









































Pa
ge

 
1



Pa
ge

 
13

1

American Journal of  Economics and 
Business Innovation (AJEBI)

Dynamic and Static Effects of  Digital Services Trading on Economic 
Growth of  World Regions

Richard Mulenga1*

Volume 2 Issue 3, Year 2023
ISSN: 2831-5588 (Online), 2832-4862 (Print)

DOI: https://doi.org/10.54536/ajebi.v2i3.2051
https://journals.e-palli.com/home/index.php/ajebi

Article Information ABSTRACT

Received: September 25, 2023

Accepted: October 22, 2023

Published: October 27, 2023

The study examines the effect of  digital services trading on the economic growth of  panel 
data from five (5) world regions, namely, Africa, Asia, Latin America, Europe, and the 
Organization for Economic Cooperation and Development (OECD) between 2005 and 
2021. The study employs quantitative experimental methods of  Panel vector autoregression 
(PVAR) models as dynamic estimators and the Fixed Effects (FE) models as static estimators. 
The PVAR model analyses indicate that the digital services trade has a significant positive 
long-run effect on Gross Domestic Product (GDP) growth in all five regions. For every 1% 
increase in digital services trade, GDP in the OECD region increases by 6.75%, followed by 
Europe at 3.19%, 2.11% in Latin America, 1.02% in Asia, and is lowest in the African region 
panel at 0.82%. The static FE analyses also shows similar results. These findings confirm the 
hypothesis that deep internet/digital penetration positively impacts the efficiency of  digital 
services trade. The study, therefore, recommends that policymakers from developing world 
regions should increase investments in digital deepening infrastructure which should include 
among others, promoting, policy and regulatory measures that augment digital infrastructure 
installations in rural places, developing the digital skills to mitigate digital illiteracy, and 
promoting the adoption of  cutting-edge digital technologies. 

Keywords
Dynamic and Static Effects, 
Digital Services Trading, World 
Regions, Economic Growth

1 Faculty of  Economics, ZCAS University, Lusaka, Zambia
* Corresponding author’s e-mail: richardmulenga2@gmail.com

INTRODUCTION
Since the development of  the steam engine and the 
telegraph in the latter half  of  the 19th century, the 
evolution of  container shipping around the 1960s, and 
the developments in transportation and information 
communication technologies (ICT) in the twenty-first 
century have fundamentally altered international trade and 
the basket of  goods and services that trading countries 
specialize in (Cosar and Demir, 2018; Steinwender, 2018; 
Zhu, Shang, & Li, 2023).
The global digital economy, a core element of  the 4th 
Industrial Revolution, ushered in digital trade (United 
Nations Conference on Trade and Development, 
UNCTAD, 2020). According to UNCTAD (2020), digital 
trade refers to all international transactions that are 
delivered remotely in electronic format, using computer 
networks. It involves all trade that is digitally ordered 
and/or digitally delivered.
From the demand side point of  view, the digitization 
processes have a positive effect as observed by the increase 
in the variety and volume of  production, exchange as well 
as consumption. Business to Consumer (B2C), Business 
to Business (B2B), Consumer to Consumer (C2C), and 
Business entities and Governments (B2G) can all engage 
in digital commerce transactions (OECD, 2020). It is 
simple to note that trade in products and services has 
been significantly disrupted and is increasingly moving 
from physical to digital forms in certain areas, including 
but not limited to, among others, education, financial 
services, entertainment, software, logistics, as well 
as health service provision (Tele-doctor services, for 
instance). Some goods that used to be traded in physical 

form such as movies, music, and books have been 
digitally transformed and can now be digitally transmitted 
as video streaming, music downloading, and e-books 
respectively (Organization for Economic Cooperation 
and Development, OECD, 2020; UNCTAD, 2022).
The Organization for Economic Cooperation and 
Development, OECD, (2020) explains that from the 
supply side, digitization procedures have had a tremendous 
impact on how goods and services are produced and 
delivered. Cloud computing, 3D printing, Artificial 
Intelligence (AI) and the App industry (App developers 
and vendors) are integrating digital technologies in the 
production and delivery of  goods and services making 
the flow of  data the ‘life blood’ of  the 21st-century 
international trade. As opposed to traditional trade, digital 
technology-driven trade makes it simple to coordinate 
international supply chains, making it relatively quicker, 
cheaper, and increases trade volumes, which subsequently 
results in higher economies of  scale, reduced trade time, 
and significantly cuts variable costs via lowered entry 
barriers (OECD, 2020; Mulenga and Mayondi, 2022). 
The digital economy in general, and digital trading in  
particular, enable Small and Medium Enterprises (SMEs) 
and emerging economies to easily engage in international 
commerce transactions, which were traditionally 
controlled by huge multinational enterprises (UNCTAD, 
2022). The COVID-19 Crisis seems to have elevated 
the importance of  digital trade in that there was a 
significant increase in the use of, and development of  
online platforms for purchasing products and services. 
UNCTAD (2022) revealed that despite the worldwide 
economic crisis that emanated from COVID-19 in 



Pa
ge

 
13

2

https://journals.e-palli.com/home/index.php/ajebi

Am. J. Econ. Bus. Innov. 2(3) 131-144, 2023

2019, global digital services were remarkably resilient, 
declining by only 1.8% relative to global traditional 
services exports, which decreased by 20%. 
Although empirical and theoretic literature is replete with 
assertions of  the positive impact/effect of  digitization 
processes on various economic activities, a literature 
review of  related literature (for example, Zhang,et al., 2023; 
Mulenga & Mayondi ,2022; Ying & Gao 2022; Pingfang, 
2021, Wang & Choi,2019; Thomas-Mini, 2018) seems 
to suggest that research gaps exist regarding empirical 
studies on the digital services trading-economic growth 
nexus of  world regions. Therefore, this topic remains an 
open question. The scarcity of  research in this area can be 
partially explained by the apparent absence of  a universal 
definition of  digital trade and the scarcity of  digital 
trade measures in official national accounting systems. 
To this end, therefore, this study conducts the dynamic 
and static analyses of  the effects of  digital services trade 
on economic growth (GDP per capita) in five (5) world 
regions from Africa, Asia, Europe, the Latin Americas, 
and the Organization for Economic Cooperation and 
Development (OECD). The study is predicated on the 
hypothesis that regions that have deeper internet/digital 
penetration experience higher economic growth driven 
by digital services trade.

LITERATURE REVIEW
Conceptual and Theoretical Background
Whereas the digital economy refers to a plethora of  terms 
ranging from digital computing including e-commerce 
and gig economy, digital trade basically focuses on how 

the internet is revolutionizing domestic and international 
trade and reshaping comparative advantage (World 
Economic Forum, WEF, 2022, OECD, 2020). In simple 
terms, a digital economy is any economic activity done 
or facilitated by digital technologies. There seems to 
be a positive correlation between the growth of  digital 
technologies and digital trade. For instance, digital 
technologies help small and medium enterprises (SMEs) 
overcome the challenges of  scale and distance and thus 
make it relatively easier to access global markets previously 
dominated by large multinational corporations (MNCs).
Although digital trading has positive economic benefits, 
barriers exist. Some of  the barriers are tariffs and quotas 
imposed on imports of  routers and servers, cross-border data 
flow restrictions, discriminatory national and international 
laws/standards that depart from recognized international 
standards (WEF, 2020, Department for International 
Development DFID, 2020, UNCTAD, 2022).

Situational Analysis: Internet Penetration
Precondition for Digital Trade
The potential benefits that accrue to a country or region 
are determined largely by the digital infrastructure, and 
regulatory framework (digital/internet penetration, 
hereafter) tenable in that region or country. It is 
hypothesized that deep digital/internet penetration 
augments the long-run economic growth through 
increased economies of  scale and the reduction of  
variable costs. As can easily be seen in Figure 1, there are 
great variations among world regions in terms of  digital 
penetration.

Figure 1: Internet Penetration in World Regions-2009-2022
Source: Author’s elaboration on data from the International Communication Union. Note that CIS denotes Commonwealth independent 
States from the former Union of  Soviet Socialist Republic (USSR).

