Exchange rate pass-through in CIS countries.
Beckmann, Elisabeth ; Fidrmuc, Jarko
INTRODUCTION
Having experienced periods of hyperinflation in the 1990s, the
countries of the Commonwealth of Independent States (CIS), (1) to
differing degrees, moved towards stabilized economic development with
lower and less volatile rates of inflation. Recently, however, CIS
members have become more vulnerable to global shocks (EBRD, 2012). In
particular, during the global financial crisis, output declined and the
exchange rate depreciated in all CIS countries. This led to concerns of
a rebound to instability and high inflation.
The majority of the CIS members operate a managed float exchange
rate regime. With the exception of Russia, they are predominantly small
and import-dependent economies (especially regarding their energy
consumption). Understanding exchange rate pass-through, defined as the
impact of the exchange rate on domestic prices, is therefore important
to policymakers with regard to the stabilization of CIS economies
following the global financial crisis.
Using monthly data, we investigate exchange rate pass-through from
the US dollar (USD) and the Euro on the consumer price index (CPI)
between January 1999 and March 2010. First, we compare the exchange rate
pass-through in the short run using VAR models for the individual
countries and in the long run by applying panel cointegration methods.
Second, we compare the exchange rate pass-through from the USD and from
the Euro.
Although the members of the CIS share a common history of nearly 70
years in the Soviet Union with central planning, there are enormous
differences among the economies in their transition to a market economy
(Pomfret, 2003). We exclude Belarus, Tajikistan, Turkmenistan, and
Uzbekistan from the empirical analysis, as some prices in these
countries are administered, which might cause a significant estimation
bias of the exchange rate pass-through. Azerbaijan had a currency reform
in 2006, which leaves a very short stable period for the estimation of
exchange rate pass-through up until the global financial crisis.
Therefore, our empirical analysis focuses on Armenia, Georgia,
Kazakhstan, Kyrgyzstan, Moldova, Russia, and Ukraine. We do not include
the remaining former Soviet Union countries (Baltic States) because they
face very different international and institutional conditions within
the European Union.
We build upon the work by Korhonen and Wachtel (2006), which is the
only study that estimates exchange rate pass-through in more than one
CIS country and has a comparative focus. We extend their analysis in
several dimensions. First, we update the original results that are based
on the years 1999-2004. This is of particular relevance, as we want to
investigate the impact of the economic stabilization and increased
vulnerability to global shocks, which had not occurred within the time
period of Korhonen and Wachtel's analysis.
Second, we provide a first estimate of the long-run exchange rate
pass-through in a panel cointegration framework for CIS countries.
Although theories of exchange rate pass-through contain a long-run
steady-state relationship between domestic prices and the exchange rate,
the long-run dimension is often disregarded in the empirical literature.
Our estimation of long-run exchange rate pass-through follows on from
the work of de Bandt et al. (2008), who study long-run pass-through for
euro area countries.
We estimate exchange rate pass-through in the short run to be
between 30% and 50% with regard to the USD, and around 20% for the Euro.
Our results confirm that the application of the long-run approach to
exchange rate pass-through is highly important in the CIS countries. In
the long run, we confirm a higher extent of the exchange rate
pass-through (above 60%) for both USD and the Euro.
The paper is structured as follows. The next section provides a
brief but comprehensive literature review with regard to publications
estimating exchange rate pass-through in one or more CIS countries. The
section 'Empirical analysis' presents the empirical analysis,
and the section 'Conclusions' concludes.
LITERATURE REVIEW
Exchange rate pass-through is defined as the elasticity of domestic
prices with respect to the exchange rate (Goldberg and Knetter, 1997).
The empirical research has shown that exchange rate pass-through is
incomplete and gradual, meaning that exchange rate changes are only
partially transmitted to domestic prices and with a significant delay.
Moreover, exchange rate pass- through is most pronounced for import
prices, is lower for producer prices, and is lowest for consumer prices
(Egert and MacDonald, 2009). In general, exchange rate pass-through is
higher in poorer and smaller countries. Furthermore, there is some
evidence that exchange rate pass-through has declined over the past
decades. We focus on those studies that examine one or more of the CIS
countries. We also refer the reader to the survey by Egert and MacDonald
(2009) of the research on monetary transmission channels in transition
economies and the general literature surveys by Menon (1995), Goldberg
and Knetter (1997), and Mishkin (2008).
Empirical studies for the CIS estimate either single equation OLS
or VAR models using CPI data. Only Korhonen and Wachtel (2006) studied
more than one CIS country. They found that exchange rate movements were
transmitted comparatively quickly and fully to consumer prices. The
pass-through was found to be higher in CIS countries than in other
transition economies. World prices (proxied by the US inflation) were
not a significant factor in determining exchange rate pass-through.
