Cohesion Policy in the European Union
An empirical impact assessment
© European Union, 1995-2015.
University of Amsterdam
Faculty of Economics and Business
Master in Economics, specialisation International Economics and Globalisation
Student: M.J. van der Burg, 5941180
Supervisor: Dr. M. Micevska Scharf
Statement of Originality
This document is written by M.J. (Martien) van der Burg who declares to take full responsibility for the contents of this document.
I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.
The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
Abstract
This paper evaluates the effect of EU cohesion policy on economic growth and employment rates at the member state level. The cohesion policy was developed by the EU to reduce the large economic disparities that exist within the Union. A data set containing information on the EU-15 for the period 1978-2012 is analysed using an augmented Solow growth model.
The data suggest that the long-term effect (five to fifteen years) on economic growth is not positive, but rather insignificantly negative. No direct impact could be identified. The policy’s impact on employment rates requires a comparable prolonged run-up time and is found to be conditional upon the level of corruption and economic freedom of the recipient member state. The data suggests that only in sufficiently non-corrupt and economically free countries will the policy result in its desired employment increasing effects. Unfortunately for the proponents of cohesion policy, the key recipient member states tend to have relatively corrupt and non-free economies. The estimated effect of the policy on employment rates in those countries therefore is mostly negative for the period under investigation.
Cohesion Policy in the European Union:
An empirical impact assessment
Contents
List of abbreviations... 5
1.
Introduction ... 6
2.
Cohesion Policy ... 9
2.1
What is cohesion policy? ... 9
2.2
How does cohesion policy work? ... 10
2.3
Economic growth and convergence ... 11
2.3.1 Neo-classical growth theory ... 11
2.3.2 New growth theory ... 12
2.3.3 Agglomeration theory and new economic geography... 12
2.3.4 Technology gap theory ... 14
2.3.5 Heckscher-Ohlin model ... 14
2.3.6 Measuring convergence ... 15
2.4
Historic overview ... 16
2.5
Previous findings ... 17
2.5.1 Convergence ... 17
2.5.2 Contribution of cohesion policy to convergence ... 18
3
Methodology ... 21
3.1
Model specification ... 21
3.2
Data description ... 23
4
Empirical analysis... 26
4.1
Economic growth impact ... 27
4.1.1 Fixed effects regressions ... 27
4.1.2 Lagged effects ... 30
4.1.3 Conditional effects ... 32
4.2
Labour market impact ... 35
4.2.1 Fixed effects regression and lagged effects ... 35
4.2.2 Conditional effects ... 38
5
Conclusion ... 41
6
References ... 43
7
Appendices ... 45
7.1
Detailed data description ... 45
7.2
Correlation matrix of the lagged values ... 45
7.3
Correlation matrix of the different government efficiency variables ... 45
List of abbreviations
CAP Common Agricultural Policy
CF Cohesion Fund
CPI Corruption Perceptions Index
DG REGIO Directorate-General for Regional and Urban Policy
EAGGF European Agricultural Guidance and Guarantee Fund
EC European Commission
ECU European Currency Units
EEC European Economic Community
EFW Economic Freedom of the World index
EMU Economic and Monetary Union
ERDF European Regional Development Fund
ESF European Social Fund
EU-13 Set of countries that joined the EU on or after 1 May 2004
EU-15 EU as constituted on 1 January 1995
FIFG Financial Instrument for Fisheries Guidance
ILO International Labour Organization
… the Community shall aim at reducing disparities between the levels of development of the various regions and the backwardness of the least favoured regions or islands, including rural areas.
Treaty establishing the European Community, 1957
1. Introduction
When it comes to economic activity, major differences exist in the EU. Statistics from Eurostat reveal how the 2014 per capita GDP of the Union’s richest member state (Luxembourg, €79.500), is approximately fifteen times higher than that of the EU’s poorest member (Bulgaria, €5.400). Similar data from the U.S. Bureau of Economic Analysis show how the per capita GDP in the richest state (Alaska, $66,160) is only two times higher than that of the poorest state (Mississippi, $31,551). While these figures are extremes, they hint at economic differences within the EU that are significantly larger than those in the U.S. The existence of such economic disparities is confirmed by figure 1.1, which displays regional per capita GDP figures in the EU member states, Norway, Turkey and Macedonia.
Large disparities in per capita GDP are considered unwanted as increased economic activity is associated with desirable characteristics such as high quality of life and low unemployment (Baldwyn & Wyplosz, 2012, p.276). The insight that balanced economic development is a prerequisite for functioning European integration, in both economic and political terms led the European Commission (EC) to seeking to reduce these disparities by supporting regional development in Europe’s ‘less favoured regions’ (Tondl, 2007, p.171). The EU does so via its cohesion policy: a set of funds that predominantly finance projects in poorer European regions. A region’s eligibility for such funding is dependent upon its average GDP per capita. The EU never fully finances a project but always requires the member state or region to co-finance a project. The required ‘co-financing rate’ is also dependent upon the region’s average per capita GDP. Funding the construction of a motorway in an economically backward region would be just one example of a project. Co-funded projects can also cover a range of other themes such as environment, health, tourism, culture, etc. Many projects are highly visible as they are marked by the familiar billboards bearing the EU flag and reference to the EU fund(s) involved in the project (Baun & Marek, 2014, p.2).
While regional policy seeks to reduce economic disparities, its goal is not to make all regions economically homogeneous. Rather, the policy wants to target the diverse development needs in all regions to support employment, business competitiveness, sustainable economic growth and improve the quality of life of all European citizens (EC, 2014c, p.3).
In 2013, the EU spent 54.5 billion euro on this cohesion policy, amounting to approximately 35% of total expenditures in that year (EC, 2014a, p.22). After agricultural spending, cohesion spending is the largest expenditure on the EU budget, signifying the importance of regional policy to the EU. Money is funnelled to projects in the member states in order to ‘strengthen economic, social and territorial cohesion between regions and the EU member states’ (EC, 2014a, p.63).
This cohesion spending is, however, not without controversy. It is often questioned whether money is spent efficiently (that is, whether it is spent in the best possible manner) and whether it is effective (does it accomplish its purpose). Due to a high expenditure error rate the European Court of Auditors, for example, was unable to grant a Statement of Assurance in 2015 for the twentieth consecutive year (Netherlands Court of Audit, 2015a). Whether cohesion spending actually succeeds in decreasing economic differences is therefore still a topic of debate.
Figure 1.1 Regional GDP (PPP) per inhabitant by NUTS 2 regions (2011)
Source: Eurostat
This paper seeks to investigate the long-term effectiveness of EU cohesion policy in the EU-15. It analyses effectiveness at the member state level by identifying the effect of cohesion policy on real GDP per capita growth and employment rates. As will be discussed the cohesion policy has a multitude of policy goals. Stimulating economic growth in economically backward regions is likely the goal that has received most attention from scholars evaluating the policy, but it is not the only policy objective. Increasing job opportunities is another aim and it is also investigated in this paper.
