Economic Modelling 61 (2017) 93–101

Contents lists available at ScienceDirect

Economic Modelling
journal homepage: www.elsevier.com/locate/econmod

Growth, human development, and trade: The Asian experience MARK

a b,⁎ c,1
Ghulam Mustafa , Marian Rizov , David Kernohan
a
Federal Urdu University, Pakistan
b
University of Lincoln, UK
c
Middlesex University Business School, UK

A R T I C L E I N F O A BS T RAC T

JEL classification: This study looks at the three-way relationship between economic growth, human development, and openness to
F43 trade in a large panel of developing Asian economies. Using a theoretically motivated simultaneous equations
O10 system, we find that although human development contributes positively to economic growth, in the case of our
O15 Asian sample growth does not appear to have had a positive influence on human development. Uneven growth
O19
accompanied by lagging institutional development, preventing human capital formation, might have inhibited
O53
human development in the short to medium run. Complementary to the literature showing that growth is
Keywords: sustainable only when accompanied by human development, we confirm a role for trade liberalisation policies in
Economic growth
achieving higher growth as well as human development.
Human development
Trade
Openness
Asia

1. Introduction child labour (Dagdemir and Acaroglu, 2010; Neumayer and De Soysa,
2005). We build on the recent literature, notably Suri et al. (2011), that
The subject of this paper is the relationship between economic has uncovered subtle causal interactions between HD and EG in
growth (EG) and human development (HD). Recent work on develop- developing countries. But we also build OT into our analysis, since it
ment and growth has suggested that human capital accumulation may has long been at the core of economic orthodoxy in development policy.
be important in enhancing economic growth as well as human By examining this three-way link between EG, HD, and OT, the
development (Suri et al., 2011). We widen the debate by also more complete model is capable of addressing not only outcomes but
considering the role of trade liberalisation, which has a long pedigree also the factors that drive those outcomes. Our approach is consistent
in the policies of development organisations such as the World Bank, with the recent literature that emphasises the socio-economic role
IMF and WTO (Wang et al., 2004). played by institutions (education, governance quality, social develop-
The empirical literature on the relationship between openness to ment, etc.) as long-run determinants of development and growth
trade (OT) and economic growth has had somewhat mixed results (see (Acemoglu et al., 2005). Our findings suggest that development policy
Frankel and Romer, 1999; Greenaway et al., 2002; Falvey et al., 2012). can be considered as a three-way mix of openness, growth and
Most authors conclude that openness has generally improved economic development. Focussing on human development earlier in the process
growth in developing countries, however the precise channel through can help sustain growth, while openness to trade may be appropriate in
which it can help achieve balanced economic growth does not appear to cases where socio-economic conditions and the quality of institutions
be straightforward.2 are at an adequate level.
The ‘conventional’ economic approach to development holds that Despite strong arguments (Acemoglu et al., 2005) that political
trade liberalisation has a generally positive impact on poverty allevia- institutions underlie the poverty traps besetting many countries growth
tion. A more sceptical view has seen globalisation as a channel for records, there has been relatively little analysis or agreement on
exploiting developing countries' low labour costs, for example through whether inadequate HD has a role in sustaining such traps. Barro

⁎
Correspondence to: Lincoln International Business School, Brayford Pool, Lincoln LN6 7TS, UK.
E-mail address: [email protected] (M. Rizov).
1
The paper draws on Ghulam Mustafa's doctoral research; nevertheless, the authors' contributions to this paper are equal. We thank Michela Vecchi and John Grahl for contributions
to earlier versions of the paper and the journal editor and anonymous referees for constructive comments.
2
For example Cooray et al. (2014) show that the impact of openness on growth is importantly moderated by the gender-specific levels of primary and secondary education.

http://dx.doi.org/10.1016/j.econmod.2016.12.007
Received 26 January 2016; Received in revised form 8 December 2016; Accepted 9 December 2016
Available online 14 December 2016
0264-9993/ © 2016 Elsevier B.V. All rights reserved.
G. Mustafa et al. Economic Modelling 61 (2017) 93–101

(2000), for example, sees HD as a ‘good’ which wealthier countries

.6
choose to supply to their population. Against this, we can set Amartya
Sen's (1999) argument in favour of all types of HD. This approach

.5
found empirical support in Blume and Voigt (2007), who found
positive relationships between elements of HD and economic develop-
ment. Econometric modelling by Suri et al. (2011) has shown that bi-

.4
directional causality can exist between HD and EG. Thus the former
can be viewed not only as an outcome of EG, but also as an essential
precondition for achieving it. Our goal is to test further whether such

