European Structural and Investment Funds: A Positive Impact on the GDP Per Capita Growth Rate

University of Groningen

Faculty of Economics and Business

MSc in International Economics and Business (IE&B) Master Thesis

European Structural and Investment Funds: A Positive Impact on the GDP Per Capita Growth Rate

of EU Member States?

Author: Joshua Indenhuck

Student number: S3302180

E-mail address: j.indenhuck@student.rug.nl Supervisor: Dimitrios. Soudis, PhD

2018

Abstract

This paper studies the impact of the European Structural and Investment Fund (ESI) on the EU27 countries in the period 2000 - 2013. Conducting an OLS regression based on the neoclassical growth model by Burnside and Dollar (1997), it seeks to find out if ESI Funds have a positive impact on the GDP Per Capita Growth Rate of the EU27 countries. First, I analyzed the net effect of ESI funds on the GDP Per Capita Growth Rate. Second, I analyzed if the impact of ESI funds is conditional to institutional quality. Third, I analyzed if the impact of ESI Funds on growth is larger in the core EU countries than the peripheral EU countries. The results suggest that there is little evidence that ESI funds have an effect on the GDP per Capita Growth Rate. Furthermore, the effectiveness of ESI funds is conditional to the Government Effectiveness and Corruption Perception Index. Moreover, the ESI Funds have the same impact on growth in the core and peripheral EU member states.

Keywords: European Structural and Investment Fund, Growth, Neoclassical growth model

Table of Contents:

1. Introduction………....1

2. Literature Review………..2

2.1 Growth Theory………...2

Hypothesis 1 2.2 Trade and NEG theory approach to structural funds……….3

Hypothesis 2 2.3 Empirical Findings………4

2.4 Positive Impact………..5

Hypothesis 3 2.5 No Net Impact through Institutions………..6

3. Methodology………..7

4. Data………8

5. Regression Results……….. 10

5.1 Effectiveness of Structural Funds on the GDP per Capita Growth Rate……….10

5.2 Conditional Effectiveness………11

5.3 Robustness Analysis………16

5.4 Arellano and Bond (1991): Two-Step GMM (Structural Funds)……...…………...18

6. Conclusion………23

7. Bibliography……….23

8. Appendix………...26

List of Acronym’s:

European Union (EU)

European Structural and Investment Fund (ESI) European Regional Development Fund (EFRD) Cohesion Fund (CF)

European Social Fund (ESF)

European Agricultural Fund for Rural Development (EAFRD) European Maritime and Fisheries Fund (EMFF)

New Economic Geography (NEG)

Corruption Perception Index (CPI)

Rule of Law (RoL)

1. Introduction:

Greater equality across the European Union (EU) in income has been at the forefront of EU policy since the creation of the European Union. Its motives might have changed over time but its importance has stayed. The main tool through which the EU is trying to achieve this goal is through the European Structural and Investment Fund. Since, the EU is allocating a large sum of its budget to this instrument and promoting it across all channels and institutions it is important to analyze if it is effective in what it is trying to achieve.

Assessing the impact which the European Structural and Investment

(ESI) Fund has on the GDP per Capita Growth Rate is a broad research topic. Nevertheless, it is an important and contradictory topic to investigate. Economists have found mixed and controversial results regarding the ESI Funds impact on growth. For instance, Beugelsdijk and Eijffinger (2005) find a very large positive effect of Structural Funds on economic growth, while others such as Dall’erba and Le Gallo (2008) find no positive effect at all on regional growth. However, even though numerous renown economists have published on this topic, there are still several issues that require further investigation.

First of all, assessing the impact of the ESI Fund on recent GDP per Capita Growth Rate data. This is important since the ESI Fund has changed, adapted, and progressed since its introduction in 1957 during the Treaty of Rome. Second of all, the impact has not been assessed on new member states such as the countries that have joined during the 2004 and 2007 EU enlargements.

This paper aims to investigate whether the European Structural and Investment Fund has a positive impact on the GDP per Capita Growth Rate of the EU27

countries for period of 2000 to 2013 and what conditions may affect its success. This paper arrives at three fundamental conclusions. First, there is little evidence that ESI Funds have an effect on the GDP per Capita Growth Rate. Second, its effectiveness is conditional to institutional quality indicators such as Government Effectiveness and the Corruption Perception Index. Third, the effect of ESI Funds on growth is equal in the core and periphery countries.

This paper is structured as follows. Section 2 briefly reviews the relevant literature. The methodology is described in section 3, followed by a detailed description of the origins and composition of the data in section 4. The empirical results are discussed in section 5, before concluding in section 6. The appendix can be found at the end of the paper in section 8, after the bibliography.

1 ESI funds are also defined as Structural Funds in some papers.

2. Literature Review:

The European Union allocates more than one third of the total EU budget for every Cohesion Policy cycle (1994-1999, 2000-2006, 2007-2013, and 2014-2020) of the European Structural and Investment Fund (ESI). The ESI fund is a regional policy tool and is designed to target all regions and cities in the European Union (EU), with the aim of promoting “job creation, business competitiveness, economic growth, sustainable development, and improve the citizen’s quality of life” (Europa…c2016), according to the EU. In order for the EU to reach their predetermined goals the ESI fund has been fragmented into 5 separate funds, with each one focusing on a separate sector; the European Regional Development Fund (EFRD), the Cohesion Fund (CF), the European Social Fund (ESF), the European Agricultural Fund for Rural Development (EAFRD), and the European Maritime and Fisheries Fund

(EMFF). Since, the ESI fund makes up a very large portion of the total allocated EU budget it is important to examine the effectiveness of this fund in promoting economic growth in the EU. This question has become of even higher importance due to the recent EU enlargements in 2004, 2007, and 2013.

Considering, that the majority of the EU budget is allocated towards the ESI fund, it is of importance to discover if the redistributed EU funds have positively influenced the economic growth rates in the EU member states. Isolating the effectiveness of cohesion policies on regional growth and the convergence of countries, has been a frequently published topic, with some economists finding structural funds to have a positive impact on growth rates, other economists only finding partial evidence and some find none at all. Aside from the results of previous papers, the theoretical approaches that analyze the impact of economic integration, growth, and cohesion policies can be sorted into either growth or trade theories, differentiating between new and old approaches.

2.1 Growth theory:

The core theory of my thesis is the neoclassical growth theory by Solow (1956) and Swan (1956). It attempts to explain long-run economic growth by observing capital accumulation, labor or population growth, and increases in productivity (technological progress). The theory suggests that regions with similar saving rates, population growth, depreciation rates and access to technology will converge in time. Financing capital scarce regions has the effect that its growth rate is temporarily elevated above the steady-state growth rate, which helps the region converge. However, the decreasing marginal product of capital, will only permit the region to converge faster to its own steady state. Yet, still not all countries will converge to the same GDP per capita. Higher investment rates in the form of higher savings and capital may aid the region to reach its steady state faster, but it can only do so temporarily.

