CASE Network Studies and Analyses 248 - Exchange Rate Regimes and the Real Sector: a Sectoral Analysis of CEE Countries

S t u d i a i A n a l i z y
S t u d i e s & A n a l y s e s
C e n t r u m A n a l i z
S p o ł e c z n o – E k o n o m i c z n y c h
C e n t e r f o r S o c i a l
a n d E c o n o m i c R e s e a r c h
2 4 8
Przemek Kowalski, Wojciech Paczynski, Lukasz Rawdanowicz
Exchange Rate Regimes and the Real Sector:
a Sectoral Analysis of CEE Countries
W a r s a w , J u l y 2 0 0 3

Materials published here have a working paper character. They can be subject to further
publication. The views and opinions expressed here reflect the author(s) point of view and not
necessarily those of the CASE
This research was supported by a grant from the CERGE-EI Foundation under a program of the
Global Development Network. Additional funds for grantees in the Balkan countries have been
provided by the Austrian Government through WIIW, Vienna. All opinions expressed are those of
the author(s) and have not been endorsed by CERGE-EI, WIIW, or the GDN.
Keywords: exchange rate regimes, Central and Eastern European countries,
dynamic panels
Acknowledgements:
We are grateful for comments from participants of the summer workshop at CERGE in Prague and
of the seminar at the University of Dortmund. All remaining errors are our own. We would also like
to thank numerous people from statistical offices for providing us with data, in particular to
Zsuzsanna Boros, Ana Gorišek, Karmen Hren, Sona Seginova, Małgorzata Wojciechowska.
© CASE – Center for Social and Economic Research, Warsaw 2003
Graphic Design: Agnieszka Natalia Bury
ISSN 1506-1701, ISBN: 83-7178-301-9
Publisher:
CASE – Center for Social and Economic Research
12 Sienkiewicza, 00-944 Warsaw, Poland
tel.: (48 22) 622 66 27, 828 61 33, fax: (48 22) 828 60 69
e-mail: case@case.com.pl
http://www.case.com.pl/

Contents
Abstract ..............................................................................................................................5
1. Introduction ....................................................................................................................6
2. Literature review ............................................................................................................7
2.1. Theory....................................................................................................................7
2.2. Models using aggregated output............................................................................8
2.3. Introducing sectoral differentiation .......................................................................10
2.4. Theoretical lessons ..............................................................................................13
2.5. Empirical evidence...............................................................................................14
3. Empirical analysis........................................................................................................15
3.1. Classification of exchange rate systems..............................................................15
3.2. Data description ...................................................................................................16
3.3. Stylised facts – descriptive statistics ....................................................................16
3.4. Methodology and results......................................................................................18
4. Conclusions .................................................................................................................24
References........................................................................................................................26
Appendix I – Classification of exchange rate regimes .................................................29
Appendix II – Data............................................................................................................30
Appendix III – Estimation results ...................................................................................36

Studies & Analyses CASE No. 248 – Przemyslaw Kowalski, Wojciech Paczynski, Lukasz Rawdanowicz
Przemyslaw Kowalski
Przemyslaw Kowalski is an economist at the OECD Secretariat in Paris. He graduated in
economics from the Universities of Warsaw (Poland) and Sussex (UK) and is a doctoral candidate
at the University of Sussex. His areas of interest include: international trade theory and policy,
investment and international finance.
Wojciech Paczynski
Wojciech Paczynski holds Master degrees in International Economics from the University of
Sussex (1998), Warsaw University's Department of Economics (1999) and Warsaw University's
Department of Mathematics and Computer Science (2000). He is currently participating in a PhD
programme at the University of Dortmund. His research interests concentrate on monetary and
fiscal policy, institutional economics, game theory and political economy of transition. He has been
an economist at CASE since 2000.
Lukasz Rawdanowicz
Lukasz W. Rawdanowicz holds an MA in International Economics from Sussex University (UK) and
an MA in quantitative methods from Warsaw University (Poland). His main area of interest is
applied international macroeconomics. He has dealt with issues related to trade liberalisation,
currency crises propagation, exchange rate misalignments and exchange rate regime choice. His
empirical research focuses primarily on transition economies. He is a co-author of CASE’s
quarterly publications: Polish Economy – Trends, Analyses, Forecasts and Global Economy and
deals with macroeconomic forecasting.
4

Studies & Analyses CASE No. 248 – Exchange Rate Regimes and the Real Sector….
5
Abstract
This paper analyses the impact of exchange rate regimes on the real sector. While most
studies in this field have so far concentrated on aggregate variables, we pursue a sectoral
approach distinguishing between the tradable and nontradable sectors. Firstly, we present a
survey of the relevant theoretical and empirical literature. This demonstrates that evaluations of
exchange rate regimes and their impact on the real economy are largely dependant on specific
assumptions concerning, in particular, the parameters of a utility function, the nature of the price
adjustment process and the characteristics of analysed shocks. Secondly, we conduct an empirical
analysis of the behaviour of the tradable and nontradable sectors under different exchange rate
regimes for seven Central and Eastern European countries. We find no firm evidence of a
differential impact of given exchange rate regimes on the dynamics of output and prices in the two
sectors. We proffer a conceptual and technical interpretation of this.

1. Introduction
6
The choice of an exchange rate regime is one of the key economic policy choices. After the
wave of financial crises in the 1990s, the introduction of the euro and in view of euro-zone
enlargement by new members from Central and Eastern Europe (CEE), this choice is not only an
academic bone of contention, but also a real policy dilemma. The decision on the exchange rate
system is affected by many considerations, ranging from economic arguments related to the
stabilisation properties of regimes, existing institutional frameworks, trade and investment
concerns, to political arguments pertaining to political preferences in given international contexts.
On of the most intriguing aspects of choosing an exchange rate regime is its impact on the real
economy (Stockman, 1999; Harris, 2002). Half a century after Friedman (1953) formulated his
hypothesis of a flexible exchange rate as a facilitator of adjustments in an economy characterised
by nominal rigidities, there is still much controversy concerning the optimal choice of regimes for
particular countries or groups of countries. After the collapse of the Bretton-Woods architecture of
fixed exchange rates in 1973, real exchange rates became more volatile. At the same time, such
observable volatility did not result in any directly observable changes in other real variables.
Dornbusch’s (1976) infamous overshooting model demonstrated that while unanticipated monetary
contractions lead to temporary declines in output, the extent of exchange rate volatility under a
floating exchange rate regime may be much larger than that of any underlying shocks. However,
the lack of real effects in terms of increased real exchange rate volatility in the longer term
remained a puzzle. “That productive and allocative effects of the exchange rate system are virtually
undetectable is generally viewed as surprising, particularly in light of evidence that real exchange
rates (defined as the relative price of overall bundles of goods in two countries) vary substantially
more under a system of floating rates than pegged ones” (Stockman, 1999).
Theoretical and empirical investigations into exchange rate regimes to date have dealt
primarily with aggregate variables and have paid little attention to sectoral issues. This is surprising
given that the distinction between tradables and nontradables has important implications for
thinking across a whole range of issues in international economics. While this distinction emerges
naturally in discussions of real exchange rates, the question of the differential impact of exchange
rate regimes on the relative performance of sectors of the economy producing tradables and
nontradables has been neglected. This is an important shortcoming of the literature given that
nontradables comprise a predominant share of most economies. “Aggregate” approaches assume
that national output is traded internationally and hence, implicitly, focus only on tradables. In our
view, analysis at a disaggregated level provides new insights into the debate on the choice of
exchange rate regimes. Given the close relationship between nominal and real exchange rates
even in the medium term (see e.g. De Grauwe, 1997), it is probable that the choice of exchange
rate arrangement will impact upon the allocation of resources between the two sectors (tradables
vs. nontradables) and, in general, upon their growth rates. In view of the higher labour intensity

being usually characteristic of nontradables this may have not only economic but also political
effects (Fischer, 2001).
In dealing with exchange rate regime issues the problem of a clear-cut classification emerges,
especially for regimes in the ‘floating’ corner: pure free-float regimes are not that common and
many countries prefer managed floats. As Calvo and Reinhart (2002) have suggested, many
economies exhibit ‘fear of floating’ and in practice use interest rate and intervention policies whose
effect is to steer the exchange rates. At the other extreme, fixed pegs are more transparent given
their institutional frameworks in the form of either currency boards or dollarisation. The problem of
exchange rate regime classification, as such, boils down to a comparison of de facto and de jure
exchange rate policies.
Against this background, this paper provides a detailed critical survey of the theoretical and
empirical literature dealing with the sectoral dimension of the exchange rate regime debate, as well
as a new empirical contribution to the debate. We investigate the ex post impact of exchange rate
regimes on tradables and nontradables in a sample of selected Central and Eastern Europe
transition economies. We demonstrate that the pegged regimes in our sample have failed to curb
the volatility of nominal and real effective exchange rates. We find no firm evidence of a differential
impact of exchange rate regimes on the dynamics of output and prices in tradables and
nontradables. We suggest various conceptual and technical interpretations for this.
The remainder of the paper is structured as follows. In Section 2, we review the main
theoretical contributions on the choice of exchange rate regime. This section also focuses on
empirical evidence for the differential impact of exchange rate regimes on the real economy. In
Section 3, we outline our empirical methodology and results. Conclusions are presented in
Section 4.
7
2. Literature review
2.1. Theory
The discussion about choice of exchange rate regime dates back to Friedman (1953), who
suggested that in the case of nominal price rigidities a floating exchange rate could alleviate
adjustments to shocks. Friedman’s notion of an insulating role played by floating exchange rates in
circumstances of foreign demand shocks was developed further in the 1960s. Mundell (1960), for
example, proposed a simple analytical framework demonstrating that the relative effectiveness of
an exchange rate regime in helping a country adjust to shocks depends in principle on government
policy rules, capital mobility and the speed of price adjustment to excessive or deficient demand.
Numerous subsequent contributions focused on concrete evidence of the impact of regime
choices. Despite these efforts, as Stockman (2000) points out, “economists lack strong evidence
on how exchange rate systems affect the economic variables that people care about, such as long-run
growth, avoidance of dislocations from business cycles, and so on”.

