IIMB Management Review (2018) 30, 91–103

a v a i l a b l e a t w w w. s c i e n c e d i r e c t . c o m

ScienceDirect

j o u r n a l h o m e p a g e : w w w. e l s e v i e r. c o m / l o c a t e / i i m b Production and hosting by Elsevier

Information communication technology (ICT)

infrastructure and economic growth: A
causality evinced by cross-country panel data
Rudra P. Pradhan a,*, Girijasankar Mallik b, Tapan P. Bagchi a

a
Vinod Gupta School of Management, Indian Institute of Technology Kharagpur, West Bengal, India
b
School of Business (Parramatta Campus), University of Western Sydney, NSW 1797, Australia

Received 26 September 2014; revised 28 December 2014; accepted 10 January 2018; available online 12 February 2018

KEYWORDS Abstract This study examines certain long-run relationships hypothesised to be present among
ICT infrastructure; per capita real GDP, information and communication technology (ICT) infrastructure, consumer
Per capita economic price index, labour force participation rate, and gross fixed capital formation manifest in G-20
growth; countries recorded for the 2001–2012 period. Using panel cointegration, the study finds that the
Panel cointegration; variables are cointegrated and do not drift apart in the long run. Methodology using vector error
G-20 countries correction models (VECM) further confirms that embellishment of ICT infrastructure – an appar-
ent imperative in an economy’s information technology (IT) policy formulation – for both fixed
broadband and internet users causes a boost in the per capita GDP.
© 2018 Production and hosting by Elsevier Ltd on behalf of Indian Institute of Management
Bangalore. This is an open access article under the CC BY-NC-ND license (http://
creativecommons.org/licenses/by-nc-nd/4.0/).

Introduction a leading growth enabler in countries which have realised its

importance. Not surprisingly, therefore, many developing
Information and communication technology (ICT) infrastruc- nations are working hard to internalise ICT, balancing the
ture plays a substantial role in catalysing economic growth, limited allocation of their revenues, to catch up rapidly with
especially in today’s era of internet and mobile telecommu- the developed economies (Bankole, Osei-Bryson, & Brown,
nication (Lee & Brahmasrene, 2014; Ishida, 2015; Rohman & 2013; ASEAN, 2011; Kooshki & Ismail, 2011; Bollou, 2010;
Bohlin, 2014; Shahiduzzaman & Alam, 2014; Pradhan, Arvin, Shirazi, Gholami, & Higon, 2009; Kuppusamy, Raman, & Lee,
Norman, & Bele, 2014; WTO, 2008; Jorgenson & Stiroh, 1999). 2009; Thompson & Garbacz, 2008; Jorgenson & Vu, 2005;
Information and communication technology infrastructure is Kuppusamy & Santhapparaj, 2005; Chakraborty & Nandi,
2003). In fact, adoption of ICT-enhanced policies is one of the
top agendas for governments today in most developing
* Corresponding author. Tel.: 913222282316. countries.
E-mail address: [email protected] (R.P. Pradhan). Information and communication technology infrastruc-
Peer-review under responsibility of Indian Institute of Management ture now encompasses digital telephone network, mobile
Bangalore. phones, internet capability, internet servers and fixed
https://doi.org/10.1016/j.iimb.2018.01.001
0970-3896 © 2018 Production and hosting by Elsevier Ltd on behalf of Indian Institute of Management Bangalore. This is an open access article
under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
92 R.P. Pradhan et al.

broadband, and other technologies. In this paper we particu- The rationale that links ICT infrastructure and
larly probe the extent of the use of broadband and internet economic growth
in the hands of their users. These two innovations have re-
markably impacted communication and interaction among de-
cision makers in the conduct of business and administration In the conduct of modern day business, the importance of ICT
– the two key drivers of economic prosperity. infrastructure, otherwise called information highways, is un-
Broadband, defined as always-on internet access with deniable – supply chain management, B2C and B2B transac-
transmission speeds equalling or exceeding 256 Kbps for tions, and instantaneous transfer of funds are some examples.
downstream connections and 64 Kbps for upstream connec- One important feature of broadband infrastructure, which is
tions (Organization of Economic Cooperation and Development not present in other types of infrastructure, is the presence
(OECD), 2001), is, at present, the most common mode of of network externalities: the greater the number of users,
internet access (Lin & Wu, 2013). Both kinds of broadband the more value is derived by other users. For instance, such
adoption (256 Kbps for downstream connections and 64 characteristics do not exist in other types of public infra-
Kbps for upstream connections) by internet users have structure such as transport, drainage and sewage systems.
emerged as a vital component of the new marketplace Thus, investment returns (in terms of higher economic growth)
infrastructure. This has even led to economic restructuring are likely to be higher in broadband infrastructure than in
(Rajabiun & Middleton, 2013; Prieger, 2013; Grimes, Ren, & other types of infrastructure. Furthermore, the returns may
Stevens, 2012; OECD, 2003; Dabinett, 2002; Gillen & Lall, not accrue as a linear function of the value of infrastruc-
2002). Broadband adoption has influenced the economy ture investment. One can thus expect a positive link between
widely, particularly impacting economic growth, employ- broadband infrastructure and economic development in all
ment and national competitiveness (Ng, Lye, & Lim, 2013; countries (Jorgenson, 1991). While there are many ways broad-
Lin & Wu, 2013; Bojnec & Ferto, 2012; Czernich, Falck, band infrastructure can contribute to economic develop-
Kretschmer, & Woessmann, 2011; Bauer, 2010; Qiang, ment (Van Gaasbeck, 2008), perhaps the greatest impact of
Rossotto, & Kimura, 2009; International Telecommunication broadband infrastructure is on information diffusion and
Union (ITU), 2011; Cambini & Jiang, 2009; Holt & Jamison, organisational efficiency (Hardy, 1980).
2009; Majumdar, 2008; Van Gaasbeck, 2008; Dwivedi & Lal, Many economists have asserted that broadband infrastruc-
2007; Arbore & Ordanini, 2006; Horrigan, Stolp, & Wilson, ture affects economic growth directly and also indirectly.
2006; Firth & Mellor, 2005; Group, 2004; Roller & Waverman, Others state that the development of broadband infrastruc-
2001; Krueger, 1993). ture is a prerequisite to exploit other infrastructure devel-
Recent literature addressing the interaction of broad- opments (such as transportation, education, and remote
band (and internet) and economic growth has noted a sensing) which are necessary for economic growth
reinforcing relationship between broadband (and internet) (Koutroumpis, 2009). These instances imply that a positive
availability, and economic growth and local employment relationship between broadband infrastructure and eco-
(Bertschek, Cerquera, & Klein, 2013; Kolko, 2012; Gillett, nomic growth (often measured by total number of business
Lehr, Osorio, & Sirbu, 2006; Stenberg et al., 2009; Katz, establishments) would be but “natural.” On the contrary,
2009; Koutroumpis, 2009; Crandall, Lehr, & Litan, 2007). there are three reasons why one might not always observe a
Countries with wide broadband availability have seen higher positive relationship between the two. First, a total count of
economic growth and lower unemployment rates, even at business establishments does not capture the number of firms
the height of the recent recession (Jayakar & Park, 2013). entering and exiting the market. If increased broadband use
However, these studies only investigate the magnitude of does lead to a total sectoral shift, it is possible that no change
the impact of broadband infrastructure on economic growth would show in the number of total business establishments.
and employment, without looking at the direction of causal- Second, a related issue is that increased broadband use leads
ity. The primary objective of this paper is to overcome this to an increase in self-employment, telecommunication, and/
deficiency and to examine more broadly the causes and or easier importation of goods and services from outside of
consequences of broadband use. In particular, the paper a given country. Third, a business may draw the advantage
uses a macroeconomic growth framework to investigate the of economies of scale through the adoption of physical capital,
long-run relationship between broadband infrastructure and such as broadband. These factors would tend to reduce the
economic growth experienced in G-20 countries, using a real number of business establishments though causing an in-
panel data approach. The ceteris paribus assumption for crease in productivity (Barua, Whinston, & Yin, 2000;
the present, therefore, is whether broadband infrastruc- Thompson & Garbacz, 2008; Van Gaasbeck, 2008).
ture (and internet users) has contributed to economic growth, Currently, broadband infrastructure arises and grows on
or whether the expansion of the broadband infrastructure a global scale. This trend has heightened the debate about
(and those who use internet) itself is a consequence of the benefits of broadband internet (Ng et al., 2013). This has
economic growth. caused broadband infrastructure to be of much interest and
This paper is organised as follows: Section 2 provides the also of concern to governments and industry (Sommers &
theoretical background for the possible interaction between Carlson, 2000; Wieman, 1998). However, there has been in-
ICT infrastructure and economic growth; the third section pres- creasing consensus among experts that broadband infrastruc-
ents a review of the related literature; the data and empiri- ture should be contrasted with other types of infrastructure.
cal models are presented in Section 4; an econometric analysis Infrastructure, irrespective of type, cannot be all things to
of the scenario and the empirical results thus obtained are all communities – it is not the silver bullet. Hence, cities that
discussed in the fifth section. Section 6 concludes the paper are vying to be the next Silicon Valley need to understand this
with suitable policy implications. varied role of broadband infrastructure and know its limits.
ICT infrastructure and economic growth: A causality evinced 93

