The Dynamics of Trade in Central
and Eastern European Countries
Imre Ferto
We describe the evolving pattern of Central European countries' trade
using recently developed empirical procedures based around the clas-
sic Balassa index. Despite significant changes in Central European
economies during transition to a market economy, the distribution of
the indices did not change radically over the 1990s. Our results suggest
that the trade pattern converged in Czech Republic, Hungary, Poland
and Slovenia, while it polarised in Estonia Latvia, Lithuania and Slo-
vakia over the period. For particular product groups, the indices display
greater variation. They are stable for product groups with comparative
disadvantage, but product groups with weak to strong comparative ad-
vantage show significant variation. At the product level different de-
velopment can be observed in the changes of trade specialization. The
comparative advantages are still based largely on primary and natural
resource intensive product groups in the Baltic countries, while CEE5
countries show a successful upgrading process in technological and hu-
man capital intensive products.
Key Words: international trade, revealed comparative advantage,
Central Europe
jEL Classification: F14, F15, E23
Introduction
Recently, there has been renewed interest for trade dynamics (Proudman
and Redding 2000; Brasili et al. 2000; Redding 2002; Stehrer and Wörz
2003; Wörz 2005; Hinloopen and van Marrewijk 2004b; Zaghini 2005).
The theoretical literature on growth and trade stresses that comparative
advantage is dynamic and develops endogenously over time. In particu-
lar, one strand of the literature (Lucas 1988; Young 1991; Grossman and
Helpman 1991) has demonstrated that the growth rate of a country may
be permanently reduced by a 'wrong' specialisation. Another strand em-
phasises the role of factor accumulation in determining the evolution of
international trade (Findlay 1970; 1995; Deardorff1974).
Although there is a wealth of literature on the trade between Central-
Eastern European countries and the eu member states, this has tended
Imre Ferto is Senior Research Fellow and Research Director at the
Institute of Economics of the Hungarian Academy of Sciences.
Managing Global Transitions 5 (1): 5-23
not to deal with evolution of trade patterns, except Zaghini (2005). The
dynamics of trade pattern often reflects deep structural changes in the
whole economy of a particular country. It takes usually a long time,
since comparative advantages may not change in the short run. But, there
may occur sudden external and internal shocks influencing production,
diffusion of new technology and institutional systems. During the last
decade, the economies of Central European countries have been con-
siderably transformed, including transition from planned economy to
market economy, increasing trade openness, fdi etc. Therefore, it is rea-
sonable to assume that these changes may affect on the trade pattern over
time. In other words, Central European countries represent exceptional
cases, when powerful changes in the economy should have effects on the
evolution of trade pattern.
In this paper we apply recently developed empirical methods to in-
vestigate the dynamics of trade patterns in Central European countries
(Czech Republic, Estonia, Hungary, Latvia, Lithuania, Poland, Slovakia
and Slovenia). The novelty of the paper is that we extend the traditional
empirical trade analysis by the application of the factor abundance ap-
proach (Hinloopen and van Marrewijk 2004b) to identify the changes in
the pattern of comparative advantage. The paper is organised as follows.
The second section briefly reviews some of the theoretical literature con-
cerning the dynamics of trade patterns. The third section outlines the
employed methodology. Results are reported in two stages. First, in the
fifth section we present results on the dynamics of trade pattern in the
fourth section. Second, we show the results on the structural changes
in comparative advantage based on a factor abundance approach. The
summary and some conclusions are presented in the last section.
Trade Dynamics
The standard Heckscher-Ohlin model implies that the pattern of trade
specialisation changes only if trading partners experience a change in
their relative factor endowments. This suggests that the existence of per-
sistent trade patterns is perfectly consistent with the model, if the relative
factor endowments of countries do not change significantly with respect
to their main trading partners.
The New Trade Theory emphasises the importance of increasing re-
turns scale in explaining trade flows, which complicates the predictions
of trade theory, because they depend on the specific assumptions about
the nature of return to scale. One strand of this literature assumes that
economies of scale are internal to the firm (e. g. Krugman 1987; Helpman
and Krugman 1985). In this case, the main implications of the factor pro-
portions theorem basically do not change.
If national external economies of scale exist, trade patterns dynam-
ics depend on the effects of the external economies of scale on the slope
of the production possibility frontier. Kemp (1969) and Markusen (1981)
have proven that, if external economies of scale are negligible with re-
spect to the factor intensity differences between two sectors, then a rel-
ative supply curve is positively sloped, and yields similar implications as
in the standard Heckscher-Ohlin model.
If national external economies of scale are relevant, the predictions of
the model will change substantially. Wong (1995) has shown that in the
presence of national external economies of scale, the world trade pattern
is determined by initial comparative advantage.
However, Either (1979; 1981) argues that increasing returns depend on
the size of the world market. He demonstrates that in the case of interna-
tional external economies of scale, increasing returns of scale do not in-
fluence the pattern of international trade. Grossman and Helpman (1990;
1991), under the assumption that knowledge spillovers are international
in scope, have shown that the history of the production structure of a
country do not affect on its long-run trade pattern, which only depends
on the relative factors endowments.