8
At a high level of generality, an evaluation of an exchange rate regime is possible when
specific criteria and an economic model are defined. With regards to evaluation criteria, most of the
literature focuses on measures of stability of macroeconomic variables, such as aggregate output,
inflation, real exchange rates, etc. relative to given ‘trend’ values. Definitions of such trends are
rarely explained, despite the fact that their specific character is a key factor in determining final
results (see e.g. Woodford (2002)).
In optimising models based on microfoundations, welfare evaluation of alternative exchange
rate regimes and monetary policies tends to be explicit and depends on the specification of the
utility function (usually involving specification of preferences in household consumption and
leisure). The specification of the model also affects optimality conditions. For instance, Tille (2002)
shows that the widely asserted benefits of a floating exchange rate regime in terms of its shock
absorbing properties, are severely diminished if one includes incomplete sectoral specialisation
(i.e. firms from one sector can be located in various countries) and sector-specific shocks (as
opposed to commonly assumed country-specific shocks). Finally, other specific assumptions
regarding the functioning of the world economy are vital for the predictions. In particular,
assumptions on how exchange rate fluctuations are transmitted into prices (exchange rate pass-through)
turn out to be critical for the results (see e.g. Lane and Ganelli (2002)).
Current theoretical models of exchange regime choice analyse primarily only two alternatives:
fixed vs. floating. This is motivated by difficulties in modelling so-called ‘intermediate’ or ‘hybrid’
regimes, but is also closely related to actual trends in exchange rate regime patterns. The last
decade has witnessed a general trend away from intermediate exchange rate regimes, especially
in the aftermath of the ERM and emerging markets crises in the 1990s. Fischer (2001)
demonstrated that the number of countries with ‘corner’ exchange regimes has increased
significantly since 1990, and that free floaters are now preponderant. Fischer’s explanation of this
finding was that for countries open to capital flows, soft pegs (i.e. intermediate solutions) do not
provide a viable option in the longer run due to their vulnerability to currency crises.
2.2. Models using aggregated output
The traditional workhorses of policy analysis: the Mundell-Fleming and AS/AD models, by
assumption, treat output as a homogenous good which is produced, consumed, and traded
internationally. Their results can, therefore, be cautiously interpreted as indicative of the impact of
exchange rate regimes on tradables. This caution is necessary, however, since the inclusion of
nontradables may change the properties of the models. Any explicit analysis of exchange rate
regimes in these models is constrained to assessment of their ability to stabilise output and
inflation around some natural level under a given structure of shocks. Deviations of output (and
possibly also prices and exchange rates) from their natural paths are considered undesirable and
are thought of as a proxy for costs associated with the adoption of a particular exchange rate
regime.

A simple illustrative example of such analysis is provided in Calvo (1999a). His basic version of
9
the Mundell-Fleming model can be reduced to the following equations:
y = a * e + u , a > 0 (1)
m = y + v (2)
where y denotes output, e the nominal exchange rate, m money (all in logarithms) and u and v
stochastic disturbances. Domestic prices are assumed to be constant and are normalized to 1
(thus zero in logarithm). Equation (1) represents an IS curve, and (2) an LM curve. In both
equations interest rate effects are included in stochastic terms and elasticity of money in relation to
output in (2) is set to unity.
A fixed exchange rate regime (Fix) is defined as one in which var(e) = 0, and a floating regime
(Flex) by the condition var(m) = 01. Thus, under Fix var(y) = var(u) (as var(e) = 0), whereas under
Flex var(y) = var(v) and var(e) = (var(v) + var(u) + 2cov(u,v)) / a2.
If we identify u with real shocks and v with nominal shocks, we arrive at the standard
conclusion that, provided the loss function is proportional to the variance of output, a fixed
exchange rate regime is superior (or inferior) to a floating exchange rate regime, if the volatility of
nominal shocks is larger (or smaller) than the volatility of real shocks. It could be argued, however,
that an excessive variation of the real exchange rate (stemming purely from nominal exchange rate
volatility, due to the fixed price assumption in this model) is also potentially harmful and, thus, that
variance of the exchange rate should also enter the loss function. This inclusion improves the
attractiveness of the fixed exchange rate regime, which increases with the correlation between v
and u (i.e. between real and nominal shocks)2.
The model can be also modified by allowing for a full and immediate pass-through from the
exchange rate to the price of domestic output (full indexation). Under these conditions, equations
(1) and (2) can be rewritten as
y = u (3)
m = y + e + v (4)
indicating that variance of output is equal to the variance of real shock u, independent of the
exchange rate regime, i.e. exchange rate regimes are equivalent in terms of the loss function
proportional to the variance of output3.
1 Note, that a floating regime is defined by a money stock target. In contrast, it is a form of inflation targeting that has
become the most commonly used system in floating exchange rates in the real world. Dehejia and Rowe (2001)
elaborate on this point.
2 Calvo (1999a) argues that the discussed correlation coefficient is likely to be positive.
3 Since volatility of the nominal exchange rate does not feed into the real rate, inclusion of this factor in the loss
function is not justifiable.

2.3. Introducing sectoral differentiation
10
While the aggregate Mundell-Fleming and AS/AD models do not explicitly tackle adjustments
in tradables and nontradables, they do provide predictions about short-term changes in income
levels and real exchange rate changes, as illustrated above. It can be argued that the analysis of
underlying shocks and resulting adjustments in real exchange rates and incomes is only a step
away from analysis of adjustments in tradables and nontradables.
Assuming equal income elasticity of demand in the two sectors4 and imperfect substitutability
between tradables and nontradables, we can make some approximate predictions about the likely
relative exposure of sectors producing tradables and nontradables to different shocks under
floating and fixed exchange rate regimes. If, as a result of a given shock, only income changes,
both sectors should experience proportionate changes in demand. In this case, conclusions related
to the desirability of a given exchange rate regime drawn from aggregate analysis can be extended
to the sectoral dimension. However, if income shifts coincide with exchange rate changes, the two
sectors will be affected differently.
This line of reasoning, coupled with the Calvo (1999a) model described above, confirms this
intuitive result: under a fixed exchange rate regime demand in both sectors will be affected
proportionally [var(y)=var(u); var(e)=0]. The same finding is obtained with the full exchange rate
pass-through. Under a floating exchange rate regime with less than full pass-through, however,
var(y) = var(v) and var(e) = (var(v) + var(u) + 2cov(u,v) / a2), implying that tradables and
nontradables will be affected proportionately only under the condition that var(v) + var(u) = -
2cov(u,v), which is a special case of perfect negative correlation of shocks to IS and LM curves
and cannot be presumed a priori.
These results from the very general variation of the IS/LM model in Calvo (1999a) can be
extended to more traditional versions of this model in which shocks to fiscal policy, foreign income,
foreign price level and interest rates can be analysed explicitly if one identifies real exchange rate
(in)stability with the relative (in)stability of sectors. This idea has been implemented by Dehejia and
Rowe (2001), who considered real exchange rate stability as one of the criteria for an assessment
of monetary regimes.
Dehejia and Rowe (2001) propose a version of the flexible price Mundell-Fleming model with
stochastic components in aggregate demand and supply and compare the variance of aggregate
output and of the expected real exchange rate and the expected interest rate, all relative to their
natural rates under different monetary rules. They consider the variance of the expected real
exchange rate as a proxy for stabilisation of tradables and the variance of the expected real
interest rate as an indicator of stability in the interest rate sensitive sector. Natural rates are defined
as those that would prevail if no price surprises occurred (i.e., when the actual price level meets
4 If a representative consumer makes the choice between tradables and nontradables according to the Cobb-
Douglas type of utility function, shares of expenditure devoted to tradables and nontradables will be constant, and the
unit increase in income, ceteris paribus, will increase demand in both sectors by the same amount.

expectations that have been formed in previous periods). The main finding of this study is that a
policy of price level targeting is most successful in stabilisation of output and expected real
exchange rates and interest rates relative to natural levels. Such a policy would appear, however,
to be a purely theoretical concept given that its implementation would imply the need for repeated
central bank activity to lower the price level (deflation). In the framework developed, price level
targeting means that a bank commits to reversing shocks to the price level that push it above (or
below) a declared level. The ranking between the two actual policy options analysed – fixed
exchange rate and inflation targeting – is parameter dependent.
While analysis of the Mundell-Fleming type of model provides a simple framework for thinking
about aggregate and sectoral stability it also leaves many questions unanswered. Because the
supply and demand sides are not modelled explicitly, we cannot account for either the effects of
the relative size of tradable and nontradable sectors or the size of the domestic tradable sector in
comparison to the foreign tradable sector5. Both of these factors clearly determine the relevance of
real exchange rate stability as a criterion for the assessment of adopted exchange rate regimes.
Discussion on the relevance of the size and openness of economies when choosing exchange rate
regime brings us back to the influential work of McKinnon (1963). He argues that flexible exchange
rates are less effective in checking the external balance and are more damaging to internal price
level stability the more open the economy is. This proposition is confirmed in a very different
setting by Calvo and Kumhof (2002), who developed and calibrated a general equilibrium model
with detailed microfoundations. Adopting the highly plausible assumptions of greater labour
intensity and greater price rigidity in the nontradable sector, the benefits of greater exchange rate
flexibility are reported to be higher for relatively closed economies6. This follows from the
observation that in a relatively closed economy (i.e. one with a larger nontradables sector),
exchange rate adjustment to foreign real shocks impacts on the labour market in nontradables.
The extent of this impact increases with the size of nontradables sector, as well as when nominal
rigidities in this sector are greater than or equal to those in the rest of the economy.
The new open economy macromodels, with explicitly specified microfoundations and general
equilibrium closures, seem better suited to deal with sectoral issues. Thus, it might come as a
surprise that relatively few models in the NOEM tradition include disaggregation of output into
tradables and nontradables. Obstfeld and Rogoff (1996) introduce nontradables in their small open
economy model. Their choice of the utility function, with consumption of tradables and
nontradables entering additively, implies that agents choose consumption of tradables
independently of nontradables’ production and consumption. Thus, in the absence of shocks
affecting productivity of tradables, the current account is always balanced. In this formulation, there
are no long-term effects of money shocks on the level of production. Only the price level in both
sectors is proportionally affected. In the short run, the price stickiness in the nontradable sector
5 The general equilibrium impact of e.g. expansion of the tradable and nontradable sector on the economy-wide
11
wage will depend not only on proportional expansion of their demands but also on the initial sizes of the two sectors.
6 This could also be viewed as giving some support to the Optimal Currency Area test of trade openness.

