THE DUAL ROLE OF HUMAN CAPITAL FOR ECONOMIC GROWTH: AN EMPIRICAL INVESTIGATION USING PANEL DATA FOR DEVELOPING COUNTRIES

Transcription

1 THE DUAL ROLE OF HUMAN CAPITAL FOR ECONOMIC GROWTH: AN EMPIRICAL INVESTIGATION USING PANEL DATA FOR DEVELOPING COUNTRIES Miethy Zaman* Abstract The paper investigates the effect of human capital on economic growth. Under a hybrid specification, for a sample of 75 developing countries, using panel data for the period , the empirical outputs provide considerable evidence for a dual role of human capital in determining economic growth. Human capital aids in economic growth in accumulation term and as a facilitator of technology diffusion. The sensitivity of the findings is evaluated by subjecting the estimations to different sub-samples and using different technology leaders. Some specific problems relating to the measures of human capital and technology gap are also outlined. Index Terms: developing countries, economic growth, human capital, hybrid model *The author is with Department of Economics North South University Bangladesh Phone: miethyzaman@northsouth.edu ; miethy.zaman@gmail.com 1. Introduction Human capital has been considered as an important determinant of economic growth both theoretically and empirically. Aghion and Howitt (1998) have categorized the modelling of human capital as a factor of growth into two alternative frameworks. The first is the standard approach based on the Lucas (1988) and also shared by the neoclassical growth theory which treat human capital as an ordinary input in the aggregate production function. Under this structure, economic growth is determined by the accumulation of human capital. The alternative framework, first proposed by the Nelson and Phelps (1966) suggests that human capital aids in domestic innovation and adaptation of foreign technologies. Under this structure, a country s economic growth is determined by the level of human capital, and it suggests a higher stock enhances productivity through faster innovation and through speeding technological diffusion throughout the economy. With many studies finding a weak relationship between the accumulation of human capital and economic growth, the Nelson-Phelps approach to modelling human capital for growth started to gain importance both theoretically and empirically. Benhabib and Spiegel (1994) and Engelbrecht(2003) confirm some considerable evidence for human capital s role in assisting technology diffusion. However the combined effect of human capital for domestic innovation and assisting technology diffusion is found to be negligible. Following this argument, few papers Engelbrecht(2003) and Papageorgiou(2003) focused on empirically testing a hybrid model which incorporates both the growth and level of human capital for economic growth. This paper uses Benhabib and Spiegel (1994) as a starting point and following Engelbrecht(2003), it uses a structural specification to examine the possibility of a dual role of human capital in determining economic 435

2 growth for a group of developing countries. Empirical estimation shows that using IV estimation technique for panel data, the hybrid model gives considerable evidence for both the level and accumulation of human capital in explaining economic growth. At first the technology diffusion fails to give a positive relationship with economic growth when USA and other regional countries are used as the technology leader. But when OECD countries collectively are considered as the technology leader, estimation generates a positive statistical relationship between diffusion and growth. Thus the result reinforces the dual role of capital for economic growth. This paper is organised as follows. Section 2 reviews the existing theory and evidence on the technology catchup. Section 3 describes and outlines the standard methodologies used by Benhabib and Spiegel (1994).It also discusses the data, data limitations and estimation procedure used by us for the purpose of our study. Section 4 presents and analyses the estimated results and finally section 5 concludes the paper. 2. Human Capital and growth: Theory and evidence Human capital has been considered to be an important determinant of economic growth in many empirical studies. Early literatures have mostly focused on human capital s role as an accumulation factor input. Mankiw et al. (1992) and the Barro (1991) based their studies on the convergence framework. The work of Mankiw et al.(1992) studies the role of human capital under the augmented Solow model framework that incorporates accumulation of human capital in the production function. Following the Mankiw et al. (1992) framework, Islam (1995) extends the study using a different measure of human capital. The author use the Barro and Lee (1993) data on the educational attainment instead of the enrolment rates used in the Mankiw et al. (1992) paper. In addition to the cross-sectional and pooled regressions, Islam (1995) also implements panel techniques that allow taking into account the cross country differences in the aggregate production function. Using different estimation techniques and various sub-samples, the author could not generate a convincing positive relationship between human capital and growth. Contrary to the previous theoretical approach, Benhabib and Spiegel (1994) base their study on a growth accounting framework. They investigate the link between human capital and economic growth using the generic Cobb-Douglas production function in which income per capita is a function of labour force, physical and human capital and the unobserved level of technology. For a sample of 122 countries for the period , and for cross sectional regression, the result depicts that the accumulation of human capital enters the equation insignificantly or with a negative coefficient in determining economic growth. Nelson and Phelps (1966) pointed that treating human capital simply as a factor input in the production function is a misspecification of the relationship between growth and human capital. They focus on human capital as a facilitator of domestic innovation and technology catch-up. The theory implies that the educated people are the bearers of human capital, which aids in domestic innovation in technology development and also helps to adapt the technology developed elsewhere. The point to note here is that the catch-up is not to some exogenously growing technology but to technology of the leading country. For the Nelson-Phelps approach, Benhabib and Spiegel (1994) analysed a structural specification to allow the alternative mechanism through which human capital affects growth. Using ordinary least squares (OLS) estimation and accounting for cross-sectional heteroskedasticity, for a sample of 78 countries, the technology diffusion term is statistically significant with a positive sign when USA has been represented as the technology leader. 436

