An interregional impact analysis of the EU structural funds in Spain (1995-1999)

Page 1

An interregional impact analysis of the EU structural funds inSpain (1995–1999)*

Julián Pérez1, Milagros Dones1, Carlos Llano2

1 Applied Economics Department. L.R.Klein Institute (Centro Stone) and CEPREDE. Facultad de CC.EE y EE.Módulo E-XIV. Universidad Autónoma de Madrid. Campus Cantoblanco. 28049 Madrid, Spain

2 Economic Analysis Department and L.R.Klein Institute (Centro Stone). Facultad de CC.EE y EE. Módulo E-I.Universidad Autónoma de Madrid. Campus Cantoblanco. 28049 Madrid, Spain.(e-mail: [email protected], [email protected], [email protected])

Received: 25 October 2007 / Accepted: 21 August 2008

Abstract. This paper uses an interregional input output model to estimate the economic impactof the EU structural funds received by the Spanish regions during the period 1995–1999. Weattempt to cast further light on the interregional effects that the funds have produced in termsof output, value added and employment, not just in the regions where they were originallyallocated, but also in the rest of the regions. This analysis offers additional information than theone attained using macroeconomic models which do not take interregional spillovers intoaccount. The results are relevant regarding the discussion about the effectiveness of EU cohesionpolicy, and the share of output effects that are captured by the richest regions through theirintersectoral linkages.

JEL classification: C67, R11, R15, R58

Key words: Interregional input-output model, interregional trade, EU structural and cohesionfunds, origin and destination matrices, regional convergence

1 Introduction

According to the European Commission’s Third Report on EU Cohesion (EC 2004), “disparitiesin income and employment across the EU have narrowed over the past decade, especially sincethe mid-1990s”. Despite this process of convergence among the EU member states, it has alsobeen observed an increase in disparities among regions within countries (Esteban 1994, 2000).According to different studies (Dall’erba and Le Gallo 2003; Le Gallo and Dall’Erba 2006,2008), the EU integration seems to have benefited mainly the richest regions in the poorest

* We acknowledge helpful comments from Antonio Pulido, Emilio Fontela, Geoffrey Hewings, Jan Oosterhaven andtwo anonymous referees on early versions of this paper. We would also want to thank all the people and institutions thatsupplied the required statistics and participated in the INTERTIO Project where the first version of the interregionalinput-output model was developed.

doi:10.1111/j.1435-5957.2008.00212.x

© 2008 the author(s). Journal compilation © 2009 RSAI. Published by Blackwell Publishing, 9600 Garsington Road,Oxford OX4 2DQ, UK and 350 Main Street, Malden MA 02148, USA.

Papers in Regional Science, Volume 88 Number 3 August 2009.

Page 2

countries, rather than the poorest regions in the poorest countries. In fact, the European Com-mission (EC 2004) admits that although “GDP per head in Objective 1 regions has convergedtowards the EU average . . . since 1994, the extent of convergence has varied markedly betweenregions, mostly reflecting their relative importance in the Member States in which they aresituated. In the four Cohesion countries (Ireland, Greece, Portugal and Spain) . . . growth ofGDP per head was much higher than in the rest of the EU. The increase was particularly largein Ireland and was even larger in Spain”.

Apart from the EC official analysis, a number of academic works focused on the Spanishregions have demonstrated the important contribution of EU Funds to regional convergence interms of per capita income and employment during the period 1986–2006 (Herce and Sosvilla1994, 1995, 2004; Sosvilla 2005; De la Fuente 2002, 2003). These results partially contradict theprevious series of results obtained for the period 1986–1994 which appeared to show how theEU structural and cohesion funds could have actually fostered regional divergence (Boldrin andCanova 2001; Canova 2001; Martin 2001).1

All these works have primarily focused on the estimation of the macroeconomic effectsinduced by the EU funds inflow within the Objective 1 Spanish regions using macroeconomicmodels. Although these models provide a correct approach to analyse such effects since theycapture the interaction between supply and demand, they fail specifically to take into consid-eration the interregional spillovers that could either accelerate or restrain the catch-up process.

In this respect, alternative studies analysing the convergence among the European regionshave convincingly argued for the specific employment of spatial econometric approaches,emphasising the convenience to control the spatial autocorrelation effects induced by theinterregional spillovers (Quah 1996a, 1996b; López-Bazo et al. 1999, Dall’erba and Le Gallo2003, 2008; Armstrong 1995; Fingleton 1999; Moreno et al. 1999; Le Gallo and Dall’Erba,2006; Le Gallo, 2004, 2003b; Baumont et al. 2003, Geppert and Stephan 2008, for Europeanregions; Rey and Montouri, 1999; Rey, 2001, for US states). Furthermore, Le Gallo andDall’erba (2008) argue that “in the absence of interregional input/output tables”, spatial econo-metrics is the most accurate approach to model spatial effect among European regions.

Although in many of these works the interregional spillovers are not specifically defined,they are usually assimilated to the productivity gains that the investments in one region mayinduce in their neighbours as a consequence, for example, of better transport infrastructures,which may improve accessibility, reduce travel times and transport costs. While these effects arecritical, they are sometimes mixed together with other sources of spillovers that may not be onlyconditioned by contiguity or distance. This is the case, for instance, of interregional spilloverderived from the inter-sectoral linkages between any pair of regions.

Based on this debate about the real effects of EU funds on regional convergence and theconvenience of including the interregional effects induced by the funds, an interregional input-output model is used in this paper to estimate the economic impact of the EU Structural andCohesion Funds (EUSCF) received by the Spanish regions for the period 1995–1999. With thisapplication, we attempt to cast further light on the intersectoral interregional effects that thestrong inflow of funds has produced during the period, showing additional findings to thoseobtained from macroeconomic models which do not take interregional spillovers into account.We also expect to offer a complementary view to the results obtained by the use of spatialeconometric models, where intermediate sectoral linkages are hardly considered and spatialspillovers are reduced to contiguity or closest neighbours. In this regard, we lend further supportto the argument advanced by other authors (Nazara and Hewings 2007) about how interregional

1 Among the four southern countries, only Spain displays a weak form of reduction in regional income inequalitiesbetween 1986–1994, while in Italy, Portugal and Greece, territorial disparities have not decreased. Additionally, newlosers have appeared within some of the richest countries, e.g., Germany, France and the Benelux.

510 J. Pérez et al.

Papers in Regional Science, Volume 88 Number 3 August 2009.

Page 3

spillovers usually exceed contiguity relations, since they are also linked to the upward/downward sectoral linkages driving the regional economies.

Finally, we also expect to offer new results to contribute to the on-going discussion about theeffectiveness of EU cohesion policy. More specifically, we will attempt to quantify how muchshare of the output effects induced by the funds originally attributed to the poorest regions iseffectively captured by the richest ones through their intersectoral linkages.

The paper is organized as follows. The main features of the EU regional and cohesion policywill be briefly reviewed in section 2. The main underpinnings of the 1995 Spanish interregionalinput-output model (1995 Spanish IRIO model) will be described in some detail in Section 3.Finally, Section 4 will explain the process of estimating the economic impact of the EUstructural funds in the Spanish regions during the period 1995–1999, offering detailed results interms of output, value added and employment.

2 The EU regional policy: Definitions, instruments and effectiveness

The main goal of regional policy in the EU is the attainment of both economic and socialcohesion. The funds are distributed among regions according to their specific difficulties.Although the criteria have changed over the years, the funds are mainly allocated to two priorityobjectives: the goal of flows assigned to Objective 1 is to promote the development andstructural adjustment of regions whose income level is below 75% of the EU average; incontrast, the flows assigned to Objective 2 are aimed to support the economic and socialconversion of areas experiencing structural difficulties. To date, Objective 1 has accounted forapproximately 70% of the funds. Their allocation among the eligible regions within eachobjective is based on the detailed regional development plans presented to the EuropeanCommission by each member state. The funds must not be used to replace existing publicinvestment (additionality principle) and have to promote the financing effort of national publicand private agents (co-funding principle). As a result, the contribution of the structural fundsunder Objective 1 to a specific project cannot exceed 75% of the total eligible volume.

In quantitative terms, the EU regional policy is one of the most important in the Union andcurrently accounts for more than 35% of the Union’s budget. For 1989–1993, the first twopackages allocated 68 billion ECU (at 1997 prices) and 200 billion ECU (1994–1999). The thirdpackage (2000–2006) accounted for another 200 billion EUROs (at 1999 prices). (Table 1 andTable 2 summarise the economic impact of the EU funds estimated by the EC, 2004 in thecohesion country).