However, other families of models find that dynamic scale economies
arising from learning by doing are country specific and suggest a lock-in
effect for the pattern of specialisation. Krugman (1987) and Lucas (1988)
demonstrate that in the presence of dynamic economic scale the long-
run trade pattern is determined by initial comparative advantage. The
main implications of these models are that international trade patterns
tend to be more specialised.
Proudman and Redding (2000) built a model focusing on interna-
tional trade and endogenous technical change which illustrates that
a precisely specified model yields ambiguous conclusions concerning
whether international trade patterns display persistence or mobility over
time. They conclude that it is ultimately an empirical question.
Methodology
The most popular indicator of a country's trade specialisation is the Re-
vealed Comparative Advantage (rca) index first proposed by Balassa
(1965):
B=i,	d)
xrs
where x represents exports, i is a commodity, j is a country, r is a set
of commodities and s is a set of countries. B is based on observed trade
patterns; it measures a country's exports of a commodity relative to its
total exports and to the corresponding export performance of a set of
countries. If B > 1, then a comparative advantage is revealed, i. e. a sector
in which the country is relatively more specialised in terms of exports.
Many researchers have attempted to refine revealed comparative ad-
vantage (see Vollrath 1991). Despite many criticisms, e. g. the asymmetric
value, problem with logarithmic transformation, etc. (see De Benedictis
and Tamberi 2001) the B index remains the popular tool in empirical
trade analysis. The main advantage of the B index against alternative
measures is its theoretical foundation. Earlier, Hillman (1980) had in-
vestigated the relationship between the B index and comparative advan-
tage as indicated by pre-trade relative prices, abstracting from consider-
ations caused by the possibility of government intervention on exports.
He showed that the B index is not appropriate for cross-commodity com-
parison of comparative advantage, because in this case the value of B
is independent of comparative advantage in the Ricardian sense of pre-
trade relative prices. Furthermore, Hillman developed a necessary and
sufficient condition to obtain a correspondence between the B index and
pre-trade relative prices in cross-country comparisons for a given prod-
uct:
1 - — > — (l - —V	(2)
Wi Xj \ W)'
where Xj is exports of commodity i by country j, Xj is total exports of
country j, Wi is world exports of commodity i, and W is the world's
total exports. Assuming identical homothetic preferences across coun-
tries, the condition in (2) is necessary and sufficient to guarantee that
changes in the B index are consistent with changes in countries' relative
factor-endowments. This condition guarantees that growth in the level of
a country's exports of a commodity results in an increase in the B index.
For an empirical test, Marchese and Nadal de Simone (1989) transformed
Hillman's condition into:
1 _ £2.
HI=	(3)
xM1 Wi
x
•I
If HI is larger than unity, the B index used in cross country compari-
son will be a good indicator of comparative advantage. The authors ar-
gued that Hillman's index should be calculated in any empirical research
attempting to identify the long-term implications of trade liberalisation
using the B index. However, only two studies appear to have applied Hill-
man's index: Marchese and Nadal de Simone (1989) show that Hillman's
condition is violated in less than 10 per cent of exports of 118 develop-
ing countries in 1985; and in the data set used by Hinloopen and Van
Marrewijk (2001) Hillman's condition was not valid for only 7 per cent
of export value and less than 1 per cent of the number of observations.
Furthermore, Hinloopen and Van Marrewijk (2004a), using a compre-
hensive dataset between 1970-1997, find that violations of the Hillman
condition are small as a share of the number of observations, but may
be considerable as a share of the value of total world exports. The au-
thors argue that the Hillman condition should be included as a standard
diagnostic test for empirical analysis of comparative advantage.
We focus on the stability of the Balassa index over time. One can dis-
tinguish at least two types of stability (Hinloopen and Van Marrewijk
2001): (i) stability of the distribution of the indices from one period to
the next; and (ii) stability of the value of the indices for particular prod-
uct groups from one period to the next.
The first type of stability is investigated in two ways. First, after Dalum
et al. (1998) we use the B index in regression analysis:
Bj = ai + ß^ + ej,	(4)
where superscripts ti and t2 describe the start year and end year, respec-
tively. The dependent variable, B at time t2 for sector i in country j, is
tested against the independent variable which is the value of B in year
ti; a and ß are standard linear regression parameters and e is a residual
term. Ifß = i, then this suggests an unchanged pattern of B between pe-
riods ti and t2, i. e. there is no change in the overall degree of specializa-
tion. If ß > i, the existing specialisation of the country is strengthened.
If 0 < ß < i, this indicates despecialisation, i. e. commodity groups with
low initial B indices grow over time, while product groups with high ini-
tial B indices decline. The special case where ß < 0 indicates a change in
the sign of the index. However, Dalum et al. (1998) point out that ß > i is
not a necessary condition for growth in the overall specialisation pattern.
