Introduction
Human capital is a measure of skills and attributes of labour in which education plays a crucial role to increase efficiency and productivity. Theoretical discussions on human capital are in broad consensus that education is the main source of human capital (Becker Reference Becker1964; Griliches Reference Griliches1997; Mincer Reference Mincer1974 among others). Odit et al. (Reference Odit, Dookhan and Fauzel2010) describe education as an investment in human capital due to the fact that benefits accrue to an educated individual over a lifetime of activities. Hanushek and Wößmann (Reference Hanushek, Wößmann, Peterson, Baker and McGaw2010) describe education as a key determinant of economic well-being, which leads to an increase in the human capital inherent in the labour force of a country. In addition, a great number of empirical studies have also addressed the importance of education in promoting economic growth (see, for example: Temple Reference Temple2002; Afzal et al. Reference Afzal, Rehman, Farooq and Sarwar2011; Marquez-Ramos and Mourelle Reference Marquez-Ramos and Mourelle2019 among others).
According to the Mediterranean Strategy on Education for Sustainable Development (MSESD) Report (2014), the Mediterranean is regarded as an exceptional eco-region not only due to its geographical characteristics, but also the fact that it brings countries and peoples together from different levels of economic and social development, religions, languages, and cultures. Based on the vision presented in the report, education is addressed as a key factor for achieving sustainable development. Indeed, the main objective of the MSESD is to integrate sustainability into education policies and systems in the Mediterranean region while promoting a holistic and transformative approach to education that fosters sustainable development values, knowledge, skills, and behaviours among learners and educators. The strategy also seeks to promote dialogue and cooperation among Mediterranean countries on ESD and to mobilize resources to support ESD initiatives. Given the explicit role of education in the development process, the Mediterranean seems to be an interesting region to study.
Despite a variety of single-country cases from the region,Footnote a we still do not know how education affects growth in an encompassing sample. Therefore, the purpose of this study is to investigate the impact of education as a channel for human capital accumulation on two different growth concepts in a panel group of eight Mediterranean countries over the period 2000–2017.
Apart from the aforementioned motivation, the contribution of this study to the existing literature is two-fold. First, this study does not only investigate the impact of education on output growth, it also focuses on the impact that it has on inclusive growth. Inclusive growth is a broader concept that considers the dependence between macroeconomic and microeconomic dynamics and economic growth.Footnote b Experiencing higher output growth is not a sufficient achievement given the high level of extreme poverty and therefore could take place without a poverty reduction (Mulok et al. Reference Mulok, Kogid, Asid and Lily2012). Marquez-Ramos and Mourelle (Reference Marquez-Ramos and Mourelle2019) also discuss that increasing the level of education is not necessarily associated with higher economic growth in countries where underemployment arises. Given the high unemployment rates in Mediterranean countries, one would expect education level to have a higher impact on inclusive growth in the region. In addition, education can also help promote social mobility and reduce inequality by providing opportunities for individuals to improve their socioeconomic status as well as greater social cohesion by reducing discrimination and promoting understanding among different cultural and ethnic groups.
Second, we also wish to contribute to the existing literature on an empirical front. Unlike the literature on multi-country studies,Footnote c this study employs the Augmented Mean Group (AMG) to take the possible heterogeneity and cross-sectional dependence into account.
The structure of this article is as follows: theoretical framework, literature survey, data and model presentation, empirical methods and results, policy discussions, and concluding remarks.
Theoretical Framework
The human capital concept has attained a place and provided great convenience to the economic literature since the late 1950s. Influential works (Mincer Reference Mincer1958; Becker Reference Becker1962, Reference Becker1964; Schultz Reference Schultz1961; Denison Reference Denison1962; Nelson and Phelps Reference Nelson and Phelps1966) have integrated the role of education in the accumulation of human capital for economic growth. Within these models, human capital has contributed to improved labour force productivity and subsequent growth in national income. It has assumed that education’s contribution to growth is attributable to several separate but interdependent functions (Daren Reference Daren2017).
