The impact of population ageing on economic growth in the Visegrad group countries

(1)

The Impact of Population Ageing on Economic Growth

in the Visegrad Group Countries

Master Thesis

Author: Alexandra Hezselyova Thesis supervisor:

Stud. Nr.:11416572 Drs. Naomi J. Leefmans

Email: alexandra.hezselyova@student.uva.nl Second reader:

Study programme: MSc Economics Dr. Dirk J.M. Veestraeten

Track: International Economics and Globalization Number of words: 14901

(2)

Statement of Originality

This document is written by Alexandra Hezselyova 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 of the work, not for the contents.

(3)

1. Introduction

During the last three decades, the Central European population has significantly begun to age. Since 1960 the share of the population over 65 years doubled and it is predicted to rise even more in the upcoming years (World Bank, 2018). This thesis focuses on four Central European countries, namely the Czech Republic, Hungary, Poland and the Slovak Republic. These countries belong to the so-called Visegrad group or ‘V4’. The group was formed in 1991 to successfully finish social transformation for accomplishing a market economy and to prepare for the European integration process. Its main goal is to strengthen the stability through cooperation and integration (Visegrad Group, 2018). According to scholars, the median age of the population is increasing at a faster rate in Central and Eastern European countries than in EU-15. It is mainly due to lower levels of the fertility rate and outward migration of young people (Sinnott and Koettl-Brodmann, 2015). For instance, the Slovak Republic is ranked today among the youngest countries in the European Union and it will become the second oldest nation in the EU in the middle of the 21st_{century (Infostat, 2014).}

This demographic change might have a huge effect on a country’s economic growth. Therefore, it is essential to study the impact of the population ageing on economic growth in this region. Moreover, in 1960 the fertility rate in Poland and in the Slovak Republic was around 3 children per woman and in the Czech Republic and Hungary it was around 2 children per woman (see Figure 1). However, since 2000, the fertility rate has been below EU average in all four countries and it has been lower than the replacement fertility level of 2.1 children per woman. If this trend persists, the population is expected to keep on ageing and shrinking in these countries.

Figure 1: Fertility rate in the V4 countries (1960-2015)

Source: Self-elaborated with data from World Bank (2018) 1 1.5 2 2.5 3 3.5 1960 1970 1980 1990 2000 2010 2015

(6)

6

Furthermore, the share of old and young dependency ratios and the share of the working-age population are often used in empirical studies as ageing proxies. The old dependency and the young dependency ratios are the ratios of people older than 64 years, respectively younger than 15 years, to the working age population (population between 15 and 64) (World Bank, 2018). According to Bloom, Canning and Finlay (2010); and Bloom, Canning and Fink (2010) a higher share of the old and young dependency ratios has a negative impact on economic growth as these categories of people are considered to be a dependent part of the population. In contrast, a larger share of the working-age population (productive part) increases economic growth. In the V4, the share of the working age population was increasing during the period 1960-2010. However, since 2010 this ratio has been declining. Thus, there are less people that are able to work which may negatively influence the countries’ economic performance (see Appendix A). On the other hand, the old dependency ratio has increased on average by 13% since 1960 (see Figure 2). In the upcoming 34 years, this ratio is expected to almost triple in Poland and the Slovak Republic, whereas in Hungary and the Czech Republic it will double. Therefore, the ageing will accelerate in Central Europe in the next decades. Moreover, the young dependency ratio has decreased since 1960 by more than 30% in Poland and the Slovak Republic and by 17% in the Czech Republic and Hungary.

Figure 2: Old and Young dependency ratios in the V4 countries

Source: Self-elaborated with data from United Nations (2018) 0 10.0 20.0 30.0 40.0 50.0 60.0 70.0 1960 1970 1980 1990 2000 2010 2015 2020 2030 2040 2050

CZE - old dependency ratio CZE - young dependency ratio HUN - old dependency ratio HUN - young dependency ratio POL - old dependency ratio POL - young dependency ratio SVK - old dependency ratio SVK - young dependency ratio

(7)

7

The aim of this thesis is to estimate the impact of the above-mentioned ageing on economic growth in the selected countries and to make a projection from 2017 until 2050. Hence, the research question for this study is: “How has the ageing population influenced economic growth in the Visegrad group countries since 1991 and what impact will ageing have on economic growth in the upcoming 34 years?”

Several studies have been conducted to explore the impact of population ageing on economic performance in developed countries. However, the impact of ageing on economic growth and the projection for the upcoming decades have not yet been examined within the V4 countries. That is, indeed, the main contribution of this thesis.

Firstly, panel data analysis is applied to capture the impact of ageing on economic growth. Bloom, Canning and Finlay (2010) estimated the effect of ageing on economic growth in Asian countries. Their regression analysis found a negative effect of both old and young dependency ratios on economic growth. Bloom and Finlay (2009) measured the effect of demographic changes on economic growth in Asia by constructing a panel for the period 1960-2005 and using 5-year averages of country level data. They found a positive relation between the share of the working-age population, life expectancy and economic growth. In this paper, a similar methodology is applied by using 3 ageing proxies. In one regression, old and young dependency ratios is taken into consideration as ageing proxies, whilst the share of the working-age population is used in the second regression. Different ageing proxies help us identify the impact that ageing has had and will have on economic growth in the selected countries. In addition, this thesis includes some other additional explanatory variables such as: gross domestic investments, foreign direct investments and the government expenditures on R&D, that were not used in the paper of Bloom and Finlay (2009).

A panel using annual data from 1991 to 2016 is constructed. No earlier data are available for the selected economies. This is explained by the fact that these countries were part of the Eastern block and statistics were measured differently or often not published back then. Moreover, an additional 16 countries from the same region (Central, Eastern and South-eastern Europe) are included, in order to increase the variation in the data, which is fundamental to find robust results. The dependent variable is economic growth and the explanatory variables are: the level of income (measured as real GDP per capita in constant 2010 US$), technology (measured as government expenditures on R&D), the ratio of domestic investments to GDP, ratio of foreign direct investment to GDP, trade openness, population growth, political instability, human capital (expressed as the average years of

(8)

8

secondary education), old dependency ratio, young dependency ratio, the share of the working-age population, and life expectancy.

Secondly, the coefficient of the working-age share is used together with the actual demographic changes to calculate the impact of the population ageing on economic growth in each of the V4 countries. After that, the same coefficient is combined with the projections for future demographic changes to estimate the impact of ageing on economic growth in the upcoming 34 years.