Table 1. Economic effects of the structural and cohesion funds

Greece Ireland Portugal Spain EUR4

% GDP

1989–93 2.6 2.5 3 0.7 1.41994–99 3 1.9 3.3 1.5 22000–06 2.8 0.6 2.9 1.3 1.6

% Gross fixed capital formation

1989–93 11.8 15 12.4 2.9 5.51994–99 14.6 9.6 14.2 6.7 8.92000–06 12.3 2.6 11.4 5.5 6.9

Source: Structural and cohesion funds: commitment data up to 1999; forecastfor 2000–2006 (EC 2004).

511An interregional impact analysis of the EU structural funds in Spain (1995–1999)

Papers in Regional Science, Volume 88 Number 3 August 2009.

Page 4

Despite the large amount of funds invested in the poorest EU areas, inequalities betweenregions within each Member State have increased. According to the first results observed for the1988–1999 period, even the European Commission (EC) acknowledges the existence of atrade-off between the convergence process at the national and regional level within the EU. Inthis respect, the EC argues that internal disparities in terms of income between regions appearto widen initially as the economic activity tends to agglomerate in the most efficient areas. Asa result, the first stages of an economic catching up tend to foster specialization, polarization andregional disparities (EC 2004).

From a more pessimistic viewpoint, some authors (Boldrin and Canova 2001; Canova 2001)have suggested that the allocation of investments within the less developed regions is inefficientand hampers the catch-up process of the nation with the European core. This argument appearsto be further reinforced when one takes into account the additionality principle which conditionsthe regional allocation of national public and private investments on a redistributive rather thanefficiency basis.

Furthermore, Puga (2002) has pointed out other additional forces that could contribute toreduce the effectiveness of the EU regional policy. According to his argument, income dispari-ties have been traditionally explained on the basis of differences between regions in their factorendowments. In this context, the removal of obstacles to the movement of goods and/orfactors would by itself cause the convergence of factor returns and living standards. By con-trast, other authors (Krugman 1991; Krugman and Venables 1995, 1996; Puga 2002) maintainthat integration caused by a reduction in transport cost may foster sectoral specialization andincome inequalities between the countries or regions. This process of divergence can beexplained by the tendency of firms to agglomerate around big markets. This may create acumulative causation process that tends to increase regional differences. Moreover, Puga(2002) questions the effectiveness of the EU strategy towards the allocation of a major partof the EU funds in the improvement of transport infrastructure (i.e., large investments in theSpanish High Speed Train system). From the perspective of new economic geography, it isnot obviously clear that lower transport costs facilitate convergence: since it is easier for firmsin richer regions to supply poorer regions at a distance, this can accordingly harm, rather thanbenefit, the industrialisation prospects of less developed areas. From another perspective, arecent paper by Montolio and Solé-Ollé (2008) has examined the relation between investmentin road infrastructures and the performance of the Spanish provinces in terms of total factorproductivity. The authors found that road investment is more productive in those provinceswhere there is a higher level of private inputs that directly use the services provided by roadinfrastructures, and suggest that this result seems to favour ‘efficiency’ arguments for the‘optimal’ allocation of public resources.

From an empirical point of view, there exists a large and often conflicting number ofanalyses. According to the official analysis from the EC (EC 2004), the impact of the EUSCF

Table 2. Economic effects of the structural and cohesion funds

Greece Ireland Portugal Spain

GDP UR GDP UR GDP UR GDP UR

1989 4.1 -3.2 2.2 -1.4 0.8 -0.5 5.8 -3.61993 4.1 -2.9 3.3 -1 1.5 -0.8 7.4 -4.11999 9.9 -6.2 3.7 -0.4 3.1 -1.6 8.5 -42006 7.3 -3.2 2.8 0.4 3.4 -1.7 7.8 -2.82010 2.4 0.4 2 0.5 1.3 -0.4 3.1 -0.1

Source: ESRI estimations based on HERMIN model 2000 (EC 2004). UR:Unemployment rate.

512 J. Pérez et al.

Papers in Regional Science, Volume 88 Number 3 August 2009.

Page 5

is far from negligible. Based on the HERMIN model (EC 2004), the contribution of thestructural funds in Cohesion countries in the 1993–1999 period ranks from the 1.5% of GDP inSpain (1993) to the 9.9% in Greece (1999) (see Table 2 and Table 3). The impact in terms ofunemployment rate is also significant, ranging from -0.8% in Spain (1993) to -6.2% in Greece(1999). In the case of Spain, according to the same estimation, the funds lead to an averageincrease of 1.4% in GDP, 1.5% in employment and 9.1% in fixed investment. Within thisinvestment effort, the main improvement was directed towards promoting accessibility as theinvestment in transport infrastructures absorbed almost 40% of the funds.

Regarding Objective 2 regions a total of 82 regions with 62 million inhabitants (17% of EU15 population) received during the period 1994–1999 Objective 2 assistance, primarily aimed athelping areas affected by industrial decline. The funds concentrated in a large number of smallareas, mainly located in UK, France, Spain and Germany. The EC estimates that interventionsunder Objective 2 during the period led to the creation of 700,000 jobs in the whole of the EU.However, the growth in GDP per capita in these regions was below EU average over the period(2.1% between 1995 and 2000 as opposed to 2.4%), inducing lower growths in labour produc-tivity compared to the rest of the EU (EC 2004).

Apart from the EC official analysis for the whole EU, a series of recent works focused on theSpanish regions have estimated the effects of EU Funds in terms of per capita income andemployment during the period 1986–2006 (Herce and Sosvilla 1994, 1995, 2004; De la Fuente2002, 2003; Sosvilla 2005).

For example, by means of a macroeconomic model that captures the leverage effect of theEU funds on public and private national investments, De la Fuente (2002, 2003) have estimatedthat the cumulative contribution until 2000 of the EU funds alone to the GDP in Spain accountsfor 4.82% in the Objective 1 regions and 2.4% in Spain as a whole. Furthermore, when thepublic expenses mobilized by the EU funds (additionality and co-funding principles) are alsotaken into account, the cumulative effect reaches a 6.92% of the GDP in the Objective 1 regionsand 3.44% in Spain. Despite this clearly positive effect, De la Fuente (2002, 2003) questions thetrade-off between reducing regional disparities within Spain and the catching-up process of thecountry towards the EU. By means of three alternative scenarios, he compares the evolution ofSpain and the Objective 1 regions if the investments were allocated following efficiency criteria.In this context, he found that the re-distributive allocation of the EU funds and the nationalbudget conditioned by them have a considerable opportunity cost. In fact, an alternative allo-cation of the same investment following efficiency strategies would have led to a one-thirdincrease in the convergence process of Spain with the EU average, although it would have alsoresulted in an increase in the internal regional differences.

Linked to this controversy about the optimal allocation of funds, little has been said aboutthe interregional distribution of the effects within the country. In this respect, although the ex

Table 3. Ex-post macroeconomic effects of structural policy 1994–1999:HERMIN simulation results (% difference from baseline without policy

in 1999)

Greece Ireland Portugal Spain

GDP 2.2 2.8 4.7 1.4Manufacturing output 3.4 4.7 10.6 3.7Market service output 2.4 2.4 4.8 1.2Fixed investment 18.1 1.1 24.8 9.1Labour productivity 2.3 2.2 6.6 2.1Employment 1 4.7 3.7 1.5

Source: DG Region (EC 2004).

513An interregional impact analysis of the EU structural funds in Spain (1995–1999)

Papers in Regional Science, Volume 88 Number 3 August 2009.

Page 6

ante allocation of the funds is based on a detailed analysis of the projects and the regions thatwould manage the funds, the effective allocation of the effects in terms of output and employ-ment has been hardly considered. Paradoxically, while the EC usually points out to the indirecteffects that EU funds assigned to poor countries bring to rich countries, described by the EC asthe ‘leakage effect’ (EC, 2004), nothing is said about the interregional effects that investmentsin poor regions may induce in the richest regions within the poor country. Given the high levelof regional integration within a country, one may expect that, at least in the short run, a large partof the effects induced by the projects funded in the poorest regions (Objective 1) would migrateto the richest (Objective 2), where the big companies (from sectors with the higher added value)tend to be located.

Based on this assumption, we will briefly describe in the next section the first interregionalinput-output model for the Spanish economy that allows estimating the effective allocation ofthe effects induced by the EUSCF received by the Spanish regions during the period 1995–1999.