With the proliferation of Neoclassic economists, the concept has gained popularity. Economic growth theories have highlighted the role of human capital and the different other factors through which it may impact economic performance. The main theoretical approaches focusing on the relationship between human capital and economic growth have been Solow’s augmented neoclassical approach (see Mankiw et al. Reference Mankiw, Romer and Weil1992) and the endogenous growth theories of Romer (Reference Romer1990) and Lucas (Reference Lucas1988). Solow’s (Reference Solow1957) growth model extends the basic production function by adding human capital to the aggregate production function (Holland et al. Reference Holland, Liadze, Rienzo and Wilkinson2013).
Finally, supplements by the endogenous growth theories have corrected the concept’s shortcomings, and human capital has been fully integrated into the production functions as a crucial driver. Theories of endogenous growth regard education and knowledge as fundamental determinants of economic growth and assign a crucial role to education. The production of knowledge by the education sector results in self-reliant economic growth. Because the marginal return of this new factor (human capital) does not decrease. The endogenous growth models such as Lucas (1998), Romer, (Reference Romer1990), and Aghion and Howitt (Reference Aghion and Howitt1998) have included education in demonstrating its role in increasing the economy’s innovative capacity for innovation by developing new ideas and technologies (Holland et al. Reference Holland, Liadze, Rienzo and Wilkinson2013; Monteils Reference Monteils2002; Hanushek and Wößmann Reference Hanushek, Wößmann, Peterson, Baker and McGaw2010).
A high level of per capita income per se is not adequate for sustainable development in a country. Despite rising production across the globe, the increasing number of countries where social problems are of prime importance requires a proper framework in which the relationship between economic growth and human development is better explained (Delhey and Steckermeier Reference Delhey and Steckermeier2020; Easterlin and Angelescu Reference Easterlin, Angelescu, Land, Michalos and Sirgy2012; Skidelsky and Skidelsky Reference Skidelsky and Skidelsky2013 among others).
Education is an effective tool that triggers a number of channels in an economy, all of which can affect sustainable development. Education is a fundamental component of the process of economic growth through the use of subjective knowledge (Monteils Reference Monteils2002). According to Nelson and Phelps (Reference Nelson and Phelps1966), education can facilitate the diffusion and transmission of knowledge needed to understand and process new information and successfully implement new technologies devised by others, which promotes economic growth. Education can increase the economy’s innovative capacity, and new knowledge about new technologies, products, and processes encourages economic growth (Hanushek and Wößmann Reference Hanushek, Wößmann, Peterson, Baker and McGaw2010). First of all, education increases labour productivity by increasing access to new information and/or knowledge on a specific issue (Rosenzweig Reference Rosenzweig2010). Lucas (Reference Lucas1988) shows that education and the creation of human capital could be responsible for both the differences in labour productivity and the overall level of technological progress in an economy (Daren Reference Daren2017). Notice that the more educated labour ends up with better-implemented technological advances (Odit et al. Reference Odit, Dookhan and Fauzel2010). Therefore, a well-designed and qualified education system promotes economic development by increasing productivity. Revoredo and Morisset (Reference Revoredo and Morisset1999) state that education is the most efficient factor affecting the life-cycle income. In addition, the unemployment risk also decreases as the level of education increases (Zimmer Reference Zimmer2016). On the other hand, the contribution of education to welfare is essential given the positive externalities for individuals or societies (Hall Reference Hall2006).