In my case, the redistribution of wealth in the European Union in the form of ESI funds should in theory provide regions with more capital for investments, which in turn should have a positive impact on the growth rate.

3 In the regional policy cycle of 1994 to 1999 and 2000 to 2006 the fund was named Financial Instrument for Fisheries Guidance (FIFG) and in the regional cycle of 2007 to 2013 it was named European Fisheries Fund (EFF).

However, by definition the FIFG, EFF, and EMFF are all the same fund.

Hypothesis 1: ESI funds have a positive impact on the GDP Per Capita Growth Rate of EU27 member states.

The majority of all economists that have examined the impact of EU structural funds on regional growth and the convergence of countries have based their papers on the neoclassical growth theory. The original model uses investments, in our case investments are substituted with Structural Funds. Burnside and Dollar (1997) have demonstrated that investments can be substituted with financial aid in the standard neoclassical growth model. However, even though almost all empirical studies in this field are based on the neoclassical growth theory, it doesn’t come without shortcomings. For instance, if you assess it from an econometric point of view, the use of regional datasets, can increase the occurrence of spatial heterogeneity and spatial autocorrelation among the investigated regions (Baltagi, 2008). For this reason, I have chosen to investigate the impact of structural funds on a country-level, using country-level data instead of regional-level data. Assessing the impact of structural funds on a country-level has a few merits. For instance, the analysis is less sensitive to leakage or spillover effects than for instance regional growth regressions may be. Considering the possibility of crowding out

it is safer to assess the impact at a country-level. Country-level analysis also allows one to control for variables that would otherwise not exist on a regional-level, such as the educational attainment.

2.2 Trade and NEG theory approach to structural funds:

New economic geography (Krugman, 1991; Krugman and Venables, 1995) predicts that economic integration through the reduction in transportation costs could potentially lead to a spatial “core-periphery” pattern. This means that, industries that are located in the “core”

experience increasing returns to scale, while industries in the periphery would experience constant returns to scale. According to NEG, regional policy may therefore only project a positive outcome on regional convergence under specific policy measures. For instance, through investments in public infrastructure. Countries would experience a reduction in barriers and transport costs, between core and peripheral regions. However, an intensification of investments into public infrastructure might also have the effect of stimulating the agglomeration of economic activity away from less-developed regions and towards more developed “core” regions. In addition, Vickerman et al. (1999) indicate that governments tend to finance new infrastructure projects between already rich regions, where the demand in this sector is the highest. Additionally, their results state that interregional transport infrastructure benefits the “core” regions more in terms of transport cost reduction than the “peripheral”

regions. Consequently, investment into infrastructure will not automatically lead to a reduction of interregional disparity.

This would have been an interesting hypothesis to test for, however given my data availability it is not possible to precisely isolate investments into infrastructure from the total sum of ESI fund investments. As stated earlier the ESI fund is a cumulative fund made out of 5 separate funds. Three out of the five funds are not meant for investments into infrastructure,

4 The crowding-out effect occurs when a country will replace the money for a project for which the state has already allocated tax paper money for with ESI funds.

while the two remaining funds (Cohesion Fund and European Regional Development Fund) allow for investments into infrastructure, however the funds can also be used for investments into non-infrastructure projects. Therefore, given the availability of a more detailed dataset it would have been an interesting hypothesis to test for. I have mentioned these findings, even though I cannot properly test for it, because I would like to contribute a first rather general approach towards analyzing if Krugman and Venables (1995) NEG theory can hold up within the EU27.

Hypothesis 2: The net effect of ESI funds on the GDP Per Capita Growth Rate of core

EU member states is stronger than for the EU peripheral

member states.

Finally, the new endogenous growth theory by Romer (1986, 1990) predicts that regional economic policy will be effective if its funds are directed towards research and development or human capital resources. As stated earlier, the ESI fund also covers investments in human capital (ESF) and research and development (EAFRD, EMFF, ERDF, and ESF) (Europarl…c2015). However, I won’t test this hypothesis on my sample of countries since I only have aggregate ESI fund data.

2.3 Empirical findings:

There is no universal consensus amongst economists of what the concrete impact of the ESI fund is on economic growth. Economists are divided on whether the effects of ESI funds on economic growth are positive, zero, or only conditionally affective.

Before analyzing the findings of past papers it is necessary to mention three distinct characteristics of the ESI fund in order to understand its motivation and features. Firstly, the ESI fund should be perceived as a form of fiscal federalism, which allows for the transfer of income across jurisdictions. Other national federations that have this capacity besides the European Union, are for instance the United States of America or the German States (Länder).

As stated by Ma (1997) the most important goal of income transfers are to establish economic convergence between the participating jurisdictions and foster economic growth in capital scarce regions. Secondly, projects that have acquired ESI funds have to be co-funded by the receiving country. Thirdly, ESI funds are frequently allocated to pre-determined projects with little room for modification. Therefore, conditional to which project structural funds are diverted too, the effect can be positive, negative, or zero. Thus, even though the total net effect of ESI funds might be positive, negative, or zero it doesn’t necessarily imply that all the projects have the same impact. Furthermore, there are many other factors that affect the effectiveness of Structural Funds.

5 Austria, Belgium, Denmark, Finland, France, Germany, Ireland, Italy, Luxembourg, Netherlands, Portugal, Spain, and Sweden.

6 Bulgaria, Cyprus, Czech Republic, Estonia, Greece, Hungary, Latvia, Lithuania, Malta, Poland, Romania, Slovenia and Slovakia.

2.4 Positive Impact through Institutions

Amongst all the economists that have published on this topic Beugelsdijk and Eijffinger (2005) indicate the largest positive impact. However, they raise the issue of moral hazard and the substitution effect in the distribution of the funds. The main factor that determines the amount of ESI fund a region receives is the welfare level of its inhabitants. A country with a GDP per Capita far lower than the EU average will receive more ESI funds than a country that is on par with EU average. Therefore, politicians or the ruling elite may not be inclined to raise the welfare level of its population. In short, this means that it might be possible that ESI funds are not invested into projects that would have the largest direct or indirect impact on the GDP Per Capita growth rate. On the other-hand the substitution effect implies that a country will replace the money for a project for which the state has already allocated tax paper money for with ESI funds, thereby resulting in a crowding-out effect. The authors state that there are two ways of testing the existence of a moral hazard and the substitution effect. Firstly, you can create two separate regressions, one with countries with strong institutional structures and one with relatively corrupt countries and then observe for which group of countries the impact of ESI funds on GDP per capita growth is larger. Secondly, you can include a corruption variable in the regression and interact it with the log of structural funds. This interaction variable would test if the level of corruption has an impact on the effectiveness of structural funds. The assumption here is that countries with a higher corruption index also exhibit higher levels of moral hazard and the substitution effect. I will not directly test the existence of moral hazard or the substitution effect, however I will be testing if the effectiveness of structural funds is conditional to the level of corruption in a country.

