Teixeira 2016

Research Policy 45 (2016) 1636–1648
Contents lists available at ScienceDirect
Research Policy
journal homepage: www.elsevier.com/locate/respol
Economic growth, human capital and structural change: A dynamic

panel data analysis
Aurora A.C. Teixeira a,∗ , Anabela S.S. Queirós b
a
CEF.UP, Faculdade de Economia, Universidade do Porto, INESC TEC, OBEGEF, Portugal
b
Faculdade de Economia do Porto, Universidade do Porto, Portugal
a r t i c l e i n f o a b s t r a c t
Article history: Human capital is identified as one of the main determinants of economic growth and plays an impor-
Received 20 June 2015 tant role in the technological progress of countries. Nevertheless, existing studies have to some extent
Received in revised form 25 February 2016 neglected the importance of human capital in the growth process via the interaction it can have with a
Accepted 20 April 2016
country’s industrial specialization. Additionally, the emphasis is mainly placed on supply-side determi-
Available online 6 May 2016
nants, while demand-side factors are neglected, particularly the relevance of the processes of structural
change. Thus, using a growth model which integrates variables from both the supply side and demand
JEL classification:
side, we assess the direct and indirect effects of human capital on economic growth, including in the latter
J24
O3
the interaction of human capital with the industrial specialization of countries. Based on dynamic panel
O4 data estimations, we found that human capital and the countries’ productive specialization dynamics
O47 are crucial factors for economic growth. Moreover, the interaction between human capital and structural
change in high knowledge-intensive industries impacts significantly on economic growth. However, the
Keywords: sign of this effect depends on the type of country and the period of analysis. Specifically, over a longer
Economic growth
time span (1960–2011) and for more highly developed (OECD) countries, the impact of the interaction
Human capital
between human capital and structural change is positive. When we also include transition and Mediter-
Structural change
Dynamic panel data ranean countries over a shorter time period (1990–2011), we find that human capital significantly and
positively impacts on the countries’ economic growth but the effect of human capital via specialization
in high-tech and knowledge-intensive activities is negative. The latter result indicates that the lack of
industrial structures able to properly integrate highly educated individuals into the productive system
leads countries to experience disappointing economic returns.
© 2016 Elsevier B.V. All rights reserved.
1. Introduction theless, human capital stock stands as one of the most frequent
determinants included in such analyses (Barro, 1991; Hanushek
The analysis of the determinants of economic growth has been and Wößmann, 2012; Aisen and Veiga, 2013).
the subject of extensive literature, especially since the beginning The concept of human capital can be interpreted as the set
of the 1990s. Some authors (e.g., Barro, 1991; Mankiw et al., 1992; of intangible resources embedded in the labor factor which have
Mauro, 1995) have estimated the impact of certain variables on eco- improved its productivity (Goldin, 2016). These are associated to
nomic growth through cross-section analysis and concluded that knowledge and skills acquired through education, experience and
human capital plays an important role in economic growth. health care (Schultz, 1961; Becker, 1962).
Neoclassical and endogenous growth theory identified and ana- Human capital has a direct effect on economic growth because
lyzed some determinants of economic growth such as foreign trade, individuals with more education are more productive and inno-
government consumption and geography, as well as institutions, vative leading to the creation of new products and improving the
namely the case of political instability (Barro, 1991; Levine and productivity of factors (Romer, 1990; Benhabib and Spiegel, 1994;
Renelt, 1992; Acemoglu et al., 2001; Moral-Benito, 2012). Never- Teixeira and Fortuna, 2011; Bodman and Le, 2013). On the other
hand, human capital enhances technology adoption from neigh-
boring countries through the absorption of ideas and equipment
∗ Corresponding author at: Faculdade de Economia do Porto, Rua Dr Roberto Frias, imports (Nelson and Phelps, 1966; Benhabib and Spiegel, 1994;
Porto 4200-464, Portugal. Teixeira and Fortuna, 2011). Human capital also has indirect effects
E-mail addresses: ateixeira@fep.up.pt (A.A.C. Teixeira), 120411078@fep.up.pt namely via interaction with the productive structure of coun-
(A.S.S. Queirós).
http://dx.doi.org/10.1016/j.respol.2016.04.006
0048-7333/© 2016 Elsevier B.V. All rights reserved.
A.A.C. Teixeira, A.S.S. Queirós / Research Policy 45 (2016) 1636–1648 1637
tries. Concretely, the specialization of a country in technologically income. Human capital improves the quality of labor, increasing its
advanced activities improves the impact (positive) of human capital productivity (Mankiw et al., 1992; Wö␤mann, 2003; Bodman and
on economic growth (Silva and Teixeira, 2011). Le, 2013). It is usually considered that an additional school year
Theoretical approaches related to evolutionary economics have will increase the productivity and efficiency of workers, and conse-
revealed a need to add demand-side factors to economic growth quently, their income (Hall and Jones, 1999). Likewise, differences
analysis (Witt, 2001; Metcalfe et al., 2006; Dietrich, 2012; Teixeira, in the average schooling of countries are related to different eco-
2012). Certain changes in demand that favor more diverse and com- nomic growth rates (Benos and Zotou, 2014). For example, Easterly
plex products lead to structural changes (Hidalgo and Hausmann, and Levine (1997) found that the low economic growth observed
2009), i.e., changes in sectoral composition and in economic spe- in African countries is due, in part, to low rates of schooling.
cialization, by boosting technological innovation and creating new Human capital is the driver of Research and Development (R&D),
products (Silva and Teixeira, 2011; Saviotti and Pyka, 2012). In this which enhances innovation and technological progress, thus lead-
line of thought, ‘high-tech’ industries have higher growth rates of ing to increased productivity and creation of new products (Romer,
productivity and therefore contribute more than proportionally to 1990; Benhabib and Spiegel, 1994; Teixeira and Fortuna, 2011;
economic growth (Silva and Teixeira, 2011). This contribution tends Bodman and Le, 2013). This means that the more educated the
to increase the more intense the absorption capacity and innova- workforce of a country, the greater the benefits of the R&D activ-
tion becomes, related to higher levels of human capital (Nelson and ities in terms of economic growth. Human capital promotes the
Phelps, 1966; Teixeira and Fortuna, 2011). absorption of new ideas (absorption capacity) and products already
This paper intends to integrate into a single model supply-side created by other countries. This fosters a faster convergence of
variables linked to the endogenous growth theory, and demand- economies by importing equipment and technologies (Nelson and
side variables linked to structural and evolutionary approaches, Phelps, 1966; Benhabib and Spiegel, 1994; Bodman and Le, 2013).
namely the specialization pattern of countries. More specifically, Through the mechanisms described above, human capital will
the aim of this paper is to estimate the direct effects of human cap- encourage greater investment in physical capital (Benhabib and
ital on economic growth as well as indirect effects embodied in the Spiegel, 1994).
interaction between human capital and the countries’ productive Finally, human capital also affects economic performance indi-
structure, while controlling for other determinants reported in the rectly, most notably through its interconnections with institutions.
literature. Our hypothesis is that a country with a higher level of Human capital accumulation contributes to shaping efficient poli-
human capital will grow faster the higher the level of specialization cies, less violence and more political stability (Lipset, 1960; Glaeser
in high-tech and knowledge-intensive industries, in which skilled et al., 2004) and, therefore, fosters economic growth. Consistent
labor plays an important role. with this perspective, Sianesi and Reenen (2003) show that human
In methodological terms, we apply the latest dynamic panel data capital, specifically in its educational dimension, besides stimulat-
techniques based on the generalized method of moments (GMM) ing the productivity of workers, tends to improve health levels,
developed by Arellano and Bover (1995) and Blundell and Bond environmental conditions, criminal rates, social cohesion and civic
(1998), to two distinct sets of data: one including only OECD coun- participation. Therefore, investment in education (i.e., human capi-
tries over a long time span (52 years), 1960–2011; and another tal accumulation) has an impact not only on individual returns, but
including (21) OECD plus (9) transition and Mediterranean coun- also drives a spillover effect that produces social benefits (see Dias
tries over a shorter period of time (22 years), 1990–2011. and Tebaldi, 2012).
The paper is organized as follows. The following section presents Albeit some authors (e.g., Sunde and Vischer, 2015) acknowl-
a literature review on the relationship between economic growth edge difficulties in clearly assessing the empirical effects of human
and the three main variables of this study: human capital stock, capital on growth, most studies show a positive and significant
structural change and the interaction between the two. Section relationship between human capital and economic growth (Barro,
3 presents methodological considerations as well as a statistical 1991; Mankiw et al., 1992; Easterly and Levine, 1997; Hall and
description of the data for the relevant variables. In Section 4, we Jones, 1999; Bodman and Le, 2013), regardless of the proxy used
discuss the empirical results. The final section presents the main for human capital (e.g., the average schooling of the working pop-
contributions of this study, policy implications, limitations and ulation or initial enrollment rate) (see Benos and Zotou, 2014).
paths for future research. From the above, we assume that:
H1. Countries with a higher stock of human capital tend to grow
faster than others.
2. Determinants of economic growth: a review
Notwithstanding the enormous importance of human capital
Over the last few decades, a large body of literature has been pro- accumulation, the differentials in economic growth across coun-