From the scatter plot in Figure 1, it is easy to note that 
digital penetration or digital connectivity, defined as 
the ability to access and use technologies to conduct 
economic activities including digital services trade (World 
Bank 2016; Brookings 2017) has been growing across the 
world regions from 2009.  However, it appears that the 
African Region (light blue dotted scatter plots) has been 
lagging behind other regions since 2009 even though 
the percentage of  the population’s access to digital 
technologies in Africa has grown from 7.6% in 2009 to 

40% in 2022. The European region (black scatter plots) 
ranking first in terms of  internet penetration, grew from 
59.6 % to 89 % in the same period. The Americas rank 
2nd; it ranges between 46.3% in 2009 and 83% in 2022.
 
Conceptual Framework of  Digital Trade
Data and digital platforms are two essential ingredients 
of  the digital economy in general and digital trade in 
particular. Conceptually, the fundamental driver of  digital 
trade is data which can be traded and is a means through 



Pa
ge

 
13

3

https://journals.e-palli.com/home/index.php/ajebi

Am. J. Econ. Bus. Innov. 2(3) 131-144, 2023

which services are traded (World Economic Forum, 
WEF, 2020, González & Jouanjean,2017). Figure 2 shows 
the conceptual framework.
The digital trade enablers refer to the hard and software 
infrastructure. Data flows support trade by enabling 
control and coordination in the international trade 
networks or by facilitating a wide range of  trading 
activities. Digital trade is done via physically delivered and 
digitally delivered platforms (OECD, 2020). 
In our analyses, Gross Domestic Product per capita 
(GDP per capita) is the dependent variable (also referred 
to as economic growth) whereas digital exports and 
digital imports (collectively called digital services trading), 
goods exports and goods imports (as control variables), 
and the number of  people using the internet (% of  the 
population) and Secure Internet Servers are independent 
variables (Regressors). The level of  digital/internet 
penetration is considered an intervening variable.

Review of  Related Literature
For the years 2013 through 2020, Zhang, Ye, and Sun 
(2023) evaluate the Static and Dynamic Efficiency of  the 
Chinese Digital Economy. They make use of  models from 
the Three Stage Malmquist Index-Based Approach. Their 
findings show that the effectiveness of  the digital economy 
differs by province. Brazil, Russia, India, China, and South 
Africa (BRICS) countries’ economic growth was examined 
by Wang and Choi (2019) in relation to the effects of  the 
digital economy for the years 2000–2016. They used the 
Gravity, Fixed Effects (FE), and Random Effects (RE) 
models in their investigation. Findings demonstrate that 
the digital economy has significantly improved. They 
recommend that the BRICS countries should invest 
massively in digital trade infrastructure to reap maximum 
benefits from the effects of  the digital economy.
Mulenga and Mayondi (2022) evaluated how the trade in 
digital services impacted panels of  developing, emerging, 
and developed nations’ economic growth for the period 
2005 to 2021 using the panel vector autoregression models 
as dynamic estimators. Results show that even though all 
the country panels’ GDP growth was positively impacted 
by digital services exports, the developing countries’ 
panel lagged behind. Thus, the study recommends that 
developing countries should enhance their investments in 
digital infrastructure. Using fixed effect (FE) models as 
static estimators for the years 2013 to 2019, Zhu, Shang, 
and Li (2023) undertake research on the effects of  the 

Digital Economy on global trade. The study finds that the 
digital economy significantly promotes the development 
of  urban international trade. The study recommends that 
policymakers should be proactive in increasing investment 
in digital infrastructure.
In their examination of  the strategies for developing 
digital trade and methods of  governance in United 
States, the European Union, China, and India, Ying, 
and Gao (2023), use comparative taxonomy models. 
The study findings reveal that the four economies in 
the survey employ strategic policies or mechanisms in 
global digital commercial activities that enable them to 
substantially derive large amounts of  rent. Mini-Thomas 
(2018) examined how the trade in services impacted India’s 
economic growth using both the Balance of  Payments 
Constrained Growth (BPCG) and the Autoregressive 
Distributed Lag Co-integration (ARDL) models. The 
results show that the growth of  India’s economy was driven 
largely by the goods trade than the goods services trade.
Simon and Pingfang (2021) use cross-sectional data from 
53 nations for the years 2000–2018 to evaluate the effects 
of  the digital economy on economic growth in Africa. 
They use Fixed Effects (FE) for static models and the 
System General Method of  Moments (Sys-GMM) for 
dynamic models. The results demonstrate that the digital 
economy significantly enhances commerce and economic 
development in Africa. Using the vector auto-regression 
(VAR) method, Maune (2019) examined how trade in 
services impacted the economic growth of  a panel of  
ten (10) Southern African nations. In contrast to trade 
in commodities exports, the study revealed that trade in 
services had a higher positive coefficient than goods trade. 
To ascertain the effect of  the digital economy (digital 
index) on provincial economic growth in China, Pan et 
al (2022) employed pooled ordinary regression models 
(POL). They discovered that although the digital economy 
had a significant positive effect on China’s economy, 
regional differences in the digital infrastructure penetration 
caused variations in economic growth. Therefore, to 
reduce the economic disparities of  the effects of  the digital 
economy, the study recommends the integration of  the 
digital economy in all the regions in China.

Theoretical Framework
The augmented Solow growth model might theoretically 
be conceived as increasing total factor productivity (TFP) 
arising from the digital economy’s growth (Pan et al, 2022; 

Figure 2: Conceptual Framework
Source: Author’s elaboration



Pa
ge

 
13

4

https://journals.e-palli.com/home/index.php/ajebi

Am. J. Econ. Bus. Innov. 2(3) 131-144, 2023

Zhang et al, 2021; Mulenga & Mayondi,2022). Solow’s 
augmented growth model as postulated in Mankiw, 
Romer, and Weil 1992):
Yit=Kit

α Hit
β (AL)(1-α-β)                                                    (1)

Where Y_it is regional GDP growth over time (t) with 
regards to the variations in physical capital K, human 
capital H, Total Factor Productivity,A. Labor implies 
the embodiment of  natural skills in people while human 
capital may be defined in terms of  skills built in people 
through channels of  education, training, and experience 
Mankiw, Romer and Weil, 1992; Erken et al., 2016). 
Therefore, labor may be expressed as shown in equation 2:
Yit/Lit =A(1-α-β ) ((Kit/Lit ))

α ((Hit/Lit))
β                                                                  (2)

When expressed in natural logarithms, equation 2 
transforms into equation 3:
ln(Yit/Lit)=(1-α-β)+ln(Ait)+αln((Kit/Lit)+βln(((HitH)/
(LLit ))                         (3)
It can easily be observed from equations 2 and 3, that 
the productivity of  labor is measured as capital-labor 
ratio (K/L), human per capital labor ratio(H/L) as well 
as the residual term (1-α-β) ln(A). The residual term in 
the context of  this study is a parameter that measures 
the digital/internet penetration in a given world region 
(Erken et al., 2016).

MATERIAL AND METHODS 
Research Design and Data Description
The panel experimental quantitative study design is used 
in this study. Yearly panel data from 2005-2021 on five 
(5) world regions (excluding high-income countries) 
namely, Africa (Sub-Saharan Africa, SSA) Europe 
(Eastern Europe), Asia (South Asia),the Americas (Latin 
America & the Caribbean) and the Organization for 
Economic Cooperation and Development, OECD2 are 
used. The OECD region is included as a benchmark 
region because it represents the high-income and deepest 
internet penetration region among the regions in this 
study. The panel regional data, Gross Domestic Product 
per capita at current USD3 (as a dependent variable), and 
explanatory/regressor variables are Goods export namely 
goods exports (denoted as Goods_EXP,) Goods imports 
(denoted as Goods_IMP), Number of  People Using 
the Internet  (denoted as Internet_Users, expressed in 
terms of  per centage  %  of  the population) and Secure 
Internet Servers (denoted as Sec_Servers, per 1 million 
people-these ensure the confidentiality and protection 
of  data from hackers) were extracted from World Bank’s 
Development Indicators database (WDI) of  the World 
Bank (2023). However, the Digital Services Trade Exports 
regressor variable data were extracted from the United 
Nations Conference on Trade and Development database 
(UNCTADstat). The digital trade imports data were not 
available on the UNCTAD database (UNCTADstat), so 
the author proxied digital services imports with Computer 
Communications Services (as a % of  commercial service 
imports) obtained from the World Development Index 
(2023) database. This study follows Mini-Thomas (2018), 
Maune (2019) and Mulenga and Mayondi (2022), in using 

goods exports and goods imports regional panel data as 
control variables for digital services trade.

