37 (2013) 302–329308

3.2. The dynamic panel data analysis

The earlier literature on pass-through focused on studying the behavior of import prices from amicroeconomic perspective, based on the pricing behavior of exporting firms. In this context, it isuseful to consider a simple static profit-maximization problem faced by a foreign exporting firm (thisfirm exports its products to the domestic country):

maxq

p ¼ s�1 pq � CðqÞ (1)

where p denotes profits (in foreign currency), s is the exchange rate measured in units of domesticcurrency per unit of foreign currency, p is the price of the good (in domestic currency), Cð:Þ is the costfunction (in foreign currency units) and q is the quantity demanded for the good.

The first condition obtained when solving Eq. (1) is:

p ¼ sCqm (2)

with Cq being the marginal cost and m the markup of price over marginal cost. The markup is furtherdetailed as m � h=ðh � 1Þ, where h is the price elasticity of demand for the good. According to Eq. (2),the local currency price of the good can vary as a result of a change in the exchange rate, a change in thefirm’s marginal costs and/or a change in the firm’s markup. One should note that the firm’s marginalcost and markup may change independently of the exchange rate. To properly isolate the effects ofexchange rate changes on import prices, it is thus important to take into account the movements inother price determinants when estimating the pass-through.

A log-linear, reduced-form equation may be expressed as:

pt ¼ a þ bst þ lwt þ tyt þ et (3)

where wt measures the marginal cost of the exporter and yt measures the demand conditions in theimporting country, respectively. In the above equation, b measures the exchange rate pass-through.Different versions of Eq. (3) are used in the pass-through literature (see Goldberg and Knetter, 1997).

3.2.1. Econometric specification

I modify the standard pass-through specification (Eq. (3)) in order to estimate the pass-through forthree prices indices (the consumer, the producer and the import price indices):

D pi;t ¼ ai þ ht þX2

j¼1

f jD pi;t� j þ bDsi;t þ bregðDsi;t � regimei;tÞ þ lDulci;t þ tga pi;t þ ei;t (4)

with D pi;t being the rate of change in the relevant aggregate price index for country i in time period t,ai a country-specific effect, ht a time dummy, Dsi;t the rate of change in the nominal effective exchangerate for country i and time period t,10 regimei;t a dummy variable that captures the exchange rateregime (it takes the value 1 for countries in a fixed exchange rate regime and 0 for flexible exchangerate regimes),11 Dulci;t and ga pi;t control variables capturing changes in foreign producer cost anddomestic demand conditions for country i and time period t, ei;t an independent and identicallydistributed error term.

Inflation persistence is captured by the autoregression on inflation,12 namely by the first two lags ofinflation (D pi;t�1 and D pi;t�2).13

10 Even though some countries have opted for a fixed exchange rate regime (thus limiting nominal exchange rate movements),

there is still a considerable degree of variation in the effective nominal exchange rate that allows the exchange rate pass-

through estimation for these countries.11 By including the interaction term between the change in nominal effective exchange rate and the exchange rate regime, we

account for the potential influence of a change in the exchange rate regime on prices.12 According to Fuhrer (2009), several measures of the reduced-form inflation persistence exist: conventional unit root tests;

the autocorrelation function of inflation series; the first autocorrelation of inflation series; the dominant root of the univariate

autoregressive inflation process; the sum of autoregressive coefficients for inflation; unobserved components decompositions

of inflation that estimate the relative contributions of ‘‘permanent’’ and ‘‘transitory’’ components of inflation.13 The inflation persistence is measured by the sum of coefficients of the two lags of inflation.

R. Jimborean / Economic Systems 37 (2013) 302–329 309

The aggregate pass-through may be a function of the exchange rate regime environment(regimei;t). According to Darvas (2001), in an exchange rate targeting environment, a change in theexchange rate might be regarded as more permanent than in a floating regime, implying a higherpass-through.

Movements in the costs of foreign producers that export to the domestic market are captured bythe foreign exporters’ unit labor cost (ULC), defined as ulct ¼ reerulct � st þ ulcdomt (reerulct is theULC-based real effective exchange rate, st is the nominal effective exchange rate and ulcdomt is thedomestic country’s ULC). Both nominal and real effective exchange rate series are trade weighted(since they are calculated against 36 trading partners), so that Dulci;t effectively measures the rateof change in ULC of exporters relative to the domestic country. The output gap is used as a proxy forchanges in domestic demand conditions (i.e. it accounts for possible demand side shocks toinflation).

In Eq. (4) above, there are two coefficients of interest: the coefficient of the rate of change in theexchange rate (b)14 and the coefficient of the interaction term between exchange rate changes andexchange rate regime (breg). The former captures the average rate of short-run exchange rate pass-through (for each price index), while the latter captures the incremental effect due to a change in theexchange rate regime over the period of analysis.

Before estimating Eq. (4), several issues have to be considered, namely the non-stationarity and themeasurement of the domestic demand conditions:

� T

1

co1

on

ag

do1

di1

At

to

G1

as

be

he aggregate price level and the exchange rate follow non-stationary processes. The Fisher typeunit root test for panel data15 shows that both variables are best described as I(1) series; weconsequently use the two variables in their first-difference form.

� T he domestic demand conditions are proxied by the output gap, which is the difference between the

actual and the estimated potential output.16 We apply the Hodrick Prescott (HP) filter on seasonallyadjusted GDP series. In order to mitigate the well-known end-of-sample problem of the HP filteringprocedure,17 we extend the output series by eight quarters by means of forecasts using ARIMAmodels (see Kaiser and Maravall, 1999). A brief overview of the method can be found in Appendix B.For robustness checking, we use the OECD output gap series while bearing in mind the limitationsdue to data availability, since these series exist only for the OECD member states: Czech Republic,Estonia, Hungary, Poland, Slovenia and Slovakia.

3.2.2. Estimation technique and results

I apply the two-step system generalized method of moments (GMM), designed by Arellano andBover (1995) and fully developed by Blundell and Bond (1998). The use of this method is due to theinclusion of the lagged dependent variable as an explanatory variable18 and to the potentialendogeneity of some variables (as is probably the case with the exchange rate term). The Kivietestimator is suggested for estimating panel data models with small N and large T. It is an efficientapproximation of the bias of the least square dummy variable (LSDV) estimator for dynamic panel datamodels, but its main drawback is the fact that the endogeneity of the explanatory variables is not

4 Normally, a depreciation of the currency should be followed by an increase in inflation. Consequently, a positive befficient is expected, since st is the index of foreign currencies per unit of domestic currency.5 The Fisher test combines the p-values from N independent unit root tests, as developed by Maddala and Wu (1999). Based

p-values of individual unit root tests, the Fisher test assumes that all series are non-stationary under the null hypothesis

ainst the alternative that at least one series in the panel is stationary. Unlike the Im-Pesaran-Shin (2003) test, Fisher’s test

es not require a balanced panel.6 Conceptually, the potential output is the level of output achieved when prices and wages are fully flexible; it cannot be

rectly observed, so that any estimate of potential output is subject to considerable uncertainty.7 In the middle of the sample, the HP is a symmetric two-side filter as both leads and lags of output appear in the loss function.

the beginning or the end of the sample, some leads and lags will be unavailable, requiring either the transformation of the HP

a one-side filter toward the edges of the sample, or generating forecasts of output outside the sample of observations. See

uarda (2002) for further details.8 The presence of the lagged dependent variable among the regressors in a specification which considers the individual effect

well brings about a correlation between the error term and the right-hand regressor. In such a case, the OLS estimation would

inconsistent and biased.

R. Jimborean / Economic Systems 37 (2013) 302–329310

resolved. Because of this drawback, I did not estimate the Kiviet estimator. The Arellano–Bond (1991)estimation starts by transforming all regressors, usually by differencing, and uses the generalizedmethod of moments (Hansen, 1982). The Arellano–Bover/Blundell–Bond estimator augmentsArellano–Bond by making an additional assumption, i.e. that the first differences of instrumentvariables are uncorrelated with the fixed effects. This allows the introduction of a larger number ofinstruments and can improve the efficiency of the estimators. I apply the Windmeijer (2005) finite-sample correction, without which the standard errors in two-step estimation tend to be significantlydownward biased because of the large number of instruments. A crucial assumption for the validity ofGMM is that the instruments are exogenous, tested by the Sargan/Hansen test (for the joint validity ofinstruments). The GMM validity also depends on the assumption that the model is not subject to serialcorrelation in ei;t .