(prices in the tradable sector are fully flexible) means that nontradable output is demand
determined, e.g. it rises in response to an unanticipated permanent money shock.
12
Devereux and Lane (2001) propose a small open economy model allowing for, among other
effects, a different degree of exchange rate pass-through in import prices (where export prices
adjust instantaneously). The model is calibrated by setting parameter values using various results
found in the literature and data on East Asian emerging economies. Two types of shocks (changes
in the world interest rates and terms of trade) are considered under alternative monetary/exchange
rate regimes. The simulation, based on linear approximation, is used to study the response of
some model variables to shocks. In particular, impulse response graphs for output of tradables and
nontradables are presented. The impact of interest rates shock (increases in foreign interest rates)
on sectoral output is similar under the peg and CPI inflation targeting if exchange rate movements
are immediately passed on to import prices. Allowing for a delayed pass-through from exchange
rates to import prices changes the results. Under such conditions, the inflation targeting rule makes
tradables’ and nontradables’ output less volatile (i.e. the initial expansion in the tradables sector
and fall in the nontradables sector is smaller). Analysis of terms of trade shocks leads to similar
conclusions. The authors note that in their formulation a negative terms-of-trade shock is
essentially equivalent to a negative productivity shock in the export sector (the decline in return on
investment). Overall, the peg and the inflation targeting regimes produce similar outcomes for
tradable and nontradable output (and other variables) via immediate pass-through from exchange
rates to import prices. The delayed pass-through results in the superiority of the inflation targeting
framework as far as stabilisation properties of both aggregated and nontradables’ output are
concerned. However, this comes at the price of destabilising tradables’ output slightly more than
under the peg.
Finally, it is also worth mentioning another category of models that yield interesting results on
sectoral differentiation under alternative exchange rate regimes. Vartiainen (2002) introduces a
model of an open economy with two sectors – one producing tradables and the other producing
nontradables and sheltered from international competition. All workers are assumed to be
organised in trade unions that co-ordinate their wage claims within sectors. Two unions
representing tradable and nontradable sectors bargain over their wages each aiming at maximising
the sum of wages within the sector above the reservation wage, taking into account the labour
demand function. Under such conditions, in equilibrium, there are differences in relative prices,
output shares and wages in the sectors, depending on the exchange rate regime. The fixed
exchange rate regime induces lower prices for tradables and lower relative wages in the tradable
sector. This, however, does not indicate that a floating regime is preferred by the tradable sector,
since the impact on employment can also go in the opposite direction, i.e. a fixed exchange rate
regime can, while dampening relative wages, boost employment in the tradables sector.

13
2.4. Theoretical lessons
The above survey illustrates the lack of clearly defined relationships between the choice of the
exchange rate regime and its sectoral implications. The advances in modelling techniques and the
development of optimising models in the NOEM tradition did not substantially change the insights
provided by the Mundell-Fleming type of models. Specifications with respect to the parameters of
the utility function, the nature of the price adjustment process and characteristics of analysed
shocks ultimately determine the model’s predictions. For instance, the predictions of Devereux and
Lane (2001) are broadly in line with the findings of the model sketched by Calvo (1999a). The
immediate pass-through of exchange rate movements to prices makes free float and fixed regimes
similar in terms of their stabilisation properties.
Also, the explicit inclusion of tradables and nontradables in selected models does not seem to
result in any strong predictions on regimes’ characteristics with respect to the performance of the
tradable and nontradable sectors. Some models are able to illustrate the trade-off between
stabilisation of total output and stabilisation of output components (tradables in Devereux and Lane
(2001)). This point, however, can be also demonstrated using the Mundell-Fleming framework.
Calvo (1999a) rightly observes that the theoretical work on exchange rate regimes appears to
make one important simplification, which leads to certain reservations. Namely, it is commonly
assumed that random shocks hitting a particular economy are not related to the exchange rate
regime that has been adopted. However, given the way in which financial markets operate, it
seems plausible to assume that at least some shocks (originating in capital accounts) are likely to
be dependant on the choice of exchange rate regime. For instance, Calvo (1999b) argues that
dollarization may lower capital account volatility.
At this point, the relative robustness of theoretical findings from various categories of models
should be brought to attention. While the NOEM clearly have an advantage over Mundell-Fleming
models in that they allow for an explicit welfare analysis (rather than the use of ad hoc criteria
focusing primarily on output and inflation variance), it is not clear whether they are capable of
producing more valuable and realistic policy recommendations. Given the still early stages of
development of NOEM models and the sensitivity of their results to particular assumptions,
claiming their superiority for policy recommendations might appear premature7. In fact, no single
theory (or a set of theories) seems unquestionable in this respect. In this context, Taylor (1993)
noted that “...policy evaluation results cannot be obtained from pure theoretical considerations.
They depend on the empirical nature of the economic relations and on the size and correlation of
the shocks to these relations”.
7 See Lane (2001) and Lane and Ganelli (2002) for a comprehensive overview of NOEM models’ strengths and
deficiencies.

2.5. Empirical evidence
14
Having surveyed theoretical developments, we now turn to discussion of available empirical
evidence on the impact of the exchange rate regime on the relative performance of the tradable
and nontradable sectors. The non-neutrality of exchange rate regimes has been well documented
in empirical work. This, however, mostly pertains to the impact of regimes on the volatility of real
and nominal exchange rates. The standard fact one repeatedly comes across in the literature is
that floating regimes tend to induce higher volatility of exchange rates. Mussa (1986) and Flood
and Rose (1995) document this for OECD countries and Broda (2002) provides results for
developing countries. At the same time, no corresponding significant impact on other key
macroeconomic variables has been observed. Baxter and Stockman (1989) and later Flood and
Rose (1995) credibly proved this assertion. A theoretical framework allowing for (a partial)
explanation of this puzzle is provided by Dedola and Leduc (2002). Assuming differential speeds of
price adjustments between the two sectors, Dedola and Leduc were able to closely mimic the
characteristics of the actual data for G7 countries during the Bretton Woods era and after the
demise of this system.
Hau (2002) investigated the empirical relationship between openness to trade (measured by
the size of the tradable sector) and real exchange rate volatility (measured by standard deviations
over 3-year periods). The study is motivated by the theoretical predictions from Hau (2000) and
Obstfeld and Rogoff (2000) suggesting that more open economies should exhibit less volatile real
exchange rates since imported goods provide a channel for a quick adjustment in the domestic
aggregate price level in case of shocks occurrence. The postulated relationship is indeed found to
be statistically significant in two samples; one consisting of 23 OECD countries and the other
consisting of 48 countries. Interestingly, from the perspective pursued in this paper, the inclusion of
the dummy variable for the fixed exchange rate commitment visibly reduced the effect of openness
on real effective exchange rate volatility. This is broadly in line with findings from other studies that
this volatility tends to be reduced under fixed exchange rate regimes.
Broda (2002) assessed the different consequences of macroeconomic shocks depending on
the exchange rate regime. Semi-structural panel vector autoregression methodology is used to
analyse the responses of real GDP, real exchange rates and prices to negative terms of trade
shocks for 75 developing countries in the post-Bretton Woods period. The estimated impulse
response functions for the case of the fixed exchange rate regime picture a moderate, 3-year
decline in real GDP following a 10% permanent fall in terms of trade. No such statistically
significant response is found for the floating exchange rate. The real exchange rate experiences a
prolonged depreciation (primarily due to nominal depreciation as prices increase) in the floating
regime and no such effect is observed for countries with fixed exchange rates. This finding is
consistent with Friedman’s (1953) arguments.
A shortcoming of Broda’s analysis is that it concentrates only on one very specific shock and
thus does not provide general insights concerning the performance of alternative exchange rate

regimes. To our knowledge, there has to date been no empirical investigation concerning the
volatility of output and prices at sectoral levels under alternative exchange rate regimes. This is the
subject of the empirical exercise presented in the remainder of the paper.
15
3. Empirical analysis
In the empirical part of the paper we focus on selected countries in CEE. There are several
reasons why we chose the transition economies of CEE. Firstly, CEE countries exhibit diversity of
exchange rate regimes. At the same time they are characterised by many common features, such
as a similar political and economic legacy, EU accession process, trade patterns, economic
structure, etc. These factors make controlling, the region-specific effects, less important than would
be the case in large panels of more diversified countries. Secondly, the choice was also partly
determined by availability of sectoral data needed for the study. Thirdly, there exists relatively less
empirical research on transition economies than on industrialised countries.
3.1. Classification of exchange rate systems
Prior to testing any differences among exchange rate regimes it is necessary to classify the
observed exchange rate regimes. This is not a straight-forward exercise and several approaches
are possible. One possibility is to follow a de jure classification based on the public declarations
made by each countries’ authorities. Such information is also provided, for instance, by the IMF
(1999), which distinguishes eight categories, ranging from the adoption of another currency
(dollarisation, euroisation) to an independent float. However, it is widely known that de facto
regimes often differ from officially declared regimes (Calvo and Reinhart, 2002).
The differences between declarations and reality have stimulated efforts to find algorithmic
methods for classifying regimes based on available data (usual parameters include nominal
exchange rates, international reserves and interest rates). Levy-Yeyati and Sturzenegger (2002)
undertook such an approach and have provided a classification of exchange rate regimes for a
large set of countries. One should also, however, acknowledge that such methods have drawbacks
and in particular cases may fail to provide realistic descriptions. Given the relatively small and
specific sample of countries used in our research, combined with the need for classification on a
quarterly basis, we propose our own classification that comprises three exchange rate regimes
(fixed exchange rate, intermediate and floating exchange rate regime). This classification utilises
both official declarations made by monetary authorities (an overview can be found in de Souza,
2002) and data from Levy-Yeyati and Sturzenegger (2002). In our opinion, such an approach
provides a good description of the de facto exchange rate regimes pursued by the group of CEE
countries in the period 1994-2002. The detailed classification can be found in Appendix I.

