Economic Policy Uncertainty and International Tourism Demand: A Global Vector Autoregressive Approach

Abstract

In a globalized environment, a major shock or event can reverberate across tourism interdependent countries over time. This paper aims to empirically examine how economic policy uncertainty (EPU) emanating from a large country may influence inbound and outbound tourism in other countries. Using Global Vector Autoregressive (GVAR) model and two alternative measures of EPU, the study finds the modeled effects of EPU are dependent on the source of EPU measures, level of internationalization, and type of trade-weights used. Notably, the results show the “decline-rebound-overshoot” effect of EPU shock on tourism demand, and capture the link between capital flight and outbound travel from China in times of surge in EPU. This research highlights that GVAR is able to capture previously unobserved empirical insights because assessing the international impact of shocks such as economic crises, pandemics, and political instability necessitates accounting for cross-sectional interdependence in tourism flows between many countries.

Keywords

global vector autoregressive (GVAR) model economic policy uncertainty (EPU)tourism demand interdependence globalization China

Introduction

Tourism is vulnerable to uncertainty arising from natural disasters (typhoons, volcanic eruptions, earthquakes, etc.) (Rosselló, Becken, and Santana-Gallego 2020); outbreak of infectious diseases (COVID, SARS, avian flu, etc.) (Farzanegan et al. 2021); security threats (terrorism, crimes, and corruptions) (Fourie, Rosselló-Nadal, and Santana-Gallego 2020); geopolitical risks (Gozgor et al. 2021; Tiwari, Das, and Dutta 2019); and economic/financial crises (Khalid, Okafor, and Shafiullah 2020; Lee and Chen 2021). The tourism sector has a high level of requirement on the stability of institutional environment, which consists of a set of rules, laws, regulations, customs, practices, and procedures that shape people’s behavior and decision making within the economic system (Lee and Chen 2021). Risk, uncertainty, and instability of the destination jeopardize the institutional environment, on which tourism development relies heavily.

Uncertainty about fiscal, regulatory, and monetary policies contributes to steep economic downturn and slow recovery afterward (Baker, Bloom, and Davis 2016). “The uncertainties in economic policies include uncertainty in the decisions of economic policy makers, which influence the decisions on economic units such as consumption, investment, saving, and lending. Accordingly, the whole economy may be negatively influenced by the uncertainty in policies” (Wu and Wu 2020, 1). In the measures identified by Altig et al. (2020), such as the newspaper-based policy uncertainty, they have captured a noticeable increase in uncertainty during the coronavirus disease pandemic (COVID) and its subsequent economic crisis. Apparently, COVID has triggered an economic policy uncertainty (EPU) surge with size larger than the 2008/09 Global Financial Crisis and similar to the 1930s Great Depression (Baker et al. 2020).

EPU is relevant to the study of tourism demand. When countries implement economic policy measures to counteract the outbreak of diseases, natural disasters, wars, terrorism or economic downturn, the outcome is uncertain and inbound tourists may respond by avoiding these destinations. To investigate the global or regional influences of EPU on tourism demand, the well-established vector autoregressive (VAR) model is a suitable starting point. This method allows for examining the dynamic relationships between variables of research interest without being too restrictively underpinned by economic theory of demand. The well-known impulse response function under the VAR model is commonly used to demonstrate the temporal evolutions of variables in response to a scenario (the EPU shock in our study). This provides a tool to analyze the impact of a simulated shock using a model developed based on historical data. However, VAR has limitations in handling panel data with large cross-sections as the number of parameters to be estimated grows exponentially with increase in the cross-section sample (Assaf et al. 2019; Cao, Li, and Song 2017).

A natural extension of the VAR for large cross-sections is the Global Vector Autoregressive (GVAR) modeling approach. Specifically, the method harnesses the country-to-country (inter)dependence in the model error terms. Essentially, GVAR can be used to break down the “global” system into a number of “local” models for individual estimations before they are stacked again to form the “global” system. The usefulness of GVAR has been demonstrated in examining the impact of economic variables such as income and price on tourism demand (Assaf et al. 2019; Cao, Li, and Song 2017). GVAR approach is used by governments and central banks to investigate the global or regional impacts of a specific counterfactual shock (e.g., Bussiere, Chudik, and Sestieri 2009; Inoue, Kaya, and Ohshige 2015). However, despite its natural fit (as tourism demand between countries are interrelated in temporal and spatial dimensions), to the best of the authors’ knowledge, there has been no prior study that uses GVAR to empirically examine how EPU (and its shock) affects tourism demand.

Against this background, the paper aims to empirically examine how an EPU shock emanating from a country affects the tourism demand to/from other countries. More specifically, using a GVAR model covering 31 countries across a 20-year period, the research examines to what extent the transmissions of the shock propagate across time and how these effects vary across countries. In doing so, this study demonstrates how changing cross-sectional samples (the implicit assumptions about the set of countries the uncertainty shock transmits) influence the estimated tourism impact of EPU. This paper constructs a model to simulate the influences of country-specific EPU shock using the index of economic policy uncertainty (EPU) developed by Baker, Bloom, and Davis (2016) to investigate how the policy-related economic uncertainty affects the tourism markets regionally and internationally.

The EPU index is measured by an automated search for articles pertaining to “economic AND policy AND uncertainty” through a series of relevant keywords in certain leading newspapers. According to Baker, Bloom, and Davis (2016, 4–5), the newspaper-based “EPU aims to capture uncertainty about who will make economic policy decision, what economic policy actions will be undertaken and when, and the economic effects of policy actions (or inaction) including uncertainties related to the economic ramifications of ‘non-economic’ policy matters, e.g., military actions.” EPU is a forward looking (ex-ante) measure which “reflects uncertainty in the minds of consumers, traders, managers, and policymakers about possible futures, and covers events like terrorism, natural disasters, wars, and climate change” (Rogers and Xu 2019, 1). In the past decade, movements in economic uncertainty closely tracked movements in policy-related uncertainty and the countercyclical nature of EPU is empirically observed (Baker, Bloom, and Davis 2012).

The study focuses on the EPU shock from Mainland China (hereafter China) as a scenario-based simulated shock and examines how it impacts the tourism markets of other countries. China is recognized as an important research subject in GVAR studies (e.g., Cao, Li, and Song 2017; Cashin, Mohaddes, and Raissi 2017; Han, Qi, and Yin 2016; Sznajderska 2019; Tam 2018). Empirical evidence has demonstrated that any hard shock from China can easily and rapidly impact the world, particularly the emerging economies (Sznajderska 2019). For instance, Sznajderska (2019) (Cashin, Mohaddes, and Raissi 2017) finds that a 1% negative shock to China’s GDP leads to 0.22% (0.23%) fall of global growth during 1995–2016 (1981–2013). In tourism trade, China is arguably the most significant source market in global tourism. Chinese visitors account for 20% of global tourism spending in 2019 (United Nations World Tourism Organization [UNWTO] 2020). Between 2009 and 2018, Chinese outbound visitors grew more than three times, while total spending grew more than six times (UNWTO 2019).

However, while the impact of China’s EPU on merchandise trade has been examined (e.g., Tam 2018), the literature is yet to empirically investigate the transmissive impact of China’s EPU on tourism demand across countries using GVAR. Due to its capacity to analyze cross-sectional transmissive impacts of shocks, as will be seen, GVAR approach creates the opportunities to reveal previously unattainable empirical insights to the patterns of tourism recovery following a shock. Thus, although using pre-COVID data, the paper is expected to be of relevance to the post-COVID tourism setting from both the methodological and managerial perspectives.

The rest of the paper is structured as follows. In the next section, we review the extant literature on the EPU-economic activity nexus and the EPU-tourism demand nexus. The importance of accounting for interdependence in tourism demand is illustrated and the appropriateness of adopting GVAR model is discussed in this section. In the third section, we describe the application procedures of the GVAR model to the internationalized and regionalized systems. The ensuing sections present the results and discussions before concluding.

Literature Review

EPU and Economic Activities

Identifying the “first mover” from the empirical correlation between the policy-related economic uncertainty and economic activity (including tourism) is extremely challenging because the causation can be mutual (Baker, Bloom, and Davis 2012; Cesa-Bianchi, Pesaran, and Rebucci 2017). On one hand, high uncertainty is associated with high probability of left-tail default and thus high borrowing cost, which reduces the economic growth. Also, people prefer to adopt the “wait and see” strategy, delaying investment/consumption plans due to high uncertainty in the outcomes. Additionally, high uncertainty makes people become more cautious and sluggish to stimulus policy, thus delaying the recovery (Baker, Bloom, and Davis 2012; Barrero and Bloom 2020). On the other hand, it has also been theoretically and empirically demonstrated that spike in uncertainty may be a response to, rather than a driver of, adverse economic conditions (Cesa-Bianchi, Pesaran, and Rebucci 2017).

Nonetheless, substantial empirical evidence supports the effect of EPU on economic activity as well. Baker, Bloom, and Davis (2016) have empirically investigated a panel dataset across 12 main countries worldwide using a VAR model and demonstrated the persistently negative effects of EPU shock on GDP, industrial production, investment, and employment. Additionally, a number of studies investigating the EPU-economic activity nexus through the GVAR model (e.g., Han, Qi, and Yin 2016; Tam 2018; Trung 2019) observe findings generally in line with Baker, Bloom, and Davis (2016). The impulse response graphs in their GVAR model demonstrate the countercyclical nature of EPU, of which the shock can significantly influence the economic activities for at least a few years. Importantly, all the above mentioned GVAR studies have empirically highlighted how impact of EPU shock transmits across countries; that is, how EPU shock in one country affects economic activities of another country.

In the literature examining the effect of EPU shock on economic activities, a number of transmission channels have been identified: (i) exchange rate channel (Abid and Rault 2020; Tam 2018); (ii) merchandise trade channel (Alam and Istiak 2020; Caggiano, Castelnuovo, and Figueres 2020; Han, Qi, and Yin 2016; Luk et al. 2020; Trung 2019); (iii) financial market channel (Luk et al. 2020; Prüser and Schlösser 2020; Trung 2019); (iv) consumption channel (Prüser and Schlösser 2020); (v) investment channel (Alam and Istiak 2020; Luk et al. 2020; Prüser and Schlösser 2020); (vi) capital flow channel (Trung 2019); and (vii) employment channel (Luk et al. 2020). As inbound and outbound tourism demand are significantly determined by the economic fundamentals of the origin and destination countries, it is reasonable to hypothesize that tourism demand between countries are interdependent. When considering EPU, this is important because EPU shock propagation depends on the level of interdependence.