.3
1970 1980 1990 2000 2010
positive, bi-directional effects are robust to the inclusion of a third
Year
explanatory factor - openness to trade – since this has been an hd openness
important factor in standard growth equations (e.g., Cooray et al.,
Fig. 1. Openness and human development in Asia. Notes: Openness is measured by
2014).
economic globalization calculated as in Dreher (2006) and human development (hd) is
The literature on trade liberalisation has generally taken the view authors own calculations.
that it increases economic growth (e.g., Frankel and Romer, 1999;
Greenaway et al., 2002; Wang et al., 2004; Falvey et al., 2012). By
contrast, an influential strand of international economics increasingly
concerns itself with socio-economic phenomena. Hence we extend the 2. Data and method
analysis of HD to include trade openness in line with work by Nunn
(2007), which has looked at the relative quality of national institutions 2.1. Data
(security, law, governance) in trade performance. But also this
approach is in keeping with work that has looked at the role of social, For this paper we assembled a dataset including panel observations
institutional and political factors in EG (e.g., Acemoglu et al., 2005; from twelve developing Asian countries, over forty-two years (1970–
Acemoglu et al., 2008; Tabellini, 2010). In this stylised view, the social 2011). The countries are Bangladesh, India, Nepal, Pakistan, Sri Lanka,
and institutional components inherent in HD are often not only ‘deeply Indonesia, Malaysia, Philippines, Singapore, South Korea, Thailand,
embedded’ but usually also long-run in nature. These long-run, deeply- and China. The data come from several sources. Real GDP at PPP
embedded processes may play a part in how EG, HD, and OT interact exchange rates and employment data is collected from the Conference
in the development process. Furthermore, such subtle relationships Board (2011). We complement this data with information from
may not have been easily picked up in ‘conventional’ economic studies. Deininger and Lyn (1996), Dreher (2006), WIDER (2008), Barro and
And this oversight may have been largely due to their use of single Lee (2010), IMF (2011), UNDESA (2011), UNDP (2011), and World
equation frameworks, shorter data sets, and pervasive endogeneity Bank (2012). Table 1 provides a brief description, summary statistics,
problems. Taken together, these difficulties may have served to conceal and sources of the variables used in the analyses that follow.
the economic significance of some deep lying, socio-economic phenom- We use the UNDP (2011) methodology to construct a time-varying
ena. HD index (HDI) as an indicator of human development. This index has
Asian economic development has generally been characterised by a been designed to emphasize the role of human welfare as a develop-
disparity between levels of human development and economic growth ment policy goal (and outcome) rather than focussing only on
(Suri et al., 2011). Not only has the literature on improving HD, and economic growth (Klugman et al., 2011). The HDI aims to measure
that on generating EG, tended to proceed on separate lines but also the human development and capabilities in three dimensions: (i) long and
HD literature has tended to view development mainly as an output of healthy life; (ii) knowledge and human capital; and (iii) a decent
economic growth rather than a potential contributing factor. standard of living. The HDI is based on the human capital measure
We focus on estimating a three-way relationship between EG, HD, used by Cohen and Soto (2007), for which we obtained data from Barro
and OT in the context of Asian economic development. Even if there is and Lee (2010).3
no simple association between openness and growth (e.g. Cooray et al., To measure trade openness, we use a globalization sub-index from
2014), improvements in human development may be a pre-requisite the KOF Globalization Index (Dreher, 2006) as a broad measure of
for sustained growth (Ranis et al., 2000; Suri et al., 2011) since trade trade openness (OP1) which is our preferred OT measure. The KOF
openness may interact with both these variables. Our sample of Globalisation Index is a composite index comprising an economic
developing countries is highly relevant to investigating this three-way globalization index, a social globalization index, and a political
relationship. China and India are countries which adopted trade globalization index. To check the robustness of our results we also
liberalisation policies only after achieving higher rates of economic use a trade volume measure of openness (OP2), from the Penn World
growth, while the East Asian smaller economies are often cited as Tables and a final measure (OP3) from the World Bank (2012).
successful examples of export-led growth. Furthermore, Fig. 1 shows a
strong positive association between openness and human development
in the Asian economies. 2.2. Analytical framework and estimation methodology
Among the key relationships we set out to test are: is trade
liberalisation a pre-requisite for economic growth, or the result of The starting point of our analytical framework is the standard
sustained output growth? Further, are there any systematic links Cobb-Douglass country-level production function with constant re-
between trade openness and economic growth and are the welfare turns to scale as used in Cooray et al. (2014):
consequences from trade liberalisation reflected in the level of human Yit =Ai 0 Kitα Lit(1− α ) e φZit , (1)
development?
The rest of the paper is organized as follows. In Section 2 we where Yit is aggregate output of country i in period t, Aio is total factor
provide a description of the dataset and then set out a theoretically productivity, Kit is the stock of physical capital, and Lit is the labour
motivated framework for the empirical analysis and econometric
methodology. Section 3 reports our estimation results, provides 3
Human capital stock (H) is constructed using Cohen and Soto (2007) methodology
robustness checks, and includes a discussion of our main findings. and employing Barro and Lee (2010) data. We use a depreciation rate of 5% following
Section 4 provides a brief summary in the context of the literature and Wang and Yao (2003). Details on the calculation methods for H and for HDI are available
draws some broader conclusions. on request.

94
G. Mustafa et al. Economic Modelling 61 (2017) 93–101

Table 1
Description and source of regression variables.

Variable Definition Source Mean S.D.

ΔY (real GDP) Growth rate in GDP in 1990 US$ (Geary Khamis Conference Board (2011), World Bank (2012) 0.054 0.036
PPPs)
ΔL (labour force) Growth rate in employment Conference Board (2011) 0.024 0.025
ΔK (physical capital stock) Growth rate in capital, constructed series using World Bank (2012) 0.092 0.059
PIM method
H (human capital index) Index constructed using Cohen and Soto (2007) Authors own calculations using Barro and Lee (2010) data 4.00 1.357
methodology
OP1 (openness) Economic globalization Index Dreher (2006) 0.422 0.225
OP2 (openness) Ratio of exports plus imports to GDP Penn World Tables 7.0 0.807 0.907
OP3 (openness) Ratio of exports plus imports to GDP World Bank (2012) 0.816 0.897
GLOB Globalisation index Dreher (2006) 0.431 0.179
HDI (human development index) Composite index of income, health and education Authors own calculations using UNDP (2011), UNDESA (2011), 0.519 0.134
indices Barro and Lee (2010)
EDU Education index A sub index of HDI, based on Cohen and Soto methodology 0.403 0.125
LEI Life expectancy index A sub index of HDI 0.713 0.124
IMR (infant mortality rate): per Mortality rate, under-5 (per 1000) World Bank (2012) 0.700 0.586
thousand
FDI (foreign direct investment) Ratio of annual net inflows of FDI to GDP World Bank (2012) 0.020 0.035
MS (market size) Log total population World Bank (2012) 17.977 1.611

Notes: Number of observations is 492.

