Internet tradition and tourism development: A causality analysis on BRI listed economies

Abstract

The study aims to explain the economic impact of Internet implication in tourism sector by taking sample of mega project listed countries (which provide big pitch to boost tourism business). Our work find the volatility cause of tourism revenue at country i, by examining the inbound tourist expenditures as a factor of technological infrastructure. We deploy data ranging from 1990 to 2017 and uses error correction model as representative of Autoregressive-Distributed Lag (ARDL) model after addressing diagnostic tests (for data reliability concern). We found long- and short-run association between tourism expenditure and information and communication technology (ICT) proxies in case of developed economies, while only short-run association in underdeveloped countries. The startling scenario about underdeveloped economies are also confirmed by one-way causation in our analysis. After sensitive analysis at each slot, the study concludes that tourism revenue is streaming low across those boundaries where tourists are suffered by paying more due to technological inaccessibility and its underdeveloped infrastructure. The suffered economies are recommended to upgrade their ICT sector to facilitate inbound tourist.

Keywords

ARDL Belt and Road Initiative (BRI)causality ECM Internet usage tourist expenditure

Introduction

In the late 1990s, the proliferation of information and communication technology (ICT) in tourism industry across intercontinental boundaries discerns the indispensable path to economic growth for both the developing and developed countries. The ICT role in tourism management and satisfying tourist behavior is briefly studied in literature (Bethapudi, 2013; Chiappa and Baggio, 2015; Da Xu, 2016; Mariani et al., 2014; Mihajlović, 2012; Tsokota et al., 2017). Researchers recognizes that tourism sector has the potential to leverage the economic health of any economy in the modern competitive era (e.g. WTCC (2018) report states that tourism yield covers 9.9% unemployment gap in 2017, which elucidate major role in poverty alleviation).

Tourism-related product/services seems to be fascinating when the associated traits are sensationally depicted on electronic environment. Internet reduces information asymmetry which help tourist to compare and select the different spots on different basis like venue attractiveness, associated expenses, product/services. With this regards, literature reveals favorable postulations when tourism management considers technological aspect (Czernich, 2014; Gruber et al., 2014; Ishida, 2015; Lechman and Marszk, 2015).

Our study investigates the impact of ICTs on inbound tourist inflow stream by considering mega project (Belt and Road Initiative (BRI)) as intervening role. Our focus is to analyze how proper ICT structure plays its role in minifying inbound tourist expenditures, which resultantly surge the tourist inflow stream as well. As the tourists from one country are unfamiliar with the site parameter of the other country, the online penetration can help them plan time, trace location, budget, and necessary stuff should be carried.

The significance of the Internet to tourism industry has progressively surged over the past few years. Literature postulates that proper ICT structure is plying significant role to persuade the diversified tourist behavior (Aguilo et al., 2017; Brida and Scuderi, 2013; Olya and Mehran, 2017; Smolčić and Soldić, 2017). Almost all developed nations use different ICT applications in order to facilitate international tourists and create synergies. Marcussen (2009) postulates that most of the services or products are on tap through different ICT applications in Spain.

According to World Tourism Organization (WTO Report, 2018), the number of total international tourist arrival is 1322 million which is 7% increase compared to previous record. The World Travel & Tourism Council (WTCC Report, 2018) reveals that tourism inflow contributes 10.4% to gross domestic product (GDP) in 2017 which subsidize 9.9% employment gap. The WTCC (2018) also stated that investment of the government in promoting tourism administrative, informative, and other public services is helping in attracting inbound tourist.

Some studies reveal that the government plays a significant role in implementing the ICT application for the tourism industry (Bethapudi, 2015; Tsokota et al., 2017; Utz, 2017). Government tries to construct a digital economy which has benefits like user-friendly, reliability, and security. In the digital economy, products and services are traded on conveniently (24/7) like TaoBao and Jingdong software in China. E-Government is playing an integral part in nurturing e-services which simultaneously facilitating tourism.

Figure 1 shows the tourism revenue-wise distribution of inbound tourist to all BRI listed countries in relation with Internet penetration rate. It is vivid from the above depiction that some countries like Poland, Singapore, or Thailand are competitive in such race and also successful in achieving the strategic goals of high revenue from the source. Low spending countries like Afghanistan, Gaza, Turkmenistan, and so on has stumpy potential lucrative source due to weak governance. The countries like Romania–Russia, Georgia–Vietnam–Armenia, Hungary–Israel, and so on are depicted similar dots in the above Figure 1, which shows their similar revenue level in the tourism industry.

Figure 1.

Tourism revenue-wise distribution of BRI listed countries. BRI: Belt and Road Initiative.

We find the short- and long-run effect of Internet usage on economic growth, tourism, and financial development in BRI listed countries. Until now, no work is made to analyze the impact of ICT structure on tourism expenditures in different economies tethered with mega project (BRI).

The study outcomes will be beneficial to construe the factor responsible for inbound tourist expenditures. For better understanding, we investigate the structural role in disparate nations and visualize the results distinctively. For this purpose, we especially focuses on 65 BRI (which is initiated by Chinese government) listed economies on income and region basis.

The rest of the work is structured as follows: The second section outlines the brief literature which enlighten the on-hand topic by different researchers. The third section elaborates the data and methodology and study hypothesis used in the research. The fourth section reveals the results and their respective interpretations. The fifth section ends the article by concluding and limitation while the sixth section recommends some future direction and makes policy implication of the research agenda.

Literature review

Literature of the study is segmented in three major parts: The first part is about descriptive nature literature while the second part is about empirical perspective of ICT influence over tourism development. The third part specifically targets only those passionate researchers, who analyze the Internet influence over tourism development.

Descriptive literature

Tourists tend to travel for mental satisfaction but only on the condition when they have sufficient information about the place where they are visiting. Kozak and Rimmington (2000) pointed out that tourist satisfaction level is associated with the information of the destination even in the off season. Devesa et al. (2010) studied the tourist motivation and satisfaction level and find tourist information as one of the significant attribute that effect tourist satisfaction level globally.

Information related to tourist expenditure is highlighted more significantly in the literature. Researchers like Brida and Scuderi (2013), Smolčić and Soldić (2017), Aguilo et al. (2017), and Olya and Mehran (2017) point out tourism expenditure as a significant factor in reshaping tourist behavior. Our study tend to visualize if these expenditures are explicit through technological visuals, it will resultantly surge inbound tourist stream.

No doubt, information technology is considered animus for the tourism industry, but it also escalate competitive environment globally (Abou-Shouk et al., 2013; Assaf and Tsionas, 2018; Lu and Stepchenkova, 2015; Sigala, 2015). New destinations and organization were instigated in this industry (with extra features) and gives tuff time to the existing one as a prime mover. Survival for whacked destinations and organizations is possible to reformate the provided facilities layout especially the ICT (Internet) structure for depicting quality product and services at ease and low cost. A continuous business process reengineering is needed to grasp and sustain the tourist attentions.

WTO and WTCC divulge the economic benefit of tourism industry in their annual report 2018. Their report suggested that reengineering process is needed in the information section by introducing new technologies and techniques via Internet. Researchers like Benhabib and Spiegel (2005) exhibit that generation and distribution of ideas and information at low cost can help the adaptation of new technologies (learned from others) which facilitating the economic performance. Jorgenson et al. (2008) also propose that transformation of knowledge across business organization build other networks which generally promote global competitive culture.

The technological trend in tourism industry expedites end user with easy access to updated information related to tourist-based product/services (e.g. web, fixed telephone line connected to rental agencies like car, hotels, online payment system, etc.). This technological trend in tourism industry facilitates inbound tourist by minifying unnecessary manual operational expenditures, which are causing for their attractiveness.

Bethapudi (2015) highlights the role of ICT in rural tourism development by using primary and secondary data. Researcher identifies the key role of technology and posits recommendation to government specifically in avoiding power failure, installing Wi-Fi facility, and embarking mobile commerce trend.

Literature postulates that market is reshaped due to ICT critical role at tourism industry which resultantly embark competitiveness (Abou-Shouk et al., 2013; Assaf and Tsionas, 2018; Lu and Stepchenkova, 2015). Sigala (2015) states that due to Internet, major customization is provided to the end user and enables tourist to analyze/compare the niche market traits. This customization phenomenon favors tourist empowerment, association, and liberation in choices, which is the key driving force in budgeting expenditures.

Pesonen (2013) segmented the tourist market in two parts: travel motivations and socio demographics. Researcher reports that desired financial goals are achieved by such segmentation and recommend to work more in this area. Other researchers like Filieri and McLeay (2014) posits that increasing dependency on Internet for making decision about tourist place, product, and services gives research direction in buzz marketing.

Visitor consumption (which is essential component of total tourism demand) is composed of multifarious items (Laimer et al., 2006). Our research mainly focuses on the role of ICT structure which if factually depicts the outlay details, will motivate tourist to visit. The next part of the study reveals empirical literature which comprises of facts and figures about the topic agenda. Empirical literature also postulates the tools and techniques used by different researcher during their study analysis.

Empirical literature

There is a variety literature present on the role of ICT reshaping tourism industry especially when inbound tourist attraction is targeted by online application systems (Sigala, 2015). To distinguish the significant changes made in ICT sector regarding tourism promotion, researchers like Navío-Marco et al. (2018) re-examine the Buhalis and Law (2008) work after 10 years about consumer demand dimensions, ICT role, industry function, and their respective implications. Researchers reported that technological trend which is pinpoint by Buhalis and Law (2008) is altered by new tendencies like Internet of things (IoTs) as their study contribution. Furthermore, the service connectivity factor is significant in tourism growth sustainability.

Lee and Jan (2019) examine the tourism growth sustainability by analyzing questionnaires data of six Taiwanese communities to know their perceptions based on tourism area life cycle theory (Butler, 2006). Researchers collect 849 questionnaires in total depicting the environmental, socio-cultural, and economic sustainability of the tourism sector. Researchers reveals that perception is a variant factor which depends on the development stage of the sector.