Predominantly using data from the World Bank and the EC the paper determines the effectiveness of the policy. Via a panel regression analysis the conclusion will then answer the main question: does cohesion policy meet its objectives of stimulating economic development and increasing employment?
To answer this question, the remainder of this text is structured as follows: section two provides a literature review. Here, cohesion policy and its instruments are introduced. The literature review subsequently explains the need for cohesion policy via economic theory and provides a brief historic overview. Then the results of previous assessments of the policy’s effectiveness are presented. Section three discusses the methodology of the analysis and the data used. Subsequently, in section four, a quantitative analysis will follow using the policy’s goals of increasing GDP per capita growth and increasing employment in lagging regions. Lastly, the conclusion will summarize the key findings and answer the research question.
2. Cohesion Policy
This section elaborates on the cohesion policy, explains what it is, how it functions and why it was developed. A historic overview and literature review of previous studies of the policy’s effectiveness are also provided.
2.1 What is cohesion policy?
Cohesion policy is a strategic investment policy that targets all EU cities and regions and seeks to boost their economic growth and improve people’s quality of life (EC, 2014b). Since its inception it has developed into the EU’s primary policy for strengthening economic and social cohesion across the Union. Economic and social cohesion seek to boost competitiveness, green growth, employment and quality of life in regions (EC, 2014c, p.5). In 2007 territorial cohesion was added as a third official aim. It refers to promoting transnational cooperation of regions so that they draw advantages from their respective strengths (EC, 2014c, p.5). This objective collects only a handful of projects as the amount of funding available for territorial cohesion is limited. In 2013 less than 3% of all cohesion funding was spent on such projects (EC, 2014a).
The concept of ‘cohesion’ can be interpreted in various ways. Baun and Marek (2014, p.2) argue that for the purposes of cohesion policy, ‘cohesion’ typically is defined as the promotion of ‘convergence’, the reduction of disparities between regions and member states within the Union.
Via planned investments in predominantly less developed regions and member states, the policy aims at promoting economic development and growth-enhancing structural change. Regional policy is certainly not the only EU policy that affects economic growth: policies like the Common Agricultural Policy (CAP), energy policy, fisheries policy and the single market also influence economic development and cohesion. The cohesion policy is however the Union’s most important effort and the only policy that is directly aimed at reducing disparities (EC, 2014b).
Cohesion policy is considered an expression of solidarity between EU member states as the bulk of EU funding is dedicated to the Union’s less developed regions (EC, 2014b). The EU has co-funded tens of thousands of projects over the years. The EC (2014b) claims that between 1988 and 2013 over 800 billion euros was spent on projects that bring about regional economic development.
The EC’s Directorate-General for Regional and Urban Policy (DG REGIO) is the main body responsible for the policy and uses a wide variety of indicators to demonstrate the success of the policy. Amongst others, these include the amount of start-ups supported, length of new co-sponsored roads and railways, amount of research projects funded, increase in broadband coverage etc. These diverse indicators demonstrate the ubiquity of the policy that, according to Baun and Marek (2014),: ‘seems to be everywhere and to do all things’ (p.1). This ubiquity also emerges from the long list of officially stated goals of cohesion policy: boosting competitiveness and green economic growth, providing people with better services, more job
opportunities, better quality of life and connecting regions so that they utilize each other’s strengths and work together (EC, 2014b).
Baun and Marek mention that cohesion policy also serves an important unofficial purpose as a compensatory mechanism to facilitate intergovernmental bargaining amongst member states. The creation of cohesion policy and its ever increasing budget are generally considered a ‘side payment’ made by the wealthier member states to the poorer countries in exchange for agreement to further economic liberalization and integration (Baun & Marek, 2014, p.5). The policy seeks to help economically backward regions to better cope with the increased competitive pressures caused by European integration.
2.2 How does cohesion policy work?
Cohesion policy does not directly funnel money from the EU to the budgets of less developed regions and/or member states but rather co-finances individual projects. The policy is conducted via a variety of funds.
The European Social Fund (ESF) was established in 1960, with the goal of supporting workers affected by industrial restructuring. It focuses on increasing employment and contributing to the improvement of living conditions. The fund has a dual aim to realise this: via the creation of jobs and by assisting training (Nello, 2005, p.276). Programs focus on teaching new skills and preparing the labour force for changing professional situations (EC, 2014c). The ESF also funds projects that combat discrimination and seek to improve the quality of public administration.
Created in 1975, the European Regional Development Fund (ERDF) accounts for the largest share of the cohesion budget and focuses on economic and social cohesion. It was incepted as a means of assistance to regions facing industrial decline (Nello, 2005, p.277). The ERDF still seeks to correct imbalances between European regions by focussing on investment in key priority areas such as innovation and research, digital agenda, supporting small and medium-sized enterprises (SMEs) and the low-carbon economy. By investing in growth-enhancing sectors it aims to boost competitiveness and create jobs throughout the EU (EC, 2014c).
Both the ERDF and ESF are considered ‘Structural Funds’. The Structural Funds also include the European Agricultural Guidance and Guarantee Fund (EAGGF) and the Financial Instrument for Fisheries Guidance (FIFG). Although all these funds impact convergence and economic growth, only the ERDF and ESF are considered key funds for cohesion spending. Funding is available for individual projects throughout the entire Union, yet the largest amount of EU funding is used in regions that are considered ‘less developed’ (EC, 2015). In the current policy configuration, the level of support and the required national contribution is adapted to each region’s level of economic development: less developed (per capita GDP <75% of the EU average), transition (between 75%-90%) and more developed (>90%) (EU, 2014c). The required percentage of co-financing increases as a region is more economically developed.
The Cohesion Fund (CF) is strictly not classified as a structural fund but is closely related to them (Nello, 2005, p.278). The CF was created in 1994 to support member states that feared the additional competitive pressures that result from the economic and monetary union. Member states with a per capita GDP below 90% of the EU average are eligible for CF assistance. In exchange for financial support, countries are obliged to adopt economic policies that are conducive to convergence (Nello, 2005, p.278). The fund mainly invests in transport networks and environmental projects (EC, 2014c).
The focus of the ERDF and ESF is regional, while that of the CF is at the member state. Moreover, each fund has its own programmatic focus and supports projects connected to the aim of the fund. Nonetheless, the eligibility requirements and focuses of the funds do show significant overlap. The combination of ERDF, ESF and CF is the key instrument for delivering EU cohesion support.
2.3 Economic growth and convergence
As mentioned, regional economic differences are deemed important as high economic activity in a region typically is associated with desirable characteristics. This is relevant as Badlwyn and Wyplosz (2012, p.276) found that European integration on the whole has stimulated economic convergence amongst member states yet simultaneously led to economic divergence within member states. Income gaps amongst regions hence increased within member states. This is in line with Tondl’s (2007, p.172) finding that income disparities became less pronounced at the member state level, but that regional income disparities are a reason for great concern. Before this paper provides an original assessment of the effectiveness of the EU’s regional policy, first various economic growth theories will be discussed. This subsection concludes with two formal definitions of convergence.
