Government Expenditure, Education, and Productivity in the European Union: Effects on Economic Growth

2.1 The Relationship Between Human Capital and Economic Growth

The acquisition of knowledge and skills by individuals is a form of capital holding that, like other types of capital, is acquired through an investment process. As Becker (1962, 1964) said, education and training are the most important investments in human capital.

A large number of studies focusing on the theory of human capital confirm its relevance. The relationship between human capital and economic growth has been studied by authors such as Barro (2001), Becker (1964), De la Fuente and Doménech (2006), Frank (1960), Galor and Tsiddon (1997), Krueger and Lindahl (2001), Lucas (1988), Mankiw et al. (1992), Mincer (1984), Romer (1986, 1989), Schultz (1961), Stokey (1991) or Weisbrod (1962), among others.

Among the results obtained by empirical approaches, it is found that education generates positive externalities that give rise to private and social returns. A wide range of studies found that increases in school enrolment, as well as the increases in highly skilled workers, raise GDP per capita and productivity. For example, Angrist and Krueger (1991), Bassani and Scarpetta (2001), Card (2001), Denison (1985), Dickson (2013), Griliches (1997), Gurney and Englander (1994), Jenkins (1995), Nelson and Phelps (1966), and Pritchett (2001) to name a few.

Regarding the fact that education can promote both private and social benefits, Krueger (1999) stressed that the literature shows two important conclusions. First, the initial stock of human capital is more relevant than that of physical capital. Second, secondary and post-secondary education are more important than primary education.

Along these lines, Funke and Strulik (2000) observed that physical capital contributes greatly to the growth of per capita income in the early stages of development, but it is through the accumulation of knowledge (through continuing education and training) that progress is made towards higher stages of economic development. Later, Krueger and Lindahl (2001) showed that technological progress is related to an expansion of human capital when starting from high levels of it, while crime reduction or well-informed political decision-making are related to the development of human capital from lower levels.

As Barro (2002) stated, the difference between prosperity and poverty for a country depends on how fast it grows in the long run. This author revealed that for a specific level of initial per capita GDP obtained with a higher proportion of human capital than physical capital, it tends to promote higher growth through two channels: the absorption of more developed technologies facilitated by more developed human capital and the difficulty of adjusting human capital, as opposed to physical. This implies that countries with a higher proportion of human capital relative to physical capital tend to grow faster by adjusting the amount of physical capital upwards.

Fleisher et al. (2010) established a positive link between investments in human capital, specifically higher education, and economic growth, particularly in the outer regions of China. In a different context, Hanushek (2013) emphasized the significance of cognitive skills and educational quality in driving long-term economic growth. More recently, Lim et al. (2018) introduce a comprehensive global measure of human capital, assessing the combined impact of education and health as critical determinants for economic growth.

From a labour market perspective, Teixeira and Queirós (2016) addressed the need to align a well-educated workforce with specific industries for sustained economic development. In this context, Cinnirella and Streb (2017) reveal the vital role of boosting human capital, particularly through basic education, for the ongoing economic growth, improved labour productivity and accelerated technological change.

From another point of view, it is worth noting that public institutions play a central role in the investment in human capital. Sylwester (2000) reveal that income inequality primarily drives public education expenditures, positively correlated with future economic growth but negatively impacting contemporaneous growth. Blankenau and Simpson (2004) argue that while public expenditures positively affect economic growth, this impact may diminish or negate under certain conditions, depending on factors like government spending, tax structure, and production technologies. The expansion of higher education globally, driven by institutional processes, underscores a societal shift where education is acknowledged as indispensable “human capital” for unrestricted progress (Schofer & Meyer, 2005). This evolution highlights the vital link between institutions and the interdependence of human capital with economic development. Tsai et al. (2010) corroborate these findings, noting differences between developed and developing countries. Similarly, Nissan et al. (2011) identifies a direct and indirect association between institutional quality and human capital development. From the same perspective, Dias and Tebaldi (2012) delve into the relationship between human capital, institutions, and economic growth. Ramos et al. (2012), from a novel perspective, discover, based on microdata in the EU, that human capital, measured as overeducation, contributes to economic growth.

Qadri and Waheed (2013) highlight positive returns on investment in human capital for both lower-income and higher-income countries, with the former experiencing greater benefits. Churchill et al. (2015) find that government education expenditure positively influences economic growth in developed countries but not in less developed ones, with government health expenditure exhibiting a negative growth effect. In this regard, Affandi et al. (2019) assert that human capital is a key driver for economic growth.