The structure of this paper is as follows. Chapter 2 presents the literature review analysing existing research papers. The aim of this chapter is to provide sufficient theoretical and methodological background. Chapter 3 presents and explains the variables, which are used for the regressions. Chapter 4 describes the methodologies used. Subsequently, chapter 5 displays the results obtained from the panel data regressions. Lastly, chapter 6 summarises this master thesis and provides a conclusion.

(9)

9

2. Literature Review

This chapter provides a discussion about the existing literature regarding the effect of population ageing on economic growth. In the first part, section 2.1, the general relation between ageing and economic growth is reviewed. It elaborates the theoretical channels through which ageing has an impact on economic growth. In the second part, section 2.2, the population ageing in the V4 countries and also in other regions is examined. Additionally, the main proxies of ageing are discussed to explain why they are used in the panel data analysis. The last part, section 2.3, examines other determinants of economic growth, that are important to know as they are included in the growth regressions.

The theoretical channels from ageing to economic

growth

This subsection provides a theoretical background in order to understand the effect of ageing on economies’ performance. It examines the theoretical channels through which ageing affects economic growth.

Recent literature described the theoretical channel through which ageing may have an impact on economic growth. Bloom, Canning and Fink (2010) defined the population age groups and their different behaviours and needs. Older generations usually work and save less than younger (working-age) generations. Thus, when the old dependency ratio is increasing, the economies have fewer capital and labour available. Increasing older population also requires bigger investments in healthcare, that may place the burden on working populations as they pay the taxes. On the other hand, people tend to live longer and have healthier lives. Therefore, they can work longer and save more for their retirement. This may have a positive effect on economy. Kögel (2005) argued that a high young dependency ratio reduces aggregate savings. Lower aggregate savings have a negative impact on investments in research and development and, therefore, reduce total factor productivity growth. Furthermore, the paper of Cruz and Ahmed (2018) states that a higher share of the working-age population increases the savings in the short-run. In addition, it also increases investments in human capital in the long-run, therefore, it has a positive impact on the economy and

(10)

10

economic growth. This is caused by a higher share of the working-age population which increases the production and it leads to more available resources in the economy since there are more people that are able to work.

According to Maestas, Mullen and Powell (2016) 1/3 of the total effect of ageing on economic growth is derived by slower labour force growth and 2/3 is caused by lower productivity growth (measured as the 10-year growth rate in GDP per worker).

Empirical studies that investigate the impact of

ageing

The first part, subsection 2.2.1, presents the descriptive studies that analysed the ageing in the V4 countries. After that, the second part, subsection 2.2.2, discusses other empirical studies that concentrate on ageing in other regions. It also examines ageing proxies that have been used in the literature.

2.2.1 Ageing in V4 countries

For what is available in the literature there are no empirical papers that concentrate specifically on ageing in the Czech Republic, Hungary, Poland and the Slovak Republic. However, there are some descriptive studies that analysed ageing in Central and Eastern Europe including our 4 countries of interest. These are presented hereafter.

Durasiewicz (2015) included V4 countries in the so-called “graying society” that is characterized by an increasing share of elderly. Hence, in these countries there is not a sufficient number of young people that will be able to replace the current working-age population when these people retire in the future. This article predicts that a higher share of the elderly population will have a direct negative impact on public finance since these economies will need to cope with a higher healthcare spending. Moreover, negative net migration is another challenge that these countries have to deal with. Due to an unfavourable economic situation and worse quality of education many young people are moving abroad. Although migration is a very relevant factor, it is not taken into consideration in this master thesis because of data limitations.

According to Chawla, Betcherma and Banerji (2007), ageing in Central and Eastern European countries (including V4 countries) has different characteristics than ageing in

(11)

11

Western Europe and/or in Asia, therefore it deserves special attention. These countries went through political transition at the beginning of 1990s. Nowadays, they have to face another transition that is called the “third transition”. It is characterized by fast ageing and shrinking populations. The authors expected that, in 2050, every fifth person in this region will be older than 65 years. Hence, these countries will have to invest more into health care systems and also assure pensions for an increasing number of elderly. This argument is in line with Durasiewicz (2015). On the other hand, a decreasing population will allow to save more on education as countries will provide education to a reduced number of young people. This paper also explains that population ageing affects economic growth through 3 main channels in these countries: labour input and productivity, consumption and saving and financial markets. In order to eliminate the negative impact of ageing on economic growth, countries will have to adjust their policies and also undertake many reforms (e.g. pension and healthcare reforms).

2.2.2 Ageing in other regions

There are many empirical studies that analysed the relationship between ageing and economic growth and most of them concentrated on Asia or Western Europe. All the papers mentioned here, considered economic growth to be the dependent variable. As indicated earlier, the methodology used in this thesis is based on Bloom and Finlay (2009). Their results were estimated by using ordinary least-squares (OLS) and instrumental variable (IV). The authors studied the role of demographic change on economic growth in Asia by constructing a panel for the period 1960-2005 and using 5-year averages of country level data. As demographic variables they used the share of the working-age population, life expectancy, population growth and population density. The authors found a positive and significant effect of the changes in the working-age share on GDP per capita growth. They argued that a higher share of working-age populations together with a lower fertility rate increases income per capita in a country, since output per worker stays unchanged but the share of the young dependency ratio is lower. Similarly, a positive and significant effect of life expectancy on economic growth was found. The explanation is that people live longer and have healthier lives, therefore they can also work longer, which allows them to save more.

The paper written by Cuaresma, Lábaj and Pružinský (2014) investigated the impact of ageing on economic growth in 23 European Union countries (including the Czech Republic and Poland) between 1970-2010 by using ordinary least-squares (OLS) and country

(12)

fixed-12

effects. The panel dataset was defined on 5-year averages of country level data. The old age dependency ratio and the prospective old age dependency ratio were used as ageing proxies. A negative coefficient was found for both. The prospective ageing indicator was used as income growth determinant and it was defined as “the ratio of the population whose age is such that the remaining life expectancy is 15 years or less (the old-age threshold) to the number of people of age 20 to that old-age threshold” (Cuaresma, Lábaj and Pružinský, 2014). They observed that the impact is different across countries and the strongest negative effect of ageing was discovered in poorer countries. Specifically, a significant difference between Western and Eastern Europe in the prospective ageing was identified. This can be explained by the different levels of life expectancy in these two regions. They also showed that for Eastern European countries, the ageing poses a challenge to sustainable income growth, due to their lower level of income per capita.

Furthermore, Bloom, Canning and Finlay (2010); and Cruz and Ahmed (2018) used old and young dependency ratios and the share of the working-age population as ageing proxies (the same proxies are used in this paper).