3 The interregional input-output model

Following classical notation for the single region input-output model, we consider an economyof n sectors and one region R. The output of sector i is denoted by xi(i = 1, . . . n), which satisfiesequations 1 and 2, where f denotes the n element vector of final demands, and A, the matrix oftotal input coefficients obtained as a z xij ij j= , or in matrix terms as x Zx= −ˆ 1, (Z being thematrix of inter-industry demand).

x Ax f= + (1)

x I A f Lf= −( ) =−1 (2)

Based on this single-region model, Isard (1951) developed the ‘interregional input-output model(IRIO), where the spatial and sectoral origin and destination of the final and intermediatedemand are captured. Following classical notation (Miller and Blair 1985), an IRIO model withN regions and n sectors is denoted as:

x

x

x

A A A

A A A

A A AN

N

N

N N NN

1

2

11 12 1

21 22 2

1 2

�

��

� � � ��

⎡

⎣

⎢⎢⎢⎢

⎤

⎦

⎥⎥⎥⎥

=

⎡

⎣

⎢⎢⎢⎢⎢

⎤

⎦

⎥⎥⎥⎥

×

⎡

⎣

⎢⎢⎢⎢

⎤

⎦

⎥⎥⎥⎥

+

⎡

⎣

⎢⎢⎢⎢

⎤

⎦

⎥⎥⎥⎥

x

x

x

f

f

fN N

1

2

1

2

� �(3)

Z

Z Z Z

Z Z Z

Z Z Z

A

A AN

N

N N NN

=

⎡

⎣

⎢⎢⎢⎢

⎤

⎦

⎥⎥⎥⎥

=

11 12 1

21 22 2

1 2

11 12��

� � � ��

�

;

AA

A A A

A A A

x

x

x

x

N

N

N N NN N

1

21 22 2

1 2

1

2�� � � �

��

⎡

⎣

⎢⎢⎢⎢

⎤

⎦

⎥⎥⎥⎥

=

⎡

⎣

⎢⎢⎢⎢

⎤

⎦

;⎥⎥⎥⎥⎥

=

⎡

⎣

⎢⎢⎢⎢

⎤

⎦

⎥⎥⎥⎥

; f

f

f

f N

1

2

�

For Z, we have a block matrix with ZRL matrices that capture the inter-industry relationsbetween regions R and L. The ZLL on-diagonal elements capture the intermediate flows withinthe L region (intraregional intermediate flows). By contrast, ZRL off-diagonal elements containthe inter-industry interregional flows: zRL

ij is the amount of product generated by sector i inregion R that is being used as an intermediate good by sector j in region L (interregionalinter-industry flow).

514 J. Pérez et al.

Papers in Regional Science, Volume 88 Number 3 August 2009.

Page 7

The A matrix is also a block matrix containing the intraregional input coefficients

az

xijLL ij

LL

jL

= in the on-diagonal elements, and the interregional input coefficients, az

xijLM ij

LM

jM

= , in

the off-diagonal ones.The solution of the model is also obtained by applying the Leontief matrix to the x, A and

f matrices, which now contain sectoral and regional data organized by blocks:

x I A f

x

x

I 0

0 I

A A

A

L

N

LL LN

N

= −( )

⎡

⎣

⎢⎢⎢

⎤

⎦

⎥⎥⎥

=⎡

⎣

⎢⎢⎢

⎤

⎦

⎥⎥⎥

−

−1

��

� � ��

�� � �

LL NN

L

NA

f

f��

⎡

⎣

⎢⎢⎢

⎤

⎦

⎥⎥⎥

⎛

⎝

⎜⎜⎜

⎞

⎠

⎟⎟⎟

×⎡

⎣

⎢⎢⎢

⎤

⎦

⎥⎥⎥

−1

(4)

One of the critical elements on the estimation of the IRIO model is the availability of data on theinterregional flows. Usual statistical systems do not generally satisfy the sectoral and spatialdetail that is required by the pure IRIO model. As a result, a number of alternative specificationsthat require less amount of information have been developed. This is the case, for example, ofthe MRIO model (Chenery-Moses approach) or the ‘Pool-approach’ (see Batten 1983; Ooster-haven 1984; Miller and Blair 1985 for a detailed description). Despite these classical develop-ments of the interregional framework, we find recent examples of MRIO and IRIO models in theliterature for estimation of impact analysis at the sectoral and regional level (Ichimura and Wang2003; Ishikawa and Miyagi 2004; Park 2006; Park et al. 2007). In the next section, we willbriefly review the estimation procedure used for the first Spanish interregional input-outputmodel (SIRIO), which was described in detail by Llano (2004a, 2004b).2

3.1 The 1995 Spanish interregional input-output model

Following Van der Linden et al. (1995), the estimation of the 1995 SIRIO table may be seen bothas the disaggregation of the 1995 National IO table, or as the interconnection of a full-set of 181995-Single-Region IO tables (SRIO), one per each of the Spanish regions at the NUTS 2 level.This interconnection would remain on a parallel database on interregional trade flows byproducts.

Regarding the estimation of the 18 SRIO tables, since not all of the 18 Spanish regions hada survey 1995 SRIO, the bi-proportional RAS procedure (Miller and Blair 1985) was used forupdating old ones or for estimating them using SRIOs from regions with similar sectoralstructure. Finally, all the SRIO integrated in the IRIO were harmonized with the 1995 nationalinput-output table (NIO-95) and the Spanish regional accounts (SRA), using RAS techniques.3

With regards to the interregional trade database, a complete set of 26 interregional trade matrices(I = 26 sectors) was estimated. The estimation of the commodity flows (Llano 2004a, 2004b)combined the available data on Spanish transport flows of goods (road, rail, ship and aircraft)with additional information used to estimate specific export prices, one per region of origin,

2 Based on the similarities between our model and the EU IRIO (Van der Linden et al. 1995), we consider the Spanishmodel as an interregional IO model rather than a multiregional IO model. An argument in favour of this definition comesfrom the fact that our model includes full-survey data for the intraregional block matrices, while a combination of surveyand non-survey data for the interregional ones. In this regard, the model could be considered as a hybrid between theso-called ‘multiregional-columns-only input-output table’ and its ‘interregional-columns-only’ equivalent (Oosterhaven1984).

3 The model was built for 1995 because the widest set of survey regional input-output tables was available for thisyear.

515An interregional impact analysis of the EU structural funds in Spain (1995–1999)

Papers in Regional Science, Volume 88 Number 3 August 2009.

Page 8

transport mode and type of product. The methodology also included a process for debugging theoriginal transport flows database, which allows the identification and reallocation of multi-modal movements and international transit flows hidden in the interregional trade. By contrast,interregional flows of services were estimated using gravity models. Once the 26 Ti tradematrices (18 ¥ 18) had been estimated for 1995, we proceeded to the interconnection of the 18SRIO tables considering the assumption of equal spatial structure for every intermediate andfinal interregional import along rows (Oosterhaven 1984; Van der Linden and Oosterhaven1995). See Table 10 in the Appendix for a description of the 26 sectors considered.

4 Application: The impact of the EU structural funds in the Spanish regions(1995–1999)

In this section, the 1995 Spanish IRIO model is used for the estimation of the impact of the EUstructural and cohesion funds (EUSCF) received by each of the 18 regions during the period1995–1999.

First, a database with the EUSCF was built using the information published by the FinancialAccounts of the Central Bank of Spain. To capture the total amount of funds, non-regionalizedfunds (36% of total) were allocated according to the regional distribution of gross fixed capitalformation in the 1995 Spanish IRIO. Thus, the final value of the funds assigned to each regionr in each year t, was:

tf rfgfkf

gf kf

nrftR

tR t

R

tR

R

t= + ∗

=∑

1

18 (5)

where tftR is the total funds received by each region R in year t, rft

R the regionalized fundsreceived directly by each region R in year t; gfcft

R the gross fixed capital formation in each ofthe regions R, in year t, and nrft the non-regionalized funds per year t.

Using matrix notation, the vector of total funds received by each region in year t (forconvenience, time index is omitted) can be expressed as tf = rf + nrf, where nrf = [gfkf*k]*nrf,k being the n element summation vector of the gross fixed capital formation vector, and nrf thetotal amount of non regionalized funds in year t.

Table 4 shows the total funds allocated in each region between 1995 and 1999, obtained bymeans of Equation 5. The table also includes the regional structure of the average GDP along theperiod (third column) and the difference between the regional shares in terms of both variables(fourth column). Additionally, a last column has been added with a per capita GDP index(1995–1999 average).