duced examining the role of human capital in determining the level tries should (also) be traced to structural change and the complexity
and growth of GDP per capita (Goldin, 2016). The so-called ‘growth underlying their productive structures. Indeed, several studies
accounting’ literature emphasizes the importance of measuring show that the productive structure of an economy and especially
changes in the quality of labor, as indicated by improved quali- its dynamics, i.e., “structural change” (shifts in sectoral composition
fications and higher skills, when trying to account for economic where certain industries gain relative shares in economy) emerge
growth over the long term, whereas ‘new growth theories’ high- as an important determinant of economic growth (Montobbio,
light the determinants of economic growth in the broadest sense, 2002; Saviotti and Frenken, 2008; Silva and Teixeira, 2011).
concentrating on human capital inputs (Wilson and Briscoe, 2004). The influence of structural change on economic growth has been
In endogenous growth models, economic growth can continue a highly disputed and controversial issue in the literature (Hartwig,
indefinitely because the returns on investment in (both physi- 2012).
cal and) human capital goods do not necessarily diminish over On the one hand, studies associated with supply-side
time. Spillovers of knowledge across producers and external ben- approaches based on Baumol’s (1967) cost disease generally advo-
efits from improvements in human capital are part of this process cate that structural change causes aggregate growth to decline
because they offset tendencies to diminishing returns. (Nordhaus, 2008; Hartwig, 2012, 2015). Being relatively silent
Acquiring skills and knowledge is a means of capital forma- regarding the demand side of the economy, Baumol’s argument
tion by delaying consumption with the aim of increasing future posits that because the composition of output has shifted away
1638 A.A.C. Teixeira, A.S.S. Queirós / Research Policy 45 (2016) 1636–1648
from (‘progressive’) sectors with rapid productivity growth (e.g., available capabilities to develop additional products (Hidalgo and
manufacturing) toward those (‘non-progressive’) with stagnant Hausmann, 2009).
technologies (e.g., public services, construction), aggregate produc- Despite being relatively neglected in traditional economic
tivity growth has slowed down. growth analyses, which analyze predominantly the supply-side
On the other hand, evolutionary and economic development- (see Witt, 2001; Metcalfe et al., 2006; Dietrich, 2012), the evo-
related approaches, which emphasize the co-evolution of demand lution of demand is an important driver of structural change to
and innovation in the dynamics of economies, usually find a pos- ensure the need to reinvent and produce new goods and innova-
itive relation between structural change and economic growth. tion is the underlining concept of this process (Saviotti and Pyka,
Within structural change evolutionary theories, the market is the 2012). Changes in consumption patterns reflect changes in learn-
result of a continuous process of transformation (Metcalfe et al., ing, knowledge and specialization (Witt, 2001; Saviotti and Pyka,
2006) fruit of innovation (Justman and Teubal, 1991; Saviotti and 2012).
Pyka, 2012). Under Schumpeterian conceptions, radical changes According to Saviotti and Pyka (2013), the growing productiv-
in innovation contribute significantly to economic development ity efficiency of incumbent sectors combined with the employment
(Saviotti, 2001). effect associated to the emergence of new sectors helps to explain
Structural change can be instigated by both exogenous and the rise in the disposable income of consumers. As societies become
endogenous forces, such as the life cycle dynamics of sectors and more developed, a growing share of their income is spent on
changes in consumption patterns in favor of goods with high new, higher quality and more differentiated goods with structural
income elasticity of demand. As sectors go through a life cycle change evolving towards increasingly varied and, in particular,
from birth to maturity, one can expect that some variables, namely more sophisticated goods and complex economic systems (Hidalgo
the rate of demand growth or the profit rate will fall as a sector and Hausmann, 2009). The co-evolution of demand and innova-
matures. Such dynamics can propel Schumpeterian entrepreneurs tion is thus a “fundamental mechanism of economic development,
to start creating new sectors where they could have a temporary which is essential to understand long-run processes of structural
monopoly (Saviotti and Frenken, 2008; Quatraro, 2009). This win- change and transformations” (Saviotti and Pyka, 2016: 21).
dow of opportunity for the creation of new sectors tends to be Empirically, Nelson and Pack (1999) note the rapid catching-
greater in an economic system in which many resources are allo- up of Asian countries associated to increasing specialization in
cated to R&D and to other research activities (Bogliacino and Pianta, activities with high technological content, and argue that the fast
2013; Saviotti and Pyka, 2016).1 Sectors that have higher income growth in these economies is the result of the assimilation of
elasticity of demand see their relative importance in the economy technological innovations and structural changes. Using network
rise (Peneder, 2003) and more productive industries gain a greater theory methods, Hidalgo et al. (2007) show that the development
relative share in the economies because they offer better wages, path of a country is determined by its capacity to accumulate
attracting more qualified and skilled individuals. These phenomena the capabilities that are required to produce diverse and more
cause direct effects on growth by creating new ways of produc- sophisticated/complex products. Also linked to demand issues,
tion that will result in a more efficient reallocation of resources Aditya and Acharyya (2013) find that, in dynamic terms, export
and greater earning capacity (Zagler, 2009; Noseleit, 2013). There- specialization in high-tech goods has a positive relationship with
fore, structural change in favor of a specialization in technologically economic growth. Relating the growth experiences of 21 countries
more advanced sectors leads to economic growth (Peneder, 2003; (20 OECD members plus Taiwan) which had relatively similar eco-
Silva and Teixeira, 2011). nomic structures in the late 1970s, with changes occurring in these
In this line, Structuralist theories stress that economic spe- countries’ structures between 1979 and 2003, Silva and Teixeira
cialization by itself is not sufficient to generate economic growth (2011) demonstrate that structure influences productivity growth.
(Aditya and Acharyya, 2013), depending instead on the sector that From the above it is conjectured that:
dominates the economy. According to these approaches, economic
growth can be the result of a shift in specialization towards indus- H2. Countries that experience changes in productive structures
trial products (Aditya and Acharyya, 2013). In fact, it appears that towards a greater share of technology/knowledge-intensive activ-
most industrialized regions observe greater economic growth than ities will tend to observe higher economic growth.
mainly agricultural regions (Marelli, 2004). Furthermore, special-
ization in agricultural products makes the country more vulnerable The structuralist approach considers human capital as a major
to external shocks, according to Aditya and Acharyya (2013). determinant of economic growth since this factor enhances struc-
Recent contributions related to the literature on economic tural change (Justman and Teubal, 1991). Thus, human capital is
development (e.g., Hidalgo et al., 2007; Hidalgo and Hausmann, critically important in the evolution of a country’s specialization
2009) and grounded in the ideas of the literature on structural (Krishna and Levchenko, 2013). The productive specialization of
transformation pioneered by Kaldor (1967), among others, during economies depends on their endowment factors, whereby techno-
the 1950s and 1960s, emphasize that the direction of structural logically more advanced industries will tend to locate in countries
change has been steered towards increasing complexity and differ- with a high stock of human capital. Ciccone and Papaioannou (2009)
entiation of economic systems. Accordingly, changes in a country’s also emphasize the positive relationship between education and
productive structure (structural change) are the result of two joint ‘virtuous’ structural changes, i.e., when technology-intensive activ-
processes involving the creation of new products from hitherto ities gain a relative share in the economy.
unexplored combinations of capabilities they already have, and the The technological catching-up process and structural change
accumulation of new capabilities interacting with other previously related to the transfer of technology from developed coun-
tries to developing ones may be enhanced (in economic growth
terms) when the country has a higher level of human capital,
which increases its absorption capacity (Nelson and Phelps, 1966;
Benhabib and Spiegel, 1994). Through this process, developing
1
Following Schumpeter, to the extent that entrepreneurs have to teach con- countries can have productive structures with more technologi-
sumers to consume new goods and services, the partial endogeneity of the
emergence of new sectors would make the emergence of new types of demand
cal content via imitation. However, a successful imitation needs a
partially endogenous (Saviotti and Pyka, 2013). The authors acknowledge one of the minimum threshold of human capital (Vandenbussche et al., 2006;
referees for bringing up this point. Teixeira and Fortuna, 2011). A fortiori, creative and innovative pro-
cesses require greater stocks of human capital (Vandenbussche capital adopt complementary technologies in order to achieve max-
et al., 2006). imum efficiency. It follows that the accumulation of human capital
The interaction between human capital and structural change enhances the role of Research and Development (R&D) in the
is clearly explained by Gürbüz (2011) who highlights the situa- economies by promoting the creation of new products (Caselli and
tion of North and South countries. Southern countries, during their Coleman, 2006; Bodman and Le, 2013). Therefore, human capital
tertiarization, specialized in labor-intensive services and absorbed affects the technological progress of countries (Nelson and Phelps,
low-skilled workers from rural migration; in contrast, northern 1966; Romer, 1990; Benhabib and Spiegel, 1994; Vandenbussche