force. The vector Zit contains control variables that affect growth such growth in both the augmented neoclassical growth model (Mankiw
as human capital, policies, and institutions identified in the literature. et al., 1992) and the endogenous growth model (Lucas 1988; Romer,
Notably, in the Zit vector we include our human development and 1990). Empirical growth studies have often found it difficult to show
openness indicators. The standard production function is a natural the strong positive impact of human capital on economic growth
theoretical framework for our analysis given that it is the foundation of predicted by theoretical models.5 The difficulties encountered in
the neoclassical (Solow) growth model, where economic growth is linking H variables to EG growth may stem from methodological
determined by investments in physical and human capital and employ- issues, such as the inclusion of skills in the measurement of human
ment growth which, taken together, ultimately influence human capital, and the identification of channels through which it affects EG
development. (Sianesi and Van Reenen, 2003; Cooray et al., 2014; Qadri and
Following our discussion of the bi-directional causality between Waheed, 2014). We follow Mankiw et al. (1992) and include human
economic growth (EG) and human development (HD) and the moder- capital as an additional input, as it is expected to produce long-run
ating effects of openness to trade (OT), we opt for a system of growth even in the absence of technological advancements (Lucas,
simultaneous equations as our empirical specification. We believe this 1988). Our a priori expectations are that α1, α2, α3, α4, and α5 are all
estimation strategy, in essence an instrumental variables approach, is a positive.
reasonable way of dealing with the severe endogeneity and reverse In setting up our empirical HD equation, we draw from the
causality problems that characterise single equation specifications capabilities approach (Anand and Sen, 1994, 2000). This postulates
containing EG, HD, and OT. Opting for a structural, multi-equation that the accumulation of human capital and health facilities are
empirical framework allows us to study the determinants of each of the important for both economic growth and human development.
three variables of interest rather than trying to control for and limit Openness to trade may also affect human development by directly or
their impacts on each other. indirectly facilitating access to goods and services, through income
We set up a three simultaneous equations empirical model based on growth. Although trade liberalisation can raise growth in exports and
the production function (1) to study the interrelationships between EG, imports, the balance of payments consequences depend upon its
HD, and OT in a panel of twelve major Asian countries.4 Our empirical relative impact and on any relative shifts in the prices of traded
EG Eq. (2) closely resembles a neoclassical growth equation, derived commodities (e.g., Thirwall, 2012).
from Eq. (1) but augmented with indicators of trade openness and We specify the HD Eq. (3) based on development theory and
human development: evidence from recent empirical research (Anand and Sen, 2000; Binder
and Georgiadis, 2011).
∆Yit = α1 ∆L it + α2 ∆Kit + α3 ∆Hit + α4 OPit + α5 HDIit + μi + T +εit ,
(2) HDIit =β1 ∆Yit + β2 OPit + β3lnHit + β4 IMRit +μi + T + εit , (3)

where ∆Y is the growth rate of output, ∆L is the growth rate of where HDI is the human development index, IMR stands for infant
employment, ∆K is the growth rate of physical capital stock, ∆H is the mortality rate and other variables are previously defined. Wagstaff
growth rate of human capital stock, OP is the level of trade openness (2002) suggests that there is two-way causation between poverty and ill
(the OP1 measure), and HDI is the level of human development. The
term μi is the individual country effect, T is a time trend, and εit is the 5
In a survey of macroeconomic literature on the link between education and growth
zero mean error term which varies across countries and time. Sianesi and van Reenen (2003) conclude that there is compelling evidence on the positive
Human capital plays an important role in stimulating economic impact of human capital on productivity growth; the evidence is also consistent with
findings by Cameron et al. (2005) and Bournakis (2012). However, the empirical
evidence on both OECD and developing countries in favour of new growth theories is
4
In our robustness analysis we also estimate single equation specifications using weak. Moreover, there is still no consensus on whether the stock of human capital
GMM SYS which is a popular alternative instrumental variables approach for dealing influences the level of income in long run (augmented neoclassical models) or the long
with endogeneity problems. run growth rate suggested by endogenous growth theories.

95
G. Mustafa et al. Economic Modelling 61 (2017) 93–101

health. IMR is an important indicator linked closely to individuals’ Table 2

health conditions, levels of poverty and human development. Base Specification using OLS, 2SLS and 3SLS.
Generally, developing countries with low levels of income are expected
Variable OLS 2SLS 3SLS
to have high IMR rates while lower IMR rates would reflect improve- EG (ΔY) EG (ΔY) EG (ΔY)
ments in economic development in more than one dimension. There is
also recent evidence that, on balance, more open countries have higher ΔL 0.202*** 0.204*** 0.176***
(0.062) (0.064) (0.056)
levels of human development and that returns to schooling are positive
ΔH 0.138 0.209* 0.188*
(Human Development Report, 2015). To allow for the possibility of (0.102) (0.111) (0.097)
reverse causality and multicollinearity in our empirical analysis we ΔK 0.253*** 0.290*** 0.225***
experiment with lagged IMR and step-wise introduce the explanatory (0.040) (0.043) (0.040)
variables. Our a priori expectations are that β1, β2, and β3 are positive, OP -0.063* 0.075 0.195***
(0.034) (0.061) (0.058)
while β4 < 0.
HDI 0.101 0.477*** 0.260**
The specification of our empirical OT Eq. (4) is based on a version (0.080) (0.138) (0.131)
of gravity model modified by Guttmann and Richards (2006) and our HD (HDI) HD (HDI) HD (HDI)
preceding discussion on the interrelationships between EG, HD, and IMR -0.008 -0.027** -0.020**
(0.007) (0.013) (0.010)
OT. The equation links openness with economic growth, human
lnH 0.038*** 0.100*** 0.057***
development, foreign direct investment and market size for the sample (0.007) (0.015) (0.012)
of Asian countries. Thus, ΔY -0.040 -0.324*** -0.514***
(0.024) (0.107) (0.101)
OPit =γ1 ∆Yit + γ2 HDIit + γ3 FDIit + γ4 MSit +μi + T + εit , (4) OP -0.032 0.396*** 0.422***
(0.020) (0.069) (0.065)
where OP is our preferred trade openness measure, OP1, FDI stands
OT (OP) OT (OP) OT (OP)
for foreign direct investment, MS represents the market size, measured MS -0.333*** -0.322*** -0.277***
by population, and other variables are previously defined. Rigobon and (0.023) (0.027) (0.026)
Rodrik (2005) among others demonstrate that income has a positive FDI 0.010 0.130 0.212***
impact on openness and that variables related to geography such as (0.094) (0.110) (0.078)
ΔY -0.115** 0.423*** 0.455***
market size are the most important determinant of openness. Tsen (0.049) (0.140) (0.131)
(2006) found that during the liberalisation period 1978–1999 in China, HDI -0.557*** -0.327** 0.006
economic growth and openness Granger cause each other in both (0.086) (0.156) (0.143)
directions. In the light of the above discussion we expect that γ1, γ2,
and γ3 are positive, and γ4 < 0. Notes: ***, **, and * denote statistical significant at 1%, 5%, and10% level respectively;
figures in parentheses are the standard errors. Number of observations is 492. Country
In all equations institutional factors that evolve slowly or remain
fixed effects and time trend are included in each equation.
fixed over the period of analysis are accounted for by country fixed
effects. In addition, a time trend is included in all specifications to
control for technological progress and business cycles.
We estimate the system of three simultaneous equations specified
above using a three-stage least squares estimator (3SLS). The 3SLS is with a p-value of 0.02. Thus, we cannot reject the null hypothesis of
superior to the two-stage instrumental variables estimator (2SLS) as it validity of instruments, while we do reject the null hypothesis of
is a combination of 2SLS and seemingly unrelated regression estimator underidentification. We also run Breusch-Pagan LM diagonal covar-
(SURE) and is consistent and more efficient than the 2SLS (Kennedy, iance matrix test of 3SLS validity; the LM test statistic is 384.26 (χ2(3))
2009). We include only the most important and theoretically motivated with a p-value of 0.001. Therefore, we reject the null hypothesis of
variables in each equation, to reduce problems of misspecification. To diagonal disturbance covariance matrix – a result in support of the
identify potential endogeneity issues we step-wise introduce explana- choice of 3SLS.
tory variables in all equations. To investigate potential problems of For all three simultaneous equations we consider specifications
omitted variable bias we conduct robustness checks, where we augment with the three alternative measures of trade openness (OP1, OP2 and
the base specifications outlined above with additional relevant vari- OP3) discussed earlier but report only results from the specification
ables according to theory. with our preferred measure, OP1 while results from specifications with
the two alternative (partial) measures are available from the authors.
3. Results