Figure 1 Impact of ICT infrastructure on economic growth. Source: Thompson and Garbacz, 2008.

It is expected that the relationship between broadband and economic growth (Bacache, Bourreau, & Gaudin, 2013;
economic growth is likely to be complex and also mutually Bertschek et al., 2013; Bojnec & Ferto, 2012; Bouras,
reinforcing. Broadband infrastructure can potentially con- Giannaka, & Tsiatsos, 2009; Brewer et al., 2005; Crandall
tribute to economic development through the lowering of et al., 2007; Czernich et al., 2011; Dwivedi, Lal, & Williams,
transactions costs (such as through faster provision of finan- 2009; Gillett et al., 2006; Holt & Jamison, 2009; Koutroumpis,
cial services), creating new opportunities for innovation, pro- 2009). Some studies use a structural model to isolate the effect
viding access to new markets (for instance through of ICT infrastructure on economic growth by controlling for
e-commerce and better exchange of information), lowering a number of macroeconomic variables such as employment,
the cost of capital (through increased efficiency in the func- trade openness, inflation, government expenditure, educa-
tioning of financial markets), closing regional discrepancies tion, urbanisation, and regulation (Bojnec & Ferto, 2012; Caroli
in incomes and productivity, providing access to human capital & Van Reenen, 2002; Czernich et al., 2011). Although these
(through tele-networking), and generating positive exter- studies investigate the correlation between ICT infrastruc-
nalities. Hence, as noted in advanced economies, a solid ture and economic growth, a causal connection has been
broadband infrastructure is a key pre-condition to enhanc- harder to prove due to the presence of simultaneous bias and
ing economic development through supporting industry and spurious correlation (Koutroumpis, 2009). More recent studies
manufacturing, marketing and sales, improving agriculture, have utilised advanced econometric tools to eliminate many
education, health, social services, and transportation, as of these problems (Czernich et al., 2011). For instance, a study
well as contributing to macroeconomic stability (Gasmi & by Cronin, Parker, Colleran, and Gold (1991) showed that ap-
Recuero Virto, 2010; Hackler, 2003; Narayana, 2011). Fig. 1 propriately lagged telecommunications penetration vari-
hypothesises the possible links between broadband infra- ables are causally linked to economic growth, and vice versa.
structure and economic growth. The presence of a correlation between telecommunications
deployment and economic growth is now, virtually, taken for
granted. However, while some studies assert significant posi-
Literature review tive effects, others find non-significant or even negative effects
(Cronin, Colleran, Herbert, & Lewitzky, 1993; Fornefeld,
Explaining the causes and consequences of ICT infrastruc- Delaunay, & Elixmann, 2008; Jayakar & Park, 2013). However,
ture in general, and broadband deployment and internet use all such studies completely miss establishing the direction of
in particular has been a central focus of recent empirical lit- causality between ICT infrastructure and economic growth
erature, particularly since Jipp’s seminal work in 1963 (Jipp, – a key requirement for policy makers. The present study,
1963). Numerous studies confirm a positive relationship therefore, particularly investigates the direction of causal-
between broadband infrastructure (and internet users) and ity between the two. It is expected that this will answer the
94 R.P. Pradhan et al.

Table 1 Summary statistics.

Panel A: For level variables
BBAND CPI GFCF INET LFPR PCGDP
Mean 11.74 109.79 22.91 282.24 68.47 19,468.22
Maximum 37.47 185.85 46.82 2751.54 81.60 45,431.03
Minimum 0.00 61.18 11.96 0.12 49.20 596.25
Std. dev. 11.21 20.81 6.12 474.12 7.86 14,837.93
Panel B: For growth variables
DlnBBAND DlnCPI DlnGFCF DlnINET DlnLFPR DlnPCGDP
Mean 32.51 4.24 0.77 25.90 0.24 2.43
Maximum 201.49 23.00 23.60 211.50 5.64 12.76
Minimum −4.83 −1.36 −19.38 −20.72 −3.26 −12.48
Std. dev. 33.82 3.38 6.12 20.23 1.03 3.50
Note: PCGDP: Per capita economic growth; CONSPI: Consumer price index; LAFPR: Labour force participation rate; GDFCF: Gross do-
mestic fixed capital formation; BBAND: Fixed broadband adoption; and INNET: Internet users.

question whether development of broadband infrastructure age (variable: PCGDP), broadband users in percentage of total
(and growth of internet users) drives economic growth or if population in numbers (variable: BBAND)1, internet users in
it is a consequence of growth. As we state below, hypotheti- percentage of total population in numbers (variable: INNET)2,
cally, one may have four possible causality nexus existing gross domestic fixed capital formation in percentage of GDP
between ICT infrastructure and economic growth. (variable: GDFCF), labour force participation rate in per-
The supply-leading hypothesis (SLH) contends that ICT in- centage of the adult population (variable: LAFPR), and Con-
frastructure is a necessary pre-condition to economic growth. sumer Price Index (variable: CONPI). These variables are
Thus, the causality runs from ICT infrastructure to eco- converted into their natural logarithms for use. Tables 1 and
nomic growth. This hypothesis maintains that ICT infrastruc- 2 present the descriptive statistics and their correlations for
ture induces economic growth by directly supporting other the variables under study.
infrastructures and factors of production, thereby improv- As the data is for individual countries ranging from 2001
ing economic growth. to 2012, it is not possible to perform a time series analysis
A second proposition is the demand-following hypothesis for an individual country. Panel data analysis would then be
(DFH), which suggests that causality runs instead from eco- most appropriate. We used the following models to de-
nomic growth to ICT infrastructure. Supporters of the demand- scribe the long-run relationships among the variables under
following hypothesis suggest that ICT infrastructure plays only study.
a minor role in economic growth and that it is a by-product
or an outcome of economic growth. The idea is that as an LnPCGDPit = α1 + β11LnICTINit + β12LnCONPIit + β13LnLAFPRit
economy grows, additional ICT infrastructure emerges in the + β14 LnGDFCFit + ε1it [1]
economy.
Third, the feedback hypothesis (FBH) suggests that eco-
nomic growth and ICT infrastructure can complement and re- LnPCGDPit = α 2 + β 21LnICTINit + β 22LnCONPIit + β 23LnLAFPRit
inforce each other, making economic growth and ICT + ε 2it [2]
infrastructure mutually causal. The argument in favour of such
bidirectional causality is that ICT infrastructure is indispens- where:
able to economic growth and economic growth inevitably re- ICTIN: ICT infrastructure (used for both BBAND and INNET
quires ICT infrastructure to be integrated into the economy. separately);
Fourth is the neutrality hypothesis (NEH), which sup- i = 1,. . ..N; and t = 1,. . .. T; and
ports the view that economic growth and ICT infrastructure ε1it and ε 2it are random error terms, normally distributed
are independent of each other. The neutrality hypothesis avers with zero mean and a finite heterogeneous variance.
that the two are independent of each other in the process The first task here was to estimate the parameters in equa-
of economic development. tions (1) and (2) using panel cointegration and conduct some
panel tests using the appropriate method of the variables in