2.3.1 Neo-classical growth theory
The most influential theory on economic growth was forwarded by Robert Solow (Ederveen, Gorter, de Mooij, & Nahuis, 2002, p.23). His neo-classical growth theory builds upon a capital accumulation function and an aggregate production function. Key assumptions of the model are diminishing returns on capital and technological development that occurs external to the economy and therefore is exogenous (Solow, 1956). An important concept in the neo-classical growth theory is the so-called steady state. When an economy is in steady state, the amount of investments is exactly such that the ratio output-capital is constant. In the steady state, growth of the output per capita is entirely attributable to technological advancements. In this situation the growth rate of per capita GDP is therefore equal to the rate of technological progress.
The neo-classical growth theory states that economies that are not in steady state will eventually converge towards this situation due to the assumption of diminishing returns on capital. With capital being relatively scarce it will find more profitable investments that temporarily stimulate growth so that an economy converges towards the steady state. The theory also predicts that an economy will grow faster as it is further from its steady state. In the neo-classical perspective all economies converge
towards a steady state, yet different economies will reach different equilibria. The steady state of an economy depends upon investment rates and the rate of population growth. Investments in either human or physical capital increase the steady state whereas population growth decreases the steady state. Therefore, provided that countries are similar when it comes to investment rates, population growth and levels of technology they will converge to the same steady state. While initial differences in per capita GDP may exist these will gradually disappear due to convergence.
2.3.2 New growth theory
Unlike exogenous growth theories, new growth theory considers technological progress as the result of investment in human or physical capital (Romer, 1986). Technological advancement therefore is considered to be endogenous to the economy. While the neo-classical growth theory assumes that technology is ‘simply there’, the new growth theory seeks to explain the economic forces that drive technological advancement. Moreover, new growth theory assumes a constant return to capital, so that the productivity per workers increases linearly as more capital is being made available per worker. This assumption is built upon the idea that investment in capital can take many forms like investment in human capital, physical capital or technology capital (Ederveen et al., 2002, p.24)
New growth theory states that a savings increase this will permanently increase the growth of an economy. Investment in capital results in returns that benefit the entire society and lead to innovation. This innovation in turn results in increased profits and the possibility to further increase capital investments. In the endogenous growth theory there is no force to stop this mechanism so that an economy can continue to grow. In other words, technology creates even more technology. Developed economies that developed state-of-the-art technology hence will have the advantage when generating further ideas and therefore may further increase their lead. The result is that poor countries remain poor, while developed countries continue to grow wealthier. In the new growth theory technology thus can be an important cause for economic divergence amongst countries.
2.3.3 Agglomeration theory and new economic geography
New economic geography explains the relocation of economic activity within regions of a particular member state (Baldwyn & Wyplosz, 2012, p.276). It thus provides an explanation for the fact that economic activity is often geographically clustered (Ederveen et al., 2002, p.25). The new economic geography provides a framework that can be used to model the trade-off between dispersal and agglomeration forces (Neary, 2001, p.536).
Agglomeration forces lead to a concentration of economic activity within a country (Baldwin, 1999). Such forces are caused by workers that relocate to where employers are, while employers prefer to locate their production close to where demand is. This induces a circularity in which economic activity concentrates in an economic core. To a firm it is attractive to produce close to where demand is, as it allows the company to gain a better understanding of customer demands and save on transportation costs. After a firms relocates in the economic core the amount of jobs in the periphery falls while employment possibilities in
the core rise. Workers then face an incentive to move to the core region in search of jobs which reinforces this demand-linked circular causality.
Locating in the economic core close to competitors also allows firms to reap economies of scale advantages (Nielsen, Madsen, & Pedersen, 1994). External economies of scale exist if the production costs per unit decrease when the production of the total industry increases. This may occur when a cluster of firms can be serviced cheaper due to having suppliers of semi-manufactures or raw materials nearby. This allows companies to save on transportation costs or to save by buying bulk quantities so that all firms profit. Companies form a cluster in which they all benefit from firm-specific progress. The relocation of one firm to the economic core can then induce other companies to follow suit. Internal economies of scale exist when the long-run unit cost curve decreases (Nielsen et al., 1994). Thus, while external economies of scale profit the industry as a whole, internal economies of scale are firm-specific.
The mechanisms above stand in contrast to dispersion forces that cause economic activity to spread within a country. Land prices in the economic core typically are higher than in the periphery so that firms and workers have an incentive to prefer a location in rural areas. Other negative side effects of economic concentration like traffic congestion or pollution also contribute to the appeal of the periphery. Local competition force is another dispersion force linked to integration of the EU: as competitors relocate to the core region this increases the attractiveness of staying in the rural areas so as to avoid local competition.
Figure 2.1 illustrates the locational effects of European integration. Plotted on the horizontal axis is the share of firms located in the economic core. The grey lines resemble a situation in which no European integration took place. As the amount of firms in the economic core increases, it becomes more attractive for firms to relocate toward the periphery due to the low amount of competitors there. Hence, the dispersion force curve is upward sloping. The agglomeration force curve also increases due to the circular causality that was mentioned above. The share of firms in the economic core is determined by the intersection of both grey curves. When the share of firms in the economic core is to the left of the intersection the agglomeration force outweighs the dispersion force and subsequently periphery firms will relocate towards the economic core until an equilibrium is established. To the right of the intersection the dispersion force is larger than the agglomeration force so that firms relocate towards the periphery until equilibrium is restored.
One can think of European integration as the lowering of trade costs (Baldwyn & Wyplosz, 2012, p.284). Due to integration it becomes cheaper and easier for firms to service remote areas. Firms in the periphery are no longer ‘protected’ by trade costs and the dispersion force decreases. European integration will also reduce agglomeration forces. After all, these forces partly stemmed from saving on transportation costs due to producing close to where demand is. As both forces decrease the share of firms in the economic core in the new equilibrium could shift both left and right compared to the old situation in the diagram. Baldwyn and Wyplosz (2012, p.286), however, indicate that European integration reduces agglomeration
forces less than the dispersion forces so that the new locational equilibrium implies that more firms are located in the economic core.
Figure 2.1
2.3.4 Technology gap theory
The technology gap theory considers technology to have public good properties (Fagerberg, 1987). However, unlike endogenous growth literature, the technology gap literature predicts that these public good properties actually result in economic convergence rather than divergence. The mechanism behind this convergence is fairly straightforward: technology is considered an international public good. Less developed economies profit from a situation where new technology is already invented by developed economies. Rather than having to invest heavily into generating and implementing ideas, laggard countries can simply replicate these inventions at much lower cost.
