Revisiting the EU Cohesion Policy 2007-2013: Did it bring convergence?

(1)

Revisiting the EU Cohesion Policy 2007-2013: Did it bring convergence?

Master Thesis

Jelmer Otter │ S2958155 │ TAE37V

University of Groningen │ Corvinus University of Budapest 03-07-2020

Abstract

The 2007-2013 EU Cohesion Policy put forward a Convergence objective that aimed to stimulate growth and employment in the least developed regions of the EU. Throughout this programming period, the European Commission provided in excess of €280 Billion in Convergence funding to various eligible EU regions. As a considerable amount of time has passed since the last funds were transferred, it is time to research whether these 2007-2013 convergence funds actually brought convergence to the recipient regions. Earlier studies provided evidence for convergence funds positively affecting GDP per capita growth in NUTS-1, 2 and 3 regions during earlier programming periods. This paper however only finds a positive effect of 2007-2013 convergence funds on the convergence of GDP per capita levels of

NUTS-1 and NUTS-2 regions towards the EU-28 average. Moreover, this effect was only found three

(2)

2 Table of Contents

I. Introduction 3

II. Institutional Background 4-7

- Eligibility 5-6

- Distribution of Funds 6-7

III. Literature Review 7-10

IV. Methodology of Research 10-18

- Method 10-11

- Convergence 11

- Convergence rates of eligible NUTS-regions 12

(3)

3

I. Introduction

With the creation of the Cohesion Policy, the EU has tried to improve the economic welfare of EU member states and regions within, as well as to lessen the economic divergence between regions. It does so by putting forward multiple objectives, which subsequently provide funding to promote the development of backward regions. The EU Cohesion Policy of 2007-2013 put forward three objectives, of which the ‘Convergence’ objective became the main part of the Cohesion Policy. This objective provided over €280 Billion in funding, which was over 80% of the total amount of Cohesion Policy funding available during this round.

So far, the existing literature has found certain positive effects for funding on economic growth of recipient regions, albeit the main focus remains to be earlier Cohesion Policy rounds. The amount of research on the 2007-2013 round remains relatively scarce to date, most likely due to the fact that when funds are invested, it takes time before both data becomes available, as well as before investments start to affect economic measures such as GDP per capita levels. In the earlier programming periods, the main objectives of the EU Cohesion Policy were relatively similar. The programming period of 2007-2013 however seems to have created a break in the trendline, by changing the name of the main objective (from a simple ‘Objective 1’ to the name ‘Convergence’), as well as the aim of the EU Cohesion Policy overall. The aim seems to have shifted from equity to competitiveness, which might be at odds with the name ‘Cohesion Policy’. Additionally, this round is the first one in which all EU regions were able to obtain EU Cohesion Policy funds; when regions were ineligible for ‘Convergence’ objective funds, they were able to apply for funds from another objective instead. Furthermore, the enlargement of the EU in 2004, 2007 and 2013 led to a shift of where EU Cohesion Policy funding flowed towards; during the 2007-2013 round, the majority of these funds went to regions which were not part of the EU before 2004.

Lastly, as the European Commission has changed the name of the main objective from ‘Objective 1’ to ‘Convergence’, this would imply that the funds allocated through this ‘Convergence’ objective aim at convergence in itself. While earlier research has mainly focused on the effect of funds on the economic growth of regions in terms of GDP per capita growth, it might be interesting to research whether the convergence funds actually bring convergence, rather than ‘just’ economic growth.

(4)

4

II. Institutional Background

Back in 1975, the European Economic Community (EEC) established a regional fund called the European Regional Development Fund (ERDF), which had the goal of investing in underdeveloped regions of the EEC (European Commission, 2020a and 2020b). This ERDF had an annually assigned budget, and would fund development projects in poor EEC regions till 1989 (Mohl, 2016). After the implementation of the Single European Act in 1986, the EEC put forward a Cohesion Policy, of which the aim was to ‘‘reduce disparities between the various regions and the backwardness of the least-favoured regions’’ (European Commission, 2020a). The already established ERDF became one of the main funds of this new Cohesion Policy. Since the Lisbon Treaty came into force in 2009, the Cohesion Policy additionally focuses on ‘‘economic, social and territorial cohesion’’ within the EU (European Commission, 2020a). The Cohesion Policy of the EEC (and later the EU) has been part of the multi-annual EEC/EU programming periods since 1989 (Mohl, 2016), and provides funding to numerous projects inside the European Union. Besides the already mentioned ERDF, there are two other main funds that make up the totality of the Cohesion Policy: The Cohesion Fund (CF) and the European Social Fund (ESF) (European Commission, 2020a). By 2020, five programming periods have (almost) passed; 1989-1993, 1994-1999, 2000-2006, 2007-2013 and 2014-2020, of which the latter is soon coming to an end and will be replaced by the 2021-2027 budgetary round. These six periods came with different objectives, different amounts of available funding, as well as differences in eligibility of regions that are able to apply for specific funds.

In the round of 1994-1999, so-called Objectives were put forward that defined a specific target, which were financed by EU funds. In that round, six objectives were introduced, with the main ones being Objective 1 (promote development of regions that lag behind), Objective 2 (convert regions affected by industrial decline) and Objective 3 (tackle long-term unemployment and labour market mismatches), which together made up 88.1% of the EU Cohesion Policy funding (European Commission, 2008 and European Commission, in Mohl, 2016). In the programming period of 2000-2006, the objectives of the previous round were revised and shortened to four in total, that made use of six instruments (European Commission, 2007). While Objective 1 remained similar, Objective 2 now supported social and economic transformation in regions that faced structural (economic) problems, while Objective 3 supported modernisation in various regions in regards to e.g. educational systems (European Commission, 2008).

(5)

5

Table 1: Amount of funds of the three EU Cohesion Policy objectives,

including subdivision and eligibility criteria that need to be met to be able to apply for funds. In % In % € in Billion € in Billion Eligibility criteria Convergence 81.5 283.139 Of which:

- Regions with a per capita GDP less than 75% of EU average

57.5 199.761 NUTS-2 regions with a GDP per capita of <75% of EU-25 average

- Phasing-out 4 13.896 Transitional support for regions who would comply if the convergence threshold was set at <75% of EU-15 average - Cohesion Fund 20 69.482 EU member states with a GNI of <90% of EU-25 average + transitional support for regions which would comply if the threshold was set at <90% of EU-15 average. Regional Competitiveness and

Employment

16 55.586

Of which:

- Regional competitiveness and employment

12.6 43.774 All NUTS-1 and NUTS-2 regions not covered by convergence funds

- Phasing-in 3.4 11.812 Transitional support for regions eligible for convergence funds, but with a GDP of >75% of EU-15 average European Territorial cooperation 2.5 8.685

Of which:

- Cross border cooperation 1.8 6.253 All NUTS-3 regions that have a land-based or maritime* internal or external border (*max distance of 150km) - Transnational cooperation 0.5 1.737 All regions

- Interregional/network cooperation 0.1 0.347 All regions

- ‘PEACE’ programme 0.1 0.347 Northern-Ireland only

Total 100 100 347.410 100

Source: European Commission, 2007. Table by author.

Eligibility

The government of every EU member state that meets certain eligibility criteria is able to apply for Cohesion funds. These criteria (can) change with every programming period, as well as the specific regions that meet these criteria and therefore are able to apply for funds. The eligibility criteria for the 2007-2013 programming period can be found in Table 1.

(6)

6 the EU-25 (see Table 1). Nevertheless, even when including the regions that were able to receive ‘transitional support’ under the convergence objective, there was a significant geographic shift down- and eastwards in regards to which countries were eligible to receive convergence funds in comparison to previous rounds.