Divisekera and Nguyen (2018) investigate the determinants of innovations in tourism industry by deploying panel data of marketing and service as innovation outputs for tourism growth. Researcher reported that ICT and human collaboration, environment, foreign ownership, and size as innovation inputs are the key determinants.

Tsokota et al. (2017) examine the influence of ICT on tourism sector within Zimbabwe region by extending the Yin (2014) case study as research methodology guideline. Researcher pinpoint that lack of eservices, infrastructure, poor system integration, low human development, weak ICT governance, and lack of finances obstructing the sustainability of Zimbabweans tourism sector.

Brida et al. (2016) investigate the ICT role in curbing the passenger rush at airport using face-to-face 28 set of survey questions from 995 travelers. Researcher uses odds ratios from Principal Component Analysis (PCA) and apply Logit model to testify the study hypothesis. Researcher finds ICT plays significant role and suggests to reformate the announcement structure to communicate the flight information in a way that hinders rush.

Howard and Hussain (2011) postulated that the insurrection of Arab success is due to the dominant role of mobile usage in economic operations. Researchers expose that how indulging mobile usage can assemble societies in a controlled and systematic manner.

Choi and Murray (2010) uses the generalized method of moment (GMM) econometric tool due to the endogeneity problem in a panel of 151 countries from 1990 to 2006. The researcher tries to reveal the Internet influence over the service trade graph. The study concluded an increase from 0.23% to 0.42% in service trade is occurred due to spurring 10% surge in Internet usage.

Myung and Changkyu (2009) used data from 1991 to 2000 of 207 panel countries and analyze economic development associated with Internet usage by indulging inflation, GDP as controlling variables. In order to grasp the endogeneity issue (T > N), the researcher employs GMM econometric tool supported by diagnostic tests like cointegration, cross-dependency test. Researcher reveals the significant part of Internet usage in the governmental operations by economic development.

Duncan (2009) posits a descriptive nature study upon the remarkable impact of Internet in tourism industry by targeting three major areas, industry and commerce structure and tourism planning. Researcher concluded that if Internet structure is well established, tourists will not be uncertain about their visit plan.

In a study of Noh and Yoo (2008), researchers analyzes 60 countries panels from 1995 to 2002 to know the association between Internet implementation, economic development, and income disparity. Researcher puts mix results by revealing that Internet and economic growth has an adverse correlation in high-income disparity countries. The researcher selects less developed and low-income countries from every region where only rich have knowledge and skills of using modern tools of the Internet while poor have don’t even access to those tools, for example, IoTs.

Internet role in tourism development

The evolution of Internet put versatility in consumer as well as producer operation. Internet makes customer sophisticated and determined regarding variety of products and services and perhaps the value of money and time. Tourists, now a days heavily depends on media to procure information about visit destination (Bethapudi, 2013). Standing et al. (2014) did meta-analysis from 2001 to 2010 to investigate the Internet impact on consumer and service provider. Researcher finds that out of 228 articles, 47 articles were found Internet related which shows necessity for further work. Researcher also highlights that Internet demand in operation is increasing because of its vulnerability and cost effectiveness.

The tourism industry needs to provide accurate information by adopting information technological tools. Huang et al. (2016) exposes that Web 2 marketing travel is a tactic used for variant destinations in order to attract a large amount of tourist from all over the world. The researcher also postulates that the blog appears to be exceedingly rated search engine for tourist keywords.

From the previous 10 years, Springer Link listed journal named Journal of Information Technology and Tourism consecutively published a large number of a research paper on subject mentioned topic. Wahab et al. (2017) contended that implanting the online tourism portals can make convenience to the tourist, which also proved effective in matching the offer and demand parity. The tourism sector is concentrated customer of ICT to provide accurate, reliable, and on-time information to the end user (Law et al., 2009; Law et al., 2010).

Currently, China implemented one feature of the global competitive mega project “Belt and Road Initiative (BRI)” which hold intensive excitement of its surroundings. The project aims to tighten the control by reinforcing the ICT features like global access through the Internet. Such globalize mechanism also deem the tourism sector on the prior basis as of its augmented role in economic development. Chinese government also installing underground Internet cable planning to build digital Silk Road initiative (Aly Song, 2017)

The on-hand study analyzes the advantageous role of ICT infrastructure and its usage in curtailing tourist expenditure especially when dealing with a mega project, for example, Chinese initiative regarding One Belt One Road mega project. The intervening impact of mega project is deemed due to its adverse impact on environmental degradations. Our work exposes that how different income-level countries took advantage from Internet usage when dealing with a globalized mega project (BRI). The study also analyzes Internet inculcation impact on tourism development in the short and long run at BRI listed region-wise nations.

Da Xu (2016) investigates the prospective challenges and trends of IoT applications in business operations. Researcher stretch the (Pretz, 2013) work and reported that things will be connected to Wi-Fi network through smart sensors. Researcher suggested that related to Belt and Road (BRI) success, technological innovation in shape of IoT needs to be intrigued across the mega project.

The remainder of the article is arranged as follow: The third part is about methodology used in the article, for example, data description, study hypothesis, or econometric tools. The fourth part is about results and discussion. Diagnostic tests are also included in the fourth part. The fifth part concludes the article in light of existing literature and on-hand results. Also, the conclusions are associated with policy implication in the sixth part.

Methodology

Introduction

The study deploys long-rang data (1990–2017) of 65 BRI listed economies which are further categorized into two groups: (1) income-level countries and (2) regional-base countries (see Table 1 and Table 2). The reason of taking long-rang data is to gauge long-run and short-run causalities for the below listed variables.

Table 1.

Countries list by their income-level line up with BRI.

No.	Lower income countries	No.	Lower middle-income countries	No.	Upper middle-income countries	No.	High-income countries
1	Afghanistan	1	Armenia	1	Albania	1	Bahrain
2	Nepal	2	Bangladesh	2	Azerbaijan	2	Brunei Darussalam
		3	Bhutan	3	Belarus	3	Czech Republic
		4	Cambodia	4	Bosnia and Herzegovina	4	Estonia
		5	Egypt, Arab Rep.	5	Bulgaria	5	Hungary
		6	Georgia	6	China	6	Israel
		7	India	7	Croatia	7	Kuwait
		8	Indonesia	8	Iran, Islamic Rep.	8	Latvia
		9	Jordan	9	Iraq	9	Lithuania
		10	Kyrgyz Republic	10	Kazakhstan	10	Oman
		11	Lao PDR	11	Lebanon	11	Poland
		12	Moldova	12	Macedonia, FYR	12	Qatar
		13	Mongolia	13	Malaysia	13	Saudi Arabia
		14	Myanmar	14	Maldives	14	Singapore
		15	Pakistan	15	Montenegro	15	Slovak Republic
		16	Philippines	16	Romania	16	Slovenia
		17	Sri Lanka	17	Russian Federation	17	United Arab Emirates
		18	Syrian Arab Republic	18	Serbia
		19	Tajikistan	19	Thailand
		20	Timor-Leste	20	Turkey
		21	Ukraine	21	Turkmenistan
		22	Uzbekistan
		23	Vietnam
		24	West Bank and Gaza
		25	Yemen, Rep.

Note: BRI: Belt and Road Initiative.

Table 2.

Countries list by their region-wise line up with BRI.

No.	Central Asia	No.	East Asia	No.	South East Asia	No.	South Asia	No.	Middle East and North Africa (MENA)	No.	Europe
1	Kazakhstan	1	China	1	Brunei Darussalam	1	Afghanistan	1	Bahrain	1	Albania
2	Kyrgyz Republic	2	Mongolia	2	Cambodia	2	Bangladesh	2	Egypt, Arab Rep.	2	Armenia
3	Tajikistan			3	Indonesia	3	Bhutan	3	Iran, Islamic Rep.	3	Azerbaijan
4	Turkmenistan			4	Lao PDR	4	India	4	Iraq	4	Belarus
5	Uzbekistan			5	Malaysia	5	Maldives	5	Jordan	5	Bosnia and Herzegovina
				6	Myanmar	6	Nepal	6	Kuwait	6	Bulgaria
				7	Philippines	7	Pakistan	7	Lebanon	7	Croatia
				8	Singapore	8	Sri Lanka	8	Oman	8	Czech Republic
				9	Thailand			9	Qatar	9	Estonia
				10	Timor-Leste			10	Saudi Arabia	10	Georgia
				11	Vietnam			11	Syrian Arab Republic	11	Hungary
								12	United Arab Emirates	12	Latvia
								13	Yemen, Rep.	13	Lithuania
										14	Macedonia, FYR
										15	Moldova
										16	Montenegro
										17	Poland
										18	Romania
										19	Russian Federation
										20	Serbia
										21	Slovak Republic
										22	Slovenia
										23	Turkey
										24	Ukraine

Note: BRI: Belt and Road Initiative.

Vector error correction model (VECM) approach is used to testify the short- and long-run heterogeneous shocks of large linear panel data having 28 years and 65 cross-sections. For robustness, the study employee dynamic ordinary least square (DOLS), fully modified ordinary least square (FMOLS) and visualize graphically by the impulse response function CUSUM graph, fitted curve, and residual plot techniques, respectively. Reliability issues are governed by using diagnostic test, for example, cross-dependency (CD) test and weak cross-dependency (CD) test by Pesaran (2004, 2015), second generation unit root tests, White test for heteroscedasticity, VIF test for multicollinearity, and Westerlund error correction model (ECM) panel cointegration tests (Westerlund, 2008), respectively.