To what extent a country can imitate these invention depends on its ability to absorb and adapt new technology. If the technological leader can be followed successfully the theory predicts convergence. Like the neo-classical growth theory the technology gap theory is associated with the prediction of convergence, yet the mechanism that causes economically less-developed countries to catch-up is different.
2.3.5 Heckscher-Ohlin model
The Heckscher-Ohlin model of comparative advantages provides a possible explanation for the outcome that income disparities became less pronounced at the member state level. European integration led to trade liberalization and allowed for economic resources to shift between and within member states. According to Baldwyn and Wyplosz (2012, p.278), integration caused economic resources to shift between sectors in each nation and production to reallocate sector by sector across countries. European
integration would allow a member state with a relatively low-educated population to specialise in the production of simple products like clothing, while a relatively highly educated population could specialize in the development of high-tech products such as pharmaceuticals. The reallocation of factors of production towards the clothing or pharmaceutical sector would then induce other sectors to shrink. The Heckscher-Ohlin model therefore predicts national economies to become more specialised and their overall income to increase. Market integration then leads to a better resources allocation that does benefits all member states, but not to the same extent (Tondl, 2007, p.174). The adjustment towards a more efficient allocation (during which some sectors shrink) may impose a hefty burden on some countries, for which they might seek compensation.
In summary, the Heckscher-Ohlin model provides an explanation for the convergence amongst member states and can explain why some European countries require monetary compensation in the form of cohesion support in exchange for European integration. The model, however, does not explain why regional income gaps increased.
2.3.6 Measuring convergence
Recall that in the context of the cohesion policy ‘convergence’ is typically defined as the reduction of economic disparities between regions and member states within the EU. Given this somewhat vague description of convergence it is no wonder that the concept can be formally defined in various ways. From economic growth literature, two key definitions have emerged.
β-convergence refers to the situation in which poor regions grow faster than wealthy regions and therefore catch-up (Monfort, 2008, p.4). The concept is closely related to the neo-classical growth theory. β-convergence occurs when the relation between a region’s initial economic development and its economic growth is negative. Stated differently: countries that initially are economically less developed (and hence have a low GDP per capita) will exhibit higher growth rates than countries that already have a developed economy. The name β-convergence stems from the regression via which this concept is measured. Scholars regress the economic growth rate on a constant, the initial per capita GDP and an error term:
𝑌𝑔𝑟𝑜𝑤𝑡ℎ𝑖,𝑡= 𝛼 + 𝛽𝑌𝑡,0+ ∈𝑖,𝑡
In this regression specification, coefficient β measures β-convergence. The sign of the coefficient indicates whether convergence is taking place, while the size of the coefficient indicates the pace of convergence (or in the case of a positive β: the speed of divergence).
Alternatively, σ-convergence looks at the dispersion of per capita incomes between the economies under investigation. While β-convergence focusses on detecting catching-up processes, σ-convergence simply measures disparities over regions in time (Monfort, 2008, p.5). Again, the name stems the how σ-convergence is measured: testing for σ-σ-convergence is usually based on the standard deviation of regional
GDP per capita. β-convergence investigates the mobility of income within a distribution whereas σ-convergence simply studies the evolution of the income distribution itself (Ederveen et al., 2002, p.27).
2.4 Historic overview
Unlike the CAP or common commercial policy, cohesion policy is not an original policy of the EU (Baun & Marek, 2014, p.11). While the 1957 Rome Treaty did already mention reducing disparities as an aim of the European Community, the EU founders initially did not envisage a large role for the Community in this sphere (Nello, 2005, p.271). Over the years, however, cohesion policy gradually evolved into a policy that currently makes up a significant part of the EU budget.
During the early years the least-developed regions were typically rural communities that profited from the above market-clearing prices guaranteed by the CAP (Baldwyn & Wyplosz, 2012, p.290) so that economically backward regions were indirectly supported anyway. Moreover, the original six member states formed a fairly homogenous group (Nello, 2005, p.279). The need for cohesion therefore was limited. The accession of Ireland, a relatively poor member state, and the United Kingdom, a country facing significant industrial decline that was expected to be a large net contributor to the EU budget, led to establishment of the ERDF in 1975. This fund funnelled resources from the EU budget to the poorest regions, yet its budget was limited.
In 1985 the EC announced its ambition to complete the internal market, raising fears among less favoured member states of not being able to meet the additional competitive pressures of the single market (Nello, 2005, p.279). The 1987 Single European Act therefore not only prepared Europe for a single market but also introduced ‘economic and social cohesion’ as explicit goals of the European Economic Community (EEC). In 1988 the European Council agreed on the regulation that put existing EU funds in the context of economic and social cohesion (Goulet, 2008). The EC therefore considers 1988 the official starting point of its cohesion policy.
The increased focus on regional economic disparities gained momentum due to the accession of Portugal, Spain and Greece, poor countries that benefited little from the CAP due to the characteristics of their agricultural sectors (Baldwyn & Wyplosz, 2012, p.290). Since 1988 the budget of cohesion policy has continued to expand, both in absolute numbers and relative to the total EU budget. In anticipation of the 1999 introduction of the euro, in 1994 the CF was added as an instrument to deliver cohesion funding. Over the years cohesion policy has been continuously critiqued for its complex procedures, lack of coordination and insufficient decentralisation (Nello, 2005, p.282). In response to such criticism and to prepare the EU for the accession of additional countries, the cohesion policy has been reformed several times with a focus on simplification of instruments, concentration of objectives and better monitoring and evaluation of projects.
Since 2006 cohesion policy has been given the additional task of promoting the EU’s broader economic growth and competitiveness. Cohesion funding spending is therefore closely linked to the ‘Europe 2020 strategy’ that seeks to support ‘smart, sustainable, inclusive growth’ for the entire Union. Consequently, cohesion policy now has dual goals of promoting convergence and boosting growth and competitiveness (Baun & Marek, 2014, p49).
2.5 Previous findings
As was mentioned, cohesion spending has not been without controversy. It therefore is not surprising that the effects of the policy on economic development and convergence have been the topics of many studies. Mendez (2011, p.521) even states cohesion policy is probably the most evaluated of all EU policies. This section will present the outcomes of previous assessments of the economic impact of cohesion support. From a methodological perspective, evaluating the impact of cohesion policy is ‘an extremely challenging task’ (Hart, 2007, p.296). This is partially caused by the ambiguity of concepts like ‘cohesion’ and ‘convergence’. Beyond such definitional difficulties, impact evaluation is also complex as the effects of the policy need to be disentangled from underlying trends and from the effects of others policies (Barca, 2009, p.46). According to Hart (2007, p.299) isolating the cohesion policy’s effect from that of other EU policies is especially problematic.
