The Role of Openness in Regional Economic Growth. The Case of Polish and Spanish NUTS-2 Regions

Abstract

With the use panel data techniques, we estimate an empirical growth model for Polish and Spanish NUTS-2 regions – two similar-in-size European economies with the inferior initial level of development and at the same time major recipients of EU structural funds. The analysis is carried out for 16 Polish voivodeships and 19 NUTS2 level municipalities, provinces and autonomous communities observed over the period 2000-2014. Within the joined group of regions, we observe a clear beta-absolute and sigma-convergence. Within countries, the evidence points to divergence. The level of regional sigma convergence is similar. Of particular interest to us is the assessment of the role of broadly defined economic openness in the process of regional economic growth. The initial analysis points to the bidirectional relationship. We then estimate a dynamic panel data model with the use of two-step GMM due to non-stationary nature of the key variables. We control for potential interactions of openness with regional human capital endowments as well as other major determinants postulated by theoretical models. The obtained results are in line with theoretical predictions.

Full text

Tomasz Brodzicki1
The Role of Openness inRegional Economic Growth.
The Case of Polish and Spanish NUTS-2 Regions2
Summary
With the use panel data techniques, we estimate an empirical growth model for
Polish and Spanish NUTS-2 regions –two similar-in-size European economies with
the inferior initial level of development and at the same time major recipients of EU
structural funds. The analysis is carried out for 16 Polish voivodeships and 19 NUTS-2
level municipalities, provinces and autonomous communities observed over the period
2000–2014. Within the joined group of regions, we observe aclear beta-absolute and
sigma-convergence. Within countries, the evidence points todivergence. The level of
regional sigma convergence is similar. Of particular interest tous is the assessment
of the role of broadly dened economic openness inthe process of regional economic
growth. The initial analysis points tothe bidirectional relationship. We then estimate
adynamic panel data model with the use of GMM due tonon-stationary nature of the
key variables. We control for potential interactions of openness with regional human
capital endowments as well as other major determinants postulated by theoretical
models. The obtained results are inline with theoretical predictions.
Keywords: regional development, economic growth, panel data, Poland, Spain
JEL Classication Codes: C23, R11, F43, O18, O4
1 Instytut Rozwoju, Sopot; Uniwersytet Gdański, Wydział Ekonomiczny, Katedra Ekono-
miki Integracji Europejskiej; e-mail: t.brodzicki@ug.edu.pl
2 Acknowledgments: The study has been supported by agrant from the National Science
Centre [grant number 2015/19/B/HS4/01704] ‘Regional Exporting Activity. Assessment of
Determinants inLight of Contemporary Foreign Trade Theory for Poland and Spain’ super-
vised by Prof. Stanislaw Umiński and carried out by Institute for Development. The data for
the regional trade of Spain have been retrieved by Marcin Skurczyński and Anna Fornal-
ska-Skurczyńska. We appreciate the comments from two peer reviewers as well as the com-
ments from the participants of the conference at Warsaw School of Economics on 12 May
2017 and the comments and suggestions from Stanisław Umiński, Jarosław Nazarczuk and
Dorota Ciołek.

44 Tomasz Brodzicki
1.Introduction
With ongoing globalisation, the degree of openness of national and thus
regional economies is increasing, even inthe aftermath of the global nancial
crisis. Liberalisation of trade ows and greater mutual openness lead todefrag-
mentation of production and emergence of the so-called global value chains,
which increases mutual interdependence of economies. Greater openness
increases exposure tointernational shocks, which could be considered as one
of its major costs.
Despite the dominance of the viewpoint of the benecial impact of open-
ness (liberalisation) on economic growth, the review of both theoretical and
empirical literature does notbring clear results. Nonetheless, the signicance
and direction of causality inthe relationship between openness and economic
growth is an important issue both intheoretical and empirical economic liter-
ature. On the theoretical ground, asignicant progress was made inthe 80 s
and 90 s,with the emergence of the new growth theory and the new economic
geography. On the empirical ground, the major development was ashift from
standard cross-sectional regressions (ala Barro) tomore sophisticated panel
data, including dynamic panel data models. Most of the empirical studies sofar
have been carried out at national economies level and notthe within the coun-
try or inter-country regional level of analysis.
The purpose of the paper is toidentify the role of trade openness indeter-
mining the growth of Polish and Spanish NUTS-2 regions, controlling for other
signicant factors affecting economic growth process. With the use panel data
techniques, we estimate an empirical growth model for NUTS-2 regions of Poland
and Spain –two similar-in-size European economies with the inferior initial
level of development and at the same time major recipients of the EU structural
funds. The analysis is carried out for 16 Polish voivodeships and 19NUTS-2 level
municipalities, provinces and autonomous communities of Spain, observed over
the period 2000–2014. Assuming that Spanish and Polish regions share the same
steady-state point inthe long run and noting the higher present mean level of
development of Spanish regions, we can learn alot from their experience inthe
EU (Spain entered the EU –then EEC back –in1986).
The remainder of the paper is constructed as follows. Section 2 reviews the
theoretical literature, while Section 3 reviews the empirical literature. Section4
describes the data. Section 5 discusses the dependent variable, potential beta
and sigma convergence of income per capita, the changes inthe openness rate