The aggregates for Objectives 1 (in italics) and 2 at the bottom of the table, shows how thepoorest regions have, at this first step, clearly benefited by the funds: in agreement with ECregulation, poor regions receive more than 70% of the funds while they just produce less thanhalf of the national GDP. Based on the original allocation of the funds, the regions that benefitedmost in relative terms are Ceuta and Melilla, followed by Andalusia and Galicia. In contrast, theregions with the largest difference between the Funds and GDP structure are the richest ones,namely, Madrid and Catalonia. Consequently, the convergence objective of the EU Funds seemsto be guaranteed at this level of the analysis.

Once the funds were territorialized, a complete set of primal impact vectors of final demandby regions was obtained.

It is then considered that part of the funds had been expended in taxes or foreign products,which induced no direct impacts in national sectors. Thus, total funds allocated in each region

516 J. Pérez et al.

Papers in Regional Science, Volume 88 Number 3 August 2009.

Page 9

were corrected by tax effect (VAT and other production taxes) and foreign imports. Bothcorrections were made using the corresponding shares on the GFCF from the Spanish 1995IRIO. Using the same information, the funds invested in national demand were split in twocategories – the ‘own region’ and the ‘rest of Spain’ – and allocated to the corresponding 27sectors in each region. Thus, a matrix of set of impact column vectors of net domestic demandDt, with 513 rows ([18 regions +1 extra-regio area] ¥ 27 sectors4) and 19 ([18 regions +1extra-regio area] columns was obtained. Then by the summation along the rows, a set of impactcolumn vectors of net domestic demand dt = Dti, with 513 elements ([18 regions +1 extra-regioarea] ¥ 27 sectors), was obtained. Using matrix notation, the impact vector of net nationaldemand originated by the funds in a specific year is expressed in Equations 6, 7 and 8.

nd tf tx fm= − −( ) (6)

end nd I i= ⊗( ) (7)

d end gfcf= ∗ (8)

Where for nd, we have the column vector of the funds invested in each region in year t minustaxes (tx) and foreign imports (fm). Then, the nd vector is expanded to another column vectorwith order (N*n,1), which contains the allocation of the net funds invested in each region (N) and

4 Although there are 18 regions in Spain at the NUTS 2 level, in coherence with the Spanish Regional Accounts, the1995 Spanish IRIO includes an additional element, called ‘extra-regio’, which captures the output of non territorializedactivities.

Table 4. Total EUSCF per region in 1995–1999 (in millions of Euros of 1995)

Total fundsreceived

(Million €)

Fundsstructure

GDPstructure

Differences Per capita GDP

Funds-GDP Spain = 100

Total 34,635 100.0% 100.0% 0.0% 100.0Andalusia 6,068 17.5% 13.4% 4.2% 74.0Aragon 914 2.6% 3.2% -0.6% 107.0Asturias 1,290 3.7% 2.3% 1.4% 85.4The Balearic Islands 358 1.0% 2.4% -1.4% 122.5Canary Islands 1,725 5.0% 3.9% 1.1% 96.1Cantabria 610 1.8% 1.2% 0.5% 92.2Castile-León 2,975 8.6% 5.9% 2.7% 93.4Castile-La Mancha 1,887 5.4% 3.5% 1.9% 81.2Catalonia 3,438 9.9% 19.0% -9.0% 122.2Valencian Com. 3,352 9.7% 9.6% 0.1% 96.0Extremadura 1,336 3.9% 1.7% 2.2% 63.5Galicia 3,366 9.7% 5.4% 4.3% 80.2Madrid 1,915 5.5% 17.1% -11.6% 132.8Murcia 896 2.6% 2.3% 0.2% 83.1Navarre 508 1.5% 1.7% -0.3% 126.8Basque Country 1,825 5.3% 6.3% -1.0% 120.2Rioja 151 0.4% 0.8% -0.3% 114.3Ceuta Melilla 2,007 5.8% 0.3% 5.5% 84.8Extra-regional 16 0.0% 0.1% 0.0%

Total Objective 1 (italics) 25,510 73.7% 49.5% 24.19% 84.5Total Objective 2 9,125 26.3% 50.5% -24.19% 120.8

Source: Own elaboration from Bank of Spain data. Financial Accounts 1995/2000 III and INE National Accounts.

517An interregional impact analysis of the EU structural funds in Spain (1995–1999)

Papers in Regional Science, Volume 88 Number 3 August 2009.

Page 10

sector (n). Using matrix notation, the estimation of this vector is summarized in Equation 7 and8: first, the nd column vector of net funds by regions (N,1) is transformed in an expanded endcolumn vector with order N*n. This expanded end vector is obtained by the addition along therows of the matrix obtained by the multiplication of nd and an identity matrix I of order (n*n)using the Kroeneker product; second, end is multiplied by the vector gfcf (N*n), which capturesthe column vector of the gross fixed capital formation included in the 1995 Spanish IRIO table,expressed in terms of sectoral percentage over the N region subtotals.

Thus, for each year, we obtain a column impact vector d that captures the initial demandshock which producers in each sector and region were likely to have received as a consequenceof the funds originally allocated by the EC according to their redistributive strategy. Table 5summarizes the allocation of aggregate funds expanded during the period 1995–1999 in con-stant euros of 1995, assuming the territorial and sectoral structure from the Spanish IRIO 1995.

According to Table 5, the average distribution of the 34,634.86 millions of euros investedduring the period 1995–1999 was: 5% in taxes, 12% in international imports, 66% in productsfrom the same region and 18% from other regions. Furthermore, Table 5 shows how different isthe distribution of total funds depending on the region that received them, at least regarding twodifferent aspects: first, the amount of final demand imported from the rest of the world, andsecond, the final demand imported from other regions.

In accordance with a progressive tax system, the proportion of the funds that goes totaxes in Objective 1 regions (the poorest ones) is lower than in those included in the Objective2 category (4.9% versus 5.3%). Regarding the spatial distribution of the demand inducedby the original funds, the share of funds expended in foreign imports seems to be higher in

Table 5. Regional allocation of net funds. Average 1995–1999 (in millions of 1995 Euros)

Funds Taxes Imports National demand

Total(1) + (2)

Own region(1)

Rest of Spain(2)

National total 34,635 5.0% 11.7% 83.3% 65.7% 17.6%Andalusia 6,068 4.2% 2.8% 93.0% 73.1% 19.9%Aragon 914 4.7% 15.6% 79.7% 62.9% 16.8%Asturias 1,290 5.2% 4.2% 90.6% 72.5% 18.1%The Balearic Islands 358 3.1% 4.4% 92.6% 55.9% 36.7%Canary Islands 1,725 4.0% 12.3% 83.6% 68.9% 14.8%Cantabria 610 5.2% 14.4% 80.4% 64.1% 16.3%Castile-León 2,975 5.6% 13.5% 80.8% 59.7% 21.1%Castile-La Mancha 1,887 4.7% 3.2% 92.2% 74.8% 17.4%Catalonia 3,438 5.3% 21.0% 73.7% 59.4% 14.3%Valencian Com. 3,352 4.0% 12.6% 83.4% 61.3% 22.1%Extremadura 1,336 5.2% 1.2% 93.6% 90.2% 3.4%Galicia 3,366 5.4% 12.3% 82.3% 62.2% 20.2%Madrid 1,915 6.6% 23.9% 69.5% 67.6% 1.8%Murcia 896 4.8% 12.9% 82.3% 66.5% 15.8%Navarre 508 6.0% 11.9% 82.1% 60.3% 21.7%Basque country 1,825 4.3% 13.8% 81.9% 62.2% 19.7%Rioja 151 5.8% 11.0% 83.3% 65.5% 17.8%Ceuta Melilla 2,007 7.7% 20.9% 71.4% 50.3% 21.1%Extra-regional 16 5.1% 23.2% 71.7% 0.2% 71.5%

Total Objective 1 (italics) 25,510 4.9% 9.3% 85.7% 67.0% 18.7%Total Objective 2 9,125 5.3% 18.3% 76.4% 61.9% 14.5%

Source: Own elaboration.

518 J. Pérez et al.

Papers in Regional Science, Volume 88 Number 3 August 2009.

Page 11

the richest regions (14% for Objective 1 and 24% for Objective 2), while the dependenceon imports from the rest of Spain is higher in the poorest (19% for Objective 1 and 15% forObjective 2).

Based on the impact vectors deduced in the previous steps and the first redistribution effectthat arises when taxes and international imports are discounted from the original funds received,the next section analyses the territorial distribution of the interregional spillovers induced by thefunds invested in Spain during the period.