Regional Panel Unit Root Diagnostic Tests
Green (2003) advises that since economic data seldom 
exhibit stationarity and frequently gravitates towards 
the unit root, it is necessary to conduct stationarity 
and unit root tests. These tests mitigate the incidences 
of  conducting spurious regressions, inferences, and 
recommendations. Therefore, the researcher carried 
out four (4) regional panel unit root tests namely, the 
Common root-Levin, Lin & Chu (2002, LLC,2002 
hereafter), the Individual root-Im, Pesaran & Shin (2003, 
IP & S, 2003 hereafter), Individual unit root-Augmented 
Dickey-Fuller, (1979 ADF, 1979 hereafter) and Individual 
root-Phillips and Peron, (1988, PP, 1988 hereafter).
Theoretically, regional panel root tests can be conceived 
simply as applying multiple panel data time-series unit 
root test diagnostics (Levin et al., 2002). Panel unit root 
tests evaluate whether the autoregressive (AR) process 
is constrained throughout the cross sections of  the 
panel data models (Woodridge, 2000; Dickey & Fuller, 
1979). In terms of  unit root test algorithms, this study 
follows Mulenga and Mayondi (2022). Starting with an 
autoregressive procedure, AR (1):
yit=ρi y(it-1)+Xit δi+εit                                      (4)
Where i=1,2,…,N cross sectional series observed over 
periods t=1,2,…,Ti
The Xit  refers to a matrix of  external/exogenous 
factors in the model which include, among others, fixed 
effects, and individual trends. ρi measures autoregressive 
coefficients, whereas the ε(it)is a matrix or vector of  white 
noise error terms.
It follows that given that; |ρi |<1, yi becomes trend/
weakly stationary. If, however, |ρi |=1, yi is said to be 
non- stationary/has a unit root (Phillips and Peron, 
1988). The two assumptions normally taken regarding 
ρi  in the process of  conducting the panel unit root test 
diagnostics are that, firstly, we assume that persistence 
parameter are common across-sections such that ρi=ρ 
for all. The LLC (2002) common unit root test is done 
in accordance with this assumption. Secondly, we assume 
also that ρi  has the freedom to vary across the sections in 
the panels. The individual IP&S (2003) panel unit root test 
is conducted in the backdrop of  the second assumption. 
The ADF (1979) and the individual PP (1988) unit root 
test diagnostics ride on this assumption. 
The basic ADF (1979) panel unit root test may be denoted 
as shown in equation 5:
∆yit=αy(it-1)+∑(pi)

(j=1) βij ∆y(it-j) +X’it δ+εit                         (5)
This test rides on the assumption that α=ρ-1 and that 
the lag order, ρ_i, varies freely across the sections. We 
express the respective null and alternative hypotheses for 
the ADF (1979) unit root tests as follows: 
H0 : α=0, (Implies presence of  a unit root is a unit root)
H1 : α<1 (Implies absence of  a unit root/ data stationarity) 
Given that all the five (5) panel unit root procedures 
carried out in this study began with the basic ADF 



Pa
ge

 
13

5

https://journals.e-palli.com/home/index.php/ajebi

Am. J. Econ. Bus. Innov. 2(3) 131-144, 2023

(1979), we only report the test statistics algorithms or 
model expressions in order to conserve space. The LL& 
C (2002) common unit panel unit root test, assumes that, 
under the null, a modified test-statistic for the resultant α 
is asymptotically normally distributed:

              (6)

on economic growth, our study evaluates the impact of  
digital services trading on five (5) diverse world regions. 
Additionally, whereas Maune (2019) applied P-VAR on 
a panel of  countries from one region, specifically from 
the Sub-Saharan Africa, (SSA), this study evaluates the 
effect of  digital services trade on growth of  five (5) 
world regions. Furthermore, whereas Mulenga and 
Mayondi (2022) applied PVAR models to determine the 
digital services-economic growth  nexus on panel data 
from Developed, Emerging and Developing countries, 
this study employs PVAR dynamic models to examine 
the effect of  digital services trading on five (5) world 
regions from Africa, Asia, Europe, the Latin America and 
the OECD cluster, and extends the data analysis by one 
additional variable (Secure Internet Servers) as well as 
the time frame from 2005-2021. Additionally, we follow 
Simon and Pingfang (2021) in employing static models 
of  Fixed Effects (FE) or Least Squares Dummy Variables 
(LSDV) models. We employ FE estimators to conduct 
static regressions because the Hausman test rejected the 
null hypothesis that supports the use of  Random Effects 
(RE) models.
In compact (in log-difference) format, we can express the 
PVAR dynamic model as shown in equation 9:
dlnYit=αt+β1dlnX it+β2dlnX it+βkdlnXkt+μXit(μM it)+ 
ωXit+ωMit +ε(it)                (9) 
Where the variables Xit,...,Xik  refer to a matrix/ vector of  
independent(regressor)variables,γ_it refers to the regional 
economic/GDP growth, the coefficient estimates 
β1, β2,..., βk refer to the quantitative effects that each 
respective regressor in the model has on regional GDP 
growth(γit), holding other regressors constant i refers to 
region  i, t refers to  time,  and μxit,(μMit,) refers regional 
i’s unobservable individual effects on export (import) 
equation, ωXit, ωMit are unobservable time fixed effects 
for exports and imports respectively and, εit,refers to 
the stochastic error term parameters αt refer to a vector 
of  intercepts in each time series(year). We can write the 
compact expression under expression 9 more specifically 
as shown in equation 10:
dlnYit= αt + β1dlnDigserv_EXPit + β2dlnDigserv_IMPit + 
β3dlnG_EXPit + β4dlnG_IMPit + β5dlnINT_Usersit + β5dlnSEC_
INTSERVit+ μXit(μMit) + ωXit + ωMit + εit                                (10)
Where Yit is GDP per capita in current US Dollars for-
ith region at time t, αt denotes various intercepts for 
different years in each time series, β1 Digserv_EXPit is 
digital services exports trading for region i at time t, β2 
Digserv_IMPit denotes digital services trading imports 
variable for region i at time t, β3 G_EXPit denotes goods 
exports trading variable for region i at time t, β4 G_IMPit  
refers to the goods imports variable for region i at time 
t, β5 Int_Usersit refers to the number of  internet users 
measured as per centage points (%) of  the population 
for region i at time t and SEC_SERVit refers to the secure 
internet servers per 1 million people. ωXit, ωMit denote 
unobservable time fixed effects for exports and imports 
respectively and, εit  denotes a stochastic disturbance term 

Where; tα is the  standard test statistic â=0. â2  refers to the 
estimated variance of  the stochastic error term, εit  SE(â) 
is the standard error of  â, and T=t-(∑i (ρi/N))-1. The 
terms μ(m(T* )), and σ(m(T* )), refer to respective adjustments 
for the mean and variance. The respective null & 
alternative hypothesis for the IP& S (11) unit root test 
can be conveniently expressed as shown in equation 7:
H0=0,∀ i

              (7)

The average of  the t-statistics αi from the individual ADF 
regressions t(i(Ti)) (ρi):

IP&S (2003) assert that a properly standardized 
has an asymptotic standard normal distribution, N(0,1).  
As an alternative diagnostic panel set of    unit root tests, 
Fisher’s test may be employed. This unit root test process 
tends to combine the p-values from the individual unit 
roots test processes. Now, Wooldridge (2000) suggested 
that if   πi  is defined as the p-value for any individual unit 
root test for cross section i, it implies that under the null 
unit root for all N, for all cross sections, the asymptotic 
result may be written again in a convenient format as 
shown in equation 8:
-2∑N

(i=1)log(πi)→X2
N                                    (8)

IP&S (\2003) demonstrate that;
z=1/(√N) ∑N

(i=1) Φ
(-1)(πi)→N(0,1),

Where Φ(-1) denotes the of  the standard normal 
Cumulative Distribution Function. 
E-views 9.5, statistical software employed in this study, 
yields the asymptotic Fisher chi-squared (X2) and the 
standard normal statistics by applying and reporting 
the ADF 1979, i.e ADF(1979) & PP 1988, i.e PP(1988) 
individual unit roots diagnostic results. The null and 
alternative hypotheses are the same as those in individual 
root-IP& S (2003), panel unit root test procedures as 
expressed in equation number 7.