Nowadays, regarding our research goal, the case of the EU-27 countries, we could obtain some information on the relationship between spending on education and the evolution of GDP per capita, looking at the data.

In Figures 1–4, we can see the evolution of public spending on education in the European countries and the EU average. Given the reference of the EU average, it is easy to perceive the heterogeneous pattern of the expenditure in education. In the Western group, the data from Ireland follow a descendent evolution. In the Norther group, most of the countries exhibit figures above the EU average. We found the opposite when looking at Central and Southern countries. However, with the exception of Cyprus, Malta, Poland, and Slovenia, the rest of the countries show figures below the EU average for most of the years.

Figure 1

Public spending on Education in Western Europe (2009–2020). Source: Own elaboration based on Eurostat data.

Figure 2

Public spending on Education in Northern Europe (2009–2020). Source: Own elaboration based on Eurostat data.

Figure 3

Public spending on Education in Central Europe (2009–2020). Source: Own elaboration based on Eurostat data.

Figure 4

Public spending on Education in Southern Europe (2009–2020). Source: Own elaboration based on Eurostat data.

Figures 5–8 show the growth of GDP per capita for the same groups of European countries. The impressive increase of GDP for the case of Ireland, reaching the maximum in 2015, is remarkable. Even in 2020, the year of the crisis of COVID-19, Ireland show a figure above the rest of the countries and that of the UE average. Regarding the Norther countries, the evolution of GDP per capita reveals a rapid recovery after the financial crisis of 2008, reaching positive values around the EU average for the rest of the period. For Central and Southern countries, the recovery begins some years later, around 2012 and 2013. But the evolution is irregular when we look at the entire period. In all the countries or that two groups, the GDP per capita figures sharply decline in 2019.

Figure 5

Real GDP per capita growth rate in Western Europe (%) (2009–2020). Source: Own elaboration based on Eurostat data.

Figure 6

Real GDP per capita growth rate in Northern Europe (%) (2009–2020). Source: Own elaboration based on Eurostat data.

Figure 7

Real GDP per capita growth rate in Central Europe (%) (2009–2020). Source: Own elaboration based on Eurostat data.

Figure 8

Real GDP per capita growth rate in Southern Europe (%) (2009–2020). Source: Own elaboration based on Eurostat data.

2.2 The Human Capital and the Reduction of Inequalities

Barro and Lee (2013) argued that the level and the distribution of educational achievements also have effects on social progress. Hence, the acquisition of human capital and its contribution to reducing inequalities has also been studied by various authors such as Benabou (1996), Castelló and Doménech (2014), Chani et al. (2014), Checchi (2006), Glomm and Ravikumar (1992), or Murphy and Topel (2016). According to Le Grand (2018), education is the great equalizer of the conditions of men and women.

But despite the enormous development of economies achieved by many countries, manifested in aspects such as convergence in the increase in life expectancy, or the accumulation of human capital, there are still notable differences between rich and poor. Schultz (1961) stated that investment in human capital accounts for most of the impressive increase in real income per worker. On this subject, Becker (1964) addressed that the incomes of the most educated people are almost always much higher than the average, although profits are generally higher in the least developed countries. This could be explained by the greater inequality of this group of countries. The starting assumption is that human capital development contributes to the reduction of inequality, which directly influences the economic growth.

As Castelló and Doménech (2008) showed, while the secondary school enrolment rate was almost 100% in rich countries, more than 70% of children in sub-Saharan Africa were not enrolled in secondary school and, therefore, entered the labour market as unskilled persons, workers since childhood.

One of the main objectives of growth in education spending is to reduce inequality, which has a significant impact on aggregate output. Gradstein (2003) found that inequality in the distribution of public spending on education slows down long-term growth and that income inequality is also harmful in this sense. Viaene and Zilcha (2003) found, within a general equilibrium framework with human capital accumulation, the cross-country differences in income distribution. Specifically, among other variables, they found that education technology generates a cross-country variation in the equilibrium income distributions. Coady and Gupta (2012) revealed that income inequality has been expanding in many advanced economies since 1980. This is due to an array of variables such as the effects of globalisation on the wages of low-skilled workers, which are also adversely affected by technological change. According to Fredriksen (2012), inequality has increased within a “core” of eight European countries, where large income increases, among the top 10% of those earning, appear to be the main driver of this evolution. Along the same lines, Fournier and Johansson (2016) showed that public investment and education boost potential growth, while other social expenditures such as pensions and public subsidies reduce it. In addition, reforms aimed at improving the efficiency of the public sector, as well as educational reform that promotes completion of secondary education, would contribute to reducing income inequality.