Table 2 provides comparisons of the estimates of our system of 3.1. EG equation

equations using pooled OLS, 2SLS, and 3SLS. The three sets of results
are comparable; less comparable but still qualitatively similar are the Table 3, column (1) reports 3SLS estimates of the EG Eq. (2) within
results from fixed effects estimations which are available from the our three-equation system. The estimated coefficients of employment
authors. The estimates produced by 3SLS are the closest to the a priori and physical capital growth rates are positive and statistically signifi-
(theoretical) expectations. Furthermore, 3SLS provides relatively more cant in all specifications, as predicted by neoclassical growth theory.
precise estimates. The estimated coefficient of growth in human capital (H) is positive and
Given that 3SLS is effectively an instrumental variables (IV) also significant at conventional levels. Thus, we find strong evidence
estimator we formally test if HD and OP equations accurately identify that the growth in human capital stimulates economic growth in Asian
the EG equation; the significance threshold p-value is set at 0.05. The economies. A one percent growth in human capital stock is associated
Hansen J-statistic of the Sargan-Hansen overidentification test is 5.53 with about a 0.2 percent increase in income growth. Thus far, our
(χ2(2)) with a p-value of 0.07, while the Kleibergen-Paap Lagrange results provide support for the human capital theory and endogenous
Multiplier (LM) statistic of the underidentification test is 9.84 (χ2(3)) growth models of Lucas (1988) and Romer (1990) and are consistent

96
G. Mustafa et al. Economic Modelling 61 (2017) 93–101

Table 3 openness measure (results available from the authors) they are likely
3SLS estimates of the base specification. to also capture the imprecision in the openness measure.6
Variable EG (ΔY) Variable HD (HDI) Variable OT (OP)
3.2. HD equation
(1) (2) (3)

