The impact of the USD/EUR exchange rate on inflation in the Central and East European countries.

Jankov, Ljubinko ; Krznar, Ivo ; Kunovac, Davor 等

INTRODUCTION

During the last few years, there has been growing empirical support for the idea that external factors might have a leading role in explaining business cycles in small open economies. (1) In particular, import prices and exchange rates have been the focus of empirical studies trying to determine the main sources of inflation in small open economies. This paper suggests that the USD/EUR exchange rate might be considered as an additional important determinant of inflation in the Central and East European countries (CEEC), an idea that has not been explicitly analysed in previous studies.

Our motivation comes from the empirical evidence shown in Figure 1. We see a strong correlation between the first principal component of the CEEC's annual consumer price inflation rates and the annual change in the USD/EUR exchange rate. Despite the existence of quite different monetary and exchange rate regimes in the CEEC, it seems that there are some similarities in their inflation paths that might be accounted for by USD/EUR exchange rate fluctuations.

Most previous studies of pass-through in the CEEC focus on effective exchange rates and assume that the individual country can influence effective exchange rates through monetary policy (for a survey, see Egert and MacDonald, 2006). Contrary to that, we distinguish between the exchange rate of the domestic currency against the euro and the USD/EUR exchange rate and analyse which portion of the variation in inflation in the CEEC can be attributed to the USD/EUR exchange rate, as an external shock. In addition, we study to what extent USD/EUR exchange rate shocks influence inflation. Finally, we attribute the different levels of impact of the USD/EUR exchange rate on inflation among the CEEC to the different exchange rate regimes.

To measure the impact of the USD/EUR exchange rate on domestic producers and consumer inflation across countries, we employ the empirical model of pricing along a distribution chain, as in McCarthy (2007). The advantage of this model is that it has a vector autoregression (VAR) representation that allows us to trace the impact of exchange rate fluctuations on inflation at each stage along the distribution chain (importers, producers, and consumers). While McCarthy (2007) studies a large open economy that can influence external factors, we adopt a small country assumption where domestic variables cannot influence external variables. In other words, we represent the model of pricing along the distribution chain in the CEEC with a VAR model with block exogeneity restrictions (external variables) in the spirit of Cushman and Zha (1997). (2) The imposition of block exogeneity seems a reasonable way to identify foreign shocks from the perspective of the small open economy.

[FIGURE 1 OMITTED]

Our empirical exercise shows that the USD/EUR exchange rate accounts for the largest share of inflation volatility in the CEEC with stable exchange rates of the domestic currency against the euro. Furthermore, the extent of the USD/EUR exchange rate's influence on inflation in the CEEC is the largest in the economies with stable exchange rate regimes. These results might be important in the context of the price stability requirement of the Maastricht Criteria: in addition to the internal challenge of keeping low inflation and dealing with the difficulties of the price convergence process, the applicant countries could face problems beyond their influence. Given that most of the CEEC peg their currencies to the euro, (3) either because of the conditions of the ERM-II or because of their domestic issues (euroisation in particular), and taking into account the high volatility of the USD/EUR exchange rate, our findings suggest that the CEEC under a fixed or heavily managed exchange rate might face substantial problems in achieving a high degree of price stability.

The decision to include the USD/EUR exchange rate as a separate external factor is motivated by the monetary and exchange rate regimes in the CEEC. These countries are primarily concerned with fluctuations of their exchange rate against the euro: while all countries (will) have to participate in the ERM-II, some countries use the exchange rate against the euro (previously the Deutsche Mark) to reduce imported inflation and anchor inflation expectations. Since the USD/EUR exchange rate is determined on the global financial market, neither an individual country is able to influence it nor can it influence the world prices. Hence, it cannot simultaneously manage both its bilateral exchange rate against the euro and against the dollar. For this reason, we refrain from using the effective exchange rate, which combines the managed exchange rate against the euro and the exchange rate against the dollar. (4) Therefore, for countries with heavily managed exchange rates to the euro, the USD/EUR exchange rate in fact represents an external shock. By focussing on the stability of their domestic currencies against the euro, the CEEC effectively reduce the exchange rate pass-through of goods priced in euros to domestic inflation. However, since a number of commodities are priced in dollars, there is still a pass-through from the dollar, which is amplified by the USD/EUR exchange rate fluctuations.

This paper is organised as follows. The second section illustrates the model of pricing along the distribution chain applied to the CEEC. The third section describes the VAR methodology with block exogenous restrictions. The fourth section describes the data used and provides a basic description of monetary and exchange rate regimes in the CEEC. Results are presented in the fifth section, along with a discussion of the impact of the USD/EUR on disaggregated data to confirm our understanding of the transmission channel. The special case of the regime change in Lithuania, where the currency peg was changed from the dollar to the euro, is also presented. The final section concludes.