3 Engelbrecht (2003) paper using the Benhabib and Spiegel (1994) structural equation display a very small combined impact of human capital on growth for the OECD countries. The empirical evidence seems to support the micro-macro paradox claim by Pritchett (2001) which implied poor empirical evidence for the human capital-growth relation. The paradox is counterviewed by Engelbrecht (2003) as he proposes that the findings are possible indications of model misspecification. The author then adopts an alternative hybrid model which integrates both the major approaches of human capital growth theory suggested by Aghion and Howitt (1998).Thus the econometric framework takes into account both the level effect and accumulation effect on growth, where the former represents the impact for its role in technological diffusion and domestic innovation and the later represents its effect through the accumulation of human capital. For OECD countries using cross sectional data, estimation with OLS shows that for the catch up component and the human capital growth, the coefficients are positively significant. The findings support the author to conclude that there is a conceivable need to construct hybrid models to understand the comprehensive effect of human capital on growth. Even though Engelbrecht( 2003) did not find any evidence of combined impact of technology diffusion and domestic innovation, his findings strongly support the Nelson-Phelps contribution of the technology diffusion and the catch up phenomenon along with the standard accumulation theory of human capital. Papageorgiou (2003) also uses an alternative model specification to integrate both the accumulation and the absorption theories of human capital. Following the claim by Baumol et al (1989) concerning the difference in contribution to growth by the different level of educational attainment, Papageorgiou (2003) also used subcategories of education data to study their impact on economic growth. Regression estimates from the study present considerable support for the dual role of human capital and the author also argues that the alternative specification performs better relative to the specifications incorporating only either one of the theories. For instance, the study found significant positive effect for the accumulation of human capital which is contrary to most of the findings in the literatures that have solely used the human capital growth in the standard growth accounting process. For sub-sample estimations, Papageorgiou (2003) finds evidence for differences in the relative importance of human capital as a factor input and as the facilitator of innovation and technology adaption. In addition, the estimation reveals that human capital contribution to technology diffusion is proportional to country s income. Papageorgiou explains this as a possibility that countries may need to reach a threshold economic development first to benefit from the technology diffusion. The domestic innovation term based on the empirics seem to have a significant effect only on the high-income countries, whereas the human capital accumulation is significant only for the low-income countries. 3. Model, data and methodology This section outlines the standard methodologies used in the previous empirical literatures and how they are extended for the empirical part of this study. The section starts with the description of the growth accounting methodology used in the Benhabib and Spiegel (1994) paper which uses human capital as a factor of production. Then from the same paper the alternative model is shown which accounts for the Nelson-Phelps approach which emphasized on the technology catch up. In section 3.3 shows how the two models can be merged to derive the hybrid model following Engelbrecht (2003). The model extensions and addition implemented by this paper are discussed in section