Another important observation is that the eligibility criteria to receive any EU Cohesion Policy funds are set at different levels of regions for different types of funds. Whereas for the Convergence objective (without the Cohesion fund) eligibility is set at the NUTS 2-level, the Cohesion funds are set at the NUTS-1 level, and the eligibility for Cross-border cooperation funding is set at the NUTS 3-level. These levels refer to the Nomenclature des Unités Territoriales Statistiques (NUTS), which is a classification system developed by Eurostat to divide the territory of (prospective) EU member states for statistical purposes, and was given a legal status in 2003 (Eurostat, 2020). This system consists of 4 levels: NUTS-1, 2, 3 and local administrative units. The current system (NUTS 2016-2021, EU-28) sees 104 regions at the NUTS-1 level, 283 regions at the NUTS-2 level and 1345 at the NUTS-3 level; NUTS-3 regions are subdivisions of the NUTS-2 level regions, which are subdivisions of the NUTS-1 level regions (Eurostat, 2020). The NUTS levels are divided based upon population, as can be seen in Table 2. This subdivision of regions lead to various outcomes, one of them being that countries such as Cyprus, Estonia, Latvia, Luxembourg and Malta are one complete region at both the NUTS 1 and NUTS 2 level. In comparison, Germany has 16 and 39 regions for those levels respectively.

Table 2: NUTS-levels and their respective cut-off points in terms of population size.

Level Min. Pop. Size Max. Pop. Size

NUTS-1 3.000.000 7.000.000

NUTS-2 800.000 3.000.000

NUTS-3 150.000 800.000

Source: Official Journal of the European Union, 2003

Distribution of funds

(7)

7 are then transferred to the respective NUTS-1/2/3 regions; this means that a transfer of funds from the European Commission to an EU member state does not necessarily take place in the same year as when these funds are invested in particular NUTS-1/2/3 regions.

III. Literature Review

In regards to whether the EU Cohesion Policy is actually able to realise its pre-set Objectives, various studies have tried to answer whether the EU Cohesion Policy funds had any (positive) effects, and if so in what form. Overall, studies that focused on the EU Cohesion Policy funds from earlier rounds (with Objective 1 in particular), seem to have found positive and significant effects for these funds on average annual GDP per capita growth of recipient NUTS-1, 2 and 3 regions. While this paper will focus on convergence rates instead of GDP per capita growth, higher growth (or a lower than EU-28 average decline) is needed in order for convergence to take place. On the other hand, more recent studies show that the 2007-2013 EU Cohesion Policy shifted its aim towards efficiency over equity. This shift came with serious implications for the distribution of funds, as the fund eligibility criteria favoured more developed regions over backward regions, which could impact the effect of convergence funds on convergence rates. Research by Dall’erba and Le Gallo (2008) claims that the funds of the EU Cohesion Policy in the 1989-1993 and 1994-1999 rounds positively affected economic growth in the regions that were targeted by these policies, although spill over effects were small, and the effect was not similar across all regions. Similar research by Cappelen, Castellacci, Fagerberg and Verspagen (2003), which also focused on pre-2000 EU Cohesion Policy funds, stated that the funds did have a positive and significant effect on GDP growth of EU regions, with a side note that the effects of these funds were larger in more developed regions amongst the targeted regions. Boldrin and Canova (2001) also researched the effects of the EU’s Cohesion Policy pre-2000, but they stated that these regional policies pursued by the EU were merely redistributional in nature, as they were unable to find evidence for convergence taking placing due to the distribution of Cohesion Policy funds. Nevertheless, they do not rule out that EU Cohesion Policy funds might lead to higher economic growth in backward regions. An interesting side note here is that they acknowledge the existence of the six different Objectives which were part of the 1994-1999 programming period, but forewent the fact that these Objectives did not necessarily have a clear and similar link with economic growth, nor convergence.

(8)

8 Moreover, they checked for robustness of these results by including testing on the NUTS-3 level as well, which yielded similar positive results.

Fiaschi, Lavezzi and Parenti (2017) went a little further, and researched the effectiveness of the overall EU Cohesion Policy between 1991-2008 (which besides Objective 1 also included other objectives, and therefore also other funds than the previously mentioned ERDF, CF and ESF). They found similar overall (positive) effects of EU Cohesion Policy funds on GDP per capita growth, although they do state that the effect of some of these funds on GDP per capita growth is ambiguous at first, due to an unclear link between these particular funds and GDP growth. In regards to the 2007-2013 programming period, the objectives were modified and therefore diverged from the objectives used in earlier programming periods. These changes were implemented in order to simplify the policies overall, although the aim of these policies seem to have changed as well. Overall convergence between backward regions and the EU GDP per capita average, as well as a reduction in disparity-levels used to be the main goal of the EU Cohesion Policy (hence the name Cohesion Policy), but the objectives put forward in the 2007-2013 programming period seem to be a little contradictory. The European Commission stated that for the ‘Convergence’ objective, the aim is to ‘‘stimulate growth and employment in the least developed regions … and targets the least well-developed Member States and regions’’ (European Commission, 2007), which would indicate a policy of stimulating convergence of the backward regions of the EU towards the EU-25 average in regards to GDP per capita levels. On the other side, for the ‘Regional competitiveness’ objective they pursue coherent strategies that apply to whole regions, rather than ‘‘micro-zoning at borough or village level’’ (European Commission, 2007), which would indicate a policy of increasing the competitiveness of already relatively wealthy areas (e.g. regions in The Netherlands and Sweden), which would make the convergence of backward EU regions towards the EU average more challenging.

According to Bachtler and Wishlade (2011), the main aim of the Cohesion Policy shifted from increasing cohesion (equity between regions) to increased economic growth (competitiveness) within all European Union areas since the programming period of 2007-2013. This programming period was also the first to make all EU-regions eligible to apply for particular EU Cohesion Policy funding, something that is also proposed for the new 2021-2027 period (Bachtler, Mendez and Wishlade, 2019). This leaves one to wonder whether the so-called Cohesion Policy of the programming period 2007-2013 still fosters convergence of backwards EU regions towards the EU-25 average, or whether the drive for increased competitiveness of more wealthy EU regions causes the convergence of backward EU regions to cease.

(9)

9

time passes since the start of the programming period. They tested for this effect up till six years

after the start of the period, of which the sixth year showed the strongest significance and the largest impact on average annual growth of GDP per capita. Such would indicate that research on a programming period should include successive years in order to test for the effect of EU Cohesion Policy funds invested in time = t, on GDP per capita growth in time = t + *. As just over 4 years have passed since the last allocated funds of the programming period 2007-2013 were callable by EU member state governments, this moment in time would be the start to fully assess the period 2007-2013.

In addition the 2007-2013 objectives being different in comparison to the objectives stated in earlier programming periods, as well as the fact that enough time has passed for it to be suitable to research the 2007-2013 programming period, there is another reason why studying this programming period would be of interest. Between 2004 and 2013, a total of 13 countries have entered the EU, which created a significant shift in which regions obtain EU Cohesion Policy funds, and how much these regions received. Funds such as the Cohesion Fund were solely accessible to new member states (NMS) in the programming period 2007-2013, with the exception of Portugal, and Spain which was able to apply for transitional support (European Commission, 2007). More recent research into the effects of EU funds in the NMS showed that the criteria for fund eligibility for the programming period 2007-2013 was benefitting more developed NMS regions over backward NMS regions, although domestic quality of government can positively influence better dissipation of funds throughout the whole country (Medve-Bálint, 2018). In line with this is a case study that looked into the distribution of EU Cohesion Policy funds in Poland and Romania, which showed that weaker institutions in the latter country lead to greater concentration of funds within the more developed regions of Romania (Medve-Bálint and Šćepanović, 2019).