THE MODEL OF PRICING ALONG THE DISTRIBUTION CHAIN

Our model of pricing includes two stages due to the unavailability of import price data for many CEEC. (5) The stages correspond to producer price inflation and consumer price inflation, each with several components. In each stage, inflation is a function of the expected inflation based on the available information in previous period and contemporaneous shocks: a supply shock, a demand shock, an exchange rate shock (either a USD/EUR exchange rate shock in the case of heavily managed exchange rates of the domestic currency against the euro, or both a USD/EUR exchange rate shock and a shock to the exchange rate of the domestic currency against the euro in the case of a looser exchange rate regime (6)), a shock to inflation at the previous stage of the distribution chain as well as its own shock.

The supply shock is identified from the world primary commodity prices expressed in dollars. (7) The USD/EUR exchange rate shock is identified from the behaviour of the USD/EUR exchange rate after taking the supply shock into account. These two shocks make the exogenous block that is unaffected by the domestic business cycle. (8) In contrast to previous studies that combine the two external shocks to save degrees of freedom (see, eg, Mackowiak, 2007), our intention is to analyse the impact of each external factor separately to see which of the two has the dominant role. The demand shock is identified from the dynamics of the output gap after taking into account the supply shock and the exchange rate shock. The shock to the exchange rate of the domestic currency against the euro (in countries with a looser exchange rate regime) is identified from the behaviour of the exchange rate of the domestic currency against the euro after taking the supply shock, the USD/EUR exchange rate shock and the demand shock into account. The last two shocks (the demand shock and the domestic currency shock), together with the dynamics of producer and consumer prices, comprise the domestic block, which can be affected by the exogenous block.

The structure of the model suggests that it can be cast into a recursive VAR framework estimation, as described in the next section.

METHODOLOGY--VAR ANALYSIS WITH BLOCK EXOGENEITY RESTRICTIONS

In this section, we describe the VAR framework that is used to identify the shocks in the model and their impact on prices.

Let [y.sub.1] be an [n.sub.1] dimensional vector of external variables. Let [y.sub.2] be an [n.sub.2] dimensional vector of domestic (small open economy) variables. We combine both vectors in y = [[y.sub.1], [y.sub.2]]'. Now consider a dynamic system of equations:

[p.summation over (s=0)] [A.sub.s][y.sub.t-s] = [[epsilon].sub.t] (1)

where [A.sub.0] is the (regular) contemporaneous matrix of coefficients, [{[[epsilon].sub.t]}.sup.[infinity].sub.t = 0] are i.i.d. random vectors with multivariate normal distribution MVN(O, I), and [A.sub.j] are block lower triangular matrices of dimension ([n.sub.1] + [n.sub.2]) x ([n.sub.1] + [n.sub.2]), which have the following form:

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]

Submatrices [A.sup.j.sub.lk] are of [n.sub.l] x [n.sub.k] dimension for l, k = 1, ..., 2 and j = 1, ..., p.

The form of [A.sub.j] assumes block exogeneity restrictions that represent the underlaying idea that foreign shocks can affect the small open economy, but not the other way around.

After multiplication by [A.sup.-1.sub.0] equation (1) yields a corresponding reduced-form VAR model:

[y.sub.t] = [p.summation over (s=1)] [B.sub.s][y.sub.t-s] + [[eta].sub.t] (2)

where [A.sup.-1.sub.0] [[epsilon].sub.t] = [[eta].sub.t]: MVN(0, [[SIGMA]sub.[eta]]) and [B.sub.j] = [A.sup.-1.sub.0] [A.sub.j] for for j = 0, ..., p. It can be shown (see Lutkepohl (2005)) that matrices of coefficients [B.sub.s] inherit (9) block exogeneity form so that

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]

Note that this is equivalent to the statement that the domestic block does not Granger-cause the foreign block, that is, that the domestic block does not help to forecast (in the MSE sense) the variables in the foreign block. This is a standard and testable assumption when modelling the small open economy's reaction to foreign shocks.

Given the autoregressive representations (1) and (2), we can derive the corresponding moving average representations:

[y.sub.t] = [([A.sub.0] + [A.sub.1]L + ... + [[A.sub.p][L.sup.p]).sup.-1] [[epsilon].sub.t]

=([D.sub.0] + [D.sub.1]L + [D.sub.2][L.sup.2] + ... +) [[epsilon].sub.t] = D(L)[[epsilon].sub.t] (3)

and

[[y.sub.t] = (I - [B.sub.1]L - ... - [B.sub.p] [L.sup.P]].sup.-1] [[eta].sub.t]

=(I + [C.sub.1]L + [C.sub.2][L.sup.2] + ... +) [[eta].sub.t] = C(L) [[eta].sub.t]