The relationship between inequality and human capital can be analysed from a bidirectional perspective. In this regard, Sawicka and Stolarczyk (2018) found that equality of opportunities in the labour market, social life, education, and public life enhances the value of human capital, generating economic benefits while lack of such equality causes losses. As a driver of equal opportunities, education is a powerful tool for achieving social inclusion, which leads to a decrease in income inequality, promoting intergenerational mobility. Conversely, a higher per capita income, greater trade openness, and accelerated technological progress tend to contribute to heightened inequality in both income and education distribution (Lee & Lee, 2018). As an example of the importance of these arguments, the United Nations significantly incorporates these ideas into the Sustainable Development Goals promoting Quality Education as Goal No. 4, closely linked to No. 5: Gender Equality, No. 8: Decent Work and Economic Growth, and No. 10: Reducing Inequalities.

When studying inequality, most authors focus on wages as the variable to be analysed. But when measuring wages, the concentration of graduates in public sector employment is identified as a problem in growth studies. This problem lies in the fact that public sector wages do not normally reflect market wages (Pissarides, 2000). On the other hand, Psacharopoulos and Patrinos (2004) pointed out the growing concern of governments and other agencies in the educational reforms on returns, to guide macro-political decisions on the management and financing of educational reforms.

Although wages are an indicator of well-being, it is useful to analyse other factors such as job quality and satisfaction. In this sense, Albert and Davia (2005) found a strong relationship between education and satisfaction, explained by the link between education and job quality. However, the authors pointed out that this link is not conclusive due to salary expectations and characteristics. According to Castriota (2006), the higher the educational level, the less relevant is the absolute level of income for self-declaration of satisfaction with life. This explains why the pattern of southern European countries is more in line with common evidence (Albert & Davia, 2005).

2.3 Level of Education and Productivity

The traditional theory of human capital (Becker, 1962) conceives education as an investment in human capital, which favours greater workers’ productivity and greater returns in the form of better salaries. It is also worth highlighting the theory of signalling (Spence, 1973) according to which the educational level of the individual is an indicator of his productivity, and it is the cost of acquiring education (in monetary terms and effort), which differentiates the most productive workers from the rest. To contrast the predominance of those approaches, we found Chevalier et al. (2004) who ask themselves about the mechanism through which education translates into higher salaries. They found, at least in the case of the UK, that education raises productivity and that this fact predominates over the education as a signal.

A higher level of human capital provides a greater capacity to absorb new technologies and business techniques. In other words, the level of human capital, approximated as the educational level of the active population, can have an impact on productivity levels (what is known in the literature as “level-level effect”) and/or on the growth rates of this (“level-growth effect”). Specifically, the first effect is due to school returns, while the second is through innovation. Just as the accumulation of human capital produces income growth, so do the corresponding social or national aggregates (Mincer, 1984). In fact, both Barro (2001) and Lucas (1988) pointed out that the educational level can also affect productivity through different ways such as improving the capacity for innovation and the quality of the workforce.

Numerous studies have highlighted the importance of education on labour productivity. For instance, Benhabib and Spiegel (1994) found that productivity growth would be around 0.13% against a 1% increase in human capital. This fact has the strongest impact on regional productivity, with R&D stock showing a positive bi-directional causality (Bronzini & Piselli, 2009). In this sense, Becker (1964) had affirmed that the expansion of scientific and technical knowledge, which increases the productivity of labour and other inputs in production, is the reason for the persistent growth per capita in the United States, Japan, and European countries in the last hundred years. Consequently, there is a link between wage growth and education and training.

The growing number of skilled workers and the ability to absorb advanced innovations is related to productivity. For that reason, many studies have pointed to the relationship between human capital, measured as education attainment, and labour productivity. Moretti (2004) studied the magnitude of human capital spillovers and found that the productivity of firms increased as employed college graduates were employed. Bowlus and Robinson (2012) investigated human capital prices and quantities in the United States from 1963 to 2008, employing four education levels as skill types. They found a high correlation in human capital prices across different skill types, an increase in prices until the mid-1970s followed by a decline, and changes in human capital quantities explaining a substantial part of relative wage and total factor productivity evolution. Chansarn (2010) showed the suitability of education as a key factor in labour productivity by analysing 30 countries during the period 1981–2005. Later, Mason et al. (2012) found the same relationship, considering both certified and uncertified skills as drivers to improve labour productivity. However, these results are limited to individuals with higher education degrees.