45The Role of Openness inRegional Economic Growth. The Case of Polish and Spanish...
as well the relation between the two. Section 6 presents the results of an empir-
ical model. The nal sectionconcludes.
2. Openness ineconomic growth
–review of theoretical literature
In the neoclassical growth theory (Solow3&Swan4) openness does notmat-
ter inthe long-run, as growth is independent of economic policy. It could only
lead tothe level effects. In the short-run, capital deepening is the major source
of growth –as income per capita is proportional tocapital per capita. The level
of real income per capita inthe steady state is apositive function of the rate of
saving (investment), anegative function of the population growth rate n and
depreciation of capital δ. Technological progress affects the level of real GDP per
capita positively. The only factor affecting the long-run growth rate is the rate
of exogenous technological progress. In this setting the impact of an increase
inopenness due totrade policy on economic growth is temporary.
In an augmented model of Mankiw et al.
5
inaddition, the human capital
endowment is taken into account. The augmented neoclassical model by Brodzicki
6
takes further the impact of infrastructure into account. In the model, inaccord-
ance with Mincerian tradition, the average level of education may be specied
as afunction of average years of schooling and average years of experience7.
The emergence of the endogenous growth and new trade theories (Lucas8,
Romer9) has led tothe reopening of the debate on the role of trade, and more
general, the degree of openness indetermining economic growth inthe medium
3 SolowR. (1957), AContribution tothe Theory of Economic Growth, QJE 70 (1)/1956,
65–94; idem, Technical Change and the Aggregate Production Function, Review of Econom-
ics and Statistics 39, 312–320.
4 SwanT. (1956), Economic Growth and Capital Accumulation, Economic Record 32,
334–361.
5
MankiwG., Romer D., WeilD. (1992), AContribution tothe Empirics of Economic
Growth, Quarterly Journal of Economics 107 (2), 407–437.
6
BrodzickiT. (2015), Shallow determinants of growth of Polish regions. Empirical analy-
sis with panel data methods, Collegium of Economic Analysis Annals 39, 25–40.
7 BilsM., KlenowP. J. (2000), Does Schooling Cause Growth?, AER 90, 1160–1183.
8 LucasR. (1988), On the mechanics of economic development, Journal of Monetary Eco-
nomics 22 (1), 3–42.
9 RomerP. M. (1986), Increasing Returns and Long-run Growth, JPE 94/1986, 1002–1037;
idem, Endogenous Technological Change, JPE 1990, 98 (5), 71–102.

46 Tomasz Brodzicki
and long run. The models of the rst and second generation endogenized the
rate of growth of technology, either by allowing for the impact of human cap-
ital or introducing aseparate R&D sector purposefully producing knowledge
inthe form of patents. It is worth topoint out, however, that even inasemi-en-
dogenous model of Ben-David&Romer10, openness totrade through its impact
on the process of accumulation of knowledge and technology transfer leads
toendogenization of economic growth.
The new growth theory models of Rivera-Batiz&Romer
11
or Grossman&Help-
man12 differ –apolicy shift leading toagreater extent of openness, could lead
toapermanent effect –long-run growth rate could be affected but notonly pos-
itively, an adverse impact is also possible. In brief, the balance of costs and ben-
ets of greater openness (liberalisation) depends on the nature and the exact
product structure of trade.
Greater openness totrade affects the rate of accumulation of knowledge
mostly through imports. They work as achannel allowing absorption of more
advanced knowledge positively affecting overall efciency and thus growth
rates. Rivera-Batiz&Romer13 show however that whether the effect is positive
or adverse depends on the distance of economy from global technology fron-
tier and the nature of diffusion of knowledge (perfect versus imperfect). Imper-
fect knowledge ows coupled with openness can actually harm underdeveloped
states or regions.
From atheoretical standpoint, openness affects growth through anumber
of channels. First of all, it leads toreallocation of factors of production tomore
productive sectors and thus tospecialisation inaccordance with the compara-
tive or competitive advantage thus resources are allocated efciently. Secondly,
it leads toincreased diffusion and accelerated absorption of knowledge and tech-
nology (technology transfer) inparticular through imports14 or inow of FDI15.
10 Ben-DavidD., LoewyM. B. (2002), Trade and the Neoclassical Growth Model, Journal
of Economic Integration, 18, 1–16.
11 Rivera-BatizL., RomerP. M. (1991), Economic Integration and Endogenous Growth,
QJE 106 (2), 531–555.
12 GrossmanG. M., HelpmanE. (1991), Innovation and Growth inthe Global Economy,
MIT Press, Cambridge MA.
GrossmanG. M., Helpman E. (1992), Innovation and Growth: Technological Competition
inthe Global Economy, MIT Press, Boston.
13 Rivera-BatizL. et. al. (1991), op.cit.
14 CoeD. T., HelpmanE. (1995), International R&D spillovers, EER 39 (5), 859–887.
15 BranstetterL. (2006), Is foreign direct investment achannel of knowledge spillovers?
Evidence from Japan’s FDI inthe United States, Journal of International Economics 68 (2),
325–344.

47The Role of Openness inRegional Economic Growth. The Case of Polish and Spanish...
Thirdly, it stimulates the rate of innovation as it is frequently associated with an
increase inthe expenditures on research and development. Fourthly, it allows
better utilisation of scale economies and agglomeration externalities as aresult
of greater specialisation. At the same time, it leads toenhanced accumulation
of factors of production. Finally, it stimulates competition innational and inter-
national markets thus forcing companies tobe more innovative.
It is worth stressing that Rodrik16 perceives openness or the extent of inte-
gration as one of three fundamental deep determinants of economic growth
alongside the quality of institutions and geographical conditions.
Afurther insight can be brought by the new economic geography literature.
As Breinlich et al.17 stress, NEG theory is based on trade theory, and thus the
relationship between external trade, internal economic geography, and regional
disparities, is at its core. Fujita et al.18 suggest that openness could work todis-
perse manufacturing industry as awhole but also lead tothe spatial clustering
of specic industries. External trade thus affects spatial patterns of activity by
changing market access considerations19.
It is also worth addressing the direction of causality between trade openness
and economic growth. If openness affects growth than we deal with export-led
growth process through the channels described above. On the other hand, the
causality could go from growth toopenness. High productivity inthe larger
domestic market (home marker effect) translates into greater international com-
petitiveness and increase inexports. At the same time, larger domestic economy
creates alarger demand for imports. Thus abidirectional relationship is likely
toexist if these two are allowed tohold simultaneously20.
16 RodrikD. (2003), Institutions, Integration and Geography: In Search of the Deep De-
terminants of Economic Growth, in: Search for Prosperity: Analytic Narratives on Economic
Growth, Princeton University Press, Princeton.
17
BreinlichH. et al. (2013), Regional growth and regional decline, CEP Discussion
Paper1232.
18
FujitaM. et al. (1999), The spatial economy: cities, regions and international trade, MIT
Press, Cambridge MA.
19
HansonG. (1996), Localization Economies, Vertical Organization and Trade, AER, 86 (5),
1266–1278.
20 LiuX., SongH., RomillyP. (1997), An empirical investigation of the causal relationship
between openness and economic growth inChina, Applied Economics 29 (12), 1679–1686.