A Brief Literature Survey: The Mediterranean Case
Although the existing literature on the nexus between education and economic growth is well-documented, the number of studies looking into this issue in the case of Mediterranean countries is relatively limited. Jaoul (Reference Jaoul2004) finds that the impact of higher education on the gross domestic product is positive in France whereas no significant relationship is detected in Germany. Another study in the case of France by Ozatac et al. (Reference Ozatac, Taspinar, El Rifai, Eren and Procházka2018) indicates unidirectional causality running from education to economic growth. Tsamadias and Prontzas (Reference Tsamadias and Prontzas2012) show for Greece that education has a positive and statistically significant impact on economic growth, while Granger causality results reveal evidence of unidirectional long-term and short-term causality from higher education to economic growth. Radić and Paleka (Reference Radić and Paleka2020) reveal a unidirectional causality from higher education expenditure to economic growth in the short-run in Croatia. Popović et al. (Reference Popović, Erić, Stanić and Krajišnik2019) indicate that education affects economic development positively in Bosnia and Herzegovina. Yurtkuran and Terzi (Reference Yurtkuran and Terzi2015) report for Turkey that the causality runs from economic growth to the number of university graduates while the number of vocational high school and high school graduates Granger causes economic growth. Guma (Reference Guma2014) finds a long-term and uni-directional causal relationship from higher education to GDP growth in Albania. Jihène (Reference Jihène2013) fails to find a significant relationship between economic growth and higher education in Morocco and Tunisia. Another study in the case of Morocco by Hadir and Lahrech (Reference Hadir and Lahrech2015) report that a high level of education is essential in the growth process. Qutb and El Shennawy (Reference Qutb and El Shennawy2016) find for Egypt that education has a significant positive impact on economic growth in the long-run. In the case of Algeria, Mekdad et al. (Reference Mekdad, Dahmani and Louaj2014) and Becherair (Reference Becherair2014) find that education is positively associated with economic growth while the results of Boutayeba and Ramli (Reference Boutayeba and Ramli2019) indicate no causality between education and economic growth. In five different groups of EU countries in which most of the EUROMED countries are included, Son et al. (Reference Son, Noja, Ritivoiu and Tolteanu2013) address the importance of education in economic growth in the EU.
Table 1 summarizes the aforementioned studies with a special focus on the Mediterranean countries by variable definition, which attempts to emphasize the shortcoming of a broader growth definition to represent sustainable development in measuring the role of education. With the exception of Tsamadias and Prontzas (Reference Tsamadias and Prontzas2012), previous studies use output growth as the dependent variable. However, no previous study in the case of the EUROMED region has investigated the impact of education on inclusive growth, to the best of our knowledge.
Table 1. Previous studies by variable selection.
Overall, the essential inference from the proceeding discussions which have motivated this study can be given as follows: the role of education in income growth is well-documented in the region whereas the role of education in inclusive growth has not yet been addressed.
Model Specification and Data
This study investigates the impact of education on economic growth in a panel group of eight Mediterranean countries over the period 2000–2017. The Euro-Mediterranean Partnership (EUROMED) countries included in the sample are the following: Croatia, Cyprus, France, Greece, Israel, Malta, Spain and Turkey.Footnote d The production function derived from the endogenous growth theory (Lucas Reference Lucas1988; Romer Reference Romer1990; Grossman and Helpman Reference Grossman and Helpman1991, among others) is described as follows:
$$y = f\left( {k,h,t} \right)$$
where y denotes output per capita, k the capital stock per capita, h the human capital accumulation, and t the technology level. Given the crucial role of human capital on inclusive growth addressed by previous studies, we also wish to examine the impact that education has on inclusive growth, which can be described as follows:
$${y^{\rm{*}}} = f\left( {{k^{\rm{*}}},h,t} \right)$$
where y* denotes output per worker and k* the capital stock per worker. Growth (y) is measured by per capita GDP in constant 2010 US dollars. Following the studies by Raheem et al. (Reference Raheem, Isah and Adedeji2018), Oyinlola and Adedeji (Reference Oyinlola and Adedeji2019) and Kouton (Reference Kouton2020), inclusive growth (y*) is represented by output per worker, which is measured as per worker GDP in constant 2010 US dollars. Capital stock per capita (k) is given by per capitaFootnote e Gross Fixed Capital Formation in 2010 US dollars while capital stock per worker (k*) is given by per worker Gross Fixed Capital Formation in 2010 US dollars.Footnote f Education is taken as the source of human capital and measured by mean years of schooling. Finally, technology is defined by research and development expenditures as a share of GDP. All variables depending on annual observations spanning from 2000 to 2017 are converted into natural logarithms in order to interpret the slope coefficients as elasticities. Data were collected from the World Development Indicators (World Bank 2021) database except for education, which was obtained from the United Nations Development Programme Human Development Reports (2021).