Exchange rate pass-through in Armenia, Georgia, and Russia was
analyzed in individual studies by Dobrynskaya and Levando (2005), Beck
and Barnard (2009), and Kataranova (2010), with Russia, as the largest
economy, receiving particular attention. For Georgia, Samkharadze (2008)
described a decline in the importance of exchange rate pass-through. In
general, these individual country studies confirmed the comparatively
high pass-through rates for the CIS countries reported by Korhonen and
Wachtel (2006).
EMPIRICAL ANALYSIS
Data description
Similar to previous empirical studies, we use CPI data from the
International Financial Statistics (IFS) of the International Monetary
Fund, the bilateral exchange rate of the national currencies against the
USD from the IFS, the bilateral exchange rate of national currencies
against the Euro from Bloomberg, and Brent Crude as a proxy for oil
prices from the IFS. To account for foreign price pressure, we also
consider CPI data for the United States and the EU from the IFS.
We use monthly data (in logarithms) from January 1999 until the
beginning of 2010. We exclude data before 1999 because of structural
breaks following the Russian crisis in 1998. CPI indices display a
significant seasonality pattern for all countries and are adjusted using
the X12 additive monthly seasonal adjustment method.
All analyzed time series are typically considered to be
non-stationary. As a starting point, the Augmented Dickey Fuller Test
(ADF) is carried out for each of the endogenous variables. The lag
length is set according to the Schwarz information criterion. In
addition, we conduct the Dickey-Fuller GLS by Elliott et al. (1996),
which improves the power of the ADF test by detrending, and the Ng and
Perron test (2001), which uses a different information criterion to set
the lag length to improve the size of the test. In general, the tests
indicate that these variables are non-stationary in levels and
stationary in first differences, (2) a result that is confirmed by many
other papers. To account for the likely presence of structural breaks in
the time series, which would compromise the validity of the unit root
tests, we carried out an endogenous break test proposed by Andrews and
Zivot (1992). Table 1 presents the results for the price series, which
are the key variable of interest for the present analysis. The Andrews
and Zivot (1992) test fails to reject the null of simple unit root
without structural breaks for all price series except Kyrgyzstan.
On the basis of these tests, we conclude that all variables are
non-stationary in levels and stationary in first differences. Reflecting
the results of Andrews and Zivot (1992) unit root tests, we proceed in
our analysis with time series taken for the whole available time
periods. However, the robustness analysis will check the stability of
the results for a subperiod excluding the most recent period of the
financial crisis.
Estimation of short-run exchange rate pass-through
We start our analysis with the discussion of short-run exchange
rate pass- through. We estimate a VAR model for the domestic CPI, the
bilateral exchange rate, the Brent Crude price index for oil, and the US
or EU inflation. The initial lag length is set at 9 and the residuals
are examined for outliers. To allow for comparability across countries,
an outlier is identified if a residuum exceeds its three standard
deviations. Consequently, we define dummy variables for these
observations. In general, the outliers are in line with results from
Table 1 and appear to be caused by exogenous events that do not affect
other periods of the estimation. The dummy variables are found to be
highly significant in the estimations. Next, we determine the optimal
lag length according to the Hannan, Quinn, Schwarz, and Likelihood Ratio
information criteria. These are compared also with the test of lag
exclusion and the AR stability test to determine the final lag length.
Small variations to the selected lag length are made to see whether the
model would be sensitive to a different lag order. Similar to Korhonen
and Wachtel (2006), the final lag length is rather short. Finally, the
significance of the US and Euro inflation, the proxy suggested by
Korhonen and Wachtel (2006) for price pressures from the world economy,
is tested by Granger Causality and Granger block exogeneity tests. Table
2 presents the final specifications.
Due to the complex lag structure of a VAR and the simultaneous
relationships between the endogenous variables, the estimated
coefficients as such are difficult to interpret. To study the short-term
dynamics of exchange rate pass-through, we impose a 1-unit shock on the
estimated system. As all variables are in logarithms and first
differences, the 1-unit shock corresponds to a percentual innovation to
the original variables and a further normalization is not necessary. The
accumulated response of the variables after 12 months is recorded in the
second column of Table 2. Exchange rate pass- through is the fraction of
the 1% shock that is passed on from the exchange rate to the domestic
CPI.
Table 2 shows that exchange rate pass-through is relatively high in
the CIS countries for the USD, while it is lower for the Euro. Exchange
rate pass- through from the USD to consumer prices is around 30% for
Armenia. For Moldova and Ukraine, it is around 50% and for Kyrgyzstan it
is around 70%. For Georgia, the exchange rate pass-through estimate is
very low at around 2 % with respect to the USD, and is not significant.
In combination with weak results of the unit root tests for the Georgian
variables, this might point out that the VAR model is misspecified or
that important explanatory variables are missing from the estimation.
For Russia and Kazakhstan, the picture is similar. The exchange
rate pass-through estimate is very low or even negative, and is not
significant. Notably, Russia and Kazakhstan are both important energy
exporters. As already pointed out by Korhonen and Wachtel (2006), the
surprising result of negative pass-through in Russia and extremely low
exchange rate pass-through in Kazakhstan could possibly be explained by
the interaction between the oil price and nominal exchange rate and
domestic inflation. Beck and Barnard (2009), however, found that Russian
exchange rate pass-through is around 25% and Kataranova (2010) showed it
to be asymmetric.