48 Tomasz Brodzicki
3.Review of empirical literature
In the empirical literature, two strands dominate –macro approach –mostly
cross-sectional analysis of global or more homogeneous groups and the micro
approach –analysis for individual countries based on sectoral or rm level data.
Many variables are utilised as proxies of openness, nonetheless, the openness
ratio is the most popular.
Barro21 identied apositive and statistically signicant impact of the level
of openness on economic growth inacross-sectionof countries. Dollar
22
noting
apotential bias inthe index, utilised an index of exchange rate disturbances,
nding it toadversely affect economic growth. The result was conrmed by
Easterly et al.23 and Lee24 using similar approaches. Sachs &Werner25 utilised
adichotomous index of openness, conditional on meeting 5 criteria nding open-
ness tomatter for growth inacross-sectionof countries. It was also utilised by
Gallup et al.26 leading tosimilar result even if deep-rooted geographical factors
were taken into account and Vamvakidis27 nding positive and statistically sig-
nicant effects of multilateral economic integration. On the other hand, Wac-
ziarg&Welch28 found the studies applying Sachs-Werner index tobe sensitive
tothe period under analysis.
Edwards
29
(1998) inhis seminal study analysed the impact of 9 different
indices of openness/disturbances inthe exchange rate on TFP and thus indi-
rectly on real GDP per capita inacross-sectionof 93 countries. The impact was
21 BarroR. J. (1991), Economic Growth inaCross Section of Countries, QJE 106/1991,
407–443.
22
DollarD. (1992), Outward-oriented Developing Economies Really To Grow More Rapidly:
Evidence from 95 LDCs, 1976–1985, Economic Development and Cultural Change, 523–544.
23 EasterlyW. et al. (1993), Good Policy or Good Luck?, Journal of Monetary Economics
32 (3), 459–483.
24 LeeJ. W. (1993), International Trade Distortions and Long-run Growth, IMF Staff Pa-
pers, 40 (2), 299–328.
25 SachsJ. D., WarnerA. (1995), Economic Convergence and Economic Policies, NBER
Working Paper 5039.
26
GallupJ. L., SachsJ. D., MellingerA.D (1999), Geography and Economic Development,
International Regional Science Review 22 (2), 179–232.
27
VamvakidisA. (1999), Regional Trade Agreements or Broad Liberalization: Which Path
Leads toFaster Growth?, IMF Staff Papers 46 (1), 42–68.
28 WacziargR., WelchK. H. (2003), Trade Liberalization and Growth: New Evidence, Re-
search Paper 1826.
29 EdwardsS. (1998), Openness, Productivity and Growth: What Do We Really Know?,
The Economic Journal 108, 383–398.

49The Role of Openness inRegional Economic Growth. The Case of Polish and Spanish...
found tobe positive however its magnitude was found tobe inferior incompar-
ison tothe initial level of GDP per capita or the initial level of human capital.
Due topotential endogeneity IV approach is frequently utilised. For instance,
Frankel&Romer
30
proposed an instrumental variable based on geographical
factors that determine toalarge extent trade while having exogenous nature
inrelation tothe level of income. The impact of openness proved tobe insignif-
icant intwo large cross-sections. In contrast, Irwin&Tervio31 reiterated the test
by Frankel&Romer32 inaslightly modied manner inapanel of countries. The
results pointed toapositive relationship between the intensity of trade and the
level of GDP per capita. Romalis33 found similar results using the instrumental
variable approach inalarge panel of countries (135) observed over aperiod of
40 years (1960–2000).
Vamvakidis
34
tested six different measures of openness inalonger time period
(1920–1999) nding that the positive relationship between openness and growth
exists only after 1970, which could be related tooverall higher openness with
increasing extent of globalisation.
Wacziarg &Welch35 utilized adifferent approach toanalysing the effects of
cases of signicant trade-policy liberalizations and found that, on average, the
investment rate increased by 1.5 to2 percent, and the share of trade inGDP by
5percent, while the ex-postgrowth rate was higher than ex-ante growth rate
by amean of 1.5 percent.
Using the extreme bounds analysis, Levine&Renelt
36
found the index of open-
ness tobe one of the variables affecting the growth rate inacross-sectionof
countries indirectly through an impact on the process of accumulation of capi-
tal (rate of investment). The direct impact of openness was rejected. In contrast,
30
FrankelJ., RomerD. (1996), Trade and Growth: An Empirical Investigation, NBER
Working Paper 5476; FrankelJ., RomerD. (1999), Does Trade Cause Growth?, AER 89 (3),
379–399.
31 IrwinD., TervioM. (2002), Does trade raise income? Evidence from the twentieth cen-
tury, Journal of International Economics 58, 1–18.
32 FrankelJ., RomerD. (1999), Does Trade Cause Growth? op.cit.
33 RomalisJ. (2007), Market Access, Openness and Growth, NBER Working Paper W13048/
2007.
34 VamvakidisA. (2002), How Robust is the Growth-Openness Connection? Historical Evi-
dence, Journal of Economic Growth 7, 57–80.
35 WacziargR., WelchK. H. (2003), op.cit.
36
LevineR., ReneltD. (1992), ASensitivity Analysis of Cross-country Growth Regressions,
American Economic Review 82, 942–963.