The Empirical Econometric Models
The empirical models employed to assess the dynamic and 
fixed effects of  digital services trading on the economic 
(GDP) growth of  each of  the five world regions, we 
follow Mini-Thomas (2018), Maune (2019) and Mulenga 
and Mayondi (2022) in applying the panel vector auto-
regression (PVAR) empirical econometric models. This 
study, however, is distinct from that of  Mini-Thomas 
(2019) in that while Mini-Thomas (2019) evaluated the 
impact of  Information Communication technology (ICT) 



Pa
ge

 
13

6

https://journals.e-palli.com/home/index.php/ajebi

Am. J. Econ. Bus. Innov. 2(3) 131-144, 2023

i at time t. The various intercepts for different years in 
each time series denoted as αt  accounts for the change 
in GDP per capita per year. In empirical literature, it 
is recommended that panel data analyses be done in 
natural log form to rescale data thus making variance 
constant, and to reduce the effects of  positive skewness. 
Additionally, data in natural logs transforms non-linear 
data to linear (Green, 2003; Maune, 2019; Stock and 
Watson, 2001).

Pvar Versus Pvecm
Despite being widely used in the literature, VAR models 
(refer to Lutkepohl, 1991, Johansen, 1995 for detailed 
discussions), we briefly discuss it here highlighting reasons 
why we employ PVAR models in this study. Lutkepohl 
(1999) and Green (2003) explain that the use of  PVAR 
models help to capture the dynamic interdependencies 
in panel data thereby considering the cross-sectional 
dynamic heterogeneities through the incorporation of  
time variations in the coefficients and in the variance of  
innovations or shocks. The same algorithms used to create 
ordinary VARs are used to create PVAR models, but PVAR 
models have cross section dimensional features added to 
them (Green,2003). VAR is a description of  the evolution 
of  the set of  k (endogenous) variables in the same sample 
period as a linear function of  their past changes. In other 
words, VAR model can be described as n-equation, with 
n-variables explained by their own lagged(past) and 
current values of  the remaining n-1 variables (Stock and 
Watson, 2001).  Thus, the basic Panel VAR (PVAR) or 
PVAR at levels can be expressed as shown in equation 11:

            (11)

VAR at levels may not be an appropriate estimator for 
cointegrated system of  equations. To this end, this paper 
employs the vector error correction models (VECM) 
given that the system of  equations in our study are 
cointegrated.
We follow Lutkepohl (1999) in expressing the panel-
VECM (PVECM) model as shown in equation 13: 

           (13)

Where p is the number of  (lags) parameters, A0 is a vector 
of  intercepts, yt is a vector of  endogenous variables, Ai  is 
k x k coefficient matrices, implying that, i=1,2,…,p,
Ut = K-dimensional denotes stochastic disturbance non 
time variant process (15). The basic PVAR model assume 
a VAR(p) process as the number of  lags equals p. It is 
assumed further that the VAR(p)  is generalizable enough 
to factor in probabilistic trends predicated on the process 
that is stable given that:

            (12)
This means that, given that the polynomial determinant 
in equation 11 has a unit root, implying also z=1, it 
follows that some or all the variables in the systems may 
be cointegrated of  order 1. Lutkepohl (199) advises that 

Where π=-(I(k)-A(1)-…-Ap) , and Γi= -(A(i+1)+⋯+Ap) for 
i=1,…p-1 
It is assumed that the expression: △y(it), contains no 
stochastic trends. All the variables are thus integrated 
of  order one, (I(1)) implying that the presence of  
cointegration relations is manifested by the term, π(yit-1) 
be I(0).When y(it) is cointegrated with cointegration 
rank, r, rank(π)=r<K  and π=αÁ’ where α and A are 
Kx r matrices. The term  Γ(j) (j=1,…,p-1) is interpreted 
as short run parameters while π(yit-1) term is the long 
run association part of  the P-VECM. The unknown 
VAR order p in equations 11 and 13 is estimated using 
the Akaike Information Criterion (AIC). In estimating 
the fixed effects, we employ the model which can be 
expressed generally as shown in equation 14:

  (14)
Where; γ is a vector of  dependent variables,β0 are 
intercepts, β(1) are coefficient estimates, X is a vector of  
regressors,
i=1,2,…n, D represents dummy variables,t denotes 
time,uit  is the  white noise error term for region i at time,t. 

RESULTS AND DISCUSSION
Diagnostic Tests and Choice of  Models
The author employed Hausman tests to determine 
whether to estimate Fixed or Random effects models. The 
Hausman test rejected the null hypothesis of  estimating 
Random Effects (RE). Thus, this author used the Fixed 
Effects (FE) estimator or Least Squares dummy variable 
(LSDV) models static estimators. We follow Kao (1999) 
and Pedroni (1999) in conducting regional panel data 
cointegration tests. Both Kao (1999) and Pedroni (1999) 
confirmed that regional panel data were cointegrated. 
Consequently, we used the PVECM as dynamic models. 

Summary Descriptive Statistics
Table 1 reports summary descriptive statistics for the five 
(5) sub-panel world regions surveyed in this study.

Table 1: Descriptive Statistics of  Five Regions
Mean Max Min Std. Dev Obs.

Panel 1: African Region (SSA)
GDP_CAP 16.57 29.34 6.13 7.02 121
DIG_EXP 24454.85 30268.34 18565.59 3560.73 121
DIG_IMP 35.77 42.99 32.43 2.38 121
G_EXP 3.58 4.35 2.68 5.36 121
G_IMP 3.53 3.96 2.96 3.04 121
INT_USERS 16.57 29.34 6.13 7.02 121
SEC_INTSERV 257.75 787.44 3.63 323.39 121



Pa
ge

 
13

7

https://journals.e-palli.com/home/index.php/ajebi

Am. J. Econ. Bus. Innov. 2(3) 131-144, 2023

Table 2 reports the panel unit test results. We observe 
in Table 2 that the variables in the African, Asian, Latin 
American, and European regions show that some data 
have unit roots while the other data is stationary at levels. 
However, all the variables in the OECD regional data are 
all integrated of  order 1.
Note that Fisher tests were computed using an asymptotic 
Chi-squared distribution. All other tests assume asymptotic 

normality. The null hypothesis assumes a common unit root 
process. *, **, ***, imply that the respective regional panel 
data variable is stationary at 10%, 5% and 1% significance 
levels. Root tests included individual intercepts only. ADF 
denotes is Augmented Dickey Fuller test, PP refers to 
Phillips and Peron unit root tests whereas LLC refers to 
Levin, Lin and Chu unit root tests and IPS denotes Im, 
Pesaran and Shin unit root tests.