countries specialized in high-tech, tradable products, which require et al., 2006; Caselli and Coleman, 2006). On the other hand, it
a high level of human capital. According to this author, the accu- reduces the costs of implementing these technologies (Kim and Lee,
mulation of human capital is a necessary condition for a virtuous 2009).
structural change, i.e., an increasing share of the most productive The process of innovation requires a certain level of human
sectors. capital to actually occur (Vandenbussche et al., 2006). The sec-
Regarding the demand-side, human capital makes consumers tors that arise through structural change require the acquisition of
more sophisticated. This means that if consumers are more highly new skills by workers from the declining industries in order to be
educated, they will be more likely to seek ‘high-tech’ products, absorbed by the first ones (Zagler, 2009). Individuals acquire a set
which positively contributes to virtuous structural change (Justman of important skills to perform certain functions, particularly those
and Teubal, 1991). involving adaptation to change, through formal education (Nelson
Structural change is also enhanced by entrepreneurs (Justman and Phelps, 1966). In addition, the productivity gains resulting
and Teubal, 1991; Dias and McDermott, 2006; Saviotti and Pyka, from technological progress depend on both the accumulation of
2012), because they invest in more innovative and modern sec- physical capital and human capital, and these gains are greater
tors. Recognizing a business opportunity, entrepreneurs create new the higher the stocks of these factors (Ciccone and Papaioannou,
technologically advanced and efficient firms, which contribute to 2009). Physical capital and human capital are complementary since
structural change (Noseleit, 2013). Entrepreneurs are, in general, increases in the stock of physical capital lead to an increase in the
more ‘talented’ than other workers and they invest in their human marginal productivity of human capital and vice-versa (Caselli and
capital, via professional experience, in order to enhance their own Coleman, 2006). In this context, high-tech activities observe a great
talent. The skills acquired by entrepreneurs allow them to cre- increase in productivity which leads to the creation of value added
ate new ideas and generate new businesses (Iyigun and Owen, and employment growth in industries (Ciccone and Papaioannou,
1999). Entrepreneurs also foster the reallocation of productive fac- 2009).
tors (such as labor) among sectors (Noseleit, 2013). The latter look The technological catching-up process and structural change
for more productive agents so that they can integrate the most related to the transfer of technology from developed coun-
innovative and technologically advanced activities, which require tries to developing ones may be enhanced (in economic growth
certain skills and knowledge. Therefore, entrepreneurs create an terms) when the country has a higher level of human capital,
attractive environment for the accumulation of human capital by which increases its absorption capacity (Nelson and Phelps, 1966;
workers through education, since the expected return is higher Benhabib and Spiegel, 1994). Through this process, developing
(Dias and McDermott, 2006; Noseleit, 2013). This means that eco- countries can have productive structures with more technologi-
nomic growth will be higher the more numerous and talented the cal content via imitation. However, a successful imitation needs a
entrepreneurs and the higher their stock of human capital (Iyigun minimum threshold of human capital (Vandenbussche et al., 2006;
and Owen, 1999; Dias and McDermott, 2006). Teixeira and Fortuna, 2011). A fortiori, creative and innovative pro-
cesses require greater stocks of human capital (Vandenbussche
H3. The impact of human capital on the economic growth of a
et al., 2006).
country is greater the more specialized the economy is in high
The interaction between human capital and structural change
technology/knowledge-intensive activities.
is clearly explained by Gürbüz (2011) who highlights the situa-
Beside human capital and structural change, there are a set of tion of North and South countries. Southern countries, during their
determinants of economic growth recurrently reported by sev- tertiarization, specialized in labor-intensive services and absorbed
eral relevant studies which we will also include in our empirical low-skilled workers from rural migration; in contrast, northern
analysis. These include: population growth (Dreher, 2006; Aisen countries specialized in high-tech, tradable products, which require
and Veiga, 2013), physical capital accumulation/investment (Barro, a high level of human capital. According to this author, the accu-
1991; Benhabib and Spiegel, 1994; Fabro and Aixalá, 2009; Aisen mulation of human capital is a necessary condition for a virtuous
and Veiga, 2013), public consumption (Barro, 1991; Levine and structural change, i.e., an increasing share of the most productive
Renelt, 1992; Moral-Benito, 2012; Afonso and Jalles, 2014), and sectors.
institutions (Mauro, 1995; Hall and Jones, 1999; Acemoglu et al., Regarding the demand-side, human capital makes consumers
2001; Dreher, 2006; Hanushek and Wößmann, 2008). more sophisticated. This means that if consumers are more highly
Ciccone and Papaioannou (2009) also emphasize the positive educated, they will be more likely to seek ‘high-tech’ products,
relationship between education and ‘virtuous’ structural changes, which positively contributes to virtuous structural change (Justman
i.e., when technology-intensive activities gain a relative share in and Teubal, 1991).
the economy. Therefore, structural change and the virtuous spe- Structural change is also enhanced by entrepreneurs (Justman
cialization of economies depend on new information, new skills and Teubal, 1991; Dias and McDermott, 2006; Saviotti and Pyka,
and workers’ productivity, factors which are closely related to the 2012), because they invest in more innovative and modern sec-
stock of human capital (Justman and Teubal, 1991; Krishna and tors. Recognizing a business opportunity, entrepreneurs create new
Levchenko, 2013). technologically advanced and efficient firms, which contribute to
The accumulation of human capital enables new business structural change (Noseleit, 2013). Entrepreneurs are, in general,
opportunities and endows the agents with management skills and more ‘talented’ than other workers and they invest in their human
technological knowledge (Justman and Teubal, 1991). Countries capital, via professional experience, in order to enhance their own
with highly skilled workers tend to be more efficient in activities talent. The skills acquired by entrepreneurs allow them to cre-
that incorporate more advanced technologies. From a micro per- ate new ideas and generate new businesses (Iyigun and Owen,
spective, firms that have employees with a high level of human 1999). Entrepreneurs also foster the reallocation of productive fac-
tors (such as labor) among sectors (Noseleit, 2013). The latter look is inevitable. Apart from that, the lagged dependent variable of this
for more productive agents so that they can integrate the most empirical model will lead to the interaction between explanatory
innovative and technologically advanced activities, which require variables and random disturbances (meaning that explanatory vari-
certain skills and knowledge. Therefore, entrepreneurs create an able is endogenous), which renders the traditional “fixed effects”
attractive environment for the accumulation of human capital by and “random effects” panel estimators inconsistent.
workers through education, since the expected return is higher In order to mitigate the endogeneity problem with explanatory
(Dias and McDermott, 2006; Noseleit, 2013). This means that eco- variables, Arellano and Bond (1991) (AB) propose the use of instru-
nomic growth will be higher the more numerous and talented the mental variables to deduce the generalized method of moments
entrepreneurs and the higher their stock of human capital (Iyigun (GMM) of corresponding moment conditions, namely difference
and Owen, 1999; Dias and McDermott, 2006). GMM. The basic idea of the method is: eliminating the individual
fixed effect by proceeding with the first difference of regression
H3. The impact of human capital on the economic growth of a
equation in the first place. Then, the lagged variable will be regarded
country is greater the more specialized the economy is in high
as the corresponding instrumental variable of endogenous vari-
technology/knowledge-intensive activities.
ables in the difference equation. However, in finite samples AB may
Beside human capital and structural change, there are a set of suffer a severe “weak instruments” problem and thus poor preci-
determinants of economic growth recurrently reported by sev- sion (Bond et al., 2001). Arellano and Bover (1995) and Blundell
eral relevant studies which we will also include in our empirical and Bond (1998) propose a solution to this problem (combin-
analysis. These include: population growth (Dreher, 2006; Aisen ing additional moment restrictions with those in AB), providing
and Veiga, 2013), physical capital accumulation/investment (Barro, a “system-GMM” estimator in which GMM is applied to a sys-
1991; Benhabib and Spiegel, 1994; Fabro and Aixalá, 2009; Aisen tem of two equations: an equation in differences instrumented by
and Veiga, 2013), public consumption (Barro, 1991; Levine and lagged levels, and an equation in levels instrumented by lagged
Renelt, 1992; Moral-Benito, 2012; Afonso and Jalles, 2014), and differences.
institutions (Mauro, 1995; Hall and Jones, 1999; Acemoglu et al., System GMM is able to correct unobserved country hetero-
2001; Dreher, 2006; Hanushek and Wößmann, 2008). geneity problems, omitted variable bias, measurement error and
potential endogeneity issue that frequently affect growth models
3. Methodology when using pooled OLS and fixed effect methods (Bond et al., 2001).
It is also capable of reducing potential bias and imprecision asso-
3.1. Econometric specification and option for dynamic panel data ciated with a simple first-difference GMM estimator (Arellano and
estimation Bover, 1995; Blundell and Bond, 1998).
Therefore, in this study, we apply the system-GMM estimation
Our empirical specification is based on the typical cross-country method for the dynamic panel data model. The dynamic panel GMM
catch-up equation (cf. Barro and Sala-i-Martin, 2003), which can be estimation can be divided into one-step and two-step estimations
formulated as follows: in accordance with different choices of the weight matrix. Bond
et al. (2001) suggest that, with limited samples, the standard error