Data and empirical model 1

Our period of analysis (2001–2012) was dictated by data availabil-
ity for ICT infrastructure, particularly for fixed-broadband
Annual data ranging from 2001 to 2012 for the selected G-20 operations.
countries were obtained from World Development Indica- 2
We are considering the variable INET for robustness check. If the
tors of The World Bank. Appendix 1 lists the selected G-20 coefficients of both BBAND and INET are found to be positive signifi-
countries included in the analysis. The multivariate frame- cant we conclude that ICT has positive and significant effect on per
work encompasses per capita economic growth in percent- capita real GDP.
ICT infrastructure and economic growth: A causality evinced 95

Table 2 Correlation matrix.

Panel A: For level variables
BBAND CPI GFCF INET LFPR PCGDP
BBAND 1.00 0.07 -0.08 0.73 0.39 0.68
CPI 0.07 1.00 0.16 0.02 −0.17 −0.24
GFCF −0.08 0.16 1.00 -0.10 0.14 -0.29
INET 0.73 0.02 −0.10 1.00 0.37 0.60
LFPR 0.39 −0.17 0.14 0.37 1.00 0.42
PCGDP 0.68 −0.24 −0.29 0.60 0.42 1.00
Panel B: For growth variables
DlnBBAND DlnCPI DlnGFCF DlnINET DlnLFPR DlnPCGDP
DlnBBAND 1.00
DlnCPI 0.20 1.00
DlnGFCF 0.28 0.12 1.00
DlnINET 0.08 −0.05 0.10 1.00
DlnLFPR 0.02 −0.05 0.13 0.01 1.00
DlnPCGDP 0.20** 0.21** 0.55* 0.26** −0.02 1.00
Note 1: PCGDP: Per capita economic growth; CONSPI: Consumer price index; LAFPR: Labour force participation rate; GDFCF: Gross do-
mestic fixed capital formation; BBAND: Fixed broadband adoption; and INNET: Internet users.
Note 2: * and ** indicate statistical level of significance at a 1% and 5% level respectively.
Note 3: The positive causality linkage between CPI and per capita GDP may be due to threshold effect and growth factors only (see for
instance, Eggoh & Khan, 2014; Khan & Senhadji, 2001; Bruno & Easterly, 1998; Fischer, 1993).

their first difference to find out the causal relationships of Wu (1999) and Choi (2001) respectively. These tests are widely
the growth effect of the variables. It was expected that co- used by many researchers. Hence, we do not repeat their
efficient of BBAND/INNET would be positive and significant, details here. The null hypothesis of the above three unit root
implying that an increase in BBAND/INNET would likely in- tests is that there exist unit root in the series, i.e. the vari-
crease per capita economic growth. Similarly, we would expect ables are non-stationary.
the coefficient of CONPI to be negative, the coefficients of Table 3 shows the results of the panel unit root tests for
LAFPR to be positive or negative, and the coefficient of GDFCF each variable. It may be seen that the level values of all series
to be positive and significant. Literature suggests that an in- (PCGDP, BBAND, INNET, CONPI, LFPR and GDFCF) are non-
crease in CONPI may reduce PCGDP. However, this may not stationary and all variables are stationary at the 5% signifi-
hold for G-20 countries because their average inflation is 4.24% cance level of the first difference, i.e., all variables are I (1).
and the standard deviation of inflation is 6.12. Hence this study It may be seen from the above section that all the four
aimed at testing the following: a) panel cointegration as per variables are integrated of order one, which certainly meets
equations (1) and (2) for a long-run relationship existing among the requirements of the cointegration test. Hence, the next
the variables under study; and b) if cointegration exists, would step is to test whether there is a long-run equilibrium rela-
Granger causality be found to exist among the variables (in- tionship between these variables. While there are number of
cluding the Error Correction Term (ECT)) taken in pairs in a tests available here such as Maddala and Wu (1999), Kao (1999)
multivariate dynamic framework? and Pedroni (1999), this study used Pedroni (1999) due to its
popularity.
Pedroni (1999) advocates two statistics, both based on a
Econometric analysis and empirical results group-mean approach. The Group PP is non-parametric and
analogous to the Phillips–Perron t-statistic while Group ADF
As stated above, the objective of this study was to test is a parametric statistics and is analogous to the ADF
whether there exists long-run relationship among PCGDP, ICT t-statistic.3 These two statistics are referred to as between-
infrastructure (i.e. either BBAND or INNET), CONPI, LAFPR and dimension statistics that average the estimated autoregressive
GDFCF. The testing procedure consists of two steps: panel coefficients for each country.
cointegration test, and Granger causality in a vector error cor- Under the alternative hypothesis of cointegration, the
rection models (VECM) framework. autoregressive coefficient is allowed to vary across coun-
In panel data analysis, particularly for cointegration and tries. This allows one to model an additional source of
Granger causality, an essential first step is to identify the sta-
tionary properties of the variables. While there are a number
of panel unit root tests, in this study we use three panel unit 3
The latter statistics is analogous to the IPS test (IPS: Im, Pesaran,
root tests (Levin–Lin–Chu: LLC, and Fisher-types: ADF and PP & Shin, 2003) for a panel unit root applied to the estimated residu-
tests) proposed by Levine, Lin, and Chu (2002), Maddala and als of a cointegration regression.
96 R.P. Pradhan et al.

Table 3 Unit root test for panel data.

Level 1st Difference
LLC ADF PP Order LLC ADF PP Order
lnPCGDP −1.14 16.60 27.19 I(1) −9.08*** 134.08*** 176.28*** I(0)
lnBBAND −0.70 50.71 107.57*** I(1) −25.55*** 213.96*** 291.87*** I(0)
lnINET 1.93 27.42 46.50 I(1) −8.46*** 68.59*** 98.53*** I(0)
lnCPI 5.48 11.43 7.78 I(1) −17.54*** 126.87*** 171.69*** I(0)
lnLFPR −2.02** 44.46 37.20 I(1) −4.36*** 134.11*** 219.89*** I(0)
lnGFCF −1.50 30.96 41.53 I(1) −14.63*** 260.17*** 293.97*** I(0)
Note 1: PCGDP: Per capita economic growth; CONSPI: Consumer price index; LAFPR: Labour force participation rate; GDFCF: Gross do-
mestic fixed capital formation; BBAND: Fixed broadband adoption; and INNET: Internet users.
Note 2: LLC stands Levin–Lin–Chu test (Levine et al., 2002), ADF stands for ADF- Fischer Chi-square test (Maddala & Wu, 1999), PP stands
PP-Fischer Chi-square test (Choi, 2001), I(1) stands for integrated of order one or non-stationary and I(0) stands for integrated of order
zero or stationary. ***, ** and * stand for significance at the 1%, 5% and 10% levels, respectively.

Table 4 Results for panel cointegration analysis.