Cruz and Ahmed (2018) used three different methods (first-difference, panel fixed-effects, and generalized method of moments (GMM)) in order to estimate the relation between ageing and economic growth for the period 1950-2010. They showed that a higher share of the working-age population and a lower child dependency ratio are associated with a rise in GDP growth. For the old dependency ratio, the results were insignificant. Bloom, Canning and Finlay (2010) constructed an analysis for Asian countries for the period 1960-2005 by using OLS and IV regressions. They found a significant and negative coefficient for both ratios. Hence, an increase in the share of old and young dependency ratios had a negative impact on the short run growth rate. Additionally, in the long-run they also found negative but insignificant effect of the share of old-dependency ratio and a negative significant effect of the share of young-dependency ratio. The authors concluded that a rise in the share of the working-age population has a positive impact on income per capita since the rate of potential workers is higher.

On the other hand, Acemoglu and Restrepo (2017) analysed 169 countries for the period from 1990 to 2015 (similar to the time period used in this thesis) by using OLS regressions and IV estimations. The authors found a significant positive relationship between ageing and GDP per capita. As an ageing proxy, they utilised the change in the ratio of the population above 50 years to those between the ages of 20 and 49. Afterwards, the authors separately looked at 35

(13)

13

OECD counties and through the instrumental variable (IV) estimation they found again a positive and significant coefficient for ageing. They argued that the impact of technologies after 1990 can explain their results. Thanks to new technologies and artificial intelligence, firms have a wide range of opportunities to automate their production. They showed that adaptation of new technologies leads to a higher aggregate output, although the labour input is reduced. According to the authors, the countries with the rapid ageing process are the ones that adopt new technologies at the fastest rate.

Last but not least, to support the argument that a higher old dependency ratio may have a positive impact on economic growth, the empirical study of Bloom and Williamson (1998) deserves to be mentioned. They also found a positive coefficient for the old dependency ratio by using OLS and IV estimations. The authors explained that elderly do often work part-time and/or they take care of younger population. Thus, they do still positively contribute to a country’s economy.

To conclude, most empirical studies have determined a negative impact of the young dependency ratio and a positive effect of the share of working-age population on economic growth. The results for the old dependency ratio are diverse with different authors views. The most common methodologies used in the literature were OLS, country fixed-effects estimation and IV estimation therefore, these methods are also used in this thesis.

Other determinants of economic growth

When investigating the impact of ageing on economic growth, it is fundamental to control for all other variables affecting GDP growth. This subsection reviews the main control variables that have been used in the literature.

Multiple variables were used in empirical papers to control for other determinants of economic growth and one of them is human capital. Human capital was interpreted in many different forms. For instance, primary, secondary school enrolment rates and/or average years of primary and secondary schooling. Barro (1991) measured human capital in the form of primary and secondary school enrolment rates. He found that an increase of both school enrolment rates leads to a rise of economic growth. He also stated that economies with greater human capital usually have lower fertility rates and greater ratios of physical investment to

(14)

14

GDP. Barro (2000) estimated the effect of the average years of primary and secondary education on economic growth. For primary education, the results were insignificant. For the secondary education, he found a significant and positive effect. He argued that, secondary education has a positive impact on growth since it helps with technologies’ absorption. In addition, Sala-I-Martin (1997); Bloom and Williamson (1998); and Bloom and Finlay (2009) measured human capital in the form of the average years of secondary schooling. Using OLS, Bloom and Finlay (2009) found a positive and significant coefficient. In contrast, when the results were obtained by instrumental variable (IV) estimation, the coefficient was insignificant. Moreover, Bloom and Williamson (1998) also found a positive and significant coefficient when using OLS. However, in a paper of Sala-I-Martin (1997) results were not significant.

Moreover, Sala-I-Martin (1997); Barro and Lee (1994); and Cruz and Ahmed (2018) used the initial level of income (measured as log of GDP per capita) as a control variable for economic growth. They explained that including this variable in a regression is crucial, since it helps capture the concept of income convergence across countries. This concept states that lower values of starting GDP may trigger higher GDP growth. Barro and Lee (1994) also explained that an overestimated convergence effect can arise because of measurement error in GDP. This can be minimized by using a lagged value of the initial level of income. Lastly, a significant and negative correlation between the initial level of income and economic growth was found in these papers.

Trade openness is also often used as an explanatory variable for economic growth. In many papers it was measured as import plus export divided by GDP. Harrison (1996) in his cross-country analysis for developing countries tested the relation between openness and growth and found a positive and significant coefficient. The author argued that trade openness might not only be correlated with higher productivity growth but also with higher returns to human capital. Thus, it is expected that more educated people will benefit more from trade openness. Additionally, it is important to mention the causality issue that is associated with this variable. It is true that countries with greater trade openness are usually the ones with higher economic growth, but, higher growth rates may also lead to higher openness. Yanikkaya (2003) used panel analysis of more than 100 developed and developing countries for the period 1970–1997. He discussed the main channels (technology transfers, scale economies and comparative advantage) through which trade openness affects growth. He stated that openness can help a country gain access to advanced technologies. He tested

(15)

15

separately the impact of trade openness on growth in OECD countries and non-OECD countries. The results were very similar. For both, he found significant and positive coefficients. Hence, there are similar benefits of trading with developing countries as with developed countries. His conclusion stated that economies with a greater trade openness grow faster than other economies. This is in line with the empirical study of Bloom and Finlay (2009). They also found a positive and significant coefficient for the trade openness variable.

Barro and Lee (1994) tested the ratio of domestic investment to GDP in their regression analysis for 1965-1975 and 1975-1985. They found a significant and positive coefficient. In addition, the lagged value of the ratio of domestic investments to GDP was tested. The authors explained that using the lagged value of investments may lead to a lower investments coefficient if there is reverse causality between GDP growth and domestic investments. However, even when using a lagged value, a positive and significant coefficient was found. Adams (2009) analysed the effect of foreign direct investments and the domestic investments on economic growth by using both the OLS and the fixed effects estimation. He used a panel analysis for the period from 1990 to 2003. His results are consistent with Barro and Lee (1994). He found a significant and positive coefficient for domestic investments by using both OLS and also fixed effects. However, the results for foreign direct investments were not very clear. When using an OLS regression a positive and significant coefficient was found but this was not demonstrated when country-specific effects were added to the regression. Moreover, contemporaneous FDI had a negative and insignificant coefficient, whereas lagged FDI had a positive and insignificant coefficient. The importance of FDI was clearly explained by Kobrin (2005). He reported that a country can benefit from FDI by importing new technologies, knowhow, managerial skills and gaining access to new markets. It is also important to mention, that there is a higher chance that managerial and technological spillovers, that are considered to be the primary benefit of FDI, will occur more often in developed countries. It is due to the fact that the absorptive capacity of the more developed economy is higher. Moreover, Sala-I-Martin (1997) used public investment share (investment share as a fraction of GDP) in his paper. His results confirm previous studies. Therefore, countries with greater domestic investments are more likely to grow faster.