Panel data forms of the functions given in equations (1) and (2) can be shown as follows:
$${y_{i,t}} = {\alpha _0} + {\alpha _1}{k_{i,t}} + {\alpha _2}{h_{i,t}} + {\alpha _3}{t_{i,t}} + {\varphi _i} + {\delta _t} + {\varepsilon _{i,t}}$$
$${y^{\rm{*}}}_{i,t} = {\beta _0} + {\beta _1}{k^{\rm{*}}}_{i,t} + {\beta _2}{h_{i,t}} + {\beta _3}{t_{i,t}} + {v_i} + {\theta _t} + {\varepsilon _{i,t}}$$
where i denotes the country (i=1, …, N), t the time period (t=1, …, T).
$${\varphi _i}$$
and
$${v_i}$$
represent country-specific effects while
$${\delta _t}$$
and
$${\theta _t}$$
represent time effects.
$${\varepsilon _{i,t}}$$
is the random error term.
Table 2 presents the descriptive statistics of the data. Notice that the mean values of the variables described as per capita and per worker are close to each other. Realize also that the variable that has the greatest standard deviation is the technology level, which is quite high compared with the other variables.
Table 2. Descriptive statistics.
Methods and Findings
First generation applications in panel data econometrics are based on the assumption of cross-sectional independence. If observations are dependent across individuals, however, estimators which depend on this might be biased (Hsiao et al. Reference Hsiao, Pesaran and Pick2012). Therefore, one of the major concerns in current panel data econometrics is the possibility of cross-sectional dependence (Sarafidis and Wansbeek Reference Sarafidis and Wansbeek2012; Topcu and Çoban Reference Topcu and Çoban2017). Following the recent trends, we initially apply Pesaran’s (2004) cross-sectional dependence (CD) test as a preliminary step of the empirical framework, which can be shown as follows:
$$CD = \sqrt {{{2T} \over {N(N - 1)}}} \left( {\sum\limits_{i = 1}^{N - 1} {\sum\limits_{j = i + 1}^N {\mathop {\rho_{ij}}\limits^ \wedge } } } \right)$$
where T denotes the time period, N the number of countries, and
$$\rho $$
the bilateral correlation of the residuals. Panel A of Table 3 indicates the results of Pesaran’s (2004) CD test. Panel A of Table 3 indicates that the null hypothesis of cross-sectional independence is refused at the 1% level.
Table 3. Cross-section dependency and unit root results.
CIPS test is applied with a lag and constant. *** and ** represent 1% and 5% significance levels respectively.
It is necessary to determine the order of integration of variables in the system before conducting regression analysis to avoid inconsistent results and the problem of spurious regression that may arise if non-stationary data are used (Topcu and Aras Reference Topcu and Aras2017). Moreover, first generation unit root tests would produce inconsistent results given the presence of cross-sectional dependence. We, therefore, implement a widely used second generation test. The following equation represents the mathematical form of the CIPS test proposed by Pesaran (Reference Pesaran2007):
$$CIPS(N,T) = {N^{ - 1}}\sum\limits_{i = 1}^N {{t_i}(N,T)} $$
where CIPS(N,T) is an individual augmented form of the IPS test proposed by Im et al. (Reference Im, Pesaran and Shin2003) while, t i (N,T) is an individual augmented form of the DF statistic, proposed by Dickey and Fuller (Reference Dickey and Fuller1979). Unit root results illustrated in Panel B of Table 3 reveal that all variables contain a unit root, with the exception of education.
Once the presence of cross-sectional dependence and the combination of the different integration levels among the variables are determined, the choice of the estimation procedure is of particular importance. In addition, the estimation procedure is expected to allow for heterogeneity as well as addressing cross-sectional dependence.