EPU and Tourism Demand

The studies which examine the impact of EPU on tourism demand are provided in Table 1. Empirical evidence demonstrates the negative effects of destinations’ EPU on inbound tourists (e.g., Akadiri, Alola, and Uzuner 2020; Işık, Sirakaya-Turk, and Ongan 2020), as well as the tourism generating regions’ EPU on outbound tourists (e.g., Demir and Gozgor 2018; Tiwari, Das, and Dutta 2019). The study approaches vary—from an analysis of single market (e.g., Khan et al. 2021; Liu, Liu, and Wang 2021), multiple markets (each market is analyzed independently of one another (e.g., Işık, Sirakaya-Turk, and Ongan 2020; Wu and Wu 2020)), to a sample of several markets using panel data models (e.g., Demir and Gozgor 2018; Tsui et al. 2018).

Table 1.

Extant Literature on EPU and Tourism Demand.

Author(s)	Region(s)	Period	Model	Dep. Var.	Exp. Var(s).	Finding(s)
Dragouni et al. (2016)	USA	1996M1–2013M12	VAR	Outbound tourists	SP500, Index of consumer sentiment, EPU	Spillover effects from EPU to outbound tourists are found.
Gozgor and Ongan (2017)	USA	1998Q1–2015Q4	Dynamic OLS	Real tourism spending	Real disposable income, real tourism price, gasoline price, EPU	Increase of EPU results in significant reduction in tourism expenditure in the long run.
Balli, Shahzad, and Salah Uddin (2018)	8 OECD countries	1997M1–2017M8	Wavelet analysis	International tourist arrivals	Country-specific and global EPU	Impacts of EPU on tourist arrivals are stronger during peak period of uncertainties (e.g., 9/11 and Global Financial Crisis). Canada, the UK, and USA are more influenced by EPU.
Demir and Gozgor (2018)	15 main countries worldwide	Varying country-specific periods	Fixed-effects and the bias-corrected least square dummy variable estimation	Outbound tourists	GDP, inflation, population, exchange rate, EPU	Positive shock to country-specific EPU can decrease outbound tourists of the home country in the current and next year.
Ongan and Gozgor (2018)	Japanese tourists to USA	1996Q1–2015Q1	Dynamic OLS	Japanese tourist arrivals	GDP, exchange rate, EPU	One SD positive shock to Japan’s EPU results in 4.7% decrease of Japanese tourist arrivals to USA.
Tsui et al. (2018)	Top 8 trading partners of New Zealand	2008M1–2015M12	OLS	Business visitors	GDP, flight distance, tourism index, total direct flight seats, dummy variables for certain events, EPU	New Zealand’s EPU has significant negative influence on inbound business visitors to New Zealand. The spillover effects of other countries’ EPU are insignificant on business visitors to New Zealand.
Tiwari, Das, and Dutta (2019)	India	2003M1–2017M6	Wavelet analysis	Outbound tourists	Geopolitical risk index, EPU	Geopolitical risks have stronger and more persistent effects on tourist arrivals of India than the India’s EPU.
Wu and Wu (2019)	Portugal, Ireland, Italy, Greece, and Spain	1995–2015	Wavelet analysis	International tourist arrivals	GDP, EPU	There exists unidirectional causal effect from Europe’s EPU to tourism receipts of European countries in the short run, and bidirectional causality between the two in the long run.
Akadiri, Alola, and Uzuner (2020)	12 major economies in America, Europe, and Asia Pacific	1995–2016	VAR	International tourist arrivals	GDP, EPU	Unidirectional causality from EPU to tourist arrivals exists in France and Ireland, while the reverse exists in Brazil, Canada, China, Germany, Ireland, and USA.
Chen et al. (2020)	Tourists from mainland China and Japan to Taiwan	1999M1–2016M12	Markov regime-switching model	Tourist arrivals’ room demand	GDP, exchange rate, inflation, EPU	Room demand by mainland Chinese and Japanese tourists tend to be lower in the trough period due to high economy policy uncertainty.
Işık, Sirakaya-Turk, and Ongan (2020)	Tourist arrivals to USA from Canada and Mexico	1996M1–2017M9	Dynamic OLS	International tourist arrivals	Industrial production index, exchange rate, EPU	1% increase in Canada’s (Mexico’s) EPU leads to 0.37% (0.19%) decrease in tourists from Canada (Mexico) to USA.
Nguyen, Schinckus, and Su (2020)	82 countries worldwide	2002–2014	GMM	Tourist departure, tourism expenditure	GDP, remittance, inflation, unemployment, vulnerable employment, aging population, urban population, CO2 emission, forest area, percentage of individuals using internet, global EPU	EPU leads to (i) increase (decrease) in tourism departure (tourism expenditure) in lower-middle income countries (migration effect); (ii), decrease (increase) in tourism departure (tourism expenditure) in upper-middle income countries (concentration effect); (iii) increase in both tourism departure and total tourism expenditures, but decrease in per tourist’s expenditure in high income countries (inferior good effect).
Khan et al. (2021)	The UK	2000M1–2019M3	Time-varying rolling window technique	International tourist arrivals	GDP, infrastructure development, exchange rate, EPU	EPU has negative effect on inbound tourists.
Liu, Liu, and Wang (2021)	Hong Kong SAR	1982Q2–2018Q1	VAR	Visitor arrivals	GDP, EPU	EPU changes the size but not the direction of the effect of GDP shock on visitor arrivals. Different EPU (HK, China, and global) have different size of impacts.
Wu and Wu (2020)	Brazil, India, Indonesia, South Africa, and Turkey	1997–2016	Wavelet analysis	International tourism receipts	GDP, international tourism arrivals, global EPU	A generally positive relationship between global EPU and international tourism receipt is found. However, the relationship changes over time and demonstrates low-to-high frequency cycles.

As summarized in Table 1, the majority of findings confirm the negative impact of EPU on tourism demand. Three studies apply the VAR-type models (e.g., Akadiri, Alola, and Uzuner 2020; Dragouni et al. 2016; Liu, Liu, and Wang 2021), but only the study by Liu, Liu, and Wang (2021) provides an impulse response analysis. Although not directly simulating the EPU shock, the authors have recognized that the presence of EPU variable in the model changes the response of inbound tourism demand to GDP shock. Moreover, a number of studies investigate the EPU-tourism relationship using wavelet analysis (e.g., Wu and Wu 2019, 2020). However, while this technique may be superior in demonstrating the evolution of tourism demand over time, in its current form it is not suitable for exploring EPU shock in cross-country transmission.

Vatsa (2020) has found a high degree of global interconnectedness in tourism demand by outbound markets for New Zealand, which demonstrate common trends and cycles in the form of synchronous fluctuations. Other empirical evidence supports the interdependence between a number of destinations/origins (e.g., Balli and Tsui 2016; Chan, Lim, and McAleer 2005; Chang et al. 2011; Kožić, Sorić, and Sever 2019; Seo, Park, and Boo 2010), showing statistical correlations between tourist flows in one country and the tourist flows in another country. The nature of interdependence can be influenced by whether the set of tourism destinations are complements or substitutes (Cao, Li, and Song 2017; Kožić, Sorić, and Sever 2019).

In an interdependent global environment, an impact of such economic policy uncertainty is expected to reverberate through countries connected directly as well as indirectly to the source of EPU. As demonstrated in the studies examining EPU and economic activities in general, the capacity to analytically embed this cross-sectional interdependence—how tourism flows (inbound and/or outbound) to/from one country relate to the tourism flows of another—is an important methodological requisite for a more complete understanding of how EPU shock propagates. Such interdependence can be accommodated in various ways. Studies may regress the tourism demand indicators of the destinations with (i) EPU in another market (e.g., Liu, Liu, and Wang 2021) or (ii) a global EPU (e.g., Liu, Liu, and Wang 2021; Nguyen, Schinckus, and Su 2020; Wu and Wu 2020).

However, if the selection of countries/markets included are solely based on tourism trading partners (such as a number of major origin-destination pairs), the estimated EPU impacts on tourism markets in the broader macroeconomic context can be distorted. This is because interdependence between tourism markets are not only influenced by tourism flows, but also other types of economic activities with very different patterns (Song, Li, and Cao 2018). For example, general merchandise trade is viewed as the most important source of inter-country business cycle linkages (Baxter and Kouparitsas 2005; Imbs 2004; Ong and Sato 2018). China and Australia tourism and merchandise trade are a case in point. Between 2016 and 2018, on average, China (Australia) represents 16.1% (1.2%) of visitor inflows into Australia (China) (UNWTO 2019), whereas in merchandise trade volume China (Australia) represents 44.0% (8.8%) into Australia (China) (International Monetary Fund [IMF] 2019). Although larger sample size may partially alleviate this concern (such as in Khalid, Okafor, and Shafiullah 2020; Lee and Chen 2021, which conduct a global analysis involving 100+ countries), econometric model which is capable of accounting for linkages in the country-specific regressions’ error terms is also necessary. The latter is particularly important when modeling EPU shock propagation is the primary interest.

Although not specific to EPU and tourism demand, such a modeling approach in tourism research is available in Cao, Li, and Song (2017) and Assaf et al. (2019). They have examined how the country-specific shocks to income and price heterogeneously impact the tourism demand of all the observed countries using GVAR. Examining 24 main countries around the world, Cao, Li, and Song (2017) find that negative shocks to China’s income and price tend to have larger impacts on developing countries and China’s neighboring countries. Later, Assaf et al. (2019) examine nine countries in Southeast Asia and demonstrate that a country-specific positive shock to GDP growth can boost inbound tourism to surrounding countries differentially. In particular, the authors highlight that interdependence of tourism demand between border-sharing countries can be explained by cross-country social, cultural, and economic exchanges, which lead to growing travel between the countries (Assaf et al. 2019). Following this lead, the current paper constructs a GVAR system with the purpose of examining the tourism demand impact of EPU shock.