Study hypothesis

The on-hand study analyzes the following hypothesis:

H₀: Inbound tourist expenditure are irrelevant to ICT proxies

Table 3 states the study variable including ICT proxies. ICT proxies are considered in structural (fixed telephone line) and operational (use of Internet) manner. With the help of econometric tools, the combine effect is tested and then elucidate the tourism revenue status on scatter diagram (Figure 1) as robustness check. The study deploy following econometric bolts and nuts.

Table 3.

Data elaboration and sources.

Variable name	Elaboration	Source	Code (if any)
TE	Tourism expenditure in the country (inbound tourism)	UNWTO	NA
IU	Individuals using the Internet (% of the population)	WDI	IT.NET.USER.ZS
MCS	Mobile cellular telephone subscribers	World Telecommunication/ICT Indicators Data	I271
FTS	Fixed telephone subscription	World Telecommunication/ICT Indicators Data	I112
PII	% of individuals using the Internet	World Telecommunication/ICT Indicators Data	I99H
GDP	GDP per capita, current prices (US dollars per capita)	IMF	NGDPDPC
CPI	Consumer price index	WDI	FP.CPI.TOTL

VECM Granger causality estimations

Granger (1969) proposes that same order integrated variables need Granger causality estimates to know their casual association. As unit root test (Table 7) results in postulate mix stationarity level. There as per Shahbaz et al. (2013), the casual link contends precise trend for appropriate policy measures. The following equation is used in the study for VECM Granger causality estimations

Δ {lnTE}_{i, t} = β_{0 i, t} + \sum_{i = 1}^{p} β_{1 i} Δ {lnTE}_{t - i} + \sum_{i = 1}^{p} β_{1 i} Δ {lnIU}_{t - i} + \sum_{i = 1}^{p} β_{1 i} Δ {lnMCS}_{t - i} + \sum_{i = 1}^{p} β_{2 i} Δ {lnFT}_{t - i} + \sum_{i = 1}^{p} β_{3 i} Δ {lnPII}_{t - i} + \sum_{i = 1}^{p} β_{4 i} Δ {lnGDP}_{t - i} + \sum_{i = 1}^{p} β_{5 i} Δ {lnCPI}_{t - i}

In equation (1), $β_{0 i, t}$ is the constant term for country i at time t. Rest of the β are coefficients of the respective variables. ΔlnTE _i,t is the natural log change in the inbound tourist expenditures.

CD test and weak CD test for cross-sectional dependence

Social data are mostly interdependent or not independent due to its intermingling nature upon one another. In a dynamic panel modeling, if such dependence is ignored, it leads the data analysis to misleading inferences which postulate oversize under the alternative hypothesis of cross-sectional correlation. Chow test value is shown in Figure 3.

Figure 3.

Overlapped boxplot.

To deal with the phenomenon, primarily a least square regression equation is generated as mentioned below

Δ y_{i t} = \sum_{j = 1}^{p_{i}} {\hat{a}}_{i j} Δ Y_{i, t - j} + \sum_{j = - q i}^{p_{i}} {\hat{γ}}_{i j} Δ x_{i, t - j} + {\hat{e}}_{i t}

Panel unit root tests (second generation)

Pesaran (2003) proposes a Covariate-Augmented Dickey–Fuller (CADF) test in heterogeneous panels with cross-sectional dependency problem.

H₀ = all variable series are not stationary.

To eradicate the CD issue, the cross-sectional averages are increased with CADF regressions of lag and 1st difference series.

Westerlund ECM panel cointegration tests

Westerlund (2008) develop a four-panel cointegration test to verify the existence of error correction term for each (or whole) large panels under the null hypothesis of no cointegration. The test deemed all variables at stationarity to check the existence/absence of cointegration in the case of N < T ≥ 10. The fore stated association among study variables let to deem for linear equation as below

ln (H_{i t}) = μ_{i} + τ_{i} t + β_{i} t ln (Y_{i t}) + τ_{i t}

Bootstrapping approach assure the results consistency by accommodating the cross-sectional dependency issue for calculating the p value. Prior to use bootstrapping, we execute the CD test by satisfying the T < N assumption.

Error correction term w.r.t ARDL model

The study applies second generation unit root test to relax the cross-sectional independence notion. As the study found mixed outcomes regarding cointegration, stationarity, and cross-dependency biases, we approach to error correction-based model to estimate the long- and short-run coefficients. The VECM is executed to contend the speed of adjustment of the explanatory variables towards error term. Following equations are postulations of the study analysis

Cointegration equation (long-run model)

z_{i, t - 1} = {ECT}_{i, t - 1} = y_{i, t - 1} - β_{0} - β_{1} x_{i, t - 1}

Estimated VECM with Tourism Expenditure (TE) as a target variable

Δ y_{i, t} = β_{0} + \sum_{i = 1}^{n = 65} β_{i, t} Δ y_{i, t - i} + \sum_{i = 0}^{n = 65} δ_{i, t} Δ x_{i, t - i} + φ z_{i, t - 1} + μ_{i, t}

The coefficient of cointegration equation is $φ$ for the error correction term which is mention $z_{i, t - 1}$ in the above equation.

Estimated long-run coefficients

As the study found cointegration among variables, so there is a long-run relationship to support the study hypothesis. To gauge the long-run coefficients, the study employs the following model:

Δ Y_{t} = β_{0} + \sum_{i = 1}^{p} β_{i} Y_{t - i} + \sum_{i = 1}^{p} γ_{j} Δ x_{1 t - j} + θ_{0} Y_{t - 1} + θ_{1} x_{1 t - 1} + ε_{t}

Heteroscedasticity consistent variance-covariance estimators

Heteroscedasticity problem often occurs when there’s a large difference among the variable’s value. We can say that dependent variable can’t hold over the independent variable values basis on its range differences as shown in overlapped boxplot (Figure 3). Koenker (1981) suggests a test for homoscedasticity (unconditional) assumption which is robust w.r.t asymmetry of disturbance. We use Lagrange multiplier (LM) test of $γ = 0$ in the model as below

E (u_{t}^{2}) = σ_{t}^{2} = σ^{2} + γ {(x_{t}^{'} β)}^{2}, t = 1, 2, 3, \dots, n

We provide short-run ARDL outcomes by mechanizing the heteroscedastic consistent estimators of the variance-covariance matrix (Chesher and Jewitt, 1987). In Table 12, we find the existence of the heteroscedastic problem in each category except for lower income, Central Asia, East Asia, and South Asia region, respectively.

Cumulative sum of recursive residuals and square of recursive residuals

Figure 7 elucidates the CUSUM and CUSUM Sq. graph of every category of the study variables, respectively. The plot shows whether the variations of the dependent variables are stable, mechanized, or asymmetric. The graph mechanism is formulated by the below equation

Figure 7.

Cumulative sum of squares of residuals and recursive residuals (CUSUM) plots.

S_{m} = \sum_{i = 1}^{m} ({\bar{x}}_{i} - {\hat{u}}_{0}) or S_{m} = \frac{1}{σ_{\bar{x}}} \sum_{i = 1}^{m} ({\bar{x}}_{i} - {\hat{u}}_{0})

In the above equation, ${\hat{u}}_{0}$ is the control factor of estimates, m is the number of sample and $σ_{\bar{x}}$ is the sample mean standard deviations.

Results and discussions

Descriptive, correlation, and causality tests

The below Table 4 (region base) and Table 5 (income base) postulate the large categories of descriptive, correlational, and causality tests of every region economies in their respective renamed sections. Each section exposes a number of countries and on hand, observations to be dealt. Where $\bar{x}$ =mean, $\tilde{x}$ =median, $σ$ = standard deviation, $S k_{b}$ = skewness, $β_{2}$ = kurtosis, $\land$ = minimum, and $\lor$ = maximum, respectively. Correlation test exposes the variable co-movement effect in a given time while causality contends the movement of one variable past value (x) due to another variable past value (y) which resultantly predict the value of x.

Table 4.

Descriptive, correlation, and causality test: Income base.

Var.	(1)	(2)	(3)	(4)	(5)	(6)	(7)	$\bar{x}$	$σ$	$\land$	$\lor$
(A) All (65 countries) obs. 1820
TE (1)	1							4.484	3.602	0.000	10.730
IU (2)	0.474	1						18.679	25.179	0.000	97.999
MCS (3)	0.641	0.484	1					12.216	5.510	0.000	20.974
FTS (4)	0.357	0.193	0.525	1	≠		≠	13.354	2.959	0.000	19.723
PII (5)	0.509	0.718	0.545	0.168	1			1.248	2.532	−9.859	4.516
GDP (6)	0.462	0.511	0.606	0.335	0.442	1		7.442	2.406	0.000	11.532
CPI (7)	0.500	0.411	0.622	0.248	0.381	0.499	1	3.441	1.804	−5.474	5.828
(B) Lower income countries (2) obs. 56
TE (1)	1	≠	≠					3.171	2.706	0.000	6.218
IU (2)	0.203	1			≠			−0.116	2.147	0.000	2.980
MCS (3)	0.634	0.522	1	≠		≠	≠	9.145	7.449	−6.985	16.969
FTS (4)	0.510	0.069	−0.030	1		≠	≠	10.644	3.618	0.000	13.647
PII (5)	0.222	0.978	0.543	0.083	1		≠	−0.069	2.157	−6.985	2.736
GDP (6)	0.693	0.205	0.742	0.096	0.222	1		4.644	2.487	0.000	6.726
CPI (7)	0.718	0.339	0.674	0.052	0.321	0.875	1	3.174	1.986	0.000	5.142
(C) Lower middle-income countries (25) obs. 700
TE (1)	1							4.012	3.364	0.000	9.940
IU (2)	0.395	1		≠	≠			0.549	2.454	−9.103	4.262
MCS (3)	0.678	0.484	1					11.803	5.857	0.000	20.665
FTS (4)	0.391	0.115	0.498	1			≠	13.106	2.840	0.000	17.731
PII (5)	0.427	0.947	0.498	0.164	1			0.489	2.479	−9.103	3.889
GDP (6)	0.468	0.265	0.540	0.314	0.254	1		6.471	1.808	0.000	8.708
CPI (7)	0.556	0.288	0.609	0.229	0.265	0.614	1	3.501	1.665	−5.474	5.460
(D) Upper middle-income countries (21) obs. 588
TE (1)	1			≠				4.613	3.766	0.000	10.730
IU (2)	0.505	1		≠				1.318	2.618	−9.859	4.383
MCS (3)	0.637	0.520	1	≠				12.100	6.110	0.000	20.974
FTS (4)	0.338	0.148	0.623	1	≠		≠	13.688	3.760	0.000	19.723
PII (5)	0.508	0.989	0.520	0.142	1			1.322	2.601	−9.859	4.313
GDP (6)	0.488	0.393	0.708	0.459	0.397	1		7.475	2.388	0.000	9.680
CPI (7)	0.561	0.453	0.662	0.325	0.457	0.486	1	2.943	2.163	−3.755	5.828
(E) High-income countries (17) obs. 476
TE (1)	1					≠	≠	5.173	3.693	0.000	9.868
IU (2)	0.615	1					≠	2.409	2.050	−5.689	4.584
MCS (3)	0.663	0.750	1					13.327	3.259	0.000	17.864
FTS (4)	0.389	0.132	0.403	1				13.624	1.230	10.463	16.345
PII (5)	0.623	0.986	0.756	0.117	1			2.427	2.037	−5.689	4.516
GDP (6)	0.375	0.441	0.660	−0.062	0.449	1		9.158	2.081	0.000	11.532
CPI (7)	0.322	0.466	0.558	0.108	0.456	0.401	1	4.000	1.233	0.000	4.824

