Está en la página 1de 16

Ida Musiałkowska

Poznań University of Economics, Poland

Chair of European Studies
Conference: Regional Growth Agendas
University of Aalborg, Denmark
28th-31st May 2005

Gateway 8: New developments in regional theory

Regional dynamics, convergence and divergence: The European Union - Poland.

The paper will refer to the widely discussed regional convergence and divergence issues. It
will especially focus on the problem of the economies’ fluctuations and present one of the
aspects of the regional dynamics and convergence: business cycles synchronisation. The
context of the situation in the European Union regions before the enlargement will be given,
(based on presentation of some empirical results referring to the regional business cycles
synchronisation in the last 28 years) as well as the identification of main factors influencing
business cycles synchronisation. That part will be followed by the presentation of dynamics of
Polish regions in the transformation period. The key points of the regional development
strategies are going to be a part of the whole image. Main findings, conclusions form the
paper summary.

Convergence/ divergence processes have been widely discussed over last years. The
neoclassical approach seems to be applied as a standard. Synchronisation of the business
cycles is a part of discussion, in the context of convergence issues. Some authors even use the
term of the “business cycles convergence”. There is relatively large number of research
devoted to national business cycles synchronisation. In integrating Europe, however, regions
have been playing growing role. Their economies, treated as the “mini-economies”, have the
same features as the countries have. They adapt in more flexible way to changing conditions
at the markets, use of new technologies and react faster to the cultural patterns change. They
compete among themselves and adjust their structure to gain better position. Regions:
• react to exogenous shocks;
• their economies are influenced by multipliers and accelerators;
• have growth trajectories subject to shifts in world markets, trade cycles, relative price
changes and competition from outside.[ Klaassen H.L. et al., 1987]
Moreover, local and regional economies are combinations of industries, of which each
is characterised by its own cycles. Cycles duration in some areas dependent on the particular
industry is almost the same as in given industry (i.e. five years cycles in Detroit, Michigan or
Wolfsburg in Germany – areas dependent on the car industry). Such phenomenon is not
present in regions or local areas with highly diversified structure of employment. The more
diversified structure the less sensitive region is to the fluctuations.[2]
Under such conditions the national business cycle cannot be a simple sum of regional
business cycles. The evidence both in the European Union and the United States dismantled
the thesis of uniformity in regional business cycles and response to the stabilisation efforts of
the central governments. Timing and severity of business cycles varies across regions, due to
different factors. Comparing shape of national and regional business cycles one can observe
that some areas do not experience cyclical behaviour during national business cycles. In
general, the timing of the turning points in regional cycles do not correspond to those of
national cycles. In the majority of works the regional business cycles are treated the same way
as it is in case of national economies. Regions have their own smaller economic system. The
morphology of regional business cycles is characterised by: phases (expansion and recession
or upswings and downswings), turning points (upturn and downturn), amplitude of
fluctuations (which determines symmetry of fluctuations) and duration time (the whole cycle
is measured from one turning point to another of the same character i.e.: from one upturn to
the next upturn). [I. Musialkowska; Regional Economies’ Fluctuations in Europe]

Theories applied in research of national business cycles may be used as well in the researches
of the regional business cycles. In the literature the convergence (or synchronisation) of the
regional business cycles almost has not presented. The conducted researches are rather
“partial” – they refer to the regions of one or only a few chosen countries. The length of the
data series and indexes were differentiated. Comparing those results one cannot evaluate the
process of regional business cycles synchronisation. Therefore, there is necessity to examine
the whole population of the European Union regions.
Data and methodology
The ERECO database served as the main source of calculation. Research comprises regions of
all EU-14 member states.1 The EU regions are classified by EUROSTAT in six categories
(levels) - NUTS. In the article three territorial levels are analysed: countries (NUTS0),
macroregions (NUTS1) and regions (NUTS2). Too short-time series was main reason for non-
including regions of new member states into analysis.2 The analysis covers the whole
population of the NUTS1 and NUTS2 regions – 257,3 and data series – period from 1975 to
2001. The main variable used in the analysis is total GVA (in euro 1995) in particular regions.
The survey based on annual series, due to lack of comparable quarterly or monthly data.
The methodology referring to the links between regional business cycles fluctuations is based
on the methodology used in surveys of international business cycles synchronisation. It is
embedded in the theories of real business cycles and assumes that business cycles fluctuations
can be identified by decomposition of time series on a trend and deviation from a trend.
Oscillations around the trend represent the path of business cycles fluctuations, while
correlation coefficients between deviation form the trends in regions reflect the convergence
level of regional business cycles.
In order to identify trend in time series Hodrick-Prescott filter was applied. That methodology
aims at time series decomposition to growth element, which represents a trend and to cyclical
element. It is often presented as;
yt = gt + ct for t=1,…, T