Given the reduced-form residuals [[eta].sub.t] with the corresponding estimate [[SIGMA].[eta]] (with n(n + 1)/2 unique elements) and coefficient matrices [B.sub.i] and [C.sub.i], one can recover impulse responses [D.sub.i], subject to normalisation condition [[SIGMA].sub.[eta]] = [A.sup.-1.sub.0][A'.sup.-1.sub.0] In order to identify [A.sub.0], we need to impose at least n(n-1)/2 additional restrictions. For that purpose, let us define [[epsilon].sub.t] = [A.sub.0][[eta].sub.t], where [A.sub.0] is a lower triangular Cholesky factor (10) of noise covariance matrix [[SIGMA].sub.[eta]]. It follows that E[[epsilon].sub.t][[epsilon]'.sub.t]] = E[[A.sub.0][[eta].sub.t][[eta]'.sub.t] [A'.sub.0] = [A.sub.0][[SIGMA].sub.[eta]] [A'.sub.0]=I and orthogonality holds. For alternative types of identification, see Cushman and Zha (1997) and Mackowiak (2007).

The reduced-form VAR model was estimated applying the feasible least-squares estimator. Details concerning the estimation and structural analysis of VAR processes of these types of models can be found in Lutkepohl (2005).

DATA

The data were taken from the IMF's International Financial Statistics database. For the external block, which is the same for all countries, we use the IMF's Primary Commodity Price Index ([WCP.sub.t]) as a measure of world prices and the USD/EUR exchange rate (USD/[EUR.sub.t]). (11) The domestic block consists of the output gap, defined as the deviation of GDP (in constant prices) from its trend [(Gap.sub.t)], (12) the exchange rate of domestic currency against the euro (DC/[EUR.sub.t]), the producer price index [(PPI.sub.t)], and the consumer price index [(CPI.sub.t)] for each country. DC/[EUR.sub.t] was calculated as a product of the domestic currency against the US dollar rate and the USD/EUR rate. All the price and exchange rate data are in quarterly averages (prices, exchange rate) from 1998 (first quarter) to 2006 (third quarter).

The most serious problem with the CEEC data is structural breaks. The first kind of structural break pertains to the undergoing transition process that could affect parameter stability. In our analysis, we bracket this type of structural break as we analyse the late phase of the transition. However, the second kind of structural break--changes in monetary and exchange rate regime--presents more serious problems because it might affect the price formation process that we analyse. As shown in Table 1, an exchange rate regime change occurred in more than half of the CEEC in our sample.

Although we do not model the determinants of the regime changes, we group countries according to different monetary and exchange rate regimes in two ways: by type of regime currently in place and by the severity of the regime change those countries undertook during the period under study.

When looking at existing monetary regimes, we distinguish between exchange rate targeters and inflation targeters. Exchange rate targeters include countries with fixed exchange rate against the euro or those with small oscillations against the euro. The extreme example is Slovenia, which adopted the euro at the beginning of 2007. There are two currency boards (Bulgaria and Estonia), a fixed exchange rate (Lithuania), a country with a tight (1%) exchange rate band (Latvia), and a managed floater (Croatia). Those countries seem to be perfect candidates for our analysis because the USD/EUR exchange rate corresponds to their exchange rate against the dollar. The other group consists of the inflation targeters: the Czech Republic, Hungary, Poland, Romania, and Slovak Republic. However, there are significant differences among them in terms of exchange rate stability against the euro (see Table 2).

Unfortunately, some of the countries recently undertook significant regime change that altered the price formation process and therefore we cannot analyse them using the VAR. On one extreme are Slovenia and Romania, which in their attempt to achieve real exchange rate stability have gone through a gradual disinflation and depreciation before achieving price stability. A serious policy change from the perspective of our analysis occurred in two Baltic countries that changed the peg currency. The most interesting case is Lithuania, which repegged from the dollar to the euro in February 2002. This shift should lead to a change of sign in the estimated USD/EUR exchange rate pass-through coefficients. A similar case is Latvia, which repegged from the SDR to the euro in February 2004. Because of the estimation problems in the cases of Slovenia and Romania as a result of regime shifts, we exclude these two countries from our analysis. Furthermore, due to the short sample, we are unable to model the regime change in Latvia and Lithuania and therefore we also exclude them from our analysis.

The most interesting countries for our analysis are those with fixed (or managed) exchange rate to the Deutsche Mark prior to 1999 and to the euro afterwards (Bulgaria, Croatia, and Estonia). We compare their results with countries that moved from more managed to less managed regimes--usually in the form of inflation targeting (Czech Republic, Hungary, Poland, and Slovak Republic)--before or during the period under study.

Prior to the estimation, we test the block exogeneity restrictions on the constrained VAR specification in order to find out whether such constraints are supported by the actual CEEC's data. We have already mentioned that block exogeneity is equivalent to the hypothesis that the domestic block does not Granger-cause the foreign block. Given Wald test's P-values from Table 3, we conclude that a priori exogenous restrictions in the VAR specification have been well chosen.

