Evidence on the convergence theory among the member state of the Eurozone from 1995 to 2016 : do countries and regions in the Eurozone converge to the same GDP per capita level?

(1)

EVIDENCE ON THE CONVERGENCE THEORY AMONG THE

MEMBER STATES OF THE EUROZONE FROM 1995 TO 2016

Do countries and regions in the Eurozone converge to the same GDP per capita level?

Author: Toma Černiauskaitė

Student number: 11117184

Thesis supervisor: Cenkhan Sahin Finish date: 2018 06 26

UNIVERSITY OF AMSTERDAM

FACULTY OF ECONOMICS AND BUSINESS BSc Economics & Business Administration Track: Economics & Finance

(2)

2

Statement of Originality

This document is written by Toma Černiauskaitė, who declares to take full

responsibility for the contents of this document. I declare that the text and the work

presented in this document are 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

(3)

(4)

4

ABSTRACT

This paper looks into economic convergence in term of GDP per capita income across the Euro Currency Area from 1995 to 2016. The convergence is examined using two classifications, countries and

Nomenclature Units for Territorial Statistics, level 2 (NUTS2); population growth is used as a control variable. Both classifications confirm catch-up in Eurozone as a whole, its Northern geographic region and eight countries that joined the Eurozone later than 1999. The Western geographic region shows convergence across its countries and divergence across its regions, also known as ‘diverging-convergence’.

Keywords: convergence, beta-convergence, growth, Eurozone, GDP per capita

(5)

5

1. Introduction

The convergence hypothesis, or the “catch-up effect”, is the process of GDP per capita approaching a common value in a group of countries and might be used in describing a growth path of a set of economies. The theory of convergence can reveal information about income inequality, “interregional differences” or the effectiveness of economic programs (Sala-i-Martin, 1996, p.1326). For this reason, if a number of economies indeed tend to converge to the same level of GDP per capita, this theory could become a substantial base for a decision-making in international economic and financial organizations that possess the same goals for members with different characteristics. Instances of such organizations can be optimal currency areas (OCA), such as the United States (U.S.) and the European Union (EU) or the Euro currency area (Eurozone).

The monetary policy in optimal currency areas is controlled by one centralized body. For this reason, one of the main factors important to the successful functioning of the OCA is the extent to which member states across the OCA are affected by asymmetric shocks (Frankel and Rose, 1998). If there is a large correlation among the business cycles between countries, the common policy applied by the central monetary authority can suit all countries better and more effectively (Frankel & Rose, 1998, p. 1011). However, due to s high level of heterogeneity between EU countries, Eurozone cannot easily achieve proper functioning of the OCA and its fundamental goal of convergence.

To eliminate some of the differences in economic basis, the nominal Maastricht convergence criteria (1992) for interest rates, inflation, government debt and government budget deficit were established (Treaty on European Union, 1992). Besides, the monetary policy in the member states of Eurozone is transnationally regulated by the European Central Bank, while the fiscal policy is managed nationally. Despite the demand for convergence and actions taken to achieve this goal, the real outcome has not been clear yet. Moreover, literature does not examine and not provide a clear answer, whether Eurozone economies are catching-up in per capita incomes. Therefore, the purpose of this paper is to expand the existing knowledge about the empirical validity of the Economic Convergence Theory and examine whether the Eurozone gets closer to its fundamental goal of convergence. This will be done by testing whether the catch-up was seen in the Member States of the Eurozone and its regions in a period from 1995 to 2016.

The research will be conducted using panel data from 19 Eurozone states. The estimates will be obtained by regressing the initial GDP per capita level and average population growth

(7)

7

on the average GDP per capita growth rate during the period 1995-2016. To check the validity of the regression, using a sample of 19 observations, the research will be replicated on the statistical regions of same Eurozone countries based on the second level of Nomenclature Territorial Units for Statistics (NUTS2). The NUTS distinction divides the EU countries into smaller regions of similar size and is aimed to help statisticians in conducting their research. In addition, the Wald test will be performed to accompany some of the findings when different regression outcomes are observed after controlling for the average population growth.

The results, using both classifications, showed significant economic convergence in the Eurozone overall, its Northern region and countries that entered the Eurozone after 1999. There, the GDP per capita levels indeed approach a common value. Eastern and Southern geographic regions as well as the 11 countries that adopted euro in 1999 also displayed similar relationship but it was insignificant. The Western part of the Eurozone had different outcomes for regressions made on country- and NUTS2 units. The convergence among Western countries seems to exist at the same time with a slight divergence among its regions. This is further analyzed in this paper.

This paper is structured as follows. In Section 2 the existing literature and their empirical findings will be discussed. Section 3 will describe the theoretical grounds and analysis of the Convergence effect. In Part 4, methodology, research design and hypotheses of this research will be presented. This will be followed by Section 5, which analyses the results, and Section 6 that will draw a conclusion and provide an answer to the research question.

2. Literature review and empirical evidence

2.1 History of the European Union, the Economic and Monetary Union and the Eurozone

The European Union (EU) is an organization established in 1992 and currently consisting of twenty-eight European states. The members of the EU operate autonomously, however, a part of the decision making is delegated to the official EU authorities of the European Union, namely, the European Parliament, the European Council, the Council and the European Commission. The EU focuses on and takes action in about 35 areas, for instance, migration, social affairs, research, security and the special attention is given to the economic and financial stability of the EU member states (“The European Union: what it is and what it does?”, 2018). More specifically, the Maastricht Treaty explicitly declared its goals for “a higher degree of convergence of economic performance <...> and economic and social cohesion and solidarity among the Member States” (Treaty on European Union, p. 11).

(8)

8

Entering the European Union also means entering its Economic and Monetary Union, EMU. As countries, willing to enter the EU (and simultaneously EMU), were vastly different in their basics, such as wages, saving rates, national debt levels, and economic policies pursued, subsequently, burdening the attainment of a high degree of convergence, three stages were composed to smoothen the economic and financial integration of the EMU candidates. The first stage that took place from 1990 to 1994 was directed to promote the liberty of transactions and importance of collaboration between central banks that would enhance the economic convergence (Treaty on European Union, 1992 p. 11). The second stage took place from 1994 to 1999 and was focused on better coordination of monetary policies among the EU member states and improving the economic catch-up in the union (Treaty on European Union, 1992 p. 36). For this reason, the European Monetary Institute (EMI) was introduced. The EMI was responsible for establishing solid grounds and legal framework for its successor the European Central Bank (ECB), which would be able to coordinate the monetary policy in the Euro currency area (later: Eurozone) (Treaty on European Union, 1992, pp. 37-39). Finally, the third stage started in 1999, when 11 EU members irrevocably converted their local currency to a single currency of the Eurozone, the Euro.