The (G)VAR Approach

“Modeling the bidirectional causations between tourism demand and its determinants is driven by the reality of global economic integration, in addition to being a means of advancing econometric techniques” (Assaf et al. 2019, 384). The OLS method treats the explanatory variables as exogeneous, and thus may render the estimates invalid (Assaf et al. 2019). The Almost Ideal Demand system (AIDS) used in tourism demand studies mainly in the 1990s and 2000s (Song and Li 2008; Song, Qiu, and Park 2019) is underpinned by economic demand theory and applied to model the “budget” allocations by tourists over a number of destinations. However, the concept of interdependence in the AIDS, which is based on the complementary/substitutive relationships between tourism markets, is narrowly defined. Its rigid specification restricted by the economic theory makes it difficult to include other explanatory variables (Cao 2016). Treating all the variables as endogenous without the restrictive axioms of demand theory, the VAR approach is suitable for examining how EPU affects tourism demand. Commonly used in assessing the effects of policy change, the impulse response function of the VAR approach demonstrates the marginal effects of a one-time simulated shock to one endogenous variable on the evolution of other endogenous variables.

With the improvement in data collection and greater accessibility to databases (such as International Monetary Fund (IMF), World Tourism Organization (UNWTO), etc.), extensive and comparable data for many countries have become available for researchers to examine the interdependent nature of tourism demand within a region or around the world. To adequately explain the effects of shocks with substantial cross-country comparisons, Cao, Li, and Song (2017) suggest the inclusion of tourism flows and economic determinants in empirical tourism interdependence studies. However, considering a large number of countries in a standard VAR specification requires estimations of numerous parameters, usually exceeding the number of observations. To deal with the problem of overparameterization, three popular techniques are used: (i) factor augmented VAR (FAVAR); (ii) panel VAR (PVAR); and (iii) global VAR (GVAR) (Feldkircher, Huber, and Pfarrhofer 2020).

The FAVAR can largely reduce the dimensionality, in which a large set of variables is shrunk into a small number of factors. However, with the FAVAR, the economic interpretations of the factors and the interdependence across countries may be unclear (Dees et al. 2007). The PVAR model accounts for cross-country interdependence by including lagged foreign endogenous variables, and restrictions which reduce the dimensionality of the model are imposed on some parameters (Feldkircher, Huber, and Pfarrhofer 2020).

Viewed as a special case of PVAR, the GVAR, first introduced by Pesaran, Schuermann, and Weiner (2004) and extended by Dees et al. (2007), deals with a large number of countries in an effective manner. The GVAR model replaces the lagged foreign endogenous variables (in the PVAR) with a vector of weakly exogenous variables, which is constructed as cross-country weighted averages of the foreign variables (known as common correlated effects (CCE) estimators). Analogously, with GVAR, the underlying country can be viewed as trading with a particular “pseudo country” (or “rest of the world”), which is constructed by a weighted average of all foreign economies. Empirical evidence supports that taking cross-country information into account can improve the forecasting performance compared to solely relying on domestic variables (Feldkircher, Huber, and Pfarrhofer 2020). Further, based on Monte Carlo experiments, the CCE estimators (as used by the GVAR), which make use of cross-country averages, outperform the corresponding estimators based on principle components (as used by the FAVAR) (Kapetanios and Pesaran 2005). In modeling international tourism demand with GVAR, individual destinations are separately estimated but systematically linked. This is in sharp contrast with conventional single equation panel data approaches which generate “one size fits all” estimates for a sample of destinations (Dogru, Sirakaya-Turk, and Crouch 2017; Dogru, Bulut, and Sirakaya-Turk 2021). As will be seen, practical implications are enriched as a result of using the GVAR approach in tourism demand modeling. For the above reasons, our study adopts the GVAR approach to model the impact of EPU on tourism demand. The next section explains the key features of the methodology.

Methodology

GVAR Model

A GVAR model consists of a set of country-by-country VAR models that include a set of country-specific domestic, foreign, as well as global variables. The global system is segmented into different local models for individual estimations. The approach avoids the need to estimate a large group of parameters simultaneously and allows for heterogeneity across countries. The GVAR model is constructed in two steps: the country-specific models and the global VAR.

The country-specific model is a reduced-form VAR model augmented with weakly exogenous variables conditional on the rest of the countries (the so called VARX). In the first step, the country-specific VARX model is constructed and estimated for each individual country. This consists of a set of endogenous domestic variables $x_{it}$ , a set of weakly exogenous foreign variables $x_{it}^{*}$ , and weakly exogenous global observed common variables $g_{t}$ (e.g., the oil price) (for the USA equation, $g_{t}$ is treated as endogenous). The foreign variables $x_{it}^{*} = \sum_{j = 1}^{N} w_{ij} x_{jt}$ , proxy for global unobserved factors, are the weighted averages of the corresponding domestic variables across all the other countries investigated. Specifically, $w_{ij}$ is the corresponding element of the $N * N$ weighted matrix, which is the bilateral exposures between the markets under investigation (Martin and Crespo Cuaresma 2017). “This matrix plays a key role in linking up the models of the different regions together and shows the degree to which one country/region depends on the remaining countries” (Pesaran, Schuermann, and Weiner 2004, 140). And “geographical patterns of trade provide an obvious source of information for this purpose and could also be effective in mopping up some of the remaining spatial dependencies” (Dees et al. 2007, 6). Thus, the individual country under the GVAR model can be viewed as a “small open economy,” such that $x_{it}^{*}$ are unidirectional long run forcing variables for $x_{it}$ (Chudik and Smith 2013).

In the first step, the $VARX (p, q, s)$ of country i with p lagged order of $x_{it}$ , q lagged order of $x_{it}^{*}$ and s lagged order of $g_{t}$ , can be written as follows:

x_{i t} = a_{i 0} + a_{i 1} t + \sum_{j = 1}^{p} Φ_{i j} x_{i, t - j} + \sum_{k = 0}^{q} Λ_{i k} {x *}_{i, t - k} + \sum_{l = 0}^{s} Ψ_{i l} g_{t - l} + u_{i t}

(1)

where the lagged orders p, q and s are chosen by Akaike information criterion (AIC). The error term $u_{it}$ , is independent and identically distributed (i.i.d) in the context of country-specific VARX equation. But across the country-specific VARX equations under the GVAR model, it is assumed to be cross-sectionally weakly dependent, which enables the impact transmission of idiosyncratic shock from the individual country model (Dees et al. 2007). Assuming the variables are integrated of order one ( $I (1)$ ), equation (1) allows for cointegration among the domestic, foreign, and global variables. And the rank/number of cointegrating relationships can be determined by Johansen’s trace statistic or maximal eigenvalue statistic (Pesaran, Shin, and Smith 2000).

For ease of exposition, the $VARX (1, 1, 1)$ can be re-parameterized into an error-correction form $VECMX (1, 1, 1)$ with rank of order $r_{i}$ :

Δ x_{i t} = c_{i 0} - α_{i} β_{i}^{'} [z_{i, t - 1} - γ_{i} (t - 1)] + Λ_{i 0} Δ x_{i t}^{*} + Ψ_{i 0} Δ g_{t} + u_{i t}

(2)

where $z_{it} = (x_{it}^{'}, x_{it}^{*'}, g_{t}^{'})'$ , $β_{i}$ is the $(k_{i} + k_{i}^{*} + k_{g}) * r_{i}$ matrix of cointegration coefficients, $α_{i}$ is the $k_{i} * r_{i}$ error-correction or speed of adjustment matrix to the long run relationships. $x_{it}$ is a $k_{i} * 1$ vector of domestic variables of country $i$ , ${x^{*}}_{it}$ is a $k_{i}^{*} * 1$ vector of weakly exogenous foreign variables of country $i$ , and $g_{t}$ is the $k_{g} * 1$ vector of weakly exogenous global observed common variables. By partitioning $β_{i}$ as $β_{i} = (β_{ix}^{'}, β_{i x^{*}}^{'}, β_{ig}^{'})^{'}$ , the error correction terms can be written as follows:

β_{i}^{'} [z_{i, t - 1} - γ_{i} t] = β_{i x}^{'} x_{i t} + β_{i x^{*}}^{'} x_{i t}^{*} + β_{i g}^{'} g_{t} - (β_{i}^{'} γ_{i}) t

(3)

In the second step, the VARX equations are stacked to generate the global model (in this paper, the latter is used to espouse internationalization and regionalization). Let’s define $z_{it} = {(x_{it}, x_{it}^{*}, g_{t})}^{'}$ , and $x_{t} = W_{i}^{- 1} z_{it}$ , where the weighted matrix $W_{i}$ is a $(k_{i} + k_{i}^{*} + k_{g}) * k$ matrix and $x_{t} = {(x_{1 t}^{'}, \dots, x_{Nt}^{'})}^{'}$ is a $k * 1$ vector which includes all the variables (assuming $k = \sum_{i = 1}^{N} k_{i}$ ). Thus, assuming $p = q = s$ for ease of exposition, the $GVAR (p)$ model can be rearranged as:

x_{t} = G_{0}^{- 1} a_{0} + G_{0}^{- 1} a_{1} t + G_{0}^{- 1} G_{1} x_{t - 1} + \dots + G_{0}^{- 1} G_{p} x_{t - p} + G_{0}^{- 1} u_{t}

(4)

where $a_{0}$ , $a_{1}$ , $G_{0}$ , $G_{1}$ ,…, $G_{p}$ are all country-specific vectors.

All the variables in the global model are solved simultaneously. The GVAR model is characterized by the construction of impulse response to study the dynamic process of the data. The impulse responses demonstrate the evolutions of the endogenous variables in future periods in response to a counterfactual scenario, identified by a variable-specific shock from country $i$ . As suggested by Pesaran, Schuermann, and Weiner (2004), we choose the generalized impulse response function (GIRF) rather than the orthogonalized impulse response function (OIRF) because the order of the factors in the former does not influence the results of the impulse responses. Details of the GVAR model can be found in Chudik and Pesaran (2016), Dees et al. (2007), and Pesaran, Schuermann, and Weiner (2004).