Note: The symbol shows top to left, left to top, mutual, and ≠ no causal relationship between variables. $\bar{x}$ is the mean, $σ$ standard deviation, $\land$ minimum, and $\lor$ means maximum

Table 5.

Descriptive, correlation, and causality test: Regional base.

Var.	(1)	(2)	(3)	(4)	(5)	(6)	(7)	$\bar{x}$	$σ$	$\land$	$\lor$
(A) Central Asia (5 countries) obs. 140
TE (1)	1							4.484	3.602	0.000	10.730
IU (2)	0.474	1						18.679	25.179	0.000	97.999
MCS (3)	0.641	0.484	1					12.216	5.510	0.000	20.974
FTS (4)	0.357	0.193	0.525	1	≠		≠	13.354	2.959	0.000	19.723
PII (5)	0.509	0.718	0.545	0.168	1			1.248	2.532	−9.859	4.516
GDP (6)	0.462	0.511	0.606	0.335	0.442	1		7.442	2.406	0.000	11.532
CPI (7)	0.500	0.411	0.622	0.248	0.381	0.499	1	3.441	1.804	−5.474	5.828
(B) East Asia (2 countries) obs. 56
TE (1)	1	≠	≠	≠	≠	≠	≠	3.421	3.541	0.000	10.231
IU (2)	0.012	1			≠		≠	0.623	2.851	−8.687	3.994
MCS (3)	0.165	0.456	1	≠				14.102	6.150	0.000	20.974
FTS (4)	−0.115	0.210	0.764	1			≠	15.196	3.580	11.102	19.723
PII (5)	0.028	0.988	0.454	0.189	1		≠	0.675	2.864	−8.687	3.897
GDP (6)	−0.096	0.753	0.575	0.312	0.764	1		7.338	0.955	5.854	9.064
CPI (7)	0.281	0.332	0.846	0.453	0.351	0.455	1	4.028	1.087	0.000	5.148
(C) South East Asia (11 countries) obs. 308
TE (1)	1					≠		5.003	3.782	0.000	10.730
IU (2)	0.478	1		≠	≠			0.926	2.686	0.000	4.499
MCS (3)	0.694	0.423	1			≠		13.059	4.882	−9.859	19.601
FTS (4)	0.475	0.204	0.761	1		≠		12.710	3.664	0.000	17.527
PII (5)	0.481	0.967	0.421	0.243	1			0.823	2.678	−9.859	4.414
GDP (6)	0.364	0.434	0.600	0.494	0.432	1		7.309	2.358	0.000	10.963
CPI (7)	0.538	0.339	0.842	0.583	0.315	0.738	1	3.894	1.311	0.000	5.048
(D) South Asia (8 countries) obs. 224
TE (1)	1					≠		3.628	3.238	0.000	9.940
IU (2)	0.321	1		≠			≠	0.281	2.413	−9.103	4.079
MCS (3)	0.642	0.441	1	≠	≠	≠	≠	11.230	6.650	0.000	20.665
FTS (4)	0.470	−0.009	0.445	1	≠	≠		12.481	3.173	0.000	17.731
PII (5)	0.347	0.973	0.457	−0.008	1			0.350	2.430	−9.103	3.897
GDP (6)	0.363	0.349	0.493	0.137	0.372	1		6.439	1.783	0.000	9.435
CPI (7)	0.556	0.302	0.591	0.278	0.315	0.514	1	3.656	1.538	0.000	5.142
(E): Middle East and North Africa (MENA) (13 countries) obs. 364
TE (1)	1						≠	4.716	3.770	0.000	9.544
IU (2)	0.555	1					≠	1.499	2.313	−7.803	4.584
MCS (3)	0.653	0.580	1				≠	12.604	4.769	0.000	18.417
FTS (4)	0.305	0.099	0.369	1				13.715	1.356	11.430	17.238
PII (5)	0.567	0.975	0.590	0.086	1		≠	1.535	2.301	−7.803	4.516
GDP (6)	0.359	0.342	0.425	−0.185	0.339	1	≠	8.309	2.373	0.000	11.532
CPI (7)	0.291	0.300	0.418	0.140	0.279	0.309	1	3.505	1.743	−3.487	5.828
(F) Europe (24 countries) obs. 672
TE (1)	1							4.945	3.469	0.000	10.528
IU (2)	0.596	1		≠				1.742	2.500	−7.629	4.482
MCS (3)	0.737	0.650	1					12.103	5.480	0.000	19.286
FTS (4)	0.370	0.160	0.567	1	≠		≠	13.597	3.262	0.000	17.634
PII (5)	0.600	0.990	0.656	0.160	1			1.755	2.487	−7.629	4.433
GDP (6)	0.580	0.492	0.751	0.427	0.495	1		7.516	2.593	0.000	10.232
CPI (7)	0.596	0.556	0.703	0.254	0.565	0.597	1	3.363	1.929	−5.474	5.460

Note: The symbol shows top to left, left to top, mutual, and ≠ no causal relationship between variables. $\bar{x}$ is the mean, $σ$ standard deviation, $\land$ minimum, and $\lor$ means maximum. Separate table is made with parallel headings, for each country across its categorical (income-wise and regional-base) distribution.

Granger causality results postulates that tourism expenditure Granger causes net usage at every category mostly from left-to-top (except for lower income countries). This is because of not having reciprocal response from integrated technology. The absence of mutual and top-to-leftward causation supports the tourism-pull factor hypothesis which postulates that surge for ICT sector demand is due to tourism factor.

We experience causation of ICT proxies with tourism expenditure as well as rest of the variables in both tables (Standing et al., 2014). The mutual causation of net usage with rest of ICT proxies is found in each slot (except for 2C and 2E) which contend their joint distributive impact on tourism expenditure. Due to lack of technological integration with tourism industry, we found no variable reciprocal effect towards TE in low-income countries. Macroeconomic indicator like CPI tend to be exogenous in most cases especially 1C and 1F, respectively.

We found robust significant association of TE with ICT proxies at 1F, 1G, and 2E while weak at 1C, 1E, 2B, and 2C, respectively. The negative association experienced at 1C which is against theory but overlooked due to limited cross-sections. The higher variance of net proxy is due to volatility of net investment in each region. However Shapiro–Wilk normality test is used for outlier which is proposed by Royston (1983) and depicted in Figure 2 respectively.

Figure 2.

Normality of probability distribution plot for Tourism Revenue (TR).

CD test postulations

Equation (2) analyzes the data by Pesaran (2004, 2015) using CD test. The results of CD test are pasted below in Table 6, respectively.

Table 6.

CD test and weak CD test for cross-sectional dependencies.