The increased number of instruments is a common feature both of Arellano–Bond and Arellano–Bover/Blundell–Bond methodologies. According to Roodman (2008), in small samples numerousinstruments can cause different kinds of problems: the over-fitting of endogenous variables, impreciseestimates of the optimal weighting matrix, downward bias in two-step standard errors,19 and a weakHansen test of instrument validity. I seek to avoid the proliferation of instruments by collapsingthem20 and limiting the lag depth.

Both the short- and long-run elasticities of the model are presented.21 The short-run pass-throughis the immediate reaction of inflation to a change in the exchange rate, while the long-run pass-through is the overall response of inflation to an exchange rate shock. The collapsed instruments arethe third lags of the dependent variable, the first and second lags of the exchange rate term and thefirst lag of the output gap. The unit labor cost is considered to be exogenous.

I examine the results of the estimation of Eq. (4), first by abstracting away the effects of theexchange rate regime on the average pass-through (columns 1, 3, 5 and 7) and then by consideringthese potential effects (columns 2, 4, 6 and 8). The pass-through estimates, both in the short and longrun, for the price indices that we consider (i.e. the consumer price index (CPI), the producer price index(PPI) and the import price index (IPI)), are reported in Tables 2–4 below. More complete estimationresults can be found in Appendix C (Tables 11–13).

For robustness checking, I use the OECD output gap series (where available), completed by my owncalculated output gap for non-OECD members. I equally estimate Eq. (4) while treating the output gapboth as an endogenous and an exogenous variable. In Tables 2–4, I first use my own calculated outputgap (columns 1–4) and then the OECD output gap series (columns 5–8). For both indicators, I firstconsider the output gap as endogenous (columns 1, 2, 5 and 6) and then as exogenous (columns 3, 4, 7and 8). The conclusion that emerges is that of quite robust results.

In Table 2, the CPI pass-through estimates present the expected positive sign (i.e., an increase in thenominal effective exchange rate translates into a depreciation of the currency and should normally befollowed by an increase in inflation) and are statistically significant both in the short and long run. Thisapplies both when we consider only the effects of the exchange rate on the average pass-through(columns 1 and 3) and when we distinguish between the fixed and flexible exchange rate regimes(columns 2 and 4). In the short run, a 1 percent depreciation of the nominal effective exchange rateleads, on average, to an increase in consumer prices between 0.05 and 0.07 percent. In the long run, theincrease in consumer prices determined by a 1 percent depreciation of the nominal effective exchangerate is larger, being situated between 0.08 and 0.11 percent.

When I use the OECD output gap series (columns 5–8), the CPI pass-through estimates present theexpected positive and significant coefficient, but of a slightly different magnitude: a larger magnitudein the short run, where a 1 percent depreciation of the nominal effective exchange rate leads, on

19 Before the Windmeijer correction, researchers considered one-step results in making inferences (Roodman, 2008).20 One instrument is created for each variable and lag distance, rather than one for each time period, variable and lag distance.

This reduces the statistical efficiency in large samples but in small samples it can avoid the bias that arises as the number of

instruments increases with the number of observations (Roodman, 2008).21 The long-term coefficient of a variable is computed as the sum of its coefficients (of its lags and current values, where

applicable) divided by one minus the sum of coefficients of the lags of the dependent variable. This way, the measure of the long-

run pass-through is b=½1 � ðP2

j¼1 f jÞ�; it is intended to capture the effects of an exchange rate change in period t on inflation

over several subsequent periods.

Table 2GMM estimates of CPI exchange rate pass-through over 1996Q1–2011Q4.

Price index CPI CPI CPI CPI CPI CPI CPI CPI

(1) (2) (3) (4) (5) (6) (7) (8)

Short run

Dsi;t 0.058* 0.081 0.070** 0.105 0.064*** 0.087* 0.075*** 0.111**

(0.029) (0.057) (0.027) (0.064) (0.018) (0.039) (0.019) (0.047)

Dsi;t � regimei;t �0.114 �0.151 �0.120 �0.151*

(0.093) (0.100) (0.070) (0.077)

Long run

Dsi;t 0.079*** 0.099** 0.084** 0.110*** 0.074*** 0.095*** 0.078*** 0.101***

(0.023) (0.032) (0.017) (0.025) (0.016) (0.023) (0.011) (0.016)

Dsi;t � regimei;t �0.140 �0.159*** �0.131* �0.137***

(0.080) (0.048) (0.059) (0.032)

Notes: (1) Columns 2, 4, 6 and 8 report the results obtained when including interaction terms to account for exchange rate

regime shifts. The output gap is endogenous in columns 1, 2, 5 and 6 and exogenous in columns 3, 4, 7 and 8. In the first 4

columns the output gap is my own calculation, in the last four columns the OECD data is used where available, completed by my

own calculated indicator; (2) two-step system GMM with the Windmeijer (2005) correction; (3) standard errors in parentheses.* Significance at 10% level.** Significance at 5% level.*** Significance at 1% level.

Table 3GMM estimates of PPI exchange rate pass-through over 1996Q1–2011Q4.

Price index PPI PPI PPI PPI PPI PPI PPI PPI

(1) (2) (3) (4) (5) (6) (7) (8)

Short run

Dsi;t 0.130** 0.108* 0.102* 0.078 0.105* 0.084* 0.095 0.072

(0.048) (0.051) (0.054) (0.070) (0.05) (0.047) (0.054) (0.068)

Dsi;t � regimei;t 0.199* 0.221* 0.230* 0.217*

(0.108) (0.112) (0.129) (0.116)

Long run

Dsi;t 0.200** 0.166* 0.134 0.101 0.167 0.126 0.122 0.092

(0.082) (0.088) (0.088) (0.104) (0.095) (0.081) (0.085) (0.098)

Dsi;t � regimei;t 0.305* 0.288* 0.346* 0.277*

(0.163) (0.137) (0.182) (0.141)

Notes: (1) Columns 2, 4, 6 and 8 report the results obtained when including interaction terms to account for exchange rate

regime shifts. The output gap is endogenous in columns 1, 2, 5 and 6 and exogenous otherwise. In the first 4 columns the output

gap is my own calculation, in the last four columns the OECD data is used where available, completed by my own calculations;

(2) two-step system GMM with the Windmeijer (2005) correction; (3) standard errors in parentheses.* Significance at 10% level.** Significance at 5% level.

R. Jimborean / Economic Systems 37 (2013) 302–329 311

average, to an increase in consumer prices between 0.06 and 0.11 percent, while in the long run, theincrease in consumer prices determined by a 1 percent depreciation of the nominal effective exchangerates is slightly lower, being situated between 0.07 and 0.10 percent.

When taking into account the potential effect of the exchange rate regimes on the relationshipbetween prices and the exchange rate (i.e. columns 2 and 4), the exchange rate regime influences therelationship between the nominal effective exchange rates and the CPI in a statistically significant wayonly in the long run and only when considering the output gap as exogenous. In this case, in the longrun, the increase in consumer prices determined by a 1 percent depreciation of the nominal effectiveexchange rate decreases by 0.16 percent in the fixed exchange rate regimes compared to an averageincrease of 11 percent in CPI.

When considering the OECD output gap series (columns 5–8), the type of exchange rate regime hasa statistically significant effect on the link between CPI and the nominal effective exchange rates both

Table 4GMM estimates of IPI exchange rate pass-through over 1996Q1–2011Q4.