Panel A: Pedroni cointegration test
Model without lnGFCF Full model including lnGFCF
for BBAND For internet For BBAND For internet
Group PP −48.77*** −51.05*** −29.01*** −27.75***
Group ADF −11.94*** 4.25*** −6.88*** −11.53***
Panel B: Westerlund (2007) panel cointegration test
Gt 8.39*** −2.56 2.43*** −2.26
Ga 4.41 4.74 4.69 5.36
Pt −13.67*** −34.83*** −7.64*** −10.73***
Pa −6.08*** −3.22*** −4.54*** −3.19***
Note 1: PCGDP: Per capita economic growth; CONSPI: Consumer price index; LAFPR: Labour force participation rate; GDFCF: Gross do-
mestic fixed capital formation; BBAND: Fixed broadband adoption; and INNET: Internet users.
Note 2: ***,** and *stand for significance at the 1%, 5% and 10% levels, respectively.

potential heterogeneity across the panel (countries).4 Fol- formation, labour force participation rate, and consumer price
lowing an appropriate standardisation, both of these statis- index by implementing the four-panel cointegration tests (Gα,
tics tend to the standard normal distribution as N, T → ∞ Gτ, Pα and Pτ) developed by Westerlund (2007). The results
diverging to negative infinity under the alternative hypoth- are shown in Panel B of Table 4. The query would aim to test
esis. Consequently, the left tail of the normal distribution is for the absence of cointegration by determining whether the
used to reject the null hypothesis of non-cointegration. individual panel members are error-correcting. The Ga and
The results of panel cointegration test, based on Group Gt test statistics are based on a weighted average of the in-
PP and Group ADF statistics, are shown in Table 4 (see dividually estimated short-run coefficients and their t-ratios,
Panel A). It may be seen that these two statistics are respectively. As given by Pedroni (1999), the null and alter-
significant at 1% level. So, the null hypothesis of non- native hypotheses for the panel tests are H0: αi = 0 against
cointegration may be rejected. These results strongly support HA: αi < 0 for all i.
existence of long-run equilibrium relations among eco- On the contrary, Pa and Pt test statistics pool informa-
nomic growth, ICT infrastructure, gross domestic fixed capital tion over all the cross-sectional units to test the null hypoth-
formation, labour force participation rate, and consumer esis of non-cointegration for all cross-section entities. The
price index. null and alternative hypotheses for the panel tests are H0: τi = 0
We also checked cointegration existing among economic against HA: τi < 0 for all i5.
growth, ICT infrastructure, gross domestic fixed capital These tests are very flexible and allow for an almost com-
pletely heterogeneous specification of both the long- and
short-run parts of the error-correction model, where the latter
4
Pedroni (1999) also purposes four within-dimension statistics [panel can be determined from the data. The series are allowed to
v-statistic, panel ρ-statistic, panel t-statistic (non-parametric) and be of unequal length. We can see from the results of
panel t-statistic (parametric)] that effectively pool the autoregressive
coefficients across different countries during the unit root tests. In
5
these tests, a common value for the autoregressive coefficient is speci- Westerlund also points out that Gα and Pα have higher power than
fied under the alternative hypothesis of cointegration. τα and τα in samples where T is substantially larger than N.
ICT infrastructure and economic growth: A causality evinced 97

Table 5 Estimation of the long-run coefficients (dependent variable: lnPCGDP).

Panel DOLS Panel FMOLS Panel DOLS Panel FMOLS Panel DOLS Panel FMOLS Panel DOLS Panel FMOLS
lnBBAND 0.0089*** 0.0159*** 0.0045* 0.0088**
(3.59) (4.42) (1.91) (2.60)
lnIINET 0.0232** 0.0277** 0.0161* 0.0177*
(2.09) (2.23) (1.73) (1.75)
lnCPI 0.0257 0.1457 −0.0263 0.8080*** 0.027 0.1463 0.0534 0.2141**
(0.39) (1.43) (−0.26) (3.31) (0.45) (1.63) (0.63) (2.04)
lnLFPR 0.4664** 0.5974*** 0.4969** 0.1589 −0.046 0.0566 −0.2088 −0.1600
(2.29) (2.74) (2.28) (1.24) (−0.23) (0.27) (−1.06) (−0.67)
lnGFCF 0.2238*** 0.2040*** 0.2438*** 0.2423***
(6.66) (5.63) (7.65) (6.96)
Note 1: PCGDP: Per capita economic growth; CONSPI: Consumer price index; LAFPR: Labour force participation rate; GDFCF: Gross do-
mestic fixed capital formation; BBAND: Fixed broadband adoption; and INNET: Internet users.
Note 2: Considered Dgdfcf (a dummy variable for 2008 global financial crisis) as additional deterministic regressor and linear trend and
constant as deterministic regressor. ***, ** and *stand for significance at the 1%, 5% and 10% levels, respectively.

Westerlund (2007) tests that the null hypothesis of no error- in the estimation process. In Model 2, the effect of GDFCF is
correcting relationship was rejected by all four tests. positive and significant in both the cases.
Having confirmed the existence of cointegration of our On the basis of unit root and cointegration test results cited
panel, the next step was to estimate the associated long- above, the following vector error correction models (VECMs)
run cointegration parameters. In the presence of cointegration, were set up to study short-run fluctuations and long-run
the Ordinary Least Squares (OLS) estimator is known to yield equilibrium.
biased and inconsistent results. For this reason, several other
methods have been proposed (Nasreen & Anwar, 2014). In this p1 p2

study, the long-run equilibrium relationship was estimated ΔLnPCGDPit = α10 + ∑ β1ik ΔLnPCGDPit −k + ∑ λ1ik ΔLnICTINit −k
k =1 k =1
by using panel DOLS6 (Mark & Sul, 2003) and Panel FMOLS7 p3 p4
(Pedroni, 2000, 2001), where we estimated two variants of + ∑π 1ki ΔLnCONPIit −k + ∑ δ 1ik ΔLnLAFPRit −k
a long-run equation in which per capita economic growth is k =1 k =1

“explained” by ICT infrastructure (broadband and inter- + η1i Dgdfcfit + κ 1i ECT1it −1 + ξ1it [3]
net), consumer price index, labour force participation rate,
and gross domestic fixed capital formation. The results of these q1 q2
tests are presented in Table 5. ΔLnPCGDPit = α11 + ∑ β 2ik ΔLnPCGDPit −k + ∑ λ 2ik ΔLnICTINit −k
We seek to learn about the nature of the relationship (posi- k =1 k =1
q3 q4
tive or negative) of the variables, such as BBAND, INTIN,
CONPI, LAFPR, and GDFCF. This insight could give us support
+ ∑π 2ik ΔLnCONPIit −k + ∑ δ 2ik ΔLnLAFPRit −k
k =1 k =1
to claim that these variables cause per capita economic q5

growth. There are two different ways to investigate this. First, + ∑μ 2ik ΔLnGDFCFit −k + η2i Dgdfcfit + κ 2i ECT2it −1
k =1
the impact of ICT infrastructure, CONPI and LAFPR on eco-
nomic growth would be studied. Next, the impact of ICT in- + ξ2it [4]
frastructure, CONPI, LAFPR, and GDFCF would be investigated.
where
The panel DOLS results are reported in Column one, while
Δ is first difference filter (I—L);
Panel FMOLS results is reported in Column two of Table 5. It
ICTIN is ICT infrastructure and represented for two ICT in-
may be seen that the estimated coefficient of ICT infrastruc-
dicators such as fixed broadband (BBAND) and internet users
ture (both broadband and internet) and CONPI are positive
(INNET); and
for both the models. However, the effect of LAFPR is posi-
ξ1ik and ξ2ik are the random error terms and pj (for j = 1 − 4)
tive in Model 1, while it turns to negative when we add GDFCF
and qj (for j = 1 −5) are the lag lengths determined by AIC/SBC8
criteria.
Importantly, we also observed that there exists a struc-
6 tural break in 2008 attributable to the global financial crisis
DOLS stands for dynamic ordinary least squares estimation. It uses
a parametric adjustment to the errors by augmenting the static re- (GFC). We added the dummy variable Dgdfcf9 to capture the
gression with leads, lags, and contemporaneous values of the re- effect of GDFCF on per capita economic growth.
gressor in first differences (see, for instance, Kao & Chiang, 2000).
7
FMOLS stands for fully modified ordinary least squares estima-
8
tion. It is a non-parametric approach that takes into account the pos- The study uses Akaike information criterion (AIC) and Schwarz in-
sible correlation between the error term and the first differences of formation criterion (SIC) to determine the optimum lag length. Like
the regressor as well as the presence of a constant term, to dealing the standard information criteria, a smaller SIC (or AIC) indicates a
with corrections for serial correlation (see, for instance, better fit of the model to data.
Maeso-Fernandez, Osbat, & Schnatz, 2006). 9
Dgfcit = 0 from 1990 to 2008 and 1 after 2008.
98 R.P. Pradhan et al.