yt – empirical time series
gt – growth element
ct – cyclical element

Luxembourg treated as a country and one region at the same time was excluded from the research.
The data series are annual and data for 10 new members cover the period since 1990, while for the rest the data
are available since 1975.
In the database there are 262 regions, but for the 5 French outseas territories (DOMs) the data is not available.

Growth and cyclical series are reduced to:
⎧T T1
min ⎨∑ ( y t − g t ) 2 + λ ∑ [(g t +1 − g t ) − ( g t − g +1 )] }
⎩ t −1 t −2

where λ is a smoothing parameter. In the analysis, as the most frequently suggested, λ=100
for the annual data. After eliminating a trend from the time series, the first differences form
cyclical fluctuations. The next step was creation of Pearson correlation matrix, representing
the correlation coefficients ( r) for pairs of analysed regions.

r2,1 r2, 2
r257 ,1 r257, 2 r257, 257

The analysis was mainly based on the level of r≥0,75 (high correlation).
The research will serve as a base for further deepened analysis.
3. Results
The main purpose of the research was verification whether synchronisation of regional
business cycles do exist or not and to what extent. Two kinds of synchronisation was defined:
• internal
• external.
Synchronisation of regional business cycles within each country, at three following levels,
form internal synchronisation:
1. NUTS1 and NUTS2 regions’ cycles synchronisation with the NUTS0 (country) cycle
2. NUTS1 and NUTS2 regional business cycles
3. NUTS2 regional cycles.
External synchronisation comprises cross-country synchronisation of regional business cycles
at the NUTS1 and NUTS2 regional level.

Scheme 1. Level of analysis of regional business cycles synchronisation in the EU-14.

macroregion NUTS1 country NUTS0

region NUTS2

region NUTS2 macroregion NUTS1 synchronisation

other NUTS2regions
region NUTS2 in the country

Other EU regions: synchronisation
macroregion NUTS1 macrooregion NUTS1
region NUTS2 region NUTS2

Own elaboration

Observation of the internal synchronisation let formulate the following conclusions:

1) In the majority of the EU countries relatively often regional and national business cycles
are synchronised. 186 of 257 (72,37%) of regions have the cycles correlated with the
national one. At the same time a part of regions does not show correlation with the
national business cycles. Ireland and Portugal have the maximum percentage of the
regional cycles correlated with the national one, while Belgium, Greece and Austria – the
minimum. (Table 1)
2) Within the country not all NUTS2 regions have their cycles synchronised with the
business cycles of the NUTS1 regions that they belong to i.e.: in the United Kingdom,
Belgium, the Netherlands, Denmark, Greece and Italy. About 10% of all NUTS2 regions
do not synchronise their cycles with the NUTS1 regions’ cycles in a particular country
(this fact may reflect biases within particular countries). Strong internal correlation of
business cycles at those levels is in Spain, Portugal, Finland, Western Germany, Austria
and partly - in France.

Table 1. Synchronisation of the regional business cycles with the national business cycle in the EU
Number of regions, whose cycle is synchronised with
the national business cycle
Total number Medium
Country High Quite high Low
of regions correlati
correlation correlation correlation
r≥0,75 0,7≤r<0,75 r<0,5
Belgium 13 - - 3 10
Denmark 3 2 - 1 -
Germany 48 37 - - 11
Greece 16 7 3 5 1
Spain 23 21 1 1 -
France 284 18 1 7 2
Ireland 2 2 - - -
Italy 25 22 1 1 1
Netherlands 16 11 1 1 3
Austria 12 7 4 1 -
Portugal 8 8 - - -
Finland 7 6 1 - -
Sweden 8 8 - - -
United Kingdom 48 37 4 4 3
Source: Own elaboration