(10)

10

1_{The EU-28 GDP per capita average was used rather than the EU-27 or EU-25 average, as by the end of the 2007-2013 programming period}

28 countries made up the entirety of the European Union (Bulgaria and Romania joined in 2007, and Croatia joined in 2013. All three countries were eligible for EU Cohesion Policy funds in the 2007-2013 programming period).

Hypothesis 1: The EU Cohesion Policy funds of the Convergence objective positively affect convergence of NUTS-1 level regions towards the EU-281_{GDP per capita average, with the}

effect becoming larger after more time has passed since funds were transferred.

Hypothesis 2: The EU Cohesion Policy funds of the Convergence objective positively affect convergence of NUTS-2 level regions towards the EU-28 GDP per capita average, with the effect becoming larger after more time has passed since funds were transferred.

Hypothesis 3: The EU Cohesion Policy funds of the Convergence objective do not have a significant effect on the convergence of NUTS-3 level regions towards the EU-28 GDP per capita average

IV. Methodology of Research

To test the aforementioned hypotheses, I will compare the convergence level of NUTS-1, 2 and 3 regions towards the EU-28 average in terms of real GDP per capita with the amount of received convergence funds. This research will look at the regions which were eligible for funding under the convergence objective (either through the ERDF, ESF, CF, or a mixture of these three funds), and also received funding in any given year between 2007 and 2016. Method

A regression analysis will be performed to research whether EU Cohesion Policy Convergence funds (EUCN*) affect convergence levels of NUTS-1, 2 and 3 regions (EUPN*), based upon the programming period of 2007-2013. This regression analysis includes fixed effects, which account for the effect coming from NUTS-regions themselves (i) and from the different years (t) (2007-2018 for Hypothesis 1 and 2, 2013-2018 for Hypothesis 3). The reason to add NUTS-region fixed effects stems from the inability to observe certain ‘NUTS-regional traits’, such as local culture and geography, which could be seen as relatively stable over the time span discussed, yet can still affect the convergence rates of NUTS-regions. As the regression analysis would also include other independent variables besides the EU Cohesion Policy funds, the regression equation would look as follows:

ỹ

it

= β'P

it

+ β'x

̃

it + ε̃it

The ỹ stands for the convergence of NUTS-1/2/3 regions towards the EU-28 GDP per capita average, while the P denotes the transfers of EU Cohesion Policy funds towards NUTS-1/2/3 regions per capita, per year. The

x̃

portrays the other independent variables (e.g. GDP, TE, IMR, BD, WAP, MEN, AGR).

(11)

11

ỹ

it+*

= β'P

it

+ β'GDP

it + β'x

̃

it+* + ε̃it+*

The ỹ still stands for the convergence of NUTS-1/2/3 regions towards the EU-28 GDP per capita average, while the asterisk denotes the specific lead used in each model. The P denotes the transfers of EU Cohesion Policy funds towards NUTS-1/2/3 regions per capita, per year, while GDP portrays the GDP level of NUTS-1/2/3 regions in time = t + 0. The

x̃

portrays the other independent variables (e.g. GDP, TE, IMR, BD, WAP, MEN, AGR), while here the asterisks also denotes the same lead as used for ỹ.

Convergence

Convergence of a NUTS-region towards the EU-28 GDP per capita average will take place when either the level of GDP of a given NUTS-region rises more quickly than the EU-28 average, or when the decline is smaller than that of the EU-28 average. Therefore, in order to obtain a valid measure of convergence, I first looked at the GDP per capita level per region (NUTS-1, 2 and 3, in Euros). These data were derived from the Eurostat database of the European Commission. NUTS-1 and NUTS-2 level data of GDP per capita per year for the years 2006-2018 was readily available, while data for the NUTS-3 level was created by dividing GDP per NUTS-3 region-data by the average annual population per NUTS-3 region data. In order to obtain real GDP per capita data for the NUTS-1, 2 and 3 regions, the two datasets are adjusted for inflation by using a deflator figure. A limitation of using this deflator is that this figure is from the country level, which means that GDP per capita data is adjusted for country-wide inflation rates. While realising that inflation rates can be somewhat differing within countries, adjusting the regional GDP per capita data for country-level inflation still portrays a more ‘real’ picture of GDP per capita per NUTS-region due to taking into account country-wide effects.

The two types of data were then compared to the EU-28 GDP per capita average (in Euros) for the given years. When the GDP per capita of any region is taken as a percentage of the EU-28 average, it provides an overview of how much regions have ‘converged’ with the EU-28 average: the closer the number is to 100, the more converged a region is with the EU-28 average. While many regions have a figure greater than 100% due to their GDP per capita level being (much) larger than that of the EU-28 average, this is (mostly) not the case for recipients of Objective 1 funds: the whole aim of this objective is, as the title of the objective shows, ‘Convergence’.

(12)

12 Convergence rates of eligible NUTS-regions

Looking at the NUTS-1 regions that were eligible for Objective 1 funding through a combination of the ERDF/ESF and CF, all regions had a GDP per capita level that was less than that of the EU-28 average (= below 100%). This would make it possible to research convergence rates as described earlier, as the convergence figure would be made up of a GDP per capita level between 0 and 100% of the EU-28 average for a given year, while subtracting the same figure derived from the previous year.

The only exception to this idea is the region EL3/Attiki (Greece), which saw a level of 106%. This discrepancy of a NUTS-1 region being eligible to receive Objective 1 funding while already being ‘fully converged’ most likely stems from the fact that budgets for the cohesion policy objectives are finalised before the programming periods start. Historically, Attiki has received Objective 1 during earlier rounds (e.g. 1989-1993, 1994-1999 and 2000-2006 (Becker, Egger, von Ehrlich and Fenge, 2008)), but for the 2007-2013 round this region was put under the ‘phasing out’ category, which is reserved for regions that are eligible for the convergence objective transitional support system. Due to this region being hit hard during the financial crisis (European Commission, 2020b), their GDP per capita was below 90% of the EU-28 average’s GDP per capita level by 2013, which made them eligible to receive funding again in the 2014-2020 round (European Commission, 2015).

Of the 81 NUTS-2 regions that were eligible for ERDF/ESF and CF funding through Objective 1, all regions had a GDP per capita level of less than 100% of the EU-28 average, except the EL30/Attiki (Greece) region, which fully corresponds with the previously mentioned NUTS-3 region. 42 NUTS-2 regions were eligible for CF funding but not for ERDF/ESF funding, or vice versa. Of these regions, six had a GDP per capita level above the EU-28 average, with UKM6/Highlands and Islands (United Kingdom) the largest ‘outlier’ with a GDP per capita level reaching 124% of the EU-28 average. This UKM6 region was eligible for ERDF/ESF funding but not for CF funding, of which the latter is allocated on a country bases (the United Kingdom was not a recipient of CF funding during the 2007-2013 round). The other five regions were only eligible to receive CF funding, but not for ERDF/ESF funding through Objective 1. The picture looks much different in regards to the NUTS-3 regions. As ERDF/ESF funding is allocated at NUTS-2 level, and CF funding on a country basis, the funds of Objective 1 can end up in regions which are already converged with the EU-28 average in terms of GDP per capita. Of the 521 NUTS-3 regions which are eligible for either the ERDF/ESF, the CF, or both funds, 40 regions had a GDP per capita level of >100% of the EU-28 average. Again, the side note here is that only four NUTS-3 regions which were above the 100%-level were eligible for both funds: EL301, EL303, EL305 and EL306, which are all part of the aforementioned NUTS1- EL3 and NUTS-2 EL30 regions.