4.1 Analysis of direct and indirect effects of the EU structural funds by region

The estimation of direct and indirect effects induced by the increases of final demand wascalculated applying the Isard’s extension of Leontief’s model denoted in Equation 4. For eachyear, by the application of Equation 9 to the impact vectors d, a Dx vector of total output effectis obtained

Δ Δx I A d= −( )−1 (9)

Then, by the elimination of the direct effects from the total output effects Dx, the indirect effectsDo were obtained:

Δ Δ Δo x d= − (10)

Finally, by the addition of all the elements included in the Dx vector, we obtain the total outputeffect induced by the funds invested in Spain in a specific year. Moreover, this procedure allowsdistinguishing between direct and indirect output effects captured by the region that originallyreceived the funds and each of the remaining regions in the country.

In addition to the output effects, to go one step further, it is interesting to estimate the effectsinduced by the EU funds in terms of value added and employment. These effects were obtainedby the estimation of Equations 11 and 12, defined using classical notation from input-outputliterature (Miller and Blair 1985; Pulido and Fontela 1993):

Δ Δ Δg vx v I A d= = −( )−ˆ ˆ 1 (11)

Δ Δ Δe ex e I A d= = −( )−ˆ ˆ 1 (12)

In Dg and De we have, respectively, the two column vectors of increases in value added andemployment as a consequence of an increase in the final demand induced by the impact vectorin each year. Both equations are based on the two vectors v and e, which refer, respectively, tothe value-added and employment coefficients obtained according to the following expressions:

Value added coefficients: vv

xjR j

R

jR

=

Employment coefficients : 5 ee

xjR j

R

jR

=

5 Before the estimation of the employment coefficients, they have been projected during the period assuming thelabour productivity gains observed in each region in the Spanish Regional Accounts (INE).

519An interregional impact analysis of the EU structural funds in Spain (1995–1999)

Papers in Regional Science, Volume 88 Number 3 August 2009.

Page 12

Where x jR is output of sector j in region R; vj

R is the euro value of the value added that isrequired to produce one unit of product in sector j in regions R; ej

R is the number of peopleemployed that is required to produce one unit of product in sector j in regions R.

The results finally obtained have been indicated in Table 6, which shows the regionalallocation of the funds originally assigned to each region, along with information about totaloutput, value added and the net demand received, broken down into intra- and interregionalorigins. The last column in Table 6 shows the ratio between funds invested and the value addedgenerated in each region. The differences observed among these ratios will give a first quanti-fication about which regions take more returns from the EU funds managed in his region or inthe rest of the country.

According to Table 6, the average ratio between value-added and the Funds is 0.7, whichmeans that almost one quarter of the total EC Funds are absorbed by foreign imports and taxes.In this regard, it is important to bear in mind that part of this remaining 0.3% would bedistributed by the public sector (European, national, regional and local governments) in terms ofpublic expenses, while the other part would flow back to the rest of the EU (and the rest of theWorld) as intercountry spillovers (see the EC 2004 for the debate about the contribution of EUfunds to the European Single Market; see also Dietzenbacher and Van der Linden 1997,Dietzenbacher et al. 1993 and Llano 2004a for a quantification of the intercountry spilloverscaptured by EU IRIO model).

Returning to the regional figures shown in Table 6, the range of this ratio varies significantly,oscillating from the 0.3 observed in Ceuta and Melilla6 to the 1.8 in Madrid. These differencesare induced by the original structure of the funds, the size, openness and sectoral specializationof the regions. As a result, the highest ratio appears in the most powerful regions in terms ofproduction and interregional trade, which are able to capture a large part of intersectoral demandshocks due to their strong linkages with other dependent regions.

In terms of total output and value-added effects, we could assume that winners are regionswith ratios higher than the national average, while losers are regions with ratios below it. Ingeneral, Objective 1 regions show lower ratios than those shown by Objective 2 group onaverage (0.6 and 1.0, respectively). There are just two exceptions to this general rule: on the onehand, Asturias, although included in the Objective 1 regions, has a ratio above the average as itcaptures more than 80% of the funds invested as value added; on the other hand, the BalearicIslands, one of the richest regions in the country in terms of per capita income, captures only45% in return from the funds invested. In addition to this, it is also worth emphasizing the strongratio observed in the Madrid region, whose value added generated is 1.8 times the amount offunds originally received following a redistributive strategy.

The impact analysis of EC Funds can be completed with the data shown in Table 7, whichcaptures the relative effects induced by the funds in terms of VA and employment with respectto the GDP and total employment in the region. Based on the figures, when the analysis focuseson the differential performance of regions in Objectives 1 and 2, it can be observed that finaleffects produced in Objective 1 regions come close to the expected results in terms of incomeconvergence observed by other authors following different methodologies.

The last two columns of Table 7 contain the average regional growth of value addedthrough the period and a dependence ratio that measures the relation between the averagecontribution of the Funds in percentage of the GDP (column 3) and the average growth rateof the value added (in column 4). The interpretation of this ratio is relatively straightforward:since the output effects obtained each year per region are supposed to be included in the realGDP of each region, we can accordingly expect that, without these funds and the investmentslinked to them, the effective growth of the region may be reduced in the same percentage as

6 Ratio for extra-regional data is not economically relevant.

520 J. Pérez et al.

Papers in Regional Science, Volume 88 Number 3 August 2009.

Page 13

Tabl

e6.

Tota

lou

tput

effe

ctco

mpa

red

tofu

nds

inve

sted

per

regi

on.T

otal

1995

–199

9(i

nm

illio

nsof

1995

Eur

os)

Fund

sN

etde

man

dre

ceiv

edO

utpu

tV

alue

adde

dV

A/f

unds

Tota

ltf

Intr

a-re

gion

Res

tof

the

coun

try

Tota

lDd

Indu

ced

DoTo

tal

DxTo

tal

DgR

atio

Nat

iona

lto

tal

34,6

3522

,748

6,09

928

,847

25,3

2654

,173

24,4

400.

71A

ndal

usia

6,06

84,

435

203

4,63

83,

705

8,34

33,

318

0.55

Ara

gon

914

575

116

691

642

1,33

362

30.

68A

stur

ias

1,29

093

624

91,

184

1,04

52,

230

1,05

60.

82T

heB

alea

ric

Isla

nds

358

200

1121

110

331

416

20.

45C

anar

yIs

land

s1,

725

1,18

813

71,

325

742

2,06

71,

112

0.64

Can

tabr

ia61

039

160

451

512

963

393

0.64

Cas

tile

Leo

n2,

975

1,77

650

82,

284

1,64

83,

932

1,80

10.

61C

asti

leL

aM

anch

a1,

887

1,41

175

1,48

71,

059

2,54

61,

298

0.69

Cat

alon

ia3,

438

2,04

21,

323

3,36

53,

948

7,31

33,

155

0.92

Vale

ncia

nC

om.

3,35

22,

055

337

2,39

22,

380

4,77

22,

013

0.60

Ext

rem

adur

a1,

336

1,20

523

1,22

963

01,

858

957

0.72

Gal

icia

3,36

62,

093

761

2,85

41,

617

4,47

12,

166

0.64

Mad

rid

1,91

51,

294

1,80

43,

098

3,66

26,

760

3,40

51.

78M

urci

a89

659

651

647

496

1,14

252

30.

58N

avar

re50

830

695

401

443

844

333

0.66

Bas

que

coun

try

1,82

51,

135

324

1,45

92,

121

3,58

01,

433

0.79

Rio

ja15

199

1411

317

128

412

70.

84C

euta

Mel

illa

2,00

71,

009

91,

018

400

1,41

756

40.

28E

xtra

-reg

iona

l16

01

12

22

0.11

Tota

lO

bjec

tive

1(i

tali

cs)

25,5

1017

,096

2,41

219

,508

14,2

3433

,742

15,2

000.

60To

tal

Obj

ectiv

e2

9,12

55,

652

3,68

79,

339

11,0

9220

,431

9,24

01.

01

Sour

ce:

Ow

nel

abor

atio

n.

521An interregional impact analysis of the EU structural funds in Spain (1995–1999)

Papers in Regional Science, Volume 88 Number 3 August 2009.

Page 14

the ‘dependence ratio’. In relation to this idea, the strong dependence of regional growth tothe EU Funds appears to be clearly evident in regions like Ceuta and Melilla where nearly 8%of income and employment are induced, either directly or indirectly, by these Funds. Thedependence ratio is also large in regions like Extremadura, Asturias, Galicia and Castile andLeon, where over half of the economic growth registered in the period was induced by thestructural and cohesion funds.