Another variable that is often included in the regressions as a determinant of economic growth is population growth. The empirical analyses showed different results for this control variable. In the paper of Sala-I-Martin (1997) a positive coefficient for population growth was found, although it was insignificant. On the other hand, Barro (1998) explained that high

(16)

16

population growth leads to higher investments that provide capital for new workers, rather than capital per worker. According to the author, it means that high population growth can have a negative impact on economic growth. When Bloom and Finlay (2009) introduced population growth in their regression, their result was insignificant and negative while using OLS regression. When they used IV estimation and they did control for potential reverse causality between population growth and economic growth their results become significant. Their results are in line with a paper of Barro and Lee (1994). The authors also obtained a negative and significant coefficient for population growth variable.

Barro and Lee (1994) measured political instability in the form of political rights and civil liberties. They argued that including the variables that measure political stability is crucial since the government may have an impact on the economy by changing individual rights e.g. freedom of the press. They simultaneously included these variables in a regression and found for both significant coefficients. More precisely, greater political freedom had a positive impact on growth, whereas more liberties had a negative impact on growth. In addition, Barro (1999) pointed out that civil liberties and political rights are usually highly correlated, therefore it may be a better idea to not include them together in one regression. Moreover, the results of Sala-I-Martin (1997) indicated that bigger political rights and civil liberties produce lower economic growth. For both, he found negative coefficients, even though they were not significant. In his regressions, Barro (1998) used political rights and the rule of law as measurements for political instability. He converted the measurements for political rights into a scale from 0 to 1 (0 representing the lowest political rights and 1 the highest). He showed that correlation between low levels of political rights and economic growth is positive, but negative once a certain level of political freedom is reached.

Recently, many studies have highlighted the importance of technologies and their impact on economic growth. Most papers used R&D expenditures as a proxy for technology. Bayarçelik and Taşel (2012) examined the correlation between the number of researchers employed in R&D departments, the number of patents, expenditures on R&D and economic growth. The coefficient of the R&D expenditures was positive and significant. He also stressed the importance of the number of researchers employed in R&D departments as one of the most essential indicators of innovation. They found a positive and significant relation between this variable and GDP growth. Moreover, Pece, Simona and Salisteanu (2015) analysed the relation between innovation and economic growth in the Czech Republic, Hungary and Poland for the time period from 2000 to 2013. R&D expenditures were used as a

(17)

17

proxy for technology. They found that a positive and significant effect of R&D expenditures on economic growth in these countries.

The last explanatory variable that is going to be discussed is life expectancy. In empirical papers, life expectancy is considered to be a determinant for health. The higher the values of life expectancy in a country, the longer and healthier lives people have. Barro and Lee (1994) found a positive and significant coefficient. They considered life expectancy to be not only an indicator of health but also an indicator of better work habits and a higher level of skills. Authors also showed that the coefficient for life expectancy has a higher magnitude for poorer countries. Bloom, Canning and Malaney (2000) summarized the channels through which life expectancy may have an impact on economic growth (e.g. higher savings and higher productivity). As people are expected to live longer they will also save more. Lastly, healthier population will be more productive. In their research, a positive and significant coefficient for life expectancy was found.

To conclude, most studies investigated the impact of population ageing on economic growth for the period starting around 1960 and finishing by the beginning of the 21th_century.

They found a significant and negative correlation between population ageing and economic growth. All studies associated the higher share of the working-age population with the higher economic growth and the higher share of the young dependency ratio with lower economic growth. Different results were obtained for the old dependency ratio. In this thesis the same ageing proxies are used. Moreover, following the literature review, another 9 explanatory variables are used as control variables, namely: the level of income (measured as real GDP per capita in constant 2010 US$), technology (measured as government expenditures on R&D), the ratio of domestic investments to GDP, ratio of foreign direct investment to GDP, trade openness, population growth, political instability, human capital (expressed as the average years of secondary schooling), and life expectancy.

(18)

18

3. Data description

This chapter discusses the variables included in the growth regressions of this thesis. Each variable is discussed as well as defined in one separate paragraph. Consistently with the literature review, the expected impact of the selected variable on economic growth is illustrated. Afterwards, a detailed description of data trend is provided. Our attention is drawn to significant peaks and drops. All data are retrieved from the World Bank database, except for the political rights data that were obtained from the Freedom House database. Lastly, this chapter provides a table that includes the descriptive statistics for the 20 Central and Eastern European countries observed in this research.

Variables and Data

Panel data of 20 Central, East and Southeast European countries (including our 4 countries of interest) is constructed for the time period 1991-2016. Because of the Yugoslav War and political instability in the Southeast European countries (Balkan countries), some data for these countries have been obtained only from 1996. Although Austria was not part of the Eastern Block and does not share a similar history with other countries, in this thesis it is also considered to be a part of Central Europe. Austria was added in order to further increase the number of observations and the variation in the data. A list of countries used in this master thesis can be found in Appendix K. A list of variables, their definitions and data sources are summarized in a table in Appendix M.

As mentioned earlier, economic growth is the dependent variable. It is measured as an annual percentage growth rate of GDP per capita based on the constant local currency (World Bank, 2018). It is difficult to obtain data for GDP growth for some countries, because in the earlier 90s they were going through a difficult regime transition and the data were not published regularly. In addition, some countries in the beginning of the 1990s were part of one country namely Yugoslavia and they became independent in 1996. Therefore, it is important to mention that we were not able to obtain for all countries data from 1991 but only from 1996.

(19)

19

The data exhibit a similar trend for economic growth for the selected countries (see Appendix C). At the beginning of the 90s all countries, except Austria, had negative economic growth. For example, in Albania the economic growth in 1991 was -29%, in Romania it was -12%, in the Czech Republic -11%, and in Poland -7% (see figure 3 for the V4 countries). This is not surprising since these countries were going through institutional and structural transition period. They were transforming from socialism to capitalism and from centrally planned to market economies. After that, the economies had a positive growth level of GDP per capita. There was a high peak for Bosnia and Hercegovina in 1996 and 1997, when the values reached 92%, respectively 35%. This country was experiencing a boom period since it was shortly after the end of Yugoslav War. After 1997, economic growth was fluctuating in all countries. Overall, it was slowly increasing up until 2009 when the crisis hit these economies. More into detail, a very large trough was visible in Estonia, Latvia and Ukraine where GDP growth was around -15%. Furthermore, from the data it can be seen that in 2016 all countries (except Russia and Belarus) had positive economic growth.