Internationalized and Regionalized Systems

According to Held et al. (2000), the four spatially delimited processes of globalism/globalization are “localization,”“nationalization,”“regionalization,” and “internationalization.” The latter two are most relevant to this paper. “Regionalization can be denoted by a clustering of transactions, flows, networks, and interactions between functional or geographical groupings of states or societies,” whereas they note that “internationalization can be taken to refer to patterns of interaction and interconnectedness between two or more nation-states irrespective of their specific geographical location” (Held et al. 2000, 17). GVAR studies have provided empirical evidence that macroeconomic variables of the countries investigated are more influenced by shocks originating from countries within the region than outside the region (e.g., Ong and Sato 2018). Tourism markets are closely interdependent particularly within the same region (Assaf et al. 2019), such as the Asia Pacific region, which can be explained by its dominance of intra-regional travel¹ (UNWTO 2019), as well as regional economic cooperation (e.g., regional trade agreements) (Khalid, Okafor, and Burzynska 2021).

In this study, both the internationalized and regionalized (Asia Pacific) tourism systems are estimated separately using the GVAR model. Characterized by the cross-sectional interdependence mechanism embedded in the GVAR model, estimation results can change substantially as cross-sectional sample frame changes. The underlying markets may respond differently in exposure to an internationalized or a regionalized environment. In our study, the differences in tourism demand response to EPU shock under the two systems give insights on whether interdependence in an international-wide or a regional-wide environment is favorable to the resilience of the tourism sector. In the GVAR context, compared to the internationalized system, a regionalized system is simpler to specify and update.

Subject to data availability, the internationalized system includes a total of 31 tourism markets around the world with 12 in Asia Pacific (see Table 2). Note that the seven countries in Euro Area (France, Germany, Greece, Italy, Netherlands, Portugal, and Spain) are treated as a single market following the practice of GVAR literature. The data cover most tourism regions defined by the UNWTO except the Middle East, where corresponding data are inadequate.

Table 2.

The Selected Tourism Markets.

Asia Pacific				Non-Asia Pacific
East Asia	South Asia	Southeast Asia	Oceania	Africa	Euro Area	North-Western Europe	Rest of Europe	Latin America	North America
China (CN)	India (IN)	Indonesia (ID)	Australia (AU)	South Africa (ZA)	France (FR)	Denmark (DK)	Poland (PL)	Brazil (BR)	Canada (CA)
Hong Kong SAR (HK)		Malaysia (MY)	New Zealand (NZ)		Germany (DE)	Norway (NO)	Russia (RU)	Mexico (MX)	USA (US)
Japan (JP)		Philippines (PH)			Greece (GR)	Sweden (SE)	Turkey (TR)
South Korea (KR)		Singapore (SG)			Italy (IT)	UK (UK)
		Thailand (TH)			Netherlands (NL)
					Portugal (PT)
					Spain (ES)

A regionalized tourism system is estimated separately from the internationalized system. The former comprises 12 Asia Pacific markets. In 2019, aggregate tourism receipts of the 12 selected Asia Pacific markets accounted for almost 80% of market share in the region (UNWTO 2020). Unlike the internationalized system, the regionalized system only depicts intra-regional interdependence among the tourism markets. In the GVAR model, the so called “rest of the world” faced by a given market is constructed by the remaining 11 Asia Pacific markets. In other words, for modeling the transmission channels beyond the Asia Pacific region are cut off or filtered out.

Variables

Table 3 summarizes the variables included in the two systems. Quarterly data are used for estimation covering 2000Q1 to 2019Q4. We use travel credit/debit under the balance of payments (BoP) of each economy as a proxy for inbound/outbound tourism demand. United Nations (2010) considers international tourism as traded services, and recommends the use of BoP data to describe international tourism activities. Based on IMF (2005), the term travel in the BoP is synonymous with tourism. Travel credit (debit) covers receipts (expenditures) from (by) inbound (outbound) visitors on business and personal travels of less than a year in one economy (other economies) (IMF 2005, 2009). Tourism receipt/expenditure data which are collected using survey at the border and/or the airport of destination, are not widely available. Since IMF compiles travel credit/debit across all economies, it is a suitable proxy for inbound/outbound tourism demand. Similar practice can also be found in Cao, Li, and Song (2017). Real travel credit/debit ( $rcr e_{it}$ or $rde b_{it}$ ) are computed by deflating the nominal travel credit/debit (in million USD) with home CPI (2015=100). Real GDP ( $y_{it}$ ; 2015 constant price; in million USD) and real exchange rate ( ${rex}_{it}^{(1)}$ or ${rex}_{it}^{(2)}$ ) are included, which are commonly used in tourism demand studies (Lim 1997; Song and Li 2008). Real exchange rate is given by CPI-adjusted nominal exchange rate against USD, or ${rex}_{it}^{(1)} = \ln (\frac{E X_{it}}{CP I_{it}})$ . This follows the common practice of extant literature with US-included GVAR system (Bi and Xin 2021; Cao, Li, and Song 2017; Dees et al. 2007; Han, Qi, and Yin 2016). Specifically, the weakly exogenous ${rex}_{USA, t}^{* (1)}$ is only included in the USA equation, and the endogenous ${rex}_{it}^{(1)}$ only in the remaining equations. This is because real exchange rates are constructed as a closed system in which the real exchange rate of any single market can be derived from the remaining markets.

Table 3.

Variables Included in the Market-Specific VARX/VECMX Models.

Variable	Internationalized system				Regionalized system
	All markets except USA		USA		All markets
	Endogenous	Weakly Exogenous	Endogenous	Weakly Exogenous	Endogenous	Weakly Exogenous
Real travel credit	$rcr e_{it}$	${rcre}_{it}^{*}$	$rcr e_{USA, t}$	${rcre}_{USA, t}^{*}$	$rcr e_{it}$	${rcre}_{it}^{*}$
Real travel debit	$rde b_{it}$	${rdeb}_{it}^{*}$	$rde b_{USA, t}$	${rdeb}_{USA, t}^{*}$	$rde b_{it}$	${rdeb}_{it}^{*}$
Real GDP	$y_{it}$	$y_{it}^{*}$	$y_{USA, t}$	$y_{USA, t}^{*}$	$y_{it}$	$y_{it}^{*}$
Real exchange rate	${rex}_{it}^{(1)}$	-	-	${rex}_{USA, t}^{* (1)}$	${rex}_{it}^{(2)}$	${rex}_{it}^{* (2)}$
EPU	$ep u_{it}$	${epu}_{it}^{*}$	$ep u_{USA, t}$	${epu}_{USA, t}^{*}$	$ep u_{it}$	${epu}_{it}^{*}$
Oil price	-	$poi l_{t}$	$poi l_{t}$	-	-	-

Note: The variables with asterisk are the market-specific foreign variables. Subject to data availability, $ep u_{it}$ is included in only Australia, Brazil, Canada, China, Euro Area, Hong Kong SAR, Japan, Mexico, Russia, South Korea, Sweden, the UK, and USA.

In contrast, under the regionalized system which excludes USA, real exchange rate ( ${rex}_{it}^{(2)}$ ) is given by CPI ratio-adjusted nominal exchange rate against USD, adjusted by both the USA CPI and home CPI, or ${rex}_{it}^{(2)} = \ln (\frac{E X_{it} * CP I_{USA, t}}{CP I_{it}})$ (Assaf et al. 2019). This approach can eliminate the situation that an increase in the real exchange rate is purely due to the appreciation of USD, when USA is not included in the regionalized system. Meanwhile, as the real exchange rates are no longer constructed as a closed system, we incorporate both ${rex}_{it}^{(2)}$ and ${rex}_{it}^{* (2)}$ in each market-specific equation.

For the impulse response analysis, we simulate the adverse shocks of China’s EPU, as well as China’s GDP. The two measures of China’s EPU are (i) the one based on South China Morning Post (SCMP), a Hong Kong leading English-language newspaper (Baker et al. 2013); and (ii) the one based on Renmin Daily and Guangming Daily (RDGD), two national Chinese-language newspapers in mainland China (Davis, Liu, and Sheng 2019). As the EPU index is based on specific newspapers, the positions/perspectives of the targeted newspapers determine the coverages and descriptions of EPU. Statistically, time series of the RDGD-based China’s EPU has smaller fluctuations than the SCMP-based China’s EPU.

Under the internationalized system, the global common variable, given by the UK Brent oil price (in USD), $poi l_{t}$ , is considered endogenous but not weakly exogenous in the USA equation due to the relative dominance of the USA economy in the world. Following Kempa and Khan (2017), under the regionalized system which excludes USA, we do not incorporate the global common variable. Data for travel credit/debit, CPI, nominal exchange rate, and oil price are collected from IMF (data.imf.org). Data for real GDP are extracted from Oxford Economics, and the market-specific EPU data from the Economic Policy Uncertainty website (www.policyuncertainty.com). Data (except the EPU variable) are log transformed and seasonally adjusted using the X12 method.

Weighted Matrix

For GVAR estimation, the selection of weighted matrix determines the nature of the linkages of individual markets with other markets. Previous studies have used different weighted matrices in estimating the GVAR model and distinctive impulse response results are obtained (e.g., Gross 2019; Inoue, Kaya, and Ohshige 2015; Martin and Crespo Cuaresma 2017; Tam 2018).

For the internationalized system, we utilize the merchandise trade volumes (totals of export and import) to construct the weighted matrix, which is a conventional and common practice in the tourism GVAR literature (e.g., Cao, Li, and Song 2017) as well as other disciplines (e.g., Bi and Xin 2021; Dees et al. 2007; Han, Qi, and Yin 2016; Sznajderska 2019; Tam 2018). This is reasonable because the impacts of the shock are not only transmitted through visitor flows, but also other economic channels mentioned in the Literature review section of the paper. Data for bilateral merchandise trade volumes between tourism markets are available from IMF (data.imf.org). We take the averages of 2016–2018 to construct the weighted matrix to avoid the effect of idiosyncrasy for a certain year.

For the regionalized system, considering the dominance of intra-regional travel within Asia Pacific, bidirectional origin-destination (O-D) pair visitor volume-based weighted matrix (averages of 2016–2018) is utilized for estimation (similar practice can also be found in Assaf et al. 2019), in addition to the conventional merchandise trade volume-based weighted matrix. This enables us to understand how the impacts of EPU shock differ when the assumptions about the transmission channels are different. Annual data for O-D pair visitor volume² (including inbound and outbound) are collected from UNWTO Tourism Statistics (www.unwto.org/statistics). Due to data unavailability, the visitor volume-based weighted matrix cannot be constructed for the internationalized system. Table 4 shows the extent to which the weights can differ, although the lists of top trading partners in these two matrices tend to be similar (as noted by the underlined figures in Table 4).