In summary, for energy-importing countries, rising oil prices can
influence core and overall inflation via higher production costs and by
extension the expectation of inflation in wages and prices (further
discussed in the section 'Estimation of long-run exchange rate
pass-through'). It appears that for the energy-exporting countries
oil prices seem to have a primary influence on the exchange rate rather
than directly on domestic prices. It is therefore very difficult to
disentangle these effects in a simple VAR model.
For the Euro, results are less robust. The estimates for Armenia,
Ukraine, and Russia do not differ much in the sensitivity analysis from
the original estimate. For Armenia, exchange rate pass-through from the
Euro is similar to that from the USD at around 30%. For Ukraine,
exchange rate pass-through from the Euro is around 20 percentage points
lower than from the USD. The estimate for Russia is insignificant, but
appears to be robust at around 2%. For the rest of the countries,
estimates of exchange rate pass-through from the Euro are not robust.
It might have been expected that the Euro would gain importance as
a currency especially for the European CIS countries. However, exchange
rate pass-through is important for these countries mainly in relation to
the USD and less in relation to the Euro, with Ukraine as a possible
exception. Korhonen and Wachtel (2006) came to a similar conclusion, so
there does not appear to have been an increase in the importance of the
Euro for the CIS countries. However, they found a positive correlation
between average inflation and exchange rate pass-through, which we find
has reversed for the longer time period and is now -0.17. Although the
correlation coefficient is very low, the negative correlation is more in
line with the results for the rest of the world.
We also look at the speed of exchange rate pass-through defined by
the ratio of the accumulated impulse response after 6 months to that of
the accumulated impulse response after 12 months. For all countries, the
speed of exchange rate pass-through is very high at above 0.7. This is,
however, also influenced by the short lag length selection.
As an alternative indicator of exchange rate pass-through, we
estimate the variance decomposition after 12 months. We apply the
Cholesky factorization and order the oil price first, followed by
foreign inflation, then exchange rate, and lastly domestic inflation.
The results from the variance decomposition (Table 3) are largely in
line with those from the accumulated impulse response (Table 2). For
those countries where the impulse response is insignificant, the effect
on domestic prices, which is due to the exchange rate, is also low
(between 0.15% and 3.2%). By contrast, the percentage explained by the
shock is much higher for countries with significant impulse responses,
ranging from around 5% for Ukraine to 34% for Kyrgyzstan.
To check the robustness of our results, we carried out several
sensitivity checks including estimation over a reduced sample period,
estimation with nominal effective exchange rates, and estimation with
the price index for all primary commodities instead of the oil price.
(3) The magnitude of exchange rate pass-through does not change greatly
if the period of the financial crisis is excluded. It also does not
change significantly if alternative variables are used. The results
remain robust over different specifications.
Estimation of long-run exchange rate pass-through
While VAR estimations are a standard tool for the discussion of
exchange rate pass-through in the literature, they require that all time
series are differenced in order to achieve their stationarity. However,
this may result in an underestimation of the long-run relationship
between time series. There are several cointegration analyses of
exchange rate pass-through, but few using panel cointegration
techniques. Dobrynskaya and Levando (2005) estimated cointegration
models for disaggregate CPI data in Russia. Larue et al. (2010) used
threshold cointegration for meat prices in Canada and Japan. De Bandt et
al. (2008) applied time series and panel cointegration tests for
exchange rate pass-through in import prices of selected OECD countries.
We examine long-run exchange rate pass-through in a panel of CIS
countries. The choice of panel data methods for this analysis reflects
the fact that we only have short time series. Panel data methods, in
comparison to time series models, provide a more powerful tool to
analyze the long-run steady state, as suggested by theories of exchange
rate pass-through, and overcome the difficulties of identifying a
long-run relationship in the data, as explained in de Bandt et al.
(2008).
Although we confirmed that all times series are I(1) in previous
sections, adding a cross-section dimension to unit root tests can
potentially improve the quality of these tests significantly by
increasing their power (Banerjee, 1999). Table 4 presents selected panel
unit root tests recommended in the literature. In general, the panel
unit root tests confirm that the variables contain a unit root. The
heterogeneous versions of the Dickey-Fuller test, that is, the IPS test
according to Im et al. (2003) and the so-called Fisher's ADF test
proposed by Maddala and Wu (1999) perform better than the homogenous
version of the Dickey Fuller test (LLC test) by Levin et al. (2002).
Finally, the panel version of the PKPSS test by Kwiatkowski et al.
(1992), proposed by Hadri (2000), clearly rejects the null of panel
stationarity for all variables.