ΔL 0.176*** IMR -0.020** MS -0.277*** Table 3, column (2) reports the regression estimates from the
(0.056) (0.010) (0.026) human development Eq. (3). As per a priori expectations, a higher
ΔH 0.188* lnH 0.057*** FDI 0.212*** infant mortality rate (IMR) leads to lower levels of human development
(0.097) (0.012) (0.078)
as the IMR coefficient is negative and statistically significant. A one
ΔK 0.225*** ΔY -0.514*** ΔY 0.455***
(0.040) (0.101) (0.131)
standard deviation increase in IMR is associated with a 0.01 (which is
OP 0.195*** OP 0.422*** HDI 0.006 about 2 percent of the mean) fall in the level of the human development
(0.058) (0.065) (0.143) index. This result shows that the poor health environment indicated by
HDI 0.260** high IMR hampers human development in Asian countries even though
(0.131)
the magnitude of the effect is small.
Trend -0.002*** Trend 0.001** Trend 0.012***
(0.001) (0.001) (0.001) Our findings in Table 3, column (2) show that an increase in the
Bangladesh 0.070* Bangladesh -0.173*** Bangladesh -0.813*** level of human capital (H) has a statistically significant positive impact
(0.038) (0.014) (0.081) on the level of human development. A one percent increase in the level
India 0.059* India -0.160*** India -0.143*** of human capital is associated with a 0.0006 improvement in the level
(0.031) (0.009) (0.036)
Nepal 0.044 Nepal -0.178*** Nepal -1.250***
of human development, which represents a more than one percent
(0.038) (0.016) (0.122) increase in the HDI's mean for a ten percent increase in H. The result
Pakistan 0.054 Pakistan -0.181*** Pakistan -0.665*** provides support for the proposition that human capital stock accu-
(0.033) (0.010) (0.077) mulation is an important factor in enhancing human development.
Sri Lanka 0.003 Sri Lanka -0.089*** Sri Lanka -1.163***
Sianesi and Van Reenen (2003) call such positive effects ‘positive
(0.013) (0.007) (0.110)
Indonesia -0.012 Indonesia -0.185*** Indonesia -0.392*** educational externalities’, as the educated labour force is associated
(0.022) (0.010) (0.056) with increases in technological progress, improvements in productivity,
Malaysia -0.065*** Malaysia -0.190*** Malaysia -0.795*** and further investments in human capital which in turn further raise
(0.022) (0.022) (0.106) productivity.
Philippines -0.024 Philippines -0.168*** Philippines -0.659***
(0.018) (0.013) (0.079)
Importantly, our results also suggest that economic growth and
Singapore -0.132*** Singapore -0.213*** Singapore -1.059*** human development may be substitutes, at least in the case of the
(0.034) (0.036) (0.148) developing Asian countries from our sample. Results in Table 3,
Korea -0.038*** Korea -0.039*** Korea -0.773*** column (2) suggest that economic growth may have hampered human
(0.011) (0.009) (0.083)
development in the Asian economies studied. However, the economic
Thailand -0.013 Thailand -0.134*** Thailand -0.708***
(0.015) (0.010) (0.080) significance of the effect is quite small: a one percentage point increase
in income growth is associated with a 0.005 (about 1 percent of the
Notes: ***, **, and * denote statistical significant at 1%, 5%, and 10% level respectively; mean) fall in the level of HDI. By focusing on faster growth, Asian
figures in parentheses are the standard errors. Number of observations is 492. China is economies may have lost some ground in human development. In
the reference country. The test statistic for Breusch-Pagan LM test is 384.26 with a p-
prioritising growth, in the presence of unfavourable institutional
value of 0.001 in favour of 3SLS.
quality, an unfair distribution of assets and income may prevent the
transformation of EG into better HD performance. Potential political
and economic instability may ensue (Ranis et al., 2000), in turn
hampering, economic performance given our results from the EG
with empirical findings at both industry and country level (Mason et al., equation.
2012 and Sunde and Vischer 2015 respectively). Openness to trade is often associated with implications for income
Our results also provide evidence that openness fosters growth in generation and distribution in developing countries. Our results
Asian countries, consistent with studies by Wacziarg and Welch (2008) suggest a one S.D. increase in openness is associated with about a
and Shahbaz (2012). The estimated coefficient of openness is positive 0.1 improvement in the level of the human development index - almost
and significant at the 1 percent level. This indicates that a one standard 20 percent of the mean. The explanation here is that trade reforms in
deviation increase in openness is associated with a four percentage Asian economies have created new markets with diversified commod-
points (about 75 percent of the mean) increase in the growth rate. ities and better access to products, and thus improved consumer
Interestingly, our results also provide evidence that improvements welfare (Winters et al., 2004).
in the level of human development (HDI) enhance growth in the Overall, the coefficients on the country dummies are negative and
region. The estimated coefficient of HDI is positive and highly significant suggesting that the large majority of Asian countries in the
statistically significant. This suggests that improvements in HDI, sample are not better off in terms of human development as compared
reflecting socio-economic factors, institutions, and freedom have to China.
increased economic growth in the Asian economies. A one standard
deviation improvement in the level of human development is asso-
3.3. OT equation
ciated with a 0.02, or a two percentage points (which is almost 40
percent of the mean) increase in the growth rate.
Table 3, column (3) presents regression estimates of the trade
The country dummies capture fixed and unobservable effects such
openness Eq. (4). Market size (MS) has a negative and significant
as institutional factors, relative to China, not captured by other
impact on openness, consistent with the argument that large econo-
explanatory variables. The dummy variables for India and
mies are less open than small ones. FDI has a positive and statistically
Bangladesh have positive and significant coefficients suggesting that
our model predicts higher growth in these countries relative to China,
had the unobserved country conditions been more favourable. Given 6
Besides this point, while pre-1978 China experienced an annual real GDP growth of
that the coefficients on the dummies change with the choice of 3.8% per year, post-1978 China saw real GDP growth of 8.7% per year. This shift in
regime could also partially explain the varied performance of the dummy variable set.

97
G. Mustafa et al. Economic Modelling 61 (2017) 93–101

Table 4 Table 6
Globalisation, economic growth, and human development. Openness, economic growth, and life expectancy.

Variable EG (ΔY) Variable HD (HDI) Variable OT (GLOB) Variable EG (ΔY) Variable HD (LEI) Variable OT (OP)

(1) (2) (3) (1) (2) (3)

ΔL 0.159*** IMR -0.014 MS -0.247*** ΔL 0.167*** IMR -0.145*** MS -0.289***

(0.061) (0.009) (0.020) (0.056) (0.012) (0.025)
ΔH 0.317*** lnH 0.076*** FDI 0.318*** ΔH 0.206** lnH 0.016 FDI -0.018
(0.104) (0.009) (0.074) (0.093) (0.014) (0.081)
ΔK 0.318*** ΔY -0.354*** ΔY -0.105 ΔK 0.229*** ΔY -0.552*** ΔY 0.444***
(0.042) (0.101) (0.106) (0.038) (0.128) (0.148)
GLOB -0.129 GLOB 0.525*** HDI -0.293** OP 0.181*** OP 0.289*** LEI 0.032
(0.089) (0.089) (0.115) (0.057) (0.080) (0.082)
HDI 0.474*** LEI 0.246***
(0.132) (0.059)

Notes: ***, **, and * denote statistical significant at 1%, 5%, and 10% level respectively; Notes: ***, **, and * denote statistical significant at 1%, 5%, and10% level respectively;
figures in parentheses are the standard errors. Number of observations is 492. Country figures in parentheses are the standard errors. Number of observations is 492. Country
fixed effects and time trend are included in each equation. The test statistic for Breusch- fixed effects and time trend are included in each equation. The test statistic for Breusch-
Pagan LM test is 183.17 with a p-value of 0.001 in favour of 3SLS. Pagan LM test is 411.73 with a p-value of 0.001 in favour of 3SLS.