50 Tomasz Brodzicki
Doppelhofer, Sala-i-Martin&Miller37 using the Bayesian Averaging of Classical
Estimates for abalanced panel of 88 countries and 68 variables founding the
time since the opening of the economy (impact of liberalisation) topositively
affect economic growth. The overall openness was found tomatter less.
The studies on the impact of openness on growth at the regional level are
rather rare. In recent years anumber of studies have been performed on Asian
economies. And thus Sun et al.38 show inastudy of Chinese regions at man-
ufacturing industries level that openness totrade (trade orientation and FDI)
have apositive effect on technical efciency. Leong39, analysing the impact of
special economic zones as cases of liberalisation on regional economic growth
inChina and India, found that both FDI and export topositively affect growth.
The presence of SEZs increases regional growth, however, an increase inthe
number of SEZs has anegligible effect on growth. Leong nds greater openness
(wider liberalisation) as aprecondition of further growth. Wei et al.40 inapanel
of Chinese regions over the entire period 1979–2003 proved that FDI inows
were one of the forces behind the observed regional discrepancies ingrowth.
The authors claim however that FDI cannot be blamed for inequality as it was
due tothe uneven distribution of FDI and notthe FDI itself.
Anwar&Nguyen41 using simultaneous equations model found inapanel of
61 provinces of Vietnam from 1996–2005, amutually reinforcing two-way pro-
cess between FDI and regional economic growth. The benets of FDI inow
could be further strengthened by more investments into education and train-
ing, development of the nancial market and reducing technology gap between
foreign and local rms.
According toKanbur&Venables
42
, rising spatial disparities inregional devel-
opment inmany developing states are mostly due touneven impact of increased
trade openness and globalisation. It leads toefciency gains mostly due to con-
37 DoppelhoferG., MillerR. I., Sala-i-MartinX. (2000), Determinants of Long-term Gro-
wth: ABayesian Averaging of Classical Estimates (BACE) Approach, NBER Working Paper
W7750.
38
SunH. et al. (1999), Economic Openness and Technical Efciency: ACase Study of Chi-
nese Manufacturing Industries, Economics of Transition 7 (3), 615–636.
39 LeongC. K. (2013), Special Economic Zones and Growth inChina and India: An Em-
pirical Investigation, International Economics and Economic Policy 10 (4), 549–567.
40
WeiK., YaoS., LiuA. (2009), Foreign Direct Investment and Regional Inequality inChina,
Review of Development Economics 13 (4), 778–791.
41
AnwarS., NguyenL. P. (2010), Foreign Direct Investment and Economic Growth inViet-
nam, Asia Pacic Business Review 16 (1–2), 183–202.
42 KanburR., VenablesA. (2005), Rising Spatial Disparities and Development, UNI-WI-
DER Policy Brief 3.

51The Role of Openness inRegional Economic Growth. The Case of Polish and Spanish...
certation of economic activity inmajor cities and coastal districts, adversely
affecting inland regions. In astudy on Latin America, Serra et al.43 argue that
regional disparities modestly increased, at least temporarily, inthe wake of trade
liberalisation. It was especially marked for Mexico.
When analysing the nexus between openness and economic growth at
regional level we have tonote the direct or indirect impact of other accompany-
ing variables or processes. For instance, Sachs et al.44 studying σ -convergence
and ß-convergence show that more than 80 percent of the cross-state variation
ingrowth rates among Indian states can be explained solely by an urbanisation
variable. Agglomeration factors are also strongly postulated by NEG theories.
The role of human capital accumulation is clear on theoretical and empirical
grounds. However, the scope of the denition of human capital differs. For exam-
ple, inthe study by Boschma&Fritsch
45
points inline with Florida toan important
contribution of the so-called creative class for regional growth in7European
countries. They are however notable todetermine whether human capital as
measured by the creative occupation, outperforms standard indicators based
on formal education and whether formal education has astronger impact. The
creative class endowment is positively affected by the regional climate of toler-
ance and openness as well as regional job opportunities.
The economic structure could matter as well including the size and share of
the industrial sector. For instance, the study by Hansen&Zhang46 points tothe
key role of the industrial sector inexplaining the regional variation ingrowth
among Chinese provinces. The result supports the Kaldorian approach toregional
economic growth with cumulative causation between trade liberalisation, the
rise inexport demand, the growth of industrial sector (industrialisation) and its
impact on overall productivity and thus increases ininternational competitiveness.
One of the issues that cannot be overlooked is the issue of path-depend-
ency inregional development. For instance, Felice&Vecchi47 indicate that the
regional North-South variation inItaly was already present the moment the
43 SerraM. I. et al. (2006), Regional Convergence inLatin America, IMF Working Paper
06 (125).
44 SachsJ. et al. (2002), Understanding Regional Economic Growth inIndia, Asian Eco-
nomic Papers 1 (3), 32–62.
45 BoschmaR. A., FritschM. (2009), Creative Class and Regional Growth: Empirical Evi-
dence from Seven European Countries, Economic Geography 85 (4), 391–423.
46 HansenJ. D., Zhang J. (1996), AKaldorian Approach toRegional Economic Growth
inChina, Applied Economics 28 (6), 679–685.
47
FeliceE., VecchiG. (2015), Italy’s Modern Economic Growth, 1861–2011, Enterprise & So-
ciety 16 (2), 225–248.

52 Tomasz Brodzicki
country was unied and then increased. The explanation of the present vari-
ation involves endogenous factors –natural resources, human capital endow-
ment, and social capital.
In an article Brodzicki
48
published inthe Annals, attempted toidentify shal-
low determinants of growth of Polish regions as well the sign and magnitude
of macroeconomic’ education –externality and macroeconomic infrastructure
externality. We constructed accordingly an augmented neoclassical growth
model incorporating aMincerian approach tohuman capital accumulation, fur-
ther assuming adirect impact of infrastructure on the overall productivity. The
estimated panel model, accounting for xed region-specic effects, was robust
and explained approx. 90 percent of observed variation inGDP per capita. The
return tothe accumulation of human capital through education and experience
for Polish regions was found tobe statistically signicant, robust and positive.
The macroeconomic infrastructure externality proved tobe, inturn, positive
–however overall insignicant with the impact of quality of railway.
4.Dataset
In the empirical partof the paper, we utilise foremost the data from the QoG
EU Regional dataset (Charron et al.
49
2016). The trade data for Polish and Spanish
regions have been obtained from the Polish Customs Chamber (Izba Celna) and
retrieved from DataComex Español database
50
. They cover the period 2005–2015.
QoG EU Regional database is adataset consisting of approximately 450 var-
iables covering three levels of European regions NUTS0, NUTS1, and NUTS2.
The data is given intime-series version (from 1990 to2015) and the unit of
analysis is region-year. The data on GDP per capita are available for the period
2000–2014 only.
48 BrodzickiT. (2015), op.cit.
49
CharronN. et al. (2016), The Quality of Government EU Regional Dataset, version Sep.16,
University of Gothenburg: The Quality of Government Institute, http://www.qog.pol.gu.se.
50 http://datacomex.comercio.es/principal_comex_es.aspx.