(A) CD test by Pesaran (2004)
Variables	All (65)	Lower income countries (2)	Lower middle-income countries (25)	Upper middle-income countries (21)	High-income countries (17)	H₀
TE	NA	2.66^a	NA	40.35^a	36.98^a	No cross-section dependence (correlation) in residuals
IU	204.08^a	2.30^b	61.74^a	57.20^a	54.75^a
MCS	220.25^a	4.89^a	84.42^a	69.20^a	57.92^a
FTS	137.55^a	−1.34	61.47^a	51.09^a	31.02^a
PII	166.12^a	2.30^b	56.68^a	58.84^a	56.43^a
GDP	166.36^a	4.48^a	54.27^a	55.42^a	48.56^a
CPI	NA	3.84^a	NA	NA	42.99^a
Variables	Central Asia (5)	East Asia (2)	South East Asia (11)	South Asia (8)	Middle East and North Africa (13)	Europe (24)
TE	5.96^a	0.31	24.62^a	17.57^a	28.53^a	59.16^a
IU	11.16^a	3.84^a	18.93^a	17.10^a	35.46^a	72.74^a
MCS	15.90^a	5.08^a	36.51^a	25.56^a	43.44^a	79.91^a
FTS	5.67^a	4.87^a	26.98^a	16.88^a	40.34^a	35.13^a
PII	11.69^a	4.00^a	13.93^a	17.24^a	37.41^a	73.91^a
GDP	14.09^a	4.50^a	33.86^a	26.43^a	24.21^a	66.62^a
CPI	NA	5.21^a	35.05^a	25.11^a	21.60^a	NA
(B) Weak CD test by Pesaran (2015)
Variables	All (65)	Lower income countries (2)	Lower middle-income countries (25)	Upper middle-income countries (21)	High-income countries (17)	H₀
TE	187.944^a	3.763^a	69.814^a	58.259^a	50.322^a	errors are weakly cross-sectional dependent
IU	215.426^a	2.275^b	58.330^a	57.844^a	57.721^a
MCS	229.725^a	5.112^a	88.114^a	72.175^a	61.180^a
FTS	233.835^a	4.727^a	89.428^a	72.400^a	61.691^a
PII	154.685^a	2.282^b	52.505^a	58.817^a	58.378^a
GDP	229.092^a	4.225^a	86.955^a	72.377^a	59.900^a
CPI	196.802^a	3.932^a	77.696^a	54.665^a	57.548^a
Variables	Central Asia (5)	East Asia (2)	South East Asia (11)	South Asia (8)	The Middle East and North Africa (13)	Europe (24)
TE	10.804^a	2.635^a	33.219	22.266^a	36.515^a	76.136^a
IU	10.953^a	3.916^a	17.368	14.903^a	35.218^a	72.271^a
MCS	16.085^a	4.984^a	37.502	26.259^a	44.664^a	84.125^a
FTS	16.608^a	5.288^a	37.302	27.238^a	46.718^a	83.462^a
PII	11.512^a	4.062^a	10.906	14.911^a	36.367^a	73.338^a
GDP	16.484^a	5.277^a	37.203	26.576^a	43.757^a	84.273^a
CPI	9.388^a	5.057^a	36.914	24.943^a	38.822^a	74.155^a

Note: NA values in some slots is due to the data insufficiency issue. There are total four categories of the sample at income-level section and six categories at regional-base section. The whole sample is adjusted in income section by naming it “All.” Due to large panels, Pesaran (2004) and (2015) both are conducted to testify the cross-sectional dependence hypothesis. Pesaran (2015) reveals the consistency of the study variables across time while Pesaran (2004) analyzes the strong CD which reveals real problem in the shape of omitted variable bias (Table 6, panel A). In above Table 6, “a” indicates 99% significance level, “b” indicates 95% significance level while “c” indicates 90% significance level respectively.

Table 6 evidences strong cross-sectional dependence at 0% significance level. This shows robust dependency of the study variables.

From Table 6, the dependency problem misleading our study inferences. To grasp over the scenario, the study mechanizes the bootstrapping approach by designing Westerlund cointegration test with 300 iterations.

Table 7 shows the result of t-test for wide-ranging panels which is applied on CD effected data. To eliminate CD effect from the data, Z[t-bar] value 4.1 is considered normal for ruminating the H₀: non-stationarity hypothesis. From tables, we experience mixed outcomes in circumventing the misrepresentation by linear trend and serial correlations. We use Westerlund ECM panel cointegration test based on residual simulation approach.

Table 7.

Pesaran’s CADF panel unit root tests (second generation).

(A) Income basis
Variables	All (65)	Lower income countries (2)	Lower middle-income countries (25)	Upper middle-income countries (21)	High-income countries (17)
Observations	1803	51	692	583	471
Level (augmented by lags average) Z[t-bar] p value
TE	−7.455^a	0.279	−3.447^a	−4.709^a	−1.633^c
IU	15.741	−0.694	−6.636^a	−6.691^a	−5.154^a
MCS	−8.955^a	−0.744	−5.880^a	−1.921^b	−9.386^a
FTS	−0.360	−1.272	0.040	−1.689^b	0.656
PII	−12.494^a	−0.690	−11.130^a	−10.249^a	−6.686a
GDP	−8.808^a	−0.598	−1.384^b	−7.602^a	−2.133^b
CPI	−1.408^c	−2.724^a	0.493	−1.558^c	−6.337^a
First difference cross-sectional average in the first period extracted and extreme t-values truncated
TE	−24.019^a	−4.514^a	−15.505^a	−16.602^a	−14.531^a
IU	−11.397^a	−4.390^a	−18.457^a	−16.224^a	−12.760^a
MCS	−22.905^a	−4.063^a	−15.201^a	−11.298^a	−10.629^a
FTS	−19.704^a	−0.607	−13.036^a	−9.842^a	−9.194^a
PII	−31.334^a	−4.379^a	−19.336^a	−18.247^a	−14.951^a
GDP	−22.604^a	−3.122^a	−11.889^a	−12.518^a	−12.364^a
CPI	−17.606^a	−1.573^c	−11.723^a	−7.976^a	−10.374^a
(B) Regional basis
Variables	Central Asia (5)	East Asia (2)	South East Asia (11)	South Asia (8)	The Middle East and North Africa (13)	Europe (24)
Obs.	135	52	297	216	351	648
Level (augmented by lags average) Z[t-bar] p value
TE	−0.811	0.127	−1.199	0.026	−1.566^c	−6.547^a
IU	−1.486^c	−2.230^b	−1.121	−3.084^a	−2.662^a	−6.610^a
MCS	−1.108	−0.991	−3.978^a	−2.131^b	−4.077^a	−7.203^a
FTS	−1.422^c	3.850	−0.459	0.106	−0.876	−0.234
PII	−3.340^a	−2.692^a	−0.931	−2.642^a	−6.361^a	−10.986^a
GDP	−3.091^a	−0.557	−0.113	−3.858^a	−3.858^a	−12.319^a
CPI	2.611	−2.801^a	−0.059	−5.284^a	2.399	−9.281^a
First difference cross-sectional average in the first period extracted and extreme t-values truncated
TE	−6.553^a	−5.978^a	−10.185^a	−7.798^a	−14.200^a	−16.763^a
IU	−7.761^a	−3.770^a	−11.722^a	−8.290^a	−10.984^a	−18.944^a
MCS	−6.416^a	1.078	−8.427^a	−7.342^a	−10.487^a	−14.450^a
FTS	−6.011^a	−1.658^b	−6.726^a	−3.768^a	−11.463^a	−9.872^a
PII	−8.308^a	−4.366^a	−12.401^a	−9.181^a	−13.238^a	−19.501^a
GDP	−7.689^a	−5.437^a	−7.801^a	−6.655^a	−9.513^a	−17.411^a
CPI	−1.060^a	−5.137^a	−8.065^a	−6.070^a	−7.239^a	−11.549^a

Note: CADF: Covariate-Augmented Dickey–Fuller. In above Table 7, “a” indicates 99% significance level, “b” indicates 95% significance level while “c” indicates 90% significance level respectively.

Westerlund test outcomes

One of the underlying assumption of cointegration (especially when dealing with social data) is there is equilibrium among variables in the long run. Therefore, if cross-sectional nodes are expected to be correlated, the bootstrapping yield robust critical value. The Westerlund (2007) approach is deemed because of the mix outcomes from previous diagnostics (e.g. CD test, unit-root test). Westerlund test consider H₀ of cointegration like other techniques (e.g. Kao, Pedroni), but H₁ is different as “not necessarily all panels are integrated,” which support our mix outcomes notion. Results in Table 8 reveal significance at all nodes except for lower income countries, East and South Asia, respectively (due to N < T). We design 300 iteration in Westerlund (2007) test to mechanize the CD issue as below:

Table 8.

Westerlund ECM panel cointegration tests.

Variables	Description	All (65)	Lower income countries (2)	Lower middle-income countries (25)	Upper middle-income countries (21)	High-income countries (17)
H₀: No cointegration
G_t	Value	−3.908	−0.180	−3.187	−3.554	−3.043
G_t	Z Value	−11.710^a	3.114	−3.730^a	−5.065^a	−2.491^a
G_a	Value	−9.239	−0.592	−9.776	−8.674	−6.222
G_a	Z Value	4.464	2.249	2.446	2.848	3.774
P_t	Value	−23.916	−0.257	−13.430	−14.251	−9.701
P_t	Z Value	−5.680^a	2.303	−2.347^a	−3.779^a	−0.785
P_a	Value	−7.691	−0.551	−8.009	−8.219	−5.322
P_a	Z Value	2.133	1.527	1.141	0.936	2.206
	Description	Central Asia (5)	East Asia (2)	South East Asia (11)	South Asia (8)	The Middle East and North Africa (13)	Europe (24)
G_t	Value	−5.129	−2.993	−5.045	−2.462	−2.805	−3.416
G_t	Z Value	−5.925^a	−0.786	−8.512^a	−0.098	−1.338^c	−4.752^a
G_a	Value	−22.240	−9.330	−10.693	−6.356	−8.392	−6.718
G_a	Z Value	−2.247^b	0.768	1.258	2.544	2.363	4.193
P_t	Value	−7.431	−4.655	−10.752	−7.402	−10.160	−12.753
P_t	Z Value	−2.244^b	−1.382^c	−3.102^a	−1.165	−2.091^b	−1.960^b
P_a	Value	−7.851	−6.601	−10.022	−6.251	−8.163	−5.816
P_a	Z Value	0.551	0.550	−0.006	1.213	0.759	2.345

Note: ECM: Error correction model.

Due to the variance in the structure of each category, Table 8 postulates dissimilar results regarding G_a and G_t as well as P_a and P_t , respectively. There is no cointegration of the under consideration series of East Asian and South Asian region. The study also accepts the null hypothesis at lower income-level countries at p < 0.001 confidence level which leads us to structural bias and make regression spurious. ECM is deemed for model fitness in the short and long run by using more replicas.

ECM outcomes

ECM is design to calculate stochastic simulation of 5000 to 20000 replications to intuit the minor probability toward equilibrium point (Table 9). The approach for interpretation for regression outcomes is held for short- and long-run fluctuation toward equilibrium point. The coefficients of EC term is statistically significant at p < 0.01 and also negative at all cases except for East Asian and lower income categories (significant at p < 0.05). The result is congruent with Rauf et al. (2018), Cialani (2017), and Dinda and Coondoo (2006) for adjustment parameter significance and negative value in all strewn categories. It shows that in how much speed, tourism expenditure rectify itself to balance with respect to ICT proxies.