THE IMPACT OF THE USD/EUR EXCHANGE RATE ON INFLATION IN THE CEEC

Owing to the short data series available and the low power of unit root tests, (13) we estimated the model in the first differences (as in McCarthy, 2007). This way we studied only the short-term effects, while possible long-run relations were not identified.

When estimating the VAR, we examined several different setups. Most importantly, we tried to estimate the VAR with and without the exchange rate of the domestic currency against the euro. The reason for this is that in some cases (Bulgaria, Croatia, and Estonia) the oscillations in the domestic currency against the euro were too small to have any material impact on inflation. Although the VAR model with the domestic currency produced impulse responses with the expected direction, the results were not statistically significant. To save degrees of freedom, we removed the domestic currency from the VAR model for Bulgaria, Croatia, and Estonia. The VAR lag length of two quarters was a compromise between the length of the series and the time needed for the exchange rate shock to manifest itself on prices. After checking for all the necessary diagnostics (Table 4), we estimated (2) for three CEEC (Bulgaria, Croatia, and Estonia) with the exogenous block [y.sub.1t] = [[WP.sub.t], USD/[EUR.sub.t]]' and the domestic block [y.sub.2t] = [[Gap.sub.t], [PPI.sub.t], [CPI.sub.t]]'. For the inflation targeters (Czech Republic, Hungary, Poland, and Slovak Republic), a somewhat richer specification with the domestic currency was used ([y.sub.2t] = [[Gap.sub.t], DC/[EUR.sub.t], [PPI.sub.t], [CPI.sub.t]]').

The variance decomposition of the specified VAR model presented in Tables 5 and 6 shows that external shocks have a large impact on the variation of domestic variables. With a 2-year horizon (eight quarters ahead), shocks in world commodity prices and the USD/EUR on average account for about half of the variation of the PPI (51%) and the CPI (42%). (14) The USD/ EUR seems to cause more variation in consumer prices than the world commodity prices, while the world commodity prices seem to have the more prominent role in the determination of the producer prices.

The variance decomposition indicates that external shocks account for a large share of price volatility (both PPI and CPI) in all countries, regardless of the policy regime. This is, however, due to the movement of the world commodities prices. The impact of the USD/EUR in explaining inflation variance is greater in countries with a stable exchange rate against the euro (Bulgaria, Croatia, and Estonia), where it explains 28 % of the variance in CPI and 18% of the variance in PPI. Countries that retain a higher degree of independent monetary policy seem to be able to use it to protect themselves from such shocks, as the USD/EUR fluctuations explain a smaller share of price variance (8% of CPI and 13% of PPI).

The size of the impact of different shocks is measured using the impulse responses for each country (Table 7). The impulse responses show that the shock in the world commodity prices affects domestic variables through various channels. Producer costs (PPI), and to some extent consumer prices, are immediately affected. With a time lag, the producer price shock is further transmitted to consumer prices in the form of higher costs. A similar channel also works for the USD/EUR exchange rate shock: appreciation of the euro against the dollar instantly reduces producer costs and to a lesser extent consumer prices, which suggests that prices of goods that represent a significant share of the consumer basket react strongly to movements in the world market. This is also confirmed by the disaggregated data (see the next section). Here, an important channel goes from the producer costs to the prices of consumer goods, which is in line with theory and the logic that the USD/EUR exchange rate to a large extent works as an important cost factor. Since we use quarterly frequency, it is possible that there is an immediate effect of the PPI on the CPI.

Directions of the impulses are as expected for most countries. Only one (Slovakia) shows a wrong sign of the impact of the USD/EUR shock on the CPI. In all other countries, euro appreciation against the dollar leads to a drop in prices. The size varies: 2 years after, the shock ranges from -0.08 for Poland to -0.3 for Bulgaria, with an average of -0.14. Again, larger effects are found in countries with stable exchange rates against the euro (-0.22 versus -0.09). This result is partially supported by the impact of the domestic currency shock on inflation. Again, for all countries it has expected sign and ranges from 0.10 for the Czech Republic to 0.56 for Hungary.

The result that countries with stable exchange rates against the euro are most susceptible to USD/EUR fluctuations is expected, since they are unable to compensate for this change in import cost through the domestic exchange rate.

Evidence from the disaggregated price data

Price movements of the individual items (categories) in the consumer basket can increase our understanding of how the USD/EUR influences domestic inflation. For that purpose, we calculate simple correlations between the annual inflation of individual components in the consumer basket and the annual rate of change of the USD/EUR exchange rate. We expect that there is a (strong) negative correlation between the USD/EUR rate and tradable products whose prices are expressed in dollars (and whose prices became cheaper when the euro appreciates against the dollar).

We use Eurostat data collected for the HICP, aggregated into categories. Thus, it is sometimes difficult to distinguish between imported and the domestically produced goods and services that make up an individual consumption category (eg, recreation and culture). For that reason, we only report correlations for the main categories.