53The Role of Openness inRegional Economic Growth. The Case of Polish and Spanish...
5.Convergence inregional incomes and the openness ratio
The empirical analysis is carried out for agroup of 16 Polish and 19 Span-
ish NUTS-2 regions within the period 2000 to2014.
The dependent variable is the present study is anatural logarithm of GDP
per capita (ln_y). The other key variable is an openness index of regions is
measured using the standard openness index –the ratio of exports and imports
toGDP (ln_open).
If we treat two regions jointly the relation of the initial log of GDP per capita
and the mean growth rate of GDP per capita over the observed period is nega-
tive and points tobeta convergence. Polish regions are clearly catching up with
Spanish regions interms of the level of development. If we treat both countries
separately, the data are less conclusive pointing to weak regional divergence
inPoland and weak regional beta-convergence inSpain, however, the results
are notstatistically robust.
We know from economic growth theory that beta-convergence is anecessary
however notsufcient condition for sigma-convergence. Thus the above result
should be indicative of sigma-divergence inboth countries at NUTS 2 level. We
test is by plotting the evolution of standard deviation of the log of GDP per cap-
ita for both countries over the analysed period.
The result points toclear sigma-divergence inPoland over the analysed period
and U-shape pattern for Spain –with the initial sigma-convergence and then
divergence inthe aftermath of the nancial and eurozone crises. It seems that
less developed Spanish regions have been more adversely affected by the crises.
The openness ratio increased inmost of the analysed regions from 2005
to2014 (on average by 9 percent). The openness ratio dropped only inthe case
of Mazowieckie, Illes Balears, Canarias and Comunidad de Madrid.
On the other extreme, the highest increases have been reported inAndalucía,
Łódzkie, Dolnośląskie and Opolskie (by more than 15 percent), Lubuskie by
approx. 25 per cent, Región de Murcia 33 percent and Pomorskie by 34.3 percent.
We nowwill investigate the relationship between income per capita and
openness. The correlation between the two is rather weak. We have tonote that
within apanel, non-stationarity and cross-sectional dependence could exist. At
the same time, we deal with aheterogeneous panel data model that is amodel
inwhich all parameters (constant and slope coefcients) vary across regions
analysed (we thus assume conditional convergence tohold).

54 Tomasz Brodzicki
We rst apply Im–Pesaran–Shin test (Im et al.51 2003) as we cannot infer
that all panels share acommon autoregressive parameter. Cultural, other insti-
tutional and deeper rooted factors make this assumption rather feeble. The two
key variables, anamely log of GDP per capita and alog of openness ratio, are
non-stationary and we cannot reject the null hypothesis of nocointegration. In
the further econometric analysis, we thus utilise the standard solution inthe
empirical literature of the subject thus applying adynamic panel data model
estimated with the use of GMM (Arellano-Bover52 & Blundell-Bond53).
The results of Pesaran’s test of cross-sectional independence (29.653,
Pr = 0.0000) indicate that we have to reject the null hypothesis of cross-sec-
tional independence and thus we deal with cross-sectional dependence.
Finally, we analyse whether there exists acausality relationship among the
key variables using the causality test developed by Dumitrescu&Hurlin54. The
authors proposed asimple Granger
55
non-causality test for heterogeneous panel
data models. Under the null hypothesis of Homogeneous Non-Causality (HNC),
there exists nocausal relationship for any of the cross-sectionunits of the panel.
Under the alternative, one subgroup of cross-section unit is characterised by
causal relationships and the other subgroup indicates nocausal relationship.
The test statistic depends on the individual Wald statistics of Granger non-cau-
sality averaged across the cross-sectionunits. Dumitrescu&Hurlin proposed
ablock bootstrap procedure implemented inSTATA todeal with cross-sectional
dependence.
The value of panel standardised statistic ZHNC, based on the assumption of
asymptotic moments, allows us toreject the null hypothesis of noGranger-cau-
sality, infavour of the alternative hypothesis that there is Granger-causality
inat least one panel. The results point tobidirectional causality between GDP
per capita and openness inour sample of Polish and Spanish NUTS-2 regions.
This is inline with some of the theoretical postulates described inSection 2 and
empirical results inSection 3.
51 ImS. K., PesaranM., ShinY. (2003), Testing for Unit Roots inHeterogeneous Panels,
Journal of Econometrics 115, 53–74.
52 ArellanoM., BoverO. (1995), Another Look at the Instrumental Variable Estimation of
Error-Components Models, Journal of Econometrics 68 (1), 29–51.
53 BoschmaR. A., FritschM. (2009), Creative Class and Regional Growth: Empirical Evi-
dence from Seven European Countries, Economic Geograph 85 (4), 391–423.
54
DumitrescuE. I., HurlinC. (2012), Testing for Granger Non-causality inHeterogeneous
Panels, Economic Modelling 29 (4), 1450–1460.
55
GrangerC. W. (1969), Investigating Causal Relations by Econometric Models and Cross-
-spectral Methods, Econometrica 37 (3), 424–438.

55The Role of Openness inRegional Economic Growth. The Case of Polish and Spanish...
6.The empirical model & discussion of the results
Noting the non-stationarity of the dependent variable, we utilise the dynamic
panel data approach estimated with GMM using the xtdpdsys command. The
command ts dynamic panel-data estimators with the Arellano–Bover/Blun-
dell–Bond system estimator. Noticing problems with one-step GMM (the high
values of the Sargan test of overidentifying restrictions) we apply the two-step
GMM estimator. The initial results are presented inTable 2, where we estimate
the models for ajoint sample of Polish and Spanish NUTS-2 regions. Analyses
are performed for anumber of different specications of the model with avar-
ying selection of explanatory variables.
Our analysis is restricted by the availability of data inour dataset. We, unfor
-
tunately, have been unable sofar tocontrol the investments rates or regional
physical endowments (apart from transport infrastructure). We control for the
population growth rate (n)as well as the human capital endowment (ln_h –log
of population share with tertiary education).
As we do notuse xed effects method due tothe utilised econometric approach
(dynamic panel model based on rst differences) we cannot assume that initial
differences inthe level of technology are included inthe region-specic xed
effects. In order toaccount for potential differences, we take into account the
evolution of the ratio of General Expenditures of Research and Development
to GDP (d_gerd).
Similarly toBrodzicki56, we take the quality of infrastructure into account
based on the methodology proposed by Careijo et al 57. The index of infrastruc-
ture quality ICQ relativizes the infrastructure endowment by normalising the
infrastructure endowment by population and land area and simultaneously com-
paring it toabenchmark. In the present article, we take the mean for Polish and
Spanish regions as the respective benchmark. ICQ is calculated inaccordance
with the following formula:
ICQr=
Xr
Nr
Xa.PLES
Na.PLES
⎛
⎝
⎜
⎜
⎜
⎜
⎞
⎠
⎟
⎟
⎟
⎟
0,5 Xr
Ar
Xa.PLES
Aa.PLES
⎛
⎝
⎜
⎜
⎜
⎜
⎞
⎠
⎟
⎟
⎟
⎟
0,5
(1)
56 BrodzickiT. (2015), op.cit.
57 CareijoE. et al. (2006), Indicadores de Convergencia Real Para los Países Avanzados,
Estudios de la Fundación, FUNCAS, Madrid.