Table 9.

Error correction term representative of ARDL model.

Variables	All (65)	Lower income countries (2)	Lower middle-income countries (25)	Upper middle-income countries (21)	High-income countries (17)
ECT_(t−1)	−0.1992^a	−0.1656^b	−0.1830^a	−0.2188^a	−0.2178^a
ΔIU_(t−1)	0.0249^a	0.0640	−0.1028	0.0651	0.0645
ΔMCS_(t−1)	0.2668^a	0.0269	0.2727^a	0.2883^a	0.3763^a
ΔFTS_(t−1)	−0.0166	0.0952	−0.0168	−0.1029^c	1.1910^a
ΔPII_(t−1)	0.1155^a	−0.4996^c	0.0921	0.1838	0.3066
ΔGDP_(t−1)	0.0219	−0.0735	0.1251^c	0.0560	−0.0032
ΔCPI_(t−1)	0.1601^a	0.2788^b	0.2886^a	0.1499^a	−0.0858
Obs.	1819	55	699	587	475
Variables	Central Asia (5)	East Asia (2)	South East Asia (11)	South Asia (8)	Middle East and North Africa (13)	Europe (24)
ECT_(t−1)	−0.1252^a	−0.1728^b	−0.2367^a	−0.2169^a	−0.2210^a	−0.2365^a
ΔIU_(t−1)	−0.0596	−0.0871	0.0416	0.0681	0.0621	0.0447
ΔMCS_(t−1)	0.3210^a	−0.0262	0.3937^a	0.0526^b	0.4121^a	0.2517^a
ΔFTS_(t−1)	−0.0350	−0.0640	−0.0242	0.1603^b	−0.0407	−0.0414^b
ΔPII_(t−1)	0.2305	−0.0726	0.1149	−0.2558	0.0724	0.1389
ΔGDP_(t−1)	0.0723	0.0462	−0.0818	−0.0233	0.2224^b	0.1204^b
ΔCPI_(t−1)	0.1783^a	0.4446	0.0049	0.1694^b	−0.0713	0.2634^a
Obs.	139	55	307	223	363	671

In Table 9, we find that EC term has negative significant value for each slot, which indicates long-run equilibrium change in TE. We find the significant absolute highest value for upper middle-income countries (−0.218) in income-level regions and South East Asia (−0.236) in regional level at p < 0.01 level. The phenomenon of robust significance reveals that explanatory variables need EC term time to restore the equilibrium point if any deviation occurs in the long-run equilibrium of TE in the fore stated economies (Figure 5).

Figure 5.

Residual plots simulations.

{ECT}_{t - 1} = {TE}_{t - 1} - 0.018 Δ {IU}_{t - 1} - 0.301 Δ {MCS}_{t - 1} + 0.083 Δ {FTS}_{t - 1} + 0.075 Δ {PII}_{t - 1} - 0.110 Δ {GDP}_{t - 1} - 0.274 Δ {CPI}_{t - 1}

Δ {TE}_{i, t} = - 0.006 {ECT}_{t - 1} - 0.141 Δ {TE}_{t - 1} + 0.019 Δ {IU}_{t - 1} + 0.020 Δ {MCS}_{t - 1} - 0.010 Δ {FTS}_{t - 1} - 0.018 Δ {PII}_{t - 1} + 0.0002 Δ {GDP}_{t - 1} + 0.040 Δ {CPI}_{t - 1}

The coefficient of cointegration equation is $φ$ = −0.006 for the error correction term which is mention $z_{i, t - 1}$ in the above equation.

Long-run estimation outcomes

From equation (6), we calculated the below Table 10 which expose the tourism expenditure association with ICT and macroeconomic proxies in long run only. We find that 1 unit increase Internet usage will effect TE by 4.73 (Europe), 8.74 (South Asia) and 4.48 (MENA) units in the long run (in case of Europe).

Table 10.

Estimated long-run coefficients using the ARDL approach.

Variables	All (65)	Lower income countries (2)	Lower middle-income countries (25)	Upper middle-income countries (21)	High-income countries (17)
IU	0.0181	14.4266^b	−0.5617	2.0785	4.7305^b
MCS	0.3015^a	0.1627	0.3894^a	0.1572	0.8535^a
FTS	−0.0837	−0.0933	−0.0918	−0.0846	−0.3393^c
PII	−0.0753	−15.4055^b	0.5030	−2.1014	−4.6110^b
GDP	0.1101	−0.4441	−0.1401	0.2562	−0.0150
CPI	0.2741^c	1.6835^b	0.4846^c	0.6850^a	−0.3939
Obs.	1819	55	699	587	475
Variables	Central Asia (5)	East Asia (2)	South East Asia (11)	South Asia (8)	The Middle East and North Africa (13)	Europe (24)
IU	−0.4766	−0.5041	0.1759	8.7453^a	4.4825^a	0.1892
MCS	0.0030	−0.1518	0.6664^a	0.2429^b	0.4445^a	0.3922^a
FTS	−0.2796	−0.3706	−0.1024	−0.0409	−0.1841	−0.1751^b
PII	0.2073	−0.4204	−0.0074	−8.9580^a	−4.3065^a	−0.2624
GDP	0.5773	0.2678	−0.3457	−0.1075	0.3280^c	0.1784
CPI	1.4238^a	2.5730	0.0211	0.7811^b	−0.3226	0.4360^b
Obs.	139	55	307	223	363	671

From the above Table 10, we experience MCS, FTS, and PII significance in developed countries because of proper infrastructure and their integration in the tourism related operations. The negative association of FTS and PII with tourism expenditure support the study hypothesis by inverse association.

Short-run ARDL estimates based on White’s Heteroscedasticity adjusted SEs

Short-run volatility w.r.t long-run equilibrium point based on White’s Heteroscedasticity adjusted Standard Errors is testified and depicted in Table 12. The heteroscedasticity of every categorical data shows its variable range volatility which regresses the dependent variable. Except for lower income, Central Asia, East Asia, and South Asia region, the rest of the categories are facing with the heteroscedastic issue. However, the phenomenon is tackled by heteroscedasticity consistent variance-covariance estimators.

Table 12 exposes less short-run association of TE with Internet only in All, lower middle-income and Europe region, respectively. However, the mobile user seems to be more significant in the short run in most of the region as depicted in Table 12 (Dinda and Coondoo, 2006). The All and European region seem to be more oriented in supporting the study null hypothesis (as experienced from its variables significance level).

On the basis of upper and lower bound, (by F and W statistics), we are able to decide the null hypothesis consideration of variables association in the long run. The critical bound values are computed by stochastic simulations using 5000 to 20000 replications.

Our analysis (in Table 11) reveals that only Central Asia and East Asia has F value less than lower bound critical values, which is an indication for the null hypothesis of no level effect can’t be rejected. Rest of the categories (including All) reveals F and W values more than upper bound critical value which reject the no level effect null hypothesis. The analysis postulates the existence of a long-run relationship of tourism expenditure with study ICT proxies.

Table 11.

Testing for existence of a level relationship among the variables in the ARDL model.

Variables	All (65)		Lower income countries (2)				Lower middle-income Countries (25)				Upper middle-income countries (21)				High-income countries (17)
F-statistic	29.92		7.72				10.83				11.23				9.97
W-statistic	209.49		54.06				75.84				78.64				69.80
	F-stat	W-stat	F-stat		W-stat		F-stat		W-stat		F-stat		W-stat		F-stat		W-stat
95% lower bound	2.04	14.34	2.22		15.55		2.06		14.42		2.03		14.27		2.04		14.33
95% upper bound	3.26	22.86	3.55		24.86		3.28		22.99		3.27		22.93		3.24		22.68
Variables	Central Asia (5)		East Asia (2)				South East Asia (11)				South Asia (8)				The Middle East and North Africa (13)				Europe (24)
F-statistic	1.78		1.43				6.14				15.57				9.02				13.68
W-statistic	12.48		10.01				43.02				109.03				63.18				95.78
	F-stat	W-stat	F-stat	W-stat		F-stat		W-stat		F-stat		W-stat		F-stat		W-stat		F-stat		W-stat
95% lower bound	2.12	14.84	2.22	15.55		2.09		14.66		2.08		14.59		2.07		14.52		2.01		14.09
95% upper bound	3.32	23.30	3.55	24.86		3.24		22.73		3.31		23.21		3.27		22.95		3.20		22.45

Note: The F and W statistics are analyzed in lower and upper bound critical values at 95% significance level. If the F-value less than lower bound critical values, it is an indication for the null hypothesis of no level effect cannot be rejected.

Table 12.

Short-run ARDL estimates based on White’s heteroscedasticity adjusted SEs.