Price index IPI IPI IPI IPI IPI IPI IPI IPI

(1) (2) (3) (4) (5) (6) (7) (8)

Short run

Dsi;t 0.705*** 0.661*** 0.652*** 0.650*** 0.667*** 0.756 0.644*** 0.663***

(0.102) (0.063) (0.112) (0.131) (0.074) (1.231) (0.099) (0.099)

Dsi;t � regimei;t 0.278 0.060 �0.657 0.026

(0.390) (0.185) (8.880) (0.171)

Long run

Dsi;t 1.134*** 1.043*** 0.921*** 0.906*** 1.052*** 1.182 0.916*** 0.912***

(0.230) (0.191) (0.138) (0.186) (0.149) (1.676) (0.136) (0.122)

Dsi;t � regimei;t 0.438 0.084 �1.027 0.036

(0.571) (0.255) (13.657) (0.236)

Notes: (1) Columns 2, 4, 6 and 8 report the results obtained when including interaction terms to account for exchange rate

regime shifts. The output gap is endogenous in columns 1, 2, 5 and 6 and exogenous otherwise. In the first 4 columns the output

gap is my own calculation, in the last four columns the OECD data is used where available, completed by my own calculations;

(2) two-step system GMM with the Windmeijer (2005) correction; (3) standard errors in parentheses.*** Significance at 1% level.

R. Jimborean / Economic Systems 37 (2013) 302–329312

in the short and long run in the same way as above, i.e. that of a decrease in CPI for fixed exchange rateregime countries.

Table 3 presents the PPI pass-through estimates. When estimating Eq. (4) without distinguishingbetween exchange rate regimes (columns 1 and 3), the PPI pass-through presents a positive andstatistically significant coefficient. In the short run, a 1 percent depreciation of the nominal effectiveexchange rate leads, on average, to an increase in producer prices between 10 and 13 percent. In thelong run the increase in producer prices determined by a 1 percent depreciation of nominal effectiveexchange rates is larger, being situated between 0.16 and 0.20 percent. When we consider thedifferent exchange rate regimes (columns 2 and 4), the PPI pass-through estimates continue to bestatistically significant.

When I use the OECD output gap series (columns 5 to 8), the PPI pass-through estimates present theexpected positive and significant coefficient, but of a slightly different magnitude. It is of a lowermagnitude in the short run, where a 1 percent depreciation of the nominal effective exchange rateleads, on average, to an increase in producer prices between 0.08 and 0.10 percent, while in the longrun, the PPI pass-through estimates are no longer significant.

When taking into account the potential effect of the exchange rate regime on the relationshipbetween prices and the exchange rate (i.e. columns 2 and 4), the exchange rate regime influences therelationship between the nominal effective exchange rates and the PPI in a statistically significant wayboth in the short and long run. Thus, in the short run the increase in producer prices determined by a 1percent depreciation of nominal effective exchange rates increases in the fixed exchange rate regimesby 0.19–0.22 percent compared to an average increase between 10 and 13 percent in PPI. At the sametime, in the long run, the increase in producer prices determined by a 1 percent depreciation ofnominal effective exchange rates increases in the fixed exchange rate regimes by 0.28–0.30 percentcompared to an average increase in PPI situated between 16 and 20 percent.

When considering the OECD output gap series (columns 5–8), the exchange rate regime type has astatistically significant effect on the link between PPI and the nominal effective exchange rates both inthe short and long run, in the same way as above, i.e. that of an increase in the PPI for fixed exchangerate regime countries.

Concerning the IPI pass-through estimates (Table 4), both the short- and long-run exchange ratepass-through are statistically significant and present the expected positive sign. In the short run, a 1percent depreciation of the nominal effective exchange rate leads to an increase in import prices of

R. Jimborean / Economic Systems 37 (2013) 302–329 313

0.65–0.70 percent. In the long run, the increase in IPI is more pronounced at about 0.90–1.13 percent.When I distinguish between fixed and flexible exchange rate regimes (columns 2 and 4), the IPI pass-through estimates continue to be statistically significant.

When I use the OECD output gap series (columns 5–8), the IPI pass-through estimates present theexpected positive and significant coefficient of a slightly similar magnitude. In the short run, a 1percent depreciation of the nominal effective exchange rate leads, on average, to an increase in importprices between 0.64 and 0.75 percent, while in the long run, the increase in import prices determinedby a 1 percent depreciation of the nominal effective exchange rate is situated between 0.91 and 1.05percent.

When taking into account the potential effect of the exchange rate regime on the relationshipbetween prices and the exchange rate (i.e. columns 2 and 4), the exchange rate regime does notinfluence the relationship between the nominal effective exchange rates and the IPI in a statisticallysignificant way either in the short or in the long run.

When considering the OECD output gap series (columns 5–8), the exchange rate regime type doesnot have a statistically significant effect on the link between IPI and the nominal effective exchangerate either in the short or in the long run.

The interaction term between the rate of change in the exchange rate and the dummy variablesthat capture the type of exchange rate arrangement (Dsi;t � regimei;t) is statistically significant only inthe case of CPI and PPI estimations. This implies that the effects of exchange rate movements onconsumer and/or producer prices are affected by the exchange rate regime, while the effects ofexchange rate movements on import prices are not affected by the exchange rate regime.

I conclude by assessing the existence of a significant average pass-through to the consumer priceindex (CPI), the producer price index (PPI) and the import price index (IPI), both in the short and longrun. An issue that deserves further attention in this first step of my analysis is the increasedheterogeneity in the sample of countries. As is known, the panel data estimations impose slopehomogeneity across countries, a hypothesis which is rather unrealistic. An individual analysis at thelevel of each country could provide us with some useful information on the differences in pass-through estimates, and we therefore proceed with it in the following subsection.

3.3. Country-by-country analysis

As in Brun-Aguerre et al. (2012), I estimate individual country-by-country equations and thenaverage the exchange rate pass-through coefficients across countries in order to obtain the meangroup (MG) estimator. This is the simplest way to exploit the panel structure of the sample of countrieswhile allowing for full heterogeneity; by doing so, all parameters are country-specific. Thus, I estimatethe exchange rate pass-through to CPI, PPI and IPI for each of the 10 countries in the sample by usingthe same specification as in the dynamic panel data analysis. In this way I obtain a baseline set of pass-through estimates that I can compare with those of previous studies.

D pt ¼ a þ ht þX2

j¼1

f jD pt� j þ bDst þ lDulct þ tga pt þ et (5)

with D pt being the rate of change in the relevant aggregate price index for each country at time periodt, a a country-specific effect, ht a time dummy, Dst the rate of change in the nominal effectiveexchange rate for each country and time period t, Dulct and ga pt control variables capturing changesin foreign producer cost and domestic demand conditions for each country and time period t, et anindependent and identically distributed error term.

3.3.1. Statistical preliminaries

Before proceeding to the empirical estimations, several empirical tests are computed in order tochoose the right specification model.

I first compute the Dickey–Fuller unit root test, then test for the autocorrelation of residuals (theBox–Ljung Q test) and for the serial correlation (the Breusch–Godfrey or Lagrange Multiplier (LM)test). I also compute the Breusch–Pagan/Cook–Weisberg heteroskedasticity test.

Table 5The exchange rate pass-through to consumer prices.

Country DCPIt�1 DCPIt�2 DNEER Output gap Dulct No. obs. R2

Bulgaria 0.200** �0.022 �0.271** �0.168*** �0.012 58 0.535

Czech Republic �0.195** �0.085 �0.034 0.175 �0.029** 60 0.169

Estonia 0.428*** �0.009 �0.288*** �0.075*** �0.002 61 0.533

Hungary 0.752*** 0.018 �0.009 0.006 0.003 61 0.644

Latvia 0.718*** �0.188* 0.051 �0.055** �0.023** 61 0.754

Lithuania 0.404** 0.182* 0.060 �0.050 �0.003 61 0.544

Poland 0.465** 0.297 �0.003 �0.066 �0.007 61 0.712

Romania 0.447*** 0.419*** �0.022 0.017 0.001 47 0.878

Slovakia 0.177*** 0.343** 0.003 �0.046 0.019 60 0.190

Slovenia 0.417*** 0.207** �0.178* �0.011 �0.022 61 0.519

Mean group �0.070* Statistical significance at the 10% levels.** Statistical significance at the 5% levels.*** Statistical significance at the 1% levels.

Table 6The exchange rate pass-through to producer prices.