Examination of the disaggregated price data available for the CEEC shows that there is a negative correlation between movement of the USD/EUR and prices of most consumer goods and services (averaged across CEEC), as shown in Figure 2. The strongest negative correlation is present for goods and services, in the group Transport and Recreation and culture, in which both have a large share of imported goods. Correlations are weaker in groups with larger share of domestic inputs such as Food, Housing and Restaurants and hotels.

A natural experiment--the case of Lithuania

Although the lack of data prevents us from conducting a proper econometric analysis, countries that changed their exchange rate policy represent natural experiments for our hypothesis. The prime candidate is Lithuania, which changed its peg from the dollar to the euro in February 2002. As shown in Figure 3, we find the expected change in the direction (sign) of correlation between the USD/EUR exchange rate from positive to negative. The depreciation of the dollar against the euro seems to have contributed to the deflation Lithuania faced after the policy shift. The euro appreciation in 2006, however, did not have an immediate effect on the Lithuanian CPI due to domestic factors (liberalisation of administrative prices in particular), and Lithuania just missed the inflation criterion for joining the Eurozone.

[FIGURE 2 OMITTED]

[FIGURE 3 OMITTED]

CONCLUSION

Our empirical analysis shows that in the countries with stable exchange rates against the euro, fluctuations of the USD/EUR exchange rate might be one of the leading factors responsible for inflation variation. This might be because the stable exchange rate managed to bring down the major external sources of inflation coming from euro-denominated goods, as well as anchoring domestic inflation expectations. Given recent large fluctuations of the USD/ EUR exchange rate, with no additional monetary instruments to contain their effects, in the stable exchange rate regimes the largest impact on price volatility comes from abroad, although the actual pass-through of the USD/ EUR is similar in size in all CEEC, regardless of the policy regime. Therefore, our findings suggest that in the case of a significant appreciation of the dollar in the run-up to the Eurozone, in countries with stable exchange rate a possible inflationary (external) shock needs to be dealt with by economic policies other than monetary policy. The 1.5% buffer in the Maastricht criteria might not be enough to accommodate rising inflation in the case of a larger dollar appreciation.

Acknowledgements

We are grateful to Nikola Bokan, Evan Kraft, and Ana Martinis for helpful comments. The views expressed in this paper are ours and do not necessarily reflect the view of the Croatian National Bank.

REFERENCES

Canova, F. 2005: The transmission of US shocks to Latin America. Journal of Applied Econometrics 20: 229-251.

Cushman, DO and Zha, T. 1997: Identifying monetary policy in a small open economy under flexible exchange rates. Journal of Monetary Economics 39: 433-448.

DeJong, DN and Whiteman, CH. 1992: Integration versus trend stationarity in time series. Econometrica 60: 423-433.

Egert, B and MacDonald, R. 2006: Monetary transmission mechanism in transition economies: Surveying the surveyable. MNB Working papers 2006/5, Magyar Nemzeti Bank (The Central Bank of Hungary).

Jones, G and Kutan, AM. 2004: Exchange rate management strategies in the accession countries: The case of Hungary. Comparative Economic Studies 46: 23-44.

Leybourne, SJ and McCabe, BPM. 1994: A consistent test for a unit root. Journal of Business & Economic Statistics 12: 157-166.

Lutkepohl, H. 2005: New introduction to multiple time series analysis. Springer-Verlag: Berlin.

Mackowiak, B. 2006: How much of the macroeconomic variation in Eastern Europe is attributable to external shocks? Comparative Economic Studies 48: 523-544.

Mackowiak, B. 2007: External shocks, US monetary policy and macroeconomic fluctuations in emerging markets. Journal of Monetary Economics 54: 2512-2520.

McCarthy, J. 2007: Pass-through of exchange rates and import prices to domestic inflation in some industrialized economies. Eastern Economic Review 33: 511-537.

Obstfeld, M and Rogoff, K. 2000: New directions for stochastic open economy models. Journal of International Economics 50: 117-153.

Schwert, GW. 1989: Tests for unit roots: A Monte Carlo investigation. Journal of Business & Economic Statistics 7: 147-159.

LJUBINKO JANKOV [1], IVO KRZNAR [1,2], DAVOR KUNOVAC [1] & MAROJE LANG [1]

[1] Croatian National Bank, Trg hrvatskih velikana 3, Zagreb 10002, Croatia. E-mail: ivo.krznar@hnb.hr

[2] Zagreb School of Economics and Management, Jordanovac 110, Zagreb 10000, Croatia.

(1) See, for example, Canova (2005), Cushman and Zha (1997), Jones and Kutan (2004), and Mackowiak (2006, 2007). Most of the empirical research on this topic was a reaction to the old Keynesian literature that was (unsuccessfully) explaining inflation solely as a domestic phenomenon in a closed economy. The above-mentioned empirical research supports a new Keynesian theory of a small open economy that, in addition to domestic factors, takes into account external factors in explaining prices. See Obstfeld and Rogoff (2000) for a basic model of a small open economy where the overall price index depends on domestic prices, import prices, and the exchange rate.