In order to enter stage three of EMU and irrevocably adopt the Euro as well as to ease the economic catch-up and ensure the stability of the Eurozone, a number of convergence requirements were set in the Maastricht criteria. For instance, long-term interest rates in the candidate countries could not be more than 2% higher than the mean of three EU states with the lowest inflation; the inflation rate could deviate at maximum by 1,5% from the mean of three EU states with the lowest inflation; government budget deficit has to be lower than 3% of GDP; government debt cannot be higher than 60% of GDP and no devaluation of the local currency should have taken place 2 years prior to the entry to the Eurozone (Treaty on European Union,1992, p. 27).

Although the Maastricht criteria were believed to increase the economic convergence among the Eurozone member states it has received some criticism amongst famous academics. For instance, De Grauwe (1996, p.1097) argued that only conditions on government debt and budget deficit have a theoretical basis to become a requirement to enter the Eurozone. The totality of the Maastricht criteria is nominal and, as a result, will not lead to the optimal currency area and the attainment of its main goals, economic and financial stability and convergence: “... Europe is more likely to lead to a setback in the existing level of economic integration” (De Grauwe, 1996, p.1097).

(9)

9

Not only the convergence criteria laid out in the Maastricht treaty but also the functioning of the whole Eurozone were criticized by Bibow (2007). He condemned the Maastricht convergence criteria and one ECB regime cannot fit a diverse set of economies. To come to this conclusion Bibow (2007, p. 19) examined the Eurozone during different time periods and argued that business cycles are different among the economies. In case of an asymmetric shock, the requirement to keep the Maastricht criteria satisfied might prevent some countries from stabilizing their economies at the time of downturn, leading to fragility and downward self-fulfilling prophecy (Bibow, 2007, p. 22). In other words, countries experiencing negative shock would be forced to keep same interest rates, inflation, could use government debt and government budget deficit to a limited extent, which would likely prevent stabilization and deepen the recession (Bibow, 2007, pp. 13-15; De Grauwe, 2012). Consequently, would eventually result in divergence amongst the EZ member states.

2.2 Traditional approach to Convergence hypothesis

A number of papers have already tested the applicability of the Convergence Theory in Economics. For instance, a study by Baumol (1986) was performed. He took a sample of sixteenindustrialized countries and tested for convergence of productivity between 1870 and 1979. His research proved that countries with lower GDP per capita rate in 1870 grew at a higher rate compared to the countries with higher initial GDP per capita indicator and indeed all countries converge to the same income per capita level (Baumol, 1986, p. 1076). Later, however, the paper was criticized by De Long (1988) for two flaws, namely, selectivity bias and measurement error. First of all, the issue of selectivity bias was identified because the sixteen countries used had been already classified as the richest and most industrialized countries in the world as found in the investigation by Maddison (1982). According to De Long (1988, p. 1139) it was inappropriate to use countries that had already demonstrated the stronger foundation of proficiency as well as significant development and in other words, were selected ex post. It could not show anything but convergence, thus, it would have been better if the investigation would have been made including countries that were only in the beginning phase of industrialization (De Long, 1988, p. 1139). Second of all, the regression by Baumol (1986) did not account for the estimation error that was likely to be part of income measures (De Long, 1988, p. 1143).

To correct for shortcomings in Baumol‘s (1986) an accompanying study was performed by De Long (1988). In order to avoid the selectivity bias, De Long (1988) increased Baumol’s (1986) initial sample size by six additional countries and used a different regression that

(10)

10

corrects for the measurement error. The investigation did not support the convergence theory since neither convergence of nations with high potential to grow, nor slower growth of initially wealthier countries could be seen (De Long, 1988, p. 1148).

Later Barro and Sala-i-Martin (1992) conducted an investigation on the Catch-up effect in the 48 U.S. member states and compared the result with the results from the sample, containing 98 countries. The researchers used two variables, namely the Gross state product per capita and average income per resident. The results revealed that U.S. states, further from their steady levels, and states with a relatively low initial level of income per capita, possess higher growth rates than more prosperous ones (Barro & Sala-i-Martin, 1992, p. 245) and, thus, confirmed the convergence hypothesis. The catch-up effect in the countries-based group was relatively the same only when accounted for a number of invariable factors, for instance, government expenditure ratio or school participation rates, standing as proxies for technological development or political cohesion (Barro & Sala-i-Martin, 1992, p. 246). The traditional approach demonstrated by Baumol (1996), De Long (1988), (Barro & Sala-i-Martin, 1992) was later criticized by Danny Quah (1996c). Quah (1996c) claimed that conventional regressions on cross-sectional data do show the evidence on convergence or its absence while leaving out the explanation behind the observed results. “Are there costs of adoption that lead to leap-frogging, where it is the temporarily follower economies that jump to being leader, because they find it easier to exploit new discoveries? Or, do persistent advantages accrue to the leader, richer countries, simply by virtue of their already being leader and richer? Do poorer economies need to overcome poverty-trap barriers before they can hope to catch up with richer ones? “– these and similar theoretical questions are not explained by the traditional empirical researches and therefore, could be improved (Quah, 1996c, p. 1047). For this reason, he suggested accounting for the distribution of GDP per capita among countries as well as intra-distribution since this would reveal more information on the country’s initial state, its path towards the common convergence level of a sample and international capital/technological movements (Quah, 1996c, p. 1048). This criticism was later opposed by Sala-i-Martin (1996b, p. 1020) because, firstly, the traditional empirical approach serves for a different purpose (approval or disproof of the catch-up effect per se). Secondly, the alternative method does not have the grounds to provide an explanation to the questions that Quah (1996c) raised.

Nevertheless, Sala-i-Martin (1996b) conducted another research, in which he made a distinction between the absolute -, the conditional - and σ-convergence, and tested a number of samples. The absolute -convergence is said to be present when higher income economies

(11)

11

grow at the smaller rates compared to those of lower income economies; the conditional -convergence exists when there is a positive relationship between the country’s distance from its steady state and the level of growth rate; σ-convergence is present if variance between a set of economies is decreasing over time (Sala-i-Martin, 1996b, pp. 1020, 1027). A major finding is that the conditional convergence coincides with the absolute convergence when the economies are similar in the set of parameters, which is likely to be the case when a number of countries are part of the same international organization (Sala-i-Martin, 1996b, p. 1027). My research will rely on this finding since the Eurozone states do not only depend to the same organization but have common characteristics from the Maastricht convergence criteria.

The existing researches and findings were later summarized and compared by Islam (2003). He analyzed the differences between the neo-classical and new growth theories, reviewed the existing types of convergence, such as beta-convergence, sigma-convergence, conditional convergence, etc. and collated the results obtained from the differences in data, namely, time series, cross-sectional and panel data. One of the main results shows that independently of the method chosen, the evidence of conditional beta catch-up, when the distance from the steady state is positively related to the economy’s growth rate, is robust among both, small and large sample sizes (Islam, 2003, p. 341). Islam also confirmed the fact, discovered by Sala-i-Martin (1996b) that the decreasing variance of GDP per capita levels (sigma convergence) cannot be found if beta convergence is not present (2003, pp. 313-314). Finally, the commonly observed trend exists, stating that controlling for variables increases the degree of catch-up, especially when accounted for the technological differences (Islam, 2003, p. 326).