Figure 2 Proposed model and hypotheses. Note 1: PCGDP: Per capita economic growth; ICTIN: ICT Infrastructure; CONSPI: Con-
sumer price index; LAFPR: Labour force participation rate; and GDFCF: Gross domestic fixed capital formation.Note 2: ICT infra-
structure involves the involvement of fixed broadband (BBAND), and internet users (INNET) separately.Note 3: H1A, B: ICT infrastructure
Granger-causes economic growth and vice versaH2A, B: Consumer price index Granger-causes economic growth and vice versaH3A, B:
Labour force participation rate Granger-causes economic growth and vice versaH4A, B: Gross domestic fixed capital formation Granger-
causes foreign direct investment and vice versa

Here ECT-1s 10 are one-period lagged error correction terms hypothesis H0: λ1ik = 0 for all i and k. The rejection of this hy-
used to detect the long-run causality between the vari- pothesis implies that ICTIN (BBAND/ INNET) is causing PCGDP
ables. We obtain the ECT-1 (the residuals) from the estima- in the short run. A similar hypothesis testing procedure was
tions based in the panel DOLS – our preferred long-run employed to test various hypotheses. The significance of the
estimators. error correction terms in each set of equations would be tested
ECT-1s here represent the long-run dynamics, while the dif- using t-tests. On the contrary, the short-run dynamics would
ferenced variables represent the short-run dynamics between be tested using F-tests.
the variables. Short-run causality is estimated here by testing Fig. 2 presents various possible hypotheses concerning the
various hypotheses. For instance, short-run causality from causal relationships among the variables.
ICTIN (BBAND/ INNET) to PCGDP is calculated by testing Table 6 reports the estimated result of vector error cor-
rection models.
We first estimated the panel-type ECM model using
ΔPCGDPit as the dependent variable. It may be seen that
10
ECT1it −1 = LnPCGDPit − LnICTINit −1 − LnCONPIit −1 − LnLAFPRit −1 and the error correction term ECMit-1 impacts it negatively and
ECT1it −1 = LnPCGDPit − LnICTINit −1 − LnCONPIit −1 − LnLAFPRit −1 is highly significant at 1% significance level. This suggests
− nGDFCFit−1 that, through error correction, a short-run disequilibrium
ICT infrastructure and economic growth: A causality evinced 99

Table 6 Panel Error correction model (dependent variable: Δ ln PCGDPit ).

Variables Model without Full model Model without Full model
Δ lnGFCFit Δ lnGFCFit
ECMi (t−1) −0.8126*** −0.8687*** 0.9094*** −1.0117***
(−10.62) (−9.52) (−10.87) (−4.70)
Δ ln PCGDPi (t−1) 0.3065*** 0.3679*** 0.2654*** 0.3727***
(5.19) (5.10) (4.18) (4.53)
Δ ln PCGDPi (t-2) 0.0945* 0.1142* 0.0903* 0.0967
(1.93) (1.90) (1.71) (1.46)
Δ ln PCGDP ⇒ Δ ln PCGDP 14.24*** 13.73*** 9.07*** 10.30***
Δ lnBBANDi (t−1) −0.0054 −0.0035
(−1.35) (−0.83)
Δ lnBBANDi (t− 2) −0.0043 −0.0052
(−1.38) (−1.61)
Δ ln BBAND ⇒ Δ ln PCGDP 4.29*** 3.55**
Δ ln INETi (t−1) −0.0181** −0.0169**
(−2.36) (−2.06)
Δ ln INETi (t− 2) −0.0201*** −0.0139*
(−2.74) (−1.78)
Δ ln INET ⇒ Δ ln PCGDP 7.57*** 4.34***
Δ lnCPIi (t−1) −0.2544*** −0.2664*** −0.1747** −0.2152**
(3.27) (−3.25) (2.10) (−2.42)
Δ lnCPIi (t− 2) 0.1173** 0.1702*** 0.0390 0.0430
(2.05) (2.84) (0.55) (0.53)
Δ ln CPI ⇒ Δ ln PCGDP 5.40*** 6.07*** 2.22 2.96*
Δ lnLFPRi (t−1) −0.4917*** −0.2155 −0.4534*** −0.0671
(−3.40) (−1.41) (−3.18) (−0.43)
Δ lnLFPRi (t− 2) 0.1558 0.3023** 0.0565 0.2838*
(1.15) (2.09) (0.41) (1.85)
Δ ln LFPR ⇒ Δ ln PCGDP 6.28*** 3.16** 5.09*** 1.81
Δ lnGFCFi (t−1) −0.0738** −0.0861***
(−2.56) (−2.75)
Δ lnGFCFi (t− 2) −0.0747*** −0.0744**
(−2.64) (−2.53)
Δ ln GFCF ⇒ Δ ln PCGDP 7.31*** 7.17***
Constant 0.03099*** 0.0290*** 0.0415*** 0.0384***
(6.37) (5.54) (6.69) (5.55)
Dgfc −0.0119*** −0.0157*** −0.0121*** −0.0150***
(−3.49) (−5.54) (−4.01) (−4.70)
Note 1: ECM, error correction term; PCGDP: Per capita economic growth; CONSPI: Consumer price index; LAFPR: Labour force partici-
pation rate; GDFCF: Gross domestic fixed capital formation; BBAND: Fixed broadband adoption; and INNET: Internet users.
Note 2: *, ** and *** indicate statistical level of significance at 1%, 5% and 10% levels respectively.
Note 3: First the residual series from the long run estimates was extracted using Eviews 8 and then we incorporated the series in the
ECM model. Δ ln BBAND ⇒ Δ ln PCGDP implies broad band growth Granger Cause per-capita regal GDP growth.
Note 4: We used fixed effect using STATA 13.

may eventually be turned into a long-run equilibrium and proximately 91% of the short-run disequilibrium is adjusted
the adjustment is stationary. The magnitude of the esti- towards the long-run equilibrium value per year. However,
mated speed of adjustment coefficient equals to −0.81 (see for Case 2, the adjustment coefficient increases to −1.01,
Column 1 of Table 6). This suggests that approximately 81% suggesting that approximately 100% of the short-run disequi-
of the short-run disequilibrium is adjusted towards the long- librium is adjusted towards the long-run equilibrium value
run equilibrium value per year. However, for the full model, per year.
the adjustment coefficient is −0.87, which suggests that In summary, the significant and error correction terms in
approximately 87% of the short-run disequilibrium is ad- equations (3) and (4) confirm the presence of long-run cau-
justed towards the long-run equilibrium value per year. sality running from ICT infrastructure (BBAND and INNET), con-
Similarly, for Model 2, the estimated speed of adjustment sumer price index, labour force participation rate and gross
coefficient is −0.91 and significant, which suggests that ap- domestic fixed capital formation to per capita economic
100 R.P. Pradhan et al.