Convergence funding

(13)

13 EU funds on the NUTS-1 and NUTS-2 level: ERDF, CF, EAFRD/EAGGF and the ESF. The EAFRD/EAGGF are agricultural funds which do not fall under the EU Cohesion Policy; therefore this type of data are not used. The other three funds are part of the EU Cohesion Policy, and therefore useful to answer Hypothesis 1 and 2.

The second source of data is an expenditure study carried out by the European Commission in 2015, which looks at the ‘expenditures’ of EU Cohesion Policy funds of the 2007-2013 programming period at the NUTS-3 level (European Commission, 2015). Expenditures are recorded as transfers of EU Cohesion Policy funds towards NUTS-3 regions, as reported by these regions themselves. This study contains data on annual ‘expenditures’ for the three EU Cohesion Policy funds per NUTS-3 region for the years 2013 and 2014. The latter can be seen as a limitation of this dataset, as only two years of data is available. However, by using said data, it is still possible to research whether for the given years the EU funds had any significant effect on convergence levels in NUTS-3 regions.

The two types of data used for Hypothesis 1/2 and Hypothesis 3 provided two types of ‘measures’ for the amount of EU Cohesion Policy funds received per region, per year (in Euros). For the first dataset this comprises of ‘Annual EU Payments’ and ‘Modelled Annual Expenditures’. The first type provides a measure of EU funds transferred to NUTS-1/2 regions, of which these payments are recorded in line with the moment that funds of the EU Cohesion Policy are transferred towards EU member states. Such recordings can however negatively affect analytic work, as ‘real expenditures’ from national governments to NUTS-1/2 regions does not necessarily follow this path: the funds provided by the EC to the EU member states might not be handed out to the NUTS1/2 regions immediately by said member states, but might rather take a while to be paid out. This is also underlined by the European Commission (2019), which therefore also provided a second measure. This second measure provides a modelled version of annual EU funds received by NUTS-1/2 regions, which should provide a better estimate of real expenditure of EU funds per NUTS-1/2 region per year. For this reason, the data derived from this second measure-type is used to answer Hypothesis 1 and 2. For the second dataset similar measures are given; a figure for allocated EU funds per year, and a figure for EU funds expenditures per year. As the latter gives a better overview of the amount of EU funds received per NUTS-3 region per year, this data is used to answer Hypothesis 3.

(14)

14 Other variables

Apart from EU payments and the fixed effects, other independent variables (partially) explain the movement of the dependent variables EUCN1, EUCN2 and EUCN3. Due to the nature of the research, it is relatively difficult to obtain valid independent variables which provide data on the NUTS-1, NUTS-2 and/or NUTS-3 level. Most data which is reported on these levels is created by either the European Commission itself, or by Eurostat (which is a branch of the European Commission), which also leads them to be the primary sources for the independent variables as one can observe in Table 3.

Table 3: Dependent and independent variables

Abbreviation Name Explanation Source Used for

EUCN* EU-28

Convergence

The convergence rate of a NUTS-regions’ real GDP per capita level towards the

EU-28 average in absolute % Eurostat, calculations by author Hypothesis 1, 2 and 3 EUPN* EU Cohesion Fund Payments

The real amount of funding received from the EU Cohesion Fund (Objective 1) per

NUTS-region, per capita per year

European Commission and Eurostat, calculations by author Only displayed in Table 4 EUPN*log EU Cohesion Fund Payments, logged

The logged amount of funding received from the EU Cohesion Fund (Objective 1)

per NUTS-region, per capita per year

European Commission and Eurostat, calculations by

author

Hypothesis 1, 2 and 3

GDP GDP per capita, The level of GDP, adjusted for inflation, per NUTS-1/2/3 region, per capita, per

year Eurostat, calculations by author Hypothesis 1, 2 and 3 TE Tertiary Education

The share of the population aged 25-64 with tertiary education, per NUTS-1/2

region

Eurostat Hypothesis 1 and 2

IMR Infant Mortality Rate

The share of deaths of children <1 year per 1000 live births, per NUTS-1/2 region

Eurostat Hypothesis 1 and 2

BD Business

Demography

The amount of active enterprises per 1000 inhabitants, per NUTS-1/2/3 region

Eurostat, calculations by author

WAP Working Age

Population

The amount of people of working age in % of total population, per NUTS-1/2/3 region

MEN Men – Working age population

The share of men in terms of % of the working age population

Eurostat, calculations by author Hypothesis 1, 2 and 3 AGR Employment in Agriculture

The share of people employed in agriculture in % of the total amount of

people employed

Hypothesis 3

While Table 3 provides a concise overview of the used variables, the following section will provide a more thorough explanation on the inclusion (and exclusion) of certain independent variables in the econometric models.

GDP

(15)

15 negative effect of the level of GDP on GDP growth, and therefore also on convergence rates; the higher the level of GDP, the lower the convergence rate.

TE

The TE variable shows the %-share of the population (aged 25-64) with tertiary education, per NUTS 1 or 2 region, per year. Research has shown that differences in the cognitive skills base can (partially) explain why economic growth rates differ across countries (Hanushek, 2016), although the simple addition of more years of schooling does not positively affect economic growth if the cognitive skills base does not increase. While tertiary education levels might not fully reflect the knowledge capital-level of a region, this variable is included to check whether it influences convergence levels of NUTS-1 and 2 regions.

IMR

The IMR variable shows the share of deaths of children below the age of 1 per 1000 live births, per NUTS 1 or 2 region, per year. As health (Bhargava, Jamison, Lau and Murray, 2001 and Weil, 2013) and healthcare expenditure (Ndedi, Metha and Nisabwe, 2017, and Raghupathi and Raghupathi, 2020) can positively affect economic growth and therefore convergence rates, it would be important to include a variable which tells something about the level of health per NUTS region. However, due to limited availability of health related data per NUTS-1/2/3 region for the years 2007-2018, I decided to use infant mortality rates as a measure of ‘health’. While this variable may be a relatively ‘flawed’ measure of ‘health’ as infant mortality rates are low across the EU, the statistics do provide variation that is wide enough for possible effects to be picked up by the analysis.

BD

The BD variable shows the amount of active enterprises per 1000 inhabitants, per NUTS-1, 2 or 3 region, per year. Entrepreneurship has been observed as an important factor for economic growth; a 2020 study that researched entrepreneurship in 22 European countries, found that entrepreneurship positively affects economic growth for all tested countries (Stoica, Roman and Rusu, 2020). On the other side, the type of active businesses might be more important than the overall number (Isenberg and Fabre, 2014). Howbeit, as data on types of businesses active in NUTS-1/2/3 regions is not available, a measure of the total amount of active businesses is used for the regression analyses.

WAP

The WAP variable shows the %-share of people of working age (25-64) of the total population, per NUTS-1, 2 or 3 region, per year. An economist working for the European Commissions’ Directorate-General for Employment, Social Affairs and Inclusion stated in 2015 that a declining working age population could negatively affect the economic growth prospects of the EU (Peschner, 2015). This would imply that when the %-share of people of working age is

high, more citizens would have the ability to participate in the labour force and therefore higher

(16)

16

MEN

The MEN variable shows the share of males in terms of %-share of the working age population, per NUTS-1, 2 or 3 region, per year. This variable is added to the regression analysis in order to account for the effects of the division of the working age population by gender on convergence rates.

AGR

The AGR variable shows the share of people employed in agriculture in % of the total amount of people employed. As the TE variable discussed above is not available for the NUTS-3 level, another variable is used to account for the effect of educational attainment on convergence rates in NUTS-3 regions. As the educational attainment of employees in the agricultural sector can be considered relatively low (Arcury, Estrada and Quandt, 2010 and Gertel and Sippel, 2014), a higher share of people employed in agriculture could imply a lower level of people with tertiary education, and therefore could negatively affect convergence rates.