In 2007 a research was made by E. Marelli, testing various aspects of integration to the European Union. Analyses were made on various economic variables, for instance, employment, productivity, Gross domestic product, and sector specialization (Mareilli, 2007, p.150) for different classifications of the EU member states in periods 1980-2005. The results showed that structural convergence, based on employment, value-added and sector specialization, was found in many research groups and periods, except for new members of the European Union during the latest periods (Marelli, 2003, p. 176). On the other hand, the late entrants to the European Union possessed substantially higher regional catch-up among themselves compared to older EU members.

Following the previous discussions on the relationship between beta- and sigma- convergence by Barro and Sala-i-Martin (1992), Sala-i-Martin (1996b), Young, Higgins and Levy (2008) have examined the data from U.S. states to explore this relationship in depth. The

(12)

12

paper contains both, theoretical and actual, proofs that sigma convergence cannot be found if the beta-convergence is not observed but beta convergence does not imply that the sigma-convergence exists (Young, Higgins & Levy, 2008, pp. 1086, 1088). Thus, it confirmed the primary conclusion by Barro and Sala-i-Martin (1992, p. 227).

2.3 Convergence club

In additional to the general convergence theory, Baumol (1986) also touched upon the idea of convergence clubs. The definition of convergence club stands for a group of countries with similar foundations that eventually converge to one income per capita (also GDP per capita) level and there can be more than one convergence club (1986, p.1080). Later investigation on the convergence clubs were conducted by Quah (1996a) and Canova (2004).

In his paper “Empirics for economic growth and convergence” Quah (1996a) analysed the data on real GDP per capita from two samples. The first sample contained data from 118 world countries during a time span of 23 years, while the second sample dealt with data from U.S. states during 41 years. The results showed that Baumol’s (1986) hypothesis on the presence of convergence clubs was true in the worldwide case only: The polarization can be observed, where two directions can be seen: namely, poor countries are becoming poorer and rich becoming richer; there are two levels of convergence corresponding to two convergence clubs (Quah, 1996a, p. 1372).

Canova (2004) established a unified model for checking the existence of convergence clubs and used European regional and OECD national income per capita data. The existence of convergence clubs was confirmed as four poles of convergence were found in the European sample (Canova, 2004, p.65), similarly, two poles of convergence were found in the OECD countries (Canova, 2004, p.69).

3. Theoretical Framework

In 1956 Robert M. Solow presented a Neo-classical growth model so called: Solow-growth model. The aggregate production function is defined by:

, [1]

where Y stands for output, K – capital stock, A – labour productivity index, L – labour, t – time and AL is also known as an effective labour. Technology is assumed to be

(13)

Harrod-13

neutral, meaning: exerted through the labour force. According to the extension of the Neo-classical growth model by Inada (1963) the production function has decreasing marginal productivity with respect to capital since and . To be more precise, the smaller the stock of capital, the larger it’s marginal product; alternatively, the marginal product of capital will become zero when the stock of capital gets substantially larger. In other words, capital features zero growth when approaching the steady state. Now taking the total derivative of the aggregate production function with respect to time results in:

, [2]

Where Y represents derivative of output with respect to time. After some simplification, the following equation is obtained:

, [3]

here stands for proportion invested in capital, and R = , also known as Solow residual. After some more arithmetic, this results in:

[4]

To put in perspective, the output growth per capita is dependent on capital-labour ratio and the Solow residual. Based on studies by Solow (1956) and Inada (1963), two conclusions can be drawn: i) if two countries have similar characteristics in terms of labour, capital and technological progress, country that is further from their balanced growth path will grow at higher rate than a country closer to the balanced growth path; ii) all countries will converge to the same value of income per capita rate. An important assumption is that fundamentals (labour, capital and technological progress) are fixed in the short run but can change in the long-run. Therefore, the common balanced growth path (implication ii) can only be achieved in the long run, after there was a change in the fundamentals.

(14)

14

4. Methodology

4.1 Research design

To recall, beta (β) convergence is said to hold if lower beginning income countries possess higher income growth rates than the countries with initially higher income levels (Sala-i-Martin, 1996, pp. 1020, 1027). In order to test if all Eurozone member states converge to the same income per capita level, an empirical econometric analysis will be performed. Based on the traditional convergence model (Baumol, 1986; Mankiw et al., 1992; Quah 1996; Sala-i-Martin 1996b) I will regress the primary level on GDP per capita on the GDP per capita growth rate using the Ordinary least squares method (OLS) by Gauss-Markov theorem. The estimated regression line is presented below:

[5]

where is the intercept, known as constant; is the regression parameter, that shows the change in the dependent variable as a result of a change in the independent variable; stands for the standard error with the expected value of zero and is assumed to be independent across the periods. t defines the start of the period, in this case, 1995, and T is the duration of the period estimated, namely 21 years. The ) is the natural logarithm of the GDP per capita estimate at time t; denotes the average growth rate of GDP per capita during the period t to t+T in logarithmic terms:

[6]

To put the equation 5 in perspective, if the country has high initial income and smaller growth rate (β-convergence is present), then coefficient has to be negative. It is because the higher GDP per capita level at the beginning of the period, , will result in higher average growth rate, , if > 0 (Inada, 1963). That is the opposite of the beta convergence identified by Sala-i-Martin (1996b, p. 1023)

In addition, I am going to use the average population growth as the control variable as, it might affect the GDP per capita estimate if there is a high degree of population movement (if increase in GDP is not big enough, the GDP per capita estimate will decrease due to high immigration and vice-versa) (Mankiw et al., 1992; Islam, 1995). The purpose is to avoid the

(15)

15

fluctuations in GDP per capita that are caused by other factors than labour productivity, technological progress and change in capital stock. The mean population growth rate will be denoted as , where

. [7]

The ultimate regression, that will be estimated, is therefore:

[8]

This research will be based on the key papers described in the literature review. As a base, for testing I will use a standard model used by De Long (1988), Barro & Sala-i-Martin (1992), Sala-i-Martin (1996b) and others. Since all countries of interest belong to the European Union and the Eurozone, I will rely on the finding of Sala-i-Martin (1996b, p. 1027) that in a group of countries, belonging to the same international organisation and, thus, implying similar economic parameters, absolute and conditional beta-convergence are the same. For this reason, if beta convergence was found, its type would not be distinguished. Finally, Panel data will be used as according to the Islam (2003, p. 341), this should produce similar findings, despite the sample size.