The allocation of ESI funds to productive and unproductive projects is affected by efficiency of transactions in the market. In other words, the higher the efficiency of transactions are in the market the higher the likelihood that ESI funds are invested into productive projects.

Efficient markets are characterized by institutions and laws that ensure strong contract enforcement. Corruption (CPI) and weak government institutions can undermine this (Azfar et al. 2001). Furthermore, corruption as specified by Robinson (1998) reduces the effectiveness of aid-funded development projects.

The idea of its conditionality is taken from Ederveen et al. (2002) paper which hypothesized that some countries which receive structural funds should not be recipients, because of inefficient allocation. However, contrary to the assumption his paper shows that countries with higher levels of corruption do not exhibit an inefficient allocation of ESI funds.

This conclusion therefore supports the continuation of income transfers to less wealthier regions in the EU. Even though, the literature states that the effectiveness of structural funds are not influenced by the levels of corruption, I have still decided to include a corruption indicator (CPI) since I have increased the country sample to include member states in the Baltic and Balkan region, that are known to have higher levels of corruption than the core EU countries.

The conditional effectiveness of structural funds is further supported by Mohl and

Tagen which examined the growth effects of EU structural funds on 126 EU NUTS-1 and EU

NUTS-2 regions. Their findings are particularly interesting since their results indicate that only

Objective

1 payments promote regional economic growth, whereas the sum of Objective 1, 2 and 3 payments do not have a positive and significant impact on the regional economic growth.

Moreover, Burnside and Dollar (2000) assess the effectiveness of international aid on the growth of developing countries and find that the effectiveness of aid is conditional to the strength of domestic policies. Furthermore, Ederveen et al. (2006) used country-level data for thirteen EU countries and covered the time period of 1960 to 1995. Their results suggest that structural funds do not explain growth differentials among the member states and that structural funds are inefficiently allocated unless given to countries with good institutions (Corruption Perception Index, Rule of Law, Government Effectiveness, and Political Stability).

Hypothesis 3: The impact of ESI funds is conditionally to the quality of institutions (Corruption Perception Index, Rule of Law, Government Effectiveness, and Inflation).

2.5 No Net Impact

Economists such as Dall’erba and Le Gallo (2008) have found that structural funds have have no positive effect on regional growth and convergence. In their paper they evaluated the impact of structural funds on the convergence process of 145 European regions over the period of 1989 to 1999. Their results suggest that the regions under investigation have converged over time, however not exclusively due to structural funds. Ederveen et al. (2002) states a similar outcome, however he identifies three reasons why ESI funds lack effectiveness. Firstly, there are no EU mechanisms that impede regional governments from designing projects that meet the EU criteria, but which are not designed to stimulate growth (rent seeking). Secondly, regional governments may allocate their ESI funds specifically to low-productive projects, in order for their region to stay eligible for Objective 1 payments (moral hazard). Thirdly, Ederveen et al (2002) find that every euro transferred to the region, through the ESI fund, reduces state support to regions by seventeen cents. As stated earlier previously this is evidence for the substitution effect (crowding-out).

Despite, the conflicting views the “Third Cohesion Report” by the European Commission (2004) has observed an unconditional convergence rate of 0.5% for the period between 1980 to 1988. Furthermore, the EC estimates a convergence rate across the EU of 0.7 and 0.9% respectively for the period between 1989 to 1993 and 1994 and 2000. My paper is not investigating if the ESI funds aids in the convergence of regions in the EU. However, economic growth is a prerequisite for convergence, therefore if it turns out that ESI funds do not have a positive effect on growth than there also won’t be evidence of regions converging in the EU.

This paper will estimate the impact of structural funds on country-level growth using a pooled cross-section as a baseline model. Furthermore, I will base this study on the neoclassical growth model. Against this background, I extend the current literature by two aspects. First, I am investigating the period 2000 until 2013, a time period that hasn’t yet been investigated by

7 Regions fall under Objective 1 payments if the per Capita GDP is less than 75% of the EU average. Objective 2 payments are directed areas affected by industrial decline, and Objective 3 payments combat long-term unemployment.

other economists. Second of all, I am investigating a larger set of countries (EU27). Previous papers have not assessed the impact of structural funds on the new member states of the 2004 and 2007 EU enlargement. The principal hypothesis I will be testing for is if structural funds have had a net effect on the GDP Per Capita growth rate in the 27 EU member states.

Additionally, I will be testing if structural funds have a larger net effect in countries with higher government effectiveness, lower levels of corruption, a stronger rule of law, or higher/lower levels of inflation. Furthermore, I will test if EU structural funds have a higher net effect in the core EU countries than the peripheral EU countries. I will isolate this effect by adding a robustness check, in order to rule out issues of endogeneity, reverse causality and other factors that might influence the GDP per capita growth rate.

3. Methodology:

This section will provide a detailed overview on the methods that have been used for the empirical analysis in the chapters to follow. This paper investigates the impact of ESI fund expenditure on EU27 countries by firstly estimating an unconditional convergence model based on the neoclassical growth theory model as used by Burnside and Dollar (1997). Burnside and Dollar (1997) have demonstrated in their paper that investments can be substituted with financial aid and incorporated into the standard neoclassical growth theory model. The most natural way to include ESI funds into the neoclassical growth model is by replacing them with the investment rate, given that the fund is mostly made of investments into physical capital. As stated earlier, subsequent works have added other regressors to this model, such as policy measures (Esposti and Bussoletti, 2008) and institutional quality indicators (Ederveen et al.

2006). These works can be interpreted as first attempts to analyze the conditional effect of Structural Funds on growth. Barro and Sala-I-Martin (1992) and Mankiw et al. (1992) laid the ground works for this estimation by deriving the linear regression specification from the transition dynamics of the neo-classical growth model. This method still has GDP per Capita growth as the dependent variable, however it allows the inclusion of conditioning variables.