Unfortunately, due to the relatively scarce amount of data available on the NUTS-1, 2 and 3 level, there are also variables which I would have preferred to add to the regression analysis, but were unable to due to a lack of data. An example is a variable that shows the strength of institutions at the NUTS-1, 2 and 3-level, as research by Medve-Bálint and Šćepanović (2019) found that weaker institutions can lead to a higher concentration of EU funds in particular regions, which could distort convergence rates. Another variable that I was unable to add was one that shows the business environment of the NUTS-1, 2 and 3- regions, which could be important as a more supportive business environment can positively influence economic growth in EU regions (Głodowska, 2017)

N-number

(17)

2_{The EUPN* logged variables are the ones used in the regression analysis. The EUPN* variables are for display only, and merely portray the}

variation in sizes of real amounts of EU Cohesion Policy Convergence funds per capita per year, across NUTS-1, 2 and 3 regions. 17

A side note here is that naturally, the number of observations will go down after including the second lead. This is due to the fact that most data is available up till 2018; therefore, the effect of EU convergence funds recorded in 2016 on the convergence rates of NUTS regions in 2019 is not able to be tested, which means that the EU convergence funds data for 2016 is dropped when testing the models that include a third lead. This continues for the fourth lead (that drops EU convergence funds data for 2015 and 2016), and so forth.

Descriptive statistics

In Table 4, the descriptive statistics of the used2_{variables are displayed per NUTS-region. What}

can be observed is that the variation found for NUTS-1 and 2 regions seem to be relatively similar, while the variation for NUTS-3 regions seems to be much larger in comparison to the other two. This most likely stems from the fact that NUTS-3 data is more localised due to the regions being (much) smaller, and therefore will show higher variation.

Table 4: Descriptive statistics of used variables

Data on NUTS-1 regions which received EU Cohesion Policy Convergence funds at any moment between 2007-2016

Variable Mean Std. Dev. Min Max

EUCN1 0.104 1.803 -11.097 5.015 EUPN1 EUPN1 logged 176.882 2.031 177.842 0.524 0.304 0.115 1170.39 3.069 BD 59.726 21.594 24.920 115.247 GDP 11583.75 5354.997 3689.337 30268.2 IMR 5.524 2.497 1.1 13.6 MEN 49.795 0.850 47.991 51.567 TE 21.509 7.296 7.3 38.7 WAP 68.296 1.811 63.439 72.230

(18)

18

Table 4, continued: Descriptive statistics of used variables

EUCN3 0.184 2.150 -15.705 23.430 EUPN3 765.851 671.975 0.027 4671.332 EUPN3 logged AGR 2.663 12.870 0.543 12.997 0.012 0 3.670 63.787 BD GDP 56.550 14271.28 25.569 7687.83 12.871 2680.889 209.697 44783.28 MEN WAP 50.494 66.335 1.082 2.762 46.312 55.490 53.212 74.245 V. Results

The regression analysis’ results for Hypothesis 1 can be found in Table 5. For the year in which the Convergence funds were transferred to the NUTS-1 regions (in time = t, Model 1), these funds do not show to have a statistically significant effect on convergence rates.

When a lead is added to the equation (time = t+1, Model 2), the results become statistically significant, albeit them being negative. This would indicate that Convergence funds would negatively affect Convergence rates one year after those funds are transferred. A possible explanation for this phenomenon could be the selection of particular recipient regions and the allocation of funds upfront, in which NUTS-1 regions which see lower economic growth than the EU-28 average (and therefore diverge from the EU-28 average, and hence see the gap between them and the average EU-level become larger) are allocated larger amounts of EU Convergence funding. From a general perspective this would make sense, as the EU convergence funds should lead to convergence, and therefore should target those regions which ‘struggle’ the most with converging. The European Commission does target areas with lower GDP per capita than the EU-28 average (see Table 1), although this is not exactly the same as areas with relatively lower or even negative convergence rates. What is striking, is that when a second lead is added in order to research the effect of EU Payments made in time = t on convergence in time = t+2, the effect stays negative and statistically significant. Moreover, the effect itself increased between the first and second lead as well.

(19)

19

Table 5: Regression analysis results for Hypothesis 1

Dependent Variable EUCN1 (Convergence of NUTS-1 regions towards the EU-28 GDP per capita average)

Model 1 (Base Year) Model 2 (Lead 1) Model 3 (Lead 2) Model 4 (Lead 3) Model 5 (Lead 4) Model 6 (Lead 5) Model 7 (Lead 6) EUPN1 logged -0.436 -0.594*** -1.132*** -1.130 1.338** 2.457*** -0.523 (0.323) (0.224) (0.379) (0.697) (0.534) (0.571) (0.558) GDP 0.000656 -0.00138*** -0.00124*** -0.000568*** -0.000621*** 0.000210 9.38e-07 (0.000399) (0.000280) (0.000251) (0.000214) (0.000198) (0.000215) (0.000198) TE -0.137* 0.00586 0.111* 0.257** 0.0196 0.278*** 0.245** (0.0738) (0.0504) (0.0667) (0.115) (0.101) (0.1000) (0.101) IMR 0.265* 0.0389 0.109 0.0106 0.161 0.160 -0.0815 (0.156) (0.119) (0.117) (0.124) (0.122) (0.121) (0.115) BD 0.00149 0.0974*** 0.0285 -0.0327 -0.0172 -0.103*** -0.0183 (0.0398) (0.0301) (0.0320) (0.0348) (0.0271) (0.0351) (0.0389) WAP -0.583** -0.566*** -0.448*** -0.348* -0.367** 0.156 -0.224 (0.223) (0.162) (0.168) (0.187) (0.184) (0.224) (0.220) MEN -2.508** 0.446 1.450 -0.242 -1.287* -1.777** -1.393* (1.092) (0.864) (0.967) (0.882) (0.693) (0.726) (0.803) Constant 159.8*** 25.96 -31.03 40.38 92.39** 70.55 82.04 (57.33) (47.24) (53.01) (51.31) (43.11) (46.01) (51.55)

Region x Year Fixed Effects

Yes Yes Yes Yes Yes Yes Yes

No. of Observations 95 115 116 113 110 96 82

No. of N1 Regions 18 20 20 20 20 20 20

Prob > F 0.0028 0.0000 0.0000 0.0000 0.0000 0.0000 0.0093

R-squared 0.259 0.371 0.365 0.303 0.432 0.477 0.277

Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1

(20)

20 since the transfer of these EU Cohesion Policy funds. A side note here is that convergence in absolute percentages as tested here, is not completely comparable to GDP per capita growth as used by e.g. Becker, Egger and Von Ehrlich (2010).