In order to test the economic convergence in the member states of the Eurozone from 1995 to 2016, I will examine and analyze convergence on a few levels, namely:

 Eurozone as a whole,

 Geographic regions,

 Time of entering the Euro area.

The Eurozone level analysis will be made to investigate whether the Euro currency area as a whole does show signs of convergence and whether its member states move to the same steady-state. Due to similarities in fundamentals, such as interest rates, inflation rate, government debt and budget deficit, it is expected that the member states of the Eurozone are converging to the same steady state (Solow, 1956; Inada, 1963)

The geographical regional level study is aimed to account for similarities on geographical as well as to some extent economical/developmental factors. Regions of interest contain Northern, Eastern, Western, Southern parts of Europe. This will allow for a better comparison of countries within the particular regions and comparison of different regions.

(16)

16

Regional separation was chosen as according to Quah (1996b, pp. 953-954) geographical location accounts for spill-over effects and is more important in convergence than individual macro-economic factors.

The time of entering the Euro area is chosen to investigate whether countries, that were the first to enter the Eurozone show any trends in their economic growth and convergence. Since all the countries had to satisfy the Maastricht convergence criteria, early entrants of the Eurozone are supposed to be converged more and, therefore, are concentrated around lower GDP per capita growth rates.

4.2 Testing method

In order to look for possible causes and relations between the GDP per capita growth rate and the population growth rate, as well as the relation between GDP per capita growth rate and its particular regions, the Wald test will be performed. The Wald test for single variable tracks the relationship between independent and dependent variables. It tells if an explanatory variable is significant in predicting the dependent variable. To get the test statistic, the comparison is made between the beta coefficient value predicted by the maximum likelihood ( ) and the hypothesized value . The difference is then tested for significance, where se stands for the standard error of the estimate:

[9]

Regression analyses and tests will be performed on overall Eurozone level as well as on different Groupings of its member states. These groupings include and account for regional and time of entry differences.

4.3 Hypotheses

In the theoretical growth model Solow(1956) and Inada(1963) showed that a production growth has a decreasing marginal productivity with respect to capital. As confirmed by the previous researchers (Baumol, 1986; Sala-i-Martin, 1996; etc.) beta-convergence occurs when countries with higher initial level of income per capita should have lower rates of income growth. In other words, initial level of GDP per capita is inversely related to its growth rate.

(17)

17

For this reason, the extent to which starting level of GDP per capita, , affects the GDP per capita growth rate should be negative. Thus,

Hypothesis 1: The lower the initial level of income per capita is, the bigger the growth rate

will be (=absolute Beta convergence). Or equivalently,

.

Hypothesis 2: Countries that fulfilled the convergence criteria earlier are in the further stage

of catch-up and, therefore, has lower growth rates. In other words, average growth rates of early Eurozone members should be smaller than that of later members. Or equivalently,

.

The Hypothesis 1 will be tested for all three groupings: Eurozone as a whole,

geographic regions and time of entry. Whereas the Hypothesis 2 will be tested only for time of entry group.

4.4 Data collection

To conduct the analysis on the country classification, GDP per capita estimates (in US $) from 19 Eurozone countries between 1995 and 2016 were collected. To keep problem-geared data and problems with rounding and choice of an exchange rate, estimates were kept in US dollars. According to the World Bank, the GDP contains the revenues earned by the residents of the country after adding back sales tax and deducted for the subsidies. Reductions from depreciation, impairment and depletion were not taken into the account. Consequently, the obtained GDP estimates were adjusted for the middle year population. Estimates of the population size present the mid-year population of individual countries. To adjust for scaling differences between estimates natural logarithmic values were used. Values of natural logarithms and average growth rates using equation 6 were calculated in excel. For all calculations 7 decimals were used.

To conduct the analysis on the classification of Nomenclature for Territorial Units for Statistics, data was obtained from the Eurostat database. GDP per capita estimates for NUTS2 classification were not found thus it were obtained by dividing yearly GDP (in euros) at current market prices by yearly population (on the first of January). Again, to account for differences in scaling, logarithmic values of GDP per capita were used and calculated in excel using 7 decimal points.

(18)

18

5. Results

5.1 Beta-Convergence Among Countries (Country classification) 5.1.1 Summary statistics

Firstly, the summary statistics of Eurozone member states will be analyzed. Table 1, in the Appendix II: Tables and Figures, contains the main statistical indicators listed for different classifications of the Eurozone countries. A geographical classification was made according to the EuroVoc multilingual thesaurus agency. EuroVoc geographical division is used as the agency itself is an official information supplier to the EU institutions. A diversification, regarding the time of entry, was separated into two main categories. The first category contains the early entrants or in other words 11 countries that entered the Eurozone in 1991 (denoted as EZ11). For the sake of simplicity, all states that joined the currency area at the later time belong to the same research group (denoted EZ8) despite their different entry times. Additionally, Figure 1 shows the GDP per capita movement across the Eurozone Member states from 1995 to 2016. Except for Luxembourg and Ireland, seventeen other countries have income per capita estimate between approximately 15000 to 45000 US $.

From the Table 1 heterogeneity in income per capita estimates between the EZ19 countries can be seen. In 1995 Lithuania was the poorest country and its GDP per capita was approximately 26 times lower than GDP per capita of the richest country, Luxembourg. Secondly, the Northern region has the second lowest initial GDP per capita estimator but shows the highest average GDP per capita growth, approximately 7.13%. Moreover, Western Countries are the richest and have the smallest rate, 2.49%. Fourthly, the early entrants of the EZ had on the average larger starting point in terms of income per capita and show lower economic growth rate compared to the late entrants, while creating some positive yet not significant signs about the convergence of GDP per capita level in the Eurozone.

5.1.2 Results from the whole Eurozone

In order to analyze convergence in the sample of all Eurozone member states, I performed a linear regression analysis in STATA. Heteroskedasticity-robust regression was used to account for variations in standard errors. The findings are presented in Table 2. The first column represents the outcomes of regression where only the natural logarithm of GDP per capita in 1995 was used as an independent variable. The second column presents the results of regression where the average population growth from 1995 to 2016 was used as a control

(19)

19

variable. In both cases coefficients are negative and significant with alpha of 1%. This means that countries, poorer in 1995, show higher growth rates compared to rich countries. Thus, Eurozone countries are indeed converging to one GDP per capita rate and being in one convergence club.

Figure 2 contains the graphical representation of the results. On the X-axis there is the logarithmic value of income per capita in 1995. The Y-axis plots the average of income per capita growth during from 1995 to 2016. Indeed the initial GDP per capita levels are inversely related to its growth rate, seen by negative sloping regression line, and confirm the economic a catch-up of lower-income countries.