However, all tests with the exception of the PKPSS test indicate
stationarity for both exchange rates. This surprising result, which
contradicts our previous findings for the individual time series, may be
caused by exchange rate depreciations during the Russian financial
crisis in 1998 and the global financial crisis in 2008. Therefore, we
perform panel unit root test also for a shorter time period between
January 2000 and August 2008, and, for this less volatile period, we
find that all series are non-stationary.
Given non-stationarity for our CIS panel, we estimate the long-run
relationship for selected models including the exchange rates for USD
and Euro, respectively. The control variables include the US or Euro
consumer prices and oil prices. We also include country fixed effects
and time effects (these specifications exclude foreign CPI and oil
prices as they are multicollinear to time effects) in selected
specifications. Moreover, we present results for selected specifications
between January 2000 and August 2008, that is, excluding the Russian and
the global financial crises. Finally, we consider specifications without
the oil exporting countries (Russia and Kazakhstan) in order to analyze
the sensitivity of our results to the oil price (Pomfret, 2011).
Table 5 presents results of the eointegrating relationship and
panel cointegration tests for selected specifications. The long-run
exchange rate pass-through is estimated to be higher than in the
short-run VAR models. The basic specification in Column 1, which
includes foreign inflation and the oil price, puts the long-run
coefficient at 57% for the USD exchange rate. It is only slightly lower
for the Euro exchange rate, which is 47%. However, we estimate the
coefficient of foreign inflation to be above 1 for both the US and the
Euro area, while oil prices are insignificant. Although this may
correspond to high inflation pressures in the CIS countries, we prefer
the specification without foreign prices, which is similar to our final
VAR specification. This model shows a higher exchange rate pass-through
(60% for the USD and 75% for the Euro) and a significant coefficient for
the oil price. Alternatively, we include time effects that cover the
impact of oil prices, world inflation in general and other effects that
are time specific. As time effects are common for all countries, this
specification uses a balanced panel. The results reported in Column 3
show even higher exchange rate pass-through for both USD and Euro
exchange rates (72%-77%). The exchange rate pass-through is found to be
higher for the period excluding the financial crisis (January
2000-August 2008), see Columns 4 and 5, compared with the basic
specification. If we exclude Russia and Kazakhstan from the sample in
Column 6, exchange rate pass-through is lower, but the oil price
coefficient remains fairly unchanged. The long-run exchange rate
pass-through of the Euro is generally lower than for the USD, but the
difference to USD exchange rate pass-through is less pronounced than in
VAR models.
The lower part of Table 5 presents the results of selected panel
cointegration tests. Following Engle and Granger's approach,
Pedroni (1996 and 2001) and Kao (1999) proposed several tests based on a
panel version of the residual Engle-Granger test. Maddala and Wu (1999)
proposed an alternative approach to test for cointegration in panel data
by combining tests from individual cross-sections to obtain a test
statistic for the full panel.
We can see that there is some ambiguity (4) between the different
tests in some specifications. For the basic specification, Kao's
and panel Johansen tests confirm a cointegrating relationship between
the variables, but not so the Pedroni test. However, for the less
volatile time period (January 2000- August 2008), Pedroni's test
fails to reject the null of panel cointegration. Thus, we conclude that
the panel estimations present a long-run relationship, which confirms
the extent of the exchange rate pass-through above the level indicated
by short-run VAR models. Moreover, currency effects are less important
for the long-run extent of the exchange rate pass-through. In
particular, the long-run effects of Euro exchange rate developments are
only slightly lower than the effects of the USD exchange rate.
Discussion of results
Before drawing any conclusions, several limitations of the present
analysis must be taken into account. First, the time series are
comparatively short and may include structural breaks caused by events
in the CIS and global influences such as the financial crisis. Even
though results are robust in sensitivity checks, it is possible that the
presence of structural breaks still influences exchange rate
pass-through estimates.
Second, the analysis does not distinguish between the types of
shocks that hit the economy (Mishkin, 2008). In particular, Dreger and
Fidrmuc (2011) found that the importance of regional shocks in the CIS
had decreased but that the CIS had become more vulnerable to global
shocks. The parsimonious VAR models do not attempt to distinguish
between global and regional factors in exchange rate pass-through. By
contrast, panel cointegration models stress the common trends, while
country-specific factors are accounted for by country effects.
Third, the development of the equilibrium exchange rate should be
accounted for when estimating exchange rate pass-through (Egert and
MacDonald, 2009). Transition economies in particular exhibited a
long-term appreciation. Darvas (2001) first estimated the equilibrium
exchange rate and then pass-through, allowing for time varying
parameters in his investigation of exchange rate pass-through. His
approach naturally requires more detailed data, which are not available
for the CIS. However, after 2008, exchange rates in the CIS depreciated
so that the overall exchange rate pass-through estimate should not be
significantly biased by permanent exchange rate appreciations.