Model 1 (Base Year) Model 2 (Lead 1) Model 3 (Lead 2) Model 4 (Lead 3) Model 5 (Lead 4) Model 6 (Lead 5) Model 7 (Lead 6) EUPN2 logged -0.0730 -0.398*** -0.105 0.557* 1.225*** 1.636*** 0.00915 (0.186) (0.143) (0.268) (0.326) (0.357) (0.434) (0.485) GDP 0.000647*** -0.000578*** -0.000410*** -0.000589*** -0.000439*** -0.000251* -0.000528*** (0.000121) (9.82e-05) (0.000102) (0.000101) (0.000112) (0.000131) (0.000150) TE -0.0623 -0.0305 0.00853 -0.0309 -0.0520 -0.0656 0.0362 (0.0398) (0.0346) (0.0379) (0.0396) (0.0421) (0.0467) (0.0560) IMR 0.0760 0.0342 0.00954 0.00248 0.0961 0.101 0.102 (0.0625) (0.0553) (0.0549) (0.0508) (0.0605) (0.0649) (0.0687) BD -0.00522 0.0653*** 0.0827*** 0.0772*** 0.0417* 0.0196 0.0661** (0.0174) (0.0148) (0.0160) (0.0158) (0.0234) (0.0264) (0.0287) WAP -0.344*** -0.430*** -0.325*** -0.201** -0.144 0.00418 0.0321 (0.109) (0.0899) (0.0935) (0.0934) (0.120) (0.141) (0.153) MEN -1.401*** -1.143*** -1.060*** -0.825** -0.862** -0.720* 0.841 (0.405) (0.335) (0.372) (0.330) (0.367) (0.421) (0.527) Constant 87.24*** 90.45*** 74.84*** 56.50*** 53.58** 35.35 -43.60 (24.02) (20.06) (21.20) (19.77) (23.96) (27.21) (33.95)

No. of N2 Regions 73 74 70 70 68 68 68

Prob > F 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000

R-squared 0.176 0.176 0.222 0.286 0.266 0.207 0.146

(21)

21 with the results from aforementioned other studies, which found a positive effect of EU payments on GDP per capita growth, of which one also found an increasingly larger effect after more time had passed since the transfer of EU funds (Becker, Egger and Von Ehrlich, 2010). An additional sixth lead (Model 7) displays statistically insignificant results, and therefore no more leads are added.

In regards to the magnitude of the coefficient estimates, increasing the logged EUPN2 variable by 1 increases the convergence rate of a NUTS-2 region by almost 0.6 %-point in the third year after convergence funds are transferred, by 1.2 %-point in the fourth year, and by 1.6 %-point in the fifth year. These results are considerably large, although they are smaller than the results found for the NUTS-1 regions. Nevertheless, an increase by 1.6 %-point per year could still significantly reduce the gap between the GDP per capita level of a backward NUTS-2 region and the EU-28 average over the course of multiple years.

When the N-number would be kept constant throughout Models 1-4, similar results are shown for all Models 1 to 7 in relation to whether they are significant and their signs (see Table 9 in the appendix). Differences to note are that Model 2 would be slightly less significant, while Model 4 is more significant. Moreover, when the N-number is kept constant, the coefficient estimates of EUP on EUC are considerably larger (up to 2.2 %-point in the Model 5) throughout every model in comparison to when the N-number is not kept constant.

(22)

3_{For most variables data is available up to 2018, while EU payment data for the NUTS-3 regions was solely available for 2013 and 2014.}

Adding a fifth lag, to test for the effect of EU payments made in time = t on convergence in time = t+5, the EU payment data for 2014 was dropped as a whole, due to no convergence data being available for 2019. 22

Model 1 (Base Year) Model 2 (Lead 1) Model 3 (Lead 2) Model 4 (Lead 3) Model 5 (Lead 4) Model 6 (Lead 5) EUPN3 logged -3.783*** 0.309 1.779 -1.192 -12.09 -0.623* (0.827) (1.433) (1.807) (1.010) (12.53) (0.345) GDP 0.00371*** -0.00540*** 0.00233*** -0.000695*** -0.000641 9.16e-06 (0.000229) (0.000316) (0.000410) (0.000249) (0.00102) (1.09e-05) BD -0.121** 0.00972 -0.149 0.0925 N/A (0.0515) (0.0115) (0.126) (0.0584) N/A WAP -0.325 -3.204*** 1.929** -0.450 -10.67 (0.369) (0.618) (0.799) (0.465) (7.728) MEN 1.574* -5.909*** 2.218 2.868** 6.718 (0.868) (1.682) (2.243) (1.230) (11.95) AGR 0.0863* -0.0101 -0.140** 0.0963 7.990 (0.0478) (0.0433) (0.0708) (0.0615) (2.947) Constant -87.26* 568.4*** -257.9* -110.2 302.5 1.906** (45.13) (107.0) (142.8) (78.10) (847.7) (0.952)

Yes Yes Yes Yes Yes No

No. of Observations 454 418 429 539 443 108

No. of N3 Regions 233 238 244 295 436 108

Prob > F 0.0000 0.0000 0.0000 0.0005 0.4182 0.0755

R-squared 0.587 0.644 0.168 0.095 0.805 0.040

(23)

23 When the N-number would have been kept constant, similar results are obtained throughout Models 1-6, both in terms of significance and the sign, as well as the to which degrees certain models were significant (see Table 10 in the appendix). Moreover, while Model 6 again shows a significant and negative effect of EUP on EUC in NUTS-3 regions, the amount of observations is even lower in comparison to when the N-number is not kept constant, which makes it once more rather impossible to derive any valuable conclusions from these results. The results of the regression analysis overall seem to support Hypothesis 3, if one takes into account the presumed selection process which yields a statistically significant negative effect of the EUPN3 variable on EUCN3 in the base year. Adding additional leads in subsequent models yield insignificant effects for Convergence funds transferred in time = t on convergence of NUTS-3 regions in time = t+1 till t+5, and it therefore seems that there is no statistically significant relationship between the transfer of Convergence funds and convergence on the NUTS-3 level.

Robustness of results

Due to the nature and set-up of the data, it was impossible to create a panel data regression which accounted for both fixed country and years effects, rather than fixed NUTS-region and year effects. What however was possible, was to perform a regression with country-fixed effects, without the year fixed effects. Such analysis was performed as a side-quest of the main analysis, and yielded similar results in terms of outcome (sign, size of effect) and significance for all three hypotheses.

An inherent flaw of the Convergence funds dataset which was used to answer Hypothesis 3 is that it records payments in four different categories. Category 1 displays the payments made under the convergence objective/Objective 1, which is the data that was used in the performed analysis. Category 2 and 3 are showing payments made under the ‘Regional competitiveness and employment’ and the ‘Cross-border cooperation’ objectives (Objectives 2 and 3), which are not part of this paper nor part of the performed analysis. Category 4 however recorded payments which were ‘Multi Objective’, which are payments coming from a combination of Objective 1, 2 and/or 3. Although this category recorded relatively few payments (2.218 out of 51.977 payments recorded), this does mean that Objective 1 payments do not fully reflect all the convergence funds transferred to NUTS-3 regions, as some of these funds were recorded under Objective 4 payments instead.

(24)

24 this new variable, as well as other variables under an adjusted N-number for NUTS-3 regions, can be found in Table 11 in the appendix. The regression analysis showed similar (insignificant) results as the main analysis for Hypothesis 3, which used the EUPN3 variable. Full results can be found in Table 12 in the appendix.

VI. Conclusion

The EU Cohesion Policy of 2007-2013 brought forward changes in terms of the stated objectives in comparison to earlier rounds, in which the main objective was renamed to ‘Convergence’, which aimed at stimulating growth in the least developed EU regions. Simultaneously however, the ‘Regional competitiveness and employment’ objective fostered competitiveness of EU regions that were ineligible for Convergence objective funds, which could make it harder for recipients of convergence funds to actually converge towards the EU-28 average in regards to GDP per capita levels.

Therefore, in this paper I aimed to answer the question whether EU Cohesion Policy Convergence funds actually lead to the convergence of recipient-regions in terms of their GDP per capita level towards the EU-28 average. I proposed in the first and second hypothesis that convergence funds would positively affect recipient NUTS-1 and NUTS-2 regions’ convergence rates, as the majority of convergence funds are allocated on the NUTS-2 level, and the focus on efficiency over equity would make it likely that the regions with more growth potential would also receive larger amounts of funding. On the other hand, I expected no significant effect for convergence funds on the convergence rates of recipient NUTS-3 regions. This was due to the fact that e.g. the allocation of funds on the NUTS-2 level can forego funds to flow to high-growth potential NUTS-3 regions, when these regions are part of NUTS-2 regions that overall see a more moderate potential for growth.