3. Business cycles of the NUTS2 regions are apparently synchronised. Strong correlation is
shown in a half of analysed countries (Spain, Portugal, Finland, Sweden, Ireland, Germany,
Austria and France). In Germany and Austria cycles are correlated in two groups: Eastern and
Western part.
Business cycles synchronisation of many regions within a country does not mean existence of
one “single” cycle in all those regions. The level of synchronisation is differentiated with the
values of correlation coefficients. The highest values (over 0,90) are present in Spain,
Germany and Portugal. In the rest of the countries, apart form the strong correlation between
regional cycles, there are also regions that are characterised by different fluctuations than the
others. Those regions are “specific”, especially due to different structure (i.e. with a high
percentage of employed in agriculture) or specialised functions (i.e. manufacturing). The
character of regional economy is often conditioned by geographic location. All these features
influence non-numerous correlation of their business cycles i.e.: mountainous regions of
Scotland, Western Austria or the Northern Ireland.

External synchronisation analysis gave the following results:

1. 228 of 257 (88,72%) of NUTS1 and NUTS2 regions have the synchronised cycles.
Values of the correlation coefficients are varied, however (from 0,75 to 0,98).

Ther are total number of 33 NUTS regions, but for 5 of them the data is not available.

Table 2. The number of regions, the cycle of whose is correlated with the other regional business
r≥0,95 0,95>r≥0,90 0,90>r≥0,85 0,85>r≥0,80 0,80>r≥0,75
coefficient r≥0,75
Number (%) of EU
regions 46 (17,9%) 135 (52,35%) 180 (70,04%) 205 (79,77%) 225 (87,55%)
Source: Own elaboration
The correlation coefficients were divided into classes upon their values. In the first class of
correlation coefficient r≥0,95 there are 46 regions of 9 European countries (17,9%). The most
represented are Spanish and German regions, whose business cycles’ correlation is also very
In the second class 0,95>r≥0,90 there are 135 (52,35%) regions of all EU-14 member states,
except for Ireland. There are also relatively few regions of Finland, Greece and United
Kingdom. In the next three classes there are regional cycles of all analysed countries.
2. The maximum number of cycles synchronized with one regional cycle is 98, while the
average is 32,37 – and above average correlation is shown by 119 regions (46,3 %).
3. Synchronisation of the regional business cycles was analysed separately in the groups of
NUTS1 and NUTS2 regions. Tendencies in both groups are similar. The most open
regions (with the highest number of correlated cycles) are in Spain, Western Germany,
Italy and Portugal. There are also single French, Swedish, Belgian and Austrian regions.
As far as the number of correlated cycles is concerned, at the bottom are regions from the
Netherlands, United Kingdom, Greece and Italy.
4. In the research two groups of regions, where the cycles are synchronised in each group,
were identified.
The first one – the northern area – comprises regions from the countries of Anglo-Saxon
model of their economies. Those are regions of southern and middle parts of the United
Kingdom, Ireland and Scandinavia – Sweden, Denmark and Finland (Fig. 1.).
The single regions from the remaining EU member states have their cycles correlated with
those enumerated peripheral ones. All regions are characterised by high level of
urbanisation and high percentage of employed in services. There are also very often
clusters (and new technologies industries) located on their territories. The latter may
explain high correlation of their cycles. At the same time those regions are also very well

The second group – continental Western European area – consists of regions from
Germany, France, Spain and Portugal (Fig. 2.). A part of Spanish and Portuguese regions
has an agricultural character, while German and French are dominated by the service
sector. The factor of the level f economic development does not explain their cycles’
correlation. Synchronisation can be explained, however, by neighbouring of countries:
Spain and Portugal, Spain and France, France and Germany, and trade between those
countries and regions.5
There are also regions that do not have the cycles correlated with the others i.e.: Scotland,
Northern Ireland, parts of Belgium, the Netherlands, Denmark, Italy, Greece, Western
Austria and Eastern Germany. Very high percentage of employed works in agriculture
sector or manufacture. Regions are of the islander or mountainous character. Both, rich
and poor regions are between them. Those regions do not show correlation within the
country they belong to, as well.
The hypothesis on the regional business cycles synchronisation was also verified in the
analysis of total synchronisation (external and internal), excluding synchronisation with the
national business cycle. Two levels of the correlation coefficient were compared:
1. for the correlation coefficient r≥0,75 results that at this level of correlation:
- correlation is shown by 249 (96,87%) of 257 regional business cycles
- the average number of cycles correlated with one regional business cycle is 45,8
- above average correlation of their business cycles is shown by 119 regions (46,3%).
2. At the level of correlation coefficient r≥0,50
- there is no region with non-synchronised business cycle
- the average number of the business cycles correlated with one regional business cycles is
- above average correlation is shown by 152 regions (59,14%).
The regions with the highest and most numerous synchronisation belong to the following
countries: Spain, Germany and Italy (mainly its northern and southern parts). At the opposite
side are regions from: northern part of the United Kingdom, Eastern Germany, central Italy,
western Austria, part of Greece and single regions from Belgium, the Netherlands, Denmark
and Sweden.
The explanation of such a situation can be found in acting of the factors, described in the
literature.6 The most significant for the regional business cycles correlation were:

Based on the calculation made by S. Barrios and J.J. de Lucia.
In the article no econometric calculation estimating influence of the factors was made, due to lack of

- geographic location
- structure of economy and regional specialisation
- neighbouring of countries (while smaller importance had regional neighbourhood)
- urbanisation (the metropolitan regions’ business cycles are generally synchronised, as well
as those of the capital ones)
- trade and historic linkages (observation on the basis of literature)
Influence of European integration was not clearly visible7. The strength of correlation of the
regional business cycles not always is bigger in case of the pairs of regions from the “core”
countries in comparison to the pairs of regions from the “core” and peripheries.
The level of economic development seems to be not very important for the process of
synchronisation. Regions that are better developed show the similar number of the business
cycles correlated within the group of 30 of the richest regions, as – in 30 of the poorest ones. 8
(Table 3.)

Table 3. Regions with the highest and the lowest level of GVA per capita in PPS and the number of
the business cycles synchronised within each group, 2002
Regions of the highest GVA Regions of the lowest GVA
1. IT3 9 ES43 10
2. IT32 9 PT12 9
3. IT6 8 ES42 9
4. UKI1 7 PT2 8
5. IT31 7 PT11 8
6. DE5 7 ES61 8
7. UKJ1 7 ES6 8
8. IT1 7 PT15 8
9. UKH1 7 GR41 7
10. UKI 7 GR14 7
11. DE6 6 GR1 7
12. FR1 6 GR2 7
13. DE71 6 ES11 7
14. BE1 5 GR13 7
15. DE21 5 GR11 6
16. IT11 5 GR25 6
17. IT33 4 PT14 4
18. AT13 3 GR22 4
19. FI16 3 PT3 3
20. IE02 3 DEE1 3
21. IT51 3 DE4 3
22. IT13 2 DED1 3
23. IT4 1 DE8 3
24. IT12 1 GR23 2
25. SE01 1 GR21 2

comparable regional data. The analysis has descriptive character.

The influence of integration is understood as the period of the EU membership.
The ranking of 30 richest and poorest regions bases on the values of GVA per capita, compared to the rankings
by the European Commission and clusters analysis.

26. FI2 1 ES63 1
27. AT32 1 BE32 1
28. DK01 1 UKK3 1
29. IT2 0 UKM4 0
30. NL31 0 UKL1 0
Średnia 4,4 Średnia 5,06
Source: Own calculations
4. Polish case
As a supplement to the research of the whole population of the EU-14 regions, short
description of the Polish regions’ situations will be given. Even in the era of non-market
economy in Poland the first sights of cyclical regional development was perceived.
[Domański R., 1989] However questions about regions’ growth trajectory and fluctuations
have arisen after 1989, starting point of economic transformation in Central and Eastern
European countries. The structural changes and economic fluctuations are imposed on each
other. Moreover, short term since transformation beginning and lack of regional statistics in
current prices make dynamics observation impossible. However, there are some initial
researches on the issue of regional fluctuations in Poland.9
As far as the enterprise profitability index in Poland in 1992-1997 is concerned we obtain
the following stages [Domański R., 2002]:
1. basic structural changes (1994);
2. expansion in which the boom in the national and European economies was superimposed
on the results of the initial impact of the reforms (1994-95)
3. inflection and contraction of the economy (1995-97)
4. recession (1997-98)
In analysed period economic fluctuations were superimposed on the structural changes.
Poland nowadays is divided into sixteen regions (voivodships) that correspond to NUTS2
classification and in all of them GDP is below 75% of the EU average. The best-developed
regions are: mazowieckie, wielkopolskie, dolnośląskie, zachodniopomorskie. At the bottom
are: podlaskie, podkarpackie, świętokrzyskie, warmińsko-mazurskie. As far as differentiation
of the value added per capita in the regions one may distinguish the following groups:
- the biggest dynamics was visible in: mazowieckie, wielkopolskie, małopolskie
- the lowest - in: lubelskie, lubuskie, kujawsko-pomorskie, śląskie, opolskie.
The majority of regions are concentrated around the national average level, however.
Such differentiation allows for indicating the cyclicity in regions’ development. In the given
example, the statistic population was limited to four regions: two from Western (poznańskie
and wrocławskie) and two from Eastern Poland (białostockie and lubelskie). In the research