3.2. Data description
16
For the purpose of empirical analyses a quarterly database for seven CEE countries has been
constructed, including data on the Czech Republic, Estonia, Hungry, Lithuania, Poland, Slovenia
and the Slovak Republic and spanning from the first quarter of 1993 to the last quarter of 20028.
The economic indicators covered include: value added in constant prices in tradables and
nontradables; the corresponding price indices of tradables and nontradables; industrial production;
real and nominal effective exchange rates; domestic interest rates; as well as foreign demand
proxied by GDP in the EU and the index of world commodity prices. All variables, except interest
rates, were transformed to normalised constant-base indices (100=base period), seasonally
adjusted using the Eurostat (2002) ARIMA X-12 procedure and transformed into logarithms.
Detailed information on data sources and transformation, as well as a broader discussion on data-related
issues, can be found in Appendix II.
3.3. Stylised facts – descriptive statistics
One of the well established stylised facts about industrialised countries points to a significant
increase in the volatility of nominal and real exchange rates in the post-Bretton Wood period that
was not, however, accompanied by increased volatility in other macroeconomic variables (see
Section 2.5) Thus, prior to model estimations, descriptive statistics of several macroeconomic
variables under the pegged and floating exchange rate regimes are analysed with a view to
investigating potentially similar patterns in our database. The results are presented in Table 1.
The first, puzzling, observation from Table 1 is that while the real effective exchange rates
seem to be more volatile under the floating exchange rate regime, the nominal effective exchange
rates tend to exhibit somewhat higher (at least not lower) volatility under the fixed exchange rate
regime. This appears to contradict what might have been expected intuitively as well as evidence
from industrialised economies. The tests for equality of variance9 in the case of nominal effective
exchange rates give mixed results, whereas in the case of real effective exchange rates the null
hypothesis is rejected. The explanation of high volatility and high average growth rate in NEER for
fixed exchange rates is that it is primarily driven by exchange rate dynamics vis-à-vis the Russian
rouble. The Russian currency and inflation were extremely volatile until 2000. For Estonia and
Lithuania, two countries that maintained fixed regimes over the entire period under investigation,
the shares of foreign trade with Russia, and in turn the shares of the rouble in their currency
baskets10, were high – at least until 1998. In the case of countries with floating exchange rate
regime this effect was less acute due to smaller trade exposure. An interesting lesson that
emerges from this analysis is that pegged exchange rates do not shelter a country from NEER (or
8 The length of particular time series varies significantly among countries and variables.
9 Test statistics (F-test, Barlett, Levine and Brown-Forsythe tests) are not reported but available upon request.
10 REER and NEER are trade weighted indices.

REER) volatility if the country trades with economies using currencies that are volatile against the
domestic country’s anchor currency (e.g. the US dollar and the euro).
Table 1. Average growth rates and volatilities under the peg and floating
17
VA VA_T VA_NT NEER REER PT PNT
Average growth under
peg (%)
0.5 0.6 0.6 0.9 0.6 0.7 1.0
Average growth under
float (%)
0.3 0.2 0.4 0.2 0.5 0.3 0.5
Standard deviation
under peg (%)
0.76 1.72 0.80 1.66 1.14 1.36 1.40
Standard deviation
under float (%) 0.32 1.38 0.60 1.24 1.45 1.05 0.72
Ratio of standard
0.42 0.8 0.75 0.75 1.28 0.77 0.52
deviations (float/peg)
Source: Authors’ calculations.
Note: Growth rates in percentages, changes quarter-on-quarter. Country-period observations were classified as fixed or
floating exchange rate regime according to Table A. 1. For the description of abbreviations see Appendix II.
In terms of comparison of statistical properties of output and prices, simple measures of
average growth and volatility can be misleading since time-specific and country-specific factors
may bias the results. The evidence based on these measures is, however, that fixed exchange rate
regimes were characterized by faster growth in output and inflation, as well as by higher volatility of
these variables. The latter effect can be explained to some extent by higher average growth rates
among fixers. Indeed, when volatilities of sectoral output and prices growth relative to aggregate
output and price growth were calculated, differences between regimes vanished (see Table 2).
With the exemption of nontradable prices, where tests give mixed results, in all other cases the null
hypothesis of variance equality between exchange rate regimes cannot be rejected at reasonable
significance levels. This result is similar to that for industrialised countries where exchange regimes
do not seem to affect the behaviour of macroeconomic variables other than the exchange rate. Our
results suggest that this might also be the case at the sectoral level.
Table 2. Volatility of sectoral output and prices relative to aggregate output and price under
the peg and floating
T output NT output T prices NT prices
Fixed exchange rate regime 1.5 0.7 1.0 0.8
Floating exchange rate regime 1.3 0.5 1.0 0.5
Note: The table reports standard deviations for variables defined as a difference between quarterly growth rates (in
percent) of total value added (value added deflator) in tradables (T) and nontradables (NT) and growth rates of
aggregate value added (deflators).

3.4. Methodology and results
18
The task of empirical testing for differences in the dynamics of tradable and nontradable output
between different exchange rate regimes can be divided into two parts. One refers to finding the
appropriate framework for modelling the behaviour of output in the two sectors, and the second
refers to testing the differences among exchange rate regimes in this framework. As far as the
former issue is concerned, the natural approach is to employ a VAR methodology. VAR models are
useful for investigation of historical data dynamics in an economy and do not require a formal
specification of the underlying theoretical model. In principle, this approach boils down to a choice
of appropriate variables and length of lags (Enders, 1995). In the exchange rate regime context,
the VAR framework was recently utilised by Broda (2002) in his quest to detect differences in
reactions of macroeconomic variables to terms-of-trade shocks between exchange rate regimes.
As far as testing of differences among exchange rate regimes is concerned, panel techniques
are natural. Pooling data on countries with one exchange rate regime allows one to gain insights on
common trends in this regime, whereas pooling all regimes together facilitates testing of differences
among regimes. Nevertheless, formal tests of the differences are often not feasible and hence not
attempted. For instance, Broda’s (2002) focus is on comparing the impulse response functions rather
than deriving formal tests. This simplified approach arises primarily from problems with combining
VAR analysis and panel data techniques which complicate the designing of formal tests.
The empirical part of this paper sets off with an approach similar to Broda (2002). As the
analysis in Section 2 indicates, under specific assumptions, differences among exchange rate
regimes in the short-run can be expected due to varying degrees of accommodation to shocks.
Thus, the derivation of impulse response functions in the VAR framework is an appropriate
approach. In particular, the objective is an assessment of differences in output response in
tradables and nontradables to various shocks hitting the economy. For this purpose a VAR system
capable of incorporating a wide spectrum of shocks has been constructed. The following is the
reduced form of the model:
A(L)* Yt = t (5)
where Yt is a vector of system variables (comprising output in tradables and nontradables, the real
effective exchange rate, relative prices of nontradables and tradables, the interest rate, GDP
growth in the EU, and world commodity prices), A(L) is lag polynomial and t is a vector of error
terms.
The choice of the variables in (5) is designed so as to incorporate the most important
indicators determining supply and demand of goods and services in open economies. The
objective is to draw a general picture consistent with basic macro models. Variables used in the
model include output in tradables and nontradables (proxied by value added in industry and value
added in services sectors), corresponding relative prices, real effective exchange rates and

Studies Analyses CASE No. 248 – Exchange Rate Regimes and the Real Sector….
nominal interest rates (for more detailed discussion see Appendix II). Inclusion of the EU’s GDP
and commodity prices facilitates control over potentially important exogenous shocks. Commodity
prices are treated as a proxy for terms of trade11. In our view, this set of variables enables
investigation of a rich structure of relationships.
An important and debated issue in the VAR methodology is whether stationary variables
should be used. This, at a high level of generality, boils down to the trade-off between less
efficiency against less information. For example, Sims (1980) argues that variables should not be
differentiated even if they contain unit roots. This recommendation has been followed in many
empirical studies – see for instance Ramaswamy and Slok (1998). On the other hand, other
authors suggest that if variables are neither non-stationary nor cointegrated, a VAR system should
be estimated in first differences. However, if variables are cointegrated through r cointegrating
relationships, the system should be modelled as a VAR in the r stationary combinations and (n-r)
differences of I(1) variables (Maddala and Kim, 1998).
A visual inspection of the data as well as standard time series and panel unit root tests (i.e.
augmented Dickey-Fuller and Pedroni (1999)) indicate that most of the considered variables are
nonstationary. Given the focus on testing differences in reactions to shocks between sectors and
exchange rate regimes in this exercise, the VAR models are estimated in first differences. We are
concerned more about efficiency than potential loss of information. Estimations in levels with
imposed wrong long-term restrictions could be equally undesirable for the results. In the light of
many reservations with regard to unit root and cointegration tests this would be very likely. The unit
root tests suffer from low power and cointegration analysis in a model containing both I(0) and I(1)
variables is problematic (Maddala and Kim, 1998). Moreover, as Campbell and Perron (1991) have
argued, distinguishing a process that exhibits cointegration from this that does not, could prove
very difficult.
The VAR model as defined in system (5) is estimated separately for each exchange rate
regime according to the classification provided in Appendix I. Given the short time series for
countries within each exchange rate regime, the condition of homogeneity of coefficients between
countries is imposed (i.e. it is assumed that country-specific effects do not affect the short-term
dynamics under a particular exchange rate regime)12. This is admittedly a restrictive assumption. It
could not, however, be avoided given the nature of available data. This consideration led us to
estimate VAR models with two lags13.
In order to obtain the structural form of system (5) and to draw impulse response functions we
applied the Choleski decomposition method, which provides a minimal set of necessary
11 Data on the terms of trade was not readily available for all countries in the sample. However, the dynamics of the
19
commodity prices well mimics the dynamics of terms of trade in countries where such data was available.
12 Equations with varying constant coefficients have been also estimated.
13 There is no commonly agreed method of selecting the lag length in VAR models. Usually it is recommended to
use between one to three full years of observations as a too small lag order may hinder capturing the true data dynamics
and relationships among variables. Given the number of observations and regressors in our models this was not feasible.
Thus, we perceive the two-lag structure as the only feasible trade-off between securing sufficient degrees of freedom and
misspecification of the model.