Panel 2: Asian Region (SOUTH ASIA)
GDP_CAP 5303.88 6553.21 4057.73 804.70 132
DIG_EXP 446727.01 634458.50 277006.31 106447.94 132
DIG_IMP 49.62 56.94 41.64 3.911 132
G_EXP 3.65 4.11 2.82 3.61 132
G_IMP 5.67 6.77 4.42 6.56 132
INT_USERS 16.76 38.53 7.17 8.77 132
SEC_INTSERV 3.02 5.94 0.33 1.98 132
Latin American Region
GDP_CAP 14663.39 15972.73 12904.98 857.51 132
DIG_EXP 55599.11 61069.64 41868.39 5484.09 132
DIG_IMP 26.98 35.14 24.98 2.63 132
G_EXP 8.60 9.45 7.19 6.86 132
G_IMP 8.73 9.68 7.19 7.49 132
INT_USERS 53.19 73.39 34.01 12.24 132
SEC_INTSERV 508.19 1676.19 19.15 599.25 132
Panel 4: European Region (EASTERN)
GDP_CAP 9133.79 10962.47 7640.69 1063.64 132
DIG_EXP 82019.53 109716.44 57255.33 16289.40 132
DIG_IMP 29.02 38.97 25.95 3.34 132
G_EXP 1.12 1.26 8.81 1.34 132
G_IMP 1.06 1.19 8.73 1.12 132
INT_USERS 60.78 80.25 40.07 12.39 132
SEC_INTSERV 94.67 381.76 1.39 130.78 132
Panel 5: OECD Region
GDP_CAP 9133.79 10962.47 7640.69 1063.64 132
DIG_EXP 82019.53 109716.41 57255.33 16289.40 132
DIG_IMP 29.02 38.97 25.96 3.34 132
G_EXP 1.12 1.26 8.81 1.34 132
G_IMP 1.06 1.19 8.73 1.12 132
INT_USERS 60.78 80.25 40.064 12.39 132
SEC_INTSERV 94.66 381.75 1.39 130.78 132

Source: Authors computations using data from UNCTADstat  & the World Bank’s World Development Indicators

Table 2: Panel Unit Root Test Results
Variable LLC (t-stat) IPS (W-t-stat) ADF-Fisher X2 PP-Fisher X2 Order of  Integration
Panel 1: African Region
GDP_CAP -3.10* -2.51** 56.43** 145.78** I(0)
DIG_EXP -2.12** -1.13** 80.79* 145.22** I(1)
DIG_IMP  -3.14*       -1.11** 51.12** 123.78* I(1)
G_EXP -4.07* -2.8** 69.17** 419.37** I(1)



Pa
ge

 
13

8

https://journals.e-palli.com/home/index.php/ajebi

Am. J. Econ. Bus. Innov. 2(3) 131-144, 2023

G_IMP -2.97** -2.32** 190.10* 401.207** I(0)
INT_USERS -3.01** -2.31** 45.05** 201.92** I(1)
SEC_INTSERV -3.01** -1.02** 69.17* 117.75** I(1)
Panel 2: Asian Region
GDP_CAP -5.09***  -4.67** 90.34* 201.61*** I(1)
DIG_EXP -2.79** -1.54*** 112.89* 173.71*** I(1)
DIG_IMP -4.67** -3.72* 144.78* 69.17** I(1)
G_EXP -2.73* -1.79** 95.11** 130.10** I(1)
G_IMP -7.91** -6.09** 86.45** 88.95*** I(1)
INT_USERS -3.22** -5.79*** 111.23* 71.09** I(1)
SEC_INTSERV -5.13*** -4.67*** 205.12* 91.65** I(0)
Panel 3: Latin America Region
GDP_CAP -4.67** -2.028* 79.67* 319.37** I(1)
DIG_EXP -2.73** -1.78** 290.14* 201.21* I(1)
DIG_IMP -9.09** -6.71** 45.05* 201.92** I(1)
G_EXP -8.79* -7.00** 169.17* 174.78** I(0)
G_IMP -4.67* -3.50* 98.45 ** 209.56** I(0)
INT_USERS -4.67* -2.79** 79.14**  86.45* I(1)
SEC_INTSERV -5.66** -4.71** 211.57*  208.78* I(1)
Panel 4:  European Region
GDP_CAP -9.66* -7.91** 204.23* 312.97** I(1)
DIG_EXP -11.21* -7.79** 117.89* 214.78* I(0)
DIG_IMP -8.96** -6.47** 211.56* 305.11*** I(0)
G_EXP -7.11** -4.67** 189.2* 86.45*** I(1)
G_IMP -10.71* -3.44** 156.3* 121.38* I(1)
INT_USERS -12.79** -5.62** 94.9* 79.19** I(1)
SEC_INTSERV -8.76** -4.02* 78.9** 87.23** I(1)
Panel 5:  OECD Region
GDP_CAP -6.34*** -4.89** 82.29* 122.99** I(1)
DIG_EXP -8.79** -5.71* 101.34* 164.78** I(1)
DIG_IMP -4.26* -1.54** 98.27* 95.11** I(1)
G_EXP -6.93** -3.72* 120.23* 86.45** I(1)
G_IMP -7.91*** -1.79** 107.18 111.23** I(1)
INT_USERS -4.71* -6.09*** 69.27* 97.0*** I(1)
SEC_INTSERV -3.76** -1.67** 98.14* 133.99*** I(1)

Test for Linear Correlations among the Regional 
Panel Data Variables
The author conducted the correlational relationships 
so that we identify the linear relationships among the 
variables.  We report the results in Table 3.
For the African Region, digital services exports, digital 
services imports, goods exports, and secure Internet 
Servers variables exhibit a strong positive correlation 
with economic growth (GDP). The number of  people 
using the internet variable is positively and perfectly 
correlated with GDP. However, goods imports variable 
is weakly correlated with GDP. For the Asian Region, 
all the variables show a strong positive correlation with 
GDP except the goods exports variable which is weakly 

correlated with GDP. The Latin American Region shows 
that all the variables have a strong positive correlation 
with GDP except the goods export variable which shows 
a weak negative correlation. In terms of  the European 
Region, all the variables are weakly correlated with GDP, 
with digital exports, goods exports, goods imports and 
the number of  people using the internet variables being 
positively correlated. Digital services imports and secure 
internet servers have a negative correlation. Finally, in 
the OECD Region, all the variables have a positive and 
strong correlation with GDP except digital imports and 
goods imports variables which show a strong negative 
correlation with GDP.



Pa
ge

 
13

9

https://journals.e-palli.com/home/index.php/ajebi

Am. J. Econ. Bus. Innov. 2(3) 131-144, 2023

Table 3: Correlations Matrix
GDP_
CAPITA

DIG_EXP DIG_IMP G_EXP G_IMP INT_USE SECURE_
INTSER

Panel 1- Africa Region
GDP_CAPITA 1
DIG_EXP 0.98 1
DIG_IMP 0.95 0.87 1
G_EXP 0.66 -0.56 -0.66 1
G_IMP -0.41 -0.31 -0.46 0.8 1
INT_USERS 1 0.98 0.95 -0.66 -0.39 1
SECURE_INTSERV 0.91 0.933 0.85 -0.39 -0.32 0.67 1
Panel 2-Asia Region    
GDP_CAPITA 1
DIG_EXP 0.96 1
DIG_IMP 0.82 0.82 1
G_EXP 0.52 0.5 0.06 1
G_IMP 0.91 0.33 -0.03 0.93 1
INT_USERS 0.84 0.93 0.83 0.26 0.07 1
SECURE_INTSERV 0.98 0.96 0.87 0.44 0.31 0.89 1
Panel 3-Latin America Region
GDP_CAPITA 1
DIG_EXP 0.90 1
DIG_IMP 0.14 0.01 1
G_EXP -0.62 0.66 -0.06 1
G_IMP 0.59 0.77 -0.32 0.92 1
INT_USERS 0.88 0.68 0.59 0.35 0.22 1
SECURE_INTSERV 0.73 0.45 0.65 0.38 0.14 0.91 1
Panel 4-Europe Region
GDP_CAPITA 1
DIG_EXP 0.14 1
DIG_IMP -0.29 0.51 1
G_EXP 0.37 0.28 -0.08 1
G_IMP 0.44 0.37 0.07 0.98 1
INT_USERS 0.35 0.96 0.45 0.05 0.15 1
SECURE_INTSERV -0.42 0.93 0.41 -0.01 0.09 0.98 1
Panel 5-OECD Region
GDP_CAPITA 1
DIG_EXP 0.98 1
DIG_IMP -0.73 0.79 1
G_EXP 0.82 0.74 0.45 1
G_IMP -0.77 0.68 0.39 0.99 1
INT_USERS 0.99 0.96 0.80 0.81 0.77 1
SECURE_INTSERV 0.84 0.88 0.82 0.54 0.48 0.84 1