Table 4.

Weighted Matrix (%) of the Observed Markets (Visitor Volume and Merchandise Trade Volume).

Market	AU	CN	HK	ID	IN	JP	KR	MY	NZ	PH	SG	TH
Australia (AU)		16.1%	6.7%	11.5%	4.9%	7.3%	3.5%	5.9%	22.1%	3.1%	11.7%	7.3%
Australia (AU)		44.0%	2.7%	2.9%	4.8%	17.2%	8.5%	4.7%	4.5%	0.7%	4.4%	5.5%
China (CN)	1.2%		72.9%	1.6%	0.6%	5.8%	5.7%	2.1%	0.3%	1.2%	2.3%	6.1%
China (CN)	8.8%		20.4%	4.3%	5.6%	20.2%	18.9%	6.6%	1.0%	3.5%	5.3%	5.5%
Hong Kong SAR (HK)	0.6%	91.7%		0.4%	0.3%	2.4%	1.5%	0.4%	0.1%	0.7%	0.8%	1.0%
Hong Kong SAR (HK)	1.5%	69.3%		0.7%	3.8%	6.4%	4.9%	2.6%	0.1%	1.7%	6.1%	2.8%
Indonesia (ID)	8.1%	15.4%	3.1%		3.4%	4.9%	3.6%	29.6%	0.8%	1.8%	25.3%	4.0%
Indonesia (ID)	4.1%	29.0%	2.2%		8.1%	16.1%	7.6%	6.6%	0.6%	3.5%	14.6%	7.7%
India (IN)	8.6%	14.2%	5.8%	8.3%		4.9%	3.8%	12.5%	1.6%	2.2%	18.6%	19.5%
India (IN)	7.1%	36.5%	11.3%	8.4%		6.9%	8.6%	6.6%	0.4%	1.0%	8.7%	4.5%
Japan (JP)	3.1%	33.0%	10.9%	2.9%	1.2%		30.0%	2.7%	0.5%	3.3%	3.9%	8.3%
Japan (JP)	8.3%	45.2%	5.6%	5.1%	2.2%		12.1%	4.8%	0.8%	3.2%	4.7%	8.0%
South Korea (KR)	1.5%	34.8%	7.3%	2.3%	1.0%	31.9%		3.0%	0.4%	7.2%	2.9%	7.7%
South Korea (KR)	6.0%	48.7%	8.3%	3.6%	3.9%	16.1%		3.5%	0.5%	2.7%	4.1%	2.6%
Malaysia (MY)	2.4%	11.4%	1.6%	16.6%	2.8%	2.6%	2.6%		0.3%	1.7%	41.3%	16.6%
Malaysia (MY)	4.7%	26.4%	6.3%	6.3%	5.4%	12.2%	6.2%		0.7%	2.2%	20.5%	9.1%
New Zealand (NZ)	61.8%	12.4%	3.4%	3.1%	2.6%	3.6%	2.6%	2.4%		1.2%	4.1%	2.9%
New Zealand (NZ)	24.4%	35.7%	1.8%	2.9%	2.1%	11.4%	6.1%	4.2%		1.3%	4.7%	5.4%
Philippines (PH)	4.4%	23.9%	10.9%	3.6%	1.8%	11.3%	23.0%	6.0%	0.6%		9.9%	4.6%
Philippines (PH)	1.8%	24.6%	10.6%	6.3%	2.0%	20.2%	10.3%	5.1%	0.5%		9.3%	9.2%
Singapore (SG)	4.9%	13.0%	3.7%	14.9%	4.4%	3.9%	2.7%	43.2%	0.6%	2.9%		5.9%
Singapore (SG)	3.7%	23.4%	12.2%	10.0%	4.5%	9.5%	7.9%	19.4%	0.6%	3.3%		5.4%
Thailand (TH)	3.4%	38.7%	5.0%	2.6%	5.2%	9.2%	8.0%	19.4%	0.5%	1.5%	6.5%
Thailand (TH)	6.1%	29.6%	5.9%	6.6%	4.1%	22.3%	5.1%	9.0%	0.9%	4.1%	6.4%

Note: Figures in bold refer to visitor volume and figures in italic refer to merchandise trade volume. The underlined elements are the top 4 trading partners for each market by visitor volume or merchandise trade volume. Data for visitor volume are from UNWTO, whereas data for merchandise trade volume from IMF. All data are averages of 2016–2018.

Empirical Results

Cointegration Tests

In our study, the GVAR model is estimated using the GVAR toolbox 2.0 (Smith and Galesi 2014). The market-specific equations of the GVAR model are written in error correction form (VECMX) if there exists cointegration among the underlying variables. The reduced-rank regression techniques used in the estimations of the market-specific VECMX equations are based on the assumption that the underlying variables included are approximately integrated of order one (or I(1)) (Dees et al. 2007; Pesaran, Schuermann, and Weiner 2004). We employ two types of unit root tests, namely the standard ADF test and the weighted symmetric ADF test (Park and Fuller 1995). Based on the unit root test results (see Tables A1 and A2 in Appendix), the domestic variables, foreign variables, and global common variable are predominately $I (1)$ , which ensure the appropriateness of the GVAR framework.³ The Johansen trace statistics have detected that 28 of the 31 markets contain at least one cointegrating relationship (those markets are specified by the VECMX equations), except for India, South Africa, and the UK (these markets are accommodated in the VARX equations with variables in first difference) (see Table A3 in Appendix; corresponding results for the regionalized system are provided upon request).

Weak-exogeneity Tests

Next, we focus on the weak-exogeneity assumption of the foreign ( $x_{it}^{*}$ ) and global common ( $g_{t}$ ) variables, which is a main characteristic of interdependence across markets. Weak-exogeneity assumes there exists unidirectional long run force from $x_{it}^{*}$ (or $g_{t}$ ) to $x_{it}$ without ruling out the lagged mutual effects between the two in the short term. Importantly, the GVAR model assumes that idiosyncratic shocks of the market specific equations should be cross-sectionally weakly correlated such that $COV (x_{it}^{*}, u_{it}) \to 0$ ( $u_{it}$ is the residual of the equation) when $N \to \infty$ (Dees et al. 2007). The weak-exogeneity tests for each foreign/global variable in the market-specific equations show that four of 110 statistics reject the null hypothesis of weak-exogeneity. More importantly, we do not reject the null hypothesis of weak-exogeneity for the equations of dominant economies like Euro Area or USA (see Table A4 in Appendix). Our results are consistent with those of Dees et al. (2007), as evidence to support the weakly exogenous foreign and global common variables specification in our study.

Impulse Responses Under the Internationalized System

We simulate one standard deviation (1SD) adverse (positive) shock to China’s EPU and demonstrate the impulse responses of inbound and outbound tourism demand of the observed markets. For ease of analysis, we group some of the tourism markets into regions as shown in Table 2 and examine the regional responses to shocks.⁴ Following Dees et al. (2007), the impulse response functions for individual markets within the same region are aggregated based on the weights of GDP (PPP) obtained from The World Bank (2019) (data.worldbank.org). We treat China, Hong Kong SAR, and India (South Africa), the only market in South Asia (Africa) region, as stand-alone markets. The impulse response graphs for the four individual markets are shown alongside all the other regions in Figure 1. The figure shows the bootstrapped medium estimates of responses to China’s EPU shock, which translates into a 53% increase in China’s EPU (information on the 90% confidence interval is provided upon request).

The globalization is mapped from the spatial-temporal dimensions by four elements, namely “extensity,”“intensity,”“velocity,” and “impact” (Held et al. 2000). These four elements correspond to “scope of affected markets” (which markets are affected); “magnitude of response” (to what extent the market is affected); “speed of tourism demand change” (whether the change is sudden or gradual); and “level of persistence” (how quick is the recovery), respectively, which are the four considered aspects in our impulse response analysis. Key findings/observations are that 1SD adverse (positive) shock to the SCMP-based China’s EPU leads to statistically significant decrease in inbound and outbound tourism demand⁵ within the first 10 quarters, followed by a gradual rebound between the 10th and 25th quarters. The variation in the impact is mainly with respect to the maximum decline and whether overshoot occurs during the rebound period. The maximum decline in inbound tourism demand of the individual markets varies between −8.4% (China) and −3.3% (Latin America).

Interestingly, during the rebound phase, an overshoot in inbound tourism demand can be observed in a number of markets (e.g., Hong Kong SAR, India, Oceania, and Southeast Asia). First identified by Bloom (2009), uncertainty/volatility overshoot has been observed in the VAR impulse response analyses in subsequent studies examining the impacts of uncertainty/volatility shocks on economic activities (e.g., Bachmann, Elstner, and Sims 2013; Carrière-Swallow and Céspedes 2013; Cuaresma, Huber, and Onorante 2019; Gourio, Siemer, and Verdelhan 2013; Haque, Magnusson, and Tomioka 2021). In the extant tourism literature, however, uncertainty overshoot has not been empirically demonstrated until the current study. The overshoot phenomenon in tourism demand can potentially be explained by the fact that pent-up tourism demand accumulates when people adopt the “wait and see” strategy in response to the uncertainty shock; for example, tourism professionals expect to observe a strong “decline-rebound-overshoot” effect during the post-crisis (e.g., COVID) period (Jin, Bao, and Tang 2021; McKercher 2021). In spite of its vulnerability to uncertainty shock, the tourism sector is also recognized as adaptive and resilient to external pressure (Reddy, Boyd, and Nica 2020).

Another key result in this study is the observation of an increase in outbound tourism demand from China (7.9%). As the dependent variable (outbound tourism demand) is proxied by real travel debit (not actual volume of outbound visitors), the increase in Chinese outbound travel in response to a shock to EPU is consistent with Wong (2017) in that capital flight occurs in search of more certain investment environment. Based on Wong (2017), reported property purchases by Chinese residents, as a form of capital flight, are mainly in Australia, Canada, and USA. Moreover, East Asia (Japan and South Korea) experiences a sudden increase (3.9%) in inbound tourism demand, along with a sharp recovery to the original level. This phenomenon is similar to the findings in Cao, Li, and Song (2017) in which an adverse shock to China’s real income induces a 1.7% increase in tourism export (or inbound tourism) of South America.