Moreover, the policy’s effects are often indirect rather than direct, and it is difficult assessing the economic impact of investing in educational programs or the construction of a motorway. The difficulty of creating a ‘without cohesion policy intervention scenario’ and the complex set of cause-effect relationships lead some experts to asserting it is impossible to determine the impact, especially in quantitative terms (Baun & Marek, 2014, pp.181-182). A member state or region that exhibits below average economic growth while receiving substantial cohesion support after all does not prove the ineffectiveness of the policy due to lack of a counterfactual.
While the amount of policy evaluating literature continues to expand, some experts stress that the state of empirical evidence remains limited and unsatisfactory (Barca, 2009, p.XXII). While econometric analyses are suited for ex-post analyses, such econometric studies do not offer a conclusive verdict of the policy’s effectiveness (Barca, 2009, p.86). Moreover, econometric analyses are relatively scarce due to the lack of reliable data, particularly at the regional level (Baun & Marek, 2014, p.183). Impact assessments have therefore also used simulation models, detailed case studies and reports of local authorities.
Below, the outcomes to two potential questions will be discussed: did convergence occur since the inception of the policy? And, to what extent was this convergence, if any, caused by cohesion policy?
2.5.1 Convergence
The EC publishes three-yearly ‘Reports on Cohesion Policy’ that build upon both programme evaluation reports and simulation models to assess economic impact. These reports have consistently found that
income convergence is occurring, both at the member state and regional level (Baun & Marek, 2014, p.186). In spite of that, the global financial crisis seems to have reversed this trend as economic disparities widened between 2008-2011 (EC, 2014b, p.3). Nonetheless, the EC remains optimistic, claiming it is likely that the long-run conversion process will continue after the crisis comes to an end (EC, 2014c, p.6). As the main body responsible for cohesion policy, one must however be cautious before accepting the EC’s generally positive cohesion reports (Baun & Marek, 2014, p.184).
The independent Barca Report, published in 2009 at the request of the EC, found that a strong regional convergence took place in the EU during the past 25 years. For the EU-15, however, this trend stopped at the end of the 1990s. This overall convergence was predominantly due to convergence between member states; since the mid-1990s within-country gaps widened. Moreover, during the last decade the convergence process has been slow (Barca, 2009, pp.81-85).
In line with the Barca Report, the 2003 Sapir Report concluded that whether convergence occurred in the EU between 1980-2000 largely depends on the level of analysis: convergence was found at the member state level but grew at the regional level. Inter-country inequality accounted for half of total EU regional inequality in the 1980s. By the mid-1990s this had risen to about two-thirds while intra-country inequality fell by about a third during that same time period. Convergence took place at the macro-level, yet the amount of convergence at this level varied widely. The rapid economic growth of Ireland and East Germany were the key drivers for this overall convergence. The growth of Spain, Portugal and Greece was only slightly above the EU-average and South Italy did not show any sign of catching up (Sapir Report, 2003, pp.59-60).
In summary, the evidence of convergence appears somewhat mixed (Baun & Marek, 2014, pp.190-191). On the whole, convergence seems to have taken place, yet the process appears to have slowed down and was even reversed by the financial crisis. Moreover, there is a large variety in the amount of convergence with some regions showing remarkable successes while other regions hardly picked up growth.
2.5.2 Contribution of cohesion policy to convergence
The EC is among the main proponents of the view that cohesion policy has had a positive impact on convergence (Baun & Marek, 2014, p.192). The outcomes of several different simulation models suggest beneficial effects, the size of which largely depends on the amount of support received as measured by the ratio cohesion support relative to total GDP. The Cohesion Reports claim this is due to the cohesion policy boosting investment in both human and physical capital. To give just one numerical example, according to a simulation, cohesion funding received by Spain between 2000-2009 increased Spanish GDP growth with an expected 1% during the program itself and up to 1.9% in 2015 (compared to a no cohesion policy baseline) (EC, 2014b, p.232). The simulations also linked cohesion support to increased employment. A report by the British House of Lords (2008, pp.26-27) also claims the policy has helped to reduce disparities in the Union. The authors found that between 1995 and 2005, the economic growth rates of the
four countries that received support from the CF (Greece, Spain, Ireland and Portugal) in addition to SF funding exceeded the EU-15 average growth rate. This seems to suggest a ‘Cohesion Fund effect’. Especially Ireland showed remarkable growth, resulting in a GDP per capita that in 2008 was second highest of the entire Union (House of Lords, 2008, pp.26-27).
Other analyses, however, are less optimistic about the policy’s impact. A meta-analysis by the Dutch Bureau for Economic Policy Analysis found a potential average growth elasticity of cohesion support of 0.18. This would indicate that additional cohesion support equal to 1% of GDP would yield 0.18% additional GDP growth per capita. The actual growth effect, again based on meta-analysis is, however, a mere 0.04. The report attributes this large gap to caveats such as crowding out, inappropriate spending and rent seeking. The EU at times funds projects that would have also been undertaken without European funding and hence crowds out national investment. In an original analysis the authors estimate one euro of policy funding to crowd out seventeen cents of national investment (Ederveen et al., 2002, pp.60-65). Illegitimate spending due to corruption also reduces the clout of the budget. The Netherlands Court of Audit concluded that while the EU keeps reasonable track of the status of EU-funded projects, there is little research done into the effectiveness of such projects. This means the EU does, for example, monitor whether an EU-funded community centre is actually built, but does not verify whether this contributed to increasing the liveability of the neighbourhood it was built in (the underlying goal of the project) (Netherlands Court of Audit, 2015b, p.6).
Some experts even argue cohesion policy may be counterproductive as it protects dominant economic interests in backward regions and so prevents structural reforms. These experts claim the policy distorts the location of knowledge intensive industries by attracting R&D firms into areas with a lacking human capital endowment, which inhibits growth (Baun & Marek, 2014, p.202). The result is that cohesion policy is counterproductive by restraining structural change so that economic convergence is actually repressed rather than boosted.
The idea that the impact of cohesion policy on convergence is neither negative nor positive, but rather is only conditionally effective has gained momentum. Ederveen et al. (2002, p.52) found policy impact to depend upon the openness of a country’s economy, with more open economies typically profiting more. Other reports found the quality of administration of importance (Sapir Report, 2003, p.61). For example, low quality local institutions may struggle to actually spend the funding that is allocated to them. Low quality institutions typically are accompanied by corruption, local elitism and political clientelism, which impose a break on economic growth (Baun & Marek, 2014, p.205). This conditionality would be a major problem as the economically backward regions, where the policy is most needed, often suffer from poor public administrations. Related to the aforementioned is Trón’s (2009, p.161) findings that rent-seeking regional governments prefer to maximize their cohesion funding receipt over actually using this funding efficiently.
In summary, despite an ever-increasing literature still no conclusive proof for cohesion policy’s impact on convergence exists. The vague definitions and ubiquity of the policy lead to very different research methodologies. Model simulations seek to identify potential policy outcomes, whereas econometric studies study actual outcomes (and tend to be more pessimistic) (Trón, 2009, p.174). Moreover, according to Trón (2009) ‘the origin of most disagreement concerning evaluations is that they are based on differing philosophical foundations’ (p.155). For example, due to differing assumptions, endogenous growth theory and neoclassical growth theory lead to very different predictions about the potentials for convergence (Baun & Marek, 2014, p.207).