Source: Author’s computations using data from UNCTADstats & the Word Bank’s World Development Indicators (WDI) databases

Panel VECM Results-Dynamic Model Estimates
Table 4 reports a summary of  the dynamic PVEC model 
estimates of  the long run cointegration coefficients 
along with their respective speeds of  adjustments 

obtained by employing equation 13. The variables are in 
natural log format.
The PVECM results for the Africa region panel show 
that, over the long term, exports of  digital services have 



Pa
ge

 
14

0

https://journals.e-palli.com/home/index.php/ajebi

Am. J. Econ. Bus. Innov. 2(3) 131-144, 2023

an impact on GDP that is statistically significant at the 
5% level. That is, every 1% increase in exports of  digital 
services results in a 0.82% significant rise in GDP. A 1% 
increase in the number of  internet users and the number 
of  secure internet servers causes an increase in GDP of  
0.52% and 0.32%, respectively. Digital services exports 
variable returns to a long run equilibrium after experiencing 
a shock with a speed of  adjustment of  0.14%. 
The PVECM estimates for the Asian region show that 
digital services exports and digital services imports 
variables have significant but opposite effects on the 
regional GDP growth in Asia. In specific terms, every 
1% increase in digital services exports, GDP per capita 
significantly rises by 1.02%, and while digital services 
imports results in significant decrease of  GDP per capita 
by 0.23% in the long run. Additionally, a rise of  1% 
in the number of  internet users and in the number of  
secure internet servers increases GDP in the long run 
by 0.31% and 0.19% respectively. The respective speeds 
of  adjustments for digital exports, goods imports, the 
number of  people using the internet and secure internet 
servers are 0.12%, 0.07%, 0.16%, 0.16% and 0.13%.
For the Latin American Region, the PVECM estimates 
indicate that, GDP increases by 2.11% and 0.27%, 
respectively, for every 1% growth in exports and imports 
of  digital services. Additionally, a 1% increase in internet 
users results in a long-term significant increase in GDP 
of  0.34%. Secure internet server variable, however, 
significantly lowers GDP by 0.08% at a 5% level. 

Digital services exports, goods exports, goods imports, 
and Secure Internet Servers demonstrate the speed 
of  adjustment of  0.13%, 0.18%, 0.05%, and 0.15% in 
terms of  long-run convergence or return to equilibrium 
following a shock.
In the European Region, the PVECM estimates show that 
the GDP is significantly impacted positively by exports 
of  digital services exports, goods, the number of  internet 
users, and the secure internet servers’ variables. For 
instance, a 1% rise in each of  these variables significantly 
increases long-term GDP by 3.19%, 1.29%, 1.22%, and 
1.15%, respectively. However, the GDP significantly 
declines by 0.13% for every 1% rise in goods imports. 
The rates of  adjustment for the models of  digital service 
exports, imports of  digital services, exports of  goods, 
the number internet users, and secure internet servers are 
0.17%, 0.13%, 0.33%, 0.27%, and 0.21%, respectively. 
According to PVECM estimations from the OECD 
region, the GDP is significantly augmented by exports 
in digital services, imports of  digital services, goods 
exports, internet usage, and secure internet servers. 
In specific terms, GDP rises by 6.75%, 1.43%, 2.25%, 
1.81%, and 0.11% for every 1% increase in each of  the 
variables, respectively. On the other hand, the goods 
imports variable significantly lowers the GDP growth 
of  the OECD region by 1.06%. Digital exports, digital 
imports, goods imports, and secure internet servers’ 
models all return to long run equilibrium at speeds of  
0.37%, 0.49%, 0.27%, and 0.06%, respectively.

Table 4: Dynamic Panel Vector Error Correction (VEC) Model Estimates
LR Cointegration Estimates ECT Speed of  Adjustment Estimates R2

Coeff t-stats Std.Errors Coeff t-stats Std.Errors
Panel 1: Africa Region
lnGDP_Capita (Dependent  1 0.03 3.5** 0.02       0.71
lnDIG_EXP 0.82 -3.1** 0.58 -0.14 -2.9** 0.39 0.53
lnDIG_IMP 0.33 -0.59 0.52 0.01 0.34 0.04 0.67
lnG_EXP 0.13 2.2** 0.71 0.06 1.7* 0.59 0.69
lnG_IMP -0.17 1.8* 0.45 0.09 1.35 0.93 0.54
lnINT_USERS 0.52 2.8** 0.21 -0.11 -3.2** 0.08 0.49
lnSECURE_INTSERV 0.32 2.5** 0.07 0.03 0.43 0.17 0.51
Panel 2: Asia Region
lnGDP_Capita (Dependent)  1 0.09                                         2.1** 0.06 0.43
lnDIG_EXP 1.02 4.8** 0.06 -0.12 -6.2** 0.04 0.74
lnDIG_IMP -0.23 -2.9** 0.08 0.04 3.5** 0.08 0.63
lnG_EXP 0.29 -1.21 0.08 0.05 2.1** 0.03 0.76
lnG_IMP -0.57 1.9* 0.17 -0.07 -2.2** 0.34 0.74
lnINT_USERS 0.31 2.2** 0.16 -0.16 -4.6** 0.04 0.56
lnSECURE_INTSERV 0.17 5.1** 0.17 -0.13 -3.1** 0.06 0.71
Panel 3: Latin America Region
lnGDP_Capita (Dependent 1 -0.32 1.56 0.04 0.41
lnDIG_EXP 2.11 1.8* 0.07 -0.13 -5.3** 0.32 0.68
lnDIG_IMP 0.27 2.4** 0.07 0.23 0.62 0.36 0.71



Pa
ge

 
14

1

https://journals.e-palli.com/home/index.php/ajebi

Am. J. Econ. Bus. Innov. 2(3) 131-144, 2023

lnG_EXP 0.09 3.6** 0.16 -0.18 -1.9* 4.38 0.71
lnG_IMP -0.19 -2.2** 0.23 0.05 4.1** 0.42 0.52
lnINT_USERS 0.34 4.3** 0.09 0.28 0.78 0.32 0.44
lnSECURE_INTSERV -0.08 -3.1** 0.01 -0.15 -3.2** 0.14 0.59
Panel 4: Europe Region
lnGDP_Capita (Dependent 1 0.13 1.45 0.01 0.61
lnDIG_EXP 3.19 4.5** 0.06 -0.17 -2.4** 0.06 0.58
lnDIG_IMP -0.03 -0.58 0.03 -0.13 1.96* 0.08 0.63
lnG_EXP 1.29 3.7** 0.08 0.33 2.5** 0.68 0.79
lnG_IMP -0.13 -2.1** 0.06 0.05 0.14 0.76 0.73
lnINT_USERS 1.22 6.1** 0.04 -0.27 -2.0** 0.45 0.79
lnSECURE_INTSERV 1.15 2.3** 0.37 -0.21 -3.5** 0.72 0.51
Panel 5: OECD Region
lnGDP_Capita (Dependent 1   1 -0.41 1.06 0.09 0.65
lnDIG_EXP 6.75 3.7** 0.01 -0.37 -3.3** 0.11 0.84
lnDIG_IMP 1.43 2.2* 0.04 -0.49 -1.9** 0.04 0.74
lnG_EXP 2.25 2.4** 0.03 0.86 0.17 0.39 0.76
lnG_IMP -1.06 -3.4** 0.01 -0.57 -2.3** 0.05 0.74
lnINT_USERS 1.81 2.1* 0.02 0.27 1.32 0.08 0.57
lnSECURE_INTSERV 0.11 3.9** 0.02 -0.06 -6.3** 0.13 0.61

Source: Author’s computations using panel data from UNCTADstat and World Bank’s World Development Indicators (37) databases. 
ECT means error correction term. The asterisks: *, ** means that the respective variable is statistically significant at 10% and 5% 
respectively. The PVECM estimators include intercepts only. According to the guidance by the Akaike Information Criterion (AIC) the 
appropriate lag length for each regional panels were as follows: Africa - lag 7, Asia -lag 5, Latin America -lag 4, Europe - lag 5, and 
the OECD-lag 6. We subtracted 1 lag from each lag length as per the VECM modeling techniques (Green, 2003)  