significant effect on the openness of Asian economies. This is in line tion on growth.
with trade theories which suggest that FDI and openness are com- To identify the net effect of globalisation, we further investigate the
plementary in nature as higher levels of FDI make economies more empirical link between an aggregate measure of globalisation, econom-
open and internationally competitive. ic growth and human development in our sample of Asian economies.
We find that economic growth promotes greater openness to In Table 4, we replace the openness (OP) variable with a globalisation
international trade. This result reflects the growth experience of many index (GLOB) and observe that the results remain stable and in line
Asian countries where trade liberalisation policies have been adopted with results reported in Table 3. The main finding is that globalisation
after achieving higher economic growth (notably, India and China). has had a positive impact on human development. A one standard
Human development seems to have a positive effect on openness deviation increase in GLOB is associated with an improvement in the
although the coefficient on HDI in our base specification is not level of human development of about 0.1, which is almost 20 percent of
statistically significant. the mean. However, globalisation has no effect on economic growth as
The coefficients on the country dummies are negative and statisti- the coefficient of GLOB in the growth equation is not statistically
cally significant suggesting that China has a more open economy than significant. One explanation could be that although the Asian econo-
other Asian countries once its size is taken into account. mies have started integrating in more recent years, historically they
have not been well integrated into the world economy. The findings
3.4. Robustness analyses taken together also suggest that the globalisation process involves
much more than improving economic growth alone.
3.4.1. System equation specifications Next we replace HDI with its knowledge and education sub-index
Existing empirical studies provide some support for the argument (EDU). This helps us test for the robustness of our results as well as in
that the net impact of globalisation has been positive. However, these directly examining the links between openness, economic growth and
studies use trade flows or other partial openness measures to proxy education. The results in Table 5 suggest that our conclusions from
globalisation and thus cannot capture the overall impact of globalisa- Table 3 remain unchanged. A major finding here is a bi-directional

Table 5 Table 7
Openness, economic growth, and education. Base specification with FDI.

Variable EG (ΔY) Variable HD (EDU) Variable OT (OP) Variable EG (ΔY) Variable HD (HDI) Variable OT (OP)

(1) (2) (3) (1) (2) (3)

ΔL 0.147*** IMR -0.013 MS -0.257*** ΔL 0.134** IMR -0.0268*** MS -0.282***

(0.053) (0.009) (0.025) (0.054) (0.010) (0.026)
ΔH 0.210** lnH 0.112*** FDI 0.156* ΔH 0.212** lnH 0.0662*** FDI -0.189*
(0.090) (0.012) (0.087) (0.093) (0.014) (0.108)
ΔK 0.215*** ΔY -0.518*** ΔY 0.547*** ΔK 0.208*** ΔY -0.881*** ΔY 0.623***
(0.039) (0.089) (0.134) (0.040) (0.132) (0.134)
OP 0.273*** OP 0.354*** EDU -0.144 OP 0.185*** OP 0.524*** HDI -0.171
(0.060) (0.058) (0.111) (0.058) (0.076) (0.147)
EDU 0.235** HDI 0.291** FDI 0.462***
(0.102) (0.133) (0.092)
FDI 0.176**
Notes: ***, **, and * denote statistical significant at 1%, 5%, and 10% level respectively; (0.086)
figures in parentheses are the standard errors. Number of observations is 492. Country
fixed effects and time trend are included in each equation. The test statistic for Breusch- Notes: ***, **, and * denote statistical significant at 1%, 5%, and 10% level respectively;
Pagan LM test is 382.52 with a p-value of 0.001 in favour of 3SLS. figures in parentheses are the standard errors. Number of observations is 492. Country
fixed effects and time trend are included in each equation. The test statistic for Breusch-
Pagan LM test is 514.62 with a p-value of 0.001 in favour of 3SLS.

98
G. Mustafa et al. Economic Modelling 61 (2017) 93–101

Table 8 equations are consistent across the two estimation approaches. In the
Comparison of system and single equation estimates for EG, HD, and OT. human development (HD) equation our original finding regarding the
negative impact of economic growth on human development is not fully
Dependent variable Specification EG (ΔY) HD (HDI) OT (OP)
(1) (2) (3) supported, as the GMM SYS estimate is positive but not statistically
significant at any conventional level. This result leads us to conclude
ΔY System −*** +*** only cautiously that economic growth may not have a positive impact
Single + +
on human development, in the context of our sample.
HDI System +** +
Single +*** +***
OP System +*** +***
Single + +** 4. Summary and conclusions