yi,t − yi,t−1 = ˛yi,t−1 + ˇXi,t + i + εi,t (1) of the two-step GMM estimation value will have a marked down-
ward bias.2 Moreover, Roodman (2009) points out that with an
In this expression, yit is the logarithm of real GDP per capita for
increasing number of periods, system GMM generates by default
country i in period t, X it represents a vector of variables influencing
a large number of instruments, which can overfit endogenous vari-
economic growth (including human capital and structural change),
ables and weaken the model specification tests. Therefore, for the
i represents the unobservable country-specific effect, and εit is
model with fewer countries and a longer time span (21 countries
the error term.
and 52 years–Model I), we use one-step system GMM, and for the
Eq. (1) can be rewritten as:
model with a larger number of countries and a shorter time span (30

yi,t = ˛1 yi,t−1 + ˇXi,t + i + εi,t (2) countries and 22 years–Model II), we use two-step system GMM.
GMM cannot be established as a consistent method of estima-
where ˛1 = 1 + ˛ (being ␣ the conditional convergence factor). tion unless there is no autocorrelation in the disturbance term.
Specifying in more detail the econometric model to be estimated In addition, to ensure that the moment conditions are not over-
(according to the literature review in Section 2 and the hypotheses constrained, the number of instrumental variables must not exceed
put forward), we have:
yi,t = ˛1 yi,t−1 + ˇ1 hci,t + ˇ2 sci,t + ˇ3 (sc × HC)i,t + ˇ4 invi,t + ˇ5 gi,t + ˇ6 popi,t + ˇ7 pri,t + ˇ8 cli,t + i + εi,t (3)
where y is the natural logarithm of real GDP per capita, hc rep-

resents the natural logarithm of human capital measure (average
years of schooling), sc is structural change (i.e., the share of high-
the number of endogenous variables.
tech/high knowledge-intensive industries in total employment),
For lagged endogenous variables and weak exogenous vari-
sc × HC is the variable measuring the interaction between human
ables to be valid as instruments, it is necessary that the transient
capital stock (HC) and structural change (sc), inv is the investment
disturbances in the base model, ␧it , are free of autocorrelation
(in physical capital) rate, g is the share of public consumption in
(Blundell and Bond, 1998). This would imply that the disturbances
gross domestic product, pop is the population growth rate, and pr
in the differenced model have significant first-order correlation
and cl are the natural logarithms of the indexes of political rights
and civil liberties, respectively.
Human capital and structural change can stimulate economic
2
growth and economic growth in turn influences human capital Bond et al. (2001) show the efficiency gains obtained with two-step GMM
and structural change. This means that the three variables may estimators, compared to one-step GMM estimators, tend to be small, two-step esti-
mators normally converge to their asymptotic distribution relatively slowly, and
have a bilateral causal relationship. Once there is an inverse causal in a finite sample, their asymptotic standard errors can be seriously biased down-
relationship between economic growth and human capital and wards, and thus making them unreliable. Thus, in such a context, a one-step GMM
structural change, the endogeneity problem with empirical models estimation might be preferred.
and insignificant second-order autocorrelation. For this purpose, (2013) asserted that by “[l]ooking purely at the educational attain-
the Arellano–Bond tests for first-order (AR(1)) and second-order ment data, we cannot find an explanation for the divergent path of
(AR(2)) serial correlation in the first-differenced residuals are used the economies of Central and Eastern Europe.” (van Leeuwen and
(see Arellano and Bond, 1991). Because the first differences of inde- Földvári, 2013: 78). An alternative proxy for human capital, more
pendently and identically distributed idiosyncratic errors will be suitable for relatively more developed countries (the ones analyzed
serially correlated, rejecting the null hypothesis of no serial corre- here), would be the average years of education adjusted to the qual-
lation in the first differenced error at order one does not imply that ity of education systems (see Benos and Zotou, 2014). Recently,
the model is misspecified. Rejecting the null hypothesis at higher Delgado et al. (2014) found evidence (within a cross-sectional
orders, however, implies that the moment conditions are not valid. framework) that educational achievement, measured by mean test
The validity of the moment condition can be directly tested by scores, may provide a more reliable measure of human capital than
Sargan and Hansen tests, whose null hypothesis is that all instru- mean years of schooling. However, data available which address
ments as a group are exogenous (thus, for the instruments to be the quality of education is of limited use for panel data analyses −
valid one should not reject the null hypothesis). However, the complete longitudinal data sets are non-existent, and the existing
behavior of Sargen test statistic is only well known when distur- datasets provide only global period averages or a few isolated years
bances can be assumed as homoscedastic (Iqbal and Daly, 2014). (e.g., Hanushek and Kimko, 2000; Altinok et al., 2014; Kaarsen,
Additionally, the Sargan test may have low power to reject the null 2014).
hypothesis, the instruments are valid when the sample size is small Structural change refers to long-term shifts in the sectors of an
(Bowsher, 2002) and tend to over-reject the null hypothesis of seri- economy. To measure these shifts, we use employment composi-
ally uncorrelated errors in the case of one-step GMM estimations tion data and its evolution over the period under review. These
(Arellano and Bond, 1991). Given the shortcomings associated with values come from the EU KLEMS (O’Mahony and Timmer, 2009)
the Sargen test and the fact that the Hansen test is the most adopted dataset, available at www.euklems.net, which gives us (among
in practice (see Chen and Sun, 2014), we opted for the Hansen test other variables) the number of employees in each industry. The
to assess the validity of the instruments. several economic activities are divided according to the ISIC REV.3
classification (‘International Standard Industrial Classification’). We
3.2. Description of the proxies and data sources use these values as a share of the total employment in each econ-
omy. Subsequently, we group the data depending on the intensity
We use two different panel data sets in order to estimate the of knowledge that each activity requires. This classification is based
specified model (Equation (3)). The first one includes 21 OCDE on Peneder (2007) who divides each industry into: ‘very low’, ‘low’,
countries, of which we have a balanced panel (the entire cross- ‘medium-low’, ‘intermediate’, ‘medium-high’, ‘high’ and ‘very high’.
section is observed every year), during the period from 1960 to Since EU KLEMS only provides data from the period between 1972
2011. The second panel adds 9 countries to the first, belonging and 2007, we estimate the values for the remaining missing years.
to Eastern Europe (Czech Republic, Estonia, Hungary, Lithuania, To this end, we assume that the share of each industry during the
Poland, Slovakia and Slovenia) and the Mediterranean (Cyprus and period between 1960 and 1971 evolves at the same average growth
Malta). However, due to data unavailability, the period analyzed rate of the next 10 years. Similarly, the values of the subsequent
is shorter (1990–2011). In the second panel, the economic growth years up to 2007 are calculated assuming the average growth rate
analysis is to some extent impaired because the time horizon is too of the previous 10 years.
short (22 years), since the dynamics analyzed change very slowly In line with the many empirical studies on economic growth
over time. Nevertheless, the Eastern European countries included (e.g., Barro, 1991; Levine and Renelt, 1992; Mauro, 1995; Moral-
in the study have experienced great structural change due to the Benito, 2012), the physical capital accumulation (I) is measured
establishment of new industries and industrial relocation resulted by investment share to GDP. Similarly, public expenditures (G) are
from FDI (Foreign Direct Investment) since the 1990s (Janicki and measured by public consumption to GDP ratio. There are many
Wunnava, 2004). Thus, the motivation to include these countries authors that resort to the Penn World Table dataset to determine
in our study lies in this last argument, despite the limited data economic growth (e.g., Easterly and Levine (1997) and Moral-
available. Benito (2012) used the Penn World Table, Version 6.2). Therefore,
In order to measure the dependent variable, economic growth, besides GDP per capita, we use the Penn World Table (Version 8.0)
we use as a proxy, in line with extant studies (see Table A1, in (Feenstra et al., 2013) to obtain investment in physical capital and
Appendix A), the natural logarithm of GDP per capita. To this public consumption data.
end, we have employed the dataset Penn World Table (version 8.0) Regarding institutional measures, this study uses the Civil Lib-
(Feenstra et al., 2013) which reports the values for real GDP at erties Index and Political Rights Index from ‘Freedom House’, as in
constant 2005 prices (in mil. US$). Moral-Benito (2012). These indexes are based on the evaluation of
The average years of schooling of adults is the most widely used human rights specialists, scholars, journalists and politicians. They
proxy of human capital (e.g., Rajan and Zingales, 2008; Easterly are measured using a scale of 1–7, where 1 the highest freedom
and Levine, 1997; Hall and Jones, 1999; Temple and Wö␤mann, level and 7 the lowest. Through these measures, we can analyze
2006; Moral-Benito, 2012; Bodman and Le, 2013). Thus, according the impact of the quality of institutions and governability of a coun-
to the available literature, we use educational attainment for the try on economic growth. We assume that freer and fairer countries
population aged 25 and over to measure the stock of human capital have higher economic growth rates.
from the Barro and Lee (2010) dataset. The Barro and Lee dataset is
widely used in the literature on economic growth, and it has been
updated over the last few years. Barro and Lee’s (2010) data cover 4. Empirical results
the period from 1950 to 2010 and are related to 146 countries. These
are disaggregated for periods of 5 years and come from UNESCO, The model specified previously (see Eq. (3)) serves to test
Eurostat, national statistical agencies, etc. For the present study, we the impact of our three core variables on the economic growth
consider the average attainment value for each five-year period for of countries: 1) human capital, measured by the average edu-
all years within the same timeframe. cational attainment of adults; 2) productive specialization in
In a study focusing on the economic growth of Central and East- ‘high-level’ industries, measured by the share of employment in
ern Europe between 1920 and 2006, van Leeuwen and Földvári knowledge/technology-intensive industries (e.g., Financial Inter-
Table 1
Dynamic panel data estimation of the relationship between economic growth, human capital and productive structure (dependent variable: GDP per capita (in ln), 1960/90-
2011).
Variable Proxy Model I Model II
21 countries; 52 years (1960–2011) 30 countries; 22 years (1990–2011)
Dynamic panel-data estimation, one Dynamic panel-data estimation,