Country DPPIt�1 DPPIt�2 DNEER Output gap Dulct No. obs. R2

Bulgaria 0.382*** �0.042 0.504** �0.017 �0.035 58 0.418

Czech Republic �0.195* �0.085 �0.034 0.175 �0.029** 60 0.134

Estonia 0.466*** �0.050 0.387*** 0.070** �0.015 61 0.512

Hungary 0.443*** �0.091 0.314*** 0.088 0.110*** 60 0.719

Latvia 0.852*** �0.457*** 0.176** 0.068 0.044** 61 0.647

Lithuania 0.028 �0.319** 0.042 �0.037 �0.007 60 0.112

Poland 0.372*** �0.083 0.115*** 0.111 0.034* 61 0.380

Romania 0.529*** 0.158 0.151** �0.066 0.006 47 0.705

Slovakia 0.274* �0.013 �0.011 0.069 �0.016 60 0.105

Slovenia �0.050 �0.636*** �0.082 0.260*** 0.004 60 0.441

Mean group 0.156* Statistical significance at the 10% levels.** Statistical significance at the 5% levels.*** Statistical significance at the 1% levels.

R. Jimborean / Economic Systems 37 (2013) 302–329314

The results of these preliminary statistical tests are presented in Appendix D, in Tables 14–16. Thenominal effective exchange rate, the CPI, the PPI and the IPI follow I(1) processes and their first-difference forms follow I(0) processes. There are problems of serial correlation in several cases: for PPIin the Czech Republic, Hungary, Lithuania and Slovenia, and, respectively, for IPI in the Czech Republicand Estonia. The aspects of autocorrelation and heteroskedasticity are also taken into account.

Regarding the estimation technique, I apply the ordinary least squares (OLS) method, except for thecases where I detect the presence of serial correlation, in which case I adjust the linear model for serialcorrelation in the error term through the Cochrane–Orcutt (1949) estimation.

3.3.2. Estimates of the pass-through coefficients

In this subsection I summarize the results of the estimation of the exchange rate pass-through toconsumer prices, producer prices and import prices.

In the case of consumer price regressions (Table 5), the estimation results show that the coefficientsof the nominal exchange rate pass-through are statistically significant and have the negative sign inBulgaria, Estonia and Slovenia (the magnitude is �27 percent in Bulgaria, �28 percent in Estonia and�17 percent in Slovenia); they are not statistically significant in the remaining countries. I computethe mean group estimator as the average of the individual coefficients estimated for each country; it islisted on the last line of the table and amounts to �7 percent. A comparison of the figures obtainedthrough individual country-by-country estimations with the average short-run exchange rate pass-through to CPI estimated in Table 2 (of 5.8 percent) shows some rather different results.

The producer price regressions (Table 6) show a positive and statistically significant coefficient forthe nominal effective exchange rate, but only for Bulgaria, Estonia, Hungary, Latvia, Poland and

Table 7The exchange rate pass-through to import prices.

Country DIPIt�1 DIPIt�2 DNEER Output gap Dulct No. obs. R2

Bulgaria 0.137 �0.088 1.484** �0.209 �0.041 41 0.237

Czech Republic 0.118 �0.157 0.487*** 0.133 0.012 52 0.496

Estonia �0.013 �0.474*** 0.733*** 0.192* �0.034 52 0.522

Hungary 0.359** �0.143 0.734*** 0.310 �0.156 33 0.673

Latvia 0.209 �0.022 0.554** 0.029 0.056 53 0.263

Lithuania 0.526*** �0.171* 0.427 �0.159 0.001 57 0.297

Poland 0.082 �0.052 0.530*** �0.189 0.010 61 0.302

Slovakia 0.330** �0.012 0.917** �0.578 0.051 60 0.211

Mean group 0.733* Statistical significance at the 10% levels.** Statistical significance at the 5% levels.*** Statistical significance at the 1% levels.

R. Jimborean / Economic Systems 37 (2013) 302–329 315

Romania. The magnitudes are 50 percent in Bulgaria, 38 percent in Estonia, 31 percent in Hungary, 17percent in Latvia, 11 percent in Poland and 15 percent in Romania. I compute the mean groupestimator as the average of the individual coefficients estimated for each country; it is listed on the lastline of the table and amounts to 15.6 percent. When comparing the mean group estimator with theaverage short-run PPI pass-through estimated in Table 3 (of 13 percent), I find quite similar results;when I look at the figures obtained through the individual analysis, these are superior to the averagepanel data estimator in Bulgaria, Estonia, Hungary and Romania, and slightly inferior in Poland.

The import price regressions (Table 7) show a positive and statistically significant coefficient forthe nominal effective exchange rate for all countries except Lithuania (less Romania and Slovenia,where there is no available data on IPI). The magnitudes are 148 percent in Bulgaria, 48 percent in theCzech Republic, 73 percent in Estonia and Hungary, 55 percent in Latvia, 53 percent in Poland and 92percent in Slovakia. I compute the mean group estimator as the average of the individual coefficientsestimated for each country; it appears in the last line of the table and amounts to 73.3 percent. Whencomparing the mean group estimator with the average short-run IPI pass-through estimated in Table 4(of 70 percent), I find quite similar results; when I look at the figures obtained through the individualanalysis, these are superior to the average panel data estimator in Bulgaria, Estonia, Hungary andSlovakia, and slightly inferior in the Czech Republic, Latvia and Poland.

The individual analysis confirms the heterogeneity in the estimated exchange rate pass-through toconsumer, producer and import prices. I wonder what the factors are that might explain thedifferences in the estimated exchange rate pass-through and proceed to their analysis in the last partof my empirical work.

3.3.3. Robustness tests and stability analysis

I investigate the possibility that the variables in Eq. (5) may co-integrate, as suggested by De Bandtet al. (2007) and Bussiere and Peltonen (2008). I thus test for co-integration with the two-step processsuggested by Engle and Granger (1987), the EG-ADF test. The results are presented in Appendix D,Table 17. Overall, there is not much evidence for a co-integrating relation among variables, as theresiduals of the long-run relationships in levels are not stationary.

I then test the stability of the estimated models by focusing on the coefficient of the exchange rate.As shown by Bussiere and Peltonen (2008), the stability of parameters is not only an issue for theemerging markets but also for the advanced economies, where a structural fall in the degree of pass-through took place. Benchmark Eq. (5) is estimated using a rolling sample of 30 quarters22 to verifywhether a decrease in the exchange rate pass-through took place over time. The results are reported inFigs. 1–3 in Appendix E. One should bear in mind that in this case a decrease in pass-through means anascending trend on the graphs, since the relationship between prices and exchange rates is positive(i.e., an increase in the exchange rates, which translates as depreciation, is followed by an increase in

22 The same estimations were performed with a rolling sample of 20 quarters for robustness checking and the results were

broadly the same.

R. Jimborean / Economic Systems 37 (2013) 302–329316

prices). Starting with the regression for consumer prices, a slight reduction in the degree of pass-through can be observed for the Czech Republic, Latvia, Lithuania and Slovenia. Stable patterns areobserved in Bulgaria, Poland and Romania, while a slight increase in pass-through is found in Estonia,Hungary and Slovakia. Turning to the producer price equation, a stable pattern is observed in Bulgariaand Poland, while the pass-through decreases in Estonia, Hungary, Latvia, Lithuania, Slovakia andSlovenia and even increases in the Czech Republic and Romania. As far as the import price regressionsare concerned, stable patterns are observed in Hungary, Poland and Slovakia, while the pass-throughdecreases in Bulgaria, Estonia, Latvia and Lithuania and increases in the Czech Republic.

3.4. The drivers of pass-through estimates

A question that arises is that of the factors explaining these evolutions in the estimated pass-through. I begin the analysis of the determinants of cross-country differences in the exchange ratepass-through with a rolling-window estimation of (5) in order to obtain time series of short-runexchange rate pass-through coefficients; I use the 30 quarter rolling windows as before (in the analysisof the stability of the individual estimates). I then pool the time series coefficients across countries andregress them on the one-quarter lagged drivers, while controlling for the unobserved country- ortime-specific factors via panel fixed or random effects formulations.

I estimate the following regression model:

bi;t ¼ ai þ ht þ g iXi;t�1 þ ei;t (6)

with bi;t being the estimated short-run exchange rate pass-through for each country i at time period t,ai the country-specific effect, ht the time-specific effects, Xi;t�1 the determinants of the exchange ratepass-through estimates, ei;t an independent and identically distributed error term.