Thus, following Cantwell (1989), they argue that:
v?2 ß2
— = —	(5)
P2
V 1 R
and hence,
4 =	(6)
IR.I
where R is the correlation coefficient from the regression and v2 is vari-
ance of the dependent variable. It follows that the pattern of a given dis-
tribution is unchanged when ß = R. If ß > R the degree of specialisation
has grown, while if ß < R the degree of specialisation has fallen.
The second way in which the stability of the distribution of B is exam-
ined seeks to measure the extent to which Central European countries'
exports have become relatively more or less specialised over the period.
This is undertaken using the Gini coefficient as a measure of concen-
tration (see, for example, Amiti 1998). The Gini coefficient is used as a
summary measure of the difference in the structure of exports between a
particular Central European country and the eu. The closer the Gini co-
efficient is to its upper bound of 1, the greater is the difference in structure
and specialisation of a particular Central European country's exports vis-
a-vis the eu.
The second type of stability, that of the value of the trade index for
particular product groups, is also analysed in two ways. First, following
a recent empirical method pioneered by Proudman and Redding (2000)
and applied by Brasili et al. (2000) and Hinloopen and Van Marrewijk
(2001), we employ transition probability matrices to identify the persis-
tence and mobility of revealed comparative advantage as measured by
the B index. There is no accepted guide in the literature for classifica-
tion of B index into appropriate categories. Most studies classify data
into various percentiles, like quartiles or quintiles. Hinloopen and Van
Marrewijk (2001) point out that this classification has several drawbacks.
First, boundaries between classes are difficult to interpret. Second, they
also differ from one country to another, therefore making cross-country
comparisons difficult. Consequently, following Hinloopen and Van Mar-
rewijk (2001), we divide the B index into four classes:
Class a: 0 < B < 1
Class b: 1 < B < 2
Class c: 2 < B < 4
Class d: 4 < B
Class a refers to all those product groups without a comparative ad-
vantage. The other three classes, b, c, and d, describe the sectors with
a comparative advantage, roughly classified into weak comparative ad-
vantage (class b), medium comparative advantage (class c) and strong
comparative advantage (class d).
Second, the degree of mobility in patterns of specialisation can be
summarised using indices of mobility. These formally evaluate the de-
gree of mobility throughout the entire distribution of B indices and fa-
cilitate direct cross-country comparisons. The first of these indices (M1,
following Shorrocks 1978) evaluates the trace (tr) of the transition prob-
ability matrix. This index thus directly captures the relative magnitude of
diagonal and off-diagonal terms, and can be shown to equal the inverse
of the harmonic mean of the expected duration of remaining in a given
cell.
m,=(7)
K-l
where K is the number of cells, and P is the transition probability matrix.
The second index (M2, after Shorrocks 1978 and Geweke et al. 1986)
evaluates the determinant (det) of the transition probability matrix.
M2 = 1 - \det(P)\.	(8)
In both indices, a higher value indicates greater mobility, with a value
of zero indicating perfect immobility.
Furthermore, to test the equality of different Markov transition prob-
abilities we apply Anderson and Goodman's (1957) test statistics, which
under null hypothesis py = ~pij, for each state i has an asymptotic distri-
bution:
_ 2	t~ 1
J	piJ	t=0
where m is the member of states, py are the estimated, pij are the prob-
abilities under null, and nt(t) describes the number of sectors in cell i at
time t.
Empirical Results
The revealed comparative advantage can be measured at the global level
(e. g. Vollrath 1991), at a regional or sub-global level (as in Balassa's orig-
inal specification) or restricted to the analysis of bilateral trade between
just two countries or trading partners (e. g. Dimelis and Gatsios 1995;
Gual and Martin 1995). Given that we are interested in the dynamics of
Central European countries' trade vis-a-vis the eu, calculation of the in-
dices is restricted to an eu context, using total merchandise exports as
the denominator (respectively, s and r in (1)). We focus on the period
1993-2002, with data supplied by unctad at the three-digit level of the
siTc for 232 product groups. Following Marchese and Nadal de Simone
(1989), the indices calculated from our data set are found to be fully con-
sistent with Hillman's condition.
dynamics of the distribution
Table 1 shows some simple summary measures of dispersion for B in-
dex at the start and end of the period (due to lack of data in the case of
the Lithuania, the end year is 2001). A general decrease in international
specialisation is evident for the Czech Republic, Hungary and Poland, in
that the all the measures have fallen. The numbers also suggest that these
countries lost their comparative advantage for some product groups. In-
terestingly, the mean of B index has decreased in Lithuania and Slovenia,
while the median of B index has grown. The other group of countries
(Estonia, Latvia and Slovakia), however report a growth in specialisa-
tion. Furthermore, Estonia and Latvia have relatively high mean values
with the lowest median values, indicating a narrow band of high special-
isation.
To evaluate the statistical significance of these changes, a two-tailed
Wilcoxon signed rank test was applied because of the non-normality in
the distributions. The null hypothesis, of no difference in the B indices
between the start and end years, can be rejected (at a level of 5% or less)
in the cases of Latvia, Lithuania, Poland, Slovakia and Slovenia.