[FIGURE 1 OMITTED]
Our estimation also does not take into account asymmetries in
exchange rate pass-through. Carranza et al. (2009) found higher exchange
rate pass-through in boom periods and lower exchange rate pass-through
in recessions. While our estimation does not account for asymmetries, it
is interesting to note that our estimate of long-run exchange rate
pass-through is higher if we exclude the period of the global financial
crisis, which resulted in an economic downturn in the CIS. This would be
in line with the result from Carranza et al. (2009).
Finally, at present, it remains difficult to analyze intra-CIS
dynamics. Russia as the largest economy continues to play an important
role for many of the other member states. Figure 1 shows that a high
share of imports from Russia or mineral imports are generally associated
with high exchange rate pass-through. However, the simultaneous
orientation of the majority of CIS exchange rates to the USD complicates
the identification of regional dynamics. The analysis of exchange rate
pass-through with regard to the rouble exchange rate could be a line of
future research, especially if Russia continues to move towards a more
flexible exchange rate regime.
CONCLUSIONS
The depreciations of the local currencies of CIS countries during
the global financial crisis have raised the question of how much of a
risk this has posed to inflation, in particular considering these
countries' history of hyperinflation in the 1990s. To shed light on
this issue, we analyze exchange rate passthrough for seven CIS countries
during the recent period of the global financial crisis. Moreover, we
compare the short-run and long-run extent of exchange rate pass-through.
By estimating VAR and panel cointegration models, we show that
exchange rates have an effect on domestic consumer prices in the
majority of the CIS countries. The small economies of the CIS, which
also display considerable levels of dollarization, have relatively high
pass-through rates. The exchange rate of the USD is of particular
significance, while the exchange rate of the Euro has a less important
effect in the short run.
In general, exchange rate pass-through is high in the CIS
countries. However, the estimates of short-run exchange rate
pass-through bear witness to the heterogeneity of its short-run
dynamics. For Kyrgyzstan, Moldova, and Ukraine, short-run exchange rate
pass-through estimates are high, with values between 50% and 70% at the
12-month horizon. Despite their peripheral geographical position
(Raballand, 2003), these countries are small open economies, which are
highly reliant on imports, both of industrial goods and also of energy.
Armenia shows lower short-run pass-through rates at 30%, which could be
linked to the extreme trade barriers with Turkey and Azerbaijan. For the
large energy exporters Russia and Kazakhstan, the primary effect seems
to stem from oil prices on the exchange rate. By contrast, energy-
importing countries generally have high exchange rate pass-through. We
cannot provide a meaningful estimate of short-run exchange rate
pass-through for Georgia as a structural break after the 'Rose
revolution' seems to influence estimates.
Moreover, we document the importance of the long-run effects.
Applying panel cointegration methods, we show that in the long run,
exchange rate pass-through converges to a high level in the CIS.
Long-run exchange rate pass-through from the USD in the CIS is around 60
% and above. It is slightly lower for the Euro at around 50%. As the
Euro does not play a significant role for short-run pass-through, this
indicates that currency effects are less important for long-run exchange
rate pass-through. These estimates suggest that monetary policy in the
CIS needs to look beyond the short-run effects of exchange-rate
pass-through in particular with regard to the persistence of inflation.
Looking at the size of exchange rate pass-through, we find that it
is negatively correlated with inflation in the CIS. Although this is
contrary to Korhonen and Wachtel (2006) who found a positive correlation
for the period up to 2004, a negative correlation is in line with the
literature on pass- through in developed economies. For the CIS, this
could indicate that price dynamics are becoming increasingly similar to
the rest of the world and that the CIS is more closely linked to shocks
in the world economy.
Acknowledgements
The authors would like to thank the participants of the conference
on Fiscal Stabilization and Monetary Union, Mendel University Brno,
November 2011, as well as Aleksandra Riedl and Campbell Leith for
valuable comments and suggestions. The views are the authors' and
do not necessarily reflect those of the Oesterreichische Nationalbank or
the Eurosystem.
REFERENCES
Andrews, DWK and Zivot, E. 1992: Further evidence on the great
crash, the oil- price shock, and the unit-root hypothesis. Journal of
Business and Economic Statistics 10(3): 251- 270.
Banerjee, A. 1999: Panel data unit roots and cointegration: An
overview. Oxford Bulletin of Economics and Statistics 61 ($1): 607-629.
Beck, R and Barnard, G. 2009: Towards a flexible exchange rate
policy in Russia. OECD Economics Department Working Paper no. 744, OECD
Publishing. http://dx.doi.org/10.1787/218428024413.
Carranza, L, Galdon-Sanchez, JE and Gomez-Biscarri, J. 2009:
Exchange rate and inflation dynamics in dollarized economies. Journal of
Development Economics 89(1): 98-108.
Darvas, Z. 2001: Exchange rate pass-through and real exchange rate
in EU candidate countries. Discussion Paper 10/01, Economic Research
Centre of the Deutsche Undesb Bank. Frankfurt a. M., Germany.
de Bandt, O, Banerjee, A and Kozluk, T. 2008: Measuring long-run
exchange rate pass-through. Economics--The Open-Access, Open-Assessment
E-Journal 2(6): 1-36.