There are three empirical estimators that relax the assumption of parameter homogeneity across panel members in the panel data econometrics: the Mean Group estimator proposed by Pesaran and Smith (Reference Pesaran and Smith1995), the Common Correlated Effects Mean Group estimator developed by Pesaran (Reference Pesaran2006), and the Augmented Mean Group estimator introduced by Bond and Eberhardt (Reference Bond and Eberhardt2013), Eberhardt and Teal (Reference Eberhardt and Teal2010), and Eberhardt and Bond (Reference Eberhardt and Bond2009). The AMG estimator is considered an alternative to the Pesaran (Reference Pesaran2006) CCEMG estimator with macro production function estimation in mind. In addition, the fact that the AMG estimator is more efficient than the MG estimator, especially when the number of individuals is small or the time period is short, makes it a better choice for small samples or short time-series data.Footnote g Given this information and that recently proposed mean group estimators are robust to cointegration and the combination of I(0) and I(1), this study employs the Augmented Mean Group (AMG) estimator in order to estimate the slope coefficients.
The AMG estimator extends equations (3) and (4) by including the cross-section means in order to address cross-sectional dependence.
$${X_{it}} = {\alpha _{1i}}Y_{it}^{\rm{*}} + {\alpha _{2i}}{R_{it}} + {\delta _{1i}}{\overline X_t} + {\delta _{2i}}\overline Y_t^{\rm{*}} + {\delta _{3i}}{\overline R_t} + {\emptyset _i}{\hat w_t} + {v_i} + {\delta _t} + {\varepsilon _{it}}$$
$${M_{it}} = {\beta _{1i}}{Y_{it}} + {\beta _{2i}}{R_{it}} + {\gamma _{1i}}{\overline M_t} + {\gamma _{2i}}{\overline Y_t} + {\gamma _{3i}}{\overline R_t} + {\emptyset _i}{\hat w_t} + {v_i} + {\theta _t} + {\varepsilon _{it}}$$
where means for individuals
$\left( {{{\overline X}_t},{{\overline M}_t},\overline Y_t^*,{{\overline Y}_t},{{\overline R}_{t,\;}}} \right)$
represent the common factors while year dummy coefficients (
$${\widehat \omega _t}$$
) are gathered from a traditional regression. The AMG estimator is a two-stage approach in which unobservable common factors follow a common dynamic process. In the first stage, the within-group estimator is used to obtain the individual-specific effects. In the second stage, the estimated individual-specific effects are used to estimate the slope coefficients in the second stage. The AMG estimator includes additional regressors that capture cross-sectional dependence, such as lagged dependent variables and lagged first differences of the independent variables. The following equation describes the mathematical notation of the AMG estimator:
$$\widehat {{\beta _{AMG}}} = {N^{ - 1}}\mathop \sum \nolimits_{i = 1}^N {\widehat \beta _t}$$
where
$$\widehat {{\beta _{AMG}}}$$
is the average of the cross-section estimators.
Table 4 reports the results attained from the AMG estimator. A 1% increase in capital per capita increases growth by 0.237%, while a 1% increase in education level increases growth by 0.061%. On the other hand, a 1% increase in capital per worker leads to an increase in inclusive growth by 0.281% while a 1% increase in education level leads to an increase in inclusive growth by 0.303%. Despite a positive estimated coefficient, the impact of technology is statistically insignificant for each specification.
Table 4. Estimation results (AMG estimator).
Notes: Parameter estimates are calculated as sample averages. The models include a constant. CD-test is a test for cross-sectional dependence while the CIPS-test is for a unit root. Wald is the chi-square test for all slope coefficients equal to zero. CIPS test is applied with a lag and constant. Significance at the 1% and 5% levels is represented by *** and ** respectively.
For robustness purposes, we also employ the CCEMG estimator proposed by Pesaran (Reference Pesaran2006). The regression results reported in Table 5 indicate that the results are robust to the estimator.
Table 5. Estimation results (CCEMG estimator).