The regression results in table 2, based on (4) show a more complete
picture. The coefficient on initial specialization is always significantly dif-
ferent from zero and one, except Slovakia. The estimations suggest that
there is no reverse change in trade pattern in Central European countries.
The coefficients are less than one for six countries (Czech Republic, Esto-
nia, Hungary, Lithuania, Poland, and Slovenia) This implies that in these
countries the international specialisation has increased for products in
which the countries were initially relatively less specialised and has de-
creased for those in which they were initially highly specialised. Nev-
ertheless, there are remarkable differences among countries: the value
of ß is relatively small for Estonia and Lithuania (below 50 per cent),
thus pointing at large changes within the distribution, it is much higher
for Czech Republic and Slovenia, then indicating a relative stability of
table 1 Descriptive statistics of B index
Country	Year	Mean	Median		Std. dev.	Maxium
Czech Republic	1993	1.81	0.88		4.80	60.92
	2002	1.62	0.81		3.92	45.7
Estonia	1993	2.47	0.38		6.38	77.48
	2002	3.22	0.48		12.93	117.35
Hungary	1993	1.63	0.57		4.70	60.17
	2002	1.28	0.56		3.24	30.5
Latvia	1993	2.16	0.18		5.78	78.89
	2002	4.16	0.33		22.28	257.92
Lithuania	1993	1.66	0.38		10.98	52.01
	2001	1.45	0.40		2.89	26.55
Poland	1993	2.81	0.57		10.95	146.50
	2002	2.40	0.56		10.43	116.09
Slovakia	1993	1.18	0.59		1.82	16.40
	2002	1.35	0.52		2.57	26.85
Slovenia	1993	1.08	0.60		2.01	15.10
	2002	1.07	0.83		1.62	12.64
table 2 Stability of B between 1993 and 2002						
	a	ß	R	ß/R	(1)	(2)
Czech Republic	0.23	0.77	0.942	0.82	0.000	0.000
Estonia	1.59	0.47	0.222	2.09	0.000	0.000
Hungary	0.34	0.60	0.828	0.72	0.000	0.000
Latvia	-0.62	2.20	0.804	2.73	0.000	0.000
Lithuania*	1.01	0.33	0.289	1.14	0.000	0.000
Poland	0.95	0.53	0.623	0.86	0.000	0.000
Slovakia	0.06	1.11	0.774	1.43	0.000	0.121
Slovenia	0.23	0.84	0.854	0.98	0.000	0.000
notes *2001. Column headings are as follows: (1) Sign. of F (H0: ß = 0); (2) Sign. of F
(Ho: ß = 1).
the specialisation pattern, and somehow intermediate for Hungary and
Poland. However, Latvia displayed a strong specialisation. The F test con-
firms that trade specialisation has not changed in Slovakia. However,
the ß/R ratios show that the pattern of revealed comparative advantage
has tended to converge for only Czech Republic, Hungary, Poland, and
table 3 Gini indices of B between 1993 and 2002
Year	Czech R.	Estonia	Hungary	Latvia	Lithuania	Poland	Slovakia	Slovenia
1993	0.642	0.7752	0.7256	0.8409	0.8257	0.8016	0.6045	0.6663
1994	0.6621	0.7737	0.691	0.8444	0.7128	0.7846	0.6154	0.6543
1995	0.6246	0.7719	0.6867	0.8995	0.718	0.7686	0.6456	0.6529
1996	0.6386	0.7775	0.6907	0.9154	0.6985	0.7443	0.6636	0.6518
1997	0.6188	0.7946	0.6822	0.9387	0.6763	0.7469	0.6257	0.6498
1998	0.6065	0.8053	0.6782	0.9561	0.7116	0.7302	0.6176	0.6353
1999	0.6082	0.8166	0.6854	0.9375	0.7451	0.7166	0.6435	0.6458
2000	0.6021	0.8259	0.6954	0.9054	0.7345	0.7029	0.6544	0.6381
2001	0.5966	0.8167	0.6739	0.8913	0.7227	0.7057	0.662	0.6306
2002	0.6418	0.849	0.6984	0.9073	n. a.	0.7561	0.662	0.621
ß	-0.006	0.010	-0.002	0.007	-0.005	-0.011	0.007	-0.006
t	-2.010	8.460	-1.360	1.690	-0.790	-3.420	2.660	-7.540
R2	0.339	0.899	0.189	0.263	0.083	0.594	0.470	0.877
Slovenia. Although Estonia and Lithuania also show 'ß-despecialization,'
their the degree of specialization has actually increased. Similarly, Slo-
vakia also experienced a growing specialisation in trade pattern.