This paper indeed found a convergence effect for the NUTS-1 and NUTS-2 regions, as long as the econometric models included certain time leads (time = t + 4 and 5 for NUTS-1 regions, time = t + 3, 4 and 5 for NUTS-2 regions). This means that according to the specified models, convergence funds received in time = t do positively affect convergence rates of NUTS-1 regions four and five years after receiving them, as well as positively affect convergence rates of NUTS-2 regions three, four and five years after receiving them. Due to the nature of the data, the N-number was not kept constant in the main econometric models, but additional testing with a constant N-number found similar effects of convergence funds on convergence rates in NUTS-1 and 2 regions.

(25)

4_{Most of the convergence funds from the 2007-2013 round were transferred to NUTS-1, 2 and 3 regions by the end of 2016, although some}

regions might have received (small amount of) funding from this programming period after this date. 25

The findings therefore confirm the assumption that EU Cohesion Policy Convergence funds do positively affect the convergence rates of recipient NUTS-1 and NUTS-2 regions, although this effect only seems to take place several years after the funds are received by the respective regions. These results are moderately similar to earlier research that focused on the link between convergence funds and GDP per capita growth in previous EU Cohesion Policy rounds, which found that convergence funds positively affected GDP per capita growth, with the size of the effect increasing when more time has passed since the transfer of the funds. Contrarily, the assumption that EU Cohesion Policy convergence funds do not significantly affect convergence rates of recipient NUTS-3 regions also seems to hold true. While this might seem contradictory to results found for earlier programming periods, the change of aims of the 2007-2013 EU Cohesion Policy most likely lead to these results, although only 2013 and 2014 data on NUTS-3 convergence funds were used to come to this conclusion.

(26)

26 VII. Appendix

Table 8: Regression analysis results for Hypothesis 1 (N-number kept constant throughout Models 1-4)

Model 1 (Base Year) Model 2 (Lead 1) Model 3 (Lead 2) Model 4 (Lead 3) Model 5 (Lead 4) Model 6 (Lead 5) Model 7 (Lead 6) EUPN1 logged -1.838 -3.099*** -3.363*** 0.583 3.952*** -0.439 -0.188 (1.129) (1.006) (1.223) (1.291) (0.832) (0.814) (2.241) GDP 0.000678 -0.00299*** -0.00123*** -0.000106 -0.000391 0.000150 0.000608 (0.000644) (0.000452) (0.000442) (0.000398) (0.000391) (0.000405) (0.00136) TE -0.132 -0.0129 0.440*** 0.504** 0.0885 0.184 1.069 (0.109) (0.0786) (0.153) (0.213) (0.222) (0.228) (0.759) IMR 0.398** 0.114 -0.0639 -0.00504 0.194 0.00801 -0.316 (0.192) (0.152) (0.146) (0.167) (0.154) (0.162) (0.372) BD -0.0190 0.0988* -0.0569 0.0154 0.0684 0.118 -0.173 (0.0534) (0.0538) (0.0658) (0.0784) (0.0644) (0.0727) (0.243) WAP -0.939*** -0.993*** -0.147 0.597 1.078** 0.351 -1.514 (0.310) (0.281) (0.330) (0.395) (0.390) (0.430) (1.868) MEN -1.786 4.819*** 3.560* 1.592 1.380 1.373 0.738 (1.644) (1.570) (1.968) (1.635) (1.396) (1.179) (1.916) Constant 151.2* -140.3* -151.4 -130.6 -152.8* -103.6 50.16 (83.11) (74.05) (99.80) (88.76) (80.39) (74.33) (191.0)

No. of N1 Regions 14 14 14 14 13 14 14

Prob > F 0.0061 0.0000 0.0013 0.0000 0.0000 0.0380 0.4126

R-squared 0.366 0.620 0.420 0.522 0.752 0.578 0.859

(27)

27

Model 1 (Base Year) Model 2 (Lead 1) Model 3 (Lead 2) Model 4 (Lead 3) Model 5 (Lead 4) Model 6 (Lead 5) Model 7 (Lead 6) EUPN2 logged 0.755 -1.249** -0.00605 1.395*** 2.176*** 1.831*** 0.365 (0.548) (0.588) (0.490) (0.439) (0.545) (0.599) (0.770) GDP 0.00143*** -0.000619*** -0.000768*** -0.000540*** _-0.000145 _-0.000131 -0.000831*** (0.000196) (0.000179) (0.000145) (0.000138) (0.000169) (0.000183) (0.000239) TE -0.181*** 0.0382 -0.0380 -0.0769* -0.113** -0.0117 -0.00755 (0.0620) (0.0587) (0.0490) (0.0461) (0.0517) (0.0619) (0.111) IMR 0.0387 0.0381 0.0700 0.146** 0.179** 0.200** -0.0594 (0.0824) (0.0838) (0.0661) (0.0693) (0.0793) (0.0808) (0.119) BD -0.0339 0.0872*** 0.0732*** 0.0678*** 0.0416 0.0328 0.00756 (0.0212) (0.0238) (0.0229) (0.0259) (0.0282) (0.0317) (0.0486) WAP -0.430** -0.724*** -0.495*** -0.217 0.109 0.257 0.167 (0.184) (0.184) (0.162) (0.161) (0.225) (0.251) (0.338) MEN -1.844*** -1.549** -1.244** -1.584*** -1.808*** -0.229 0.544 (0.655) (0.662) (0.538) (0.453) (0.545) (0.620) (1.006) Constant 108.9*** 130.2*** 100.3*** 93.95*** 79.53** -10.40 -29.34 (38.05) (37.58) (31.78) (30.16) (37.60) (43.16) (67.65)

No. of N1 Regions 51 51 51 51 51 44 44

Prob > F 0.0000 0.0000 0.0000 0.0000 0.0000 0.0002 0.0118

R-squared 0.265 0.181 0.231 0.263 0.216 0.174 0.147

(28)

28

Model 1 (Base Year) Model 2 (Lead 1) Model 3 (Lead 2) Model 4 (Lead 3) Model 5 (Lead 4) Model 6 (Lead 5) EUPN3 logged -4.215*** -0.307 2.831 0.0622 -1.2469 -0.574 (0.981) (1.449) (2.027) (1.276) (2.222) (0.548) GDP 0.00370*** -0.00557*** 0.00241*** -0.000618** -0.000214 3.17e-05 (0.000253) (0.000322) (0.000434) (0.000285) (0.000686) (2.26e-05) BD -0.114** 0.340*** -0.298 0.0583 N/A - (0.0555) (0.118) (0.180) (0.0763) - WAP -0.208 -2.890*** 2.031** -0.380 0.294 - (0.431) (0.650) (0.854) (0.559) (0.746) MEN 2.149* -5.325*** 1.791 0.662 -5.373* - (1.166) (1.877) (2.576) (1.461) (2.679) AGR 0.0679 -0.0146 -0.138* 0.104 N/A - (0.0491) (0.0437) (0.0758) (0.0646) - Constant -113.0* 495.7*** -235.5 -7.231 258.5* 1.620 (62.31) (122.7) (159.7) (92.12) (142.8) (1.404)

No. of N3 Regions 180 180 180 180 168 54

Prob > F 0.0000 0.0000 0.0000 0.0792 0.0089 0.3637

R-squared 0.599 0.668 0.183 0.067 0.233 0.031

Table 11: Descriptive statistics of used variables

NUTS-3 regions* which received EU Cohesion Policy Convergence funds, as well as a mixture of Convergence and other type of funds found under ‘Objective 4’, at any

moment between 2013-2014.