Notes: The statistical significance and the sign of the coefficients are reported; ***, **,
We set out to investigate the links between economic growth (EG),
and * denote statistical significant at 1%, 5% and 10% level respectively. Number of
observations is 492 in all regressions. System equation specifications are estimated by
human development (HD), and openness to trade (OT) for twelve Asian
3SLS. Single equation specifications are estimated by GMM SYS. economies between 1970 and 2011. An empirical strategy based on
theoretically motivated simultaneous equations framework, allowed us
to test the interrelationships between the three variables of interest.
Our results confirmed that economic growth, human development, and
association between education and economic growth. However, while openness are interrelated. While openness to trade can have a positive
greater openness to trade provides incentives and opportunities for impact on both economic growth and human development, we also find
more education, education seems to have no impact on trade openness. that economic growth alone does not have a positive impact on human
In sum, openness contributes to education which further helps boost development in our sample countries. However, human development
economic growth in the Asian countries – a finding consistent with can positively contribute to furthering economic growth. Thus, we find
Cooray et al. (2014). evidence of only a unidirectional positive link between human devel-
Similarly, we replace HDI with its life expectancy sub-index (LEI) opment and economic growth.
to directly test for the interaction between economic growth, health, Given the literature, reviewed above, the lack of support for a
and openness. The results in Table 6 provide evidence of bi-directional positive link between EG and HD warrants further discussion. One
association between health and economic growth. This suggests that a explanation could be that growth is not immediately helping human
healthy society contributes positively to economic progress and that in capital formation, and may be why we observe no positive effect on HD.
turn more resources need to be allocated to the health sector as This style of argument follows from both Tabellini's (2010) discussion
incomes increase in Asian economies. of the importance of imbedded cultural factors in good governance and
Kohpaiboon (2003) suggests that FDI affects economic growth from work by Acemoglu et al. (2005), who focus on the development of
through the diffusion of advanced technology into less developed institutional quality. Such deep, long-run affects may not previously
economies, and Agosin and Machado (2007) argue that FDI and have been picked up given the previous scarcity of long-run economic
openness are positively associated. Therefore, as a further robustness data and adequate econometric techniques. Although our data set does
check, we include an FDI variable as a complement to openness, which not allow us to control explicitly for institutional quality, our inclusion
also helps address concerns of omitted variable bias. In Table 7, first we of the infant mortality variable suggests a link to how good institutions
include FDI in the EG equation and then also add it into the HD (in this case adequate public health planning) may mitigate the
equation. This provides evidence that FDI has a strong positive impact negative effects of EG on human capital formation and HD. In support
on both economic growth and human development in Asian countries of such an argument, we find that the negative human development
independent of trade openness effects. The results established with effects of growth tend to disappear in our robustness checks, where we
previous specifications remain valid. use system GMM estimation, and when the HD equation is correctly
identified by the inclusion of infant mortality.
3.4.2. Single equation specifications Although infant mortality has generally declined in many Asian
In our system equation model, many explanatory variables are not countries, Mallick (2014) has shown how higher mortality rates may
strictly exogenous therefore as a final robustness check we use the reduce HD during periods of uneven economic growth. Displaced rural
Arellano and Bover (1995) and Blundell and Bond (1998) GMM SYS populations move into towns, as agricultural employment declines, but
estimator. This single equation approach allows us to control for the such migration takes place before the adequate urban socio-economic
endogeneity of explanatory variables in each individual equation by the infrastructure (schools, healthcare, etc. required in successful develop-
use of internal instruments. The approach uses both lagged level ment) can be put in place, and rural infrastructure remains under-
observations as instruments for differenced variables and lagged developed.
differenced observations as instruments for level variables, making Some insights into the subtle three-way linkages that exist between
them exogenous to fixed effects. EG and HD, and between EG and OT are also provided. While trade
The results of the single equation analysis are comparable to those openness considered in isolation is not an economic panacea, when the
from the system equation analysis, but also prompt us to interpret subtle interactions between the three variables are considered together
some of our empirical findings more cautiously. In Table 8 we report trade can contribute positively to both growth and human develop-
the signs and level of significance of the estimated coefficients for the ment. Such findings in our sample of Asian countries, confirm the view
three main variables of interest EG, HD, and OT from the two that trade liberalisation is a viable development strategy when applied
approaches (system and single equation) next to each other, for each with due consideration for local institutional depth and quality. Hence
of the three equations. In Table A1 in the Appendix the full GMM SYS future research may need to take the implications of the varying
estimation results are reported. The key message from Table 8 is that degrees of institutional quality found in Asian countries into account in
the relationships between EG, HD, and OT in the growth and openness a more structured way.

99
G. Mustafa et al. Economic Modelling 61 (2017) 93–101

Appendix A

See Table A1.

Table A1
GMM SYS single equation estimates of the base specification.

Variable EG (ΔY) Variable HD (HDI) Variable OT (OP)

(1) (2) (3)

ΔL 0.214*** IMR -0.145*** MS -0.051***

(0.083) (0.038) (0.020)
ΔH 0.051 lnH 0.050* FDI 1.471**
(0.070) (0.028) (0.759)
ΔK 0.133** ΔY 0.100 ΔY 0.010
(0.066) (0.137) (0.020)
OP 0.015 OP 0.113** HDI 1.027***
(0.012) (0.058) (0.195)
HDI 0.103***
(0.030)
Trend -0.000 Trend 0.001 Trend 0.002
(0.000) (0.001) (0.002)
AR(2) 0.080 AR(2) 0.073 AR(2) 0.120
Hansen J- 0.999 Hansen J- 0.999 Hansen J- 0.999
test test test

Notes: ***, **, and * denote statistical significant at 1%, 5%, and 10% level respectively; figures in parentheses are the standard errors. FDI is share in capital investment. Number of
observations is 492. For AR(2) and Hansen J-test p-values are reported.