For the case of EU and in most of the regions of Spain, De la Fuente (2011) found that social returns to investment in human capital were higher than those on physical capital. The same results were found by Benos and Karagiannis (2016), for the case of Greece, showing that human capital presents a strong positive association with productivity through upper secondary and tertiary education. In this sense, Barro and Lee (2013) argued that the level and distribution of educational achievement also have an impact on social and welfare outcomes, such as infant mortality, fertility, child education, and income distribution, in addition to social status, as Fershtman et al. (1996) had already pointed out. In addition, these authors stated that, in general, growth may be favoured by an increase in the number of workers investing in education. However, they found inefficiency in talent allocation. In this sense, and more recently, Vandenbussche et al. (2006) argued that an optimal composition of public spending on education is necessary, and this depends on the relative distance of the economy from the technological frontier. While Kampelmann et al. (2018) find that higher education credentials have a stronger effect on firms’ productivity than on wage cost. And, therefore, the room for employability increases for higher educated workers. More recently, Hansen et al. (2021) have examined the labour market returns of skill signals, finding that even when high grades initially translate into higher earnings, employers learn about workers productivity later. By exploring the linkages between innovation and economic growth, Biasi et al. (2021) address that a high-quality education, through building cognitive and non-cognitive skills, increases the productivity of students turning them into potential innovators.

After reviewing the empirical findings, we could ask ourselves about what current data, of the EU-27, could tell us about the level of education and productivity.

In Figures 9–12, we can see the evolution of labour productivity in the European countries and the EU average. The evolution of productivity shows a similar pattern to the previously observed for GDP per capita (Figure 5), drawing again attention the value reached in Ireland in 2015, and even Ireland shows figures above the EU average in 2019 and 2020. When looking at the Northern countries, they also exhibit a similar pattern to that of the GDP per capita evolution we have seen before (Figure 6). Paying attention to the Central and Southern countries, the most remarkable is the case of Greece, where labour productivity remains below the average, and below positive figures, for the whole period. The rest of the countries show trends oscillating around the EU average, but after 2019, labour productivities decline sharply in all the countries.

Figure 9

Labour productivity in Western Europe (%) (2009–2020). Source: Own elaboration based on Eurostat data.

Figure 10

Labour productivity in Northern Europe (%) (2009–2020). Source: Own elaboration based on Eurostat data.

Figure 11

Labour productivity in Central Europe (%) (2009–2020). Source: Own elaboration based on Eurostat data.

Figure 12

Labour productivity in Southern Europe (%) (2009–2020). Source: Own elaboration based on Eurostat data.

In Figures 13–16, we can see the evolution the attainment of tertiary education level in the European countries and the EU average. All the countries exhibit a growing trend of graduates in higher education level. In the Western and Northern countries, with the exception of Germany from 2013 onwards, all the countries show figures above the EU average. On the contrary, among the Central and Southern countries, only Cyprus and Spain exhibit figures above the EU average.

Figure 13

Population with tertiary education in Western Europe (2009–2020). Source: Own elaboration based on Eurostat data.

Figure 14

Population with tertiary education in Northern Europe (2009–2020). Source: Own elaboration based on Eurostat data.

Figure 15

Population with tertiary education in Central Europe (2009–2020). Source: Own elaboration based on Eurostat data.

Figure 16

Population with tertiary education in Southern Europe (2009–2020). Source: Own elaboration based on Eurostat data.

If we look jointly the figures for high education levels and those of labour productivity, we can observe that the data do not reveal a similar evolution of that variables for most of the countries.

2.4 The Efficiency of Spending on Education

When education succeeds in improving human capital endowment and thus productivity, it is concluded that education funding has been efficient. Education spending decisions are closely related to efficiency, which has been extensively studied.

Many studies have found a positive relation between education spending and outcomes. In this regard, subsides are a key factor for promoting economic growth (Blankenau, 2005). Despite many studies emphasize the positive relationship between educational expenditure and academic outcomes, it is crucial to use resources efficiently. Furthermore, an intricate understanding of the nexus between education and economic growth needs a nuanced distinction between the quantitative and qualitative dimensions of educational provision. This differentiation holds pivotal implications for research on educational efficiency, given the inherent challenges in measuring the qualitative aspects of education as opposed to its quantitative aspects.