4 3.1.1 Growth regression model with human capital as a factor of production In this methodology, human capital enters the production function as a factor input and the human capital s accumulation rate effects output growth. In the growth accounting framework, output per capita is specified as a function of following input factors: labor force, physical and human capital. Assuming a Cobb-Douglas representation, Benhabib and Spiegel (1994) show at time t, the aggregate production function can be explicitly expressed as: (1) Where Y is output, K is physical capital, L is labor force, H is human capital and A is the level of technology which grows at an exogenous rate. To express the growth relationship, Benhabib and Spiegel (1994) have taken the log differences of the above model which relates the log differences of output to the log differences of the factor inputs. It is expressed as: ( ) ( ) ( ) ( ) ( ) ( ) ( ) Based on this specification, human capital effects the output growth through its rate of accumulation Growth regression model using Nelson-Phelps approach Benhabib and Spiegel (1994) has also used the alternative specification using the Nelson-Phelps hypothesis of modelling human capital. Nelson and Phelps (1966) argued that human capital represented as a factor of production is a misspecification of the growth process. Rather they suggest countries endowed with high stock of human capital are good innovators and are quicker to adapt to new ideas and technologies. So a higher level of human capital would speed up the technological diffusion process. Hence the theory strays away from the notion of exogenous rate of technological progress and allows it to depend on the level of human capital. Benhabib and Spiegel (1994) using this approach develop the alternative growth model in which total factor productivity growth relies on: (i) the level of human capital and (ii)an interaction term allowing for the catching up to the leading country s technology. Thus for a specific country i, the growth rate of total factor productivity or the technological progress at time t is represented as: ( ) ( ) ( ) ( ) * ( ) ( ) ( ) + ( ) Where ( )is the endogenous growth rate for the innovation factor and the second component on the right hand side of the equation is the catch-up term. The equation (3) implies that a country with a high level of education has the potential to overtake the leading country s technology irrespective of the difference between each country s initial levels of technology. For a structural model following directly from the theory, Benhabib and Spiegel (1994) used the Cobb- Douglas production function: 436

5 ( ) ( ) Where, is the per capita income, is the physical capital, is the labour force and is the level of technology which is a function of human capital.in similar manner to (2), the growth equation is expressed by taking the log differences: ( ) ( ) ( ) ( ) ( ) ( ) (5) Contrary to (2) in which change in the level of technology or the growth accounting residual is exogenously determined, in (5) the first component specified as the total factor productivity growth now incorporates the Nelson-Phelps approach. Hence the change in the level of technology now depends on the level of human capital for the ability to innovate and on an interaction term that relates human capital and the technological lag for the catch-up factor. For the difficulty of observing technology, Benhabib and Spiegel (1994) used output per capita to roughly proxy for the level of technology. Thus the structural specification for the growth of total factor productivity for a country is as follows ( ) ( ) * ( ) + ( ) Where c is the exogenous technological progress, is the domestic innovation term and ( ) is the interaction term reflecting the technology diffusion from abroad. Rearranging (6) gives: ( ) ( ) ( ) ( ) (7) This is then inserted in (5) to yield: ( ) ( ) ( ) ( ) ( ) ( ) ( ) From the above equation, it can be seen that the impact of the stock of human capital on growth would depend on the cumulative effect of the domestic innovation and the technology diffusion terms The Hybrid Model: Papageorgiou (2003) and Engelbrecht (2003) stressed upon the hypothesis of multiple roles of human capital for determining economic growth. It implies that human capital is an important factor of growth for its role as an input in the production function and as an input for technology innovation and technology diffusion. If there exist more than one mechanism through which human capital effects growth, modeling based on only either one of the approaches would lead to omitted variable bias. To account for both the theories, Engelbrecht (2003) 437

6 constructed a hybrid model allowing human capital to enter as a production factor of final output and as a basis of technology innovation and diffusion. So starting with the basic Cobb-Douglas representation: ( ) ( ) The notations remain the same as the above equations. Here human capital enters the aggregate production function both as a factor input and as a determinant of technological progress. Analogously taking the log differences, the econometric framework can be represented as ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) And inserting the specification of the total factor productivity growth (7) gives the hybrid model: ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) Growth Models with panel specification and per capita capital The above models are extended into panel framework. Also with Y as the GDP per capita, instead of taking physical capital and labor as separate variables, GDP per capita is represented as a function of per capita capital, human capital and technology. So under the Cobb-Douglas representation, for the first approach with factor input human capital, nominal income of country i at time t can be expressed as and in similar manner: ( ) and ( ) for the Nelson-Phelps growth model and the hybrid model respectively, where now is the per capita capital at period t for country i. Hence the panel specification for our growth models with log differences are as follows: - With the accumulation of Human Capital: ( ) ( ) ( ) ( ) ( ) (12) -With the stock of Human Capital: ( ) ( ) ( ) ( ) ( ) (13) -With both the accumulation and stock of human capital: ( ) ( ) ( ) ( ) ( ) ( ) ( ) 438