I constructed an Ordinary Least Squared (OLS) regression that acknowledges the direct and indirect effects the ESI fund has on the GDP per capita growth rate of the EU27. The basic regression as shown below, models the direct effect of the ESI fund on the GDP per capita growth rate. Suppose that the basic regression is given by:

𝑦

= 𝛽

+ 𝛽

ln 𝑔

+ 𝛽

ln 𝑠

+ 𝛽

ℎ

+ 𝛽

ln 𝑝

+ 𝑔

+ 𝛿 + 𝛽

𝑝

+ 𝛽

ln 𝑆𝐹

+

𝑒

(1)

where (𝑦

) is the average annual growth rate of real GDP per capita growth. In line with

Mankiw, Romer and Weil (1992) work, my basic regression includes these explanatory

independent variables; initial GDP per capita in constant 2010 dollars (𝑔

), gross domestic

savings rate in percentage of GDP (𝑠

), Barro and Lee’s average years of schooling (ℎ

),

population growth rate (𝑝

), population growth dummy (𝑝

), the exogenous rate of

technological progress (𝑔

), the rate of depreciation (𝛿), and structural funds (𝑆𝐹

).

Increasing the amount of Structural Fund expenditure has a similar affect as increasing the investment rate (𝑠

) or raising the technological progress (𝑔

). First, this is due to the fact that Structural Funds increases the regional availability of capital and second because Structural Funds are also invested in areas such as R&D that has the potential of increasing the regional TFP level.

One major interest herein is to also asses if the level of institutional quality has an impact on the effectiveness of ESI funds on growth. For this reason, I have conditioned the ESI funds with five institutional quality indicators; Government Effectiveness, Rule of Law, Corruption and Inflation. The variable COND in the second equation stands for conditioning variable. The conditioning variable is made up of the Structural Fund variable and one of the institutional quality indicators. When formally derived from the neoclassical growth theory, the equation can be written as follows:

𝑦

= 𝛽

+ 𝛽

ln 𝑔

+ 𝛽

ln 𝑠

+ 𝛽

ℎ

+ 𝛽

ln 𝑝

+ 𝑔

+ 𝛿 + 𝛽

𝑝

+ 𝛽

ln 𝑆𝐹

+

𝛽

𝐶𝑂𝑁𝐷

+ 𝛽

𝐶𝑂𝑁𝐷

𝑆𝐹

+ 𝑒

(2)

Furthermore, in order to isolate the impact of ESI funds and rule out potential problems such as a sample bias or reverse causality I will perform an extensive robustness analysis. The estimator I will present in this section is the GMM estimator of Arellano and Bond (1991).

4. Data:

Unfortunately, the availability of data, limits the scope of this empirical study dramatically and therefore to some extent pre-determines the time period of investigation. The European Union statistics bureau (EUROSTAT) has only publicly published country-level data of the EU level ESI fund expenditure for the time period of 2000 to 2013. Previously published economic paper’s, such as Ederveen et al. (2006), relied on data from Vanhove (1999) and the Commission Accounting System (SIN-COM)

. Nevertheless, I am trying to identify the impact of Structural Funds on new member states, therefore data dating back to 1960 is not of much use for this particular empirical study. I solely rely on the European Union regional policy evaluation for my data on the country-level structural- and investment fund expenditure.

Furthermore, I divided the total amount of Structural Funds that a country has received in a given year, by the level of GDP (Constant 2010 US$) of the country in that year. Next, I took the natural logarithm of the ESI percentages. In my analysis, I didn’t solely restrict my attention to the European Regional Development Fund (ERDF) such as Ederveen et al. (2006) had, since the share of ERDF compared to the other four funds is no longer as big as before 2000.

For the dependent variable I use the average annual growth rate of real GDP per capita (GDP Per Capita Growth Annual in percentages), for each country over the period 2000-2013.

I have taken this data from the World Bank Development Indicators.

Furthermore, GDP per capita (in constant 2010 US$) has been taken from the World Bank Development Indicators. Since I am using a Solow-Swan neoclassical growth model, I

8 Domenech et al. (2000) granted them access to the Structural Fund database.

created an initial GDP per capita variable, because an assumption of this model is that countries with lower GDP per capita grow faster than countries with higher GDP per capita. In order for the variable to fit my regression I took the natural logarithm.

Similar, to the previous indicators I have also obtained the average gross domestic savings in percentage of GDP from the World Bank Development Indicators, which I have also taken the natural logarithm of.

For the human capital variable, I have used Barro and Lee’s five-year interval data on average years of schooling. I have then interpolated the missing years using Stata. Unlike, Ederveen et al. (2006) I didn’t use the human capital variable from De La Fuente and Domenech (2000) because it didn’t cover the entirety of my country sample.

Moreover, the variable population growth has been taken from the World Bank database. As is usual in literature, the assumption is made that the exogenous rate of technological progress and rate of deprecation add up to 0.05. Since, the variable population growth has negative values I added the value of 3 (in order to take the natural log), which is the lowest negative value in the dataset. Furthermore, I combined the population growth variable and the exogenous rate of technological progress/rate of depreciation (0.05) and took the natural logarithm. This newly produced indicator will be included into all my regression functions.

Moreover, I have also created a dummy variable for all the countries that were negative or zero before having added the value of 3, which I have also included in every regression.

To test the third hypothesis, I use a number of Worldwide Government Indicators from the World Bank. The indicators I have used to proxy for institutional quality are Rule of Law

, Government Effectiveness

, Corruption Perception Index

, and Inflation. I have collected the data for Rule of Law, Inflation, and Government Effectiveness from the World Bank Worldwide Government Indicators and from the general World Bank data website. As an alternative to the institutional quality index, in particular for the Baltic and Balkan European Union member states I have included the Corruption Perception Index from Transparency International. Other variables for institutional quality have been considered such as the institutional quality index (ICRG) from Sachs and Warner (1995). However, I have not been able to gain access to their database and have therefore reverted to using Rule of Law, Inflation (inflation GDP deflator in percentages), Government Effectiveness, and the Corruption Perception Index.

9 Reflects perceptions of the extent to which agents have confidence in and abide by the rules of society, and in particular the quality of contract enforcement, property rights, the police, and the courts, as well as the likelihood of crime and violence.

10 Government effectiveness captures perceptions of the quality of public services, the quality of the civil service and the degree of its independence from political pressures, the quality of policy formulation and implementation, and the credibility of the government’s commitment to such policies.

11 The Corruption Perceptions Index (CPI) measures the perceived levels of public sector corruption in 180 countries and territories. A composite index, the CPI is based on 13 different expert and business surveys.

Furthermore, I have also generated various dummy variables such as the financial crisis dummy, Ireland dummy, core country dummy

, and peripheral country dummy

. The financial crisis dummy omits the years 2008 and 2009 for all countries, while the Ireland dummy variable neglects Ireland from the regression. Moreover, the core country dummy omits all peripheral countries from the regression, while the peripheral country dummy omits all core countries from the regression.

My sample is composed of 27 European Union Countries (Austria, Belgium, Bulgaria, Cyprus, Czech Republic, Denmark, Estonia, Finland, France, Germany, Greece, Hungary, Italy, Ireland, Latvia, Lithuania, Luxembourg, Malta, Netherlands, Portugal, Romania, Spain, Slovakia, Slovenia, Sweden, and the United Kingdom). Croatia is not included in the Study. I have excluded Croatia from the study because it joined the EU in 2014 and thus has not been receiving ESI funds long enough to in order to properly asses impact on growth.