(2) Our approach is similar to Mackowiak (2007), who measured the impact of external shocks on some of the CEEC. He found that most of the volatility of main macrovariables comes from abroad.

(3) Exchange Rate Mechanism II (ERM-II) imposes [+ or -] 15 % fluctuations, while some countries can adopt smaller bands. Crawling pegs and pegs to currencies other than the euro are inconsistent with the ERM-II.

(4) Given that the CEEC primarily control their exchange rate against the euro, most of the variation of their effective exchange rate comes from the impact of a more volatile nominal exchange rate against the dollar, rather than a more stable price of the euro.

(5) McCarthy's (2007) model of pricing along a distribution chain includes all three stages.

(6) The value of a country's domestic currency can be expressed bilaterally against any other currency. Thus, we could include in the VAR both exchange rate of the domestic currency (DC) against the dollar and against the euro. Both bilateral rates would in this case be a part of the VAR's domestic block. However, although a country can influence any bilateral rate, the ratio of such bilateral rates is exogenously given by the USD/EUR exchange rate, which is set on the international financial market ((DC/EUR)/(DC/USD) = USD/EUR). Since all CEEC are pegged to the euro either directly or through the ERM-II, we focus on the bilateral exchange rate against the euro, and take the USD/EUR exchange rate as given.

(7) Despite the growing international role of the euro, prices of most tradables, especially commodities, are formed in dollars. An actual transaction may take place in any currency, even though the price is set in dollars, which limits the potential use of the information about the invoicing currency for determining the role of foreign currencies in a country's trade. For that reason, and in the absence of information about individual countries' import prices, we use the world commodity prices (IMF) expressed in dollars in order to model the import price inflation.

(8) Because of shortness of data and unavailability of some of the series, we were forced to adopt more a parsimonious approach by reducing the number of external variables. By focusing on the (indirect) exchange rate pass-through as a model for describing inflation dynamics, we dismiss a number of other potential external shocks that could also affect an economy (eg, foreign interest rates or foreign demand shock). However, it seems that a number of shocks are mutually correlated (eg, GDP gap in Germany, interest rate in the Euro zone, and the USD/EUR exchange rate) and the model can be reduced to save degrees of freedom from already short series for countries under study.

(9) Lower triangularity is also inherited in the MA([infinity]) representation, which implies that there is no response from the foreign variables to the domestic shocks. See Lutkepohl (2005) for details.

(10) [A.sub.0] is a lower triangular matrix such that [A.sup.-1.sub.0] [([A.sup.-1.sub.0])]' = [[SIGMA].sub.[eta]]. Such decomposition always exists for a symmetric and positive-definite matrix. It can be shown that every covariance matrix is symmetric and positive-definite.

(11) Prior to the introduction of the euro, we use the USD/DEM exchange rate and transform it into the USD/EUR using the DEM/EUR conversion rate, since the Deutsche Mark was the most important currency in the CEEC.

(12) Real GDP data are not available for Bulgaria and Romania. We use industrial production (deflated using the CPI) instead.

(13) For the evidence on the low power of unit root, see, for example, Schwert (1989), DeJong and Whiteman (1992), or Leybourne and McCabe (1994).

(14) We have also estimated a similar VAR (as in Mackowiak, 2006) with the world prices denominated in euro, and therefore the USD/EUR rate has been excluded from this specification. Results were similar as in Tables 5 and 6.

Table 1: Monetary and exchange rate regimes and inflation in CEECs

 Changes in monetary regime
 Monetary regime since 1998

Bulgaria Currency board --
Croatia Managed floating --
Estonia Currency board --
Latvia Peg to euro [+ or -]1% 2004: repegged its
 currency from SDR to EUR
Lithuania Currency board 2002: repegged its
 currency from USD to EUR
Slovenia Euro 2007: adopted euro;
 previously: managed
 floating
Czech Republic Inflation targeting --
Hungary Inflation targeting --
Poland Inflation targeting 2001: changed from managed
 to independent floating
Romania Inflation targeting 2001: changed from managed
 float to crawling bands
Slovak Republic Inflation targeting Previously: managed
 floating

Table 2: Consumer price index/exchange rates correlations and
coefficients of variation

 Bg Ee CZ Hr Hu Lv

Correlations
 CPI-(DC/EUR) -0.10 0.03 -0.26 -0.13 -0.04 -0.01
 CPI-(EUR/USD) 0.38 0.40 0.07 0.58 0.29 0.11

Coefficient of
 variation
 DC/EUR 0.00 0.00 0.08 0.02 0.03 0.09
 DC/USD 0.14 0.14 0.18 0.14 0.14 0.05

 Lt PL Ro Sk Si

Correlations
 CPI-(DC/EUR) -0.08 -0.17 0.66 0.00 0.62
 CPI-(EUR/USD) 0.29 0.52 0.41 0.03 0.25

Coefficient of
 variation
 DC/EUR 0.10 0.07 0.34 0.06 0.09
 DC/USD 0.17 0.11 0.13 0.18 0.13

Note: DC, domestic currency. Bulgaria (Bg), Croatia (Hr), Czech
Republic (Cz), Estonia (Ee), Hungary (Hu), Latvia (Lv),
Lithuania (Lt), Poland (P[), Romania (Ro), Slovak Republic (Sk), and
Slovenia (Si).