Furthermore, we simulate 1SD adverse (positive) shock to the RDGD-based China’s EPU. In China, the decline in inbound tourism demand (−5.6%) is smaller but more persistent than that in response to the SCMP-based EPU shock. On the other hand, an increase of outbound tourism demand from China, identified as a phenomenon of capital flight, is not observed under the RDGD-based EPU shock scenario. The difference in tourism demand response between the two scenarios might be attributed to the difference in the media’s global reach, and their political and economic pre-disposition. Regarding other markets, we only find a moderate and temporary decline in inbound and/or outbound tourism demand.

As a comparison, we also simulate 1SD adverse (negative) shock to China’s GDP (shown in Figure 2). While the responses of inbound and outbound tourism demand to the GDP shock work in the same direction as the EPU shock, the former impact is gradual. That is, GDP shock leads to a slower decline in inbound and outbound tourism demand in all the observed markets; it takes a longer time (15–25 quarters) to achieve the maximum decline, and the maximum decline in inbound and outbound tourism demand are smaller than the EPU shock (with few exceptions such as the inbound tourism demand of India, South Africa, and Southeast Asia). Importantly, the capital flight from China of a similar magnitude, is also observed under the GDP shock scenario.

Impulse Responses Under the Regionalized System

To reflect the strong linkages in intra-regional travel by the Asia Pacific markets, we utilize the visitor volume-based weighted matrix to estimate the regionalized system and generate the impulse responses. Figure 3 shows a comparison between the SCMP-based and RDGD-based China’s EPU shocks.

The decline in inbound tourism demand of China, Hong Kong SAR, and Southeast Asia under the SCMP-based EPU shock scenario are more noticeable in magnitude and persistence under the regionalized system than the internationalized system. This means, from a viewpoint of maintaining shock-resilient international tourism flows, it appears the globalized/internationalized environment is favorable for EPU shock generating market and its neighbors. This is because the ripple effects are absorbed and diluted across many more countries under the internationalized system.

Importantly, an increase in outbound tourism demand (proxied by real travel debit) from China in response to EPU shock is no longer observed under the regionalized model. This result may support the view that the capital flight out of China in response to EPU shock mainly targets non-Asia Pacific markets (e.g., Canada and USA). It also highlights how results can change substantially as cross-sectional sample frame changes.

Responses to SCMP-based shock and RDGD-based shock move in opposite directions in Australia, India, New Zealand, South Korea, and Southeast Asia. Notably, regarding the responses to the SCMP-based EPU shock, rebound and overshoot after the initial decline are not observed under the regionalized system. Rather, the effects of the SCMP-based EPU shock are more persistent. With respect to the RDGD-based EPU shock, partial or full recovery after the initial decline is observed in most of the markets (e.g., China, Hong Kong SAR, India, and Japan).

From Figure 4, we notice that China’s GDP shock has smaller impacts on tourism demand of all the observed markets compared to China’s EPU shock. Apart from China and Hong Kong SAR, where significantly persistent decline in inbound tourism demand are observed, the impacts of GDP shock to all the other markets are either short in duration (within 1 to 2 quarters) or statistically insignificant. The differences between the internationalized and regionalized systems may be attributed to the differences in cross-sectional sample frame and/or the weighted matrix used (recall that the internationalized system uses the merchandise trade volume-based weighted matrix while the regionalized system uses the visitor volume-based weighted matrix). Importantly, capital flight from China occurs again under the GDP shock scenario, though the magnitude is smaller than the internationalized system.

Under the internationalized and/or regionalized system(s), we also attempt to simulate the EPU shocks from other main tourism markets (e.g., Australia) and dominant economies (e.g., USA). In general, EPU shocks emanating from Australia and USA do not have significant impacts on inbound and outbound tourism demand of most markets included in the study (results are provided upon request). GVAR empirical studies have demonstrated the significant global influences of the USA EPU shock on economic activities (e.g., Tam 2018; Trung 2019). However, our study shows this result may not be directly transferred to tourism activities—measured by real travel credit/debit. Conversely, although Australia is an important destination in the Asia Pacific region, other tourism markets are not significantly affected by its EPU shock.

We also attempt to utilize the merchandise trade volume-based weighted matrix in estimation of the regionalized system, but unstable impulse responses are generated. This might be attributed to the inappropriateness of utilizing bilateral merchandise trade volumes to depict the interdependence or transmission mechanism within the Asia Pacific region, which is characterized by the tight linkages in intra-regional travel, but not necessarily in merchandise trade.

Conclusion

In a global tourism environment, the impact of economic policy uncertainty (EPU) from a country can be transmitted through a variety of channels to many countries. While the impact of EPU on inbound and/or outbound tourism demand has been investigated in the research literature, to-date the studies have been limited to a relatively small number of cross-sectional samples and they have not employed a method that can account for interdependence between many countries. As the GVAR approach has been developed for econometric advancement as well as the need to account for the contemporary reality of globalized economic activities (Assaf et al. 2019), it is capable of accounting for the inter-country dependence (or sometimes referred to as cross-sectional interdependence) while preserving the advantages of standard VAR model. This paper is the first attempt to examine how EPU from China impacts other tourism markets using GVAR. We employ the GVAR model to construct the 31-markets internationalized system and the 12-markets Asia Pacific regionalized system. Through these models, we examine how the impacts of China’s uncertainty shock are transmitted to other tourism markets.

In general, our study finds that while China’s EPU shock has significant negative effects on both the inbound and outbound tourism demand of most of the observed markets in both systems, noteworthy variations and unique empirical insights are observed. For instance, the impacts of the shock on China, Hong Kong SAR, and Southeast Asia are smaller under the internationalized system than the regionalized system. Furthermore, unlike in the regionalized system, the transmission mechanism embedded within the internationalized system includes channels for rebound and even overshoot after the maximum decline. In GVAR context, the abovementioned two disparities highlight how results vary as cross-sectional sample frame, as well as weighted matrix, changes. For policy implications, these imply that on balance, the globalized/internationalized environment may be favorable (i) in achieving shock-resilient tourism trade environment for the shock generating market and its neighbors; and (ii) in tourism recovery, as tourism is embedded in the general economic circulation system.

Notably, measured by travel debit, the model captures the initial “capital flight” out of China during times of uncertainty in the form of outbound tourism. Additionally, the study finds that EPU shock has significantly greater influence on tourism demand than GDP shock. This suggests tourism demand studies should seriously consider routinely testing and incorporating such variable (EPU) in the models. Meanwhile, it is noticed that both uncertainty (EPU) shock and economic downturn (GDP shock) might sometimes bring opportunities and the possibility of tourism demand booms following the shocks (such as the case of inbound tourism demand of Japan and South Korea). Furthermore, the impulse response results in our study also help to develop a better understanding of the degree to which the impact of the shock may persist, and the characteristics and patterns of the recovery. This provides insights in determining the type of policy measures needed to intervene in the economy. Under the internationalized system, after the simulated shock, China’s EPU index recovers to the original level within approximately 2.5 years. However, it is shown that in all observed markets the impact can last for more than five years.

In regard to tourism demand research, our study highlights the importance of widening the sample frame to include not only the countries closely related in tourism flows but also in other economic activities (e.g., merchandise trade flows) at the global level. For example, the study of how shocks propagate in the tourism demand system should not be limited to those which are major tourism generating markets for a destination—a feature that has been the dominant focus in tourism demand studies. Uncertainty shocks from many markets, especially those which are globally significant economies, should be included in the analyses. Thus, this study demonstrates the benefits of utilizing VAR-type econometric techniques (e.g., FAVAR, PVAR, GVAR, etc.) to conduct global tourism analysis which relies on large scale and high-dimensional dataset. The impulse response analysis of a scenario-based economic policy uncertainty shock, which can be induced by global health crisis like COVID, provide insights on the range and extent of the impact, as well as the potential consequence (e.g., overshoot).

While this paper shows that the GVAR model provides an appropriate framework for examining tourism interdependence at the global scale, our study faces certain limitations. It is constrained by rich data input requirement. Due to data limitations, the visitor volume-based weighted matrix could only be applied to the regionalized system. The variables in the GVAR model are supported by aggregate data. Due to the limits of aggregate data, we do not have information about the change in tourism demand for any specific O-D pair under the EPU shock scenario. Thus, a disaggregated-level analysis of a foreign EPU shock on a country’s inbound or outbound tourism markets cannot be considered. When assessing the impacts of EPU, the source medium matters. An adverse shock to China’s EPU based on the Hong Kong newspaper yields larger impacts on international tourism demand, compared with the EPU shock based on the domestic-mainland Chinese newspapers. Newspaper-based EPU can contain editorial/publisher bias but we are limited in being able to independently verify the data. Data observations amid COVID crisis are not included in our GVAR estimation. Time series or econometric models are based on large volume of historical data, and such methods may not capture the recently unanticipated events like COVID (Zhang et al. 2021). Future research may consider mixed methods (e.g., the Delphi adjustment) with the GVAR model in order to better assess post-COVID tourism recovery trajectory.

Footnotes

Appendix

Table A4.

F-statistics for Testing the Weak-Exogeneity of the Market-Specific Foreign Variables and the Global Common Variable.

Market	$y_{it}^{*}$	${rex}_{it}^{(1) *}$	${epu}_{it}^{*}$	${rcre}_{it}^{*}$	${rdeb}_{it}^{*}$	$poi l_{t}$
AU	2.26	-	3.53	0.21	0.84	1.55
BR	0.98	-	0.92	1.02	3.04	0.76
CA	0.83	-	0.00	0.06	0.47	0.04
CN	0.12	-	1.77	0.21	0.41	0.88
DK	3.25	-	0.04	6.39*	1.96	3.43
EURO	0.09	-	0.10	0.10	2.28	0.55
HK	3.41	-	0.09	0.51	2.12	0.02
ID	1.56	-	4.09*	0.27	1.83	0.03
JP	0.29	-	4.46*	0.06	0.95	0.21
KR	0.04	-	0.01	2.02	1.69	0.25
MX	0.06	-	0.48	0.10	3.43	1.22
MY	2.05	-	0.03	0.41	0.14	0.31
NO	0.58	-	3.56	1.07	0.10	8.40*
NZ	3.61	-	0.81	0.01	0.32	0.26
PH	0.12	-	0.57	1.37	1.96	0.41
PL	2.57	-	0.00	0.15	0.21	0.14
RU	0.92	-	0.82	0.16	0.18	1.04
SE	1.20	-	0.48	0.37	0.74	1.88
SG	2.63	-	0.88	0.09	0.02	0.70
TH	0.63	-	0.22	0.62	0.13	0.69
TR	0.02	-	1.19	1.26	0.00	1.01
US	2.11	0.72	0.13	0.35	0.12	-

Note: Tests for India (IN), South Africa (ZA), and the UK are not conducted because no cointegrating relationship is found for the abovementioned markets.

p < 5%.