Studies Analyses CASE No. 248 – Przemyslaw Kowalski, Wojciech Paczynski, Lukasz Rawdanowicz
restrictions. As this method is sensitive to the ordering of variables in a system, the ordering was
carefully chosen so as to reflect the expected precedence among variables.
20
The estimation of system (5) for each exchange rate regime rendered results that are generally
in line with theoretical expectations. However, impulse response functions are not statistically
significant. The results are sensitive to the inclusion of observations for the Slovak Republic, which
were partly classified as a floating exchange rate regime and partly as a fixed one. In the case of
the Slovak Republic, data quality problems were expected: data on value added in tradables were
subject to unusually high volatility and the existence of outliers. After the exclusion of this country
from the sample, the VAR analysis proved to be less sensitive to model specification. We
experimented with various model compositions and definitions of the variables listed under the
specification of system (5) (see also Appendix II).
The lack of statistical significance was judged based on standard error bands and general poor
fit (see Appendix III for examples of impulse response functions). Most coefficients were not
significantly different from zero, a block exogeneity test could not be rejected and R2 was usually
very low. In judging the significance of variables, the potentially severe problem of multicollinearity
had to be taken into account. Given the lack of significant results, no robust inferences could be
made either about output dynamics of tradables and nontradables or about differences between
regimes.
The lack of statistically significant results may have many sources. Firstly, data quality and
short times series are a problem. In general, value added data in many instances exhibited
surprisingly high quarterly volatility, even after seasonal adjustment, and in some instances there
were reasons to expect significant outliers. In addition, the large number of variables in the
specified VAR systems relative to time series length did not allow testing for higher lag length,
which could have had important bearings on the results. There is also a more general problem of
short-term dynamics in VAR. In practice, it is often found that unrestricted VAR models yield very
erratic estimates. In addition, VAR models are in general known for their overparametrisation and
low precision of coefficient estimates (Maddala and Kim, 1998). Finally, we cannot preclude that
variables were cointegrated for some model specifications and the estimation in differences
ignored important information.
This last consideration led us to investigation of short-term adjustments in the error correction
framework which allows for possible long-run relationships. For this purpose we employed the
pooled mean group (PMG) estimation procedure as proposed by Pesaran et al. (1999). This
method allows for heterogeneous short-run dynamics, which implies relaxing a restrictive
assumption adopted in the VAR exercise. Besides, as Pesaran et al. (1999) note, it is usually more
reasonable to expect long-term relationships to be similar across countries than in the case of
short-run relationships. This may stem from common forcing factors such as budget and solvency
constraints, technology and arbitrage.
The PMG estimator for heterogeneous panels allows one to estimate autoregressive
distributed lag (ARDL) models as well as their error correction representations with an explicit

estimation of long-run relationships. For each group in a panel, an ECM model is estimated with
homogeneity restriction imposed on long-run coefficients, whereas short-run coefficients are
allowed to vary and are averaged across groups. Thus, a PMG estimator can be viewed as an
intermediate approach between the mean group estimator (the average of results from separate
estimates for each group) and fixed/random effects panel models which allow only an intercept to
vary across groups - all other coefficients and error variances are constrained to be homogenous
(Pesaran et al., 1999). It is often found that a single-country estimation gives insignificant and
economically implausible results, but mean group or pooled estimates of long-run coefficients tend
to be sensible. This could be attributable to the fact that group-specific estimates are biased due to
sample-specific omitted variables or measurement errors that are correlated with the regressors
(Pesaran et al., 1999).
In estimating error correction terms we focused primarily on the relationships between output
in tradables and nontradables, exchange rates and relative prices. Different specifications for each
exchange rate regime were tested separately. A PMG is a single-equation approach (as opposed
to the multivariate VAR approach adopted beforehand) and thus normalisations with regard to both
tradables and nontradables were tested. As in the VAR approach, we experimented with various
definitions of variables. In the case of tradables, we also included EU GDP as a proxy for foreign
demand. Specifications of the EU’s GDP as a long-run forcing variable (but without imposing
homogeneity among countries) and as a lagged exogenous variables were tested.
As Pesaran et al. (1999) demonstrated, the PMG estimator is usually robust to outliers and the
choice of lag order. The latter is identified based on one information criterion. We tested models
with maximum two lags and used the Schwarz-Bayes criterion (SBC), which determined the order
of lags for the ARDL model in each country.
Generally, the statistical properties of the results obtained from PMG estimations proved to be
sensitive to definitions of variables and model specification14. Observations for the Slovak Republic
were excluded from estimations. In the case of tradables, the models with reasonable statistical
properties that could be compared between exchange rate regimes suggest that there are small
discrepancies in the speed of adjustment to long-run equilibrium (the error correction term) – see
Table 3. It is noteworthy that, overall, the adjustment seems to be fast. This seems to be consistent
with the relatively short lives of the impulse response functions in the estimated VAR models (see
Appendix III).
14 Estimations were performed using the GAUSS procedure written by Pesaran et al. (1999)
21
http://www.econ.cam.ac.uk/faculty/pesaran/jasa.exe

Table 3. Results of PMG estimations for tradables
22
Floating exchange rate regime Fixed exchange rate regime
Dependent variable: T (value
added in tradables) Coefficients t-statistic Coefficients t-statistic
Model 1
NT -0.220 -0.5 0.984 6.8
REER -0.455 -12.7 -0.382 -4.8
GDP_EU* 2.466 11.0 1.239 3.6
Phi -0.675 -2.1 -0.661 -1.9
Model 2
NT 1.804 6.3 0.807 3.2
REER -0.305 -10.2 -0.304 -3.7
Phi -0.599 -1.5 -0.478 -4.6
Model 3
GDP 1.581 6.6 0.815 7.3
REER -0.403 -5.8 -0.179 -2.8
GDP_EU* 0.346 2.1 1.176 3.8
Phi -0.746 -2.9 -0.682 -2.1
Notes: Detailed models’ specifications and results are given in Appendix III. ‘*’ denotes long-run coefficients that were
unrestricted across countries. Phi is the error correction term.
In the case of nontradables, results are less robust in terms of statistical properties and more
sensitive to alternative definitions of variables. Thus, inference concerning nontradables dynamics
or possible differences between exchange rate regimes is less compelling. For instance, in the
model, for which results are presented in Table 4, signs of relative price elasticities (prices of
nontradables to prices of tradables) are different under floating and fixed exchange rate regimes.
This could be indicative of results being driven by particular events (such as crises) or outliers in
the data.
Table 4. Results of PMG estimations for nontradables
Dependent variable: NT
(value added in nontradables) Floating exchange rate regime Fixed exchange rate regime
Coefficients t-statistic Coefficients t-statistic
Model 1
GDP 0.908 6.7 1.031 29.0
RP2 -0.318 -2.1 0.174 3.2
Phi -0.296 -1.3 -0.487 -5.6
Notes: Detailed models’ specifications and results are given in Appendix III. Phi is the error correction term.

Overall, the presented empirical analyses do not identify any significant differences in impact
of exchange rate regimes on the short-run dynamics of tradables and nontradables. However,
these findings must be treated with caution given the conceptual and technical problems with their
derivation. Apart from the problem of short-time series and the limitations of a particular
econometric estimation technique, two important conceptual considerations are worth
emphasising.
Firstly, the results could be affected by the fact that the considered panel of country-regime
observations was unbalanced. This may have led to biased results in cases where common group
effects were not accounted for in equations. In addition, in some countries (for instance the Czech
Republic, Poland, Hungary – see Appendix I) frequent changes in exchange rate regimes relative
to the sample span lead to reservations about the possibility of detecting differences among
exchange rate regimes. It is reasonable to suppose that there might have been expectations of
regime change before it actually occurred. On the other hand, economic agents or economic
relationships adjust to changes in exchange rate regimes not instantaneously but with some time
lag. The importance of such behavioural arguments for the case of corner regimes in the
considered sample is somewhat reduced, since periods of fixed and floating exchange rates were
usually separated by a period of an intermediate regime. In this context, results for countries with
fixed exchange rate regimes could be more robust: Estonia and Lithuania maintained their
currency board arrangements during the whole period under investigation.
Figure 1. NEER in Lithuania, 1995-2001 (January 1995=1)
3,5
3,0
2,5
2,0
1,5
1,0
0,5
09/1996
01/1997
05/1997
09/1997
01/1998
09/1998
01/1999
05/1999
05/2000
09/2000
12
10
8
6
4
2
Source: Authors’ calculations based on Bank of Lithuania data.
Notes: NEER (LHS axis) – nominal effective exchange rate (trade weighted – all trade partners); NEER EU (LHS axis) –
NEER calculated versus the EU; NEER CIS (RHS axis) – NEER calculated versus the CIS; The litas was pegged
to the US dollar throughout the whole period.
Secondly, in the investigated sample there is no significant difference in the NEER stability
between fixed and floating regimes. Technically, the fixed regime in our sample reflects only the
fact that the domestic currency was pegged to a single foreign currency, thus not precluding the
23
0,0
01/1995
05/1995
09/1995
01/1996
05/1996
05/1998
09/1999
01/2000
01/2001
05/2001
09/2001
01/2002
0
NEER EU NEER NEER CIS (rhs axis)

volatility against other foreign currencies – including those of important trading partners. This is the
case with Lithuania. During the period January 1994 – February 2002, the litas was pegged to the
US dollar. However, Lithuania trades mostly with the EU, the FSU and other Baltic countries.
Figure 1 depicts developments in the country’s NEER versus the EU, the FSU and all trading
partners together. The implications of this commonly overlooked fact may be far reaching. In
contrast to the usually considered theoretical two-country or small open economy models, in reality
existing fixed exchange rate regimes do not necessarily insulate against exchange rate volatility.
The distinction between pure float and pure fix is blurred, with implications for the capability of
empirical work to capture the differential characteristics between regimes. Clearly, there are other
mechanisms that are likely to determine the influence of exchange rate regimes on domestic
economic activities. In particular macroeconomic policy design and institutional framework play a
major role in macroeconomic development and the stability of any given economic situation.
Nevertheless, it is extremely difficult to identify relationships of this sort empirically.
4. Conclusions
24
This paper investigated the impact of exchange rate systems on the real economy from a
sectoral perspective. The majority of theoretical and empirical investigations to date on exchange
rate regimes have dealt primarily with aggregate variables and paid little attention to sectoral
issues. This is surprising given the important distinction between tradables and nontradables and
the consequences of this on theoretical models as well as the functioning of economies.
The detailed critical survey of both theoretical and empirical literature dealing with the sectoral
dimension of the exchange rate regime debate demonstrates that there is no commonly accepted
theory of exchange rate systems. Predictions of existing models depend crucially on specific
assumptions concerning in particular the parameters of the utility function, the nature of the price
adjustment process, characteristics of analysed shocks and openness of economies. This applies,
in particular, to NOEM models, which are viewed as superior to Mundell-Fleming type of models
and as more appropriate for explicit sectoral analysis.
The VAR and PMG estimations conducted for selected CEE countries did not identify robust
evidence of differences between exchange rate systems in terms of reactions to shocks for the
tradable and nontradable sectors. We interpret this finding as a confirmation of results for
developed countries, where the real effects of choosing exchange rate regimes are hardly
detectable at the aggregate level. This is interesting since it is expected that if real economy
differences between exchange rate regimes exist they should be more pronounced at a
disaggregated level.
The lack of strong results demonstrating real economy differences between exchange rate
regimes may also be due to technical and conceptual difficulties encountered in this type of
research. Among the technical difficulties, we should mention the short-time series, the small
number of countries in the group and limitations of econometric techniques in applied research. As