Dobrynskaya, V and Levando, D. 2005: A study of exchange rate
pass-through effect in Russia. Working Paper WP9/2005/02, Moscow State
University, Higher School of Economics: Moscow, Russia.
Dreger, C and Fidrmuc, J. 2011: Drivers of exchange rate dynamics
in selected CIS countries: Evidence from a FAVAR analysis. Emerging
Markets Finance and Trade 47(4): 5-15. EBRD. 2012: Transition region in
the shadows of the eurozone crisis. Chapter 2. Transition Report, EBRD:
London.
Egert, B and MacDonald, R. 2009: Monetary transmission mechanism in
Central and Eastern Europe: Surveying the surveyable. Journal of
Economic Surveys 23(2): 277-327.
Elliott, G, Rothenberg, TJ and Stock, JH. 1996: Efficient tests for
an autoregressive unit root. Econometrica 64(4): 813-836.
Goldberg, PK and Knetter, MM. 1997: Goods prices and exchange
rates: What have we learned? Journal of Economic Literature 35(3):
1243-1272.
Hadri, K. 2000: Testing for stationarity in heterogenous panel
data. Econometrics Journal 3(2): 148-161.
Im, KS, Pesaran, MH and Shin, Y. 2003: Testing for unit root in
heterogenous panels. Journal of Econometrics 115(1): 53-74.
Kao, C. 1999: Spurious regression and residual-based tests for
cointegration in panel data. Journal of Econometrics 90(1): 1-44.
Kataranova, M. 2010: Sviaz' mezhdu obmennym kursom i
infliatsiei v Rossii (In Russian: The link between the exchange rate and
inflation in Russia). Zhurnal Voprosy Ekonomiki 1 (1): 1-29.
Korhonen, 1 and Wachtel, P. 2006: A note on exchange rate
pass-through in CIS countries. Research in International Finance and
Business 20(2): 215-226.
Kwiatkowski, D, Phillips, PCB, Schmidt, P and Shin, Y. 1992:
Testing the null hypothesis of stationarity against the alternative of a
unit root: How sure are we that economic time series have a unit root?
Journal of Econometrics 54(1-3): 159-178.
Lame, B, Gervais, J-P and Rancourt, Y. 2010: Exchange rate
pass-through, menu costs and threshold cointegration. Empirical
Economics 38 (1): 171-192.
Levin, A, Lin, C-F and Chu, C-SJ. 2002: Unit root tests in panel
data: Asymptotic and finite-sample properties. Journal of Econometrics
108(1): 1-24.
Maddala, GS and Wu, S. 1999: A comparative study of unit root tests
with panel data and a new simple test. Oxford Bulletin of Economics and
Statistics 61(S1): 631-652.
Menon, J. 1995: Exchange rate pass-through. Journal of Economic
Surveys 9(2): 197-231.
Mishkin, F. 2008: Exchange rate pass-through and monetary policy.
NBER Working Paper no. 13889. National Bureau of Economic Research:
Cambridge, MA.
Ng, S. and Perron, P. 2001: Lag length selection and the
construction of unit root tests with good size and power. Econometrica
69(6): 1519-1554.
Pedroni, P. 1996: Fully modified OLS for heterogenous cointegrated
panels and the case of purchasing power parity. Working Paper no.
96-020, Indiana University, Indianapolis.
Pedroni, P. 2001: Purchasing power parity tests in cointegrated
panels. Review of Economics and Statistics 83(4): 727-731.
Phillips, PCB and Moon, HR. 1999: Linear regression limit theory
for nonstationary panel data. Econometrica 67(5): 1057-1112.
Pomffet, R. 2003: Economic performance in Central Asia since 1991:
Macro and micro evidence. Comparative Economic Studies 45(4): 442-465.
Pomffet, R. 2011 : Exploiting energy and mineral resources in
Central Asia, Azerbaijan and Mongolia. Comparative Economic Studies
53(1): 5-33.
Raballand, G. 2003: Determinants of the negative impact of being
landlocked on trade: An empirical investigation through the Central
Asian case. Comparative Economic Studies 45(4): 520-536.
Samkharadze, B. 2008: Monetary transmission mechanism in Georgia:
analyzing pass-through of different channels. National Bank of Georgia
Working Paper no. 02/2008. National Bank of Georgia: Tbilisi, Georgia.
(1) Armenia, Azerbaijan, Belarus, Kazakhstan, Kyrgyzstan, Moldova,
Russia, Tajikistan, and Uzbekistan are members of the CIS. Turkmenistan
and Ukraine are no official members but they participate in activities
of the CIS. We also include Georgia, who was a C1S member between 1993
and 2009.
(2) Results are available upon request from the authors.
(3) Results are available upon request from the authors.