Variables	All (65)	Lower income countries (2)	Lower middle-income countries (25)	Upper middle-income countries (21)	High-income countries (17)
TE(−1)	0.8007^a	0.8343^a	0.8169^a	0.7811^a	0.7821^a
IU	0.0249^a	0.0640	−0.1028^b	0.0651	0.0645
MCS	0.2668^a	0.0269	0.2727^a	0.2883^a	0.3763^a
FTS	−0.0166^c	0.0952^b	−0.0168	−0.1029	1.1910^a
PII	0.1155^a	−0.4996^c	0.0921^b	0.1838^c	0.3066
GDP	0.0219	−0.0735	0.1251^b	0.0560	−0.0032
CPI	0.1601^b	0.2788^b	0.2886^a	0.1499^a	−0.0858
Heteroscedasticity	28.5105^a	1.7799	6.9390^a	4.7477^b	24.7790^a
Obs.	1819	55	699	587	475
Variables	Central Asia (5)	East Asia (2)	South East Asia (11)	South Asia (8)	The Middle East and North Africa (13)	Europe (24)
TE(−1)	0.8747^a	0.8271^a	0.7632^a	0.7831^a	0.7789^a	0.7634^a
IU	−0.0596	−0.0871	0.0416	0.0681	0.0621	0.0447^c
MCS	0.3210^a	−0.0262	0.3937^a	0.0526^b	0.4121^a	0.2517^a
FTS	−0.0350	−0.0640	−0.0242	0.1603^a	−0.0407^c	−0.0414^a
PII	0.2305^c	−0.0726	0.1149	−0.2558	0.0724	0.1389^b
GDP	0.0723	0.0462	−0.0818	−0.0233	0.2224^c	0.1204^b
CPI	0.1783^b	0.4446	0.0049	0.1694^a	−0.0713	0.2634^b
Heteroscedasticity	1.1758	1.6013	6.4773^b	.0845	12.2857^a	22.1517^a
Obs.	139	55	307	223	363	671

Note: In above Table 12, “a” indicates 99% significance level, “b” indicates 95% significance level while “c” indicates 90% significance level respectively.

Robustness check

Results for DOLS and FMOLS at Tables 13 and 14 also ratify the robustness of ARDL long-run coefficients. Furthermore, the impulse response function (Figure 6) support the hypothesis by elucidating the negative association between Internet usage and tourism expenditure and vise versa.

Table 13.

Dynamic ordinary least square (DOLS).

Variables	All (65)	Lower income countries (2)	Lower middle- income countries (25)	Upper middle-income countries (21)	High-income countries (17)
IU	0.0336^a	−0.6324^c	−0.8322	2.8895	−42.6275
MCS	0.3039^a	0.1165	0.4447^a	0.3967^b	−0.0650
FTS	0.2009	0.3860^a	0.3141	−0.1916	5.1095^b
PII	−0.0569	0.2459	0.8192	−2.6439	42.6724
GDP	−0.1101	−0.3714^b	0.1312	−0.2572	−0.8504
CPI	0.1228	0.9828^a	−0.4759	0.3774	0.9503^c
Adj. R ²	0.8103	0.8491	0.8528	0.7785	0.7298
Obs.	1525	56	600	450	425
Variables	Central Asia (5)	East Asia (2)	South East Asia (11)	South Asia (8)	Middle East and North Africa (13)	Europe (24)
IU	6.0395	252.5023	−0.4528	−0.2845	4946.198^a	2.8449
MCS	0.0600	−6.4825^c	1.0781^b	0.1553	0.7036^a	0.6122^a
FTS	−4.9875	5.7516	−0.0546	−0.0530	−0.0580	−0.1214
PII	−6.8737	−250.7636	0.4715	−0.2471	−4946.371^a	−2.9017
GDP	3.7752^a	−9.1621^c	−0.1669	2.2143	−0.5375	0.0808
CPI	−0.4679	43.8178^c	−1.6265	2.1162^a	0.5039^b	−0.6734
Adj. R ²	0.8876	0.7590	0.8298	0.7552	0.7848	0.7858
Obs.	100	50	275	200	325	550

Note:Note: In above Table 13, “a” indicates 99% significance level, “b” indicates 95% significance level while “c” indicates 90% significance level respectively.

Table 14.

Fully modified ordinary least square (FMOLS).

Variables	All (65)	Lower income countries (2)	Lower middle-income countries (25)	Upper middle-income countries (21)	High-income countries (17)
IU	0.0095^c	−0.8650^b	0.0132	−0.0716	−0.2325
MCS	0.2167^a	0.2239^a	0.2456^a	0.2468^a	0.2515
FTS	0.0951	0.4196^a	0.1395	−0.0023	3.5345^a
PII	0.2995^a	0.3847	0.2340	0.3415	0.6900
GDP	0.0151	−0.5073^a	0.2247^b	−0.0928	0.0409
CPI	0.3271^a	1.1156^a	0.1352	0.4950^a	−0.0852
Adj. R ²	0.5971	0.8315	0.6707	0.5646	0.6247
Obs.	1674	54	648	513	459
Variables	Central Asia (5)	East Asia (2)	South East Asia (11)	South Asia (8)	The Middle East and North Africa (13)	Europe (24)
IU	−0.0373	−0.1900	−0.1308	−0.2757	−0.3344	0.1908
MCS	−0.1237	0.1463	0.7303^a	0.1729^b	0.1120	0.2242^a
FTS	−0.0070	1.8771	−0.1421	0.3868^a	1.6425^b	0.0027
PII	0.3649	0.2879	0.4917	0.3125	0.7593^b	0.1650
GDP	0.3471	−2.6978^b	−0.1684	−0.2610	0.5631^a	0.0155
CPI	1.1577^a	1.1529	−0.5597	0.8402^a	0.0405	0.4470^a
Adj. R ²	0.6251	0.2446	0.6099	0.5826	0.6356	0.6195
Obs.	108	54	297	216	351	621

Note: In above Table 14, “a” indicates 99% significance level, “b” indicates 95% significance level while “c” indicates 90% significance level respectively.

Figure 6.

Impulse response function (IV response to TE).

CUSUM/CUSUM2 graphs are visualized using equation (8) and adjusted in the below frame of Figure 7, respectively.

Data reliability is finally observed in above Figure 7 which exposes fluctuation of variables in a random pattern which is balanced to 0. The in-between data line depicts the stability of TE in every category except for low-income countries (due to less cross-sections—only two countries).

Conclusion and limitations

Various countries of the world integrated the use of the Internet in their operations and benefited from its optimization, economy, and reliability. With respect to BRI listed nations (65 countries), the study tends to add in the literature that how well-structured ICT proxies play its role in attracting tourist and reshaping tourism expenditures. Income and Region based BRI listed nations are selected (Table 1 and Table 2) as an analysis sample because we want to testify how mega projects play an intervening role in the technological emergency. The study also took recent and long duration data (1990–2017) and employ advanced statistical software (MATLAB and R) to enlighten the economic impact of ICT proxies on tourist expenditures in the long and short run.

For reliability issue (due to long-rang data), we execute diagnostic tests, for example, CD test, unit root, heteroscedasticity, cointegration test and also depict notions by graphs like CUSUM graph, residual and Fitness Plot (Figure 4) overlapped boxplot and normal probability plot. The long-run coefficient is noted after adjusting the heteroscedasticity variance-covariance issue and later confirmed by DOLS and FMOLS for robustness check. The speed of adjustment by error correction term is finally calculated to explain the results with the study hypothesis.

Figure 4.

Fitted values curves.

We find ICT proxies Granger causing tourism expenditure from one side in developing and mutual in developed economies. The reason is reciprocal feedback from technological integration with tourism related operations which is significant factor for prospective inbound tourists.

After authenticating data by diagnostic tests, we found (in Table 10) long-run significant association (which is cross-checked in Table 11 by F and W statistics upper and lower bound values) between ICT proxies with tourism expenditure in developed economies. We found that IU, MCS, FTS, and PII are significant in case of Europe, MENA, South Asia, and high income nodes. This support the study hypothesis that investment on ICT structure can curb on tourism expenditure in the long run. The lower income countries show insignificance because of their uncertain system and less investment on tourism sector.

Moreover, Table 12 reveals short-run associations which are mostly significant in almost all nodes of the study. This implies that low income or underdeveloped economies are get influenced by ICT proxies, but only in the short run. The short-run significance is also approved in Tables 4 and 5 by depicting one-way causation. Which means that once tourist suffered from manual operations and unnecessary expenditures in underdeveloped economies, they avoid back to cross such boundaries.

Although the study findings are inspiring, but there are also some areas which are deemed to be the limitation of on-hand research. First of all, it is argued that Internet is general purpose technology and it requires a huge investment in necessary assets like electric power consumption, installation of broadband networks or advances computer systems. Shahiduzzaman and Alam (2014) reveal that the expected return from the Internet is not sufficient as much as it’s in the long run until the fore stated investment is made on the fore mentioned areas.

Secondly, high tariff on advanced industrial technology equipment is also the barrier of low-level or underdeveloped economies. Finally, data for low-income countries are not available which also make the regression spurious in some cases.

Policy implications

The growth of Internet implications in economic operations is experienced in horizontal level economies. The study also addresses low-income countries having a small size of tourism revenue 466 compare to high-income economies which is 13,082 in 2017. Our study suggest that underdeveloped economies need to address the ICT sector upgradation and integration with tourism related product and services.

The operational expenditure related to inbound tourist is minified by curtailing agent involvement and enabling tourist to directly access to spot basic information. The protection of Internet private data become an area which need to be addressed before governing the tourism sector by ICT. Governments should train HR properly to handle the fore stated phenomenon robustly and systematic manner.

As the mega project seems to cross the listed countries, so these nations need to update the ICT sector by communication and dissemination of information for entering into a new era. Especially, the low-income countries not only need to invest more in updating the ICT infrastructure but also invest in educating their public with advanced usage of technologies and endorsing Internet usage skills.

Footnotes

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: The research work is leveraged by National Social Sciences Fund of China (18VSJ035) and National Natural Science Foundation of China (No. 71673043).

ORCID iDs

Obaid Ur Rehman

Abdul Rauf

References

Abou-Shouk

Lim

Megicks

(2013) Internet adoption by travel agents: a case of Egypt. International Journal of Tourism Research 15(3): 298–312.

Aguilo

Rossello

Vila

(2017) Length of stay and daily tourist expenditure: a joint analysis. Tourism Management Perspectives 21: 10–17.

Aly Song/Reuters (2017) China’s Politburo Standing Committee member Wang Huning attends the opening ceremony of the fourth World Internet Conference in Wuzhen, Zhejiang province, China, 3 December 2017.

Antti Pesonen

(2013) Information and communications technology and market segmentation in tourism: a review. Tourism Review 68(2): 14–30.