far as conceptual problems are concerned, exchange rate regime classification, frequent changes
in exchange rate systems and lack of differences in exchange rate volatility among different
regimes should be pointed out.
This paper demonstrates that the countries under investigation that followed pegged regimes
suffered from equally high (or even higher) volatility of effective nominal exchange rates than
countries with floating regimes. Pegging a domestic currency to a single foreign currency does not
necessarily insulate against exchange rate variability versus other foreign currencies. This finding
in our country group contrasts with the theoretical two-country or small open economy models and
could be a major obstacle in detecting empirically differences among exchange rate regimes.
In terms of further research, a closer and a more explicit empirical investigation of the
underlying assumptions of different theoretical models on the exchange rate regime would be
desirable. This could strengthen the link between empirical and theoretical research and contribute
to more valuable policy recommendations. Testing theories on exchange rate regimes is only
appropriate if the underlying assumptions are met. In addition, a deeper analysis of the importance
of various shocks could also be interesting in this respect.
As far as policy conclusions are concerned, the lack of any serious impact of the exchange
rate regime on the real economy suggests that the focus of the debate on optimal choice of
exchange rate systems should concentrate on consistency of the monetary policy framework
instead of restating standard arguments in favour of either fixing or floating exchange rate regimes.
This debate should also take into account existing institutional arrangements and credibility issues.
It is evident that exchange rate regime fit, be it peg or floating, to the overall macroeconomic
framework matters for countries’ development and stability prospects.
25

References
Baxter, Marianne and Alan C. Stockman (1989), ‘Business Cycles and Exchange Rate Regime:
26
Some International Evidence’, Journal of Monetary Economics, Vol. 23, pp. 377-400
Broda, Christian (2002), ‘Terms of Trade and Exchange Rate Regimes in Developing Countries’,
forthcoming in Journal of International Economics
Calvo, Guillermo (1999a), ‘Fixed Versus Flexible Exchange Rates. Preliminaries of a Turn-of-
Millennium Rematch’, mimeo, University of Maryland
Calvo, Guillermo (1999b), ‘On Dollarization’, mimeo, University of Maryland
Calvo Guillermo and Michael Kumhof (2002), Trade Openness and Exchange Rate Misalignments,
mimeo, University of Maryland
Calvo, Guillermo and Reinhart, Carmen, M. (2002), ‘Fear of floating’, Quarterly Journal of
Economics 117 (2), pp. 421-446
Campbell, John Y. and Pierre Perron (1991), ‘Pitfalls and Opportunities: What Macroeconomists
Should Know about Unit Roots’, in O.J. Blanchard and S. Fisher (eds.), NBER
Macroeconomics Annual, The MIT Press, pp. 141-201
Dedola, Luca and Sylvain Leduc (2002), ‘Why Are Business Cycles Alike Across Exchange Rate
Regimes?’, Federal Reserve Bank of Philadelphia Research Department Working Papers, No.
02-11, April
Dehejia, Vivek H. and Nicholas Rowe (2001), ‘Macroecnomic Stabilisation: Fixed Exchange Rates
vs Inflation Targeting vs Price Level Targeting’, mimeo, Carleton University
Devereux, Michael B. and Philip R. Lane (2001), ‘Exchange Rates and Monetary Policy in
Emerging Market Economies’, mimeo, preliminary version, June
De Grauwe, Paul (1997), International money – Postwar Trends and Theories, Oxford University
Press
de Souza, Lucio Vinhas, ‘Integrated Monetary and Exchange Rate Frameworks: Are There
Empirical Differences?’, mimeo, Bank of Estonia
Dornbusch, Rudiger (1976). ‘Expectations and Exchange Rate Dynamics’, Journal of Political
Economy, Vol. 84 (6), pp. 1161-1176
Enders, Walter (1995), Applied Econometric Times Series, John Wiley Sons, Inc.
Eurostat (2002), Demetra 2.0, advanced seasonal adjustment interface,
http://forum.europa.eu.int/irc/dsis/eurosam/info/data/demetra.htm
Fisher, Stanley (2001), ‘Exchange Rate Regimes: Is the Bipolar View Correct?’, High-Level
Seminar on Exchange Rate Regimes: Hard Peg or Free Floating?, IMF Headquarters,
Washington, D.C., January
Flood, Robert P. and Andrew K. Rose (1995), ‘Fixing Exchange Rates. A Virtual Quest for
Fundamentals’, Journal of Monetary Economics, Vol. 36, pp. 3-37
Friedman, Milton (1953), ‘The case for flexible exchange rates’, in Friedman, M. (ed.) Essays in
Positive Economics, University of Chicago Press

Harris, Richard G. (2002), ‘Asymmetric Shocks and Productivity Dynamics in a Core-Periphery
27
Integration’, mimeo, Simon Fraser University
Hau, Harald (2000), ‘Exchange rate determination: The role of factor price rigidities and
nontradeables’, Journal of International Economics, Vol. 50, pp. 421–447
Hau, Harald (2002), ‘Real Exchange Rate Volatility and Economic Openness: Theory and
Evidence’, Journal of Money, Credit and Banking, Vol. 34, No. 3, pp. 611-630
International Monetary Fund (1999), Exchange Rate Arrangements and Currency
Convertibility: Development and Issues, Washington DC
Lane, Philip R. (2001), 'The new open economy macroeconomics: a survey', Journal of
International Economics, No. 54, p. 235-266
Lane, Philip R. and Giovanni Ganelli (2002), ‘Dynamic General Equilibrium Analysis: The Open
Economy Dimension’, Trinity College, mimeo (forthcoming in S. Altug, J. Chaddha, C. Nolan,
eds., Elements in Dynamic Macroeconomic, Cambridge University Press)
Levy-Yeyati, Eduardo and Federico Sturzenegger (2002), ‘Classifying Exchange Rate Regimes:
Deeds vs. Words’, mimeo, Universidad Torcuato Di Tella
Maddala, G.S. and In-Moo Kim (1998), Unit Roots, Cointegration, and Structural Change,
Cambridge University Press
McKinnon, Ronald (1963), ‘Optimum Currency Areas’, American Economic Review, No. 51, pp.
509-517
Mundell, Robert A. (1960), ‘The Monetary Dynamics of International Adjustment under Fixed and
Flexible Exchange Rates’, Quarterly Journal of Economics, Vol. 74, pp. 227-257
Mussa, Michael, (1986) ‘Nominal Exchange Rate Regimes and the Behavior of Real Exchange
Rates’, Carnegie-Rochester Conference Series on Public Policy, No. 25, pp. 117-213
Obstfeld, Maurice and Kenneth Rogoff (1996), Foundations of International Macroeconomics, MIT
Press
Obstfeld, Maurice, and Kenneth Rogoff (2000), ‘New directions in Open Macroeconomics’, Journal
of International Economics 50, pp. 117-153
Pedroni, Peter (1999), ‘Critical Values for Cointegration Tests in Heterogeneous Panels with
Multiple Regressors’, Oxford Bulletin of Economics and Statistics, Vol. 61, pp. 653-670
Pesaran, M. Hashem, Yongcheol Shin, and Ron P. Smith (1999), ‘Pooled Mean Group Estimation
in Dynamic Heterogeneous Panels’, Journal of the American Statistical Association, 94, 1999,
pp. 621-634
Ramaswamy, Ramana and Torsten Slok (1998), ‘The Real Effect of Monetary Policy in the
European Union: What are the differences?’, IMF Staff Papers, Vol. 42, No. 2, June
Sims, Christopher (1980), ‘Macroeconomics and Reality’, Econometrica, Vol. 48, pp. 1-48
Stockman, Alan, (1999), ‘Choosing an exchange rate system’, Journal of Banking and Finance,
Vol. 23, pp. 1483-1498
Stockman, Alan (2000), ‘Exchange Rate Systems in Perspective’, Cato Journal, Vol. 20, No. 1,
pp.115-122

Taylor, John (1993), Macroeconomic Policy in a World Economy: From Economic Design to
28
Practical Operation, N.Y: Norton
Tille, Cédric (2002), ‘How Valuable Is Exchange Rate Flexibility? Optimal Monetary Policy under
Sectoral Shocks‘, Federal Reserve Bank of New York, Staff Reports No. 147
Vartiainen, Juhana (2002), ‘Relative wages in monetary union and floating’. Scandinavian Journal
of Economics, Vol. 104, pp. 277-288
Woodford, Michael (2002), ‘Inflation Stabilization and Welfare’, preliminary manuscript of Chapter 6
of the book Interest and Prices: Foundations of a Theory of Monetary Policy,
http://www.princeton.edu/~woodford

29
Appendix I – Classification of exchange rate regimes
Table A. 1. Classification of exchange rate regimes, 1Q94-4Q02
Czech
Republic Estonia Hungary Lithuania Poland Slovakia Slovenia
Q11994 0 0 … … … 0 …
Q21994 0 0 … … … 0 …
Q31994 0 0 … … … 0 …
Q41994 0 0 … … … 0 …
Q11995 0 0 0.5 0 0.5 0 0.5
Q21995 0 0 0.5 0 0.5 0 0.5
Q31995 0 0 0.5 0 0.5 0 0.5
Q41995 0 0 0.5 0 0.5 0 0.5
Q11996 0 0 0.5 0 0.5 0 0.5
Q21996 0.5 0 0.5 0 0.5 0 0.5
Q31996 0.5 0 0.5 0 0.5 0 0.5
Q41996 0.5 0 0.5 0 0.5 0 0.5
Q11997 0.5 0 0.5 0 0.5 0 0.5
Q21997 0.5 0 0.5 0 0.5 0 0.5
Q31997 1 0 0.5 0 0.5 0 0.5
Q41997 1 0 0.5 0 0.5 0 0.5
Q11998 1 0 0.5 0 0.5 0 0.5
Q21998 1 0 0.5 0 1 0 0.5
Q31998 1 0 0.5 0 1 0 0.5
Q41998 1 0 0.5 0 1 1 0.5
Q11999 1 0 0.5 0 1 1 0.5
Q21999 1 0 0.5 0 1 1 0.5
Q31999 1 0 0.5 0 1 1 0.5
Q41999 1 0 0.5 0 1 1 0.5
Q12000 1 0 0.5 0 1 1 0.5
Q22000 1 0 0.5 0 1 1 0.5
Q32000 1 0 0.5 0 1 1 0.5
Q42000 1 0 0.5 0 1 1 0.5
Q12001 1 0 0.5 0 1 1 0.5
Q22001 1 0 1 0 1 1 0.5
Q32001 1 0 1 0 1 1 0.5
Q42001 1 0 1 0 1 1 0.5
Q12002 1 0 1 0 1 1 0.5
Q22002 1 0 1 0 1 1 0.5
Q32002 1 0 1 0 1 1 0.5
Q42002 1 0 1 0 1 1 0.5
Source: Authors’ elaboration based on declared de jure and observed de facto regimes.
Notes: ‘0’ denotes fixed exchange rate regime, ‘0.5’ – intermediate regime, ‘1’ – floating regime.