7 Equation (12), (13) and (14) are the econometric specifications used in this paper for the empirical estimations. The regression technique involves using IV estimation technique. Caselli et al (1996) and Krueger and Lindahl(2001) proposed that the physical capital stock is potentially endogenous in the growth equations. The choice variable physical capital is derived from the rate of investment, and economic growth is likely to influence the optimal rate of investment which could lead to physical capital being possibly endogenous. To take into consideration the endogeneity problem, Instrumental Variable technique is used as the primary estimation technique to correct the potential bias using lagged logged differences of physical capital and output as instruments Data The estimations in this research involve a total sample of 75 developing countries for the period between 1980 and Theoretically human capital has not been defined in a specific manner. It is an extensive theoretical concept which can be summarised as the skills and resources embedded in each individual. Schütt (2003) describes human capital as the knowledge, skills and attributes of people that are pertinent for economic activity. Empirical studies have mostly resorted to using formal education as a rough proxy of human capital for the lack of availability of data encompassing a comprehensive measure of human capital. This study used the Barro and Lee (2010) dataset on educational attainment for the measure of human capital. Data on GDP per capita and labour force are retrieved from World Development Indicators (2011) and Penn World Table (7.1) respectively. Of considerable interest is the data on the stock of physical capital. It needs to be constructed from the investment flow data. The main references for the construction of the stock of capital are the papers by Bernanke and Gurkaynak( 2001) and Safdari et al. (2010). Bernanke and Gurkanayak (2001) retrieved the investment flow data from the Penn World Table (6.0) to derive the capital stock by using the perpetual inventory calculation method. Based on their approach, Safari et al. (2010) extended five-yearly data which is used for the estimations for this study. Finally, following Benhabib and Spiegel (1994), the ratio ( ) is used as an approximate for the technology gap, where is denoted as the GDP per capita of the designated technological leader i at time t respectively. For the technological leader, three different types of leaders are selected. Firstly, following previous literatures, USA is as the global leader. Then regional leaders South Africa, Japan and USA are selected for the sub sample Africa, Asia and South America respectively. And finally the OECD countries are taken cumulatively as the leader and the weighted average of the GDP per capita is calculated to proxy for the technology. 3.3 Specific limitations for the growth regressions: Before discussing the empirical results, it is important to highlight some of the specific problems related to the data, proxy measures and econometric specifications Measurement Errors of Human capital: 439