Acknowledgements

The authors would like to thank (i) the anonymous reviewers for their invaluable comments and suggestions; and (ii) the Macau Institute for Tourism Studies for library services.

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 1^st and 3^rd authors wish to acknowledge the financial support funded by University of Macau Start-Up Research Grant (File no. SRG2018-00116-FBA).

ORCID iDs

Rockie U Kei Kuok

Tay T.R. Koo

Notes

Author Biographies

Rockie U Kei Kuok is a PhD Candidate at the Faculty of Business Administration, University of Macau, and a full-time A320 airline pilot based in Macau. He is also a sessional academic with the School of Aviation, UNSW Sydney, and a Chartered Financial Analyst (CFA) Level 3 Candidate. His research interests are in tourism and aviation economics/econometrics.

Tay T.R. Koo has an academic background in geography and economics with emphasis on sustainable aviation and travel management. His research interest involves the study of aviation consumers, air route development, and spatial measurement methods. In addition to the undergraduate and postgraduate teaching portfolio in the area of sustainable aviation management, he has end-user focused projects from industry and government covering cognizant fields.

Christine Lim is Associate Professor and researcher in tourism demand modeling at University of Macau.

References

Abid

Rault

2020. “On the Exchange Rate and Economic Policy Uncertainty Nexus: A Panel VAR Approach for Emerging Markets.” Working Paper No. 8189, Center for Economic Studies and ifo Institute.

Akadiri

S. S.

Alola

A. A.

Uzuner

2020. “Economic Policy Uncertainty and Tourism: Evidence From the Heterogeneous Panel.” Current Issues in Tourism 23 (20): 2507–14.

Alam

M. R.

Istiak

2020. “Impact of US Policy Uncertainty on Mexico: Evidence From Linear and Nonlinear Tests.” The Quarterly Review of Economics and Finance 77: 355–66.

Altig

Baker

Barrero

J. M.

Bloom

Bunn

Chen

Davis

S. J.

Leather

Meyer

Mihaylov

, et al. 2020. “Economic Uncertainty Before and During the COVID-19 Pandemic.” Journal of Public Economics 191: 104274.

Assaf

A. G.

Song

Tsionas

M. G.

2019. “Modeling and Forecasting Regional Tourism Demand Using the Bayesian Global Vector Autoregressive (BGVAR) Model.” Journal of Travel Research 58 (3): 383–97.

Bachmann

Elstner

Sims

E. R.

2013. “Uncertainty and Economic Activity: Evidence From Business Survey Data.” American Economic Journal Macroeconomics 5 (2): 217–49.

Baker

S. R.

Bloom

Davis

S. J.

2012. “Has economic policy uncertainty hampered the recovery?” Working Paper (2012-003), Becker Friedman Institute for Research in Economics.

Baker

S. R.

Bloom

Davis

S. J.

2016. “Measuring Economic Policy Uncertainty.” The Quarterly Journal of Economics 131 (4): 1593–636.

Baker

S. R.

Bloom

Davis

S. J.

Terry

S. J.

2020. Covid-Induced Economic Uncertainty. NBER Working Paper 26983, National Bureau of Economic Research.

10.

Baker

S. R.

Bloom

Davis

S. J.

Wang

2013. “Economic Policy Uncertainty in China.” Unpublished paper, University of Chicago.

11.

Balli

Shahzad

S. J. H.

Salah Uddin

2018. “A Tale of Two Shocks: What Do We Learn From the Impacts of Economic Policy Uncertainties on Tourism?”Tourism Management 68: 470–5.

12.

Balli

Tsui

W. H. K.

2016. “Tourism Demand Spillovers Between Australia and New Zealand: Evidence From the Partner Countries.” Journal of Travel Research 55 (6): 804–12.

13.

Barrero

J. M.

Bloom

2020. “Economic Uncertainty and the Recovery.” Unpublished Manuscript. https://nbloom.people.stanford.edu/sites/g/files/sbiybj4746/f/jh_final.pdf

14.

Baxter

Kouparitsas

M. A.

2015. “Determinants of Business Cycle Comovement: A Robust Analysis.” Journal of Monetary Economics 52 (1): 113–57.

15.

Xin

2021. “Analysis of Economic Shock Effects Between China and ASEAN: An Empirical Study Based on Multinational VAR.” Emerging Markets Finance and Trade 57 (8): 2290–306.

16.

Bloom

2009. “The Impact of Uncertainty Shocks.” Econometrica 77 (3): 623–85.

17.

Bussiere

Chudik

Sestieri

2009. “Modelling Global Trade Flows: Results From a GVAR Model.” Working Paper No. 1087, European Central Bank.

18.

Caggiano

Castelnuovo

Figueres

J. M.

2020. “Economic Policy Uncertainty Spillovers in Booms and Busts.” Oxford Bulletin of Economics and Statistics 82 (1): 125–55.

19.

Cao

2016. “Modelling Economic Interdependencies of International Tourism Demand: The Global Vector Autoregressive Approach.” Doctoral diss., University of Surrey, Guildford.

20.

Cao

Song

2017. “Modelling the Interdependence of Tourism Demand: The Global Vector Autoregressive Approach.” Annals of Tourism Research 67: 1–13.

21.

Carrière-Swallow

Céspedes

L. F.

2013. “The Impact of Uncertainty Shocks in Emerging Economies.” Journal of International Economics 90 (2): 316–25.

22.

Cashin

Mohaddes

Raissi

2017. “China’s Slowdown and Global Financial Market Volatility: Is World Growth Losing Out?”Emerging Markets Review 31: 164–75.

23.

Cesa-Bianchi

Pesaran

M. H.

Rebucci

2017. “Uncertainty and Economic Activity: Identification Through Cross-Country Correlations.”https://conference.nber.org/confer/2017/SI2017/EFFE/Cesa-Bianchi_Pesaran_Rebucci.pdf

24.

Chan

Lim

McAleer

2005. “Modelling Multivariate International Tourism Demand and Volatility.” Tourism Management 26 (3): 459–71.

25.

Chang

C.-L.

Khamkaew

Tansuchat

McAleer

2011. “Interdependence of International Tourism Demand and Volatility in Leading ASEAN Destinations.” Tourism Economics 17 (3): 481–507.

26.

Chen

C.-M.

Hua

K.-T.

Chyou

J.-T.

Tai

C.-C.

2020. “The Effect of Economic Policy Uncertainty on Hotel Room Demand-Evidence From Mainland Chinese and Japanese Tourists in Taiwan.” Current Issues in Tourism 23 (12): 1443–8.

27.

Chudik

Pesaran

M. H.

2016. “Theory and Practice of GVAR Modelling.” Journal of Economic Surveys 30 (1): 165–97.

28.

Chudik

Smith

2013. “The GVAR Approach and the Dominance of the US Economy.” (Working Paper No. 136), Federal Reserve Bank of Dallas.

29.

Cuaresma

J. C.

Huber

Onorante

2019. “The Macroeconomic Effects of International Uncertainty.” Research Paper Series 2302, European Central Bank.

30.

Davis

S. J.

Liu

Sheng

X. S.

2019. “Economic Policy Uncertainty in China Since 1949: The View From Mainland Newspapers.” Working Paper. https://static1.squarespace.com/static/5e2ea3a8097ed30c779bd707/t/5f7f49d054a84229354fe9ab/1602177496854/EPU+in+China%2C+View+from+Mainland+Newspapers%2C+August+2019.pdf

31.

Dees

Mauro

F. D.

Pesaran

M. H.

Smith

L. V.

2007. “Exploring the International Linkages of the Euro Area: A Global VAR Analysis.” Journal of Applied Economics 22 (1): 1–38.

32.

Demir

Gozgor

2018. “Does Economic Policy Uncertainty Affect Tourism?”Annals of Tourism Research 69: 15–7.

33.

Dogru

Bulut

Sirakaya-Turk

2021. “Modeling Tourism Demand: Theoretical and Empirical Considerations for Future Research.” Tourism Economics 27 (4): 874–89.

34.

Dogru

Sirakaya-Turk

Crouch

G. I.

2017. “Remodeling International Tourism Demand: Old Theory and New Evidence.” Tourism Management 60: 47–55.

35.

Dragouni

Filis

Gavriilidis

Santamaria

2016. “Sentiment, Mood and Outbound Tourism Demand.” Annals of Tourism Research 60: 80–96.

36.

Farzanegan

M. R.

Gholipour

H. F.

Feizi

Nunkoo

Andargoli

A. E.

2021. “International Tourism and Outbreak of Coronavirus (COVID-19): A Cross-Country Analysis.” Journal of Travel Research 60 (3): 687–92.

37.

Feldkircher

Huber

Pfarrhofer

2020. “Factor Augmented Vector Autoregressions, Panel VARs, and Global VARs.” In Macroeconomic Forecasting in the Era of Big Data, edited by Fuleky

, 65–93. Cham: Springer.

38.

Fourie

Rosselló-Nadal

Santana-Gallego

2020. “Fatal Attraction: How Security Threats Hurt Tourism.” Journal of Travel Research 59 (2): 209–19.

39.

Gourio

Siemer

Verdelhan

2013. “International Risk Cycles.” Journal of International Economics 89 (2): 471–84.

40.

Gozgor

Lau

M. C. K.

Zeng

Yan

Lin

2021. “The Impact of Geopolitical Risks on Tourism Supply in Developing Economies: The Moderating Role of Social Globalization.” Journal of Travel Research. Published online April 8. doi:10.1177/00472875211004760

41.

Gozgor

Ongan

2017. “Economic Policy Uncertainty and Tourism Demand: Empirical Evidence From the USA.” International Journal of Tourism Research 19 (1): 99–106.

42.

Gross

2019. “Estimating GVAR Weight Matrices.” Spatial Economic Analysis 14 (2): 219–40.