5.1.3 Results when grouped to geographic regions

Different results are found when testing for convergence in geographic regions. Here 19 Eurozone countries are grouped into four regions. The Western region contains Austria, Belgium, Germany, France, Ireland, Luxembourg and the Netherlands. The Northern region consists of Estonia, Finland, Latvia, and Lithuania. The Eastern region is composed of the Slovak Republic, Slovenia. Finally, Cyprus, Greece, Italy Malta, Portugal, Spain are the Southern region. The graphic representation of the relationship between the logarithmic value of initial GDP per capita level and the average growth rate in the period from 1995 to 2016 for different geographic regions of the Eurozone can be found in the Figure 3.

Table 3 shows regression estimates when the average GDP growth was regressed on the initial level of income per capita (see Reg*). Negative betas were found in all cases, hence, it follows that countries in Western, Northern and Southern regions slightly tend to approach same GDP per capita level. Significant convergence can be found only in the Northern region, implying countries’ movement to the same convergence club. Because of a too small sample of Eastern countries (2) reliable estimates cannot be obtained.

When controlled for the average population growth rate (Reg**) fit of the model in different regions is slightly improved. Except for the Western region which now also shows significant convergence and the model fit is improved from to . For this reason, I decided to check for correlations between the average population growth and growth of GDP per capita (see Table 4).

Eurozone as a whole together with the Northern region demonstrates an inverse relationship between the GDP per capita and population growth, and Southern region has a relatively small positive correlation. The Western countries possess high positive correlation unlikely the countries in the North. This implies that increase in population is accompanied by

(20)

20

an increase in GDP per capita in Western countries but decrease in the Northern ones. The positive effect in the Western countries might be explained by the so-called ‘brain drain’. The brain drain is the outflow of the highly skilled labor force. Western European countries face substantial immigration of skilled and professional people, who are able to work successfully and achieve high results at a lower cost. This later results in economic growth which is higher than the increase in the population and so raising the GDP per capita level (Beine, Docquier & Rapoport, 2001; Borjas, 1989; Favell, 2008).

To examine the empirical relation between the GDP and Population growth in the Western Eurozone countries, the Wald test was performed. The F value was found to be 30.45 and is significant with p-value = 0.0053. Therefore, Western Eurozone countries are indeed (positively) affected by the increase in population growth.

5.1.4 Results when grouped according to the time of joining the Euro area

The second classification was based on the time of entering the Eurozone. In 1991, 11 countries joined the third stage of the EMU and became part of Eurozone. The early entrants are: Austria, Belgium, Finland, France, Germany, Ireland, Italy, Luxembourg, the Netherlands, Portugal, Spain. The rest 8 countries joined Euro currency area at different points in time and are Cyprus, Estonia, Greece, Latvia, Lithuania, Malta, Slovak Republic, Slovenia. The findings are seen in Figures 4&5 and Table 5.

Regressions show that early entrants do not experience much convergence among themselves. The negative betas confirm that higher GDP per capita countries have smaller GDP per capita growth rates, however, this relationship is insignificant. When controlled for the population growth rate, the fit of the model increases from 0.0415 to 0.7032 but no significant changes, regarding the convergence, appeared.

On the opposite, later entrants of the Eurozone show a substantial convergence in both regressions. These 8 countries are indeed approaching same value of income per capita, meaning that countries like Estonia, Latvia, Lithuania, that are further away from the steady state, indeed show higher growth rates than countries that had higher beginning income. When the average GDP per capita growth rates were compared (Table 1), the early entrants of EZ indeed had lower average rate than the later entrants: 0.024374 < 0.054838. This means that the early entrants have already converged more, are closer to the steady state of convergence and, hence, have smaller GDP per capita growth rates.

(21)

21

Even though the regressions made on country-units provide some significant result in terms of GDP per capita convergence, the estimations might suffer from low power due to small sample sizes. To avoid this problem and correct for possible estimation errors, additional research is done. This time I will use the data from the second level of the Nomenclature Territorial Units for Statistics (NUTS2). 1As a result, instead of 19 observations in country classification, the sample will have 176 observations, statistical regions, in NUTS2 classification. It is worth mentioning that the following regressions concern the regional instead of country convergence. The groupings, namely, Eurozone as a whole, geographical regions and time of entering the Eurozone, will stay the same. The conclusions of the NUTS2 regression are presented in the next section.

5.2 Beta-Convergence among statistical regions (NUTS 2 classification) 5.2.1 Summary statistics

Once the Eurozone countries are divided into smaller statistical units based on the second level of NUTS classification, the summary statistics are presented in the Table 6. As the data for NUTS2 units had to be obtained from different database, the estimates in the Table 6 are denoted in euro quantities unlikely the per-country data. Another difference concerns the variations in time periods, for which the regressions were made. These variations appeared due to the different starting point for reporting data. For more information, the full list of countries and their respective units can be found in Appendix I.

The statistical results are vastly similar to the country-unit observations in terms of qualitative outcome, however, different in scale. Firstly, in the Eurozone 19 classification, Luxembourg regional unit had the highest GDP per capita level in 2000, which was approximately 17 times higher than the Východné Slovensko unit. The average GDP per capita growth rates are now lower than the ones found in the dataset formed from country units, however, the trends stay the same. To be more precise, the Northern geographic region had on average on the lowest GDP per capita in 2000, and showed the highest rate of economic growth, around 4.57%. On the other hand, the Western geographic region, with the biggest average GDP per capita in 2003, had the smallest growth rate, around 1.89%. Finally, the statistical units of late entrants of the Eurozone, EZ8, once more possess lower initial starting point but higher growth rate compared to the statistical regions of early entrant countries.

1

(22)

22

5.2.2 Results from the whole Eurozone

This subsection presents the results obtained by regressing the statistical regions of all Eurozone countries in the period from 2000 to 2016 based on NUTS2 division. Out of 176 statistical units 157 were used for the research. The remaining 19 observations were dropped as the starting date of reporting the figures were later then 2000. This resulted, two Slovenian, two Italian, two French, two German and all 11 Belgian regions being disregarded. In the Table 7 the regression estimates can be found, whereas the Figure 6 is the graphical representation of the relationship between the initial GDP per capita rate in 2000 and the average growth rate from 2000 to 2016.

Both regressions showed that poor Eurozone regions, with low starting incomes, growth at a bigger rate and chase wealthy regions. Moreover, both betas are significant with significance level 1%. Therefore, there is a convergence of GDP per capita levels among regions of the Eurozone. Simultaneously, they are en route to the same income per capita level and depend to the same convergence club. The fit of the model, however, is reduced when compared to the country-unit estimation, although, the convergence persists.

5.2.3 Results when grouped to geographic regions

Before analyzing the results, a few remarks should, be made. This time Table 8 additionally contains the exact time periods for individual geographic region. The reason for these variations arose due to unavailable data either on GDP or population. Four, none, none and two NUTS2 observations were dropped from the Western, Northern, Eastern and Southern geographical regions.