8 Measurement errors in the data of the proxy for human capital are pervasive. The existing literatures have stressed upon the different possible errors in the measurement of human capital. Earlier studies use enrolment rates or literacy rates which are argued as inappropriate proxy measures by Pritchett (2001). With educational system generalised as the source of acquiring knowledge, human capital stock is mostly characterized by schooling attainment data. But whether this data constitutes to the rate of investment in human capital is questionable. Also, as pointed by Aghion and Howitt (1998), the proxy should be determined by the type of human capital that goes into the production function. Other alternative measures, for example, research and development expenditure or public expenditure on human capital may be more suitable to represent the investment in human capital in the aggregate production function. Tiruneh and Radvansky (2011) in their paper use research and development (R & D) expenditure as one of the alternative measure for human capital accumulation. For a sample of European countries, they found R & D expenditure to be positively related to economic growth. This study could not extend the investigation due to the lack of data available for the R & D expenditure of the developing countries. The educational attainment data primarily focuses on the quantity measures of human capital rather than the quality measures. Due to the heterogeneity of the countries in the data, there is a high likelihood of variation in the quality of education across the different countries that can cause differences in the stock of human capital. Hanushek and Kimko (2000) acknowledge the importance of quality measure of human capital as a key for determining a positive liaison between human capital and growth. Instead of using schooling data, they have used internationally comparable mathematics and science test scores to reflect the qualitative difference in the labour force. Estimated results show a strong positive correlation between growth and the quality measure. However the problem with the Hanushek Kimko data is it only considers the qualitative aspect of human capital not the quantitative measures. Hence for the time being, the quality adjusted schooling data is preferable for empirics Proxy for technology catch-up term: As discussed previously, Benhabib and Spiegel (1994) use per capita GDP as a rough approximation for the technology gap. Following them, Engelbrecht(2003) also applies the same approach to determine the technology diffusion factor in the hybrid model. Both the papers implement cross-sectional regression approach using data from the period between 1965 and 1985.They have assumed USA to be the technology leader and thus used 1965 s GDP per capita of USA to proxy for the technology gap. However in the context of this paper, which is using data period , USA s GDP per capita is not the highest in 1980 among the 25 OECD countries the study used to identify the technology leader. Also based on the notion of using highest GDP per capita to represent the technology leader, with the use of five yearly panel data, the catch-up component is operating at higher frequency which requires adjusting for for each period t. Clearly from the fig. 1, it can be seen the highest GDP per capita is represented by different countries GDP. For the periods 1980 and 1985 it is Switzerland and for the latter periods it is Luxembourg. GDP per capita may not be a perfect proxy for technology as there can be other factors that can enhances it; for instance, for Luxembourg, the high GDP per capita is possibly due to the financial services which account for a major portion of the nation s GDP. 440

9 Fig. 1 Year-wise GDP per capita With the view of USA being the worldwide leader of science and technology and keeping in parallel with the previous studies, this paper still represent USA as the technology leader and use its GDP per capita to approximate for the technology gap in the panel estimation framework. However, future research should be directed to this area to derive a more suitable proxy to represent the science and technology endowed in a country. The sample used in this paper contains developing countries from different regions of the world. When testing for such a diverse sample, it is important to stress on the fact that all the countries may not be catching up to the USA s technology as there could be evidence of possible growth clubs. It implies all the countries may not be converging to a single economy but rather sub-group of economies are converging or catching-up to one another or may not be converging towards any economy at all. Ben-David (1994) found some evidence for convergence clubs among the wealthier countries and among the low income countries. This issue has not been stressed upon in the previous literatures for the Nelson-Phelps catch-up factor. To take this into consideration, for estimations using the developing countries sub-samples, the study explores evidence of club convergence by taking different sub samples and regional technology leaders. Along with testing for club convergence, it is important to explore if these developing countries benefit from technology diffusion from a group of countries rather than from one single country. Few literatures have stressed on the importance of knowledge spill over from the west to the developing countries. To account for that, the study have also used OECD countries cumulatively as the technology leader to test for the catch up term. 4. Empirical Findings 441

10 For the empirical estimations, following previous studies we first evaluate the hybrid model using USA as the technology leader. Results in table 1 show that change in human capital and level of human capital are both significant under the hybrid model. The per capita capital is significant at 1% level. However the interaction term for the technology diffusion is insignificant. It can be seen that not only it is statistically insignificant; the magnitude effect of the technology diffusion is very minute. The results provide an indication that USA may not be the leader for technology diffusion from one nation to another. With the view of USA being the worldwide leader of science and technology, most previous literatures used USA to signify the leader. But conversely, the results are indicating that there should be other means for selecting the technology leader. Table 1- Estimated results using USA as the technology leader Dk * ( ) DH *** (.02473) H * ( ) H(Ymax/Y) 4.29e-06 (5.06e-06) Constant * ( ) Note: The dependent variable is DY. DX denotes the growth rate of variable X. The Standard errors are given in parentheses. (*), (**) and (***) denotes that the estimates is statistically significant at 1%, 5% and 10% level respectively. With insignificant results for technology diffusion with USA, we have looked into probable regional leaders for the developing countries. The developing countries are from different geographical regions and it can be possible that all the countries are not converging to one single country s technology but rather there can be regional technology leaders. Thus it implies that the technology diffusion takes places from the regional leader to the neighbouring less-developed countries. To empirically test for that, we have divided most of our sample into three sub-groups: Africa, South America and Asia and have taken South Africa, USA and Japan respectively as the technology leaders for each sub-group. 442