growth. In the short-run, all coefficients are found to be sig- and internet users, consumer price index, labour force par-
nificant and indicate that ICT infrastructure (BBAND/ INNET) ticipation rate, and gross domestic fixed capital formation and
Granger cause per capita economic growth; consumer price economic growth are integrated at I (1) confirmed by panel
index Granger cause per capita economic growth; labour force unit root tests. Being cointegrated, these variables do not drift
participation rate Granger cause per capita economic growth, apart in the long run. The same inference is drawn about
and gross domestic fixed capital formation Granger cause per cointegration among economic growth, ICT infrastructure
capita economic growth. (both broadband and internet users), consumer price index,
These empirically derived results are all new. These labour force participation rate, and gross domestic fixed
results provide the evidence of Granger causal relationships11 capital formation.
between per capita economic growth, ICT infrastructure The DOLS and FMOLS estimation analyses reveal a posi-
(broadband adoption/internet users), consumer price index, tive association among ICT infrastructure (both broadband
labour force participation, and gross domestic fixed capital and internet), consumer price index and economic growth.
formation. These would be important insights for policy The effect of labour force participation in this nexus is
makers in developing countries. The results suggest that positive in model specification 1 and negative in model
ICT infrastructure (broadband adoption/internet users) along specification 2. The effect of gross domestic capital forma-
with consumer price index, labour force participation and tion, particularly in specification 2, is positive and significant
gross domestic fixed capital formation increase the level of in both the broadband and internet users’ cases. Further,
economic growth in G-20 countries. Therefore, effective the results provide evidence of Granger causality relation-
utilisation/adoption of ICT infrastructure would be an im- ship among per capita economic growth, ICT infrastructure
portant intervention necessary to reap the fruits of economic (broadband adoption/ internet users), consumer price index,
growth. labour force participation and gross domestic fixed capital
formation.
An important policy implication based on the general result
of the study is that to enhance economic growth, ICT infra-
Conclusion and policy implications structure must be upgraded and expanded, and particular at-
tention should be paid to broadband adoption and internet
This paper has empirically investigated the relationships ex- users. This is not surprising, considering how ICT catalyses the
isting among ICT infrastructure (both broadband and inter- execution of business communication and decision making
net users), economic growth, consumer price index, labour today. Another policy implication is that gross domestic capital
force participation rate, and gross domestic fixed capital for- formation (GDCF) can be considered as the control vari-
mation. Data of G-20 countries spanning the period 2001– able, particularly for labour force participation, in bringing
2012 were used. In doing so, we deployed panel unit root tests high impact on the ICT infrastructure-economic growth nexus
to examine the integrating properties of these time series vari- in the G-20 countries. Therefore, the respective govern-
ables. To examine cointegration among variables, we applied ments should stabilise the economic environment by adjust-
Pedroni panel cointegration and Westerlund panel ing GDCF in order to facilitate high economic growth. Besides,
cointegration approaches. The Granger causality test was then governments should prioritise the allocation of resources to
applied to examine the direction of causality between pairs the development of ICT infrastructure and ensure the req-
of these variables recorded in the G-20 countries for the period uisite upgradation of fixed broadband adoption and inter-
studied. net users.
The study sought to know the nature (positive or nega- A key shortcoming of this study is that it has focussed only
tive) and direction of causal relationship of the chosen on G-20 countries and four macroeconomic variables. There-
variables broadband/internet users, consumer price index, fore, it is important to note that further studies are re-
labour force participation rate, and gross domestic fixed quired with expanded country data and other factors that are
capital formation. This insight would be a significant likely to affect economic growth.
input for policy makers in developing countries looking for
appropriate interventions to effect per capita economic
growth.
The study used two specifications of analysis, first, de- Appendix 1: Selected G-20 countries
termining the impact on growth of ICT infrastructure (broad- The G-20 economic group consists of 19 member countries
band adoption/internet users), consumer price index and plus the European Union (EU), which is represented by the
labour force participation rate; second, the impact of ICT in- President of the European Council and by the European Central
frastructure (broadband adoption/internet users), con- Bank. Thus, although we look at the G-20 within this group
sumer price index, labour force participation rate, and gross of developed and developing economies, we observe only 19
domestic fixed capital formation. Empirical results indicate member countries, as used here in this study. To include the
that all four variables (ICT infrastructure – both broadband EU, the twentieth member, would have meant double-
counting France, Germany, Italy, and the UK. The countries
used are Australia, Canada, France, Germany, Italy, Japan,
Korean Republic, United Kingdom, United States of America,
11
The summary of Granger causal relationships between these vari- Argentina, Brazil, China, India, Indonesia, Mexico, Russian Fed-
ables is reported in Table A2.1 (see Appendix 2). However, the dis- eration, Saudi Arabia, South Africa, and Turkey. We have also
cussions of this summary are not elaborated here due to restrictions included Spain in the list because Spain is participating as a
of space. “permanent guest”.
ICT infrastructure and economic growth: A causality evinced 101

Appendix 2: Summary of panel Granger causality tests

Table A2.1 Results and summary of panel Granger causality tests.

Dependent Independent variables
Variable (possible sources of causation)
Case 1: Models Without GDCF
Model 1: PCGDP, BBBAND, CPI, INET, LFPR
ΔPCGDP ΔBBAND ΔCPI ΔLFPR
ΔPCGDP – 3.45** 23.2* 6.46*
[–] [0.05] [0.00] [0.00]
ΔBBAND 3.79** – 38.6* 2.69***
[0.05] [–] [0.00] [0.10]
ΔCPI 6.64* 6.69* – 3.83**
[0.00] [0.00] [–] [0.05]
ΔLFPR 6.55* 2.89*** 2.59*** –
[0.00] [0.10] [0.10] [–]
Summary of findings: BBAND < =>GDP; CPI < =>GDP; LFPR < =>GDP; BBAND < =>CPI; LFPR < =>BBAND; LFPR < =>CPI
Model 2: PCGDP, INET, CPI, INET, LFPR
ΔPCGDP ΔINET ΔCPI ΔLFPR
ΔPCGDP – 4.09** 18.2* 5.98*
[–] [0.05] [0.00] [0.01]
ΔINET 9.23* – 5.65* 3.11***
[0.00] [–] [0.01] [0.10]
ΔCPI 6.02* 4.41** – 3.39**
[0.01] [0.05] [–] [0.05]
ΔLFPR 13.0* 6.99* 8.95* –
[0.00] [0.00] [0.00] [–]
Summary of findings: INET < =>GDP; CPI < =>GDP; LFPR < =>GDP; NET < =>CPI; LFPR < =>INET; LFPR < =>CPI
Case 2: Models With GFCF
Model 1: PCGDP, BBBAND, CPI, LFPR, GFCF
ΔPCGDP ΔBBAND ΔCPI ΔLFPR GFCF
ΔPCGDP – 4.02** 21.5* 5.82* 2.27***
[–] [0.05] [0.00] [0.01] [0.10]
ΔBBAND 7.79* – 36.3* 3.78** 7.87*
[0.00] [–] [0.00] [0.05] [0.01]
ΔCPI 9.27* 6.63* – 7.08* 17.4*
[0.00] [0.01] [–] [0.01] [0.00]
ΔLFPR 4.96** 2.65*** 3.04*** – 2.63***
[0.05] [0.10] [0.05] [–] [0.10]
ΔGFCF 8.56* 7.10* 4.84** 4.79** –
[0.00] [0.01] [0.05] [0.05] [–]
Summary of findings: BBAND < =>GDP; CPI < =>GDP; LFPR < =>GDP; GFCF < =>GDP; BBAND < =>CPI; LFPR < =>BBAND; GFCF < =>BBAND;
LFPR < =>CPI; GFCF < =>CPI; LFPR < =>GFCF
Model 2: PCGDP, INET, CPI, LFPR, GFCF
ΔPCGDP ΔINET ΔCPI ΔLFPR GFCF
ΔPCGDP – 3.92** 16.5* 5.21* 0.78
[–] [0.05] [0.00] [0.01] [0.35]
ΔINET 3.50*** – 5.98* 3.67** 3.04***
[0.10] [–] [0.01] [0.05] [0.10]
ΔCPI 8.99* 5.06* – 5.82* 16.4*
[0.01] [0.01] [–] [0.01] [0.00]
ΔLFPR 7.72* 7.45* 9.43* – 4.32**
[0.01] [0.01] [0.00] [–] [0.05]
ΔGFCF 5.72* 2.90*** 3.89** 5.71* –
[0.01] [0.10] [0.05] [0.01] [–]
Summary of findings: INET < =>GDP; CPI < =>GDP; LFPR < =>GDP; GFCF = > GDP; INET < =>CPI; LFPR < =>INET; GFCF < =>INET; LFPR < =>CPI;
CPI < =>GFCF; LFPR< = >GFCF
Note 1: PCGDP: Per capita economic growth; BBAND: Fixed broadband adoption; and INNET: Internet users; CPI: Consumer price index; LAFPR:
Labour force participation rate; and GFCF: Gross domestic fixed capital formation.
Note 2: * and **denote rejection of the null hypothesis at the 0.01 and 0.05 levels, respectively.
Note 3: Values in parentheses represent t-statistics.
Note 4: < =>INDICATES THE DIRECTION OF Granger causality.
Note 5: ***stands significance at the 10% level.
102 R.P. Pradhan et al.