5. Regression Results

5.1 Effectiveness of Structural Funds on the GDP Per Capita Growth Rate

The estimation results for the basic regression of the general model are presented in Table 1. According to the results of Table 1 the sign for the variable Structural Fund (-1.035) in column 3 is negative and significant to the 10% level. Although the Structural Fund coefficient is negative and significant we can still not directly reject the primary hypothesis

. This is due to the fact that the Structural Fund variable might suffer from endogeneity issues caused by reverse causality. The allocation criteria of the ESI funds are likely to be correlated with the dependent variable economic growth. In short this means that a country or region may receive less ESI Funds relative to the previous year because the GDP of the country/region grew. Therefore, I will have the variable Structural Funds undergo an xtabond2 robustness check in the section to follow.

Furthermore, to my astonishment the education variable in column 3 is negative (-0.917) and significant to the 5% level. I believe this is the case because given my sample of EU27 countries there is not much variation in education levels (min 7 and max 13). Moreover, as stated above many of the countries in my sample have high education levels such as Germany or the Netherlands but display low economic growth because the countries have reached a steady state. Therefore, the regression is not fully capable of properly capturing that higher levels in education are in theory supposed to be a catalyst for economic growth. However, this is not an issue for my investigation, since I focusing on the Structural Fund and Institutional Quality Indicators.

As expected the savings variable is positive and highly significant in all four columns.

This is not surprising since countries with a higher savings rate have more domestic capital that

12 Austria, Belgium, Denmark, Finland, France, Germany, Italy, Ireland, Luxembourg, Netherlands, Portugal, Spain, Sweden, United Kingdom.

13 Bulgaria, Cyprus, Czech Republic, Estonia, Malta, Romania, Slovakia, and Slovenia.

14 H1: ESI funds have a positive impact on the GDP growth rate of EU27 member states.

can be invested into projects that promote growth. One thing that should be kept in mind is that countries that have a high savings rate will experience a lower economic impact from ESI funds.

Therefore, high saving rates may have an effect on the Structural Fund coefficient.

Structural Funds are to a large part designed to be invested in infrastructure, development, and finance projects that may take more than a year to materialize in terms of economic growth. Therefore, I have taken the first lag of the Structural Fund (L.Structural Fund) variable in Column 4. Column 4 shows that the economic impact on growth is larger but it is nonetheless still negative (-0.916) and significant to the 10% level.

Table 1

(1) (2) (3) (4)

VARIABLES Restricted Basic Basic with SF Basic with L.Structural Fund Log of GDP Per

Capita Initial

-2.402*** -1.837** -4.492*** -3.992***

(0.610) (0.674) (1.355) (1.411)

Log of Savings 4.537*** 4.992*** 4.842*** 4.570***

(1.311) (1.323) (1.131) (1.100)

Education -0.690* -0.739* -0.917** -0.816**

(0.393) (0.394) (0.337) (0.348)

Log of (Population Growth +0.5)

-2.572 -1.878 -1.408

(1.937) (1.871) (2.024)

Dummy Variable

0.363 0.0112 0.269

(0.631) (0.675) (0.760)

Log of Structural

Fund -1.035*

(0.515)

L.Structural Fund -0.916*

(0.526)

Constant 19.32*** 15.71** 38.39*** 32.69***

(6.579) (5.933) (10.16) (10.99)

Observations 323 322 322 295

R-squared 0.123 0.140 0.166 0.143

Robust standard errors in parentheses

*** p<0.01, ** p<0.05, * p<0.1 5.2 Conditional Effectiveness

Considering the results of Ederveen et al. (2006), that structural funds are only conditionally effective, I included the Corruption Perception Index and created an interaction variable between the structural fund indicator and the Corruption Perception Index indicator in

15 This specific dummy variable omits all observations for the combined indicator population growth, technological progress, and depreciation rate that were prior to the addition of 3 negative or zero.

Table 2. Empirically, there is a broad consensus that corruption hinders the capacity of institutions to efficiently deliver services that are necessary for growth.

The first three columns of Table 2 illustrate the effect that each inclusion of a variable has on the regression results. Moreover, in column 4 I included the Corruption Perception Index in the context of a lagged Structural Fund variable in order to observe if the lagged Structural Fund variable may significantly change. However, the results in column 4 are not significantly different relative to the other columns in Table 1 or Table 2.

The coefficient of the interaction variable of Structural Funds and Corruption Perception Index in column 5 is 0.272 and highly significant. Furthermore, the structural fund coefficient in column 5 has a negative value of -3.012 and is highly significant. The negative sign implies that the effectiveness of Structural Funds is conditional to the level of the Corruption Perception Index in the country. This finding indicates that Structural Funds are positive to growth if the Corruption Perception Index is above a certain threshold. Intuitively this makes sense, since less money is siphoned off into the pockets of individuals and a larger share of the funds are used for development purposes. The results of Table 2 are therefore in line with the empirical views of Mauro (1995) and other economists in his field, which state that higher levels of corruption as measured by surveys of investors is linked with lower investment rates and economic growth. Mauro’s (1995) findings were also later confirmed by Brunetti and Weber (2008) and as a result the IMF, the World Bank and the UN began to pool additional resources into the fight against corruption.

The threshold can be calculated by this simple equation, -3.012 + (-0.272xCPI). The Corruption Perception Index ranges from 0 to 10, where 0 means that a country is highly corrupt and 10 means it is perceived as very “clean” (Transparency…c2017). In this case even if a country in my sample would have a Corruption Perception Index of 10, the Structural Fund Indicator would still be negative. It is important to mention that the Corruption Perception Index may also possibly pick up good governance (Government Effectiveness) or a strong Rule of Law, since they contribute to lower levels of corruption. The exact composition of each individual variable can be found in the appendix. Therefore, there might be other institutional quality indicators that are more capable in proving the conditional effectiveness of Structural Funds.

According to Transparency International, countries that score a Corruption Perception

Index of 10 have close to no public sector corruption. None of the countries in my sample score

a perfect 10. This brings doubt to the value of this finding since it would be mean that Structural

Funds are having a negative impact in all the observed countries. For this reason, I have also

conditioned the Structural Fund indicator with three other Institutional Quality indicators (Rule

of Law, Government Effectiveness, and Inflation) in Table 3.