step system GMM two-step system GMM
GDP pc (lagged) Gross Domestic Product per capita (in 0.826*** 0.731***
ln) in t-1
(0.034) (0.067)
Human capital Number of years of schooling of the 0.235** 1.099***
population aged 25 or more (in ln)
(0.087) (0.244)
Structural change Share of ‘high-level’ industries in total 3.480*** 8.220***
employment
(0.935) (2.286)
Interaction between Share of ‘high-level’ 0.201*** −0.645***
human capital and industries × Human Capital
productive structure
(0.063) (0.159)
Investment Investment rate 0.339** 0.482**
(0.134) (0.210)
Public Expenditures Public Consumption in GDP −0.542** −1.022***
(0.232) (0.177)
Population growth Population growth rate −2.250* 1.720
(1.212) (2.731)
Institutions Political Rights index (1: the highest 0.079*** −0.078***
level . . . 7: the lowest level)(in ln)
(0.022) (0.021)
Civil Liberties index (1: the highest −0.035*** −0.064***
level . . . 7: the lowest level) (in ln)
(0.012) (0.021)
Number of observations 1050 630
Goodness of fit Number of instruments 150 82
AR(1) p-value 0.003 0.006
AR(2) p-value 0.102 0.142
Hansen p-value 1.000 0.997
Note: dependent variable: natural logarithm of real GDP per capita, in constant prices; ‘High-level’ activities include: financial intermediation, computers, research and
development and education; standard deviations are in parentheses; *** (** )[* ]Statically significant at 1% (5%) [10%].
mediation, Research and Development (R&D), Education); and 3) 1.000 or close to 1.000 for the Hansen test (see Hall and Ahmad,
the moderating effect of human capital on the productive struc- 2014).
ture. The estimation also involves a set of control variables related The human capital presents a significant positive impact in
to investment, public consumption, population growth, and insti- both models, which supports H1 (“Countries with a higher stock
tutional environment, which the empirical literature identifies as of human capital tend to grow faster than others”) and suggests, as
relevant factors of economic growth (see Table A1 in Appendix A). postulated in the literature, that countries with a higher educa-
Table 1 presents the dynamic panel data estimation for coeffi- tional attainment of adults grow faster in the considered periods
cients corresponding to each variable for both considered panels. (1960–2011/1990–2011). This result meets the expectation that
The empirical performance of the system GMM estimation in a higher stock of human capital improves the workforce’s skills,
this study is reasonably satisfactory and robust.3 The test for first- which has a positive impact on its productivity (Bodman and Le,
order serial correlation in the residuals AR(1) shows that the null 2013).
hypothesis of no first-order serial correlation is rejected in the In addition, it is noted that the countries where structural
two models. Meanwhile, the test of second-order serial correlation change contributes to increasing the share of knowledge-intensive
AR(2) shows that all estimations have no problem of second-order activities that require high skills, (e.g., Financial Intermediation,
serial correlation since AR(2) test statistics are unable to reject the Computers, Research and Development and Education), tend,
null of no second-order serial correlation (p-values 0.102 and 0.142 on average, to grow faster. Thus, H2 (“Countries that experi-
for Model I and Model II, respectively). The Hansen test for over- ence changes in productive structures towards a greater share of
identification indicates the null of exogenous instruments is not technology/knowledge-intensive activities will tend to observe higher
rejected with p-values of 1.000 (Model I) and 0.997 (Model II). Nev- economic growth.”) is also confirmed. Knowledge-intensive activi-
ertheless, the implausibly good p-value of this range for the Hansen ties employ individuals with higher skills and knowledge because
test should be interpreted with caution since the test is apparently they are more productive (Hartwig, 2012) and capable of enhanc-
weakened by a high instrument count. Nevertheless, there are a ing the emergence of new products and processes (Zagler, 2009).
number of studies employing system GMM that report p-values of Therefore, the growth rates of countries that observe an increase in
specialization in high-level industries tend to be higher.
It is interesting to note that in the sample that includes Eastern
3
The estimates are robust particularly in the case of Model II. Using other estima- European and Mediterranean countries (Model II) the estimated,
tion techniques (available upon request from the authors), namely one-step system isolated, impact of human capital and structural change is higher
GMM, difference GMM, and fixed effects yields similar results (the signs and signif- than that of the sample which includes only OECD countries (Model
icance of the estimates do not change markedly). In the case of Model I, two-step
I). Specifically, all the other factors remaining constant, an increase
system GMM produces unreliable estimates but in all other estimations the results
are similar. of 1% in the average years of education (share of high knowledge-
intensive industries) results, on average, in an additional annual includes less developed countries, we failed to find evidence that
GDP per capita growth of 1.1 (8.2)%. The corresponding values for population growth matters for economic growth.
the OECD sample are, respectively, 0.24% (human capital) and 3.48% The institutional and political measures present statistically sig-
(structural change). nificant estimates.
Human capital also has a moderating effect, i.e., it has an indi- The ‘Political Rights’ index combines the degree of freedom in
rect impact via productive specialization. However, H3 (“The impact the electoral process, political participation and a decentralized
of human capital on the economic growth of a country is greater political system (Moral-Benito, 2012), where lower values repre-
the more specialized the economy is in high technology/knowledge- sent a higher level of democracy. In Model I, we found the opposite
intensive activities.”) is only confirmed in a much broader timeframe result (positive and significant) to the one expected (negative and
(1960–2011) and for the sample that includes only OECD coun- significant).4 This may be due to the fact that for some countries
tries (Model I). In this case particularly, the more educated the included in this panel, the first years of the period studied (essen-
adult population, i.e., the higher their level of human capital, the tially the 1960s), in which they reveal very high economic growth
higher the impact of a specialization in high-level industries on rates, correspond to periods of dictatorial regimes, thus having also
economic growth. The combination of a structural change in favor higher (i.e., worse) ‘Political Rights’ indexes. In Model II, the coef-
of more progressive, knowledge-intensive activities, with high lev- ficient estimate for the ‘Political Rights’ index is significant and
els of human capital, results in higher innovation in product and negative. Thus, we can conclude that more democratic, freer coun-
progress, which tends to generate increases in GDP per capita (see tries have, on average, higher economic growth.
Justman and Teubal, 1991; Ciccone and Papaioannou, 2009). The ‘Civil Liberties’ index includes freedom of belief and expres-
In Model II, covering a shorter period (1990–2011) and encom- sion, associative freedom, respect for individual rights and laws that
passing more countries (the 21 OECD countries of Model I plus protect them. Similarly to the ‘Political Rights’ Index, lower values
Czech Republic, Cyprus, Estonia, Hungary, Lithuania, Malta, Poland, in this index mean a greater level of freedom. In both models the
Slovakia, and Slovenia), H3 is refuted. The estimation of the mod- estimates for the ‘Civil Liberties’ index is negative (at 1% level of
erating effect of human capital on structural change is significantly significance), which means that countries with greater expression
negative. Thus, regardless of human capital having a positive direct of individual freedoms tend on average to have higher economic
(and significant) impact on economic growth, its indirect impact growth. Thus, in the line of Fabro and Aixalá (2012), the institutional
via the productive structure in favor of high-level industries is neg- variables play an important role in boosting economic growth, by
ative. We conclude that for the period in study (1990–2011) and the encouraging investment in physical and human capital and an effi-
set of countries in this model, which includes a considerable num- cient allocation of resources.
ber of Eastern European countries, economic growth is explained Finally, the coefficient estimate of the lagged GDP per capita
by the dynamics of human capital (education) and by the declin- variable is highly significant, positive and lower than 1 in both
ing share of knowledge-intensive activities (‘high-level’). It seems models, which conveys the typical conditional convergence result,
evident that in the sample with a shorter time horizon and tran- according to which countries with lower initial GDP per capita lev-
sition economies, the matching between an adult population with els present, on average, higher growth rates (see also Silva and
high educational attainment and structural change towards a spe- Teixeira, 2011).
cialization in knowledge/technology-intensive activities does not
contribute to higher economic growth. 5. Conclusions
Thus, we found that the effects of the interaction between struc-
tural change and human capital only appear in the long term, Based on two panel data sets covering highly developed coun-
yielding opposite results in shorter periods that involve samples tries (21 OECD, 7 Eastern European and 2 Mediterranean) over long
with transition economies. (52 years: 1960–2011) and short (22 years: 1990–2011) periods of
Regarding the control variables, we found that high investment time, this paper has highlighted the crucial role of human capital,
rates are related to higher economic growth, since high physical structural change processes and their interaction for the economic
capital formation contributes positively to the productivity of pro- growth of countries.
duction factors (Barro, 1991). In brief, economies with a high level of
investment tend to grow faster than others. We also verified that 5.1. Main contributions
investment/physical capital formation plays an important role in
economic growth, both for the more developed, Western countries The present study contributes to the scientific literature in three
(Model I) and the more heterogeneous sample with Eastern Euro- main ways. First, it demonstrated unequivocally the fundamental
pean emergent economies (Model II). This evidence is in line with role of formal educational attainment, reflected in improved qual-
the results of Dreher (2006), Batten and Vo (2009) and Fabro and ifications and higher skills (i.e., higher levels of human capital),
Aixalá (2009). even in sets of highly developed countries, such as the OECD coun-
Our results confirm the idea that high public consumption can tries. This corroborates findings in the scientific literature which
create market distortions, inefficiencies and crowding-out, that underlines that human capital, most notably education attain-
negatively affect economic growth (Barro, 1991; Dreher, 2006; ment, improves the quality of labor, increasing its productivity
Moral-Benito, 2012; Afonso and Jalles, 2014); in both models – (Mankiw et al., 1992; Easterly and Levine, 1997; Bodman and Le,
the estimate associated to the public consumption variable is sig- 2013 – for a meta-analysis on this topic, see Benos and Zotou,
nificant (at 5% and 1% significance level, in Model I and Model II, 2014), drives R&D activities, enhancing innovation and technolog-
respectively). These results are in line with seminal studies (e.g., ical progress and, thus, fostering productivity and the creation of
Barro, 1991) as well as more recent studies (e.g., Batten and Vo, new products (Romer, 1990; Benhabib and Spiegel, 1994; Teixeira
2009; Dreher, 2006; Afonso and Jalles, 2014). and Fortuna, 2011; Bodman and Le, 2013), and contributes to a
Model I also confirms a negative relationship between popula-
tion growth and economic growth, as suggested in the literature.
Higher population growth reduces the capital/labor ratio, which 4
This, nevertheless, is in line with the results obtained by Moral-Benito (2012).
affects negatively the GDP per capita (Mankiw et al., 1992). How- This author estimated a 73-country panel over 40 years (1960–2000) and obtained
ever, in Model II, based on a more heterogeneous sample that a significant and positive coefficient for the estimate associated with the ‘Political
Rights’ variable.
social and business-friendly environment (Glaeser et al., 2004; Dias including eastern and Mediterranean countries over the period
and Tebaldi, 2012). It also challenges recent studies which have cast 1990–2011, to simply increase the investment in education to fos-
doubt on the merits of ‘average years of education’ in explaining ter economic growth is not enough (see Kiersztyn, 2013) and more
the long-term growth of countries (e.g., van Leeuwen and Földvári, integrated and multifaceted policies, involving education-science-
2013; Delgado et al., 2014), supporting Breton’s (2011: 765) content industries, should be reassessed.
that both test scores and schooling attainment explain countries’
income differences and that “average schooling attainment implic- 5.3. Limitations and paths for future research
itly measures the quality as well as the quantity of schooling”.
Second, it irrefutably establishes that structural change is a The proxy for human capital used in this study, ‘average years
critical determinant of economic growth, contributing to overcom- of schooling’, is recognizably far from a perfect measure of human
ing the manner in which the demand side has been relatively capital. Among the most important sources of inaccuracy stand the
neglected in the literature on economic growth (see Teixeira, lack of information about sources of learning other than schooling
2012). Even though the relationship between economic growth (van Leeuwen, 2007) and the lack of information about the quality
and structural change has been highly debated in the literature of education (Hanushek, 2013). The former is almost impossible
(see Hartwig, 2012), we clearly confirm that increasing specializa- to obtain; the latter has been explored only for a limited num-
tion in knowledge/technology-intensive industries accelerates the ber of countries and for recent periods (see Altinok et al., 2014;
economic growth of countries, both in highly developed countries Kaarsen, 2014). Given that the data on the quality of education are
(such as OECD) and (particularly) in transitional and Mediterranean not available for the full sample and for the entire period analyzed,
countries. Our results strongly support demand-side, structural we had to opt for the average years of schooling. Such a proxy is
and evolutionary approaches that emphasize the co-evolution of of adjustable usefulness depending on the nature of the sample
demand and innovation in the growth of economies and sustain and on the length of the period examined. To adjust the years of
that specialization in technologically more advanced sectors (Silva schooling for differences in the quality of education is likely to
and Teixeira, 2011; Aditya and Acharyya, 2013) and the creation be more important for a sample including only highly developed
of new, diversified and more complex production sectors (Hidalgo countries over a short period than for a sample including devel-
et al., 2007; Saviotti and Frenken, 2008) lead to economic growth. oped countries at very different levels of development over a long
Third, our results reveal a possibly unexplored issue as far as period. Even taking into account the presence of Eastern European
human capital and structural change interaction is concerned: the and Mediterranean countries, the sample in this paper does not
mismatch between structural change processes and the human include countries at very low levels of development but it does
capital endowments of countries. Indeed, although for highly devel- cover a relatively long time period. Correcting for the quality of
oped countries (i.e., OECD only) the impact of human capital on education is likely to increase the explanatory power of education
GDP growth is amplified when knowledge-intensive industries, with respect to economic growth (Hanushek and Wößmann, 2007).
producing value-added goods and services, grow in importance To estimate and integrate into economic growth analyses a measure
in the composition of the economy, that is not the case when of human capital adjusted for quality would be a very hard, though
we include in the analysis, over a shorter time span, transition challenging and useful, research endeavor worth to be pursued in
and Mediterranean countries. The lack of industrial structures that future.
cannot properly integrate the highly-educated individuals into The nature of the sample used in this study – which includes
the productive system generates disappointing economic returns OECD countries, Eastern European, former Communist countries,
for this latter set of countries. This outcome is supported by the and Mediterranean countries – is not representative of countries at
evidence gathered by Kiersztyn (2013) who pointed that during very low levels of economic development. Thus, it may not be pos-
the twenty years of post-communist transition large numbers of sible to generalize the results to this type of countries. However,
highly educated polish workers are trapped in jobs with low educa- if the sample had included a wider range of levels of economic
tional requirements. Such skill mismatch/overeducation produces development, and therefore of levels of human capital, we sur-
negative impact on individual productivity and job satisfaction mise that the results could have most likely confirmed even more
(Ghignoni and Verashchagina, 2014) and reduce GDP growth strongly the positive effect of human capital on growth (see Qadri
(Jaoul-Grammare and Guironnet, 2009). and Waheed, 2013).5 This nevertheless remains to be proved and
demands further research.
5.2. Policy implications
Acknowledgments
Our results suggest that the promotion of economic growth
should not contemplate only investment in human capital through The authors gratefully acknowledge the two anonymous refer-
(formal) education, but also investment in technology/knowledge- ees for their useful comments. All the remaining errors are our own.
intensive activities, generating high value added to the economies This research has been financed by the European Regional Devel-
(e.g., Financial Intermediation, Computers, Research and Develop- opment Fund through COMPETE 2020 – Programa Operacional
ment and Education). Training and education as an investment Competitividade e Internacionalização (POCI) and by Portuguese
in human capital should consider areas of knowledge and skills public funds through FCT (Fundação para a Ciência e a Tecnologia)
required by the dominant ‘progressive’ industries in the economy, in the framework of the project POCI-01-0145-FEDER-006890.
i.e., those that experience increases in productivity and accelerate
the rates of economic growth. Therefore, it makes sense to rethink Appendix A.
the educational offer and system, and wager on vocational courses
that meet market demands. The aim will be to promote the match- Table A1.
ing between skilled labor and economic activities that require these
same qualifications.
It is important to highlight that in the presence of persis-
tent occupational mismatches, which explain, at least in part, the
significant and negative estimate of the coefficient of the interac-
tion between human capital and structural change for the sample 5
We acknowledge one of the referees for highlighting this aspect.
Table A1
Empirical studies on economic growth.
Authors (year) Period Countries (#) Estimation Method Proxy of the dependent Proxies of relevant independent variables
variable (economic
growth)
Barro (1991) 1960–1985 98 countries (Summers Cross-section approach GDP per capita growth Human Capital: Primary and Secondary
and Heston, 1988) (annual, %) enrollment rates (1960)
Initial GDP: Real GDP per capita real in 1960
Physical Capital: Average investment in
physical capital between 1960–1985 (%GDP)
Public Expenditures: Average real public
consumption between 1970–1985 (% GDP)
Institutions: Number of revolutions and coups
per year (1960–1985); Number of
assassinations per million population
A.A.C. Teixeira, A.S.S. Queirós / Research Policy 45 (2016) 1636–1648