56 Tomasz Brodzicki
where Xr iXB gives the infrastructure endowment of agiven region and the
benchmark (mean for Poland and Spain), while N and Arepresent, respectively,
population and land area.
Our base empirical model ts the data relatively well. The coefcient on
lagged dependent variable is statistically signicant, indicating the presence
of absolute (1) or conditional convergence. In (2) we introduce n and ln_h. In
most of the specications, their impact is statistically signicant and inaccord-
ance with theoretical predictions –the impact of population growth rate is neg-
ative while the impact of the human capital endowment is positive on the level
of regional income per capita. In (3) we introduce and control for variation
inregional R&D potential by the introduction of GERD (d_gerd). The impact
of general expenditure on R&D is statistically signicant, however, adverse.
Finally, in(4) we introduce our key explanatory variable –ln_open. Its impact on
the dependent variable is clearly positive and statistically signicant. Agreater
degree of trade openness boosts the economic growth of Polish and Spanish
regions, ceteris paribus. In models (5) we account for the potential joint effect
of openness and human capital endowment on the level of GDP per capita by
an introduction of an interaction term (open_h). The magnitude of the impact
of openness when we account for the interaction is signicantly stronger, how-
ever, the interaction term is negative and statistically signicant which means
that it decreases inthe human capital endowment. That is an increase inthe
extent of openness brings stronger effects on GDP per capita of regions with
initially lower levels of human capital endowment.
In the last two specications, we control for regional infrastructure endow-
ment and its quality (inicq2 we benchmark against the mean inthe group).
The impact is statistically signicant and positive inline with the results by
Cieślik&Rokicki58 for Poland or the results of Crescenzi&Rodriguez-Pose59 for
whole Europe.
As an extension, we could acknowledge the potential spatial correlation
between regions can be included inthe model through the introduction of the
agglomeration effects or the introduction of spatial weighting matrixes inamore
sophisticated spatial econometric approach.
58 CieślikA., RokickiB. (2010), Wpływ inwestycji drogowych narozwój polskich regio-
nów, w:JóźwikB., ZalewaP. (red.), Spójność ekonomiczno-społeczna regionów Unii Euro-
pejskiej, Wydawnictwo KUL, Lublin.
59 CrescenziR., Rodriguez-PoseA. (2008), Infrastructure Endowment and Investment as
Determinants of Regional Growth inthe European Union, European Investment Bank Pa-
pers132.

57The Role of Openness inRegional Economic Growth. The Case of Polish and Spanish...
7.Conclusions
With the use dynamic panel data model estimated using two-step GMM, we
have estimated an empirical growth model for Polish and Spanish NUTS-2 regions
over the period 2000–2014 inorder toidentify the dependence of regional growth
on the extent of openness. We rst review theoretical and empirical literature.
Within the joined group of regions, we observe aclear beta-absolute and sig-
ma-convergence. Within countries, the evidence points tosigma-divergence. It
holds inparticular for Spain, after the nancial and euro zone crises. Greater
openness seems overall topositively affect regional economic growth inour
sample. The results of Granger non-causality test point, however, tothe exist-
ence of abidirectional relationship between the variables.
In comparison toour previous article devoted tothe issue of determinants of
regional variation of the growth process inPoland, we have extended the anal-
ysis by using anew dataset, increasing the temporal dimension and cross-sec-
tional dimension by using data for Spanish NUTS-2 regions and nally focusing
on the role of openness totrade. Furthermore, we have utilised amore sophisti-
cated dynamic panel model, estimated with two-step GMM noting the non-sta-
tionary nature of key variables.
Our analysis has several limitations. It is mostly due tothe limited availability
of data at regional level. Nonetheless, we plan toextend our analysis inseveral
dimensions: extending the analysis further toall NUTS 2 regions of the EU28 and
accounting for potential spatial interactions with the use of spatial econometric
techniques, extending the notion of openness by considering the ows of FDIs
as well as by controlling for institutional determinants of regional development.
References
AnwarS., NguyenL. P. (2010), Foreign Direct Investment and Economic Growth
inVietnam, Asia Pacic Business Review 16 (1–2), 183–202.
ArellanoM., O.Bover (1995), Another Look at the Instrumental Variable Estimation
Of Error-Components Models, Journal of Econometrics 68 (1), 29–51.
BarroR. J. (1991), Economic Growth inaCross Section of Countries, Quarterly Jour-
nal of Economics 106, 407–443.
Ben-DavidD., LoewyM. B. (2002), Trade and the Neoclassical Growth Model, Jour-
nal of Economic Integration 18, 1–16.