The extent to which Central-European exports have become relatively
more or less specialised over the period, vis-a-vis the eu, is shown by the
Gini coefficients in table 3. Regressing the log of the Gini coefficients on
a simple time trend (see, for example, Amiti 1998), there is a significant
increase in specialisation in Estonia and Slovakia; no significant change
in Hungary, Latvia and Lithuania; and a significant fall in specialisation
in Czech Republic, Poland and Slovenia.
intra-distribution dynamics
Further information on the dynamics of the trade index can be obtained
by analysis of Markovian transition matrices, showing the probability of
passing from one state to another between the starting year (1993) and
the end year (2002). The transition matrices in table 4 suggest that val-
ues of the B index are fairly persistent from 1993 to 2002 for observations
with a comparative disadvantage (class a) for all countries. The diago-
nal elements for this class are 0.82 or above for all countries, indicating
a high probability that a product with a comparative disadvantage at the
start of the period will have that same status at the end of the period.
The persistence is relatively strong at the other ends of distribution (d,
d), the value of cells is larger than 40 per cent. This suggests that once
obtaining a large comparative advantage they will likely maintain it over
time. Note that the values relative to the ends of the distribution on the
main diagonal are larger than those in the middle of distribution for the
Czech Republic, Estonia, Latvia, Lithuania and Slovakia. In other words,
it is easier maintain a strong revealed comparative advantage than a weak
or medium one. However, indices in classes b, c and d display consider-
able variation in their pattern. The probability of a loss of comparative
advantage for those observations starting with a weak comparative ad-
vantage (class b) are high (above 50 per cent), for Hungary, Latvia and
Lithuania. There is a small chance of moving from class c (medium com-
parative advantage) to class d (high comparative advantage) in the cases
of the Czech Republic, Hungary and Slovenia.
In summary, these results suggest that the probability of an observa-
tion moving to a lower value cell (a weakening of comparative advan-
tage) is much higher than the reverse case. The limit distributions show
a more polarised distribution for Estonia, Poland, and Slovenia, whilst
asymmetry is confirmed for the Czech Republic and Hungary, tending
to a right skewed distribution.
Table 5 reports the mobility indices, M1 and M2, for each of the coun-
tries. Both indices indicate that mobility is highest in Lithuania and
lowest in Slovenia. Furthermore, Estonia, Hungary and Slovenia show
the most persistent pattern of specialisation, while Poland, Latvia and
Lithuania are the most dynamic economies. Similarly to other papers
(Brasili et al. 2000; Proudman and Redding 2000; Hinloopen and van
Marrewijk 2001; 2004b; Zaghini 2005) the two indices do not yield the
perfectly same ranking. But the Spearman rank correlation coefficient
remains high (0.95).
Finally, Anderson and Goodman's test rejects the equality of all transi-
tion matrices relative to the estimated benchmark (table 5, columns 2-3).
In other words, changes across different comparative advantage classes
were significant for all countries.
Structural Changes in Trade Specialisation
This section analyses some details of the structural change in Central
European countries' comparative advantage. After Hinloopen and van
Marrewijk (2004b) we classify most sectors at the 3-digit level according
to factor intensity on the basis of the International Trade Center's clas-
table 4 Transition matrices of B Index
B	a	b	c	d	a	b	c	d
	Czech Republic				Estonia			
a	0.77	0.16	0.05	0.02	0.84	0.11	0.04	0.01
b	0.40	0.42	0.15	0.03	0.20	0.48	0.20	0.12
c	0.29	0.35	0.26	0.10	0.26	0.32	0.26	0.16
d	0.25	0.06	0.13	0.56	0.11	0.11	0.22	0.56
Initial distribution	0.53	0.27	0.13	0.07	0.69	0.11	0.08	0.12
Final distribution	0.57	0.25	0.11	0.07	0.64	0.16	0.09	0.10
Limit distribution	0.61	0.24	0.10	0.06	0.55	0.21	0.12	0.12
	Hungary				Latvia			
a	0.85	0.08	0.03	0.03	0.80	0.11	0.05	0.04
b	0.51	0.46	0.03	0.00	0.56	0.22	0.11	0.11
c	0.33	0.33	0.29	0.04	0.35	0.09	0.26	0.30
d	0.06	0.12	0.41	0.41	0.36	0.07	0.14	0.43
Initial distribution	0.67	0.15	0.10	0.07	0.76	0.08	0.10	0.06
Final distribution	0.69	0.17	0.09	0.06	0.71	0.11	0.08	0.09
Limit distribution	0.73	0.16	0.07	0.04	0.68	0.11	0.09	0.12
Continued on the next page
sification, the joint unctad/wto organization. According to five broad
factor intensity categories as follows:
a.	Primary products (83 sectors); e. g. meat, dairy, cereals, fruit, coffee,
sand, minerals, oil, natural gas, iron ore, and copper ore.
b.	Natural-resource intensive products (21 sectors); e. g. leather, cork,
wood, lime, precious stones, pig iron, copper, aluminium, and lead.
c.	Unskilled-labor intensive products (26 sectors); e.g. pipes, various
textiles, clothing, glass, pottery, ships, furniture, footwear, and of-
fice supplies.
d.	Technology intensive products (62 sectors); e.g. various chemicals,
medicaments, plastics, engines, generators, machines, tools, pumps,
telecommunications and photo equipment, optical equipment, and
aircraft.