Levine and Renelt (1992) 1960–1989 119 countries Extreme Bounds GDP per capita growth Human Capital: Primary and Secondary
excluding oil exporters Analysis (annual, %) enrollment rates (1960)
(BM; FMI; Summers Human Capital: Primary and Secondary
and Heston, 1988) enrollment rates (1960)
Physical Capital: Investment in physical capital
by ‘World Bank National Accounts’ (% GDP)
Public Expenditures: Real public consumption
(% GDP)) by ‘World Bank National Accounts’
Institutions: Number of revolutions and coups;
Mankiw et al. (1992) 1960–1985 98 countries excluding Cross-section approach Logarithm of GDP per Human Capital: Working age population with
oil exporters (Summers working age secondary (%)
and Heston, 1988) population 1960–1985 Physical Capital: Investment in physical capital
(% GDP)
Mauro (1995) 1960–1985 58 countries excluding Cross-section approach GDP per capita growth Human Capital: Primary and Secondary
oil exporters (1960–1985) enrollment rates (1960)PIB Initial GDP: Real
GDP per capita real in 1960
(% GDP)
(%GDP)
Institutions: Corruption Index; Bureaucratic
Efficiency Ratio; Political Instability Index
Easterly and Levine (1997) 60’s, 70’s e 80’s 127 countries Cross-section approach Real GDP per capita Human Capital: Logarithm of average
growth capita real attainment
Initial GDP: Logarithm of initial income by
decade (60 ,70 80 )
Institutions: Index of political instability;
Ethnic Diversity.
Acemoglu et al. (2001) 1985–1995 64 countries OLS Regression Logarithm of GDP per Institutions: Average Protection Against
capita in 1995 (PPC) Expropriation Risk; Constraint on Executive;
Democracy index
Temple and Wö␤mann (2006) 1960–1996 76 countries Cross-section approach Logarithm GDP per Human Capital: Logarithm of average
capita (1960–1996) attainment
Structural change: Changes in employement
composition per sector
1645
1646
Table A1 (Continued)
Authors (year) Period Countries (#) Estimation Method Proxy of the dependent Proxies of relevant independent variables
variable (economic
growth)
Dreher (2006) 1970–2000 123 countries Fixed Effects Panel Data GDP growth rate per Human Capital: Secondary enrollment rate and
capita logarithm of average life expectancy by World
Bank
Initial GDP: Logarithm of GDP per capita by
World Bank
Physical Capital: Intern Investment (% GDP) by
Global Development Network Growth.
(% GDP) by World Bank
Population Growth: Logarithm of fertility rate
A.A.C. Teixeira, A.S.S. Queirós / Research Policy 45 (2016) 1636–1648