Table 2: Basic Regression Conditioned with the Corruption Perception Index

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

VARIABLES Restricted Basic with CPI

Basic with CPI and SF

Basic with CPI and L.SF

Interacted CPI and SF Log of GDP

Per Capita Initial

- 2.402***

-0.950 -3.718** -3.184** -3.836***

(0.610) (1.010) (1.383) (1.546) (1.321)

Log of Savings

4.537*** 5.332*** 5.271*** 5.106*** 5.436***

(1.311) (1.710) (1.497) (1.628) (1.375)

Education -0.690* -0.698* -0.889*** -0.810** -0.802**

(0.393) (0.357) (0.298) (0.319) (0.289)

Log of (Population Growth +0.5)

-2.787 -2.058 -1.636 -3.039

(2.046) (1.974) (2.168) (1.976)

Dummy Variable

0.674 0.372 0.657 0.281

(0.756) (0.809) (0.953) (0.791)

Corruption Perception Index (CPI)

-0.385 -0.514 -0.527 1.182*

(0.341) (0.336) (0.433) (0.582)

Log of Structural Funds

-1.194** -3.012***

(0.486) (0.931)

L.Structural Fund

-1.074**

(0.500) Structural

Funds*CPI

0.272***

(0.0970)

Constant 19.32*** 7.718 31.23*** 25.25* 21.66*

(6.579) (8.532) (11.06) (12.77) (10.89)

Observations 323 322 322 295 322

R-squared 0.123 0.151 0.183 0.159 0.201

Robust standard errors in parentheses

*** p<0.01, ** p<0.05, * p<0.1

A general look at Table 3 reveals that the regression results as far as the Initial GDP per Capita, Savings, Education, and Structural Funds (excluding the column containing Inflation) are barely affected by the use of different proxies for institutional quality. A rise in the Savings of a country has the largest percentagewise increase on the GDP per Capita Growth rate relative

16 This specific dummy variable omits all observations for the combined indicator population growth, technological progress, and depreciation rate that were prior to the addition of 3 negative or zero.

to the other indicators. Across the institutional quality indicators, a 1% increase in Savings can raise the GDP per Capita Growth rate of the EU27 from 6.006% (Rule of Law) all the way to 3.591% (Inflation). A surprising finding, as stated earlier, is that the coefficient of Education is negative and highly significant across all Institutional Quality indicators.

In the first column of Table 3 Structural Funds have been conditioned with Government Effectiveness (GE). The Structural fund coefficient in column 1 is -2.328 and highly significant.

Similarly, as for the CPI, the threshold is calculated through this equation -2.328 + (1.031xGE).

According to my results a country needs to have a Government Effectiveness level of at least 2.25 in order for Structural Funds to have a net effect of 0 on the GDP per Capita Growth Rate.

The Government Effectiveness Indicator ranges from -2.5 to 2.5. The World Bank states that countries that score a Government Effectiveness value of 2.5 have strong government institutions that reflect all the key points that are stated in the definition above. The only country in my sample that scores a 2.25 or above is Denmark.

In column 2 I provide results for the Rule of Law (RoL). The Structural Fund coefficient in column 2 is -3.112 and highly significant. Again, the threshold can be calculated by this simple equation -3.112 +(1.373xRoL). Given my findings a country needs to have a Rule of Law level of at least 2.25 in order for the net effect of Structural funds on the GDP per Capita Growth Rate to be 0. The Rule of Law indicator ranges from -2.5 (weak government performance) to 2.5 (strong government performance). Thus, my findings imply that a country needs to have a strong performing government in order for it to properly utilize and invest money from the European Union.

In the third column of Table 3 Structural Funds have been conditioned with the Corruption Perception Index, however not for further comments but merely for comparison reasons as I have analyzed the implications in Table 2.

In the last column, I report the conditionality of Structural Fund aid on Inflation.

Inflation on its own is positive and significant at the 5% level, but on its own its just a statistical artifact and should be ignored because no country will receive zero ESI Funds. However, the interaction term is small and non-significant and so is the Structural Fund indicator. This indicator should be taken with caution since some central banks raise interest rates when growth is high in order to cool down the economy. Therefore, there might be issues of reverse causality in this context.

After having tested the conditionality of the four proxies for institutional quality I will

continue this paper using the Corruption Perception Index as a conditioning variable for

Structural Funds. Furthermore, I will apply a thorough robustness analysis and GMM estimator

on the Corruption Perception Index, Structural Fund, and their interaction variable.

Table 3: Alternative Institutional Quality Indicators

(1) (2) (3) (4)

VARIABLES Government

Effectiveness

Rule of Law

Corruption Perception Index

Inflation

Log of GDP Per Capita Initial

-4.675*** -

4.033***

-3.836*** -2.364*

(1.606) (1.302) (1.321) (1.157)

Log of Savings 5.335*** 6.006*** 5.436*** 3.591***

(1.370) (1.437) (1.375) (0.837)

Education -0.864** -

0.813***

-0.802** -0.564*

(0.314) (0.269) (0.289) (0.287)

Log of (Population

Growth +0.5) -3.659** -3.537* -3.039 -1.012

(1.743) (1.977) (1.976) (1.513)

Dummy Variable

-0.290 0.148 0.281 0.930

(0.854) (0.688) (0.791) (0.594)

Log of Structural Funds

-2.328*** -

3.112***

-3.012*** -0.628

(0.741) (0.666) (0.931) (0.396)

Government

Effectiveness 6.122***

(1.532) Structural

Fund*Government Effectiveness

1.031***

(0.310)

Rule of Law 6.055***

(1.174) Structural Fund*Rule

of Law

1.373***

(0.271) Corruption

Perception Index (CPI)

1.182*

(0.582)

Structural Fund*CPI 0.272***

(0.0970)

Inflation 0.620**

(0.263) Structural

Fund*Inflation -0.00496

(0.0555)

Constant 33.74** 22.30* 21.66* 15.70

(13.43) (11.34) (10.89) (9.330)

Observations 321 321 322 322

17 This specific dummy variable omits all observations for the combined indicator population growth, technological progress, and depreciation rate that were prior to the addition of 3 negative or zero.

R-squared 0.207 0.238 0.201 0.297 Robust standard errors in parentheses

*** p<0.01, ** p<0.05, * p<0.1 5.3 Robustness Analysis:

In the models so far, the findings imply that Structural Funds are only conditionally effective. The findings are however not conclusive, because there is the danger that my variables (Structural Funds and Corruption Perception Index) suffer from endogeneity due to reverse causality. The allocation criteria of the ESI funds are likely to be correlated with the dependent variable economic growth. First and foremost, the allocation of ESI funds depends on the GDP per Capita of a region. For example, if the GDP per Capita of a region is below 75% of the EU average, the region is eligible for “Objective 1” payments. Countries or regions that fall under the Objective 1 payments receive the highest transfers relative to their GDP. In short, this means that the wealthier a country or region becomes and the less Structural Funds it will receive relative to its GDP. Furthermore, the Corruption Perception Index may also suffer from endogeneity due to reverse causality. This is due to the fact that the countries in my sample that score high on the Corruption Perception Index also have low levels of growth because most of the economies have reached their steady state. Therefore, as stated it earlier it may be that the regression doesn’t properly pick up the effect of the Corruption Perception Index. This section will therefore try to overtly deal with these issues by conducting an extensive robustness analysis. Table 4 summarizes these results.