The variables included in the vector Xi;t�1

�� are the following:

� T

he volatility of the exchange rate, whose relationship with the degree of the pass-through is notclearcut in the literature. Devereux and Yetman (2003) define a positive relationship betweenexchange rate volatility and consumer prices in a theoretical model of endogenous price stickiness.According to Gaulier et al. (2008), the trade-off of the exporter’s main strategy between stabilizingthe marginal profit and/or stabilizing the export volumes explains the ambiguous impact ofexchange rate volatility on pass-through. I compute the quarterly exchange rate volatility as theone-year moving average of the variance of the exchange rate indicator. � T he volatility of inflation, expected to present a positive relationship with the exchange rate pass-

through to domestic prices (Taylor, 2000; Choudhri and Hakura, 2006). I compute it as the movingaverage of the quarterly change in the consumer price index of each country.

� T he output gap, already calculated above. A positive gap means that the economy is running above

potential, and therefore domestic demand is expanding; a lower exchange rate pass-through mightbe observed if export companies try to gain market share by absorbing the exchange ratefluctuations in their profit margins in order to apply competitive prices.

� Im port dependence, computed as the share of imports in the GDP for each country. According

to Dornbush (1987), in highly import-dependent economies the pass-through tends to behigher.

� G lobal economic sentiment, proxied by the Global Manufacturing Purchasing Managers Index (PMI),

which might affect the pass-through dynamics.

Regarding the estimation technique, I apply the fixed-effects panel data model with Driscoll–Kraay standard errors when estimating the drivers of the exchange rate pass-through to PPI (EPRTPPI) and to IPI (EPRT IPI), and the feasible general least square (FGLS) method when estimating thedrivers of the exchange rate pass-through to CPI (EPRT CPI). The use of the first method is justified bythe presence of both heteroskedasticity and autocorrelation in a fixed effect model, while the use ofthe latter is due to random effects detected by the Hausman test. The results are presented in Table 8below.

Table 8Drivers of CPI, PPI and IPI exchange rate pass-through.

EPRT CPI EPRT PPI EPRT IPI

(1) (2) (3)

Exchange rate �0.0001* �0.001 �0.005

Volatility (0.00007) (0.001) (0.003)

Inflation 0.002 0.097** 0.198*

Volatility (0.004) (0.043) (0.108)

Import �0.009** 1.533** 3.257**

Dependence (0.004) (0.552) (1.225)

Output gap �0.0005 �0.078*** �0.121***

(0.0009) 0.020 (0.030)

Global economic 0.009 �0.699*** �1.036*

Sentiment (0.011) (0.158) (0.525)

Hausman test 0.807 0.0039 0.000

Heteroskedasticity test 0.000 0.0000 0.000

Autocorrelation test 0.000 0.0000 0.0002

No. of observations 350 350 277

R2 – 0.232 0.243

Model Random effects Fixed effects Fixed effects

FGLS With Driscoll–Kray With Driscoll–Kray

Standard errors Standard errors

Notes: (1) Null hypothesis of the Hausman test: difference in coefficients is not systematic. Null hypothesis of the modified Wald

test for groupwise heteroskedasticity: constant variance. Null hypothesis of the Wooldridge (autocorrelation) test: no

autocorrelation; (2) standard errors in parentheses; (3) observation: all the explanatory variables are lagged by one quarter.* Significance at the 10% levels.** Significance at the 5% levels.*** Significance at the 1% levels.

R. Jimborean / Economic Systems 37 (2013) 302–329 317

As can be seen, exchange rate volatility affects only the CPI exchange rate pass-through (column 1);the coefficient is slightly significant (at 10 percent confidence level) and presents a negative sign,showing that the increased volatility of the exchange rates diminishes the pass-through to consumerprices.

Inflation volatility presents the expected positive and significant coefficient, but only for PPI and IPIexchange rate pass-through (columns 2 and 3).

Import dependence presents the expected positive and significant coefficient, but only for PPI andIPI exchange rate pass-through (columns 2 and 3). Thus, the higher the share of imports (as apercentage of GDP), the higher is the pass-through to producer and import prices.

The output gap seems to affect only the PPI and IPI exchange rate pass-through (columns 2 and3); it presents a negative and significant coefficient. Thus, the higher the output gap, the lower isthe pass-through to producer and import prices. These results are in line with the expectedsign.

The global economic sentiment seems to affect the pass-through to producer and import pricesnegatively and in a statistically significant way (columns 2 and 3). The higher the PMI, the lower is thePPI and IPI pass-through. High PMI values translate into an improvement in economic conditions thatmight be accompanied by an expansion of sales, with exporters being better able to acceptfluctuations in markups, so that the exchange rate pass-through might fall.

I also tested for the potential non-linear impact of exchange rate changes on the pass-throughestimates by distinguishing between depreciations and appreciations. They do not have a significantimpact on any of the three pass-through estimates, so that I did not report them.

4. Conclusions

This paper investigates the relationship between changes in nominal exchange rates and prices inthe new EU member states over the period 1996–2011.

R. Jimborean / Economic Systems 37 (2013) 302–329318

In a first step, I examine whether changes in exchange rates affect the consumer, the producer andimport prices through an analysis at the aggregate cross-country level with quarterly data. I alsoinvestigate the potential effect of shifts in exchange regimes on the average exchange rate pass-through. I find evidence of a significant exchange rate pass-through to consumer, producer and importprices, both in the short and long run. Thus, in the case of consumer prices, I find an exchange ratepass-through between 5 and 7 percent in the short run and between 8 and 11 percent in the long run.In the case of producer prices, I find an exchange rate pass-through between 10 and 13 percent in theshort run and between 16 and 20 percent in the long run, while for import prices, the exchange ratepass-through is between 65 and 70 percent in the short run and between 90 and 110 percent in thelong run. My results are in line with those of previous studies.

In a second step, I assess the exchange rate impact on import, producer and consumer prices at anindividual level, country by country. The conclusion that emerges is that of an increased heterogeneityin the exchange rate pass-through estimates. I find evidence of a positive and statistically significantexchange rate pass-through to producer prices in Bulgaria, Estonia, Hungary, Latvia, Poland andRomania with a magnitude that is higher compared to the average producer prices exchange rate pass-through estimated through the panel data analysis, except for Poland. Regarding the exchange ratepass-through to import prices, it is positive and statistically significant in all the countries except forLithuania and its magnitude is larger than the average import price exchange rate pass-throughestimated through the panel data analysis.

I test for the stability of the estimated models, focussing on the coefficient of the exchange rate. Itherefore estimate the benchmark equations using a rolling sample of 30 quarters to verify inparticular whether there has been a decrease in the degree of exchange rate pass-through over time.Overall, the results show some rather stable patterns.

In a third step, I examine the macro drivers of the exchange rate pass-through estimates obtainedin the second step of the analysis. Among them, exchange rate volatility affects the CPI pass-throughnegatively, inflation volatility and import dependence increase both PPI and IPI pass-through, whilethe output gap and global economic sentiment have a decreasing effect on PPI and IPI pass-throughestimates.

Acknowledgements

I am grateful to Balasz Egert, Pavel Diev, Matthieu Bussiere, and an anonymous referee, as well asparticipants at the Banque de France seminar and the 28th GdRE International Annual Symposium onMoney, Banking and Finance held at the University of Reading, for valuable comments andsuggestions. Nathalie Raveau provided excellent research assistance. The views expressed in thispaper are those of the author and do not necessarily reflect the views of the Banque de France.

Appendix A. Sources and data availability

See Table 10.

Table 9Variables definitions and sources.

Variable Source Definition

ULC Eurostat Nominal unit labor cost, index, 2000=100, NSA

Own calculations for Hungary, Poland and Romania, as

a ratio of compensation on employee to labor productivity

on employee

NEER Eurostat Nominal effective exchange rate, 36 trading partners,

average index, 1999=100

REER ULC Eurostat Real effective exchange rate, deflator: ulc for the whole

economy, 36 trading partners, average, index, 1999=100

Imports (% GDP) Eurostat Imports (as a percentage of GDP)

Output Eurostat GDP in volume

Table 9 (Continued )

Variable Source Definition

PPI IFS (IMF) Producer price index or wholesale price index, index,

2005=100

IPI Datastream Import price index

HICP Eurostat Harmonized index of consumer price, overall index,

monthly index, SA, not working day adjusted

Output gap OECD Quarterly output gap

Exchange rate regime Ilzetzki, Reinhart

and Rogoff (IRR) 2008

Dummy variable (1for fixed exchange rate regime and 0 for

flexible exchange rate regime), based on IRR (2008)

database on exchange rate arrangements

PMI JP Morgan Markit

Economics

Global Manufacturing Purchasing Managers Index

(values above 50 signals improving economic conditions)

IFS, International Financial Statistics; IMF, International Monetary Fund.