11 Table 2-Estimated results for sub-sample Africa using South Africa as the technology leader Dk ( ) DH ( ) H * ( ) H(Ymax/Y) -3.12e-06 (3.98e-06) Constant ( ) Note: The dependent variable is DY. DX denotes the growth rate of variable X. The Standard errors are given in parentheses. (*), (**) and (***) denotes that the estimates is statistically significant at 1%, 5% and 10% level respectively. Table 3-Estimated results for sub-sample South America using USA as the technology leader Dk ( ) DH ( ) H ( ) H(Ymax/Y) ( ) Constant ( ) Note: The dependent variable is DY. DX denotes the growth rate of variable X. The Standard errors are given in parentheses. (*), (**) and (***) denotes that the estimates is statistically significant at 1%, 5% and 10% level respectively. 443

12 Table 4-Estimated results for sub-sample Asia using Japan as the technology leader Dk ** ( ) DH ( ) H ( ) H(Ymax/Y) ( ) Constant ** ( ) Note: The dependent variable is DY. DX denotes the growth rate of variable X. The Standard errors are given in parentheses. (*), (**) and (***) denotes that the estimates is statistically significant at 1%, 5% and 10% level respectively. Results from tables 2,3, and 4 outline that when regional technology leaders are used, technology diffusion term fails to enter the model insignificantly in explaining economic growth for any one of the sub-samples. Not only it is insignificant, in the case for Africa and Asia, the parameter estimates for the catch-up term are also negative. The growth of human capital also fails to provide statistically significant effect on economic growth. Except for Africa, the level of human capital for domestic innovation is insignificant as well. Hence the estimations fail to provide substantial evidence for human capital effecting economic growth for the regional subsamples and thus fail to provide sign of club convergence. Getting insignificant catch-up term for all the estimations raises the question on the Nelson-Phelp model s importance in human capital for technology diffusion. However it is possible that selecting the wrong technology leader can lead to insignificant catch up term even though a country in endowed with substantial level of human capital. Few literature studies such as Coe et al (1997) and Engelbrecht (2002) stressed on the notion of technological diffusion from the west to the developing countries. It is pointed that many developing countries do not have adequate levels of technology for economic development, but they benefit from the technology development in the developed countries through diffusion. Thus provided a country has adequate level of human capital, it will be able to aid in economic growth through the catch-up factor. So instead of representing USA solely as the technology leader, OECD countries are taken cumulatively as the technology leader and the weighted average of the GDP per capita is calculated to proxy for the technology of the OECD countries. Estimated results are given in table

13 Table 5-Estimated results using OECD countries as the technology leader Dk * ( ) DH *** (.03714) H * ( ) H(Ymax/Y) ** ( ) Constant * ( ) Note: The dependent variable is DY. DX denotes the growth rate of variable X. The Standard errors are given in parentheses. (*), (**) and (***) denotes that the estimates is statistically significant at 1%, 5% and 10% level respectively. The empirical output shows that per capita capital is significant at 1% level. For the human capital variables, the parameter coefficients for growth of human capital and the domestic innovation term are statistically significant. The most important finding from this estimation is that now the technology diffusion term is statistically significant at 5% level of significance. The empirical finding is a clear indication the technology diffusion may not solely take place from USA even though the country is considered to be the global technology leader. As illustrated by Coe et al (1997), technology transmission can happen from the west cumulatively to the developing countries where the developing countries can aid from the technology provided they have adequate level of human capital. The significant parameter estimate for the interaction term provides a sign that the developing countries are benefitting from technology diffusion from the developed countries collectively. Hence the result shows human capital can aid in economic growth both as in accumulation term and in level term. A country endowed with high level of human capital can benefit from absorbing technology transmitted from the developed countries. 5.Conclusion The objective of this paper is to investigate the impact of human capital on economic growth. Motivated by Engelbrecht (2003), this study uses a hybrid specification which models both the accumulation and the level of human capital simultaneously to investigate whether there is a dual role of human capital in explaining growth. Empirical results show substantial evidence to support the fact that both accumulation and level of human capital are concurrently important for economic growth and the previous insignificant or negligible outputs are likely to be results of model misspecification. The other key finding of the study is that USA solely fails to provide evidence as the technology leader for technology diffusion. There was no considerable evidence for convergence clubs when regional sub samples are used with different regional leaders. Instead a positive relationship is derived when the technology gap is denoted as the weighted average GDP per capita of the OECD countries. So the results were indicative that the 445