Static Panel Fixed Effects Regression Estimates
After adjusting for regional unobserved heterogeneity 
factors, the author used fixed effects analyses to assess 
how each regressor affected the GDP growth of  each of  
the five (5) regions. From Table 5, it is easy to observe 
that the number of  internet users, digital products 
exports, goods imports, goods exports, and secure 
internet servers have long-run, significant impact on the 
GDP in the region of  Africa. Specifically, a 1% increase 
in digital services exports, goods exports, the number 
of  people using the internet and secure internet Servers 
results in a significant long run GDP increase of  1.01%, 
0.43%, 0.17% and 0.12% respectively at 5% levels. 
However, digital services imports and goods imports 
cause a significant long run decrease in regional GDP 

by 0.83% and 0.63% respectively. For the Asian Region, 
the Fixed Effects (FE) regression estimates indicate that 
a 1% increase in exports of  digital services, imports of  
goods, and secure internet servers result in GDP growth 
of  1.42%, 0.31%, and 0.21%, respectively. However, the 
GDP plummets by 0.19% and 0.22%, respectively, for every 
1% rise in goods exports and in the number of  internet 
users. The Latin American Region is significantly positively 
impacted in the long run by imports of  goods, the number 
of  internet users, and the secure Internet server variables. 
That is, for 1% rise in each of  the foregoing variables, the 
long GDP in the Latin American region correspondingly 
increase by 1.28%, 0.18%, 0.12% and 0.20% respectively. 
An increase in digital imports into this region causes 
significant decrease in the long run GDP o by 0.14%.

Table 5: Static Panel Fixed Effects (FE) Model Estimates
lnGDP_Capita 
(Dependent variable)

Hausman Test Summary

Coeff t-stat Std.Errors X2 Stat X2 d. f
Panel 1: Africa Region
lnDIG_EXP 1.01 2.06** 0.01 19.25** 5
lnDIG_IMP -0.83 -1.88 0.03 (0.02)
lnG_EXP 0.43 3.02** 0.05
lnG_IMP -0.63 3.96** 0.01
lnINT_USERS 0.17 2.19** 0.04



Pa
ge

 
14

2

https://journals.e-palli.com/home/index.php/ajebi

Am. J. Econ. Bus. Innov. 2(3) 131-144, 2023

lnSECURE_INTSERV 0.12 4.03** 0.02
Panel 2: Asia Region
lnDIG_EXP 1.42 5.16** 0.08 15.05** 5
lnDIG_IMP -0.19  -1.53 0.08 (0.03)
lnG_EXP -0.38 -3.83** 0.09
lnG_IMP 0.31 4.76** 0.06
lnINT_USERS -0.22 -1.66 0.03
lnSECURE_INTSERV 0.21 6.49** 0.01
Panel 3: Latin America Region
lnDIG_EXP 1.28 5.19** 0.02 11.55** 5
lnDIG_IMP -0.14 -9.73** 0.06 (0.01)
lnG_EXP -0.04 -1.21 0.02
lnG_IMP 0.18 4.96** 0.07
lnINT_USERS 0.12 5.19** 0.19
lnSECURE_INTSERV 0.20 6.09** 0.12
Panel 4: Europe Region
lnDIG_EXP 1.31 13.74** 0.23 21.55** 5
lnDIG_IMP -0.51 -6.03** 0.29 (0.00)
lnG_EXP 0.32 2.02** 0.11
lnG_IMP -0.37 -2.98** 0.07
lnINT_USERS 0.45 3.02** 0.67
lnSECURE_INTSERV 0.17 1.05 0.53
Panel 5: OECD Region
lnGDP_CAPITA 33.21** 5
lnDIG_EXP 2.23 16.74** 0.14 (0.04)
lnDIG_IMP -0.18 -16.28** 0.18
lnG_EXP 1.31 2.02** 0.09
lnG_IMP -0.11 -3.78** 0.08
lnINT_USERS 0.41 16.33** 0.41
lnSECURE_INTSERV 0.15 5.97** 0.02

Source: Author’s elaboration on data from UNCTAD and World Bank’s World Development Indicators.  Notes: ** denotes statistically 
significant at 5% level. P-value are in parentheses (). d.f  denotes degrees of  freedom. X2 denotes Chi-square

The Fixed Effects (FE) regression estimates for Europe 
show that the have a long- impact on GDP. That is, 1% 
increase in goods exports, digital services trade, and 
internet users each causes 1.31%, 0.32%, and 0.45% 
growth in GDP, respectively. Meanwhile, GDP declines 
by 0.51% and 0.37 in long run when the region’s goods 
imports and digital services both rise by 1%.
We observe that, for every 1% rise in exports of  digital 
services, goods exports, the number of  people using the 
internet, and secure Internet servers, economic growth 
in the OECD region significantly increases by 2.23%, 
1.31%, 0.31, and 0.15%, respectively. Goods Imports and 
digital services trade imports both have long-run negative 
impact of  0.18% and 0.11% respectively.  

CONCLUSION
The study examines the of  digital services trading on the 
economic growth (Gross Domestic Product per capita, 

GDP) of  panel data from five (5) world regions namely, 
Africa, Asia, Latin America, Europe and the Organization 
for Economic Cooperation and Development (OECD) 
between 2005 and 2021. Quantitative experimental 
methods are employed. The Panel vector autoregression 
(PVAR) models are used as dynamic estimators while the 
Fixed Effects (FE) models are used as static estimators. 
The dynamic Panel Vector Error Correction (PVEC) 
model estimates indicate that the export of  digital services 
trade has significant long-run positive impact on GDP 
per capita (economic growth) in each of  the five regions, 
with the largest magnitude increase in GDP observed 
in the OECD region at 6.75%, followed by Europe at 
3.19%, 2.11% in Latin America, 1.02% in Asia and it is 
relatively lowest in Africa at 0.82%. In terms of  digital 
services imports, the largest long run positive significant 
effect is observed again, in the OECD region where a 
1% increase in digital imports services trade increases 



Pa
ge

 
14

3

https://journals.e-palli.com/home/index.php/ajebi

Am. J. Econ. Bus. Innov. 2(3) 131-144, 2023

GDP by 1.43% with the Africa Region ranking second 
at 0.33%. The findings from the static Fixed Effects (FE) 
estimates of  the impact of  digital services exports on 
the economic growth of  the five (5) world regions in our 
study also indicate a long run significant positive effect in 
all the world regions. Specifically, with each 1% increase 
in digital services trade variable, economic growth 
increases by 2.23% in the OECD region, 1.42% in Asia, 
1.31% in the Europe, 1.28% in Latin America and it is 
lowest in the Africa region where it increases by 1.01%. 
Digital services imports variable has a long run negative 
impact on the economic growth in the OECD, Europe, 
and Latin American regions. These findings confirm 
the hypothesis that deep internet/digital penetration 
positively impacts the efficiency of  digital services trade. 
The study, therefore, recommends that policymakers 
from the developing and emerging world regions should 
increase investments in digital infrastructure and digital 
policy development which should include among others, 
promoting policy and regulatory measures that augment 
digital infrastructure installations in rural regions, 
enhancing the digital skills, and promoting the adoption 
of  cutting-edge digital technologies to increase access to 
relatively cheap digital infrastructure and services.

REFERENCES
Brookings. (2017). Annual Report on economic Activities. 

Brookings Institution. https://www.brookings.edu/
wp-content/uploads/2017/12/2017-annual-report.
pdf

Coşar, A. K. and Demir, B. (2018). Shipping inside the 
box: Containerization and trade. Journal of  International 
Economics, 114, 331-345.

Department For International Development, DFID. 
(2020). Digital Strategy: Doing Development in a 
Digital World. https://www.gov.uk/government/
publications/dfid-digital-strategy-2018-to-2020-
doing-development-in-a-digital-world.