Table 10Sample period by country.

Country CPI, PPI, NEER IPI GDP ULC

Bulgaria 1996Q1–2011Q4 2001Q1–2011Q4 1996Q1–2011Q4 1996Q1–2011Q4

Czech Republic 1996Q1–2011Q4 1998Q1–2011Q4 1996Q1–2011Q4 1996Q1–2011Q4

Estonia 1996Q1–2011Q4 1998Q1–2011Q4 1996Q1–2011Q4 1996Q1–2011Q4

Hungary 1996Q1–2011Q4 2003Q1–2011Q4 1996Q1–2011Q4 1996Q1–2011Q4

Latvia 1996Q1–2011Q4 1998Q1–2011Q4 1996Q1–2011Q4 1996Q1–2011Q4

Lithuania 1996Q1–2011Q4 1997Q1–2011Q4 1996Q1–2011Q4 1996Q1–2011Q4

Poland 1996Q1–2011Q4 1996Q1–2011Q4 1996Q1–2011Q4 1996Q1–2011Q4

Romania 1996Q1–2011Q4 – 1998Q1–2011Q4 1999Q1–2011Q4

Slovakia 1996Q1–2011Q4 1996Q1–2011Q4 1996Q1–2011Q4 1996Q1–2011Q4

Slovenia 1996Q1–2011Q4 – 1996Q1–2011Q4 1996Q1–2011Q4

Note: All data are quarterly. One can notice the scarce data on the import price index (IPI) compared to the other price indices

(PPI and CPI).

R. Jimborean / Economic Systems 37 (2013) 302–329 319

Appendix B. Potential output – the Hodrick–Prescott filter

There exist two main approaches to measure the potential output: the univariate methods,essentially statistical, that identify the output gap solely from past behavior of output series withoutreferring to any macroeconomic variables and that are based only on some (explicit or implicit)assumptions about the dynamics of output series; and the multivariate methods that consider both thepast behavior of output series and the evolution of other macroeconomic variables, and that exploitthe relationships derived from the economic theory (such as the Phillips curve) for obtaining ameasure of potential output closer to the notion of sustainable aggregate supply capacities of theeconomy.

Both approaches imply several measurement issues. Moreover, additional challenges appear whenestimating the output gap of an emerging market economy, in which case, because of the lack ofreliable data, the potential output is mostly measured through a statistical technique (Tanaka andYoung, 2008).

The univariate filtering methods allow potential output growth to change smoothly through time.The HP filter of Hodrick and Prescott (1997) has proved the most popular univariate filtering method.It is based on the assumption that a given time series yt is the sum of a trend or growth component gt

and a cyclical component ct .

yt ¼ gt þ ct

For t ¼ 1; :::; T . The measure of smoothness of fgtg path is chosen to be the sum of the squares of itssecond difference. ct , the cyclical component, represents deviations from gt; their average is assumed

R. Jimborean / Economic Systems 37 (2013) 302–329320

to be near zero over long time periods.The growth component gt is extracted by minimizing thefollowing loss function:

MinfgtgT

t¼�1

XT

t¼1

c2t þ l

XT

t¼1

½ðgt � gt�1Þ � ðgt�1 � gt�2Þ�2

( )

The parameter l represents terms on which deviations from the trend are traded off againstvariability in the trend. The higher l, the ‘‘stiffer’’ is the trend component.23 The results can besensitive to the choice of l and no objective criterion of choosing this parameter exists. Hodrickand Prescott (1997) recommend a value of l ¼ 100 for yearly data and l ¼ 1600 for quarterlydata.24

Several pros and cons are associated with the use of this method. The pros consist in the fact that itextracts the relevant business-cycle frequencies of the spectrum and closely approximates the cyclicalcomponent implied by reasonable time-series models of output. One attraction of the HP filter is that itmay be applied to non-stationary time series (series containing one or more unit roots in theirautoregressive representation), a relevant concern for many macroeconomic and financial time series.Nevertheless, several studies raise doubts about the above-mentioned affirmations and the reliabilityof the HP filter as a mean of extracting trend components (see Harvey and Jaeger, 1993; Cogley andNason, 1995; Guay and St-Amant, 1996).

Appendix C. Estimation results for Eq. (4)

In the three tables above, columns 2, 4, 6 and 8 report the results obtained when includinginteraction terms to account for exchange rate regime shifts. The output gap is endogenous incolumns 1, 2, 5 and 6 and exogenous otherwise. In the first 4 columns the output gap is my owncalculation, in the last four columns the OECD data is used where available, completed by my owncalculations.

23 When l ! 1 the trend becomes a straight line and the HP filter gives the same result as the linear time trend method.24 These values have become ‘‘standards’’, despite several attempts to determine the value of l endogenously (see Ravn and

Uhlig, 2002).

Table 11GMM estimates of CPI exchange rate pass-through over 1996Q1–2011Q4.

Price index CPI CPI CPI CPI CPI CPI CPI CPI

(1) (2) (3) (4) (5) (6) (7) (8)

D pi;t�1 0.287 0.187 0.160 0.007 0.125 0.049 �0.006 �0.191

(0.340) (0.484) (0.317) (0.537) (0.215) (0.331) (0.273) (0.457)

D pi;t�2 �0.021 �0.0006 0.008 0.041 0.017 0.033 0.049 0.089

(0.084) (0.118) (0.076) (0.128) (0.052) (0.081) (0.065) (0.109)

Dsi;t 0.058* 0.081 0.070** 0.105 0.064*** 0.087* 0.075*** 0.111**

(0.029) (0.057) (0.027) (0.064) (0.184) (0.039) (0.019) (0.047)

Dsi;t � regimei;t �0.114 �0.151 �0.120 �0.151*

(0.093) (0.100) (0.070) (0.077)

Dulci;t �0.010 �0.012 �0.014 �0.015 �0.014* �0.015* �0.018** �0.021*

(0.008) (0.009) (0.008) (0.010) (0.007) (0.007) (0.007) (0.009)

Out putga pi;t �0.085 �0.071 �0.093*** �0.109* �0.088* �0.079* �0.065** �0.077*

(0.059) (0.049) (0.032) (0.052) (0.043) (0.037) (0.025) (0.035)

AR1 (p-value) 0.115 0.232 0.136 0.361 0.069 0.154 0.148 0.421

AR2 (p-value) 0.402 0.603 0.573 0.092 0.493 0.723 0.865 0.765

Hansen test (2nd step) (p-value) 0.312 0.268 0.246 0.227 0.478 0.362 0.394 0.401

No. of instruments 9 10 8 9 9 10 10 9

No. of observations 592 592 592 592 593 593 593 593

Notes: (1) Two-step system GMM with the Windmeijer (2005) correction; (2) standard errors in parentheses.* Significance at the 10% levels.** Significance at the 5% levels.*** Significance at the 1% levels.

R.

Jimb

orea

n /

Eco

no

mic

System

s 3

7 (2

01

3)

30

2–

32

9

32

1

Table 12GMM estimates of PPI exchange rate pass-through over 1996Q1–2011Q4.