by World Bank
Institutions: Rule of Law index by Gwartney
and Lawson (2002)
Batten and Vo (2009) 1980–2003 79 countries Fixed Effects Panel Data GDP growth rate per Human Capital: Secondary enrollment rate by
capita World Bank (2004)
Physical Capital: Gross capital formation (% of
GDP) by World Bank (2004)
Population Growth: Difference between the
logarithms in the number of inhabitants in
each year by World Bank (2004)
Hartwig (2012) 1970–2005 18 OCDE countries Granger Causality Real GDP per capita Human Capital: Expenditure on education and
Method health
Structural Change: Change in the composition
of expenditures by sector
Fabro and Aixalá (2012) 1976–2005 79 countries Panel Data estimation GDP growth rate per Human Capital: Primary and Secondary
capita enrollment rate
(% GDP)
Institutions: Economic Freedom index, Political
Rights and Civil Liberties index.
Moral-Benito (2012) 1960–2000 73 countries Bayesian Panel Data Real GDP growth rate Human Capital: Average attainment of working
Approach per capita (base year age population by Barro and Lee
2000 US$) Initial GDP: Logarithm of GDP per capita in 1960
(% GDP)
Institutions: Political Rights and Civil Liberties
index.
Iqbal and Daly (2014) 1986–2010 52 middle-income Dynamic Panel Data GDP growth rate per Human Capital: Human Development Index by
countries Models capita (annual%) United Nations Development Project (UNDP)
Initial GDP: GDP per capita
Physical Capital: Gross capital formation (% of
GDP)
Institutions: Corruption index; Freedom from
corruption; Democratic force
References Hanushek, E.A., Kimko, D.D., 2000. Schooling, labor-force quality, and the growth
of nations. Am. Econ. Rev. 90, 1184–1208.
Acemoglu, D., Johnson, S., Robinson, J.A., 2001. The colonial origins of comparative Hanushek, E.A., Wößmann, L., 2007. The role of education quality in economic
development: an empirical investigation. Am. Econ. Rev. 91 (5), 1369–1401. growth, World Bank Policy Research Working Paper 4122, February 2007.
Aditya, A., Acharyya, R., 2013. Export diversification, composition and economic Hanushek, E.A., Wößmann, L., 2008. The role of cognitive skills in economic
growth: evidence from cross-country analysis. J. Int. Trade Econ. Dev. 22, development. J. Econ. Lit. 46 (3), 607–668.
959–992. Hanushek, E.A., Wößmann, L., 2012. Schooling, educational achievement, and the
Afonso, A., Jalles, J.T., 2014. Fiscal composition and long-term growth. Appl. Econ. Latin American growth puzzle. J. Dev. Econ. 99 (2), 497–512.
46 (3), 349–358. Hanushek, E.A., 2013. Economic growth in developing countries: the role of human
Aisen, A., Veiga, F.J., 2013. How does political instability affect economic growth. capital. Econ. Educ. Rev. 37, 204–212.
Eur. J. Political Econ. 29, 151–167. Hartwig, J., 2012. Testing the growth effect of structural change. Struct. Change
Altinok, N., Diebolt, C., Demeulemeester, J.-L., 2014. A new international database Econ. Dyn. 23 (1), 11–24.
on education quality: 1965–2010. Appl. Econ. 46 (11), 1212–1247. Hartwig, J., 2015. Structural change, aggregate demand and employment dynamics
Arellano, M., Bond, S., 1991. Some tests of specification for panel data: Monte Carlo in the OECD, 1970–2010. Struct. Change Econ. Dyn. 34, 36–45.
evidence and an application to employment equations. Rev. Econ. Stud. 58 (2), Hidalgo, C., Hausmann, R., 2009. The building blocks of economic complexity. Proc.
277–297. Natl. Acad. Sci. U. S. A. 106 (26), 10570–10575.
Arellano, M., Bover, O., 1995. Another look at the instrumental variable estimation Hidalgo, C., Klinger, B., Barabasi, A.L., Hausmann, R., 2007. The product space
of error component models. J. Econom. 68, 29–51. conditions the development of nations. Science 317, 482–487.
Barro, R.J., Lee, J.W., 2010. A New Data Set of Education Attainment in the World, Iqbal, N., Daly, V., 2014. Rent seeking opportunities and economic growth in
1950–2010 (www.barrolee.com/data/dataexp.htm, accessed on January 2014). transitional economies. Econ. Model. 37, 16–22.
Barro, R., Sala-i-Martin, X., 2003. Economic Growth, 2nd edn. McGraw-Hill, New Iyigun, M.F., Owen, A.L., 1999. Entrepreneurs, professionals, and growth. J. Econ.
York. Growth 4, 213–232.
Barro, R.J., 1991. ‘Economic growth in a cross section of countries’. Q. J. Econ. 106 Janicki, H.P., Wunnava, P.V., 2004. Determinants of foreign direct investment:
(2), 407–443. empirical evidence from EU accession candidates. Appl. Econ. 36, 505–509.
Batten, J., Vo, X.V., 2009. An analysis of the relationship between foreign direct Jaoul-Grammare, M., Guironnet, J.-P., 2009. Does over-education influence French
investment and economic growth. Appl. Econ. 41 (13), 1621–1641. economic growth? Econ. Bull. 29 (2), 1190–1200.
Baumol, W.J., 1967. Macroeconomics of unbalanced growth: the anatomy of urban Justman, M., Teubal, M., 1991. A structuralist perspective on the role of technology
crisis. Am. Econ. Rev. 57 (3), 415–426. in economic growth and development. World Dev. 19 (9),
Becker, G., 1962. Investment in human capital: a theoretical analysis. J. Political 1167–1183.
Econ. 70, 9–49 (5, part 2). Kaarsen, N., 2014. Cross-country differences in the quality of schooling. J. Dev.
Benhabib, J., Spiegel, M.M., 1994. The role of human capital in economic Econ. 107, 215–224.
development evidence from aggregate cross-country data. J. Monetary Econ. Kaldor, N., 1967. Strategic Factors in Economic Development. New York State
34, 143–173. School of Industrial and Labour Relations, Cornell University, New York: Ithaca.
Benos, N., Zotou, S., 2014. Education and economic growth: a meta-regression Kiersztyn, A., 2013. Stuck in a mismatch? The persistence of overeducation during
analysis. World Dev. 64, 669–689. twenty years of the post-communist transition in Poland. Econ. Educ. Rev. 32,
Blundell, R., Bond, S., 1998. Initial conditions and moment restrictions in dynamic 78–91.
panel-data models. J. Econom. 87 (1), 115–143. Kim, Y.J. Lee, J.-W., 2009. Technological Change, Human Capital Structure, and
Bodman, P., Le, T., 2013. Assessing the roles that absorptive capacity and economic Multiple Growth Paths. ADB Economics Working Paper Series. No. 149,
distance play in the foreign direct investment-productivity growth nexus. Economics and Research Department, Asian Development Bank, Manila.
Appl. Econ. 45 (8), 1027–1039. Krishna, P., Levchenko, A.A., 2013. Comparative advantage, complexity, and
Bogliacino, F., Pianta, M., 2013. Profits, R&D, and innovation-a model and a test. volatility. J. Econ. Behav. Organ. 94, 314–329.
Ind. Corp. Change 22 (3), 649–678. Levine, R., Renelt, D., 1992. A sensitivity analysis of cross-country growth
Bond, S., Hoeffler, A., Temple, J., 2001. GMM Estimation of Empirical Growth regressions. Am. Econ. Rev. 82 (4), 942–963.
Models, C.E.P.R. Discussion Papers, No. 3048. University of Bristol, Bristol, UK. Lipset, Seymour M., 1960. Political Man: The Social Basis of Modern Politics.
Bowsher, C.G., 2002. On testing overidentifying restrictions in dynamic panel data Doubleday, New York.
models. Econ. Lett. 77, 211–220. Mankiw, N.G., Romer, D., Weil, D.N., 1992. A contribution to the empirics of
Breton, T.R., 2011. The quality vs the quantity of schooling: what drives economic economic growth. Q. J. Econ. 107 (2), 407–437.
growth? Econ. Educ. Rev. 30 (4), 765–773. Marelli, E., 2004. Evolution of employment structures and regional specialisation
Caselli, F., Coleman, W.J., 2006. The world technology frontier. Am. Econ. Rev. 96 in the EU. Econ. Syst. 28, 35–59.
(3), 499–522. Mauro, P., 1995. Corruption and growth. Q. J. Econ. 110 (3), 681–712.