In the first column I have examined the importance of a lagged variable in the context of the CPI variable. The lagged Structural Fund variable is negative and highly significant. A reason for this may be that certain projects might need even more than one year to materialize in terms of economic growth. Furthermore, a country needs to have at least a CPI index of 9.6 in order for Structural Fund projects to have a net effect of 0 on economic growth.

Table 4: Robustness Check

(1) (2) (3)

VARIABLES SF One Period

Lagged Excluding

Ireland Arellano Bond 1991 Log of GDP Per

Capita Initial -3.168** -3.817*** -8.524***

(1.407) (1.341) (1.868)

Log of Savings 5.404*** 5.542*** 26.03***

(1.495) (1.420) (4.212)

Education -0.705** -0.781** -1.483**

(0.303) (0.287) (0.742)

Log of (Population Growth +0.5)

-2.854 -3.036 -8.007**

(2.116) (1.976) (3.905)

Dummy 0.654 0.160 -3.077**

Variable

(0.893) (0.829) (1.510)

Log of Structural Funds

-3.075*** -6.375***

(0.938) (2.388) Corruption

Perception Index (CPI)

1.631** 1.245* 3.355**

(0.596) (0.613) (1.630)

Structural

Fund*CPI 0.283*** 0.595**

(0.101) (0.279) Financial Crisis

Dummy

L.Structural Fund -3.413***

(0.945) L.Structural

Fund*CPI 0.354***

(0.106)

Ireland Dummy

0.788

(0.490) L.Log of GDP

per Capita Growth

0.0440

(0.0692)

Constant 11.13 19.91* 0

(12.81) (11.65) (0)

Observations 295 322 268

R-squared 0.188 0.202

Number of ID 27

Robust standard errors in parentheses

*** p<0.01, ** p<0.05, * p<0.1

Ireland has been one of the most buoyant and successful economies in the EU exhibiting an above EU average growth rate between the 1993 and 2007 that can rival most emerging economies (World Bank…c2017). Therefore, in order to exclude the possibility that the exceptional growth rate of Ireland is the driving force for growth I included a dummy variable that ignores Ireland in the regression. However, the dummy variable for Ireland is non- significant indicating that the data is not sensitive to this variable.

18 This specific dummy variable omits all observations for the combined indicator population growth, technological progress, and depreciation rate that were prior to the addition of 3 negative or zero.

19 I have omitted the years 2008 and 2009 for all the countries in my sample.

20 I have excluded Ireland from the sample.

In the 3

column I included the GMM estimator from Arellano and Bond (1991). The Arellano and Bond estimator is a generalized methods of moments estimator that is used to estimate dynamic panel data models and is designed for situations with few time periods and many individuals. Furthermore, the difference and system GMM estimators are suited for circumstances were the independent variables are not strictly exogenous meaning they are possibly influenced by past or current realizations of the dependent variable or other parts of the regression. In my case, as stated earlier, I have the suspicion that my Structural Fund and CPI indicator are affected by reverse causality. Structural Funds are distributed according to a country’s GDP, therefore a country that receives a certain amount of Structural Funds in year X may not receive the same amount or more in year Y because the GDP as has grown in between these years. Additionally, countries with high CPI values such as Germany, Denmark, and the Netherlands also have low levels of growth because countries with high CPI values exhibit low levels of growth because in most cases they have reached their steady state.

Furthermore, the GMM estimator also eliminates time-invariant country-specific fixed effects by taking the first difference of the regression equation. In addition, the estimator also instruments the variables on the right-hand-side of the first-differenced equation. This removes the problem of omitted variable biases that are constant over time. Furthermore, through the xtabond2 approach parameters can be estimated consistently despite the presence of endogeneity of the independent variables. Lastly, it also allows in spite of measurement errors a consistent estimation of the variables.

The Arellano and Bond estimator in column 3 reveals that the effects of each of the variables are larger in the GMM estimates. The Structural Fund variable becomes relative to the other coefficients substantially smaller (-6.375) but it is still negative and significant to the 5% level. Furthermore, the interaction variable is positive (0.595) and significant to the 5%

level. This result is not very representative since even a CPI variable of 10 would suggest that ESI Funds still have a negative impact on growth. However, I have further dissected the xtabond2 approach in the section to follow.

5.4 Arellano and Bond (1991): Two-Step GMM

In this section I resorted to using the two-step instead of the one-step GMM approach in Table 5 and 6. Economists state that the two-step GMM approach is more favourable because it generates more asymptotic efficient estimates than the one-step GMM approach.

Furthermore, the bias of the standard errors in the two-step GMM are corrected by Windmeijer’s (2005) correction procedure.

Table 5 represents the results of xtabond2 for the Structural Fund indicator. In column

1 I included the standard option noleveleq, which invokes the difference GMM instead of the

system GMM. Furthermore, in order not to have any issues with consistency I included the

option robust and small. Robust is equivalent to cluster country or ID, which I have used in

previous regressions. Moreover, the option small requests small-sample corrections to the

covariance matrix estimate, which results in t instead of z test statistics and F instead of the

Wald 2 test for the overall best fit. The h(1) option specifies the incorrect assumption of

homoscedasticity. Lastly, I had to include the option collapse because the xtabond2 had generated too many number of instruments relative to my number of observations.

The Structural fund variable in column 1 is still negative (-3.319) and has become significant to the 10% level. Furthermore, the first lagged structural Fund variable has become positive (3.481) and significant to the 5% level. This is evidence for the fact that the variable had suffered from reverse causality. The deepest lagged structural fund variable had on the other hand become negative (-2.414) and significant to the 5% level.

In column 2 I ran the exact same regression as in column 1, however I removed the assumption h(1) of homoscedasticity. As you can see the results in column 2 have greatly improved. The Structural Fund coefficient is still negative (-3.951) but has become significant to the 10% level. Furthermore, as predicted the first lagged Structural Fund coefficient became positive (3.756) and significant to the 5% level. The second lagged Structural Fund similar to the previous column turned negative (-2.117) and significant to the 10% level.

In column 3 I excluded the options h(1) and noleveleq and generated a standard two- step GMM estimation. The results in column 3, contrary to column 1 and 2 are all aside from the Savings variable non-significant.