In this regression different results were found compared to the country-based classification. Significant (with 1%) convergence was noted amongst the statistical units of Northern Eurozone. The negative relationship between the starting GDP per capita and its average growth rate was recorded for the Eastern and Southern geographic regions but beta-convergence is not assumed because of insignificant coefficients. Moreover, after dividing the country units(Section 5.1) to NUTS2 statistical regions, the model fit decreased in all regressions, except for the Northern region. The fit of Northern region stayed approximately the same and significant income per capita catch-up of initially poor regions was seen compared to rich ones. The graphical representation of the relationship between the initial

(23)

23

GDP per capita and average GDP per capita growth rate for different geographic regions of the Eurozone can are presented in the Table 8.

The Western region can be a great area of interest, because of the significant convergence when accounted for the population growth in the country-units sample and insignificant divergence when accounted for the population growth amongst its regional units. Such trade-off is called a ‘diverging convergence’ and was first mentioned by (Battisti and Di Vaio, 2007).

5.2.4 Results when grouped according to the time of joining the Euro area

The grouping, according to the entry time to the Eurozone, produced similar results as in country classification. High initial income per capita levels are negatively related to income per capita growth but relationship is insignificant. Opposite results are seen for the regions of 8 later entrants (see Table 9 and Figures 8&9). This is due to the fact that 11 early entrants (and their regions) have satisfied the convergence criteria earlier and converged more in the past. Their rate of convergence is now limited as they are close to the steady state. The average GDP per capita growth rate in the period 2000-2016 of early entrants is 2.23% and is indeed lower than the one of the late entrants, 3.37% (Table 6). This again justifies the Hypothesis 2, where countries that had earlier satisfied the Maastricht convergence criteria are closer to the final convergence value and, thus, grow at lower rates.

6. Conclusion

In conclusion, the research done on 19 Eurozone countries as well as its regions based on NUTS2 showed that some of the criticism towards the convergence in the Eurozone and its potential to become an optimal currency area can be dispersed. Under both classifications, countries and statistical units, that had lower initial income per capita levels, possessed higher growth rates than their wealthier counterparts. Similarly, significant estimates, showing beta-convergence, were obtained.

The Northern geographic region was found to have the highest average growth rates but it also showed substantial catch-up amongst its countries and NUTS2 regions. The Western countries seemed to converge after controlling for average population growth. Their GDP per capita growth rate was found to have a relatively high positive correlation with the average population growth rate. This might be attributed to the inflow of professional, skilled

(24)

24

labour force incoming from other countries, also known as the ‘brain drain’. When Western statistical regions were tested for economic convergence, slight divergence was found. For Eastern and Southern countries, regression estimates were would to be insignificant.

11 countries (and their statistical regions) that joined the Eurozone in 1991 did not show significant evidence that poorer countries (and corresponding regions) were

approaching same income per capita value. This was not the case for the 8 countries that joined the Eurozone anytime after the 1999. When average growth rates of early and late entrants were compared, a proof was found that 11 early entrants are converged more than the 8 late ones.

This research is subject to a number of limitations. First of all, as mentioned in the Section 4 difference between absolute and conditional beta-convergence is not distinguished. Some data, regarding GDP or population estimates, was unavailable until 2003 or later. As a result, for some regressions the time period was only 8 years. Moreover, the sample size for estimations on countries and geographic regions was relatively small. Finally, the previous literature does not clearly define if negative beta is enough to claim convergence or does it necessarily have to be significant.

The future research could look into the ‘diverging-convergence’ case of the Western geographic region. Current economic literature does not contain much information on this topic. This is the case for both qualitative and quantitative analyses. Additionally, regional rate of economic convergence/divergence could be examined and check to what extent it influences the country-wise movements.

7. Bibliography

Barro, R. J., & Sala-i-Martin, X. (1992). Convergence. Journal of Political Economy, 100(2), pp. 223- 251.

Battisti, M., & De Vaio, G. (2008). A spatially filtered mixture of β-convergence regressions for EU regions, 1980–2002. Empirical Economics, 34(1), pp. 105-121.

Baumol, W. J. (1986). Productivity growth, convergence, and welfare: what the long-run data show. The American Economic Review, pp. 1072-1085.

Beine, M., Docquier, F., & Rapoport, H. (2001). Brain drain and economic growth: theory and evidence. Journal of development economics, 64(1), pp. 275-289.

(25)

25

Bibow, J. (2007). How the Maastricht regime fosters divergence as well as instability. In Aspects of Modern Monetary and Macroeconomic Policies (pp. 197-222). Palgrave Macmillan, London. Canova, F. (2004).

Borjas, G. J. (1989). Economic theory and international migration. International migration review, pp. 457-485.

De Grauwe, P. (1996). Monetary union and convergence economics. European Economic Review, 40(3-5), pp. 1091-1101.

De Grauwe, P. (2012). A fragile Eurozone in Search of a Better Governance. The Economic and Social Review, 43(1, Spring), pp. 1-30.

De Long, J. B. (1988). Productivity growth, convergence, and welfare: comment. The American Economic Review, 78(5), pp. 1138-1154.

Directorate-General for Communication (European Commission). (2018, January 26). The European Union: what it is and what it does. Retrieved from https://publications.europa.eu/en/publication-detail/-/publication/715cfcc8-fa70-11e7-b8f5-01aa75ed71a1/language-en

European Central Bank. (2014). Progress through crisis? Proceedings of the conference for the 20th anniversary of the establishment of the EMI. Retrieved from https://www.ecb.europa.eu/ecb/history/emu/html/index.en.html

European Union. (1992, February 7). Treaty on European Union (Consolidated Version), Treaty of Maastricht. Retrieved from http://www.refworld.org/docid/3ae6b39218.html Favell, A. (2008). The new face of East–West migration in Europe. Journal of Ethnic and

migration studies, 34(5), pp. 701-716.

Frankel, J. A., & Rose, A. K. (1998). The endogeneity of the optimum currency area criteria. The Economic Journal, 108(449), pp. 1009-1025.

Inada, K. I. (1963). On a two-sector model of economic growth: Comments and a generalization. The Review of Economic Studies, 30(2), pp. 119-127.

Islam, N. (1995). Growth empirics: a panel data approach. The Quarterly Journal of Economics, 110(4), pp. 1127-1170.

Islam, N. (2003). What have we learnt from the convergence debate? Journal of economic Surveys, 17(3), pp. 309-362.

Maddison, A. (1982). Phases of Capitalist Development. Oxford University Press, USA. Mankiw, N. G., Romer, D., & Weil, D. N. (1992). A contribution to the empirics of economic

(26)

26

Marelli, E. (2007). Specialization and convergence of European regions. The European Journal of Comparative Economics, 4(2), pp. 149-178.