Price index PPI PPI PPI PPI PPI PPI PPI PPI

(1) (2) (3) (4) (5) (6) (7) (8)

D pi;t�1 0.363** 0.378** 0.235 0.239 0.389** 0.359* 0.212 0.216

(0.142) (0.149) (0.192) (0.191) (0.145) (0.190) (0.200) (0.193)

D pi;t�2 �0.013 �0.029 0.003 �0.006 �0.019 �0.024 0.008 0.0001

(0.042) (0.041) (0.052) (0.053) (0.050) (0.055) (0.054) (0.055)

Dsi;t 0.130** 0.108* 0.102* 0.078 0.105* 0.084* 0.095 0.072

(0.048) (0.065) (0.054) (0.070) (0.05) (0.047) (0.054) (0.068)

Dsi;t � regimei;t 0.199* 0.221* 0.230* 0.217*

(0.108) (0.112) (0.129) (0.116)

Dulci;t 0.0001 �0.0008 �0.0016 0.0019 0.001 0.001 0.0001 0.004

(0.008) (0.008) (0.012) (0.013) (0.009) (0.009) (0.013) (0.012)

Out putga pi;t 0.283*** 0.342*** 0.070** 0.091** 0.249** 0.257** 0.048** 0.061**

(0.081) (0.071) (0.024) (0.029) (0.086) (0.090) (0.019) (0.023)

AR1 (p-value) 0.097 0.086 0.164 0.144 0.097 0.132 0.198 0.169

AR2 (p-value) 0.228 0.250 0.217 0.238 0.237 0.248 0.213 0.231

Hansen test (2nd step)

(p-value)

0.569 0.719 0.486 0.257 0.369 0.530 0.403 0.205

No. of instruments 10 10 8 9 10 10 8 9

No. of observations 592 592 592 592 593 593 593 593

Notes: (1) Two-step system GMM with the Windmeijer (2005) correction; (2) Standard errors in parentheses.* Significance at the 10% levels.** Significance at the 5% levels.*** Significance at the 1% levels.

R.

Jimb

orea

n /

Eco

no

mic

System

s 3

7 (2

01

3)

30

2–

32

93

22

Table 13GMM estimates of IPI exchange rate pass-through over 1996Q1–2011Q4.

Price index IPI IPI IPI IPI IPI IPI IPI IPI

(1) (2) (3) (4) (5) (6) (7) (8)

D pi;t�1 0.283 0.233 0.238 0.214 0.315** 0.295 0.233 0.157

(0.334) (0.195) (0.163) (0.291) (0.129) (0.438) (0.163) (0.178)

D pi;t�2 0.094 0.132 0.052 0.068 0.050 0.064 0.063 0.116

(0.246) (0.146) (0.133) (0.197) (0.099) (0.305) (0.139) (0.126)

Dsi;t 0.705*** 0.661*** 0.652*** 0.650*** 0.667*** 0.756 0.644*** 0.663***

(0.102) (0.063) (0.112) (0.131) (0.074) (1.231) (0.099) (0.099)

Dsi;t � regimei;t 0.278 0.060 �0.657 0.026

(0.390) (0.185) (8.880) (0.171)

Dulci;t �0.100** �0.082** �0.006 �0.004 �0.062* �0.067 �0.001 0.003

(0.039) (0.030) (0.015) (0.020) (0.029) (0.081) (0.021) (0.017)

Out putga pi;t 0.916* 0.913*** �0.060 �0.060 0.763*** 0.845 �0.023 �0.067

(0.398) (0.254) (0.096) (0.092) (0.203) (1.364) (0.038) (0.059)

AR1 (p-value) 0.304 0.113 0.151 0.265 0.017 0.378 0.162 0.141

AR2 (p-value) 0.635 0.492 0.503 0.627 0.535 0.756 0.539 0.459

Hansen test (2nd step)

(p-value)

0.514 0.520 0.476 0.476 0.592 0.580 0.470 0.654

No. of instruments 9 9 8 9 9 9 8 9

No. of observations 411 411 411 411 412 412 412 412

Notes: 1. Two-step system GMM with the Windmeijer (2005) correction. 2. Standard errors in parentheses.* Significance at the 10% levels.** Significance at the 5% levels.*** Significance at the 1% levels.

R.

Jimb

orea

n /

Eco

no

mic

System

s 3

7 (2

01

3)

30

2–

32

9

32

3

R. Jimborean / Economic Systems 37 (2013) 302–329324

Appendix D

Table 14Augmented Dickey–Fuller test for unit root.

Country Variable NEER CPI PPI IPI

Bulgaria Level 0.2857 0.5943 0.9044 0.3073

Difference 0.0000 0.0000 0.0000 0.0041

Czech Republic Level 0.8870 0.8250 0.8339 0.8650

Difference 0.0025 0.0007 0.0005 0.0061

Estonia Level 0.7432 0.5547 0.9915 0.9919

Difference 0.0022 0.0021 0.0022 0.0432

Hungary Level 0.2048 0.4013 0.9789 0.1483

Difference 0.0121 0.0310 0.0192 0.0563

Latvia Level 0.5171 0.9139 0.9791 0.9733

Difference 0.0575 0.0419 0.0594 0.0064

Lithuania Level 0.1072 0.9639 0.9761 0.2090

Difference 0.0012 0.0484 0.0005 0.0010

Poland Level 0.4750 0.6018 0.9718 0.1260

Difference 0.0072 0.0554 0.0128 0.0024

Romania Level 0.3198 0.1896 0.9833 –

Difference 0.0156 0.0000 0.0239 –

Slovakia Level 0.9469 0.2665 0.5855 0.3720

Difference 0.0611 0.0067 0.0213 0.0020

Slovenia Level 0.1409 0.3973 0.6107 –

Difference 0.0450 0.0699 0.0005 –

Notes: ADF null hypothesis: the series has a unit root. Observation: p-values reported. No IPI data available for Romania and

Slovenia. neer stands for nominal effective exchange rate.

Table 15Autocorrelation and heteroscedasticity tests.

Country Ljung–Box Q test Breusch–Pagan test

CPI PPI IPI CPI PPI IPI

Bulgaria 0.5357 0.9920 0.5912 0.0031 0.0585 0.0766

Czech Republic 0.8117 0.0949 0.8976 0.0000 0.3024 0.8082

Estonia 0.6035 0.8519 0.9345 0.8896 0.5639 0.0018

Hungary 0.1811 0.3234 0.0567 0.0414 0.0012 0.0287

Latvia 0.5240 0.2795 0.7591 0.1011 0.2331 0.4134

Lithuania 0.9874 0.7849 0.1552 0.0482 0.0131 0.7076

Poland 0.7881 0.3261 0.3815 0.0004 0.0723 0.0433

Romania 0.9385 0.6347 – 0.0876 0.0000 –

Slovakia 0.7180 0.6350 0.6173 0.0192 0.0021 0.9418

Slovenia 0.7605 0.0000 – 0.0041 0.1109 –

Notes: Ljung–Box Q test null hypothesis: the data are independently distributed (i.e. the correlations in the population from

which the sample is taken are 0, so that any observed correlations in the data result from randomness of the sampling process).

Alternative hypothesis: the data are not independently distributed. Breusch-Pagan test null hypothesis: the variance of

residuals is homogenous. Observation: no IPI data available for Romania and Slovenia.

Table 16Serial correlation test.

Country CPI PPI IPI

Bulgaria 0.6069 0.4654 0.5128

Czech Republic 0.2978 0.0071 0.0524

Estonia 0.2179 0.1613 0.0261

Hungary 0.1143 0.0691 0.7155

Latvia 0.6090 0.7882 0.8117

Lithuania 0.8448 0.0474 0.1903

Poland 0.5934 0.4690 0.4412

Romania 0.1234 0.2642 –

Slovakia 0.8121 0.2977 0.5895

Slovenia 0.1484 0.0098 –

Notes: Null hypothesis of the Breusch–Godfrey LM test: no serial correlation. Observation: no IPI data available for Romania and

Slovenia.

Table 17EG-ADF test for co-integration.

Country CPI PPI IPI

Bulgaria 0.3487 0.1023 0.2718

Czech Republic 0.1662 0.1078 0.1006

Estonia 0.1001 0.1121 0.4308

Hungary 0.1103 0.1011 0.8011

Latvia 0.3902 0.3048 0.1081

Lithuania 0.5883 0.1033 0.1147

Poland 0.1052 0.1364 0.1414

Romania 0.3877 0.3415 –

Slovakia 0.1439 0.6729 0.2558

Slovenia 0.6437 0.4288 –

Notes: The null hypothesis of the EG-ADF test: the variables are co-integrated. Observation: no IPI data available for Romania and

Slovenia.

R. Jimborean / Economic Systems 37 (2013) 302–329 325

R. Jimborean / Economic Systems 37 (2013) 302–329326

Appendix E

Fig. 1. Estimated consumer price elasticities, rolling sample with a window size of 30 quarters (end Q4/2011).