4.2 The redistributive effects of output

If we examine the full process, from the initial allocation of funds to the final generation ofthe Value Added in Table 8, it is easy to conclude that, although the EC funds have contrib-uted to the regional convergence (the final effects are higher in the group of Objective 1regions than in the Objective 2), the final effects are less redistributive than the initial allo-cation of the funds.

This progressive loss of effectiveness of the EU regional policy along the process fromallocation-to-production is derived from a net positive spillover in favour of the richest regions,which are able to capture significant shares of output effects generated in the poorest regions.

According to the figures shown in Table 8, three sources of redistribution can be identifiedin the process:

1. The first one comes from the final demand structure in each region: although rich regions aremore open to foreign imports, they show higher levels of intraregional demand and a lower

Table 7. Relative effects of VA and employment relative to GDP 1995–1999 average

EU Funds/real GDP

Employment effect/Total Employment

Value Added/real GDP

Average growthof real GDP

Dependence

National total 1.5% 1.4% 1.0% 3.5% 30.0%Andalusia 1.9% 1.5% 1.1% 3.4% 31.1%Aragon 1.2% 0.9% 0.8% 3.2% 25.0%Asturias 2.3% 3.9% 1.9% 2.6% 71.5%The Balearic Islands 0.7% 0.5% 0.3% 3.3% 9.5%Canary Islands 2.0% 1.6% 1.3% 4.0% 32.1%Cantabria 2.1% 1.6% 1.4% 3.5% 38.9%Castile-Leon 2.1% 1.6% 1.3% 2.5% 51.2%Castile-La Mancha 2.2% 2.6% 1.5% 3.6% 42.7%Catalonia 0.8% 0.8% 0.7% 3.3% 21.6%Valencian Com. 1.5% 1.0% 0.9% 3.7% 24.2%Extremadura 3.3% 3.2% 2.3% 3.1% 76.5%Galicia 2.6% 2.4% 1.7% 3.0% 56.4%Madrid 0.5% 0.9% 0.9% 4.0% 21.0%Murcia 1.6% 1.0% 1.0% 4.1% 23.2%Navarre 1.2% 0.9% 0.8% 3.6% 22.5%Basque country 1.2% 1.7% 1.0% 3.6% 27.3%Rioja 0.9% 0.8% 0.7% 4.0% 18.1%Ceuta and Melilla 30.3% 8.4% 8.5% 3.0% 284.6%Extra-regional 0.7% 0.1% 0.1% 1.7% 4.7%

Total Objective 1 (italics) 4.7% 2.6% 2.1% 3.3% 62.4%Total Objective 2 0.9% 0.9% 0.7% 3.6% 20.7%

Source: Own elaboration.

522 J. Pérez et al.

Papers in Regional Science, Volume 88 Number 3 August 2009.

Page 15

dependence on imports from the rest of the country. As a result, a larger part of the fundsoriginally allocated in Objective 1 regions would be expended in foreign and interregionalimports.

2. The second one is derived from the regional production structures of rich regions thatincrease their capability to capture induced output effects better than others. The poorestregions, by contrast, do not capture enough feedback effects induced over the richest ones.

3. Finally, the agglomeration of sectors with higher value-added/output ratios in the richestregions (headquarters, consulting and finance, public agencies, etc.) also contributes to thereduction of the redistributive effect of the EU funds as originally allocated.

Focusing on the data shown in Table 8, it can be observed that the advantage of Objective 1 overObjective 2 regions declines progressively. As a result, although the original allocation of thefunds clearly benefits the poorest regions by 47.4 points, the gap between the two groups isreduced to 35.2, 24.6 and 24.4 points as the three redistributive mechanisms come into play.The main redistribution effects are primarily caused by the first two mechanisms (demand andproduction structure), since they induce an increase in the Objective 2 group shares of 6.0 and5.3 percentage points, respectively.

By comparing the territorial distribution of EU funds and the total value-added effectsobtained with the model as shown in the last column of Table 8, it is easy to identify losers andwinners in terms of regional shares (and probably regional convergence). As a general rule,regions included in Objective 2 are gaining shares along the full process and, at the end theycapture 11.5 points of value added over the initial allocation of funds. It is particularly relevantthe gaining process registered in Madrid and Catalonia, where the total value-added effect isrespectively 8.4% and 3% higher than the amount of Funds originally managed. By contrast, all

Table 8. Redistribution effect (% of national total)

Funds received Demand received Output Value added Funds-VA

National total 100.0% 100.0% 100.0% 100.0% 0.0%Andalusia 17.5% 16.1% 15.4% 13.6% -3.9%Aragon 2.6% 2.4% 2.5% 2.6% -0.1%Asturias 3.7% 4.1% 4.1% 4.3% 0.6%The Balearic Islands 1.0% 0.7% 0.6% 0.7% -0.4%Canary Islands 5,0% 4,6% 3,8% 4,5% -0,4%Cantabria 1,8% 1,6% 1,8% 1,6% -0,2%Castile León 8,6% 7,9% 7,3% 7,4% -1,2%Castile La Mancha 5,4% 5,2% 4,7% 5,3% -0,1%Catalonia 9,9% 11,7% 13,5% 12,9% 3,0%Valencian Com. 9,7% 8,3% 8,8% 8,2% -1,4%Extremadura 3,9% 4,3% 3,4% 3,9% 0,1%Galicia 9,7% 9,9% 8,3% 8,9% -0,9%Madrid 5,5% 10,7% 12,5% 13,9% 8,4%Murcia 2,6% 2,2% 2,1% 2,1% -0,4%Navarre 1,5% 1,4% 1.6% 1.4% -0.1%Basque country 5.3% 5.1% 6.6% 5.9% 0.6%Rioja 0.4% 0.4% 0.5% 0.5% 0.1%Ceuta Melilla 5.8% 3.5% 2.6% 2.3% -3.5%Extra-regional 0.0% 0.0% 0.0% 0.0% 0.0%

Total Objective 1 (italics) 73.7% 67.6% 62.3% 62.2% -11.5%Total Objective 2 26.3% 32.4% 37.7% 37.8% 11.5%Difference between Obj 1–Obj 2 47.4% 35.2% 24.6% 24.4%

Source: Own elaboration.

523An interregional impact analysis of the EU structural funds in Spain (1995–1999)

Papers in Regional Science, Volume 88 Number 3 August 2009.

Page 16

the regions in Objective 1, except Asturias which gains 0.6 points and Extremadura that remainsalmost unchanged, registered negative differences between the amount of EU funds managedand the total effect in terms of value added.

Furthermore, the use of the Spanish IRIO allows identifying the main intra- and interre-gional total output effects, based on real data rather than assignments based on distance orcontiguity relations. Taking advantage of this key contribution, Table 9 shows the ranking of themain intra- and interregional total output effects.

According to columns 2 and 3, the regions with the higher intraregional relative effects arethe regions whose output effects are mainly induced by the funds originally allocated in theirterritory. This result is in line both with the larger amount of funds originally assigned to themas well as their weak capacity to capture spillovers from other regions. In this respect, it isinteresting to observe how some of the poorest and most peripheral regions like Extremadura,Andalusia or Castilla-la Mancha register high levels of intraregional dependence on their ownfunds. It is also interesting to find in this group the three regions located outside the IberianPeninsula, namely, Ceuta and Melilla, Balearic Islands and Canary Islands. This result appearsto indicate the difficulties faced by the regions with low levels of accessibility in terms ofself-sufficiency and ability to capture output effects from remote regions. In contrast, the lowestlevels of intraregional output effects are observed among the richest regions like Madrid,Catalonia, Navarre, La Rioja or the Basque Country. This result is in line both with the smallshare of funds that they originally received and with their higher capacity to attract interregionalspillovers from the funds originally allocated to the poorest regions.