Quah, D. T. (1996a). Empirics for economic growth and convergence. European Economic Review, 40(6), pp. 1353-1375.

Quah, D. T. (1996b). Regional convergence clusters across Europe. European economic review, 40(3-5), pp. 951-958.

Quah, D. T. (1996c). Twin peaks: growth and convergence in models of distribution dynamics. The Economic Journal, pp. 1045-1055.

Sala-i-Martin, X. X. (1996a). Regional cohesion: evidence and theories of regional growth and convergence. European Economic Review, 40(6), pp. 1325-1352.

Sala-i-Martin, X. X. (1996b). The classical approach to convergence analysis. The Economic Journal, pp. 1019-1036.

Solow, R. M. (1956). A contribution to the theory of economic growth. The Quarterly Journal of Economics, 70(1), pp. 65-94.

(27)

APPENDIX I

The list of statistical regions according to the Nomenclature for Territorial Units for Statistics level 2 (NUTS2) classification.

i

AUSTRIA : Prov. Namur Pays de la Loire Oberbayern Münster Sterea Ellada Marche Utrecht Zahodna Slovenija

Burgenland CYPRUS Bretagne Niederbayern Detmold Peloponnisos Lazio Noord-Holland SPAIN :

Niederösterreich ESTONIA Poitou-Charentes Oberpfalz Arnsberg Attiki Abruzzo Zuid-Holland Galicia

Wien FINLAND : Aquitaine Oberfranken Koblenz Voreio Aigaio Molise Zeeland Principado de Asturias

Kärnten Länsi-Suomi Midi-Pyrénées Mittelfranken Trier Notio Aigaio Campania Noord-Brabant Cantabria Steiermark Helsinki-Uusimaa Limousin Unterfranken Rheinhessen-Pfalz Kriti Puglia Limburg (NL) País Vasco Oberösterreich Etelä-Suomi Rhône-Alpes Schwaben Saarland IRELAND : Basilicata PORTUGAL : Comunidad Foral de _Navarra Salzburg Pohjois- ja Itä-Suomi Auvergne Berlin Dresden Border, Midland &

Western Calabria Norte La Rioja

Tirol Åland Languedoc-Roussillon Brandenburg Chemnitz Southern & Eastern Sicilia Algarve Aragón

Vorarlberg FRANCE : Provence-Alpes-Côte _d'Azur Bremen Leipzig ITALY : Sardegna Centro (PT) Comunidad de Madrid BELGIUM : Île de France Corse Hamburg Sachsen-Anhalt Piemonte LATVIA Area Metropolitana _deLisboa Castilla y León Bruxelles-Cap. /

Brussel Hfdst. Champagne-Ardenne Guadeloupe Darmstadt Schleswig-Holstein

Valle d’Aosta /

Vallée d’Aoste LITHUANIA Alentejo Castilla-La Mancha

Prov. Antwerpen Picardie Martinique Gießen Thüringen Liguria LUXEMBOURG

(GRAND-DUCHY)

Região Autónoma dos

Açores Extremadura Prov. Limburg (BE) Haute-Normandie Guyane Kassel GREECE : Lombardia MALTA Região Autónoma da _Madeira Cataluña Prov.

Oost-Vlaanderen Centre La Réunion

Mecklenburg-Vorpommern

Anatoliki Makedonia, Thraki

Provincia Autonoma

di Bolzano / Bozen NETHERLANDS : SLOVAKIA : Comunidad Valenciana Prov. Vlaams Brabant Basse-Normandie Mayotte Braunschweig Kentriki Makedonia Provincia Autonoma

di Trento Groningen Bratislavský kraj Illes Balears Prov.

West-Vlaanderen Bourgogne GERMANY : Hannover Dytiki Makedonia Veneto Friesland (NL) Západné Slovensko Andalucía Prov. Brabant Wallon Nord-Pas-de-Calais Stuttgart Lüneburg Ipeiros Friuli-Venezia

Giulia Drenthe Stredné Slovensko Región de Murcia Prov. Hainaut Lorraine Karlsruhe Weser-Ems Thessalia Emilia-Romagna Overijssel Východné Slovensko Ciudad Autónoma de

Ceuta

Prov. Liège Alsace Freiburg Düsseldorf Ionia Nisia Toscana Gelderland SLOVENIA : Ciudad Autónoma de _Melilla Prov. Luxembourg

(28)

APPENDIX II

Tables and Figures

Table 1

Summary statistics on the average GDP per capita level in 1995 and the average GDP per capita growth rate from 1995 to 2016 (in US$)

Summary statistics Average GDP per capita 1995 ($) Average GDP per capita growth rate 1995-2016 Min. Max. EZ19 18570.69 0.037201 2168.796 52831.25 12998.83 0.025962 Western 31222.4 0.02491 19181.4 52831.25 10352.3 0.01546 Northern 8453.979 0.071329 2168.796 26273.47 11885.75 0.031776 Eastern 7744.908 0.046247 4799.151 10690.67 4165.93 0.017882 Southern 14163.42 0.025774 9114.358 20596.39 3910.81 0.011711 EZ11 26603.56 0.024374 11782.52 52831.25 10830.81 0.012118 EZ8 7525.494 0.54838 2168.796 15098.01 5107.704 0.030155

(29)

Table 2

Regression estimates from the sample of 19 Eurozone countries. _{Linear regression analysis for Eurozone19}

Reg* Reg** β -0.02417 -0.02440 0.00318 0.00296 R^2 0.7508 0.7556 Number of observations 19 19

*Linear regression on GDP per capita 1995

**Linear regression on GDP per capita 1995 and population growth rate 1995-2016

(30)

Table 3

Regression estimates from the samples of four Eurozone geographic regions.

Linear regression analysis for Geographical Euro area regions

Western Northern Eastern Southern

Reg* Reg** Reg* Reg** Reg* Reg** Reg* Reg**

β -0.0194 -0.03198 -0.0267 -0.0303 -0.0316 -0.0316 -0.0306 -0.0312 0.0316 0.00664 -0.0008 -0.0017 Error is not given 0.0155 0.0182

R^2 0.1416 0.8778 0.9946 0.9974 0.5215 0.5757

Number of

observations 7 7 4 4 2 2 6 6

*Linear regression on GDP per capita 1995.

(31)

Table 4

Correlation between the average population growth and growth of GDP per capita between 1995-2016.

Correlation

EZ19 Western Northern Eastern Southern

(32)

Table 5

Regression estimates from the samples depending on the time of entering the Eurozone.