3 Methodology
Many of the projects funded by cohesion policy, like improving the skills of the labour force or modernizing transport networks, may require long lead times before they have any marked effect on GDP growth (House of Lords, 2008, p.26). Data sets that do not cover appropriately long periods therefore yield flawed results about the impact of cohesion support (Trón, 2009, p.174). To avoid misidentifying these long-term effects this paper makes use of data spanning the period of 1978 to 2012 in order to maximise data availability. Given the limited amount of observations for the countries that joined the Union in the 21st century (‘EU-13’) and because of the potential long-term effects of cohesion support, the
new member states have not been included in the data set. The analysis therefore only includes data on the EU-15, that is, the EU as constituted on the first of January 1995. The EU-15 comprises Austria, Belgium, Denmark, Finland, France, Germany, Greece, Ireland, Italy, Luxembourg, the Netherlands, Portugal, Spain, Sweden and the United Kingdom.
For both the structural funds and the CF, implementation depends upon the pattern of submission of payment requests and it is not unusual to see large swings in EU expenditure from one year to the other (EC, 2014a). To prevent putting too much weight on extreme investment rates in any particular year the data set has been divided into seven time periods. Each data point hence spans a five-year period and equals the average amount of cohesion spending during this interval. For the other variables included in the data set five-year averages are also used.
3.1 Model specification
The analysis builds upon the neoclassical Solow growth model and adds exogenous labour-augmenting technological change, as proposed by Mankiw, Romer & Weil (1992). The result is an exogenous growth model that predicts a natural convergence under the assumption of an open market in which capital and technology can flow freely.
The Solow growth model is founded upon a standard neoclassical production function with constant returns to scale and decreasing returns to capital:
𝑌(𝑡) = 𝐾(𝑡)𝛼(𝐴(𝑡)𝐿(𝑡))1−𝛼
This Cobb-Douglas production function follows standard notation: Y denotes output, K capital, L labour and A the level of technology. The model assumes population growth (n), saving rate (s) and technological progress (g) to be exogenous, and then determines a steady state income level. The stock of capital per effective unit of capital is denoted with k. When the level of capital reaches steady state k* the following equilibrium exists:
In this equation δ denotes capital depreciation, which is assumed constant. Different countries will reach different steady state equilibria due to different saving and population growth rates. A higher saving rate results in a higher steady state income. In contrast, the rate of population growth is negatively correlated to a country’s wealth as it spreads the available capital more thinly over the labour force. When the initial capital stock lies below the level of the steady state, capital and output will increase quicker than the labour force until the equilibrium is restored (and vice versa)(Solow, 1956). In an open market comparative advantage dynamics then result in a diffusion of capital and technology towards poorer areas in search of higher rents, leading to natural convergence. Mankiw et al. (1992) assume g to be constant across countries as the advancement of knowledge is not country-specific. Rate δ is also assumed constant as there is no reason to expect variance across countries and because there is no data available that allows for the estimation of country-specific depreciation rates. The sum of n and g is estimated to be 0.05 (Mankiw et al., 1992, pp.411-412).
By analysing a large country data set, Mankiw et al. (1992) found the Solow growth model suffers from omitted variable bias as human capital investment is correlated with savings and population growth. By including human capital investment, the model fits observations better. The authors warn that ignoring human capital leads to incorrect conclusions as scholars, for example, estimated over half of total U.S. capital stock to be human capital. Like saving in the original Solow model, the higher the investment in human capital, the higher the steady state income (Mankiw et al., 1992). Mankiw et al. (1992) also found strong evidence of a significant convergence propensity amongst the OECD countries. The authors showed that if there were no differences in investment and population growth rates there would be a strong tendency for poor countries to exhibit higher growth rates than wealthy ones. Moreover, the further a country is below its steady state, the faster it will grow (Mankiw et al., 1992).
Added to the model is the amount of cohesion support paid as a percentage of GDP and three control variables, resulting in the following regression equation:
𝑌𝑔𝑟𝑜𝑤𝑡ℎ𝑖,𝑡= 𝛽1+ 𝛽2𝑌𝑖,𝑡−1+ 𝛽3𝑠𝑘𝑖,𝑡+ 𝛽4𝑠ℎ𝑖,𝑡+ 𝛽5𝑝𝑜𝑝𝑖,𝑡+ 𝛽6𝑐𝑠𝑖,𝑡+ 𝛽7𝑜𝑝𝑒𝑛𝑖,𝑡+ 𝛽8𝑐𝑜𝑟𝑟𝑢𝑝𝑡𝑖,𝑡+ 𝛽9𝑓𝑟𝑒𝑒𝑖,𝑡+ ∈𝑖,𝑡
The paper uses a similar regression equation to analyse the impact of cohesion policy on the employment rate:
𝐸𝑖,𝑡= 𝛽1+ 𝛽2𝑌𝑖,𝑡−1+ 𝛽3𝑠𝑘𝑖,𝑡+ 𝛽4𝑠ℎ𝑖,𝑡+ 𝛽5𝑝𝑜𝑝𝑖,𝑡+ 𝛽6𝑐𝑠𝑖,𝑡+ 𝛽7𝑜𝑝𝑒𝑛𝑖,𝑡+ 𝛽8𝑓𝑐𝑜𝑟𝑟𝑢𝑝𝑡𝑖,𝑡+ 𝛽9𝑓𝑟𝑒𝑒𝑖,𝑡+ ∈𝑖,𝑡
The variable Ygrowth denotes the GDP growth rate per capita in percentage points. In the equation i stands for the various member states and t denotes five-year periods. The variable sk measures investment in physical capital while sh denotes investment in human capital. Furthermore, pop stands for population growth, cs for cohesion support relative to GDP, open for trade openness, corrupt for corruption and free for economic freedom. In the second regression equation, E denotes the employment rate. A table containing detailed information about the data and its sources is included in the appendix.
3.2 Data description
Data on GDP and GDP growth are obtained from the World Bank. The independent variable Yi,t-1 measures
the average per capita GDP during the five-year period prior to current time t. The augmented Solow growth model of Mankiw et al. (1992) predicts a natural convergence amongst countries which would result in an expected negative coefficient β2. E denotes employment rates. It measures the part of a
country’s population aged 15+ that is employed. Data are from the World Bank, which considers ages 15 and older the working population.