Table 3: Null hypothesis: domestic block does not Granger-cause
foreign block

 Bg Hr Cz Ee Hu Pl Sk

P-value 0.07 0.92 0.22 0.12 0.18 0.49 0.40

Note: Bulgaria (Bg), Croatia (Hr), Czech Republic (Cz), Estonia (Ee),
Hungary (Hu), Poland (Po), and Slovak Republic (Sk).

Table 4: Portmanteau test for autocorrelation (lag = 12, no
autocorrelation under the null hypothesis) and stability conditions

 Bg Hr Cz Ee Hu PL Sk

Portmanteau test
 (P-values) 0.08 0.57 0.1 0.1 0.12 0.08 0.04
Root's modulus
 (minimum) 1.41 1.44 1.1 1.17 1.11 1.12 1.24

In this table, we provide results from Portmanteau test for
autocorrelation. In addition, we report the minimum modulus root from
determinantal polynomial det(I--[A.sub.1]z-- ... --[A.sub.P]
[z.sup.P]), Aj denoting reduced form VAR coefficient matrices. The
VAR process is stable if this polynomial has no roots in or on the
complex unit circle (see Lutkepohl, 2005), sufficient condition for
stability is that the minimal modulus is greater than unity. Bulgaria
(Bg), Croatia (Hr), Czech Republic (Cz), Estonia (Ee), Hungary (Hu),
Poland (Pl), and Slovak Republic (Sk).

Table 5: PPI variance decomposition

 Quarters WPC USD/EUR External
 ahead shocks

Bulgaria t+1 0.70 0.00 0.70
 t+8 0.65 0.10 0.75
Croatia t+1 0.21 0.17 0.38
 t+8 0.41 0.19 0.60
Estonia t+1 0.07 0.14 0.21
 t+8 0.17 0.26 0.43
Czech Republic t+1 0.49 0.00 0.49
 t+8 0.68 0.02 0.70
Hungary t+1 0.00 0.04 0.04
 t+8 0.08 0.27 0.35
Poland t+1 0.36 0.00 0.36
 t+8 0.33 0.03 0.36
Slovak Republic t+1 0.10 0.02 0.12
 t+8 0.16 0.20 0.36

Exchange rate fixers t+1 0.33 0.10 0.43
 t+8 0.41 0.18 0.59
Inflation targeters t+1 0.24 0.02 0.25
 t+8 0.31 0.13 0.44

Average t+1 0.28 0.05 0.33
 t+8 0.35 0.15 0.51

 Quarters
 ahead Gap DC/EUR PPI CPI

Bulgaria t+1 0.00 -- 0.30 0.00
 t+8 0.06 0.19 0.01
Croatia t+1 0.01 -- 0.61 0.00
 t+8 0.03 -- 0.35 0.01
Estonia t+1 0.02 0.77 0.00
 t+8 0.02 0.50 0.05
Czech Republic t+1 0.04 0.01 0.47 0.00
 t+8 0.12 0.05 0.09 0.04
Hungary t+1 0.03 0.37 0.56 0.00
 t+8 0.18 0.18 0.26 0.04
Poland t+1 0.01 0.27 0.35 0.00
 t+8 0.04 0.29 0.21 0.10
Slovak Republic t+1 0.01 0.00 0.87 0.00
 t+8 0.04 0.04 0.56 0.01

Exchange rate fixers t+1 0.01 -- 0.56 0.00
 t+8 0.04 0.35 0.01
Inflation targeters t+1 0.02 0.16 0.56 0.00
 t+8 0.10 0.14 0.28 0.05

Average t+1 0.02 -- 0.56 0.00
 t+8 0.07 -- 0.31 0.04

Table 6: CPI variance decomposition

 Quarters WPC USD/EUR External
 ahead shocks

Bulgaria t+1 0.14 0.16 0.30
 t+8 0.17 0.23 0.40
Croatia t+1 0.24 0.42 0.66
 t+8 0.32 0.33 0.65
Estonia t+1 0.33 0.23 0.56
 t+8 0.23 0.29 0.52
Czech Republic t+1 0.03 0.06 0.09
 t+8 0.56 0.06 0.62
Hungary t+1 0.13 0.00 0.13
 t+8 0.08 0.19 0.27
Poland t+1 0.32 0.12 0.44
 t+8 0.27 0.04 0.31
Slovak Republic t+1 0.00 0.06 0.06
 t+8 0.14 0.04 0.18