Chen, F., Sun, X., 2014. Urban–rural income polarization and economic growth in Metcalfe, J.S., Foster, J., Ramlogan, R., 2006. Adaptive economic growth. Camb. J.
China: evidence from the analysis of a dynamic panel data model. Appl. Econ. Econ. 30, 7–32.
46 (32), 4008–4023. Montobbio, F., 2002. An evolutionary model of industrial growth and structural
Ciccone, A., Papaioannou, E., 2009. Human capital, the structure of production and change. Struct. Change Econ. Dyn. 13, 387–414.
growth. Rev. Econ. Stat. 91 (1), 66–82. Moral-Benito, E., 2012. Determinants of economic growth: a bayesian panel data
Delgado, M.S., Henderson, D.J., Parmeter, C.F., 2014. Does education matter for approach. Rev. Econ. Stat. 94 (2), 566–579.
economic growth? Oxf. Bull. Econ. Stat. 76 (3), 334–359. Nelson, R.R., Pack, H., 1999. The Asian miracle and modern growth theory. Econ. J.
Dias, J., McDermott, J., 2006. Institutions, education, and development: the role of 109 (457), 416–436.
entrepreneurs. J. Dev. Econ. 80, 299–328. Nelson, R.R., Phelps, E.S., 1966. Investment in humans, technological diffusion and
Dias, J., Tebaldi, E., 2012. Institutions, human capital, and growth: the institutional economic growth. Am. Econ. Rev. 56 (1/2), 69–75.
mechanism. Struct. Change Econ. Dyn. 23 (3), 300–312. Nordhaus, W.D., 2008. Baumol’s Diseases: a macroeconomic perspective, The B.E.
Dietrich, A., 2012. Does growth cause structural change, or is it the other way Journal of Macroeconomics, 8 (1) (Contributions), Article 9.
around? A dynamic panel data analysis for seven OECD countries. Empir. Econ. Noseleit, F., 2013. Entrepreneurship, structural change, and economic growth. J.
43 (3), 915–944. Evol. Econ. 23 (4), 735–766.
Dreher, A., 2006. Does globalization affect growth? Evidence from a new index of O’Mahony, M., Timmer, M.P., 2009. Output, Input and Productivity Measures at the
globalization. Appl. Econ. 38 (10), 1091–1110. Industry Level: the EU KLEMS Database (iwww.euklems.net, accessed on
Easterly, W., Levine, R., 1997. Africa’s growth tragedy: policies and ethnic divisions. February 2014).
Q. J. Econ. 112 (4), 1203–1250. Peneder, M., 2003. Industrial structure and aggregate growth. Struct. Change Econ.
Fabro, G., Aixalá, J., 2009. Economic freedom, civil liberties, political rights and Dyn. 14, 427–448.
growth: causality analysis. Span. Econ. Rev. 46, 1059–1080. Peneder, M., 2007. A sectorial taxonomy of educational intensity. Empirica 34 (3),
Fabro, G., Aixalá, J., 2012. Direct and indirect effects of economic and political 189–212.
freedom on economic growth. J. Econ. Issues 46, 1059–1080. Qadri, F.S., Waheed, A., 2013. Human capital and economic growth: cross-country
Feenstra, R.C., Inklaar, R., Timmer, M.P., 2013. The Next Generation of the Penn evidence from low-, middle- and high-income countries. Prog. Dev. Stud. 13
World Table (www.ggdc.net/pwt, accessed on January 2014). (2), 89–104.
Gürbüz, A.A., 2011. Comparing trajectories of structural change. Camb. J. Econ. 35, Quatraro, F., 2009. Innovation, structural change and productivity growth:
1061–1085. evidence from Italian regions, 1980–2003. Camb. J. Econ. 33 (5), 1001–1022.
Ghignoni, E., Verashchagina, A., 2014. Educational qualifications mismatch in Rajan, R.G., Zingales, L., 2008. Financial dependence and growth. Am. Econ. Rev. 88
Europe. Is it demand or supply driven? J. Comp. Econ. 42, 670–692. (3), 559–586.
Glaeser, E.L., La Porta, R., Lopez-de-Silanes, F., Shleifer, A., 2004. Do institutions Romer, P.M., 1990. Endogenous technological change. J. Political Econ. 98 (5),
cause growth? J. Econ. Growth 9 (3), 271–303. 71–102.
Goldin, C., 2016. Human capital. In: Diebolt, C., Haupert, M. (Eds.), Handbook of Roodman, D., 2009. A note on the theme of too many instruments. Oxf. Bull. Econ.
Cliometrics. Springer-Verlag, Berlin, Heidelberg, pp. 55–86. Stat. 71, 135–158.
Hall, S.G., Ahmad, M., 2014. Revisiting the institutions-growth nexus in developing Saviotti, P., Frenken, K., 2008. Export variety and the economic performance of
countries: the new evidence. N. Z. Econ. Pap. 48 (3), 301–312. countries. J. Evol. Econ. 18, 201–218.
Hall, R.E., Jones, C.I., 1999. Why do some countries produce so much more output Saviotti, P., Pyka, A., 2012. On the co-evolution of innovation and demand: some
per worker than others. Q. J. Econ. 114 (1), 83–116. policy implications. Revue de l’OFCE 124 (5), 347–388.
Saviotti, P., Pyka, A., 2013. From necessities to imaginary worlds: structural Temple, J., Wö␤mann, L., 2006. Dualism and cross-country growth regression. J.
change, product quality and economic development. Technol. Forecasting Soc. Econ. Growth 11 (3), 187–228.
Change 80, 1499–1512. van Leeuwen, B., 2007. Human Capital and Economic Growth in India, Indonesia,
Saviotti, P., Pyka, A., 2016. Innovation, structural change and demand evolution: and Japan: A Quantitative Analysis, 1890–2000 Ph.D. Thesis. Utrecht
does demand saturate? J. Evol. Econ., http://dx.doi.org/10.1007/s00191-015- University.
0428-2 (forthcoming). van Leeuwen, B., Földvári, P., 2013. Capital accumulation and growth in central
Saviotti, P., 2001. Special issue: variety, growth and demand. J. Evol. Econ. 11 (1), Europe 1920–2006. East. Eur. Econ. 51 (5), 69–93.
119–142. Vandenbussche, J., Aghion, P., Meghir, C., 2006. Growth, distance to frontier and
Schultz, T.W., 1961. Investment in human capital. Am. Econ. Rev. 51 (1), 1–17. composition of human capital. J. Econ. Growth 11, 97–127.
Sianesi, B., Reenen, J.V., 2003. The returns to education: macroeconomics. J. Econ. Wö␤mann, L., 2003. Specifying human capital. J. Econ. Surv. 17 (3), 239–270.
Surv. 17 (2), 157–200. Wilson, R.A., Briscoe, G., 2004. The impact of human capital on economic growth: a
Silva, E.G., Teixeira, A.A.C., 2011. Does structure influence growth? A panel data review. In: Descy, P., Tessaring, M. (Eds.), Impact of Education and Training,
econometric assessment of ‘relatively less developed’ countries, 1979–2003. Third Report on Vocational Training Research in Europe: Background Report.
Ind. Corp. Change 20 (2), 457–510. Luxembourg : Office for Official Publications of the European Communities
Sunde, U., Vischer, T., 2015. Human capital and growth: specification matters. (Cedefop Reference series, 54).
Economica 82 (326), 368–390. Witt, U., 2001. Learning to consume—a theory of wants and the growth of demand.
Teixeira, A.A.C., Fortuna, N., 2011. Human capital, R&D, trade, and long-run J. Evol. Econ. 11, 23–36.
productivity testing the technological absorption hypothesis for the Zagler, M., 2009. Economic growth, structural change and search unemployment. J.
Portuguese economy, 1960–2001. Res. Policy 39 (3), 335–350. Econ. 96, 63–78.
Teixeira, A.A.C., 2012. Reconciling orthodox and heterodox approaches to
economic growth. A modeling proposal. In: Teixeira, A.A.C. (Ed.), Technological
Change, Ch. 4. InTech.

Teixeira 2016

Uploaded by

Copyright:

Available Formats

Teixeira 2016

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Teixeira 2016

Uploaded by

Copyright:

Available Formats

Research Policy 45 (2016) 1636–1648

Contents lists available at ScienceDirect

Economic growth, human capital and structural change: A dynamic

where y is the natural logarithm of real GDP per capita, hc rep-

Variable Proxy Model I Model II

21 countries; 52 years (1960–2011) 30 countries; 22 years (1990–2011)

Dynamic panel-data estimation, one Dynamic panel-data estimation,

A.A.C. Teixeira, A.S.S. Queirós / Research Policy 45 (2016) 1636–1648

A.A.C. Teixeira, A.S.S. Queirós / Research Policy 45 (2016) 1636–1648

You might also like

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.