Fig. 2. Estimated producer price elasticities, rolling sample with a window size of 30 quarters (end Q4/2011).

R. Jimborean / Economic Systems 37 (2013) 302–329 327

Fig. 3. Estimated import price elasticities, rolling sample with a window size of 30 quarters (end Q4/2011).

R. Jimborean / Economic Systems 37 (2013) 302–329328

References

Arellano, M., Bond, S., 1991. Some tests of specification for panel data: Monte Carlo evidence and an application to employmentequations. Review of Economic Studies 58, 277–297.

Arellano, M., Bover, O., 1995. Another look at the instrumental-variable estimation of error-components models. Journal ofEconometrics 68, 29–51.

Bitans, M., 2004. Pass-through of exchange rates to domestic prices in East European countries and the role of economicenvironment. Bank of Latvia Working Paper 4, Riga.

Blundell, R.W., Bond, S., 1998. Initial conditions and moment restrictions in dynamic panel data models. Journal of Econometrics87, 115–143.

Brun-Aguerre, R., Fuertes, A.-M., Phylaktis, K., 2012. Country and time variation in import pass-through: is it driven by macroand micro factors? Journal of International Money and Finance 31, 818–844.

Burnstein, A., Eichenbaum, M., Rebelo, S., 2002. Why are rates of inflation so low after large devaluation? NBER Working Paper8748, Cambridge, MA.

Bussiere, M., Peltonen, T., 2008. Exchange rate pass-through in the global economy. The role of emerging market economies.European Central Bank Working Paper Series 951, Frankfurt/Main.

Campa, J., Goldberg, L., 2002. Exchange rate pass-through into import prices: a macro or micro phenomenon? NBER WorkingPaper 8934, Cambridge, MA.

Choudhri, E.U., Hakura, D.S., 2006. Exchange rate pass-through to domestic prices: does the inflationary environment matter?Journal of International Money and Finance 25, 614–639.

Cogley, T., Nason, J., 1995. Effects of the Hodrick–Prescott filter on trend and difference stationary time series. Journal ofEconomic Dynamics and Control 19, 253–278.

Cochrane, D., Orcutt, G.H., 1949. Application of least squares regression to relationships containing autocorrelated error terms.Journal of the American Statistical Association 44, 32–61.

R. Jimborean / Economic Systems 37 (2013) 302–329 329

Coricelli, F., Egert, B., MacDonald, R., 2006. Monetary transmission mechanism in Central and Eastern Europe: gliding on a windof change. BOFIT Discussion Papers 8, Helsinki.

Coricelli, F., Jazbec, B., Masten, I., 2004. Exchange rate pass-through in acceding countries: the role of exchange rate regimes.William Davidson Institute Working Paper Series 674, Ann Arbor, MI.

Corsetti, G., Dedola, L., Leduc, S., 2007. Optimal monetary policy and the sources of local currency price stability. CEPRDiscussion Paper 6557, London.

Darvas, Z., 2001. Exchange rate pass-through and real exchange rate in EU candidate countries. Deutsche BundesbankDiscussion Paper 10, Frankfurt/Main.

De Bandt, O., Banerjee, A., Kozluk, T., 2007. Measuring long run exchange rate pass-through. Economics: The Open-Access,Open-Assesment E-Journal 2, 2008-6, available at: http://www.economics-ejournal.org/economics/journalarticles/2008-6.

Devereux, B., Yetman, J., 2003. Price-setting and exchange rate pass-through: theory and evidence. In: Proceedings of theConference on Price Adjustments and Monetary Policy, Bank of Canada, Ottawa, ON, November, 2002, pp. 347–371.

Dornbush, R., 1987. Exchange rates and prices. American Economic Review 77, 93–106.Egert, B., 2011. Catching-up and inflation in Europe: Balassa Samuelson. Engel’s Law and other culprits. Economic Systems 35,

208–229.Egert, B., MacDonald, R., 2009. Monetary transmission mechanism in Central and Eastern Europe: surveying the surveyable.

Journal of Economic Surveys 23, 277–327.Engel, C., 2002. The responsiveness of consumer prices to exchange rates and the implications for exchange rate policy: a survey

of a few recent new open-economy macro models. NBER Working Paper 8725, Cambridge, MA.Engle, R.F., Granger, C.W.J., 1987. Co-integration and error correction: representation, estimation and testing. Econometrica 55,

251–276.Frankel, J.A., Parsley, D.C., Wei, S.J., 2005. Slow pass-through around the world: a new import for developing countries. NBER

Working Paper 11199, Cambridge, MA.Fuhrer, J.C., 2009. Inflation persistence. Federal Reserve Bank of Boston Working Paper 14, Boston, MA.Gali, J., Gertler, M., 1999. Inflation dynamics: a structural econometric model. Journal of Monetary Economics 44, 195–222.Gaulier, G., Lahreche-Revil, A., Mejean, I., 2008. Exchange rate pass-through at the product level. Canadian Journal of Economics

41, 425–449.Goldberg, P.K., Knetter, M., 1997. Goods prices and exchange rates: what have we learned? Journal of Economic Literature 35,

1243–1272.Guarda, P., 2002. Potential output and the output gap in Luxembourg: some alternative methods. Banque Centrale de

Luxembourg Working Paper No. 4, Luxembourg.Guay, A., St-Amant, P., 1996. Do mechanical filters provide a good approximation of business cycles? Bank of Canada Technical

Report 78, Ottawa, ON.Gueorguiev, N., 2003. Exchange rate pass-through in Romania. IMF Working Paper 130, Washington, DC.Hansen, L.P., 1982. Large sample properties of generalized method of moments estimators. Econometrica 50, 1029–1054.Harvey, A.C., Jaeger, A., 1993. Detrending, stylized facts and the business cycle. Journal of Applied Econometrics 8, 231–247.Hodrick, R.J., Prescott, E.C., 1997. Postwar US business cycles: an empirical investigation. Journal of Money, Credit and Banking

29, 1–16.Ilzetzki, E., Reinhart, C.M., Rogoff, K.S., 2008. Exchange rate arrangements entering the 21st century: which anchor will hold?

Mimeo, Harvard University, University of Maryland, Baltimore, MD, Cambridge, MA.Im, K.S., Pesaran, M.H., Shin, Y., 2003. Testing for unit roots in heterogeneous panels. Journal of Econometrics 115, 53–74. Earlier

version as unpublished Working Paper, Department of Applied Economics, University of Cambridge, 1997. Available at:In:http://www.econ.cam.ac.uk/faculty/pesaran/lm.pdf

Kaiser, R., Maravall, A., 1999. Estimation of the business cycle: a modified Hodrick–Prescott filter. Spanish Economic Review 1,175–206.

Kuijs, L., 2002. Monetary policy transmission mechanisms and inflation in the Slovak Republic. IMF Working Paper 80,Washington, DC.

Maddala, G.S., Wu, S., 1999. A comparative study of unit root tests with panel data and a new simple test. Oxford Bulletin ofEconomics and Statistics 61, 631–652.

McCarthy, J., 1999. Pass-through of exchange rates and import prices to domestic inflation in some industrialised economies.BIS Working Paper 79, Basel.

Mihaljek, D., Klau, M., 2001. A note on the pass-through from exchange rate and foreign price changes to inflation in selectedemerging market economies. BIS Papers 8, 69–81, Basel.

Ravn, M.O., Uhlig, H., 2002. On adjusting the Hodrick–Prescott filter for the frequency of observations. The Review of Economicsand Statistics 84, 371–375.

Roodman, D., 2008. A note on the theme of too many instruments. Oxford Bulletin of Economics and Statistics 71, 135–158.Takhtamanova, Y., 2008. Understanding changes in exchange rate pass-through. Federal Reserve Bank of San Francisco Working

Paper 13, San Francisco, CA.Tanaka, M., Young, C., 2008. The economics of global output gap measures. Bank of England Quarterly Bulletin (September),

299–305.Taylor, J., 2000. Low inflation pass-through and the pricing power of firms. European Economic Review 44, 1389–1408.Windmeijer, F., 2005. A finite sample correction for the variance of linear efficient two-step GMM estimators. Journal of

Econometrics 126, 25–51.