(4) Purchasing power parity theory predicts that real exchange
rates should be stationary, which implies complete exchange rate
pass-through. Therefore, the weak results of panel cointegration tests
for some specifications may arise because exchange rate pass-through is
incomplete. Despite the ambiguity of panel cointegration tests, the
estimated coefficients still provide consistent estimates of the
exchange rate pass-through because panel models avoid the problem of
spurious regression by using pooled data (Phillips and Moon, 1999; Kao,
1999).
ELISABETH BECKMANN [1] & JARKO FIDRMUC [2]
[1] Oesterreichische Nationalbank, Foreign Research Division, POB
61, A-1011 Vienna, Austria.
[2] Zeppelin University, Am Seemoser Horn 20, D-88045
Friedrichshafen, Germany.
Table 1: Andrews and Zivot's unit root test
Minimum Identified
Country t-statistics Significance Lags structural break
Armenia -2.787 0 April 2005
Georgia -5.129 * 0 December 2003
Kazakhstan -4.803 3 October 2006
Kyrgyzstan -7.216 ** 1 February 2007
Moldova -4.284 1 February 2006
Russia -3.992 1 July 2000
Ukraine -3.446 3 March 2006
Note: Critical values for a rejection of the null of unit
root (no structural break) are -5.57 and -5.08 at 1% and 5%
significance levels, respectively.
* and ** denote statistical significance at the 5% and 1%
significance level, respectively.
Table 2: VAR Estimates of short-run exchange rate pass-through
Country Extent of Sample size
exchange rate
pass-through
after 12 months
Armenia USD 0.28 * January 1999-March 2010
Armenia Euro 0.22 *** January 1999-March 2010
Georgia USD 0.02 January 1999-March 2010
Georgia Euro 0.08 June 2000-March 2010
Kazakhstan USD 0.03 January 1999-December 2009
Kazakhstan Euro 0.02 May 1999-December 2009
Kyrgyzstan USD 0.71 *** January 1999-August 2009
Kyrgyzstan Euro 0.12 January 2004-August 2009
Moldova USD 0.47 *** January 1999-March 2010
Moldova Euro 0.18 *** March 2002-March 2010
Russia USD -0.17 January 1999-February 2010
Russia Euro 0.02 January 1999-February 2010
Ukraine USD 0.45 *** January 1999-March 2010
Ukraine Euro 0.25 * January 1999-March 2010
Country Lag CPI *
length included (a)
Armenia USD 5 0.89/0.70
(No)
Armenia Euro 6 0.16/0.26
(No)
Georgia USD 1 0.18/0.34
(No)
Georgia Euro 1 0.52/0.19
(No)
Kazakhstan USD 1 0.21/0.85
(No)
Kazakhstan Euro 1 0.35/0.03
(Yes)
Kyrgyzstan USD 3 0.73/0.19
(No)
Kyrgyzstan Euro 4 0.13/0.89
(No)
Moldova USD 4 0.20/0.41
(No)
Moldova Euro 4 0.55/0.68
(No)
Russia USD 4 0.74/0.24
(No)
Russia Euro 1 0.13/0.31
(No)
Ukraine USD 3 0.36/0.82
(No)
Ukraine Euro 3 0.85/0.04
(Yes)
Country Dummies
Armenia USD July 2002, June 2003, January 2005,
January-December 2005, March 2009
Armenia Euro July 2002, June 2003, January 2005,
October-December 2005, March 2009
Georgia USD November 2003, November 2008
Georgia Euro November 2003, August 2008, November 2008
Kazakhstan USD December 2003, September 2007, October 2007,
October-December 2008, February 2009
Kazakhstan Euro December 2003, September 2007, October 2007,
October-December 2008, February 2009
Kyrgyzstan USD October 2003, November 2003, October 2007
Kyrgyzstan Euro October-07
Moldova USD November 1999, June 2002
Moldova Euro January 2009-March 2010
Russia USD January 2000, August-December 2008, January 2009,
February 2009, March 2009, April 2009-February 2010
Russia Euro January 2000, August-November 2008, December 2008,
January 2009, February 2009, March 2009,
April 2009-February 2010
Ukraine USD November 2008, December 2008, December 1999,
January 2009, February 2009-March 2010
Ukraine Euro December 1999, November 2008, December 2008,
January 2009, February 2009-March 2010
(a) The results are the p-values for the Granger block exogeneity,
tests. The first figure is for exogeneity with regard to domestic
CPI, the second figure is for exogeneity with regard to the
exchange rate. The dummy variables are defined as 1 in the reported
periods and 0 otherwise. Note: * and *** denote statistical
significance at the 10% and 1% significance level, respectively.
Table 3: Variance decomposition
CPI USD Oil CPI Euro Oil
Armenia 90.7 6.0 3.3 67.1 18.7 7.0
Georgia 95.0 0.8 4.2 89.4 3.2 7.4
Kazakhstan 85.3 1.7 12.9 83.4 9.2 7.4
Kyrgyzstan 62.3 33.6 4.1 94.1 4.1 1.8
Moldova 70.9 20.9 8.2 69.8 18.7 11.4
Russia 90.9 2.6 6.4 90.9 0.15 9.0
Ukraine 88.2 6.7 5.1 88.5 4.15 5.75
Note: Variance Decomposition for CPI after 12 months based on VAR
models presented in Table 2. Cholesky ordering is recursive (oil
price, exchange rate, domestic inflation).