43.

Han

Yin

2016. “Macroeconomic Policy Uncertainty Shocks on the Chinese Economy: a GVAR Analysis.” Applied Economics 48 (51): 4907–21.

44.

Haque

Magnusson

L. M.

Tomioka

2021. “Empirical Evidence on the Dynamics of Investment Under Uncertainty in the US.” Oxford Bulletin of Economics and Statistics 83 (5): 1193–217.

45.

Held

McGrew

Goldblatt

Perraton

2000. “Global Transformations: Politics, Economics and Culture.” In Politics at the Edge, edited by Carlsson

, 14–28. London: Palgrave Macmillan.

46.

Imbs

2004. “Trade, Finance, Specialization, and Synchronization.” The Review of Economics and Statistics 86 (3): 723–34.

47.

IMF (International Monetary Fund). 2005. “Balance of Payments Manual.”https://www.imf.org/external/pubs/ft/bopman/bopman.pdf (accessed November 7, 2021).

48.

IMF (International Monetary Fund). 2009. “Balance of Payments and International Investment Position Manual – Sixth Edition (BPM6).”https://www.imf.org/external/pubs/ft/bop/2007/pdf/bpm6.pdf (accessed August 3, 2021).

49.

IMF (International Monetary Fund). 2019. “Direction of Trade Statistics dataset.”https://data.imf.org/?sk=9D6028D4-F14A-464C-A2F2-59B2CD424B85 (accessed April 5, 2021).

50.

Inoue

Kaya

Ohshige

2015. “The Impact of China’s Slowdown on the Asia Pacific Region: An Application of the GVAR Model.” Working Paper No. 7442, World Bank Policy Research.

51.

Işık

Sirakaya-Turk

Ongan

2020. “Testing the Efficacy of the Economic Policy Uncertainty Index on Tourism Demand in USMCA: Theory and Evidence.” Tourism Economics 26 (8): 1344–57.

52.

Jin

Bao

Tang

2021. “Profiling and Evaluating Chinese Consumers Regarding post-COVID-19 Travel.” Current Issues in Tourism. Published online January 17. doi:10.1080/13683500.2021.1874313

53.

Kapetanios

Pesaran

M. H.

2005. “Alternative Approaches to Estimation and Inference in Large Multifactor Panels: Small Sample Results With an Application to Modelling of Asset Returns.” Working Paper No. 1416, Center for Economic Studies and ifo Institute.

54.

Kempa

Khan

N. S.

2017. “Spillover Effects of Debt and Growth in the Euro Area: Evidence From a GVAR Model.” International Review of Economics & Finance 49: 102–11.

55.

Khalid

Okafor

L. E.

Burzynska

2021. “Do Regional Trade Agreements Enhance International Tourism Flows? Evidence From a Cross-Country Analysis.” Journal of Travel Research. Published online July 14. doi:10.1177/00472875 211028321

56.

Khalid

Okafor

L. E.

Shafiullah

2020. “The Effects of Economic and Financial Crises on International Tourist Flows: A Cross-Country Analysis.” Journal of Travel Research 59 (2): 315–34.

57.

Khan

C.-W.

Xiao

Y.-D.

Zhu

Zhang

2021. “Trends in Tourism Under Economic Uncertainty.” Tourism Economics 27 (4): 841–58.

58.

Kožić

Sorić

Sever

2019. “Interdependence of International Tourism Demand for Mediterranean Countries: Impact of Demand Shocks.” International Journal of Tourism Research 21 (1): 97–107.

59.

Lee

C.-C.

Chen

M.-P.

2021. “Do Country Risks Matter for Tourism Development? International Evidence.” Journal of Travel Research 60 (7): 1445–68.

60.

Lim

1997. “Review of International Tourism Demand Models.” Annals of Tourism Research 24 (4): 835–49.

61.

Liu

Wang

2021. “Exploring the Influence of Economic Policy Uncertainty on the Relationship Between Tourism and Economic Growth With an MF-VAR Model.” Tourism Economics 27 (5): 1081–100.

62.

Luk

Cheng

Wong

2020. “Economic Policy Uncertainty Spillovers in Small Open Economies: The Case of Hong Kong.” Pacific Economic Review 25 (1): 21–46.

63.

MacKinnon

J. G.

Haug

A. A.

Michelis

1999. “Numerical Distribution Functions of Likelihood Ratio Tests for Cointegration.” Journal of Applied Economics 14 (5): 563–77.

64.

Martin

Crespo Cuaresma

2017. “Weighting Schemes in Global VAR Modelling: A Forecasting Exercise.” Letters in Spatial and Resource Sciences 10 (1): 45–56.

65.

McKercher

2021. “Can Pent-Up Demand Save International Tourism?”Annals of Tourism Research Empirical Insights 2 (2): 100020.

66.

Nguyen

C. P.

Schinckus

T. D.

2020. “Economic Policy Uncertainty and Demand for International Tourism: An Empirical Study.” Tourism Economics 26 (8): 1415–30.

67.

Ongan

Gozgor

2018. “Tourism Demand Analysis: The Impact of the Economic Policy Uncertainty on the Arrival of Japanese Tourists to the USA.” International Journal of Tourism Research 20 (3): 308–16.

68.

Ong

S. L.

Sato

2018. “Regional or Global Shock? A Global VAR Analysis of Asian Economic and Financial Integration.” The North American Journal of Economics and Finance 46: 232–48.

69.

Park

H. J.

Fuller

W. A.

1995. “Alternative Estimators and Unit Root Tests for the Autoregressive Process.” Journal of Time Series Analysis 16 (4): 415–29.

70.

Pesaran

M. H.

Schuermann

Weiner

S. M.

2004. “Modeling Regional Interdependencies Using a Global Error-Correcting Macroeconometric Model.” Journal of Business and Economic Statistics 22 (2): 129–62.

71.

Pesaran

M. H.

Shin

Smith

R. J.

2000. “Structural Analysis of Vector Error Correction Models With Exogenous I(1) Variables.” Journal of Econometrics 97 (2): 293–343.

72.

Prüser

Schlösser

2020. “The Effects of Economic Policy Uncertainty on European Economies: Evidence From a TVP-FAVAR.” Empirical Economics 58 (6): 2889–910.

73.

Reddy

M. V.

Boyd

S. W.

Nica

2020. “Towards a Post-Conflict Tourism Recovery Framework.” Annals of Tourism Research 84: 102940.

74.

Rogers

J. H.

2019. “How Well Does Economic Uncertainty Forecast Economic Activity?” Board of Governors of the Federal Reserve System Finance and Economics Discussion Series 2019-085.

75.

Rosselló

Becken

Santana-Gallego

2020. “The Effects of Natural Disasters on International Tourism: A Global Analysis.” Tourism Management 79: 104080.

76.

Seo

J. H.

Park

S. Y.

Boo

2010. “Interrelationships Among Korean Outbound Tourism Demand: Granger Causality Analysis.” Tourism Economics 16 (3): 597–610.

77.

Smith

L. V.

Galesi

2014. “GVAR Toolbox 2.0.” https://sites.google.com/site/gvarmodelling/home (accessed April 14, 2021).

78.

Song

2008. “Tourism Demand Modelling and Forecasting—A Review of Recent Research.” Tourism Management 29 (2): 203–20.

79.

Song

Cao

2018. “Tourism and Economic Globalization: An Emerging Research Agenda.” Journal of Travel Research 57 (8): 999–1011.

80.

Song

Qiu

R. T. R.

Park

2019. “A Review of Research on Tourism Demand Forecasting: Launching the Annals of Tourism Research Curated Collection on Tourism Demand Forecasting.” Annals of Tourism Research 75: 338–62.

81.

Sznajderska

2019. “The Role of China in the World Economy: Evidence From a Global VAR Model.” Applied Economics 51 (15): 1574–87.

82.

Tam

P. S.

2018. “Global Trade Flows and Economic Policy Uncertainty.” Applied Economics 50 (34-35): 3718–34.

83.

The World Bank. 2019. World Development Indicators. Washington, DC: The World Bank. https://data.worldbank.org/indicator/NY.GDP.MKTP.PP.CD (accessed April 5, 2021).

84.

Tiwari

A. K.

Das

Dutta

2019. “Geopolitical Risk, Economic Policy Uncertainty and Tourist Arrivals: Evidence From a Developing Country.” Tourism Management 75: 323–7.

85.

Trung

N. B.

2019. “The Spillover Effects of US Economic Policy Uncertainty on the Global Economy: A Global VAR Approach.” The North American Journal of Economics and Finance 48: 90–110.

86.

Tsui

W. H. K.

Balli

Tan

D. T. W.

Lau

Hasan

2018. “New Zealand Business Tourism: Exploring the Impact of Economic Policy Uncertainties.” Tourism Economics 24 (4): 386–417.

87.

United Nations. 2010. “International Recommendations for Tourism Statistics 2008.”https://unstats.un.org/unsd/publication/seriesm/seriesm_83rev1e.pdf (accessed November 7, 2021).

88.

UNWTO (United Nations World Tourism Organization). 2019. “Yearbook of Tourism Statistics Dataset.”https://www.e-unwto.org/toc/unwtotfb/current (accessed August 15, 2020).

89.

UNWTO (United Nations World Tourism Organization). 2020. “International Tourism Highlights, 2020 Edition.”www.e-unwto.org/doi/pdf/10.18111/9789284421152 (accessed August 3, 2021).

90.

Vatsa

2020. “Comovement Amongst the Demand for New Zealand Tourism.” Annals of Tourism Research 83: 102965.

91.

Wong

2017. “China’s Current Account: External Rebalancing or Capital Flight?” International Finance Discussion Paper No. 1208, Federal Reserve System.

92.

T.-P.

H.-C.

2019. “Causality Between European Economic Policy Uncertainty and Tourism Using Wavelet-Based Approaches.” Journal of Travel Research 58 (8): 1347–56.

93.

T.-P.

H.-C.

2020. “Global Economic Policy Uncertainty and Tourism of Fragile Five Countries: Evidence From Time and Frequency Approaches.” Journal of Travel Research 60 (5): 1061–73.

94.

Zhang

Song

Wen

Liu

2021. “Forecasting Tourism Recovery Amid COVID-19.” Annals of Tourism Research 87: 103149.