Regarding the interregional spillovers (columns 4, 5, 6 in Table 9), the flows are expressedas the percentage between the output effects that the funds originally allocated in the regions incolumn (4) Funds receptor and induced in the regions in column (5) Producer region, comparedto the total amount of output effect generated in the producer region as a consequence of the

Table 9. Ranking of interregional cross effects as % of total output effects (average 1995–1999)

Rank Intraregional effects Interregional spillovers

Receptor = producer % Funds receptor Producer region %

(1) (2) (3) (4) (5) (6)

1 Ceuta and Melilla 98.4% Andalusia Catalonia 18.6%2 Extremadura 94.6% Andalusia Madrid 15.9%3 Balearic Islands 88.4% Andalusia Basque country 9.7%4 Andalucía 88.1% Andalusia Aragon 9.6%5 Canary Islands 84.4% Andalusia Com.Valenciana 9.6%6 Castilla-La Mancha 83.2% Andalusia Navarre 9.4%7 Murcia 78.5% Com.Valenciana Madrid 9.4%8 Cantabria 71.7% Basque Country Navarre 8.6%9 Castilla and Leon 68.3% Galicia Asturias 8.2%

10 Com.Valenciana 65.3% Com.Valenciana Catalonia 8.2%11 Galicia 64.6% Andalusia Cantabria 7.7%12 Aragon 64.4% Andalusia Murcia 7.7%13 Asturias 60.0% Ceuta Melilla Rioja 7.6%14 Basque Country 55.3% Castilla and León Galicia 7.5%15 Rioja 53.4% Galicia Madrid 7.5%16 Navarre 53.3% Castilla and Leon Madrid 6.7%17 Catalonia 41.7% Galicia Basque Country 6.6%18 Madrid 27.9% Galicia Catalonia 6.3%

524 J. Pérez et al.

Papers in Regional Science, Volume 88 Number 3 August 2009.

Page 17

funds invested in its own region and in the rest of the country. According to Table 9, for example,the highest ‘interregional spillover effect’ in Spain goes from Andalusia (the region that origi-nally received the funds) to Catalonia (the region that effectively produced the output), andaccounted for 1,358 millions of euros, that is, to 18.6% of all the output effect generated inCatalonia by the whole EU funds invested in all over the country (an average during the period).The next highest interregional spillovers go also from Andalusia to Madrid (15.9% of the outputeffects generated in Madrid), the Basque Country (9.7%), Aragon (9.6%), Valencian Commu-nity (9.6%) and Navarre (9.4%). In this respect, it is important to bear in mind that Andalusia isthe main receptor of Funds (17.5% according to Table 4), but is located far away from theproducer regions that are showing the highest interregional spillovers (see Figure 1 for anillustration of these large inter-regional spillovers). Furthermore, the first strong interregionalcross effect between neighbours appears in the 5th (Andalusia-Valencian Community) and 8th

positions (Basque Country-Navarre). These results suggest that although the intensity of theinterregional spillovers has something to do with the distance, they are also conditioned by othervariables like the openness, the population, the accessibility or the intensity of intersectorallinkages.

Finally, it should be noted that, although our analysis has primarily focused on the impact ofthe EU funds, similar redistributive effects would be obtained from other injection of funds withthe same spatial and sectoral allocation. However, there are several reasons that make the caseof the EU funds of particular interest. Among these particularities it should be emphasized thespatial redistributive aim of the Funds, the relative short range of projects to which they can beallocated (mainly in human and physical capital) and their capability to condition the distribu-tion of private and public investments through the co-funding principle.

MADRID

ANDALUSIA

CASTILE-LA MANCHA

MURCIA

BALEARIC ISLANDS

ARAGONCASTILE AND LEON

RIOJA

SPAIN

199,5

1,9

POPULATION %CANARY ISLANDS

Objective 1 regions (1995-99)

Objective 2 regions (1995-99)CEUTA AND MELILLA

MADRID

ANDALUSIA

CASTILE-LA MANCHA

ARAGONCASTILE AND LEON

RIOJA

SPAIN

199,5

1,9

POPULATION %

CEUTA AND MELILLA

MADRID

BASQUE COUNTRYASTURIAS

COM. VALENCIANA

ANDALUSIA

CASTILE-LA MANCHA

GALICIA

EXTREMADURA

ARAGONCASTILE AND LEON

RIOJA

CANTABRIA

NAVARRE

CATALONIA

SPAIN

199,5

1,9

POPULATION %

Fig. 1. Map of Spain with the main interregional spillovers induced by the EU funds (average flows from Table 9,1995–1999)

525An interregional impact analysis of the EU structural funds in Spain (1995–1999)

Papers in Regional Science, Volume 88 Number 3 August 2009.

Page 18

5 Conclusions

This paper has used an interregional input-output model to estimate the economic impact of theEU structural and cohesion funds (EUSCF) received by the Spanish regions during the period1995–1999. To the best of our knowledge, this is the first impact analysis of the EU funds inSpain that explicitly considers the intersectoral and interregional linkages. As a result, we havebeen able to show the effects induced by the funds not just in the regions where they wereoriginally allocated, but also in the rest of the country.

Some results are particularly relevant from a policy viewpoint. Like other previous workson this topic, this paper has shown, first of all, how the large amount of EU funds receivedby the Spanish regions during the period 1995–1999 has greatly contributed towards gener-ating wealth and employment in the poorest regions and to the convergence process within thecountry. Second, in spite of this first redistributive effect, the ratio between the output gen-erated and the funds originally received is higher in the richest regions (0.6 for the Objective1 regions versus 1.0 for the Objective 2). As a result, the redistributive effect of the originalallocation of the funds seems to have been partially counterbalanced by the strong capacity ofthe richest regions to capture interregional spillovers generated by the poorest. In fact, whenthe differences between Objective 1 and 2 regions are analysed along the process that goesfrom the first allocation of the funds to the final generation of the value-added, the finaleffects are 11.5% less redistributive than the initial ones. This result reveals the importantdifference that exists between the region which originally receives the EU funds and theregion which effectively obtains the returns. Likewise, this result serves to illuminate how EUfunds may actually hinder regional convergence, since rich regions can capture higher frac-tions of the indirect output effects induced by the EU funds assigned to the poorest regions.Apart from the income level and the economic size of the regions in terms of GDP, otherfactors like the population, the accessibility or the factor endowment of each region are alsorelevant in determining the level of self-sufficiency of a region and its propensity to emit andcapture large shares of interregional spillovers. Finally, the specific features of our model havemade possible for us to identify and analyse the main intra- and interregional spillover effectsinduced by the funds. Although most of the effects remain in the region where they wereoriginally assigned (63.9% versus 36.1%), we have found strong interregional spillovers thatgo from Andalusia to the most powerful regions (Catalonia, Madrid, the Basque Country,Aragon, Valencian Community and Navarre). Furthermore, the existence of such strong inter-sectoral spillovers between distant regions has brought to our attention how interregionalspillovers may exceed contiguity relations as they are also linked to the upward/downwardsectoral linkages driving the regional economies.

526 J. Pérez et al.

Papers in Regional Science, Volume 88 Number 3 August 2009.

Page 19

Appendix

References

Armstrong H (1995) An appraisal of the evidence from cross-sectional analysis of the regional growth process withinthe European Union. In: Armstrong H, Vickerman R (eds) Convergence and divergence among European regions.Pion, London

Batten DF (1983) Spatial analysis of interacting economics. Kluwer-Nijhoff Publishing, AmsterdamBaumont C, Ertur C, Le Gallo J (2003) Spatial convergence clubs and the European regional growth process, 1980–

1995. In: Fingleton B (ed) European regional growth. Springer, BerlinBoldrin M, Canova F (2001) Europe’s regions, income disparities and regional policies. Economic Policy 32: 207–253Canova F (2001) Are EU policies fostering growth and reducing regional inequalities? Opuscle No. 8, CREI-Universitat

Pompeu Fabra. URL: http://www.econ.upf.es/crei/research/opuscles/op8ang.pdfDall’erba S, Le Gallo J (2003) Regional convergence and the impact of European structural funds over 1989–1999: A

spatial econometric analysis. REAL 03-T-14. Working Paper. Regional Applications Laboratory.Dall’erba S, Le Gallo J (2008): Regional convergence and the impact of Eurpean structural funds over 1989–1999: A

spatial econometric analysis. Papers in Regional Science 87: 219–244De la Fuente A (2002) The effect of structural fund spending on the Spanish regions: An assessment of the 1994–99

Objective 1 CSF. CEPR Discussion Papers, 36733De la Fuente A (2003) El impacto de los fondos estructurales: Convergencia real y cohesion interna. (The impact of the

structural funds: Real convergence and internal cohesion. With English summary.) Hacienda Publica Espanola/Revista de Economia Publica 165: 129–148

Dietzenbacher E, van der Linden JA (1997) Linkages in EC productions structure. Journal of Regional Science 37:235–257

Table 10. Sectoral classification of the Spanish IRIO model

N NACE Name

1 AA + BB Agriculture, fishing2 CA + CB Mining industry3 DA Food and beverages4 DB Textile5 DC Shoes6 DD Wood7 DE Paper, edition8 DG Chemical industry9 DH Plastic

10 DI Non-metallic minerals11 DJ Metallurgy and metallic products12 DK Machinery and mechanical equipment13 DL Electronic and electrical material14 DM Transport material15 DN Other industries16 DF + EE Energy; water, gas, electricity distribution17 FF Construction18 GG Trade, car reparation19 HH Hotels and restaurants20 II Transport, communications21 JJ Financial intermediation22 KK House renting23 LL Public administration24 MM Education25 NN Health26 00 + PP + QQ Other social services

527An interregional impact analysis of the EU structural funds in Spain (1995–1999)

Papers in Regional Science, Volume 88 Number 3 August 2009.