58 Tomasz Brodzicki
BilsM., Klenow, P. J. (2000), Does Schooling Cause Growth?, American Economic
Review 90, 1160–1183.
BlundellR., BondS. (1998), Initial Conditions and Moment Restrictions inDynamic
Panel Data Models, Journal of Econometrics 87 (1), 115–143.
BoschmaR. A., FritschM. (2009), Creative Class and Regional Growth: Empirical
Evidence from Seven European Countries, Economic Geography 85 (4), 391–423.
BranstetterL. (2006), Is Foreign Direct Investment aChannel of Knowledge Spillo-
vers? Evidence from Japan’s FDI inthe United States, Journal of International
Economics 68 (2), 325–344.
BreinlichH., OttavianoG. I., TempleJ. R. (2013), Regional Growth and Regional Dec-
line, CEP Discussion Paper 1232.
BrodzickiT. (2015), Shallow Determinants of Growth of Polish Regions. Empirical
Analysis with Panel Data Methods, Collegium of Economic Analysis Annals 39,
25–40.
CareijoE. et al. (2006), Indicadores de Convergencia Real Para los Países Avanzados,
Estudios de la Fundación, FUNCAS, Madrid.
CharronN. et al. (2016), The Quality of Government EU Regional Dataset, version
September 2016, University of Gothenburg: The Quality of Government Institute,
http://www.qog.pol.gu.se.
CieślikA., RokickiB. (2010), Wpływ inwestycji drogowych narozwój polskich regio-
nów, w:JóźwikB., ZalewaP. (red.), Spójność ekonomiczno-społeczna regionów
Unii Europejskiej, Wydawnictwo KUL, Lublin.
CoeD. T., HelpmanE. (1995), International R&D Spillovers, European Economic
Review 39 (5), 859–887.
CrescenziR., Rodriguez-PoseA. (2008), Infrastructure Endowment and Investment
as Determinants of Regional Growth inthe European Union, European Invest-
ment Bank Papers 132.
DollarD. (1992), Outward-oriented Developing Economies Really To Grow More
Rapidly: Evidence from 95 LDCs, 1976–1985, Economic Development and Cultu-
ral Change, 523–544.
DoppelhoferG., MillerR. I., Sala-i-MartinX. (2000), Determinants of Long-term Gro-
wth: ABayesian Averaging of Classical Estimates (BACE) Approach, NBER Wor-
king Paper W7750.
DumitrescuE.-I., HurlinC. (2012), Testing for Granger Non-causality inHeteroge-
neous Panels, Economic Modelling 29 (4), 1450–1460.
EasterlyW. et al. (1993), Good Policy or Good Luck?, Journal of Monetary Economics
32 (3), 459–483.
EdwardsS. (1998), Openness, Productivity and Growth: What Do We Really Know?,
The Economic Journal 108, 383–398.
FeliceE., VecchiG. (2015), Italy’s Modern Economic Growth, 1861–2011, Enter-
prise & Society 16 (2), 225–248.

59The Role of Openness inRegional Economic Growth. The Case of Polish and Spanish...
FrankelJ., RomerD. (1996), Trade and Growth: An Empirical Investigation, NBER
Working Paper 5476/1996.
FrankelJ., RomerD. (1999), Does Trade Cause Growth?, American Economic Review
89 (3), 379–399.
FujitaM. et al. (1999), The Spatial Economy: Cities, Regions and International Trade,
MIT Press, Cambridge, MA.
GallupJ. L., SachsJ. D., MellingerA. D. (1999), Geography and Economic Develop-
ment, International Regional Science Review 22 (2), 179–232.
GrangerC. W. (1969), Investigating Causal Relations by Econometric Models and
Cross-spectral Methods, Econometrica 37 (3), 424–438.
GrossmanG. M., HelpmanE. (1991), Innovation and Growth inthe Global Economy,
MIT Press, Cambridge, MA.
GrossmanG. M., HelpmanE. (1992), Innovation and Growth: Technological Compe-
tition inthe Global Economy, MIT Press, Boston.
HansenJ. D., ZhangJ. (1996), AKaldorian Approach toRegional Economic Growth
inChina, Applied Economics 28 (6), 679–685.
Hanson, G. (1996), Localization Economies, Vertical Organization and Trade, Ame-
rican Economic Review 86 (5), 1266–1278.
ImK. S., PesaranM., ShinY. (2003), Testing for Unit Roots inHeterogeneous Panels,
Journal of Econometrics 115, 53–74.
IrwinD., TervioM. (2002), Does Trade Raise Income? Evidence from the Twentieth
Century, Journal of International Economics 58, 1–18.
KanburR., VenablesA. (2005), Rising Spatial Disparities and Development, UNI-WI-
DER Policy Brief 3.
LeeJ. W. (1993), International Trade Distortions and Long-run Growth, IMF Staff
Papers 40 (2), 299–328.
LeongC. K. (2013), Special Economic Zones and Growth inChina and India: An Empi-
rical Investigation, International Economics and Economic Policy 10 (4), 549–567.
LevineR., ReneltD. (1992), ASensitivity Analysis of Cross-country Growth Regres-
sions, American Economic Review 82, 942–963.
LiuX., Song H., RomillyP. (1997), An Empirical Investigation of the Causal Rela-
tionship between Openness and Economic Growth inChina, Applied Economics
29 (12), 1679–1686.
Lucas, R. E., On the Mechanics of Economic Development, Journal of Monetary Eco-
nomics 22 (1), 3–42.
MankiwG., RomerD., WeilD. (1992), AContribution tothe Empirics of Economic
Growth, Quarterly Journal of Economics 107 (2), 407–437.
Rivera-BatizL., RomerP. M. (1991), Economic Integration and Endogenous Growth,
Quarterly Journal of Economics 106 (2), 531–555.

60 Tomasz Brodzicki
RodrikD. (2003), Institutions, Integration and Geography: In Search of the Deep
Determinants of Economic Growth, w:Search for Prosperity: Analytic Narrati-
ves on Economic Growth, Princeton University Press, Princeton.
RomalisJ. (2007), Market Access, Openness and Growth, NBER Working Paper W13048.
RomerP. M. (1986), Increasing Returns and Long-run Growth, Journal of Political
Economy 94, 1002–1037.
RomerP. M. (1990), Endogenous Technological Change, Journal of Political Economy
98 (5), 71–102.
SachsJ. D., WarnerA. (1995), Economic Convergence and Economic Policies, NBER
Working Paper 5039.
SachsJ. D., BajpaiN., RamiahA. (2002), Understanding regional economic growth
inIndia, Asian Economic Papers 1 (3), 32–62.
SerraM. I. et al. (2006), Regional convergence inLatin America, IMF Working Paper
no.06/125/2006.
SolowR. (1956), AContribution tothe Theory of Economic Growth, Quarterly Jour-
nal of Economics 70 (1), 65–94.
SolowR. (1957), Technical Change and the Aggregate Production Function, Review
of Economics and Statistics 39, 312–320.
SunH. et al. (1999), Economic Openness and Technical Efciency: ACase Study of
Chinese Manufacturing Industries, Economics of Transition 7 (3), 615–636.
SwanT. (1956), Economic Growth and Capital Accumulation, Economic Record 32,
334–361.
VamvakidisA. (1999), Regional Trade Agreements or Broad Liberalization: Which
Path Leads toFaster Growth?, IMF Staff Papers 46 (1), 42–68.
VamvakidisA. (2002), How Robust is the Growth-Openness Connection? Historical
Evidence, Journal of Economic Growth 7, 57–80.
WacziargR., WelchK. H. (2003), Trade Liberalization and Growth: New Evidence,
Research Paper 1826.
WeiK., YaoS., LiuA. (2009), Foreign Direct Investment and Regional Inequality
inChina, Review of Development Economics 13 (4), 778–791.