e.	Human-capital intensive products (43 sectors); synthetic colours,
pigments, perfumes, cosmetics, rubber and tires, tubes, pipes, vari-
Continued from the previous page
	Lithuania		Poland		
a	0.87 0.06	0.04 0.03	0.75 0.20	0.05	0.01
b	0.53 0.11	0.21 0.16	0.24 0.50	0.21	0.06
c	0.21 0.16	0.21 0.42	0.19 0.35	0.27	0.19
d	0.19 0.25	0.13 0.44	0.05 0.14	0.33	0.48
Initial distribution	0.69 0.16	0.08 0.07	0.65 0.15	0.11	0.09
Final distribution	0.71 0.09	0.09 0.11	0.55 0.25	0.12	0.08
Limit distribution	0.70 0.10	0.08 0.12	0.43 0.31	0.16	0.10
	Slovakia		Slovenia		
a	0.86 0.08	0.06 0.01	0.85 0.13	0.02	0.00
b	0.40 0.40	0.17 0.04	0.13 0.70	0.17	0.00
c	0.14 0.36	0.25 0.25	0.15 0.26	0.48	0.11
d	0.00 0.00	0.50 0.50	0.17 0.17	0.25	0.42
Initial distribution	0.62 0.21	0.12 0.05	0.70 0.13	0.12	0.05
Final distribution	0.63 0.17	0.13 0.07	0.64 0.22	0.10	0.03
Limit distribution	0.60 0.16	0.14 0.10	0.49 0.35	0.14	0.03
table 5 Mobility indices and test statistics for equality of Markov transition matrice					
	mi	m2		p value	
Czech Republic	0.663	0.983	107.702	0.000	
Estonia	0.619	0.979	144.287	0.000	
Hungary	0.662	0.967	131.361	0.000	
Latvia	0.762	0.993	56.595	0.000	
Lithuania	0.793	0.996	90.328	0.000	
Poland	0.669	0.994	123.504	0.000	
Slovakia	0.664	0.984	140.906	0.000	
Slovenia	0.516	0.916	190.213	0.000	
ous types of steel and iron,		cutlery, televisions, radios,		cars, watches,	
and jewellery.					
This leaves five 3-digit sectors not classified according to intensity					
which will be ignored in the remainder of this section. The complete clas-
sification can be found at following website: www.few.eur.nl/few/people
/vanmarrewijk/eta.
table 6 Share of product groups in exports B > 1 according to factor intensity
1993	2002
	a	b	c	D	E	a	b	c	D	E
Czech Republic	0.22	0.07	0.21	0.17	0.33	0.07	0.04	0.16	0.33	0.41
Estonia	0.43	0.10	0.26	0.16	0.05	0.31	0.08	0.22	0.25	0.13
Hungary	0.38	0.06	0.23	0.24	0.09	0.11	0.03	0.13	0.57	0.15
Latvia	0.29	0.22	0.21	0.08	0.20	0.45	0.18	0.24	0.02	0.11
Lithuania	0.54	0.04	0.09	0.11	0.22	0.52	0.05	0.10	0.03	0.30
Poland	0.32	0.15	0.33	0.06	0.14	0.16	0.08	0.28	0.18	0.30
Slovakia	0.17	0.10	0.17	0.21	0.35	0.12	0.07	0.16	0.16	0.49
Slovenia	0.10	0.12	0.33	0.22	0.22	0.04	0.10	0.19	0.34	0.32
Table 6 (columns 4-5) displays the breakdown of the export share by
resource intensity in 1993 and 2002, where the B index is larger than
one. The cee countries show a different pattern of revealed compara-
tive in terms of resource intensity. In 1993, the Czech Republic and Slo-
vakia have high shares (above 50 percent) of the two 'highest' categories:
technology- and human capital-intensive products during the analysed
period. Corresponding figures for Estonia, Latvia and Poland have not
exceeded the 30 percent. Between 1993 and 2002, Czech Republic, Hun-
gary, Poland and Slovenia improved their share of categories e and f by
more than 20 percentage points. Estonia and Slovakia also experienced
significant growth in these categories. However, Latvia lost its share by 15
per cent, whilst Lithuania's proportion remained the same in category e
and f.
It is interesting to compare the dynamics that we can observe in this
table and in table 5. As we already noticed above, mobility indices are
the highest in Poland, Latvia and Lithuania. However, for Latvia and
Lithuania, the mobility does not mean an 'upward' movement on the
resource-intensity scale. On the other pole, Hungary and Slovenia - al-
though showing a highly persistent pattern in specialization remain able
to increase exports' share in technology intensive and human capital in-
tensive products.