EUCN3 0.221 2.205 -15.705 23.430 EUPN3* EUPN3* logged 873.96 2.767 743.590 0.461 5.092 0.785 6723.578 3.828 AGR 13.675 13.466 0 63.787 BD 53.080 24.278 12.871 209.697 GDP 13529.43 7498.579 2680.889 44783.28 MEN 50.425 1.052 46.312 53.212 WAP 66.301 2.794 55.490 72.936

*regions that were eligible for funds from both the Convergence objective (CF) and the Regional competitiveness and employment objective (ERDF and ESF) were removed

(29)

29

Dependent Variable EUCN3 (Converge of NUTS-3 regions towards the EU-28 GDP per capita average)

Model 1 (Base Year) Model 2 (Lead 1) Model 3 (Lead 2) Model 4 (Lead 3) Model 5 (Lead 4) Model 6 (Lead 5) EUPN3* logged -4.226*** 0.149 2.156 -0.771 -12.09 -0.535* (0.946) (1.390) (1.797) (1.112) (12.53) (0.322) GDP 0.00374*** -0.00538*** 0.00228*** -0.000577** -0.000641 9.00e-06 (0.000246) (0.000306) (0.000407) (0.000254) (0.00102) (1.14e-05) BD -0.138** 0.00951 -0.122 0.0616 N/A - (0.0581) (0.0110) (0.125) (0.0576) - - WAP -0.388 -3.089*** 1.936** -0.625 -10.67 - (0.418) (0.600) (0.794) (0.471) (7.728) - MEN 1.828* -5.575*** 1.891 1.813 6.718 - (0.986) (1.623) (2.213) (1.236) (11.95) - AGR 0.0785 -0.0149 -0.130* 0.0785 7.990 - (0.0488) (0.0417) (0.0699) (0.0588) (2.947) - Constant -87.03* 538.2*** -242.0* -46.39 298.4 1.681* (52.47) (102.9) (140.9) (77.81) (847.5) (0.899)

No. of N3 Regions 200 205 211 262 393 107

Prob > F 0.0000 0.0000 0.0000 0.0102 0.4182 0.1113

R-squared 0.591 0.659 0.164 0.078 0.805 0.033

VIII. References

Arcury, T. A., Estrada, J. M., and Quandt, S. A. (2010). Overcoming language and literacy barriers in safety and health training of agricultural workers. Journal of Agromedicine, 15(3), pp. 236-248.

Bachtler, J., Mendez, C., and Wishlade, F. (2019). Reforming the MFF and Cohesion Policy 2021-27: Pragmatic Drift or Paradigmatic Shift?. European Research Paper No. 107. University of Strathclyde Publishing: Glasgow.

Barbone, L., & Zalduendo, J. (1999). EU accession of Central and Eastern Europe: Bridging the income gap. The World Bank: Washington D.C.

Bhargava, A., Jamison, D. T., Lau, L. J., and Murray, C. J. (2001). Modeling the effects of health on economic growth. Journal of health economics, 20(3), pp. 423-440.

(30)

30 Becker, S. O., Egger, P., H., and Von Ehrlich, M. (2010). Going NUTS: The effect of EU Structural Funds on regional performance. Journal of Public Economics, 94(9–10), pp. 578-590.

Boldrin, M., and Canova, F. (2001). Inequality and convergence in Europe’s regions: reconsidering European regional policies. Economic policy, 16(32), pp. 206-253.

Cappelen, A., Castellacci, F., Fagerberg, J., and Verspagen, B. (2003). The impact of EU regional support on growth and convergence in the European Union. Journal of Common

Market Studies, 41(4), pp. 621-644.

Dall'Erba, S., and Le Gallo, J. (2008). Regional convergence and the impact of European structural funds over 1989–1999: A spatial econometric analysis. Papers in Regional Science,

87(2), pp. 219-244.

European Commission. (2007). Cohesion Policy 2007-13 – Commentaries and official texts. Office for Official Publications of the European Communities: Luxembourg.

European Commission. (2015). Cohesion Policy 2014-2020: Official texts and commentaries. Publications Office of the European Union: Luxembourg.

European Commission (2019). Historic EU payments - regionalised and modelled. Retrieved

from: https://cohesiondata.ec.europa.eu/Other/Historic-EU-payments-regionalised-and-modelled/tc55-7ysv

European Commission. (2020a). Cohesion Policy. Retrieved from:

https://ec.europa.eu/regional_policy/en/faq/#1

European Commission. (2020b). Region of Attiki. Retrieved from:

https://ec.europa.eu/growth/tools-databases/regional-innovation-monitor/base-profile/region-attiki

Eurostat. (2020). History of NUTS. Retrieved from:

https://ec.europa.eu/eurostat/web/nuts/history

Fiaschi, D., Lavezzi, A. M., and Parenti, A. (2018). Does EU cohesion policy work? Theory and evidence. Journal of Regional Science, 58(2), pp. 386-423.

Gertel, J., and Sippel, S. R. (2014). Seasonal workers in Mediterranean agriculture: The social costs of eating fresh. Routledge: Abingdon, United Kingdom.

Głodowska, A. (2017). Business Environment and Economic Growth in the European Union Countries: What Can Be Explained for the Convergence? Entrepreneurial Business and

Economics Review, 5(4), pp. 189-204.

Hanushek, E. A. (2016). Will more higher education improve economic growth?. Oxford Review of Economic Policy, 32(4), pp. 538-552.

Isenberg, D. and Fabre, F. (2014). Don’t Judge the Economy by the Number of Start-Ups.

Harvard Business Review. Retrieved from:

(31)

31 Medve-Bálint, G. (2018). Implementing EU cohesion policy in the Eastern member states: quality of government balancing between equity and efficiency. In Europeanization Revisited:

Central and Eastern Europe in European Union (pp. 108-126). European University Institute:

Florence.

Medve-Bálint, G. and Šćepanović, V. (2019). EU funds, state capacity and the development of transnational industrial policies in Europe’s Eastern periphery. Review of International Political

Economy, pp. 1-20.

Mohl, P. (2016). Econometric evaluation of EU Cohesion Policy: a survey. In Empirical

evidence on the macroeconomic effects of EU Cohesion Policy (pp. 7-35). Springer Gabler:

Wiesbaden.

Ndedi, A. A., Metha, A., and Nisabwe, F. (2017). The Impact of Health Expenditures on Economic Growth of Cameroon: A Literature Review Perspective. Available at SSRN: http://dx.doi.org/10.2139/ssrn.3036510

Official Journal of the European Union. (2003). Regulation (EC) No 1059/2003 of the European Parliament and of the Council of 26 May 2003 on the establishment of a common classification of territorial units for statistics (NUTS). Official Journal of the European Union, 154, pp. 1-41. Peschner, J. (2015). Demographics and the economy: how a declining working-age population may change Europe's growth prospects. European Commission: DG EMPL. Retrieved from: https://ec.europa.eu/social/main.jsp?catId=1196&furtherNews=yes&langId=en&newsId=240 2

Raghupathi, V., and Raghupathi, W. (2020). Healthcare Expenditure and Economic Performance: Insights From the United States Data. Frontiers in Public Health, 8.

Stoica, O., Roman, A., and Rusu, V. D. (2020). The Nexus between Entrepreneurship and Economic Growth: A Comparative Analysis on Groups of Countries. Sustainability, 12(3), 1186.