Notes: Parameter estimates are calculated as sample averages. The models include a constant. CD-test is a test for cross-sectional dependence while the CIPS-test is for a unit root. Wald is the chi-square test for all slope coefficients equal to zero. CIPS test is applied with a lag and constant. Significance at the 1% and 5% levels is represented by *** and ** respectively.
Policy Implications
As indicated in the MSESD (2014), education strengthens the judgement capacity of individuals, groups, communities, organizations and, as a consequence, countries to make choices for the sake of sustainable development. It also builds up innovative knowledge which improves the quality of life. Therefore, relevant authorities in Mediterranean countries should pay special attention to education policy and planning. Education governance reforms that attempt to improve educational quality will assist policymakers to achieve sustainable development goals. As noted by Odit et al. (Reference Odit, Dookhan and Fauzel2010), schooling may reduce the cost of acquiring know-how and, therefore facilitate the adoption of innovations.Footnote h Thus, policymakers should give priority to the policies devoted to increasing educational accountability and transparency, and reducing educational inequalities. In addition, they should follow the most recent trends in educational technology closely. Notice that greater awareness and empowerment are expected to make the region a better place to live in terms of economic outcomes as well. Policies towards a knowledge-based economy are likely to bring more employment opportunities, which, in turn, increase prosperity across the region.
The findings obtained from this study do not only address straight implications such as allocating more resources towards education, including improving the quality of schools, hiring and training teachers, and providing access to technology and learning materials. In the inclusive growth process, policymakers should work to reduce barriers to education, such as cost, geographic location, and cultural or language barriers. This can include providing scholarships or financial assistance to low-income students, expanding access to online or distance learning, and promoting multilingual education. In addition, because of the well-documented evidence that investing in early childhood education has a positive impact on long-term educational outcomes and future success in life, policymakers should prioritize funding and resources towards improving access to high-quality early childhood education. Encouraging public–private partnerships is of great concern to reap the benefits of education in the inclusive growth process. To this end, policymakers should collaborate with private sector partners to improve the quality and relevance of education to the needs of the labour market, while also ensuring that education remains accessible and affordable for all. Last but not least, policymakers in the region should support innovation and entrepreneurship in the education sector, such as promoting new models of education delivery, encouraging the development of educational technology, and supporting social entrepreneurs who are working to address educational challenges in their communities.
The comparison of growth specifications is worth mentioning as well. Empirical results demonstrate that the main driver of output growth is physical capital, whereas inclusive growth is primarily triggered by human capital. The results obtained from the AMG estimator also reveal that the effect of physical capital stock on output growth is roughly comparable with that on inclusive growth. The evidence that technology does not have a statistically significant effect on both growth concepts might result from the fact that the majority of the Mediterranean countries in the sample are not better able to develop high value-adding technologies.
Conclusion
Adopting an endogenous growth model, this study investigates the impact of education as a channel for human capital accumulation on two different growth concepts in a panel group of eight EUROMED countries over the period 2000–2017. Empirical results obtained from the AMG estimator indicate that each determinant has a positive and statistically significant impact on both growth proxies, with the exception of the technology variable. In terms of our central focus, we find that the impact of education on inclusive growth is five times higher than that on output growth. Not surprisingly, this finding clearly emphasizes the role of education in building up the human capital stock in the EUROMED countries.
Considering the limitations of the existing study, some recommendations could be addressed for future studies. Potential subjects for future research could be the extension of the time period and the sample with the inclusion of more Mediterranean countries (given data availability). In particular, incorporating African EUROMED countries into the sample would be better able to reflect heterogeneity in the region. In addition, As noted by Liao et al. (Reference Liao, Du, Wang and Yu2019), different countries have different levels of investment in education due to differences in economic levels and local government policies. Thus, a comparison between middle- and high-income countries may provide useful insights into developing a comprehensive and common education policy across the region. Finally, country–time effects, such as the importance of institutional quality or corruption, would also be worthy of investigation.
About the Author
Ebru Topcu has been working as Associate Professor at the Department of Economics in Nevsehir Haci Bektas Veli University, Turkey. Ebru’s main research interest is development economics.