The first ageing proxy used in this thesis is the old dependency ratio. The old dependency ratio is defined as “the ratio of older dependents, people older than 64-to the working-age population-those aged 15-64” (World Bank, 2018). This variable is an important

Figure 3: Economic growth in the V4 countries

Source: Self-elaborated with data from World Bank (2018) -15 -10 -5 0 5 10 15

(20)

20

proxy for ageing since it measures the relation between the amount of old people (dependent population) and the amount of people that are working (productive population). Essentially, if the old dependency ratio is increasing, there are more people in pension ages compared to the amount of people that are able to work.

The old dependency ratio was used as ageing proxy in many empirical papers (Bloom, Canning and Finlay, 2010; Cuaresma, Lábaj and Pružinský, 2014; Acemoglu and Restrepo, 2017; and Bloom and Williamson, 1998). The last two papers explained that a higher ratio of elderly may positively contribute to economic growth. Retired people might help economies through small part-time jobs or through their savings. Additionally, even if older population cannot work, they can be replaced by new technologies. However, other papers found a negative effect of the old dependency ratio on economic growth. Thus, when the old dependency ratio increases, economic growth decreases. For instance, Durasiewicz (2015) pointed out that a higher number of elderly have a direct negative impact on public finance since economies have to deal with higher healthcare spending. Overall, the results from previous studies are ambiguous.

From the data, we can see a slight increase in the old dependency ratio for the period 1991-2010. However, during these years it did not significantly change in any of our countries. In the Czech Republic it increased from 19,5% to 22%, in Moldova from 13.3% to 13.8% (see Appendix D). From 2010, there is a significant upward trend in most countries. During the period 2010-2016 this ratio increased in Bulgaria from 26% to 31% and in the Czech Republic from 22% to 28%. This corresponds with the decline in the share of the working-age population that is also visible especially from 2010 (see discussion later in this chapter). According to the United Nations projections, the old dependency ratio is expected to have an upward trend also in the future for all countries. Hence, the future ageing level will be even higher. Even for the Republic of Moldova, that has not yet experienced high values of the old dependency ratio, a significant rise by more than 50% is expected in the upcoming 34 years.

The young dependency ratio is also used as an ageing proxy. The definition describes this ratio as “the ratio of younger dependents-people younger than 15-to the working-age population-those ages 15-64” (World Bank, 2018). Similarly, as the old dependency ratio, it shows the relation between the amount of dependent population and the amount of people that are working (productive population).

Many empirical papers estimated a negative effect of the young dependency ratio on economic growth (Bloom, Canning and Finlay, 2010; Cruz and Ahmed, 2018). The first

(21)

21

mentioned, explained that higher values of the young dependency ratio lower the aggregate labour supply and the productivity since there are more people that are unemployed than people that are employed. Consequently, a negative coefficient for the young dependency ratio is expected to be found.

In all 20 countries the young dependency ratio has been decreasing since 1991 (see Appendix E). Therefore, the highest values for this ratio have been reached at the beginning of period that was selected for the purpose of this thesis. In 1991, in Albania and Moldova it was 53%, respectively 43%. It is not surprising that the biggest change has been visible in these two countries where this ratio has declined by more than 50% during last 26 years. On the contrary, in Austria, Croatia and Slovenia the young dependency ratio was the lowest in 1991 and it reached the values of 24.9%, 28.5% respectively 29.6%. The United Nations projections show that this ratio will continue to decline also in the future. In most countries it will reach its lowest point around the years 2035 – 2040.

The share of the working-age population is chosen as the last ageing proxy in this thesis. According to the definition it is “the population between the ages 15 to 64 as a percentage of the total population” (World Bank, 2018).

Many empirical papers found a positive correlation between the share of the working-age population and economic growth. Cruz and Ahmed (2018) explained that if the share is higher, it implies that there are more people in an economy that are able to work and to be productive. Thus, if the production is higher, there are also more resources available in a country so that, GDP growth can increase. Hence, a positive coefficient for this variable is expected to be obtained.

One would expect that also the share of the working age population would be decreasing in the selected countries. However, that is not demonstrated. As explained, populations in these countries were not ageing rapidly during 1991-2010. The old dependency ratios were stable and therefore also the share of working-age population did not significantly vary. From the data we can see that in all countries the share of the working-age population was slowly increasing or it was stable from 1991 till 2010-2012 (see Appendix F). After those years, all countries have started to experience a decline and a decreasing trend is also expected to continue in the future. This decline is coinciding with an increase in the old dependency ratio that also started in 2010. In essence, populations are getting older so there are less people who are able to work. This ageing proxy is used in order to make projections for the upcoming 3 decades. The data were retrieved from the United Nations Database. In order to obtain the

(22)

22

projections for the share of the working-age population following calculations were made. The population aged from 15 to 64 years was divided by total population and subsequently and it was expressed as percentages values. In all 4 countries of interest the share of the working-age population is expected to decrease in the future. This may have a negative impact on economic growth.

Following the literature review, several explanatory variables for economic growth are chosen. The first one is human capital in a form of the average years of secondary schooling as used in a study of Bloom and Finlay (2009); and Bloom and Williamson (1998). These authors found a positive correlation between human capital and economic growth when using an OLS regression.

In the sample of countries that observed in this thesis, the average years of secondary schooling have not varied a lot over the last 26 years. There are only 5 countries, where the average years of secondary schooling have actually changed. For instance, in the Slovak Republic it has increased from 8 to 9 years, and in Bulgaria from 7 to 8 years. Consequently, in Albania and Slovenia it has decreased from 8 to 7 years and in Latvia from 8 to 6 years. All in all, in 2016, the duration of the secondary schooling was the longest in the Slovak Republic – 9 year and the shortest in Latvia, Estonia and Poland – 6 years.

Another explanatory variable is the level of income (measured as log of GDP per capita). GDP per capita is a gross domestic product divided by the midyear population. The data are in constant 2010 U.S. dollars.