Exchange rate fixers t+1 0.24 0.27 0.51
 t+8 0.24 0.28 0.52
Inflation targeters t+1 0.12 0.06 0.18
 t+8 0.26 0.08 0.35

Average t+1 0.17 0.15 0.32
 t+8 0.25 0.17 0.42

 Quarters
 ahead Gap DC/EUR PPI CPI

Bulgaria t+1 0.06 -- 0.07 0.57
 t+8 0.06 -- 0.09 0.45
Croatia t+1 0.00 -- 0.01 0.32
 t+8 0.06 0.13 0.15
Estonia t+1 0.00 0.10 0.33
 t+8 0.06 -- 0.24 0.19
Czech Republic t+1 0.07 0.03 0.02 0.80
 t+8 0.15 0.03 0.02 0.18
Hungary t+1 0.25 0.08 0.01 0.53
 t+8 0.24 0.21 0.11 0.18
Poland t+1 0.00 0.08 0.03 0.45
 t+8 0.01 0.21 0.07 0.39
Slovak Republic t+1 0.01 0.00 0.69 0.23
 t+8 0.04 0.18 0.45 0.16

Exchange rate fixers t+1 0.02 -- 0.06 0.41
 t+8 0.06 -- 0.15 0.26
Inflation targeters t+1 0.08 0.05 0.19 0.50
 t+8 0.11 0.16 0.16 0.23

Average t+1 0.06 -- 0.13 0.46
 t+8 0.09 -- 0.16 0.24

Table 7: CPI response to one unit residual shock

Impulse Bg Hr Ee Cz Hu

WPC t+1 0.05 * 0.01 0.06 * 0.04 * 0.03 *
 t+4 0.10 * 0.07 ** 0.07 * 0.14 ** 0.01
 t+8 0.15 * 0.08 ** 0.11 * 0.30 ** 0.02

USD/EUR t+1 0.16 ** -0.08 ** -0.12 ** -0.05 * 0.02
 t+4 -0.30 ** -0.13 ** -0.22 ** 0.07 0.05
 t+8 -0.32 ** 0.13 ** 0.20 ** -0.14 -0.18

Gap t+1 -0.03 0.01 0 -0.55 0.57 *
 t+4 0 0.19 ** 0.33 * 0.10 0.97 *
 t+8 -0.01 0.20 ** 0.22 * 2.22 ** 0.34

DC/EUR t+1 -- -- -- -0.01 0.19 *
 t+4 -- -- -- 0.1 0.61 **
 t+8 -- -- -- 0.1 0.56 **

PPI t+1 0.33 ** 0 0.50 ** 0.16 0.21
 t+4 0.44 ** 0.09 * 1.55 ** 0.43 * 0.75 *
 t+8 0.43 ** 0.11 * 1.39 ** 0.43 * 0.23

CPI t+1 0.52 ** 0.53 ** 0.56 ** 1.09 ** 1.06 **
 t+4 0.64 ** 0.65 ** 0.32 ** 1.32 ** 1.01 **
 t+8 0.58 ** 0.65 ** 0.06 1.41 ** 0.49 *

Impulse Pl Sk Fix. Target. Av.

WPC t+1 0.07 * 0.06 0.04 0.05 0.05
 t+4 0.15 * -0.01 0.08 0.07 0.07
 t+8 0.20 * -0.01 0.11 0.13 0.12

USD/EUR t+1 -0.05 * 0.06 -0.12 -0.01 0.05
 t+4 -0.07 0.08 -0.22 -0.03 -0.11
 t+8 -0.08 0.06 -0.22 -0.09 -0.14

Gap t+1 0.05 0.13 -0.01 0.05 0.03
 t+4 0.16 0.89 * 0.17 0.53 0.38
 t+8 0.16 1.55 * 0.14 0.90 0.57

DC/EUR t+1 0.10 * 0.16 0.11
 t+4 0.24 * 0.47 * -- 0.36 --
 t+8 0.35 * 0.49 * -- 0.38 --

PPI t+1 0.34 * 0.71 ** 0.28 0.36 0.32
 t+4 0.62 1.16 ** 0.69 0.74 0.72
 t+8 0.83 1.03 * 0.64 0.63 0.64

CPI t+1 0.95 ** 0.59 ** 0.54 0.92 0.76
 t+4 2.11 ** 0.86 ** 0.54 1.33 0.99
 t+8 3.17 ** 0.74 * 0.43 1.45 1.01

Note: * significance at 68% level and ** significance at 95% level.
Calculation based on 1,500 Efron-type bootstrap replications. Bulgaria
(Bg), Croatia (Hr), Czech Republic (Cz), Estonia (Ee), Hungary (Hu),
Poland (PL), and Slovak Republic (Sk). Fix., fixers; Target.,
targeters; and Av., averages.