Table 4: Panel unit root tests
Domestic Exchange Exchange
CPI rate (uSD) rate (Euro)
IPS 1.14 -3.38 ***/0.07 (a) -1.74 **/0.88 (a)
LLC -4.64 *** -2.39 **/-1.41 * (a) -2.28 **/-0.45 (a)
Fishers 17.57 45.93 ***/17.00 (a) 23.42 */6.65 (a)
ADF
PUSS 21.8 *** 7.16 ***/8.91 (a) 16.33 ***/
10.99 *** (a)
Oil CPI CPI
price US Euro area
IPS -0.63 0.70 0.68
LLC -1.05 -0.85 -1.01
Fishers 2.72 0.49 0.52
ADF
PUSS 7.70 *** 8.61 *** 8.60 ***
(a) Test period is January 2000-August 2008.
*, **, *** denote significance at the 10%, 5%, 1% level,
respectively.
Note: IPS--Im, Peseran, and Shin test, LLC--Levin, Lin, and
Chu test, PKPSS--Panel Kwiatkowski, Phillips, Schmidt, and
Shin test.
Table 5: Panel cointegration
(1) (2) (3)
A: Specifications with USD exchange rates
USD exchange rate 0.57 *** 0.60 *** 0.77 ***
The US CPI 3.59 ***
Oil -0.02 0.51 ***
Number of observations 934 934 896
Sample 1999-2010 1999-2010 1999-2009
Fixed effects Yes Yes Yes
Time effects No No Yes
Russia and Kazakhstan Yes Yes Yes
Panel PP statistics 0.31 -1.54 * -1.71 **
Panel ADF statistics -0.90 -2.19 ** -2.29 **
Group PP statistics 0.62 0.36 -0.67
Group OF statistics -0.86 -0.40 -1.48 *
Kao ADF -4.56 *** -4.40 *** -4.25 ***
P. Johansen trace 135.3 *** 132.7 *** 113.0 ***
P. Johansen ME 123.6 *** 118.2 *** 103.9 ***
8: Specifications with euro exchange rates
Euro exchange rate 0.47 *** 0.75 *** 0.72 ***
Euro area CPI 3.07 ****
Oil 0.02 0.32 ***
Number of observations 934 934 896
Sample 1999-2010 1999-2010 1999-2009
Fixed effects Yes Yes Yes
Time effects No No Yes
Russia and Kazakhstan Yes Yes Yes
Panel PP statistics 0.06 -1.17 -2.372 ***
Panel ADF statistics 0.94 -1.18 -1.939 **
Group PP statistics 1.00 -0.21 -1.509 *
Group ADF statistics 1.53 -0.30 -0.997
Kao ADF -3.24 *** -3.53 *** -3.855 ***
P. Johansen trace 97.47 *** 72.48 *** 106.9 ***
P. Johansen ME 76.35 *** 47.01 *** 96.70 ***
(4) (5) (6)
A: Specifications with USD exchange rates
USD exchange rate 0.65 *** 0.61 *** 0.52 ***
The US CPI 3.14 ***
Oil 0.08 *** 0.48 *** 0.44 ***
Number of observations 728 728 520
Sample 2000-2008 2000-2008 2000-2008
Fixed effects Yes Yes Yes
Time effects No No No
Russia and Kazakhstan Yes Yes No
Panel PP statistics 1.79 -2.08 ** -1.89 **
Panel ADF statistics 1.12 -2.71 *** -1.83 **
Group PP statistics 2.08 -0.99 -0.76
Group OF statistics 1.09 -1.22 -0.56
Kao ADF -3.88 *** -5.22 *** -5.22 ***
P. Johansen trace 36.11 *** 55.34 *** 39.63 ***
P. Johansen ME 24.84 ** 39.70 *** 28.81 ***
8: Specifications with euro exchange rates
Euro exchange rate 0.31 *** 0.55 *** 0.44 ***
Euro area CPI 2.36 ***
Oil 0.09 *** 0.30 *** 0.29 ***
Number of observations 728 728 520
Sample 2000-2008 2000-2008 2000-2008
Fixed effects Yes Yes Yes
Time effects No No No
Russia and Kazakhstan Yes Yes No
Panel PP statistics 2.06 -2.28 ** -0.62
Panel ADF statistics 1.47 -2.94 *** -0.50
Group PP statistics 2.50 -1.51 0.33
Group ADF statistics 1.92 -2.20 ** 0.40
Kao ADF -2.48 *** -4.18 *** -3.23 ***
P. Johansen trace 20.2 18.61 10.36
P. Johansen ME 15.6 13.37 8.17
Note: *, **, *** denote significance at the 10%, 5%, 1%
level, respectively.