Summary and Conclusions
In this paper we have investigated the changing pattern of Central Eu-
ropean exports to the eu. As a measure of trade specialisation we have
employed the classic Balassa index. Despite significant changes in Central
European economies during transition to a market economy, the distri-
bution of the indices has not changed radically over the 1990s. The extent
of specialisation in Central European trade exhibits a mixed trend. The
distribution of Balassa index differs markedly from one country to the
other. Our results suggest that the trade pattern has converged in the
Czech Republic, Hungary, Poland and Slovenia, whilst it polarised in Es-
tonia, Latvia, Lithuania and Slovakia over the period. Estimations based
on Gini indices reinforces the conclusion that there is a significant in-
crease in specialisation in Estonia and Slovakia; no significant change in
Hungary, Latvia and Lithuania; and a significant fall in specialisation in
the Czech Republic, Poland and Slovenia vis-a-vis the eu.
The stability of the indices for particular product groups displays more
variation. Results suggest that the indices are stable for observations with
comparative disadvantage, in all cases. But product groups with weak,
medium or strong comparative advantage show significant variation,
with a tendency to weakening comparative advantage. In other words,
the Markov matrices show a relatively high mobility for all countries,
even though changes are particularly frequent in the middle of the dis-
tribution.
How are these stylised measurements linked to the findings of other
empirical studies? An overall picture emerging from empirical studies
(Balassa 1977; Amendola et al. 1992; Laursen 2000; Proudman and Red-
ding 2000; and Brasili et al. 2000) is that one can observe a general de-
crease in specialisation, with a few exceptions. However, our study of
Central European trade only partly reinforces this result, since we ob-
serve both growth and fall in trade specialisation.
At the sector level the pattern of comparative advantage also differs
by countries. The main explanation may be that the cee countries have
different backgrounds. They liberalised and reformed their economies to
a varying degree; consequently, the differences in their earlier manufac-
turing bases, political stability, administrative reforms and geographical
locations, have resulted in different developments in comparative ad-
vantages across-countries. Recent studies have shown that the special-
ization pattern of many ceecs has changed over time, in many cases
rapidly, with the cee countries shifting production towards higher-tech
and higher-skill industries (see Havlik 2001; Landesmann and Stehrer
2002; Wörz 2003; Zaghini 2005). Consistently with these results we found
that the comparative advantages in Baltic countries are still largely based
on natural resources; whilst the Czech Republic, Hungary, Slovakia and
Slovenia are the most oriented towards human-capital and technology
intensive products, with Poland following closely. The latter countries
display a rather fast catching-up in high skill and high-technological in-
tensive products, in spite of the significant technological gap they inher-
ited from the planned economy period. However, the increasing trade
specialisation in the Baltic countries focuses mainly on primary and nat-
ural resources intensive products. Our results complement those of Dul-
leck et al. (2005), who identify two different ways in quality upgrading
for cee countries. The first group (Czech Republic, Hungary, Poland
Slovakia and Slovenia) can be characterised by successful upgrading of
their export structure, whilst the results for the second group (the Baltic
countries) was less unambiguous. In short, the Baltic states seem to be
less successful in the catch up process to the European Union.
Acknowledgments
The author gratefully acknowledges financial support from the cerge-
ei Foundation in the framework of research grant The Dynamics of
Trade in Central and Eastern European Countries. The content of the
publication is the sole responsibility of the author and does not neces-
sarily represent the views of the cerge-ei Foundation. The author is
grateful to Ira Gang, Byengjou Jeong and two anonymous referees for
useful comments.
References
Amendola, G., P. Guerrieri, and P. C. Padoan. 1992. International patterns
of technological accumulation and trade. Journal of International and
Comparative Economics 1 (1): 173-97.
Amiti, M. 1998. New trade theories and industrial location in the eu. Ox-
ford Review of Economic Policy 14 (2): 45-53.
Anderson, T. W., and L. A. Goodman. 1957. Statistical inference about
Markov chains. Annals of Mathematical Statistics 28 (1): 89-110.
Balassa, B. 1965. Trade liberalization and revealed comparative advantage.
The Manchester School of Economic and Social Studies 33 (2): 99-123.
-. 1977. Revealed comparative advantage revisited: An analysis of rel-
ative export shares of the industrial countries, 1953-71. The Manchester
School of Economic and Social Studies 45 (4): 327-44.
Brasili, A., P. Epifani, and R. Helg. 2000. On the dynamics of trade pat-
terns. De Economist 148 (2): 233-57.
Cantwell, J. 1989. Technological innovation and multinational corporations.
Oxford: Blackwell.
Dalum, B., K. Laursen, and G. Villumsen. 1998. Structural change in oecd
export specialisation patterns: De-specialisation and 'stickiness'. Inter-
national Review of Applied Economics 12 (2): 423-43.
Deardorff, A. 1974. Factor proportions and comparative advantage in the
long run: Comment. Journal of Political Economy 82:829-33.
De Benedictis, L., and M. Tamberi. 2001. A note on the Balassa index of
revealed comparative advantage. Quaderni di ricerca 158, Department
of Economics, University of Ancona.
Dimelis, S., and K. Gatsios. 1995. Trade with Central and Eastern Europe:
The case of Greece. In eu trade with Eastern Europe: Adjustment and
opportunities, ed. R. Faini and R. Portes. London: cepr.