Linear regression analysis for different entry time

Eurozone11 Eurozone8

Reg* Reg** Reg* Reg**

β -0.0062 -0.0163 -0.0373 0.0461

0.0085 0.0075 0.0027 0.0053

R^2 0.0415 0.7032 0.9612 0.9765

Number of

observations 11 11 8 8

*Linear regression on GDP per capita 1995

**Linear regression on GDP per capita 1995 and population growth rate 1995-2016

(33)

Table 6

Summary statistics on the average GDP per capita level and the average GDP per capita growth rate (in US$)

Summary statistics, using NUTS2 units

Observations Average initial GDP per capita (€) Average GDP per capita growth rate per period Min. Max. EZ19 157 20505.72 0.0239 3119.22 53226.48 2000-2016 8120.01 0.0141 Western 96 26343.10 0.0189 13696.48 58416.24 2003-2016 7608.78 0.0059 Northern 8 18422.53 0.0457 3556.59 36414.85 2000-2016 13037.07 0.0291 Eastern 6 16311.14 0.0086 8539.77 29253.20 2008-2016 8176.79 0.0051 Southern 60 16173.03 0.0180 9750.94 30066.32 2000-2016 5478.98 0.0080 EZ11 135 22260.78 0.0223 10064.87 53226.48 2000-2016 7188.92 0.0063 EZ8 22 9736.02 0.0337 3119.22 16768.84 2000-2016 4326.34 0.0334

(34)

Table 7

Regression estimates from the sample of 19 Eurozone countries.

Linear regression analysis for Eurozone19, using NUTS2 units, 2000-2016

Reg* Reg**

β -0.0150 -0.0142

0.0039 0.0038

R^2 0.2595 0.2685

Number of observations 157 157 *Linear regression on GDP per capita

(35)

Table 8

Regression estimates from the samples of four Eurozone geographic regions.

Linear regression analysis for Geographical grouping, using NUTS2 units

Western Northern Eastern Southern

Reg* Reg** Reg* Reg** Reg* Reg** Reg* Reg**

β 0.0009 0.0029 -0.0281 -0.0386 -0.0035 -0.0081 -0.0027 -0.0046 0.0034 0.0032 0.0018 0.0024 0.0057 0.0051 0.0030 0.0031 R^2 0.0010 0.0839 0.9736 0.9937 0.1077 0.2004 0.0116 0.0524 Number of observations 96 96 8 8 6 6 60 60

*Linear regression on GDP per capita

(36)

Table 9

Regression estimates from the samples depending on the time of entering the Eurozone.

Linear regression analysis for different entry time, using NUTS2 units

Eurozone11 Eurozone8

Reg* Reg** Reg* Reg**

β -0.0004 -0.0002 -0.0509 -0.0552

0.0018 0.0018 0.0038 0.0065

R^2 0.0005 0.0180 0.7362 0.7476

Number of

observations 135 135 22 22 *Linear regression on GDP per capita

(37)

Figure 1

Graphical representation of GDP per capita movement across Eurozone countries from 1995 to 2016 (in U.S. $)

0 20000 40000 60000 80000 100000 120000 140000 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016

GDP per capita estimates for 19 Eurozone countries from 1995 to 2016 (in US$)

Austria Belgium Cyprus Germany Spain

Estonia Finland France Greece Ireland

Italy Lithuania Luxembourg Latvia Malta

(38)

Figure 2

Relationship between the logarithmic value of initial GDP per capita level and the average growth rate in the period from 1995 to 2016 for 19 Eurozone countries. Austria Belgium Cyprus Germany Spain Estonia Finland France Greece Ireland Italy Lithuania Luxembourg Latvia Malta The Netherlands Portugal Slovak Republic Slovenia 0 0,01 0,02 0,03 0,04 0,05 0,06 0,07 0,08 0,09 0,1 7 8 9 10 11 GD P per c ap ita g ro w th ra te, 1 99 5-2016

Natural logarithm of GDP per capita in 1995

19 Eurozone countries

(39)

Figure 3

The graphs plot the relationship between the logarithmic value of initial GDP per capita level and the average growth rate in the period from 1995 to 2016 for different geographic regions of the Eurozone.

Austria Belgium Germany France Ireland Luxembourg The Netherlands 0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 9.5 10.5 11.5 GDP p e r cap ita gr o wt h r ate , 1995 -2016

Western region of the Eurozone

n = 7 Slovak Republic Slovenia 0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 8 9 10 GDP p e r cap ita gr o wt h r ate , 1995 -2016

Eastern region of the Eurozone

n = 2 Estonia Finland Lithuania Latvia 0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1 7 8 9 10 GDP p e r cap ita gr o wt h r ate , 1995 -2016

Northern region of the Eurozone

n = 4 Cyprus Spain Greece Italy Malta Portugal 0 0.01 0.02 0.03 0.04 0.05 0.06 8.5 9.5 10.5 GDP p e r cap ita gr o wt h r ate , 1995 -2016

Southern region of the Eurozone

(40)

Figure 4

Relationship between the logarithmic value of initial GDP per capita level and the average growth rate in the period from 1995 to 2016 for Eurozone11. Austria Belgium Germany Spain Finland France Ireland Italy Luxembourg The Netherlands Portugal 0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 9 10 11 GDP p e r cap ita gr o wt h r ate , 1995 -2016

11 Early entrants to the Eurozone

(41)

Figure 5

Relationship between the logarithmic value of initial GDP per capita level and the average growth rate in the period from 1995 to 2016 for Eurozone8. Cyprus Estonia Greece Lithuania Latvia Malta Slovak Republic Slovenia 0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1 7 8 9 10 GDP p e r cap ita gr o wt h r ate , 1995 -2016

8 Late entrants to the Eurozone

(42)

Figure 6

Relationship between the logarithmic value of initial GDP per capita level and the average growth rate in the period from 2000 to 2016 for the regions across the whole Eurozone, using NUTS2.

0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1 8 9 10 11 GDP p e r cap ita gr o wt h r ate , 2000 -2016

Regions from 19 Eurozone countries

(43)

Figure 7

The graphs plot the relationship between the logarithmic value of initial GDP per capita level and the average growth rate in the period from 2000 to 2016 for different geographic regions of the Eurozone, using NUTS2.

0 0.005 0.01 0.015 0.02 0.025 0.03 0.035 0.04 0.045 9 10 11 GDP p e r cap ita gr o wt h r ate , 2003 -2016

Western region of the Eurozone

n = 96 0 0.005 0.01 0.015 0.02 0.025 0.03 9 10 GDP p e r cap ita gr o wt h r ate , 2008 -2016

Eastern region of the Eurozone

n = 6 0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 8 9 10 GDP p e r cap ita gr o wt h r ate , 2000 -2016

Northern region of the Eurozone

n = 8 0 0.005 0.01 0.015 0.02 0.025 0.03 0.035 0.04 0.045 8.5 9.5 10.5 GDP p e r cap ita gr o wt h r ate , 2000 -2016

Evidence on the convergence theory among the member state of the Eurozone from 1995 to 2016 : do countries and regions in the Eurozone converge to the same GDP per capita level?