the former names of the regions are used, due to the fact that new administrative division of
the country has been introduced recently [Domański R., 2002].
Comparison of cyclicity between four voivodships:
1. Employment per inhabitant in 1990-1998 (1990=100)
In Poland there were seen the following stages:
• drop in the number of employed in 1990-93
• increase in 1993-1997
• decline since 1997.
The most severe employment index decrease was in the wrocławskie. The białostockie gained
its position faster than the others due to regional economy structure: traditional branches:
agriculture, forestry, FCMG industry. While in the wrocławskie many industrial branches
were deeply affected by transformation processes.
2. Industrial output per inhabitant in 1990-1998 (1990=100)
The biggest dynamics the poznańskie noticed, the lowest - the białostockie, which resulted
from the lower labour productivity. In the poznańskie during the first phase the index
fluctuated because of deeper industry transformation, which caused better market adjustment
and higher dynamics in the next years.
3. Retail sale of goods per inhabitant in 1990-1998 (1990=100)
High dynamics is observed in the białostockie and lubelskie, which stemmed mainly from
trade relations with the inhabitants from Belarus and Ukraine. After visa restrictions imposed
on those countries, the dynamics fell. Low sale index in the wrocłwskie is correlated with the
fall in the number of employed.

Research bases on Alonso, Lösch and Weidlich theoretical and methodological approach.

4. Investment outlays per inhabitant in 1990-1998 (1990=100)
Dynamics in the Western regions is higher than in the Eastern ones. High indexes of
employment and retail sale did not correspond to investment dynamics. Short-term economic
growth in 1994-1997 was caused by temporary impacts and did not transfer to other growth
factors: i.e.: industry factors. Increase in investment outlays in the Western regions was due to
foreign direct investment and capital inflow from Germany and the USA. It brought new
technologies, management and marketing systems.

Table 4. Cyclical symptoms in the Polish voivodships in 1990-1998

Poznańskie Wrocławskie Białostockie Lubelskie


fluctuations and
downswing 2 downswings 2 downswings
Employment stagnation
upswing 2 upswings 2 upswings

Industrial upswing and gradual gradual upswing
output downswing growth growth downswing
gradual growth

Retail sale growth growth growth growth

upswing and
Investment fluctuations growth
gradual growth downswing
outlays slight growth stagnation

Source: Elaborated on basis of R. Domański; Zróżnicowanie i wahania gospodarki regionów. Suplement do

dyskusji nt. teorii Augusta Loscha. “Przegląd Geograficzny” 2002, 74, 2, p. 157-174

The research and observation of other of Polish regions10 confirm that fluctuations
depend on the economies’ structure (mainly industrial one). The presence of sensitive
industrial branches is reflected in slightly appearing cyclical patterns. In general, regions
better developed (as in this case the Western regions: poznańskie and wrocławskie) develop
faster and in more stable way. They attract investors and create conditions for
Another research conducted for all new Polish voivodships based on the industrial
sales and unemployment time series in 1999-2003. Four regional business cycles groups of
similar course were identified. The main feature differentiating the course of two time series,
is higher volatility of seasonality in industrial sales growth rates than the unemployment rates.