14 human capital aids in economic growth through technology diffusion and the spill over takes place not from one single country but from the developed countries collectively. In summary, based on this study, the empirical evidence supports the hybrid specification that incorporates both the major approaches of modelling human capital indicated by Aghion and Howitt (1998) for determining economic growth. Further research should focus on studying the dual role of human capital with the use of better available human capital data and improved specification to support the findings of this paper. References [1]Aghion, P. & Howitt, P. (1998). Endogenous Growth Theory.MIT Press, Cambridge, MA. [2]Barro, R. (1991). Economic growth in a Cross Section of Countries. Quarterly Journals of 106(1), Economics, [3]Barro, R. & Lee, J-W. (1993). International Comparison of the Educational Attainment. Journal of Monetary Economics, 32(3), [4]Barro, R. & Lee, J-W. (2010). A New Data Set of Educational Attainment in the World. Working Paper 15902, National Bureau of Economic Research. [5]Baumol, W.J., Blackman, S.A. & Wolff, E.(1989). Productivity and American Leadership, MIT Press, Cambridge, MA. [6]Ben-David, D. (1994). Convergence clubs and diverging economies. CEPR Discussion Paper Series 922. [7]Benhabib, J. & Spiegel, M. (1994). The role of human capital in economic development: Evidence from aggregate cross-country data. Journal of Monetary Economics, 39, [8]Bernanke, B.S. & Gurkaynak, R.S.( 2001). Is growth exogenous? Taking Mankiw, Romer and Weil seriously. NBER Macroeconomics Annual, [9]Caselli, F., Gerardo, E. & Fernando, L. (1996). Reopening the convergence Debate: A new Look at Cross- Country Growth Empirics. Journal of Economic Growth, 1(3), [10]Coe, D.T., Helpman, E. & Hoffmaister, A.W. (1997). North-South R&D Spillovers. The Economic Journal, 39, [11]Engelbrecht, H. J. (2002). Human Capital and international Knowledge spillovers in TFP growth of a 1 [12]Engelbrecht, H.J. (2003). Human Capital and Economic Growth: Cross-Section Evidence for OECD Countries. Economic Record, 79. [13]Hanushek, E. & Kimko, D.(2000). Schooling, labour force quality, and the growth of nations. American Economic Review, 90, [14]Heston, A., Summers, R. & Aten, B. (2011). Penn World Table Version 7.1. Center For International Comparisons of Production, Income and Prices at the University of Pennsylvania. [15]Islam,N. (1995). Growth Empirics: A Panel Data Approach. Quarterly Journal of Economics, 110(4), [16]Krueger, A.B & Lindahl, M. (2001). Education for growth: Why and for whom? Journal of Economic Literature, [17]Lucas,R.(1988).On the Mechanics of Economic Development. Journal of Monetary Economics, 22, [18]Mankiw, G., Romer, D. & Weil, D. ( 1992). A contribution to the empirics of economic growth. Quarterly Journal of Economics,

15 [19]Nelson, R. & Phelps, E. (1966). Investments in humans, technological diffusion, and economic growth. American Economic Review, 51, [20]Papageorgiou, C. (2003). Distinguishing Between the Effects of Primary and Post-Primary Education on Economic Growth. Review of Developmental Economics, 7(4), [21]Pritchett, L. (2001). Where has all education gone? World Bank Economic Review, 15, [22]Safdari, M., Shahiki, M. & Sheidaee, Z. (2010). How Does Human Capital Affect on Growth in Different Economies? Journal of Social Sciences, 6(3), [23]Schütt, F. (2003).The importance of Human Capital for Economic Growth. Institute for World Economics and International Management, band no: 27. [24]Tiruneh, M. & Radvansky, M. (2011). The Contribution of Human Capital to European Economic Growth: An empirical exploration from a panel data. Working Paper in Institute of Economic Research. [25]World Bank (2011), World Development Indicators. Retrieved July 15, 2012, from World Databank: 447