References Dwivedi, Y. K., & Lal, B. (2007). Socio-economic determinants of

broadband adoption. Industrial Management and Data Systems,
107(5), 654–671.
Arbore, A., & Ordanini, A. (2006). Broadband divide among SMEs: The Dwivedi, Y. K., Lal, B., & Williams, M. D. (2009). Managing con-
role of size, location and outsourcing strategies. International sumer adoption of broadband: Examining drivers and barriers.
Small Business Journal., 24(1), 83–99. Industrial Management and Data Systems, 109(3), 357–
ASEAN (2011). ASEAN ICT Master Plan 2015. The ASEAN Secretariat, 369.
Jakarta. Eggoh, J. C., & Khan, M. (2014). On the non-linear relationship
Bacache, M., Bourreau, M., & Gaudin, G. (2013). Dynamic entry and between inflation and economic growth. Research in Econom-
investment in new infrastructures: Empirical evidence from the ics, 68(2), 133–143.
fixed broadband industry. Review of Industrial Organization, 43, Firth, L., & Mellor, D. (2005). Broadband: Benefits and problems. Tele-
1–31. communications Policy, 29, 223–236.
Bankole, F. O., Osei-Bryson, K., & Brown, I. (2013). The impact of Fischer, S. (1993). The Role of Macroeconomic Factors in Growth.
information and communication technology infrastructure and NBER Working Paper Series. National Bureau of Economic Re-
complementary factors on Intra-African trade. Information Tech- search, Cambridge.
nology for Development, 1–17. In press. Fornefeld, M., Delaunay, G., & Elixmann, D. (2008). The Impact of
Barua, A., Whinston, A. B., & Yin, F. (2000). Value and productivity Broadband on Growth and Productivity. White Paper, European
in the Internet economy. IEEE Computer, 33(5), 102–105. Commission, DG Information Society and Media.
Bauer, J. M. (2010). Regulation, public policy, and investment in com- Gasmi, F., & Recuero Virto, L. (2010). The determinants and impact
munications infrastructure. Telecommunications Policy, 34, 65– of telecommunications reforms in developing countries. Journal
79. of Development Economics, 93(2), 275–286.
Bertschek, I., Cerquera, D., & Klein, G. J. (2013). More bits- more Gasmi, F., & Recuero Virto, V. L. (2010). The determinants and impact
bucks? Measuring the impact of broadband internet on firm of telecommunications reforms in developing countries. Journal
performance. Information Economics and Policy, 25, 190– of Development Economics, 93(2), 275–286.
203. Gillen, D., & Lall, A. (2002). The economies of the Internet, the new
Bojnec, S., & Ferto, I. (2012). Broadband availability and economic economy and opportunities for airports. Journal of Air Trans-
growth. Industrial Management and Data Systems, 112(9), 1292– port Management, 8, 49–62.
1306. Gillett, S. E., Lehr, W. H., Osorio, C. A., & Sirbu, M. A. (2006). Mea-
Bollou, F. (2010). Interrogating the impact of ICT infrastructure ex- suring The economic impact of broadband deployment. Final
pansion in francophone West Africa 1993–2005. A critical theory report, national technical assistance, Training, Research and Evalu-
study using Archival data and non parametric research methods ation Project No. 99-07-13829.
(Doctoral Thesis). University of Cape Town, Cape Town. Grimes, A., Ren, C., & Stevens, P. (2012). The need for speed: Impacts
Bouras, C., Giannaka, E., & Tsiatsos, T. (2009). Identifying best of Internet connectivity on firm productivity. Journal of Produc-
practices for supporting broadband growth: Methodology and analy- tion Analysis, 37, 187–201.
sis. Journal of Network and Computer Applications, 32(2), 795– Group, B. S. (2004). Impact of Broadband- enabled ICT, Content, Ap-
807. plications and Services on the UK Economy and Society to 2010.
Brewer, E., Demmer, M., Du, B., Ho, M., Kam, M., Nedevschi, S., et al. Briefing Paper.
(2005). The case of technology in developing regions. Computer Hackler, D. (2003). Invisible infrastructure and the city. American
Society, 38(6), 25–38. Behavioral Scientist, 46(8), 1034–1055.
Bruno, M., & Easterly, W. (1998). Inflation crises and long-run growth. Hardy, A. P. (1980). The role of the telephone in economic devel-
Journal of Monetary Economics, 41(1), 3–26. opment. Telecommunications Policy, 4(4), 278–286.
Cambini, C., & Jiang, Y. (2009). Broadband investment and regula- Holt, L., & Jamison, M. (2009). Broadband and contributions to eco-
tion: A literature review. Telecommunications Policy, 33, 559– nomic growth: Lessons from the US experience. Telecommuni-
574. cations Policy, 33, 575–581.
Caroli, E., & Van Reenen, J. (2002). Skill biased organizational change? Horrigan, J. B., Stolp, C., & Wilson, R. H. (2006). Broadband utili-
Evidence from British and French Establishments. The Quar- zation in space: Effects of population and economic structure.
terly Journal of Economics, 116(4), 1449–1492. Information society, 22, 341–354.
Chakraborty, C., & Nandi, B. (2003). Privatization, telecommunica- Im, K. S., Pesaran, M. H., & Shin, Y. (2003). Testing for unit roots
tions and growth in selected Asian countries: An econometric analy- in heterogeneous panels. Journal of Econometrics, 115(1), 53–
sis. Communications and Strategies, 52(3), 31–47. 74.
Choi, I. (2001). Unit root tests for panel data. Journal of Interna- Ishida, H. (2015). The Effect of ICT Development on Economic Growth
tional Money and Finance, 20, 249–272. and Energy Consumption in Japan. Telecommunications Policy,
Crandall, R., Lehr, W., & Litan, R. (2007). The effects of broad- 32, 79–88.
band deployment on output and employment: A cross-sectional International Telecommunication Union (ITU). (2011). Broadband: A
analysis of US data. Issues in Economic Policy, 6,, 1–34. platform of progress. Geneva: ITU.
Cronin, F. J., Parker, E. B., Colleran, E. K., & Gold, M. A. (1991). Jayakar, K., & Park, E. (2013). Broadband availability and employ-
Telecommunication infrastructure and economic growth: An ment: An analysis of country- level data from the national
analysis of causality. Telecommunication Policy, 15(6), 529– broadband map. Journal of Information Policy, 3, 181–
535. 200.
Cronin, F. J., Colleran, E. K., Herbert, P. L., & Lewitzky, S. (1993). Jipp, A. (1963). Wealth of nations and telephone density. Telecom-
Telecommunication infrastructure investment and economic de- munications Journal, 30(1), 199–201.
velopment. Telecommunication Policy, 17(6), 415–430. Jorgenson, D. W. (1991). Fragile statistical foundations: The mac-
Czernich, N., Falck, O., Kretschmer, T., & Woessmann, L. (2011). roeconomics of public infrastructure investment. Paper pre-
Broadband infrastructure and economic growth. The Economic sented at the 1991 American Enterprise Institute Conference:
Journal, 121(552), 505–532. “Infrastructure Needs and Policy Options for the 1990s”. Wash-
Dabinett, G. (2002). Reflections on regional development policies in ington, DC.
the information society. Planning Theory and Practice, 3(2), 232– Jorgenson, D. W., & Stiroh, K. J. (1999). Information technology and
237. growth. The American Economic Review, 89(2), 109–115.
ICT infrastructure and economic growth: A causality evinced 103