ERECO Database, Cambridge Econometrics, 2003

There was different variation coefficients in both cases: in case of unemployment rates the
average was 0,15 and changed from 0,08 to 0,22 while for the industrial sales rates, the
average was 0,44 and changed form 0,27 to 0,84). The unemployment in the nature is more
stable, which explains the situation. [Wyżnikiewicz et al., 2004] The above mentioned groups
of regions were identified on comparison of those two time series business cycles’ course. As
resulted, those groups are the following: (Map 1.)
1. southern and eastern Poland – industrial (małopolskie and podkarpackie)
2. northern and eastern Poland (warmińsko-mazurskie, podlaskie, lubelskie and
świętokrzyskie) - industrial
3. northern and central Poland (zachodniopomorskie, pomorskie, kujawsko-pomosrskie,
mazowieckie and łódzkie) - unmeployment
4. south-western Poland (wielkopolskie, lubuskie, dolnośląskie, opolskie and śląśkie) -

Source: Wyżnikiewicz et. al.; Regional Differntiation of Business Cycles in Poland 1999-2003

The regions in the groups are in close neighbourhood – they can create particular
“macroregions”. The explanation of such “grouping” needs further analysis. The main
features influencing the process are: economic structure, historical traditions, economic
relations and geographic location and trade with the i.e. Germany (as in case of the fourth
Patterns of behaviour: convergence of industrial or unemployment time series and
transmissions of business cycles impulses are typical for transformation period in the
economies. Both research confirm the specific of the period.
In the former European Union (EU-15) the regional business cycles are synchronised. The
majority of them have their cycles synchronised with the others. Many factors influence the
process of such a convergence. Those are i.e. geographic location, economic structure,
regional specialisation, trade and historic links. However, in case of new member states there
are still research comparing the development and regional dynamics in new and old EU
countries. The EU-25 is much more diversified than EU-15. There is still question arising:
what is the optimum of economic growth and optimal amplitude of fluctuations. Those doubts
are especially important in the stage of Economic and Monetary Union, where the monetary
policy is pursued from the European level. The countries lost one of the fluctuations’
stabilisation instruments. The role of fiscal policy has increased in importance, as well as the
redistribution of resources within i.e. the structural policy.

Figure 1. The northern area – regions that have cycle correlated with many other regional business
cycles within the area.


Own elaboration

Figure 2. The continental-Western European area – regions that have cycle correlated with many other
regional business cycles within the area.


Own elaboration


Barczyk R., Kowalczyk Z. (1994): Polityka stabilizacji koniunktury gospodarczej; Akademia Ekonomiczna w
Barnes W., Ledebur L. (1998): The New Regional Economies. The U.S. Common Market and the Global
Economy. SAGE Publications, Thousand Oaks, London, New Delhi, pp.49-61.
Barrios S., Brulhart M., Elliot R.J.R., Sensier M. (2001): A Tale of Two Cycles: Co-fluctuations between UK
Regions and the Euro Zone; Econ Papers Web Site
Barrios S., de Lucia J.J. (2001): Economic Integration and Regional Business Cycles: Evidence from the Iberian
Regions; FEDEA;
Canova F. (2001): Are EU Policies Fostering Growth and Reducing Regional Inequalities?; Els Opuscles del
CREI (Universitat Pompeu Fabra, Barcelona), No. 8/2001,
Domański R. (2002): Zróżnicowanie i wahania gospodarki regionów. Suplement do dyskusji nt. teorii Augusta
Loscha., Przegląd Geograficzny, 74, 2, pp.157-174
Domański R. (2002): Gospodarka przestrzenna; PWN, Warszawa, pp. 113-114.
Domański R. (1989): Cykle regionalne w gospodarce planowej; Przegląd Geograficzny, LXL, 1-2, pp. 1-22.
ERECO Database (2003): The European Regional Database, Cambridge Econometrics
Klaassen L.H., Berg L van den, Burns L. (1987): Spatial Cycles; Aldershot, England, Brookfield, pp. 1-8, 146-
Musiałkowska I. (2004): Regional Economies’ Fluctuations in Europe; in: “Unification of European Economies:
Opportunities and Threats”, Wydawncitwo Naukowe Wydziału Zarządzania Uniwersytetu Warszawskiego,
Warszawa, pp.111-118.
Pontes J.P. (2000): Sources of Convergence in the European Union - the Case of Portugal;
Wyżnikiewicz B., Fundowicz J., Łapiński K., Peterlik M. 920040: Regional Differentiation of Business Cycles
in Poland, 1999-2003; 27th CIRET Conference, Warsaw