Page 20

Dietzenbacher E, van der Linden JA, Steenge AE (1993) The regional extraction method: Applications to the EuropeanCommunity. Economic System Research 5: 185–201

Esteban J (1994) La desigualdad interregional en Europa y en España: Descripción y análisis. Crecimiento y conver-gencia regional en España y Europa 2: 13–82

Esteban J (2000) Regional convergence in Europe and the industry-mix: A shift-share analysis. Regional Science andUrban Economics 30: 353–364

European Commission (2004) Third report on economic and social cohesion: A new partnership for cohesion. EuropeanCommission, Luxembourg URL: http://ec.europa.eu/regional_policy/sources/docoffic/official/reports/cohesion3/cohesion3_en.htm

Fingleton B (1999) Estimates of time to economic convergence: An analysis of regions of the European Union.International Regional Science Review 22: 5–34

Geppert K, Stephan A (2008) Regional disparities in the European Union: Convergence and agglomeration. Papers inRegional Science 87: 193–217

Herce JA, Sosvilla-Rivero S (1994) The effects of the community support framework 1994–99 on the Spanish economy:An analysis based on the HERMIN model. Working Paper 94-10R, FEDEA. Madrid

Herce JA, Sosvilla-Rivero S (1995) HERMIN Spain. Economic Modelling 12: 295–311Herce JA, Sosvilla-Rivero S (2004) European cohesion policy and the Spanish economy: Evaluation and prospective.

Working Paper, FEDEA. Madrid. JanuaryIchimura S, Wang H (2003) Interregional input-output analysis of the Chinese economy. World Scientific, SingaporeIsard W (1951) Interregional and regional input-output analysis: A model of space economy. Review of Economics and

Statistics 33: 318–328Ishikawa Y, Miyagi T (2004) The construction of a 47-Region interregional input-output table, and inter-regional

interdependence analysis at prefecture level in Japan. Studies in Regional Science 34: 139–152Krugman PR (1991) Increasing returns and economic geography. Journal of Political Economy 99: 483–499Krugman PR, Venables AJ (1995) Globalization and the inequality of nations. Quarterly Journal of Economics 110:

857–880Krugman PR, Venables AJ (1996) Integration, specialization, and adjustment. European Economic Review 40: 959–967Le Gallo J (2004) Space-time analysis of GDP disparities among European regions: A Markov chains approach.

International Regional Science Review 27: 138–163Le Gallo J (2003b) A spatial econometric analysis of convergence across European regions, 1980–1995. In: Fingleton

B (ed) European regional growth. Springer, BerlinLe Gallo J, Dall’Erba S (2006) Evaluating the temporal and spatial heterogeneity of the European convergence process,

1980–1999. Journal of Regional Science 46: 269–288Le Gallo J, Dall’erba S (2008) Spatial and sectoral productivity convergence between European regions, 1975–2000.

Papers in Regional Science 87: 505–525Leontief W, Strout A (1963) Multiregional Input-Output Analysis. In: Barna T (ed) Structural interdependence and

economic development. St. Martin’s Press, LondonLeontief W. et al. (1953) Studies in the structure of the American economy. Oxford University Press, New YorkLlano C (2004a) Economía sectorial y especial: El comercio interregional en el marco input-output. Instituto de Estudios

Fiscales. Colección Investigaciones, No.1, 2004.Llano C (2004b) The interregional trade in the context of a multirregional input-output model for Spain. Estudios de

Economía Aplicada 22: 1–34López-Bazo E, Vayá E, Mora A, Suriñach J (1999) Regional economic dynamics and convergence in the European

union. Annals of Regional Science 33: 343–370Martin R (2001) EMU versus the regions? Regional convergence and divergence in Euroland. Journal of Economic

Geography 1: 51–80Miller R, Blair P (1985) Input-output analysis: Foundation and extensions. Prentice-Hall, Englewood Cliffs, NJMontolio D, Solé-Ollé A (2008) Road investment and regional productivity growth: the effects of vehicle intensity and

congestion. Papers in Regional Science. Published online: 28 June DOI:10.111/j.1435-5957.2008.00167.xMoreno R, López-Bazo E, Vayá E, Artís M (1999) External effects and costs of production. In: Anselín L, Florax R

Advances in spatial econometrics: Methodology, tools and application. Springer, BerlinNazara S, Hewings G (2007) On the spatial distribution of economic impact: Some considerations of alternative

methodologies. 46th Annual Meeting of the Western Regional science Association. Newport Beach. FebruaryOosterhaven J (1984) A family of square and rectangular interregional input-output tables and models. Regional Science

& Urban Economics 4: 565–582Park JY (2006) Estimation of state-by-state trade flows for service industries, presented at North American Meetings of

the Regional Science Association International 53rd Annual Conference, Toronto, Canada, 16–18 NovemberPark JY, Gordon P, Moore JEII, Richardson HW, Wang L (2007) Simulating the state-by-state effects of

terrorist attacks on three major US ports: Applying NIEMO (national interstate economic model). In:

528 J. Pérez et al.

Papers in Regional Science, Volume 88 Number 3 August 2009.

Page 21

Richardson HW, Gordon P, Moore JEII (eds) The economic costs and consequences of terrorism. Edward Elgar,Cheltenham

Puga D (2002) European regional policies in light of recent location theories. Journal of Economic Geography 2:373–406

Pulido A, Fontela E (1993) Análisis input-output: Modelos, datos y aplicaciones. Editiones, Pirámide, SA. MadridQuah D (1996a) Empirics for economic growth and convergence. European Economic Review 40: 1353–1375Quah D (1996b) Regional convergence clusters across Europe. European Economic Review 40: 951–958Rey S (2001) Spatial empirics for economic growth and convergence. Geographical Analysis 33: 195–214Rey S, Montouri B (1999) US regional income convergence: A spatial econometric perspective. Regional Studies 33:

143–156Sosvilla-Rivero S (2005) EU structural funds and Spain’s Objective 1 regions: An analysis based on the Hermin model.

Working paper 2005-24. FEDEAVan der Linden JA, Oosterhaven J (1995) European community intercountry input-output relations: Construction

method and main results for 1965–1985. Economic System Research 7: 249–269

529An interregional impact analysis of the EU structural funds in Spain (1995–1999)

Papers in Regional Science, Volume 88 Number 3 August 2009.

Page 22

Análisis de impacto interregional de los fondos estructuralesde la UE en España (1995–1999)

Julián Pérez, Milagros Dones and Carlos Llano

Resumen. Este artículo utiliza un modelo de entrada/salida para estimar el impacto económicode los fondos estructurales de la UE recibidos por las regiones españolas durante el periodo1995–1999. Intentamos aclarar los efectos interregionales producidos por los fondos entérminos de salidas, valor agregado y empleo, no solo en las regiones dónde fueron asignadosoriginalmente sino también en el resto de las regiones. Este análisis ofrece información adicio-nal aparte de la obtenida mediante modelos macroeconómicos que no toman en cuenta efectosde derrame (spillovers) interregionales. Los resultados son relevantes con respecto a la discusiónsobre la efectividad de la política de cohesión de la UE, y el reparto de efectos de salidacapturados por las regiones más ricas a través de sus vínculos intersectoriales.

JEL classification: C67, R11, R15, R58

Palabras clave: Modelo interregional de entrada/salida, comercio interregional, fondos decohesión y estructurales de la UE, matrices de origen-destino, convergencia regional

doi:10.1111/j.1435-5957.2009.00212.x

© 2009 the author(s). Journal compilation © 2009 RSAI. Published by Blackwell Publishing, 9600 Garsington Road,Oxford OX4 2DQ, UK and 350 Main Street, Malden MA 02148, USA.

Papers in Regional Science, Volume 88 Number 3 August 2009.