Dickey, D. A., & Fuller, W. A. (1979). Distribution of  
the estimators for autoregressive time series with a 
unit root. Journal of  the American statistical association, 
74(366a), 427-431. https://doi.org/ 10.1080/016214
59.1979.10482531.   

Erken, H., Donselaar, P. & Thurik, R. (2018). Total factor 
productivity and the role of  entrepreneurship. J Technol 
Trans, 43, 1493–1521 https://doi.org/10.1007/
s10961-016-9504-5

González, L., J. and Jouanjean, M. (2017). Digital Trade: 
Developing a Framework for Analysis’, OECD Trade 
Policy Papers, No. 205, OECD Publishing, Paris. 
http://dx.doi.org/10.1787/524c8c83-en

Green, W. H. (2003). Econometric Analysis (5th ed.) Pearson 
Education, Inc. New Jersey. 

Im, K., et al. (2003). Testing for unit roots in 
heterogeneous panels. Journal of  econometrics 
115. 153-74. https://www.sciencedirect.com/
science/article/pii/S0304407603000927?casa_
token=OxSK55oiKZAAAA

Inter-American Development Bank Annual Report. 
(2020). The Year in Review https://publications.iadb.
org/en/inter-american-development-bank-annual-
report-2020-year-review

International Communication Union, ICU. (2023). 
https://unstats.un.org/unsd/accsub-public/docs/
itu.pdf

Johansen, S. (1995). Likelihood-Based Inference in 
Cointegrated Vector Autoregressive Models. Oxford: 
Oxford University Press. Journal of  Money Credit and 
Banking, 10.

Kao, C. (1999). Spurious regression and residual-based 
tests for cointegration in panel data. Journal of  
Econometrics, 90, 1–44. 

Levin., et al. (2002). ‘Unit root tests in panel data: 
asymptotic and finite-sample properties,’ Journal of  
Econometrics, Elsevier, 108(1), 1-24, May. https://ideas.
repec.org/a/eee/econom/v108y2002i1p1-24.html

Lütkepohl, H. (1991). Introduction to multiple time series 
analyses. Berlin: Springer-Verla https://cupdf.com/
document/helmut-luetkepohl-1991.html

Lütkepohl, H. (1999). Vector autoregressive analysis’, 
SFB 373 Discussion Paper, No. 1999, 31, Humboldt 
University of  Berlin, Interdisciplinary Research 
Project 373, Quantification and Simulation of  Economic 
Processes, 10056302

Mankiw, Romer and Weil. (1992). A contribution to the 
Empirics of  Economic Growth. https://ideas.repec.
org/a/oup/qjecon/v107y1992i2p407-437.html

Maune, A. (2019). Trade in services-economic growth 
nexus: an analysis of  the growth impact of  trade 
in services in SADC countries. Journal of  Economics 
and Behavioral Studies, 11(2 (J)), 58-78. https://doi.
org/10.22610/jebs.v11i2(J).2819

Thomas, M. P. (2019). Impact of  services trade on 
economic growth and current account balance: 
Evidence from India. The Journal of  International Trade 
& Economic Development, 28(3), 331-347.

Mulenga, R., and Mayondi, M. (2022). Impact of  Digital 
Services Trade on Economic Growth of  Developing,  
Emerging and Developed Countries: P-VAR 
Approach. American Journal of  Economics, 6(2), 58-85. 

OECD policy brief. (2022). https://www.oecd.org/
trade/topics/digital-trade/

OECD. (2020). Digital Economy Outlook’. https://
www.oecd-ilibrary.org/science-and-technology/
oecd-digital-economy-outlook-2020_bb167041-en

Pan, W., et al. (2022). Digital economy: An innovation 
driver for total factor productivity’, Journal of  Business 
Research, 139, 303-311, https://doi.org/10.1016/j.
jbusres.2021.09.061

Pedroni, P., (1999). Critical values for cointegration tests 
in heterogeneous panels with multiple regressors’, 
Oxford Bulletin of  Economics and Statistics, 61(special 
issue), 653-670.

Pesaran, H. M., (2003). A Simple Panel Unit Root Test 
in the Presence of  Cross Section Dependence, 
mimeo, university of  Southern California. Problèmes 



Pa
ge

 
14

4

https://journals.e-palli.com/home/index.php/ajebi

Am. J. Econ. Bus. Innov. 2(3) 131-144, 2023

économiques, 1997, Le système financier international, 
numéro double 2541- 2542.

Phillips, P. C., & Perron, P. (1988). Testing for a unit root 
in time series regression. biometrika, 75(2), 335-346. 
10.1093/biomet/75.2.335

Simon A & Pingfang D. (2021). International trade and 
economic growth in Africa: The role of  the digital 
economy. Cogent Economics & Finance, 9, 1 https://doi.
org/10.1080/23322039.2021.1911767

Steinwender, C. (2018). Real effects of  information 
frictions: When the states and the kingdom became 
united. American Economic Review, 108(3), 657-96.

Stock, J. H., and Watson, M. W. (1988). Variable Trends in 
Economic Time Series. Journal of  Economic Perspectives, 
2(1988), 147-74

UNCTAD, (2020). Handbook for measuring digital trade 
OECD policy brief, 2022. https://www.oecd.org/
sdd/its/Handbook-on-Measuring-Digital-Trade-
Version-1.pdf

UNCTAD. (2020). E-commerce and the digital economy. 
https://unctad.org/topic/ecommerce-and-digital-
economy

UNCTAD. (2022). Digitization of  services: What 
does it imply to Trade and Development. https://
unctad.org/system/files/official-document/
ditctncd2021d2_en.pdf

United Nations Conference on Trade and Development, 
UNCTAD. (2022). Digital trade: Opportunities and 
actions for Developing Countries. Policy Brief  no.92 
January 2022 https://unctad.org/webflyer/digital-
trade-opportunities-and-actions-developing-countries

Wang, M. L., & Choi, C. H. (2019). How information 
and communication technology affect international 
trade: A comparative analysis of  BRICS countries. 
Information Technology for Development, 25(3), 455–474. 
https://doi.org/10.1080/02681102.2018.1493675

Wooldridge, J. M. (2000). Econometric Analysis of  Cross 
Section and Panel Data, The MIT Press Cambridge, 
Massachusetts London, England.

World Bank (2016). World Development Report 2016. 
Digital Dividends. https://www.worldbank.org/en/

publication/wdr2016
World Development Indicators WDI, (2023) 

https://databank.worldbank.org/source/world-
development-indicators or World bank.org

World Economic Forum. (2020). WEF Report (2020), 
Shaping the Future of  Advanced Manufacturing 
and Production https://www.weforum.org/
reports?platform=shaping-the-future-of- production 
&year=2020#filter

Ying, M., and Gao. (2022) Research on the relationship 
between digital economy, technological innovation 
and industrial structure upgrading Price: Theory and 
Practice(5), 70-73

Zhang, G.; Ye, Y.; Sun, M. (2023). Assessing the Static 
and Dynamic Efficiency of  Digital Economy in 
China: Three Stage DEA–Malmquist Index Based 
Approach. Sustainability 15, 5270. https://doi.
org/10.3390/su15065270

Zhu, Y., Liu, J., et al. (2023). Research on the Impact 
of  Digital Economy on International Trade Under 
the Double Circulation Pattern. Atlantis Press 
https://www.atlantis-press.com/proceedings/
icaid-22/125977067

Appendix
Foot Notes

1Digital Enablers determine the effectiveness and 
efficiency of  digital trade. The level of  development of  
the cohort of  digital technologies determine the digital/
internet penetration (digital connectivity and digital 
depth). For a detailed discussion on the effect of  internet 
penetration on digital trade, see, for example, World Bank 
2016 , and Brookings, 2017.

2Countries in Sub-Saharan Africa (SSA), Eastern Europe, 
South Asia and Latin America (and the Caribbean) 
regions are defined as developing regions by the United 
Nations and the World Bank.  For details, see https://
www.un.org/en/development/desa/policy/wesp/wesp_
current/2014wesp_country_classification.pdf

3USD refers to the United States Dollar.