Jorgenson, D. W., & Vu, K. (2005). Information technology and the method of moments approach. Applied Economics Letters, 20,
world economy. Scandinavan Journal of Economics, 107(4), 631– 857–862.
650. Organization of Economic Cooperation and Development (OECD).
Kao, C. (1999). Spurious regression and residual based tests for (2001). The development of broadband access in OECD coun-
cointegration in panel data. Econometrics, 90, 1–44. tries. Paris: Head of Publications Service, OECD.
Kao, C., & Chiang, M. H. (2000). On the estimation and inference of OECD. (2003). Information and Communications Technologies (ICT)
a cointegrated regression in panel data. In B. H. Baltagi (Ed.), Ad- and economic growth: Evidence from OECD countries, indus-
vances in econometrics: Non-stationary panels, panel cointegration tries, and firms. Paris: Head of Publications Service, OECD.
and dynamic panels (Vol. 15, pp. 179–222). Pedroni, P. (1999). Critical values for cointegration tests in hetero-
Katz, R. L. (2009). Estimating Broadband Demand and Its Economic geneous panels with multiple regressors. Oxford Bulletin of Eco-
Impact in Latin America. Paper presented at 3rd ACORN-REDECOM nomics and Statistics, 61, 653–670.
Conference, Mexico, May 22–23. Pedroni, P. (2000). Fully modified OLS for heterogeneous cointegrated
Khan, M. S., & Senhadji, A. S. (2001). Threshold effects in the rela- panels. Advances in Econometrics, 15, 93–130.
tionship between inflation and growth. IMF Staff Papers, 48(1), Pedroni, P. (2001). Purchasing power parity tests in cointegrated
1–21. panels. Review of Economics and Statistics, 83(4), 727–731.
Kolko, J. (2012). Broadband and local growth. Journal of Urban Eco- Pradhan, R. P., Arvin, M. B., Norman, N. R., & Bele, S. K. (2014).
nomics, 71, 100–113. Economic Growth and the Development of Telecommunications
Kooshki, M. F., & Ismail, R. (2011). The impact of information and Infrastructure in the G-20 Countries: A Panel-VAR Approach. Tele-
communication technology investment on economic growth in communications Policy, 38(7), 634–649.
newly industrialized countries in Asia. Australian Journal of Basic Prieger, J. E. (2013). The broadband digital divide and the eco-
and Applied Sciences, 5(9), 508–516. nomic benefits of mobile broadband for rural areas. Telecom-
Koutroumpis, P. (2009). Broadband infrastructure and economic munications Policy, 37, 483–502.
growth: A simultaneous approach. Telecommunications Policy, Qiang, C. Z. W., Rossotto, C. M., & Kimura, K. (2009). Economic
33(9), 471–485. impacts of broadband. In World Bank (Ed.), Information and Com-
Krueger, A. B. (1993). How computers have changed the wage struc- munications for Development 2009: Extending reach and increas-
ture: Evidence from micro data 1984- 1989. The Quarterly Journal ing impact. Washington, DC: World Bank.
of Economics, 108, 33–60. Rajabiun, R., & Middleton, C. A. (2013). Multilevel Governance and
Kuppusamy, M., & Santhapparaj, S. (2005). Investment in informa- Broadband Infrastructure Development: Evidence from Canada.
tion and communication technologies and its payoff in Malaysia. Telecommunications Policy, 37(9), 702–714.
Perspectives on Global Development and Technology, 4(2), 147– Rohman, I. K., & Bohlin, E. (2014). Decomposition analysis of the tele-
168. communications sector in Indonesia: What does the cellular era
Kuppusamy, M., Raman, M., & Lee, G. (2009). Whose ICT invest- shed light on? Telecommunications Policy, 38(3), 248–263.
ment matters to economic growth: Private or public? The Malay- Roller, L., & Waverman, L. (2001). Telecommunications infrastruc-
sian perspective. The Electronic Journal on Information Systems ture and economic development: A simultaneous approach. The
in Developing Countries, 37(7), 1–19. American Economic Review, 91(4), 909–923.
Lee, J. W., & Brahmasrene, T. (2014). ICT, CO2 emissions and eco- Shahiduzzaman, M., & Alam, K. (2014). The Long-run Impact of In-
nomic growth: Evidence from a panel of ASEAN. Global Eco- formation and Communication Technology on Economic Output:
nomic Review: Perspectives on East Asian Economies and The Case of Australia. Telecommunications Policy, 38(7), 623–
Industries, 43(2), 93–109. 633.
Levine, A., Lin, C. F., & Chu, C. S. (2002). Unit root tests in panel Shirazi, F., Gholami, R., & Higon, D. A. (2009). The Impact of Infor-
data: Asymptotic and finite sample properties. Journal of Econo- mation and Communication Technology (ICT), education and regu-
metrics, 108, 1–24. lation on economic freedom in Islamic Middle Eastern countries.
Lin, M., & Wu, F. (2013). Identifying the determinants of broad- Information and Management, 46(9), 426–433.
band adoption by diffusion stage in OECD countries. Telecom- Sommers, P., & Carlson, D. (2000). Ten steps to a high tech future:
munications Policy, 37, 241–251. the new economy in Metropolitan Seattle. The Brookings Insti-
Maddala, G. S., & Wu, S. (1999). A comparative study of unit roots tute. Washington DC. Retrieved from https://www.brookings.edu/
with panel data a new simple test. Oxford Bulletin of Econom- wp-content/uploads/2016/06/SommersReport.pdf. (Accessed 2
ics and Statistics, 61, 631–651. March 2018).
Maeso-Fernandez, F., Osbat, C., & Schnatz, B. (2006). Towards the Stenberg, P., Morehart, M., Vogel, S., Breneman, V., Cromartie, J.,
estimation of equilibrium exchange rates for CEE acceding & Brown, D. (2009). Broadband Internet’s Value for Rural America,
countries: Methodological issues and a panel cointegration ERR No.78, Economic Research Service. U.S. Department of Ag-
perspective. Journal of Computational Economics, 34(3), 499– riculture.
517. Thompson, H., & Garbacz, C. (2008). Broadband impacts on state GDP:
Majumdar, S. K. (2008). Broadband adoption, jobs and wages in the direct and indirect impacts. Paper presented at the Interna-
US telecommunications industry. Telecommunications Policy, 32, tional Telecommunications Society 17th Biennial Conference,
587–599. Canada.
Mark, C. N., & Sul, D. (2003). Cointegration vector estimation by panel Van Gaasbeck, K. A. (2008). A rising tide: Measuring the economic
DOLS and long-run money demand. Oxford Bulletin of Econom- effects of broadband use across California. The Social Science
ics and Statistics, 65(5), 655–680. Journal, 45, 691–699.
Narayana, M. R. (2011). Telecommunication services and economic Westerlund, J. (2007). Testing for error correction in panel data.
growth: Evidence from India. Telecommunications Policy, 35(2), Oxford Bulletin of Economics and Statistics, 69(6), 709–748.
115–127. Wieman, C. (1998). The high-tech transition: Technology and the pros-
Nasreen, S., & Anwar, S. (2014). Causal relationship between trade pects for improving infrastructure performance. Journal of Urban
openness, economic growth and energy consumption: A panel data Technology, 5(2), 21–46.
analysis of Asian countries. Energy Policy, 69, 82–91. World Trade Organization. (2008). World trade report 2008 – Trade
Ng, T. H., Lye, C. T., & Lim, Y. S. (2013). Broadband penetration in a globalizing world. Washington, DC: World Trade
and economic growth in ASEAN countries: A generalized Organization.