Dulleck, U., N. Foster, R. Stehrer, and J. Wörz. 2005. Dimensions of quality
upgrading: Evidence from cees. Economics of Transition 13 (1): 51-76.
Either, W. J. 1979. Internationally decreasing costs and world trade. Journal
of International Economics 9 (1): 1-24.
-. 1981. National and international return to scale in the mod-
ern theory of international world trade. American Economic Review
72:389-405.
Findlay, R. 1970. Factor proportions and comparative advantage in the
long run: Comment. Journal of Political Economy 78:7-34
-. 1995. Factor proportions, trade and growth. Cambridge, ma: mit
Press.
Geweke, J., R. Marshall, and G. Zarkin. 1986. Mobility indices in continu-
ous time Markov chains. Econometrica 54 (6): 1407-23.
Grossman, G., and E. Helpman. 1990. Comparative advantage and long-
run growth. American Economic Review 80:796-815.
-. 1991. Innovation and growth in the global economy. Cambridge,
ma: mit Press.
Gual, J., and C. Martin. 1995. Trade and foreign direct investment with
Central and Eastern Europe: Its impacts on Spain. In eu trade with
Eastern Europe: Adjustment and opportunities, ed. R. Faini and R.
Portes. London: cepr.
Havlik, P. 2001. Competitiveness of ceecs' industry. In Proceedings of
the International Conference on the 10-Year Review of Transitional
Economies and Challenges in the Next Decade, 23-36. Vienna: unido.
Helpman, E. 1981. International trade in the presence of product differen-
tiation, economies of scale and imperfect competition: A Chamber-
lain-Heckscher-Ohlin approach. Journal of International Economics
11:305-40.
Helpman, E., and P. Krugman. 1985. Market structure and foreign trade.
Cambridge, ma: mit Press.
Hillman, A. L. 1980. Observation on the relation between 'revealed com-
parative advantage' and comparative advantage as indicated by pre-
trade relative prices. Weltwirtschaftliches Archiv 116 (2): 315-21.
Hinloopen, J., and C. van Marrewijk. 2001. On the empirical distribution
of the Balassa index. Weltwirtschaftliches Archiv 137 (1): 1-35.
-. 2004a. Empirical relevance of the Hillman condition and the
comparative advantage. Tinbergen Institute Working Paper 2004-
019/2
-. 2004b. Dynamics of Chinese comparative advantage. Tinbergen
Institute Working Paper 2004-034/2.
Kemp, M. 1969. The pure theory of international trade and investment.
Englewood Cliffs, nj: Prentice-Hall.
Krugman, P. 1987. The narrow moving band, the Dutch disease, and the
competitive consequences of Mrs. Thatcher: Notes on trade in the
presence of dynamic scale economies. Journal of Development Eco-
nomics 27 (1): 41-54.
Landesmann, M. A., and R. Stehrer. 2002. Evolving competitiveness of
ceecs in an enlarged Europe. Rivista di Politica Economica 92 (1): 23-
87.
Laursen, K. 2000. Trade specialisation, technology and economic growth.
Cheltenham: Elgar.
Lucas, R. 1988. On the mechanics of economic development. Journal of
Monetary Economics 22 (1): 3-22.
Marchese, S., and F. N. De Simone. 1989. Monotonicity of indices of re-
vealed comparative advantage: Empirical evidence on Hillman's con-
dition. Weltwirtschaftliches Archiv 125 (1): 158-67.
Markusen, J. 1981. Trade and the gains from trade with imperfect compe-
tition. Journal of International Economics 11 (4): 531-51.
Proudman, J., and S. Redding. 2000. Evolving patterns of international
trade. Review of International Economics 8 (3): 373-96.
Redding, S. 1999. Dynamic comparative advantage and the welfare effects
of trade. Oxford Economic Papers 51 (1): 15-39.
-. 2002. Specialization dynamics. Journal of International Economics
58 (2): 299-34.
Shorrocks, A. 1978. The measurement of mobility. Econometrica 46 (5):
1013-24.
Stehrer, R., and J. Wörz. 2003. Technological convergence and trade pat-
terns. Review of World Economics 139 (2): 191-219.
Vollrath, T. L. 1991. A theoretical evaluation of alternative trade inten-
sity measures of revealed comparative advantage. Weltwirtschaftliches
Archiv 130 (2): 265-79.
Wong, K. 1995. International trade in goods and factor mobility. Cambridge,
ma: mit Press.
Wörz, J. 2003. Skill upgrading in Central and Eastern European manufac-
turing trade. The Empirical Economics Letters 2 (6): 247-56.
-. 2005. Dynamics of trade specialization in developed and devel-
oping countries. Emerging Markets Finance and Trade 41:92-111.
Young, A. 1991. Learning-by-doing and dynamic effects of international
trade. Quarterly Journal of Economics 106 (2): 396-406.
Zaghini, A. 2005. Evolution of trade patterns in the new eu member states.
Economics of Transition 13:629-58.