Gross capital formation as a percentage of GDP measures the value of new fixed assets purchased by businesses, government and consumers. It is commonly used as a proxy for investment in physical capital. Fewer consensuses exist concerning the most suitable proxy for measuring human capital. Investment in human capital can take on many forms and this paper follows common academic practice in focussing on education, hence ignoring, for example, the effect of investing in health care on human capital. The data set uses tertiary educational attainment levels as a proxy for human capital. Data is obtained from the Barro-Lee Educational attainment data set, which is frequently used by economic growth scholars and the World Bank. The data set looks at the highest educational attainment of a country’s population aged 15 or older. The data set distinguishes between four categories: no schooling completed, primary, secondary and tertiary. Per five-year period it contains one data entry and it was chosen as the data set contains information across the entire panel. Educational expenditure as a percentage of GDP is another frequently used human capital proxy but is more complex to measure as educational spending typically takes place at multiple levels of government as well as at the family level (Mankiw et al., 1992, p.419). Reliable data on both governmental and private educational expenditure for the EU-15 is not available across the entire period studied so that this variable cannot be used.
The variable pop measures annual population growth expressed in percentage points. Data for this variable comes from the World Bank. This paper follows Mankiw et al. (1992, pp.411-412) in adding five percentage points to this variable to account for the aforementioned technological progress and capital depreciation in the augmented Solow growth model.
The cohesion support variable denotes the actual EU payments to a member state as a percentage of GDP and comprises the payments from both structural funds and the CF. As this variable is equal to the sum of all payments the separate impacts of the three funds cannot be identified. The reason for this is that the available cohesion data from the EC is not sufficiently detailed across the entire panel. While this is unfortunate, it is questionable whether the separate impacts can be identified at all, as support funding is usually highly fungible. Moreover, the three cohesion support funds are strongly positively correlated. In other words, member states typically either receive high support from all funds or limited support from all. To illustrate this, the correlation coefficients of the three separate funds for the period 2000-2013 are included in the appendix. It is important to note that this variable measures actual payments rather than EU budgetary commitments. The difference between the amount of money allocated to a member state
and the actual payment, termed the ‘disbursement ratio’, can be substantial. As an illustration: between 1994-1999 the EU average ratio was 79% yet this varied from 67% in Italy to 90% in Portugal (Sapir Report, 2003, p.61). As actual payments support economic growth and employment rather than budgetary commitments, the latter are ignored by this paper.
The regression equation also includes three control variables that were added based on the literature discussed in the previous section. The variable open denotes the sum of exports and imports of goods and services as a percentage of GDP and hence is a measure of trade openness. Trade openness was added as it might discipline markets and hence be inducive to efficient cohesion spending. In Ederveen et al.’s (2002) research trade openness had low explanatory power, yet the interaction of cohesion support and trade openness proved to be significant in explaining economic growth. The authors concluded that cohesion support only contributes to national growth in sufficiently open economies (Ederveen et al., 2002, pp.52-53). The average value for open is 83.02, varying from a minimum of 31.84 (Spain, 1978-1982) to 334.21 (Luxembourg, 2008-2012).
The variable corrupt denotes a country’s average five-year score on the Corruption Perceptions Index (CPI). This index is compiled every year by Transparency International, a nonprofit organization that monitors corporate and political corruption perceptions worldwide. Countries are scored on a scale from 0-100, on which 0 corresponds to ‘highly corrupt’. The CPI is a composite index drawing on expert opinion and corruption-related data collected by a variety of institutions (Transparency International, 2014). Corruption is added to the regression as several prior studies stressed the importance of administrative quality for cohesion support to be effective. A low score on the CPI is likely a sign of public institutions not responding to citizens’ needs and limited public trust in government and leaders, and may inhibit economic growth (Transparency International, 2014). In the light of cohesion support, high corruption perceptions may induce inappropriate spending and rent-seeking behavior of regional political leaders and therefore reduce the effectiveness of cohesion funding. The CPI is published since 1995, meaning that
corrupt is undefined for three of the seven periods studied in the data set. The variable thus can only be
included in analyses covering the period of 1993 to 2012. For the countries and time period under investigation the mean for this variable is 7.48. The lowest value in the data set is 3.80 (Greece, 2008-2012) and the highest score is 9.32 (Sweden, 1998-2002).
Free represents economic freedom and utilizes the Economic Freedom of the World index (EFW) which is
annually published by the Canadian think tank Fraser Institute. Economic freedom is a rather nebulous term that refers to societies in which economic agents can make free decisions about working, consuming and investing. Moreover, in an economically free society government protects these freedoms and does not constrain them (Gwartney, Lawson, & Hall, 2013, p.V). Countries are scored on a 0-10 scale, on which 10 corresponds to ‘more economic freedom’. The EFW is a composite index that is constructed using 42 variables in five categories: government size, legal structure and property rights, soundness of the monetary system, international trade freedom and regulation of credit, labor and business. Economic freedom is included in the regression as research has provided strong evidence that free economies
exhibit faster economic growth, have higher levels of private investment, achieve higher income levels and lower poverty rates (Gwartney et al., 2013, p.2). The average value for free in the data set is 7.21. The lowest value in the data set is 5.1 (Greece, 1983-1987) while the highest score is 8.43 (United Kingdom, 1998-2002). A variety of other indices of economic freedom exist, but the EFW is the only index that covers the entire period studied.
4 Empirical analysis
For the dependent variable economic growth, the data set has a balanced panel design, containing cross-sectional data for fifteen countries and time-dimension data for seven time periods. For the dependent variable unemployment rate, the panel data is unbalanced as for some member states no employment data is available for the earliest time period(s). Before the effects of the variables in the data set can be analysed first the appropriate panel data specification needs to be determined.
It is necessary to differentiate between the fixed effects models and random effects model. Both models are individual-specific models that assume unobserved heterogeneity across entities captured by entity-specific effects that are denoted by αi. These effects capture unobserved characteristics of the member
states that affect the dependent variables, that is – effects that influence economic growth and unemployment rates. Such characteristics could, for example, be related to geography, the political power of labour unions or national leisure-consumption preferences. These characteristics are assumed to vary across member states but to be constant over time and thus the entity-specific effects lack a time subscript. Before the most appropriate estimator can be selected, first one must check whether the individual-specific effects are correlated with the independent variables: if they are, the fixed effects estimator must be used. In case correlation is not present the random effects estimator is to be used. The fixed effects estimator provides consistent results for both models, whereas the random effects estimator is only consistent for the random effects model. The latter, however, has a higher efficiency and needs to be used if the data support it.
To test for correlation between individual-specific effects and the independent variables a Hausman test is performed. The Hausman specification test compares the estimators of the fixed effects and random effects estimators. The test has the null hypothesis that the coefficients estimated by the efficient random effects estimator are the same as the ones estimated by the consistent fixed effects estimator. The test is based on the null hypothesis assumption that no correlation exists between αi and the right hand side
variables so that both estimators are consistent but the fixed effects estimator is less efficient. If the test is significant, the random effects estimator is inconsistent and hence may not be used. This paper’s data set returns a test statistic that is only just insignificant at the 5% level. While this implies the random effects estimator should be used, it is obvious that the amount of cohesion funding received by a country is directly related to country-specific characteristics. Entity-specific effects thus are correlated with at least one of the independent variables, so that a fixed effects model will be assumed nonetheless.