According to Taylor (1994), the losses attributable to inefficiencies in secondary education likely constitute less than 1% of potential GDP. Authors such as Sutherland et al. (2007) found a weak relationship between educational achievement and increased availability of resources, which can be explained by international differences in efficiency levels in primary and secondary education. In this regard, these authors stressed that there is significant scope for improving efficiency by moving towards best practices. Hence, analyses on the efficiency of spending on education are abundant. Rajkumar and Swaroop (2008) noted that the efficiency of education spending depends on the governance of countries, measured by the degree of corruption and the quality of bureaucracy. Mandl et al. (2008) and St. Aubyn et al. (2009) studied the case of EU states and found that outcomes varied across countries depending on the independence of institutions and educational policy decisions. Blankenau and Camera (2009) pointed out that an excess of subsidies increases the number of students but decreases the incentives of these to acquire skills. Annabi et al. (2011) showed that incentives for higher education can promote the accumulation of human capital and alleviate the negative effects of unemployment. But the final impact depends on how distortive alternative fiscal instruments are and on the degree of efficiency of education spending.

Hanushek and Woessman (2011) found that it is the institutional structure that determines the incentives and that the differences in the structures of the educational systems are those that condition for the attainment achieved by the students. Agasisti (2014) and Sibiano and Agasisti (2013) found that differences in the economic development of the regions influence the acquisition of human capital and that there is no linear relationship between increases in public spending on education and increases in educational attainment. Other factors related to the composition and distribution of expenditure should therefore be addressed.

On the other hand, Heckman and Mosso (2014) noted that schooling at all stages has a positive impact on health and wages, analysing a model with multiple schooling options that recognizes the fundamentally nonlinear effect of schooling on a variety of outcomes. It is a powerful reason for governments to increase investment in education. However, the effectiveness of education is necessary and may be reduced when spending decisions are not adequate (Kingdon et al., 2014).

Wolff (2015) analysed the relationship between public spending on education and the result of students. He found that it responds to an unbalanced growth model, given that the productivity and cost of components of the educational service (use of technological advances and teacher activity, for example) do not grow at the same pace. Dissou et al. (2016) concluded that different ways of financing education led to long-term growth, but that there are differences which can be observed in the short term and the transition to the long term. Hyman (2017) found that students who were exposed to a $1,000 rise in spending per year correlates with a 3.9% higher likelihood of college enrolment and a 2.5% higher likelihood of degree attainment. More recently, Canton et al. (2018) measured the efficiency of public spending on education in the EU and found different results depending on the functioning of the institutions.

In conclusion, the intricate relationship between education spending and outcomes underscores the need for careful consideration in resource allocation. Despite positive correlations between spending and indicators like college enrolment and degree attainment, the effectiveness of education is contingent upon the efficiency of spending decisions. Studies examining the efficiency of education spending reveal diverse outcomes across countries, influenced by factors such as governance, institutional structures, and economic development. It is crucial to address the composition and distribution of expenditure to maximize the societal impacts of education. In addition, recognizing the nonlinear impact of schooling on various outcomes emphasizes the importance of aligning spending decisions with the evolving dynamics of educational service components, technological advances, and teacher activities. As governments strive to enhance the effectiveness of education in promoting health and wages, thoughtful consideration of spending decisions remains paramount in navigating the complexities of education funding.

3.1 Variables and Dataset

As we have already stated, in our first empirical analysis, we have used time series of annual data taken from Eurostat, for the years 2009 to 2020 of the 27 EU member states,^[2] for the following variables:

PRO: Work productivity per person. Real labour productivity per person, as a percentage of change over the previous year.

GPE: Public expenditure on education. Total general government expenditure on education as a percentage of GDP.

GID: Expenditure on R&D in the education sector, performed within the business enterprise sector. Total expenditure on research and development in the higher education sector as a percentage of GDP, regardless of the source of funds.

EST: Employment in technology intensive sectors. Employment in high technology and knowledge intensive sectors as a percentage of total employment.

WEM: Average wage gains. Average individual annual net earnings expressed in purchasing power parity units (PPS, purchasing power standard).

WRM: Average wage income. Income from wages and salaries as a percentage of GDP.

EDS: Educational level of the population, secondary level. Percentage of population with secondary and post-secondary education.

EDT: Educational level of the population, tertiary level. Percentage of population with tertiary education.