Appendix II – Data
30
Our quarterly database covers seven Central and East European countries: the Czech
Republic (CZE), Estonia (EST), Hungary (HUN), Lithuania (LIT), Poland (POL), Slovenia (SLO),
and the Slovak Republic (SLK). It spans from the first quarter of 1993 to the last quarter of 2002.
The length of time series differs significantly among countries and particular variables (see below).
The first data-related challenge of our research deals with definition of tradables and
nontradables and measures of their dynamics. As Obstfeld and Rogoff (1996) noted, the usual
identification of tradables (i.e. goods that can be traded internationally) with goods, and
nontradables with services is too crude. However, in practice when dealing with aggregated data in
macroeconomic analyses no good solution to this problem seems to exist. One general and
interpretable measure of the dynamics in these sectors is output. In the case of tradables it can be
measured in terms of the industrial production index, however, no such measure exists for
aggregated services. Therefore, we decided to use constant-prices value added indices for
specified sectors. We assumed industry to be the tradable sector, whereas construction, market
and nonmarket services represent the nontradables sector. It might be more appropriate to classify
some subsections of industry (like electricity, gas, and water supply) as nontradables and vice
versa for some services (e.g. part of transportation can and indeed is traded internationally).
However, data disaggregation did not allow for such division in all countries. In our classification,
we decided to exclude agriculture, as we believe the production of agricultural goods is driven
more by weather conditions and specific production cycles than by macroeconomic variables
analysed in our models.
Given our value added measures of output in tradables and nontradables we used
corresponding value added deflators as price indices (PT and PNT). However, in two cases (for
Slovenia and Hungary) deflators could not be calculated due to lack of current-price data and they
were proxied by consumer prices (as nontradables prices) and producer prices (as tradables prices).
The real effective exchange rates (REER) (i.e. exchange rates versus currencies of main
trading partners weighted by trade shares) reported by the IMF use the CPI as a relative price
measure. As consumer prices comprise a presumably large share of nontradables prices, such a
measure is not a perfect indicator for countries’ competitive position. In our view, a nominal
exchange rate deflated by tradables prices would be a preferred indicator. Therefore, we
calculated modified REER indices. These were constructed based on nominal effective exchange
rates for each country and domestic and foreign prices for tradables. Foreign prices were proxied
by a value added deflator in the euro-zone15 (RER2) or producer prices in the EU (RER1).
As for interest rates, we decided to focus on domestic rates (IR). In most cases, this was the 3-
month inter-bank interest rate, however, where data was not available, proxies of the average of
lending and deposits rates were used. The real interest rates were deflated with consumer prices
(RIR2). In our view, domestic interest rates are most relevant for determining supply and demand
15 No value added deflator was available for the EU in the tradable sector.

for tradables and nontradables in CEECs, as the access to foreign capital markets is limited and
the link between foreign and domestic interest rates in reality is far from simple. Moreover, the link
between domestic and foreign exchange rates differs between exchange rate regimes.
To take account for foreign demand we used constant-prices GDP in the five big EU
economies16 (GDP_EU5). Given the high shares of trade with the EU of all countries in our sample
(see Table A. 2) this is a fairly reasonable proxy.
Table A. 2. The EU shares in foreign trade, 1999-2002 (% of total exports and imports)
Czech
Republic Estonia Hungary Lithuania Poland Slovenia Slovakia
31
1999 67 68 70 49 67 68 55
2000 66 69 66 46 65 66 54
2001 66 62 66 47 65 65 55
2002 65 62 65 47 65 64 55
Source: Calculations based on Eurostat data.
For most countries, data on terms of trade was not readily available so we decided to proxy
terms of trade by the IMF commodity price index (COMM). Given the assumption that the countries
in our sample are price-takers in international markets (cf. e.g. Broda (2002)), this variable should
well reflect the changes in terms of trade of individual countries. An analysis performed for Poland,
where data were available, indeed indicated a high negative correlation between terms of trade
and the commodity price index.
All variables, apart from interest rates, were transformed to normalised constant-base indices
(100=base period), seasonally adjusted (using ARIMA X-12 procedure of Eurostat (2002)) and
transformed to logarithms. Detailed information on data sources and transformation as well as
broader discussion on data issues are provided below.
List of all variables (_xxx is a country identifier):
VA_xxx - Value added
T_xxx - Valued added in tradables (only industry)
NT_xxx - Valued added in nontradables (construction, market and non-market services)
NEER_xxx - Nominal effective exchange rate (increase – appreciation)
REER_xxx - Real effective exchange rate (increase – appreciation)
RIR2_xxx - Real interest rate
IR_xxx - Nominal interest rate
CPI_xxx - Consumer price index
PPI_xxx - Producer price index
DEF_xxx - Value added deflator
PT_xxx - Deflator for value added in tradables
16 No data for the entire data span was available for the EU.

PNT_xxx - Deflator for value added in nontradables
RP1_xxx - Relative domestic prices: nontradables/tradables in terms of CPI and PPI
RP2_xxx - Relative domestic prices: nontradables/tradables in terms of corresponding VA
32
deflators
RER1_xxx - NEER deflated with domestic and EU tradables prices in terms of the PPI
(PPI_xxx * NEER_xxx / PPI_EU5) increase – appreciation
RER2_xxx - NEER deflated with domestic and euro-zone tradables prices in terms of VA
deflators (PT_xxx * NEER_xxx / PT_EUR) increase – appreciation
COMM - commodity price index (all commodities), IMF
Country identifiers (_xxx)
CZE – the Czech Republic, EST – Estonia, HUN – Hungry, LIT – Lithuania, POL – Poland, SLO –
Slovenia, SLK – the Slovak Republic, EU5 – the EU, EUR – the euro zone.
Sources and definitions
Czech Republic (CZE)
Variable Period Source/definition/notes
VA, T, NT 1Q94-4Q02 Czech Statistical Office
DEF, PT, PNT 1Q94-4Q02 Czech Statistical Office
CPI / PPI 1Q93-4Q02 IFS – IMF
RP1 1Q93-4Q02 Authors’ calculations based on official data
NEER / REER 1Q93-4Q02 IFS – IMF; increase = appreciation
RER1 / RER2 1Q93-4Q02 Authors’ calculations based on official data
IR 1Q93-4Q02 Czech National Bank (CNB); 3M PIBOR
RIR2 1Q94-4Q02 CNB; 3M PIBOR deflated with the annual change in the CPI

33
Estonia (EST)
VA, T, NT 1Q93-4Q02 Statistical Office of Estonia
DEF, PT, PNT 1Q93-4Q02 Statistical Office of Estonia
NEER / REER 1Q93-4Q02 Bank of Estonia; increase = appreciation
RER1 / RER2 1Q93-4Q02 Authors’ calculations based on official data
IR 2Q93-4Q02 IFS – IMF; average of deposit and lending rates
RIR2 1Q94-4Q02 IFS – IMF; average of deposit and lending rates deflated with the
annual change in the CPI
Hungary (HUN)
VA, T, NT 1Q95-4Q02 Hungarian Central Statistical Office
NEER / REER 1Q93-4Q02 IFS – IMF, increase = appreciation
RER1 1Q93-4Q02 Authors’ calculations based on official data
Lithuania (LIT)
VA, T, NT 1Q94-4Q02 Department of Statistics to the Government of the Republic of
Lithuania
DEF, PT, PNT 1Q94-4Q02 Department of Statistics to the Government of the Republic of
Lithuania
NEER / REER 3Q93-4Q02 Bank of Lithuania; (increase = appreciation)
IR 1Q94-4Q02 IFS – IMF; money market interest rate
RIR2 1Q94-4Q02 IFS – IMF; money market interest rate deflated with the annual
change in the CPI

Poland (POL)
VA, T, NT 1Q95-4Q02 Polish Central Statistical Office (CSO);
DEF, PT, PNT 1Q95-4Q02 Central Statistical Office (CSO)
CPI / PPI 1Q93-4Q02 CSO
NEER / REER 1Q93-4Q02 IFS – IMF; increase = appreciation
IR 1Q93-4Q02 National Bank of Poland (NBP); 3M WIBOR
RIR2 1Q94-4Q02 NBP; 3M WIBOR (deflated with the annual change in the CPI)
Slovak Republic (SLK)
VA, T, NT 1Q94-4Q02 Statistical Office of the Slovak Republic
DEF, PT, PNT 1Q94-4Q02 Statistical Office of the Slovak Republic
NEER / REER 1Q93-4Q02 IFS – IMF; (increase = appreciation)
34
Slovenia (SLO)
VA, T, NT 1Q95-4Q02 Statistical Office of the Republic of Slovenia
NEER / REER 1Q93-4Q02 Central Bank of Slovenia; (increase = appreciation)
IR 1Q93-4Q02 IFS – IMF; money market interest rates
RIR2 1Q94-4Q02 IFS – IMF; money market interest rates deflated with the annual
change in the CPI