In addition to the aforementioned individual effects αi the panel data analysis also includes time fixed
effects, denoted by λt . These time fixed effects lack an entity subscript and control for time-specific shocks
that affect economic growth and unemployment figures in all member states. The global financial crisis would be one example of a shock that likely affected growth and unemployment throughout the EU-15 and hence is time-specific but not entity-specific.
Subsequently, a modified Wald test is performed to test for the presence of heteroscedasticity in the data. The test checks whether the regression residuals are uncorrelated with the independent variables in the regression equation, in which case the data is considered homoscedastic. If the null hypothesis of independence is rejected, heteroscedasticity is present. The modified Wald test rejects the null hypothesis for this paper’s data set. This outcome is in line with expectations, as it seems highly unlikely that all of the variation amongst the countries in the data set is captured by the independent variables, so that the error terms will be correlated to the regressors and heteroscedasticity must be assumed.
Lastly, a Davidson McKinnon J-test is performed to determine whether variables should be added in levels or in log levels. The test includes log fits to the model and verifies whether these have explanatory power. The J-test then adds levels to a log fit model and again verifies significance. For the variable measuring prior per capita GDP, the added log fits are not significant at any of the standard levels, whereas the level variable added to the log fit model is significant. This indicates that a level variable should be used, rather than log levels. Most of the other variables in the data set are expressed as ratios, and the interpretation of adding these variables in log levels would be awkward. Therefore this paper expresses all variables in levels.
4.1 Economic growth impact
This subsection analyses the cohesion policy’s impact on economic growth. First the policy’s direct effect is investigated. Then the lagged and conditional effects are added to the regression analysis.
4.1.1 Fixed effects regressions
Table one presents the output of three fixed effects panel regressions. The first regression includes both country-specific effects and time fixed effects. In contrast, regressions two and three respectively only include entity-specific effects and time fixed effects. Regression one hence has the largest amount of regressors and therefore the highest explanatory power: approximately 74% of the variation in per capita GDP growth can be explained by the model. Moreover, two joint tests were performed: that all entity-specific effects are equal to zero and that all time fixed effects are equal to zero. Both tests return highly significant test statistics, requiring the null hypotheses that these effects are zero to be rejected. For the remainder of this paper both types of fixed effects will therefore be included in all regression equations. The coefficient on the variable of interest, cs, is negative. Ceteris paribus, as cohesion support expressed in percentage points of total GDP increases by 1 point, the associated effect is a 0.22 percentage point decrease in the growth rate of GDP per capita. While this result is unfortunate to the EC and other proponents of cohesion policy, it must be noted that the effect is highly insignificant. Due to this insignificance the outcome cannot be interpreted as definitive proof that the policy is ineffective either. It is possible that the variable’s insignificance is caused by the policy requiring longer lead times before exerting its effect. Alternatively, it could be that the policy has a significant impact that is conditional upon another variable. Upcoming subsections investigate the existence of such lagged or conditional effects.
The remaining results of regression one are consistent with the predictions of the Solow growth model. Per capita GDP in the prior period has a negative coefficient that is highly significant, which is consistent with the model prediction of convergence. Lower initial levels of income are associated with higher growth rates. Both investment in physical capital and investment in human capital raise economic growth, yet only investment in physical capital is significant at the 10% level. The sign of population growth is in line with economic theory too: if population growth increases, expected economic growth decreases. The coefficient is however insignificant at all standard significance levels.
The effect of trade openness is significant in all three regression outputs. As the amount of imports and exports relative to GDP increases by 1 percentage point, the expected rise in per capita GDP growth is 0.068 percentage points. This effect may seem limited, but this variable displays large variance in the data set, varying from a minimum of about 32% for Spain during 1978 to 1982 to a maximum of 334% for Luxembourg during 2008 to 2012 so that openness still accounts for a substantial part of economic growth figures. Finally, better scores on the economic freedom index are associated with higher growth rates. While this result is predicted by economic theory, the positive coefficient is only significantly different from zero in regressions two and three.
Table 1: Fixed effects regressions
Dependent variable: per capita GDP growth. Displayed in parentheses are panel robust country clustered standard errors.
Both
(1) Entity-fixed effects only (2) Time-fixed effects only (3)
Yt-1 -0.302 (0.0506)*** (0.0912)** -0.265 (0.0336) -0.0359 skt 0.167 (0.0834)* (0.109) 0.185 (0.0377) 0.0560 sht 0.0277 (0.0719) (0.0887) -0.0196 -0.00192 (0.0283) popt -0.394 (0.508) (0.591) -0.116 (0.331) -0.258 cst -0.219 (0.293) (0.494) 0.364 (0.422) 0.137 opent 0.0682 (0.0164)*** (0.0181)** 0.0636 (0.00577)* 0.0114 freet 0.243 (0.433) (0.281)*** 1.355 (0.352)* 0.638 _cons -0.880 (3.372) (3.292)** -9.130 (3.380) -2.029
Time fixed effects ✓ ✕ ✓
Entity fixed effects ✓ ✓ ✕
R2 0.74 0.49 0.61
N 105 105 105
4.1.2 Lagged effects
As aforementioned, given the nature of the projects funded by cohesion policy long lead times may be required before the policy exerts its full effects on economic growth. To allow for such large time ranges table two includes lagged values of the cohesion support variable. For convenient comparison, table two first reiterates regression one from table one. The subsequent regressions replace cs by its lagged values, eventually allowing for a three period lag in regressions six and seven. As t denotes five-year periods, regressions six and seven allow for lead times up to fifteen years. As it is not clear when the effects of large infrastructural projects or educational projects fully come into being, these relatively long lags have been included to ensure that the full-effect of cohesion policy is captured by the analysis.
Rather than adding the lagged values of cs to the original regression, this paper only includes one lagged cs variable per equation to prevent multicollinearity issues. A correlation matrix of cs and its lagged values is included in the appendix and demonstrates that current and past cohesion spending are highly correlated. Including multiple lagged variables to the original regression would then distort the outcome as individual coefficients can no longer be accurately identified.
Taking lagged effects into account comes at the price of losing observations. As the lag included in the regression equation covers a more distant time interval, missing values appear for the earlier periods under investigation. The resulting decrease in the amount of observations in the data set thus not only leads to a weaker identification of the regressor coefficients but also alters the period studied. One therefore needs to be cautious when comparing the different outputs as the original regression uses data spanning the period 1978 to 2012, while regression four covers 1983 to 2012, regression five covers 1988 to 2012 and finally regressions six and seven cover 1993 to 2012. The variable corrupt is only observed for the four most recent periods as the earliest corruption perceptions index was published in 1995. Due to this limited data availability the variable cannot be included in regression equations that study long time ranges, yet it is added to regression seven which also includes a three period lag of cs.