EPL: Work experience measured by the years worked. Average years worked, duration of working life.

CRP: GDP growth. Chained volumes of GDP at market prices, as a percentage change over the previous year.

CPC: Per capita GDP growth. Chained volumes of GDP at market prices, as a percentage change over the previous year, per capita.

Table 1 shows the main statistics of the series.

If we look at the measures of position, the values for the mean and the median are quite close in all the series. It is remarkable that the maximum values for PRO, CRP, and CPC correspond to Ireland in 2015. The same occurs for Ireland in 2020, regarding EDT and EST. Those observations could anticipate, at least for the case of Ireland, that exists some relationship among tertiary education, employment in technology intensive sectors, productivity, and economic growth. On the contrary, the minimum values for those variables correspond to Malta in 2020, Estonia in 2009 (for CRP and CPC), Portugal in 2009 for EDS, and Romania in 2009 for EST. The maximum of GPE corresponds to Denmark in 2020 and the minimum to Romania in 2013. The same countries show the maximum and the minimum for GID in 2015 and 2020, respectively. Regarding the measures of WEM and WRM, for the average wage gains, Netherlands exhibit the maximum value in 2020, while Bulgaria the minimum in 2009; and for the average wage income, the maximum value corresponds to Denmark in 2009 and the minimum to Ireland in 2020. For the EDS, the maximum value corresponds to Czechia and the minimum to Portugal in 2009. Finally, for EPL, Sweden exhibits the maximum value in 2019 and 2020, while Hungary shows the minimum in 2009.Those statistics reveals the heterogeneity of the EU-27 standards along the period. That most of the minimum values correspond to 2009, the year after the outbreak of the financial crisis of 2008, and some of them to 2020, the year of the COVID-19 crisis. Regarding the dispersion of the data, the maximum value of the standard deviation can be observed for EDS, while the minimum for GID. The values for the skeweness are near zero but showing the maximum (0.9) for EST and the minimum (−0.41) for CPC. Looking at the kurtosis, the maximum corresponds to PRO and the minimum to WEM. Finally, when looking at the Jarque-Bera statistic, the null hypothesis of the normally distributed data should be rejected for most of the variables. Although given the relatively small size of the sample, we cannot reach a clear-cut conclusion about the normality of the distributions.

3.2 Some Preliminary Results

Before making the estimates, we will explore the cointegration relationships of the variables that are analysed for the group of individuals in the set (EU member states). To this end, as a preliminary study, we will check the order of integration of the series. For this, we will apply the tests conducted by Im et al. (2003), Levin et al. (2002), Maddala and Wu (1999) or Fisher type ADF and that by Choi (2001) or Fisher type PP, whose null hypothesis is the existence of unitary root. These unit root tests for panel data are based on those developed for time series, but by also taking into account cross-sectional data they increase degrees of freedom, improve estimator properties, and correct unobserved heterogeneity. The analysis is completed with the Hadri test (2000), based on the LM estimator proposed by Kwiatkowski et al. (1992), which considers the cross-dependence between individuals (EU states) and whose null hypothesis is that the series is stationary.

The results of unit root tests are presented in the Appendix. The results of Table A1 indicate that, except for the PRO, CRP, and CPC series, the null hypothesis of the presence of a unit root in the first four tests cannot be rejected, and the hypothesis of seasonality when the Hadri test is applied. Table A2 shows the root tests for variables that presented a unit root, but this time expressed in first differences. In this case, the results show that the variables in first differences no longer present a unitary root.

Next, to identify long-run relationships among the variables, we carry on the cointegration tests proposed by Pedroni (1999, 2001) and by Kao (1999). Both extend the Engle-Granger approach to test panel data, being the null hypothesis no cointegration. The tests proposed by Pedroni consider the within dimensions (panel v, panel rho, panel PP, and panel ADF) as well as the between dimension approach (group rho, group PP, and group ADF), allowing for heterogeneous intercepts and trend coefficients across cross-sections. For a relatively sample size, as ours, Pedroni (1999) indicates the group PP, the panel PP, and the panel rho as the more reliable statistics. In addition, the Kao test also considers cross section- specific intercepts and homogeneous coefficients on the first stage regressors.

The results of the cointegration tests for the equations to be estimated given by (1)–(3) described in Section 3 are shown in the Appendix. The results indicate that the null hypothesis of no cointegration may be rejected.

Once the cointegration relationship has been verified, we will proceed to estimate the long-term parameters in the next section.