Assaf

Tsionas

(2018) The estimation and decomposition of tourism productivity. Tourism Management 65: 131–142.

Benhabib

Spiegel

(2005) Human capital and technology diffusion. In: Aghion

Durlauf

(eds), Handbook of Economic Growth, vol. 1A. Amsterdam: Elsevier, pp. 935–966.

Bethapudi

(2013) The role of ICT in tourism industry. Journal of Applied Economics and Business 1(4): 67–79.

Bethapudi

(2015) Role of ICT in promoting a rural tourism product. Journal of Tourism and Hospitality 4(3): 154.

Brida

Scuderi

(2013) Determinants of tourist expenditure: a review of microeconometric models. Tourism Management Perspectives 6: 28–40.

10.

Brida

Moreno-Izquierdo

Zapata-Aguirre

(2016) Customer perception of service quality: the role of Information and Communication Technologies (ICTs) at airport functional areas. Tourism Management Perspectives 20: 209–216.

11.

Buhalis

Law

(2008) Progress in tourism management: twenty years on and 10 years after the Internet: the state of eTourism research. Tourism Management 29(4): 609–623.

12.

Butler

(ed) (2006) The Tourism Area Life Cycle, vol. 1. Bristol: Channel view publications.

13.

Chesher

Jewitt

(1987) The bias of heteroscedasticity-consistent covariance matrix estimator. Econometrica 55: 1217–1222.

14.

Choi

(2009) The effect of the internet on economic growth: evidence from cross-country panel data. Economics Letters 105(1): 39–41.

15.

Choi

Murray

(2010) Resident attitudes toward sustainable community tourism. Journal of Sustainable Tourism 18(4): 575–594

16.

Cialani

(2017) CO₂ emissions, GDP and trade: a panel cointegration approach. International Journal of Sustainable Development & World Ecology 24(3): 193–204.

17.

Czernich

(2014) Does broadband internet reduce the unemployment rate? Evidence for Germany. Information Economics and Policy 29: 32–45.

18.

Da Xu

(2016) An internet-of-things initiative for one belt one road (OBOR). Frontiers of Engineering Management 3(3): 206–223.

19.

Del Chiappa

Baggio

(2015) Knowledge transfer in smart tourism destinations: analyzing the effects of a network structure. Journal of Destination Marketing & Management 4(3): 145–150.

20.

Devesa

Laguna

Palacios

(2010) The role of motivation in visitor satisfaction: empirical evidence in rural tourism. Tourism Management 31(4): 547–552.

21.

Dinda

Coondoo

(2006) Income and emissions: a panel based cointegration analysis. Ecological Economics 57: 167–181.

22.

Divisekera

Nguyen

(2018) Determinants of innovation in tourism evidence from Australia. Tourism Management 67: 157–167.

23.

Duncan

(2009) The internet effects on tourism industry. SSRN Electronic Journal. Available at: 10.2139/ssrn.1403087.

24.

Filieri

McLeay

(2014) E-WOM and accommodation: an analysis of the factors that influence travelers’ adoption of information from online reviews. Journal of Travel Research 53(1): 44–57.

25.

Granger

(1969) Investigating causal relations by econometric models and cross-spectral methods. Econometrica 37: 424–438.

26.

Gruber

Hatonen

Koutroumpis

(2014) Broadband access in the EU: an assessment of future economic benefits. Telecommunications Policy 18: 1046–1058.

27.

Howard

Hussain

(2011) The role of digital media. Journal of Democracy 22(3): 35–48.

28.

Huang

Backman

, et al. (2016) Exploring the implications of virtual reality technology in tourism marketing: an integrated research framework. International Journal of Tourism Research 18(2): 116–128.

29.

Ishida

(2015) The effect of ICT development on economic growth and energy consumption in Japan. Telematics Inform 32: 79–88.

30.

Jorgenson

Stiroh

(2008) A retrospective look at the US productivity resurgence. Journal of Economic Perspectives 22: 3–24.

31.

Koenker

(1981) A note on studentizing a test for heteroscedasticity. Journal of Econometrics 17: 107–112.

32.

Kozak

Rimmington

(2000) Tourist satisfaction with Mallorca, Spain, as an off-season holiday destination. Journal of travel research 38(3): 260–269.

33.

Laimer

Weiß

Austria

(2006, July 19) Data sources on tourism expenditure: The Austrian experiences taking into account the TBoP requirements. In: International workshop on tourism statistics. UNWTO Headquarters, Madrid/Spain. Available at: http://unstats.un.org/unsd/newsletter/unsd_workshops/tourism/IWTS/Invited%20presentations/IWTS_Item17(Austria).pdf.

34.

Law

Leung

Buhalis

(2009) Information technology applications in hospitality and tourism: a review of publications from 2005 to 2007. Journal of Travel & Tourism Marketing 26(5): 599–623.

35.

Law

Buhalis

(2010) Progress in tourism management: a review of website evaluation in tourism research. Tourism Management 31: 297–313.

36.

Lechman

Marszk

(2015) ICT technologies and financial innovations: the case of exchange-traded funds in Brazil, Japan, Mexico, South Korea and the United States. Technological Forecasting and Social Change 99: 355–376.

37.

Lee

Jan

(2019) Can community-based tourism contribute to sustainable development? Evidence from residents’ perceptions of the sustainability. Tourism Management 70: 368–380.

38.

Stepchenkova

(2015) User-generated content as a research mode in tourism and hospitality applications: topics, methods, and software. Journal of Hospitality Marketing & Management 24(2): 119–154.

39.

Marcussen

(2009) Trends in European Internet distribution of travel and tourism services 1998–2008. Centre for Regional and Tourism Research. Available at: http://www.crt.dk/uk/staff/chm/trends.htm (accessed 30 March 2009)

40.

Mariani

Baggio

Buhalis

, et al. (eds) (2014) Tourism management, marketing, and development. The Importance of Networks and ICTs, vol. 1. Berlin: Springer.

41.

Mihajlović

(2012) The impact of information and communication technology (ICT) as a key factor of tourism development on the role of Croatian travel agencies. International Journal of Business and Social Science 3(24): 151–159.

42.

Myung

Changkyu

(2005) The effect of the internet on inflation: Panel data evidence. Journal of Policy Modeling 27(7): 885–889.

43.

Navío-Marco

Ruiz-Gómez

Sevilla-Sevilla

(2018) Progress in information technology and tourism management: 30 years on and 20 years after the internet—Revisiting Buhalis & Law’s landmark study about eTourism. Tourism Management 69: 460–470.

44.

Noh

Yoo

(2008) Internet, inequality, and growth. Journal of Policy Modeling 30: 1005–1016.

45.

Olya

Mehran

(2017) Modelling tourism expenditure using complexity theory. Journal of Business Research 75: 147–158.

46.

Pesaran

(2003) A simple panel unit root test in the presence of cross section dependence. Cambridge Working Papers in Economics 0346, Faculty of Economics (DAE), University of Cambridge.

47.

Pesaran

(2004) General diagnostic tests for cross-section dependence in panels. Faculty of Eco, CESifo Working Paper, No. 1229; 1–39.

48.

Pesaran

(2015) Testing weak cross-sectional dependence in large panels. Economic Review 34(6–10): 1089–1117.

49.

Pretz

(2013) The next evolution of the internet. IEEE Magazine The institute 50(5).

50.

Rauf

Liu

Amin

, et al. (2018) Testing EKC hypothesis with energy and sustainable development challenges: a fresh evidence from Belt and Road Initiative economies. Environmental Science and Pollution Research 25(32): 32066–32080.

51.

Royston

(1983) Some techniques for assessing multivariate normality based on the Shapiro-Wilk W. Applied Statistics 32: 121–133.

52.

Shahbaz

Hye

Tiwary

(2013) The effects of financial development, economic growth, coal consumption and trade openness in South Africa. Energy Policy 61: 1452–1459.

53.

Shahiduzzaman

Alam

(2014) Information technology and its changing roles to economic growth and productivity in Australia. Telecommunications Policy 38: 125–135.

54.

Sigala

(2015) The application and impact of gamification funware on trip planning and experiences: The case of TripAdvisor’s funware. Electronic Markets 25(3): 189–209.

55.

Smolčić

Soldić

(2017) Satisfaction as a determinant of tourist expenditure. Current Issues in Tourism 20(7): 691–704.

56.

Standing

Tang-Taye

Boyer

(2014) The impact of the Internet in travel and tourism: A research review 2001–2010. Journal of Travel & Tourism Marketing 31(1): 82–113.

57.

Tsokota

Von Solms

van Greunen

(2017) An ICT strategy for the sustainable development of the tourism sector in a developing country: a case study of Zimbabwe. The Electronic Journal of Information Systems in Developing Countries 78(1): 1–20.

58.

Utz

(2017) Assessment of the potential impact of the ICT revolution in the Pacific on economic growth. Employment, and Government Revenue. [Online] Pubdocs.worldbank.org. Available at: http://pubdocs.worldbank.org (accessed 28 June 2017).

59.

Wahab

Setiawan

Wahdiniwaty

(2017) Information distribution service of tourism and creative industry using mobile application technology. International Journal of New Media Technology 4(2): 120–125.

60.

Westerlund

(2007) Testing for error correction in panel data. Oxford Bulletin of Economics and Statistics 69(6): 709–748.

61.

Westerlund

(2008) Panel cointegration tests of the fisher effect. Journal of Applied Econometrics 23(2): 193–233.

62.

World Tourism Organization (15 January 2018); ISBN: 978-92-844-1986-9. Available at: www.unwto.org/

63.

World Travel & Tourism Council (November, 2018) Travel & Tourism Economic Impact 2018 World Report. Available at: http://www.wttc.org/

64.

Yin

(2014) Case Study Research: Design and Methods. 5th ed. Thousand Oaks: Sage.