61The Role of Openness inRegional Economic Growth. The Case of Polish and Spanish...
Table 1. Dumitrescu&Hurlin (2012) Granger non-causality test results
Direction of causality WHNC-bar ZHNC-bar ZHNC-bar title
OPEN => Y 2.2386 5.1815***
(p-value = 0.0000)
1.3028
(p-value = 0.1926)
Y => OPEN 1.8232 3.4439***
(p-value = 0.0006)
0.5702
(p-value = 0.5686)
Note: ***, **, * determine signicance at 1%, 5%, and 10% level respectively. The approximated critical
values for the average statistic Wwere computed from equation (30) for the case K = 1. The simulated
critical values were computed via stochastic simulations with 50, 000 replications. For N=25, T=10
the simulated critical value is 2.40 (Dumitrescu and Hurlin; 2012; Table 4).
Table 2. Results of estimation of dynamic panel model using two-step GMM
(1) (2) (3) (4) (5) (6) (7)
L.ln_y 0.936*** 0.813*** 0.814*** 0.859*** 0.908*** 0.763*** 0.748***
(0.000648)
(0.00508) (0.00653) (0.0124) (0.0160) (0.0289) (0.0237)
N
–0.0358*** –0.0357*** –0.0422*** –0.0408*** –0.0430*** –0.0426***
(0.000606)
(0.00150) (0.00184) (0.00221) (0.00411) (0.00570)
ln_h 0.240*** 0.277*** 0.181*** 0.847*** 0.291*** 0.320***
(0.00932) (0.00991) (0.0160) (0.156) (0.0615) (0.0739)
d_gerd
–0.0416*** –0.0563*** –0.0601***
–0.104*** –0.110***
(0.00862) (0.0127) (0.0205) (0.0111) (0.0254)
ln_open 0.0821*** 0.711*** 0.0669*** 0.0802***
(0.00561) (0.139) (0.00992) (0.00726)
open_h –0.194***
(0.0425)
ln_icq 0.157***
(0.0385)
ln_icq2 0.160***
(0.0352)
Constant 0.635*** 1.257*** 1.165*** 0.782*** –1.850*** 2.385*** 1.476***
(0.00546) (0.0326) (0.0439) (0.0879) (0.523) (0.511) (0.287)
Observations 490 311 254 169 169 119 119
No of reg_id 35 30 28 26 26 21 21
Sargan test 34.983 28.767 25.415 22.705 24.007 19.007 16.566
AR(1) –3.9494 –2.0051 –1.8532 –2.0451 –2.1912 –1.7245 –1.7704
AR(2) –2.4204 –2.4587 –2.2987 –2.1368 –1.9034 –1.5303 –1.3751
Wald chi(2) 2.09e+06 149166.58 109129.98 44602.91 44252.18 5613.77 43152.18
Note: Source: Standard errors inparentheses *** p<0.01, ** p<0.05, * p<0.1. Estimated in STATA14
(xtdpdsys).

62 Tomasz Brodzicki
Figure 1. Beta-absolute convergence inthe sample of Polish and Spanish regions
Source: Own elaboration.

63The Role of Openness inRegional Economic Growth. The Case of Polish and Spanish...
0,000
0,050
0,100
0,150
0,200
0,250
2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014
Spain
Poland
Figure 2. Sigma-convergence of GDPpc inthe sample of Polish & Spanish regions
Source: Own elaboration.
0,0
20,0
40,0
60,0
80,0
100,0
120,0
140,0
Galicia
Principado de Asturias
Cantabria
País Vasco
Comunidad Foral de Navarra
La Rioja
Aragón
Comunidad de Madrid
Castilla y León
Castilla-La Mancha
Extremadura
Cataluña
Comunidad Valenciana
Illes Balears
Andalucía
Región de Murcia
Ciudad Autónoma de Ceuta
Ciudad Autónoma de Melilla
Canarias
Łódzkie
Mazowieckie
Małopolskie
Śląskie
Lubelskie
Podkarpackie
Świętokrzyskie
Podlaskie
Wielkopolskie
Zachodniopomorskie
Lubuskie
Dolnośląskie
Opolskie
Kujawsko-pomorskie
Warmińsko-mazurskie
Pomorskie
2005
2014
Figure 3. The extent of openness of Polish and Spanish NUTS-2 regions in2005 & 2014
Source: Own elaboration on the basis of Polish and Spanish regional trade datasets.

64 Tomasz Brodzicki
* * *
Streszczenie
Przy wykorzystaniu metod estymacji modeli panelowych wartykule szacujemy
empiryczny model wzrostu polskich ihiszpańskich regionów poziomu NUTS-2, dwóch
europejskich gospodarek ozbliżonej wielkości, niskim początkowym poziomie roz-
woju, ajednocześnie głównych benecjentów funduszy strukturalnych UE. Analizę
przeprowadzono dla 16 województw Polski i19 prowincji iwspólnot autonomicz-
nych poziomu NUTS-2 Hiszpanii wlatach 2000–2014. Wpołączonej grupie regionów
obserwujemy wyraźną beta-konwergencję rozwojową isigma-konwergencję, podczas
gdy analizy wobrębie krajów wskazują na dywergencję rozwojową. Szczególnym
celem artykułu jest zbadanie wpływu szeroko deniowanej otwartości naproces roz-
woju regionalnego. Wstępna analiza przyczynowości między kluczowymi zmiennymi
wskazuje nawystępowanie zależności dwukierunkowej. Wkolejnym kroku szacujemy
dynamiczny model panelowy za pomocą dwustopniowego estymatora uogólnionej
metody momentów ze względu naniestacjonarny charakter kluczowych zmiennych.
Wprocesie estymacji uwzględniamy potencjalne interakcje otwartości zregionalnymi
zasobami kapitału ludzkiego oraz innymi ważnymi determinantami postulowanymi
przez modele teoretyczne. Uzyskane wyniki sązgodne zpodstawowymi postulatami
teoretycznymi.
Słowa kluczowe: rozwój regionalny, wzrost gospodarczy, dane panelowe, Polska,
Hiszpania