Greenaway, D., Morgan, W., Wright, P., 2002. Trade liberalisation and growth in
developing countries. J. Dev. Stud. 67, 229–244.
Guttmann, S., Richards, A., 2006. Trade openness: an Australian perspective. Aust. Econ.
References Pap. 45 (3), 188–203.
Human Development Report, 2015. Work for Human Development. UNDP, New York.
IMF, 2011. World Economic Outlook. Available through ESDS International. University
Acemoglu, D., Johnson, S., Robinson, J., 2005. Institutions as the fundamental cause of of Manchester, Manchester.
long-run growth. In: Aghion, Philippe, Durlauf, Steven N. (Eds.), Handbook of Kennedy, P., 2009. A Guide to Econometrics 5th ed.. Blackwell, Oxford.
Economic Growth 1A. Elsevier BV, North Holland. Klugman, J., Rodríguez, F., Choi, H., 2011. The HDI 2010: new controversies, old
Acemoglu, D., Johnson, S., Robinson, J.A., Yared, P., 2008. Income and democracy. Am. critiques. J. Econ. Inequal. 9 (2), 249–288.
Econ. Rev. 98 (3), 808–842. Kohpaiboon, A., 2003. Foreign trade regimes and the FDI-growth nexus: a case study of
Agosin, M., Machado, R., 2007. Openness and the International allocation of foreign Thailand. J. Dev. Stud. 40 (2), 55–69.
direct investment. J. Dev. Stud. 43 (7), 1234–1247. Lucas, R.E., 1988. On the mechanics of economic development. J. Monet. Econ. 22 (1),
Anand, S., Sen, A., 1994. Human development index: methodology and measurement. 3–42.
Human Development Report, UN, New York Mallick, S.K., 2014. Disentangling the poverty effects of sectoral output, prices, and
Anand, S., Sen, A., 2000. The income component of the human development index. J. policies in India. Rev. Income Wealth 60, 773–801.
Human. Dev. 1 (1), 83–105. Mankiw, N.G., Romer, D., Weil, D.N., 1992. A contribution to the empirics of economic
Arellano, M., Bover, O., 1995. Another look at the instrumental variables estimation of growth. Q. J. Econ. 107 (2), 407–437.
error components models. J. Econ. 68, 29–51. Mason, G., O'Leary, B., Vecchi, M., 2012. Certified and uncertified skills and productivity
Barro, R.J., 2000. Rule of Law Democracy, and Economic Performance, in 2000 Index of growth performance: cross-country evidence at industry level. Labour Econ. 19,
Economic Freedom. The Heritage Foundation, Washington, D.C, 31–49. 351–360.
Barro, R.J. Lee, J., 2010. A new data set of educational attainment in the world: 1950– Neumayer, E., De Soysa, I., 2005. Trade openness, foreign direct investment and child
2010. In: Proceedings of the NBER Working Paper, No 15902. labor. World Dev. 33 (1), 43–63.
Binder, M. Georgiadis, G., 2011. Determinants of human development: capturing the Nunn, N., 2007. Relationship-specificity, incomplete contracts, and the pattern of trade.
role of institutions. In: Proceedings of the CESifo Working Paper No. 3397. Q. J. Econ. 122 (2), 569–600.
Blume, L., Voigt, S., 2007. The economic effects of human rights. Kyklos 60 (4), 509–538. Qadri, F.S., Waheed, A., 2014. Human capital and economic growth: a macroeconomic
Blundell, R., Bond, S., 1998. Initial conditions and moment restrictions in dynamic panel model for Pakistan. Econ. Model. 42, 66–76.
data models. J. Econ. 87, 115–143. Ranis, G., Stewart, F., Ramirez, A., 2000. Economic growth and human development.
Bournakis, I., 2012. Sources of TFP growth in a framework of convergence-evidence from World Dev. 28 (2), 197–219.
Greece. Int. Rev. Appl. Econ. 26 (1), 47–72. Rigobon, R., Rodrik, D., 2005. Rule of law, democracy, openness, and income. Econ.
Cameron, G., Proudman, J., Redding, S., 2005. Technological convergence, R & D, trade Transit. 13 (3), 533–564.
and productivity growth. Eur. Econ. Rev. 49 (3), 775–807. Romer, P.M., 1990. Endogenous technical change. J. Political Econ. 98, S71–S102.
Cohen, D., Soto, M., 2007. Growth and human capital: good data good results. J. Econ. Sen, A., 1999. Development as Freedom. OUP, Oxford.
Growth 12 (51), 76. Shahbaz, M., 2012. Does trade openness affect long run growth? Cointegration, causality
Conference Board, 2011. Total economy database available at 〈http://www.conference- and forecast error variance decomposition tests for Pakistan. Econ. Model. 29,
board.org/data/economydatabase/〉 2325–2339.
Cooray, A., Mallick, S., Dutta, N., 2014. Gender-specific human capital, openness and Sianesi, B., Van Reenen, J., 2003. The returns to education: macroeconomics. J. Econ.
growth: exploring the linkages for South Asia. Rev. Dev. Econ. 18, 107–122. Surv. 17 (2), 157–200.
Dagdemir, O., Acaroglu, H., 2010. The effects of globalization on child labor in Sunde, U., Vischer, T., 2015. Human capital and growth: specification matters.
developing countries. Bus. Econ. Horiz. 2 (2), 37–47. Economica 82, 368–390.
Deininger, K., Lyn, S.A., 1996. New data set measuring income inequality. World Bank Suri, T., Boozer, M.A., Ranis, G., 2011. Paths to success: the relationship between human
Econ. Rev. 10 (3), 565–591. development and economic growth. World Dev. 39 (4), 506–522.
Dreher, A., 2006. Does globalisation affect growth? Evidence from a new index of Tabellini, G., 2010. Culture and institutions: economic development in the regions of
globalisation. Appl. Econ. 38 (10), 1091–1110. Europe. J. Eur. Econ. Assoc. 8 (4), 677–716.
Falvey, R., Foster, N., Greenaway, D., 2012. Trade liberalization, economic crises, and Thirwall, A.P., 2012. The macroeconomic reasoning in TDR, pp. 95–102, in Thinking
growth. World Dev. 40 (11), 2177–2193. Development: Three Decades of the Trade and Development Report 1981–2011,
Frankel, J.A., Romer, D., 1999. Does trade cause growth? Am. Econ. Rev. 89 (3), UN, Geneva.
379–399. Tsen, H.W., 2006. Granger causality tests among openness to international trade, human

100
G. Mustafa et al. Economic Modelling 61 (2017) 93–101

capital accumulation and economic growth in China: 1952–1999. International. groups. World Econ. 27 (4), 567–585.
Econ. J. 20 (3), 285–302. Wang, Y., Yao, Y., 2003. Sources of China's economic growth 1952–1999: incorporating
UNDP, 2011. Human Development Report. UN, New York. human capital accumulation. China Econ. Rev. 14 (1), 32–52.
UNDESA, 2011. The United Nations Department of Economic and Social Affairs. UN, WIDER, 2008. World Institute for Development Economics Research. Available at
New York. 〈http://www.wider.unu.edu/research/Database/en_GB/wiid/〉
Wacziarg, R., Welch, K., 2008. Trade liberalisation and growth: new evidence. World Winters, L.A., McCulloch, N., McKay, A., 2004. Trade liberalization and poverty: the
Bank Econ. Rev. 22 (2), 187–231. evidence so far. J. Econ. Lit. 42 (1), 72–115.
Wagstaff, A., 2002. Poverty and health sector inequalities. Bull. World Health Organ. 80 World Bank, 2012. World development Indicators. Accessed at 〈http://data.worldbank.
(2), 97–105. org/data-catalog/world-development-indicators〉.
Wang, C., Liu, X., Wei, Y., 2004. Impact of openness on growth in different country

101