As mentioned in the literature review, Barro and Lee (1994) explained that using the lagged value of log of GDP per capita is important in order to deal with the lower tendency of overestimated convergence effect. Therefore, this paper uses the lagged value of log of GDP per capita. Many empirical papers found a negative coefficient for the initial level of income. Hence, it is expected to obtain a negative correlation between the level of income and economic growth.

The data have had an upward trend in all countries except Ukraine, where the real GDP per capita was decreasing from 1991 until 1998. Since 1998 it was increasing and between years 2011 and 2016 it was stable. In Poland, this variable has tripled, and in the Slovak Republic it has doubled since 1991. For what concerns Latvia, Estonia, Moldova, Lithuania, Serbia, Montenegro, Bosnia and Hercegovina, Slovenia and Croatia only data starting from 1996 were available online.

(23)

23

Another variable that is used in this paper is trade openness. Following the literature review, trade openness is the sum of exports and imports of goods and services measured as a share of GDP (World Bank, 2018). We lag this variable by one period since we want to control for potential causality issue between trade openness and economic growth. For Croatia, Slovenia, Bosnia and Hercegovina, Montenegro, Slovenia, Lithuania, Latvia, Estonia and Moldova only data starting from 1996 have been retrieved.

Many empirical papers showed that open economies tend to grow faster than closed ones. Therefore, it is expected that also in this thesis a positive correlation between trade openness and economic growth will be observed.

Looking at the data, the trend of trade openness over the last couple of years is very evident and similar in each country. It has significantly increased in 19 countries, except for Bosnia and Hercegovina, where it has been fluctuating (see Appendix G). Especially during the recent financial crisis, trade openness declined to 73% to GDP from 113% to GDP in 2003. Nonetheless, it recovered to 88% to GDP in 2016. The biggest change has occurred in Serbia where trade openness increased from 23% to GDP in 1995 to 107% to GDP in 2016. In Hungary trade openness has tripled since 1991 and in Poland, the Czech Republic, the Slovak Republic has more than doubled.

Another important explanatory variable that needs to be discussed is investments. Specifically, the ratio of domestic investments to real GDP and the ratio of foreign direct investments to real GDP. According to its definition “gross domestic investment consists of outlays on additions to the fixed assets of the economy plus net changes in the level of inventories” (World Bank, 2018).

The same variable was also used in the paper of Barro and Lee (1994). They highlighted that higher domestic investments increase the ratio of the steady-state of output to effective worker and therefore increase the growth. The authors proved the necessity of lagging this variable by one period. There is a potential reverse causality issue between investments and economic growth, therefore lagged variable is also used.

From the data, it is visible that domestic investments have been fluctuating in most countries during last three decades (see Appendix H). The upward trend was observed in the 1990s, which was followed by a significant drop in 2003. Afterwards, investments started to rise again. This growing period ended in 2007, when investment reached its highest values. For instance, in Poland it was 25% to GDP, and in the Czech Republic 32% to GDP. Since 2008 it has been decreasing in almost all countries. In Poland investments reached 19% to GDP, in

(24)

24

the Czech Republic 26% to GDP. A huge change is identifiable in Albania, where investments have quadruplicated, from 7% to GDP in 1991 to 28% in 2016.

Further, foreign direct investments (FDI) are defined as “the net inflows of investment to acquire a lasting management interest (10 percent or more of voting stock) in an enterprise operating in an economy other than that of the investor” (World Bank. 2018).

In this thesis, FDIs are expressed as a percentage to GDP. Although, the impact of FDI on economic growth is not very clear, a positive coefficient is expected. In the selected countries, FDIs are expected to embody technological growth. Hence, they would have a positive impact on economic growth. Similarly, to domestic investments, also foreign direct investments are lagged by one period in order to control for reverse causality issue.

In the V4 countries the FDIs have been fluctuating during last years. Nevertheless, in Hungary there were two significant peaks in 2007 and 2006 and one drop in 2011 when the FDIs were -15%. With respect to the other countries, they overall reached the highest values shortly before the start of the recent financial crisis. After 2009, they have been declining and, in some countries, FDIs have fallen below zero (see Appendix I). For instance, in Austria it was -5% to GDP. However, in 2016 FDIs have recovered again and they were positive in all countries with the exception of Austria.

Another crucial variable is population growth. The population growth rate for year t is “the exponential rate of growth of midyear population from year t-1 to t, expressed as a percentage” (World Bank, 2018).

Following the literature review a negative coefficient is expected for this variable. This is because high values for population growth might lead to higher investments in capital for new workers, rather than investments in capital per worker, that decrease the economic growth (Barro, 1998). In this thesis, population growth is lagged one period, since a reverse causality issue with economic growth is expected.

Population growth has fluctuated a lot in Central and Eastern European countries. In some economies (e.g. Albania, Bulgaria, Latvia and Romania), it has been negative during the last years. It is not surprising since the overall world population growth is decreasing since 1987. Between the V4 countries, population growth has been negative in Poland and Hungary and positive in the Czech and Slovak Republic. The most significant peak took place in the Czech Republic in 2008. It was caused by high new born rates that have occurred in that year.

(25)

25

Political instability is another variable that is taken into consideration. It is expressed in the form of political rights. Political rights are measured on a one-to-seven scale, with one representing the highest degree of Freedom and seven the lowest.

Overall, the values of Political Rights have had the tendency to decrease during last 26 years that may signalize a better political situation in the selected countries. What is interesting is the situation in Hungary. In 1991, political rights had a value of 2. Afterwards, it improved to 1, plummeted in 2016 to a value of 3. Only for two countries - Russia and Belarus – this variable has significantly increased during last three decades, indicating a worsening of political stability. In addition, for Montenegro, Ukraine, Albania and Serbia, the political rights variable was stable and did not change significantly.

Technology is also an essential variable used for the purpose of this thesis. Technology is represented by the gross domestic expenditure on Research & Development, expressed as a percentage of GDP. It is crucial in this thesis to control for technology as it may have a significant impact on GDP growth in the selected ecnomies. During last two decades, when countries have been going through many transitions, technologies have played an important role.

According to Pece, Simona and Salisteanu (2015) technology in Central Europe helped improve the living conditions through the development of the private and public sectors. Therefore, a positive coefficient is expected to be derived.

For this variable only, data from 1996 till 2015 were available. However, for Bosnia and Hercegovina, Montenegro, Albania and Moldova not a complete dataset were retrieved. From the data, it is evident that in the most developed countries (Austria, the Czech Republic, the Slovak Republic, Poland, Hungary, Slovenia, Lithuania and Estonia) expenditure on R&D has been increasing (see Appendix J). As expected, among these countries the highest values have been reached in Austria. Moreover, there are some other countries such as Russia, Moldova, Bosnia and Hercegovina and Croatia, where the expenditure on R&D has not changed significantly during the last couple of years. Additionally, a decreasing trend has been visible in Romania, Montenegro, Belarus and Ukraine. The most significant drop (50%) has occurred in Ukraine. This may have been caused by country´s unstable political situation.