35
Other
PT_EUR, PN_ EUR 1Q93-4Q02 ECB, Monthly Bulletin
CPI_EU5, PPI_EU5 1Q93-4Q02 Main Economic Indicators, OECD
COMM 1Q93-4Q02 IFS – IMF

Appendix III – Estimation results
1. Impulse response functions
36
The impulse response functions were derived for VAR models as specified below with two
lags. Confidence bands are based on Monte Carlo simulations with 1000 repetitions. R1 stands for
floating exchange rate regime, R3 for fixed exchange rate regime, C is a constant.
Figure A. 1. Impulse response function for Specification 1
Floating exchange rate: VAR = D(R1_T) D(R1_NT) D(R1_REER) D(R1_GDP_EU) C
Fixed exchange rate regime): VAR = D(R3_T) D(R1_NT) D(R3_REER) D(R3_GDP_EU) C
Response to Cholesky One S.D. Innovations ± 2 S.E.
.006
.004
.002
.000
-.002
-.004
-.006
-.008
Response of D(R1_T) to D(R1_NT)
1 2 3 4 5 6 7 8 9 10 11 12
.006
.004
.002
.000
-.002
-.004
-.006
-.008
Response of D(R1_T) to D(R1_REER)
1 2 3 4 5 6 7 8 9 10 11 12
.006
.004
.002
.000
-.002
-.004
-.006
-.008
Response of D(R1_T) to D(R1_GDP_EU)
1 2 3 4 5 6 7 8 9 10 11 12
.012
.008
.004
.000
-.004
-.008
Response of D(R3_T) to D(R1_NT)
1 2 3 4 5 6 7 8 9 10 11 12
.012
.008
.004
.000
-.004
-.008
Response of D(R3_T) to D(R3_REER)
1 2 3 4 5 6 7 8 9 10 11 12
.012
.008
.004
.000
-.004
-.008
Response of D(R3_T) to D(R3_GDP_EU)
1 2 3 4 5 6 7 8 9 10 11 12

37
Figure A. 2. Impulse response function for Specification 2
Floating exchange rate regime: VAR = D(R1_NT) D(R1_GDP) D(R1_RP2) C
Fixed exchange rate regime: VAR = D(R3_NT) D(R3_GDP) D(R3_RP2) C
.005
.004
.003
.002
.001
.000
-.001
-.002
-.003
Response of D(R3_NT) to D(R3_GDP)
1 2 3 4 5 6 7 8 9 10 11 12
.005
.004
.003
.002
.001
.000
-.001
-.002
-.003
Response of D(R3_NT) to D(R3_RP2)
1 2 3 4 5 6 7 8 9 10 11 12
.006
.004
.002
.000
-.002
-.004
-.006
Response of D(R1_NT) to D(R1_GDP)
1 2 3 4 5 6 7 8 9 10 11 12
.006
.004
.002
.000
-.002
-.004
-.006
Response of D(R1_NT) to D(R1_RP2)
1 2 3 4 5 6 7 8 9 10 11 12

2. PMG results
38
The results of PMG estimations were derived based on the following parameters:
· SBC (Schwarz) has been used to select the lag orders for each group.
· The static fixed effects OLS estimates have been used as initial estimate(s) of the
long-run parameter(s) for the pooled maximum likelihood estimation.
· The number of included groups is N = 2.
The Newton-Raphson method was used to derive PMG estimates.
For the floating exchange rate regime: Group 1 is Poland and Group 2 the Czech Republic.
For the fixed exchange rate regime: Group 1 is Estonia and Group 2 Lithuania.

39
Model 1
Flexible exchange rate Fixed exchange rate
The number of time periods by groups are:
17 20
38 30
Orders of lags in the ARDL model: Orders of lags in the ARDL model:
Group 1 : 0 2 2 1 Group 1 : 1 1 0 0
Group 2 : 1 0 0 0 Group 2 : 0 1 1 1
Computations converged after 4 iterations. Computations converged after 5 iterations.
Dependent variable: T Dependent variable: T
Coef. St. Er. t-ratio Coef. St. Er. t-ratio
Long-Run Coefficients Restricted to be the Same
Across all Groups
Long-Run Coefficients Restricted to be the Same
Across all Groups
NT -0.220 0.415 -0.530 NT 0.984 0.144 6.833
REER -0.455 0.036 -12.655 REER -0.382 0.079 -4.821
Unrestricted Long-Run Coefficients Unrestricted Long-Run Coefficients
YEU 2.466 0.223 11.045 YEU 1.239 0.34 3.643
Error Correction Coefficients Error Correction Coefficients
Phi -0.675 0.325 -2.081 Phi -0.661 0.339 -1.947
Short-Run Coefficients Short-Run Coefficients
NT -0.149 0.071 -2.081 NT 0.650 0.334 1.947
REER -0.307 0.148 -2.081 REER -0.252 0.130 -1.947
YEU 1.738 0.951 1.827 YEU 0.934 0.645 1.448
dT(-1) 0.000 +DEN +DEN dT(-1) 0.000 +DEN +DEN
dNT 0.316 0.316 1.000 dNT 0.248 0.887 0.279
dNT(-1) 0.910 0.910 1.000 dNT(-1) 0.000 +DEN +DEN
dREER 0.178 0.178 1.000 dREER 0.222 0.222 1.000
dREER(-1 0.142 0.142 1.000 dREER(-1) 0.000 +DEN +DEN
dYEU -1.393 1.393 -1.000 dYEU -2.967 2.967 -1.000
dYEU(-1) 0.000 +DEN +DEN dYEU(-1) 0.000 +DEN +DEN
Inpt -1.281 0.869 -1.474 Inpt -1.340 1.010 -1.326

Model 2
40
17 20
38 30
Group 1 : 0 1 2 Group 1 : 1 1 0 0
Group 2 : 2 0 0 Group 2 : 0 1 1 1
Long-Run Coefficients Long-Run Coefficients
NT 1.804 0.285 6.323 NT 0.807 0.250 3.230
REER -0.305 0.030 -10.158 REER -0.304 0.082 -3.685
Phi -0.599 0.401 -1.493 Phi -0.478 0.105 -4.570
-Run Coefficients Short-Run Coefficients
1.080 0.724 1.493 NT 0.386 0.084 4.570
REER 0.183 0.122 -1.493 REER -0.145 0.032 -4.570
dT(-1) 0241 0.241 1.000 dT(-1) 0.000 +DEN +DEN
dNT -0. 0.979 -1.000 dNT 0.562 0.562 1.000
dNT(-1) 0.000 DEN +DEN dNT(-1) 0.000 +DEN +DEN
dREER 0.178 0. 1.000 dREER 0.000 +DEN +DEN
dREER(-1) 0.143 0.143 .000 dREER(-1) 0.000 +DEN +DEN
YEU_1 -0.036 0.036 -.000 YEU_1 0.667 0.277 2.404
C -0.490 0.300 -.633 C -0.850 0.435 -1.952

41
Model 3
T number of time periods by groups are:
1 18
37 29
Gro 1 : 2 0 0 0 Group 1 : 1 1 0 0
Gro 2 : 0 2 0 0 Group 2 : 0 0 0 2
Com converged after 5 iterations. Computations converged after 4 iterations.
GDP 1.581 0.240 6.575 GDP 0.815 0.112 7.280
REER -0.403 0.070 -5.776 REER -0.179 0.063 -2.839
Unrestricted Long-Run Coefficients Unrestricted Long-Run Coefficients
YEU 0.346 0.164 2.111 YEU 1.176 0.312 3.774
Phi -0.746 0.254 -2.934 Phi -0.682 0.318 -2.147
GDP 1.179 0.402 2.934 GDP 0.556 0.259 2.147
REER -0.301 0.102 -2.934 REER -0.122 0.057 -2.147
YEU 0.216 0.034 6.308 YEU 0.901 0.586 1.538
dT(-1) 0.032 0.032 1.000 dT(-1) 0.000 +DEN +DEN
dGDP -0.611 0.611 -1.000 dGDP 0.553 0.553 1.000
dGDP(-1) -0.276 0.276 -1.000 dGDP(-1) 0.000 +DEN +DEN
dREER 0.000 +DEN +DEN dREER 0.000 +DEN +DEN
dREER(-1 0.000 +DEN +DEN dREER(-1 0.000 +DEN +DEN
dYEU 0.000 +DEN +DEN dYEU -1.919 1.919 -1.000
dYEU(-1) 0.000 +DEN +DEN dYEU(-1) -2.564 2.564 -1.000
Inpt -0.678 0.017 -40.666 Inpt -1.299 0.929 -1.398

Model 4
42
17 18
37 29
Group 1 : 2 2 0 Group 1 : 2 0 2
Group 2 : 1 0 0 Group 2 : 1 0 1
Dependent variable: NT Dependent variable: NT
GDP 0.908 0.136 6.696 GDP 1.031 0.036 28.961
RP2 -0.318 0.152 -2.097 RP2 0.174 0.054 3.217
Phi -0.296 0.228 -1.296 Phi -0.487 0.087 -5.570
GDP 0.269 0.207 1.296 GDP 0.502 0.090 5.570
RP2 -0.094 0.073 -1.296 RP2 0.084 0.015 5.570
dNT(-1) -0.187 0.187 -1.000 dNT(-1) -0.122 0.122 -1.000
dGDP 0.270 0.270 1.000 dGDP 0.000 +DEN +DEN
dGDP(-1) -0.042 0.042 -1.000 dGDP(-1) 0.000 +DEN +DEN
dRP2 0.000 +DEN +DEN dRP2 -0.170 0.080 -2.120
dRP2(-1) 0.000 +DEN +DEN dRP2(-1) -0.032 0.032 -1.000
Inpt 0.247 0.188 1.311 Inpt -0.190 0.035 -5.480

CASE Network Studies and Analyses 248 - Exchange Rate Regimes and the Real Sector: a Sectoral Analysis of CEE Countries

More Related Content

What's hot (19)

Similar to CASE Network Studies and Analyses 248 - Exchange Rate Regimes and the Real Sector: a Sectoral Analysis of CEE Countries (20)

More from CASE Center for Social and Economic Research (20)

Recently uploaded (20)

CASE Network Studies and Analyses 248 - Exchange Rate Regimes and the Real Sector: a Sectoral Analysis of CEE Countries