Table 5: Arellano and Bond (Structural Funds Two-Step GMM)

(1) (2) (3)

VARIABLES Difference GMM H1

Difference GMM

Standard GMM

Log of GDP Per Capita Initial -2.469

(2.297)

Log of Savings 32.59*** 31.49*** 5.299**

(4.048) (3.856) (2.184)

Education -0.266 -1.258 -0.835

(1.735) (1.712) (0.552)

Log of (Population Growth +0.5)

-5.124 -6.511 -1.102

(4.203) (4.322) (4.037)

Dummy Variable

-1.410 -1.675 0.300

(2.852) (2.840) (1.322)

Log of Structural Funds -3.319* -3.951* -0.443

(1.945) (2.129) (1.261)

L.Structural Fund 3.481** 3.756** 1.071

21 This is a two-step difference GMM estimation with the assumption of heteroscedasticity.

22 This is a two-step difference GMM estimation without the assumption of heteroscedasticity.

23 This is a standard two-step GMM regression.

24 This specific dummy variable omits all observations for the combined indicator population growth, technological progress, and depreciation rate that were prior to the addition of 3 negative or zero.

(1.427) (1.528) (1.310)

L2.Structural Fund -2.414** -2.117* -1.002

(1.016) (1.114) (0.735)

Constant 17.36

(19.79)

Observations 231 231 257

Number of ID 26 26 26

Standard errors in parentheses

*** p<0.01, ** p<0.05, * p<0.1

Table 6 presents the results of the xtabond2 approach for the Structural Fund, the Corruption Perception Index, and the interaction variable. The majority of coefficients except for the log of Savings and the Log of Population Growth, first lag of Structural Funds and the first lag of the interaction variable are non-significant. This brings doubt to how reliable the significant coefficients are in this GMM estimation. The Sargan and the Hansen Test both support the weak results of Table 6. The P-value of the Sargan is 0 and therefore states that I do not have proper instruments. Furthermore, the P-Value of the Hansen Test is above 0.10 which suggests that my GMM results are weak.

I have chosen to perform the Arellano and Bond (1991) GMM estimator, with the purpose of eliminating the existence of reverse causality in the Structural Fund and Corruption Perception Index variable. I had hypothesised that the Structural Fund indicator would have turned positive and significant in the first or second lag, indicating that the Structural Fund variable does indeed suffer from reverse causality. To my prediction the Structural Fund variable became positive and significant after the first lag in Table 5.

Table 6: Arellano and Bond (Structural Funds and Corruption Perception Index Two-Step)

(1) (2) (3)

VARIABLES Difference GMM H1

Difference GMM

Standard GMM

Log of GDP Per Capita Initial 11.68

(14.35)

Log of Savings 39.62*** 34.25** 21.96**

(14.01) (13.30) (9.004)

Education 1.871 -0.547 1.030

(3.355) (2.355) (2.467)

Log of (Population Growth +0.5)

-5.854 -5.564 -17.91*

25 This is a two-step difference GMM estimation with the assumption of heteroscedasticity.

26 This is a two-step difference GMM estimation without the assumption of heteroscedasticity.

(4.314) (4.720) (9.515)

Dummy Variable

-4.525 -1.599 7.218

(7.085) (7.059) (7.220)

Log of Structural Funds 2.957 1.378 -2.967

(3.983) (4.757) (9.306)

L.Structural Fund -2.317 -1.935 -19.25*

(10.53) (8.314) (10.76)

L2.Structural Fund -2.556 -0.835 -4.522

(6.138) (5.250) (6.424)

Corruption Perception Index

(CPI) -3.904 0.128 5.825

(5.552) (4.671) (7.347)

L.CPI -3.402 -1.878 10.82

(10.42) (8.033) (8.792)

L2.CPI 1.010 -0.0925 5.566

(6.408) (5.498) (6.263)

Structural Fund*Corruption Perception Index

-0.590 -0.636 0.00761

(0.716) (0.774) (1.279)

L.Structural Fund*Corruption

Perception Index 0.959 0.979 3.456*

(1.584) (1.275) (1.782)

L2.Structural Fund*Corruption Perception Index

0.124 -0.0278 0.924

(0.978) (0.840) (0.994)

Constant -304.2

(191.4)

Observations 231 231 257

Number of ID 26 26 26

Standard errors in parentheses

*** p<0.01, ** p<0.05, * p<0.1

In Table 7 I attempted to reject the second hypothesis

. As a whole the estimation results do not show that Structural Funds are more effective in the core EU countries than the peripheral EU countries. The Structural Fund coefficient in column 1 and 3 are both significant to the 10% level and barely different in size. Furthermore, I interacted the Structural Fund variable with the CPI variable in order to test if Corruption levels play a stronger factor in the peripheral countries, which are relative to the core countries more corrupt. Interestingly, the coefficients for “Structural Fund*CPI” are exactly the same for the core and the peripheral set of countries. Therefore, the findings strongly suggest that Structural Funds have the same overall impact in the core and the peripheral EU countries.

28 This specific dummy variable omits all observations for the combined indicator population growth, technological progress, and depreciation rate that were prior to the addition of 3 negative or zero.

29 H2: The net effect of ESI funds on the GDP growth rate of core EU member states is stronger than for the EU peripheral member states.

Table 7: Core and Periphery

(1) (2) (3) (4)

VARIABLES Core Countries

Core Countries with CPI

Peripheral Countries

Peripheral Countries with CPI Log of GDP

Per Capita Initial

-4.149*** -3.636*** -4.149*** -3.636***

(1.277) (1.298) (1.277) (1.298)

Log of Savings

4.930*** 5.479*** 4.930*** 5.479***

(1.309) (1.473) (1.309) (1.473)

Education -1.003** -0.866** -1.003** -0.866**

(0.361) (0.321) (0.361) (0.321)

Log of (Population Growth +0.5)

-1.897 -3.081 -1.897 -3.081

(1.904) (1.971) (1.904) (1.971)

Dummy Variable

0.177 0.377 0.177 0.377

(0.751) (0.833) (0.751) (0.833)

Log of Structural Funds

-1.095** -3.120*** -1.095** -3.120***

(0.528) (0.872) (0.528) (0.872)

Dummy Core -0.783 -0.594

(1.020) (1.048) Corruption

Perception Index (CPI)

1.287** 1.287**

(0.527) (0.527)

Structural Fund*CPI

0.283*** 0.283***

(0.0889) (0.0889)

Dummy

Periphery 0.783 0.594

(1.020) (1.048)

Constant 35.53*** 19.61* 34.75*** 19.01*

(10.28) (10.97) (10.53) (11.13)

Observations 322 322 322 322

R-squared 0.168 0.202 0.168 0.202

Robust standard errors in parentheses

*** p<0.01, ** p<0.05, * p<0.1

30 This specific dummy variable omits all observations for the combined indicator population growth, technological progress, and depreciation rate that were prior to the addition of 3 negative or zero.