The last variable that is discussed in this master thesis is life expectancy. “Life expectancy at birth indicates the number of years a new-born infant would live if prevailing patterns of mortality at the time of its birth were to stay the same throughout its life” (World Bank, 2018).

(26)

26

A positive effect of higher life expectancy is expected on economic growth. Bloom, Canning and Malaney (2000) argued that higher values for life expectancy mean that people are able to work longer, therefore it directly influences the population’s welfare. Furthermore, higher life expectancies are associated with higher saving rate. This happens because people are able to save more (for their pension years) when they are working.

Life expectancy in Central and East Europe has increased on average from 70.8 years in 1991 to 76 years in 2016. The most significant change is noticeable in Estonia where the life expectancy has increased by 11 years during the last couple of years. Differently, smaller changes have occurred in Belarus, Russia and Ukraine where it has increased only by 3 years since 1991. In addition, population born in V4 countries in 1960 live at average 69 years, people born in 2015 are expected to live on average 77.6 years (see Appendix B). Durasiewicz (2015) predicted that this trend will continue to rise at almost the same rate in the future. In fact, people will live longer and will be healthier.

The table below (see Table 1) consists of the descriptive statistics for the 20 Central and Eastern European countries included in this research. It is evident that there are sufficient variations with respect to ageing and economic growth.

Table 1: Descriptive Statistics 1991-2016 for 20 countries

Variable Obs Mean Std.Dev. Min Max

Growth of real GDP per capita 471 3.161256 7.035335 -29.16327 92.12334 Old dependency ratio 520 20.44802 4.171534 9.194239 31.29074 Young dependency ratio 520 26.82217 6.348779 19.19955 53.66836 Working-age population share 520 67.96589 2.075386 60.89983 74.33753

Secondary years of schooling 520 7.540385 0.7489902 6 9 Real GDP per capita 482 10494.45 9531.942 698.5643 48172.24 Trade openness (EX+IM/GDP) 480 99.27249 32.83124 23.21612 185.7471

Domestic investments to GDP 479 24.7589 6.063603 0.2986439 42 Foreign direct investments to GDP 466 4.608308 5.99516 -15.98922 54.61486

Population growth 519 -0.3379081 0.7089831 -5.814339 3.732596

Political rights 489 2.390593 1.623594 1 7

Expenditures on R&D to GDP 350 0.9053896 0.5563052 0.01611 3.07181 Life expectancy 513 72.89631 3.486243 64.46707 81.49024

(27)

27

4. Methodology

This chapter provides an overview of the methodology that is used in this master thesis, in order to estimate the effect of population ageing on economic growth in the V4 countries. The first part, section 4.1, describes the regression model. The second part, section 4.2, discusses the methodology for the projections, which are made for the upcoming 34 years by using the coefficient of the selected ageing proxy.

Regression model

The regression analysis is partially based on the paper written by Bloom and Finlay (2009). The authors investigated the impact of demographic changes on economic growth in Asia over the 1960-2005 period by using OLS and IV regressions. For the regression analysis they used demographic data: the working-age shares, life expectancy, population growth and population density. The aim of this study is to examine the impact of ageing. For this purpose, we use 3 ageing proxies: old dependency ratio, young dependency ratio and the working-age shares. By using these different proxies for ageing, we are able to better identify what is the exact way through which ageing has had and will have an impact on economic growth in our 4 countries of interest. In order to avoid multicollinearity, the ageing proxies are not included together contemporaneously but in the two separate regressions. As presented below (see regressions), in the first regression we include the old dependency ratio and the young dependency ratio. In the second regression, instead of dependency ratios, we include the share of the working-age population.

𝐺𝑅𝑂𝑊𝑇𝐻_𝑖,𝑡 = 𝛽₀+ 𝛽₁𝑂𝐿𝐷_{𝑖,𝑡−1}+ 𝛽₂𝑌𝑂𝑈𝑁𝐺_{𝑖,𝑡−1}+ 𝛽₃𝑆𝐸𝐶𝑂𝑁𝐷_𝑖,𝑡 + 𝛽₄𝑅𝐺𝐷𝑃_{𝑖,𝑡−1} + 𝛽₅𝑇𝑅𝐴𝐷𝐸_{𝑖,𝑡−1}+ 𝛽₆𝐼𝑁𝑉𝐸𝑆𝑇_{𝑖,𝑡−1}+ 𝛽₇𝐹𝐷𝐼_{𝑖,𝑡−1}+ 𝛽₈𝑃𝑂𝑃_{𝑖,𝑡−1}+ 𝛽₉𝑃𝑅_𝑖,𝑡 + 𝛽₁₀𝑅&𝐷_𝑖,𝑡 + 𝛽₁₂𝐿𝐼𝐹𝐸_𝑖,𝑡+ 𝜀 𝐺𝑅𝑂𝑊𝑇𝐻_𝑖,𝑡 = 𝛽₀+ 𝛽₁𝑊𝑂𝑅𝐾_{𝑖,𝑡−1}+ 𝛽₂𝑆𝐸𝐶𝑂𝑁𝐷_𝑖,𝑡+ 𝛽₃𝑅𝐺𝐷𝑃_{𝑖,𝑡−1}+ 𝛽₄𝑇𝑅𝐴𝐷𝐸_{𝑖,𝑡−1} + 𝛽₅𝐼𝑁𝑉𝐸𝑆𝑇_{𝑖,𝑡−1}+ 𝛽₆𝐹𝐷𝐼_{𝑖,𝑡−1}+ 𝛽₇𝑃𝑂𝑃_{𝑖,𝑡−1}+ 𝛽₈𝑃𝑅_𝑖,𝑡+ 𝛽₉𝑅&𝐷_𝑖,𝑡 + 𝛽₁₀𝐿𝐼𝐹𝐸_𝑖,𝑡+ 𝜀

(28)

28

In total, this thesis consists of 7 models. In each of them, we include year fixed effects so that one may control for annual shocks that have occurred during the last 3 decades.

The first model (basic growth model) is estimated by using pooled OLS regression. It is identical to the one in the paper of Bloom and Finlay (2009) and, therefore, including similar variables: the level of income (measured as log real GDP), political instability (measured in a form of political rights), trade openness, human capital variable (measured as average years of secondary schooling). Additionally, another 3 variables are introduced: foreign direct investments, gross domestic investments, and government expenditures on R&D, all measured as a ratio to GDP. All explanatory variables, except for political rights, secondary schooling and expenditures on R&D, enter our regression in lagged forms since we want to control for potential causality issue as explained in chapter 2.

Model (2) is again similar to the regressions proposed by Bloom and Finlay (2009) and is estimated by using pooled OLS. In addition to model (1), we introduce the following variables: population growth, life expectancy, and the young and the old dependency ratio. These variables are added in order to investigate the effect of ageing.

Moreover, model (3) is again estimated by using pooled OLS. However, instead of old dependency ratio and the young dependency ratio we use another ageing proxy - the working-age share. As already mentioned, we do not include all working-ageing proxies at the same time to avoid multicollinearity. In addition, since we expect that previous values of ageing proxies may have an impact on current economic growth, we use the lagged values of ageing proxies. Furthermore, since we are dealing with panel data, pooled OLS may violate many assumptions. That is why, before performing the other models, we run the Hausman test. This test is used in order to choose between fixed or random effect. Depending on the results we estimate model (4) and (5) by using fixed or random effect. These models are identical to previous equations (2) and (3), therefore same variables are included.

Last but not least, we perform the last two models (6) and (7) by using IV estimation. The regressions are again going to be similar, but we control for potential reverse causality between population growth and GDP growth by using the lagged values of population growth and the fertility rate as instruments.

(29)

29

Forecasting model

The results from the regression analysis are used in order to estimate the future impact of ageing on economic growth in V4 countries. We use the coefficient of the working-age share for making projections from 2017 to 2050. Firstly, the coefficient is combined with the actual demographic changes that have occurred in the Czech Republic, Hungary, Poland and the Slovak Republic. More precisely, we multiply the coefficients retrieved from the regression analysis with the actual demographic changes - changes in the working-age share for the period 1991-2016. By doing that, we separately calculate for each country what has been the impact of the population ageing on economic growth during last 26 years. This helps us distinguish and compare the effect of ageing on economic growth in 4 selected countries.

After that, by combining the same coefficient (retrieved from our regression analysis) with projections for demographic changes (changes in the working-age share for the period 2017-2050), we calculate the future impact of ageing on economic growth. Data for future demographic changes are obtained from the United Nations database. As already explained (see chapter 3), the projections for the working-age share have been calculated by the author of this study. The population aged from 15 to 64 years was divided by total population and it is expressed in percentage values.

(30)

30

5. Empirical results and interpretation

In this chapter empirical results are interpreted by using 3 different methods: pooled OLS, fixed-effects estimation and IV estimation. Secondly, in subsection 5.2, we investigate what has been the impact of the population ageing on economic growth in 4 selected countries. It is done by combining the coefficient of the selected aging proxy with the actual demographic changes. Lastly, this subsection makes the projections for the upcoming 34 years using the ageing coefficient. Last but not least, it is important to mention that, the econometric package STATA is used to run the regressions in this research.

Estimated results

The results are presented in table 2. As already discussed (see methodology), we decide to include year fixed effects (in each regression) in order to control for annual shocks that have occurred during the last 3 decades.

Column (1) represents the growth model and it is similar to one proposed by Bloom and Finlay (2009). The results show a significant and positive correlation between the lagged value of trade openness and economic growth. These results are consistent with previous findings in other empirical papers. Thus, countries benefit from trade openness. We also find a positive and significant coefficient for political rights. This result indicates that a higher level of political rights decreases real GDP per capita growth. It confirms the finding of Barro (1998). He argued that only low levels of political rights enhance economic growth. Once a certain level of democracy is reached, high levels of political rights actually decrease GDP per capita growth. One may expect a positive correlation between this variable and economic growth. In contrast, a significant and negative coefficient is derived for the expenditures on R&D. As an explanation, it is essential to mention that our data are limited because we obtained 350 observations for this variable. Alternatively, the problem could arise by using an inappropriate variable in order to control for R&D. Future research might use, for instance, the number of researchers employed in R&D as a variable for research and development. Moreover, the coefficients of other variables are insignificant.

In a column (2), again similarly as in Bloom and Finlay (2009), we add the population growth, life expectancy and two ageing proxies – old dependency ratio and young

(31)

31

dependency ratios. For ageing proxies, we use lagged values as we expect that ageing from last year has an impact on current economic growth. Nevertheless, when using the pooled OLS regression, insignificant results for both are obtained. A positive but insignificant correlation is derived for log of GDP per capita and economic growth. There is a significant and negative correlation between population growth and economic growth. For life expectancy that is a health indication we obtain a negative and insignificant coefficient. The positive and significant correlation is again find between trade openness, political rights and economic growth.

In the third column (3), we replace the old dependency and young dependency ratios by another ageing proxy; working-age share. The share of the working-age population is positively correlated with economic growth, although obtained coefficient is insignificant. The coefficient of R&D stays negative and significant. Similarly, as in column (2) we obtain a negative and significant result for population growth. In essence, population growth in Central and East Europe decreases the economic growth. Our results correspond to the ones found in papers of Barro and Lee (1994); Barro (1998).

Furthermore, before performing other regressions, the Hausman specification test was run. Hausman test is used in order to choose between regressions with fixed and random effect. The main idea is that it checks whether there is a correlation between the unique errors and the regressors. Random effect model is preferred if the omitted variables are not correlated with the explanatory variables. In case of correlation, the null hypothesis is rejected. Explicitly, the null hypothesis assumes that the model is a random effect, the alternative hypothesis assumes that the model is a fixed effect. In other words, if the model includes omitted variables and they are correlated with the explanatory variables, it is advisable to use fixed effects model. The results can be interpreted as follow: if p-value is less than 0.05, we reject the null hypothesis and we choose the fixed effect model (Torres-Reyna, 2007). In our regression, we obtained the p-value that equals to 0.00 (see Appendix L). Thus, we reject the null hypothesis and we choose the fixed effect. Subsequently, in column (4) and (5) we run the same regressions as in column (2), respectively in column (3) by using the fixed-effects estimation instead of pooled OLS. This helps us control for potential endogeneity issues. R-squares increases and the coefficients for old and young dependency ratios become significant. It is not surprising that the coefficient for the young dependency ratio is negative, since also other papers described in literature review found same results. For example: Bloom, Canning and Finlay (2010); and Cruz and Ahmed (2018). They stressed that a higher

The impact of population ageing on economic growth in the Visegrad group countries