Impact of Economic Policy Uncertainty and Pandemic Uncertainty on International Tourism: What do We Learn From COVID-19?

Abstract

Uncertainty is an overarching aspect of life that is particularly pertinent to the present COVID-19 pandemic crisis; as seen by the pandemic’s rapid worldwide spread, the nature and level of uncertainty have possibly increased due to the possible disconnects across national borders. The entire economy, especially the tourism industry, has been dramatically impacted by COVID-19. In the current study, we explore the impact of economic policy uncertainty (EPU) and pandemic uncertainty (PU) on inbound international tourism by using data gathered from Italy, Spain, and the United States for the years 1995–2021. Using the Quantile on Quantile (QQ) approach, the study confirms that EPU and PU negatively affected inbound tourism in all states. Wavelet-based Granger causality further reveals bi-directional causality running from EPU to inbound tourism and unidirectional causality from PU to inbound tourism in the long run. The overall findings show that COVID-19 has had a strong negative effect on tourism. So resilient skills are required to restore a sustainable tourism industry.

Introduction

The world woke up to an alien and infectious coronavirus, now called COVID-19, that started spreading by mid-December 2019. Wuhan, one of China’s most populous cities, was the first to record an index case (Meo & Abd Karim, 2022; Si, Lu, & Aziz, 2021, 2021b), after which, like a wildfire, COVID-19 has spread worldwide, with only a few countries yet to record a case. Similar to severe acute respiratory syndrome (SARS), COVID-19 is an airborne infection that is highly contagious among humans. In January 2020, the Chinese government alerted the world and its citizens of the horrendous effects of the ravaging pandemic, followed by preventive measures like business and school closures, individual and community quarantines, and temporary closure of markets and leisure centers across China.

The pandemic has since metamorphosed into a global health emergency of enormous concern, to the extent of the World Health Organization (WHO) terming it an “unprecedented global pandemic.” In March 2020, the United States, Spain, Italy, and other European countries had become epicenters of COVID-19, with strict restrictions on human movement across and within countries. As of 28 April 2020, there were about 2,954,222 confirmed cases globally with 202,597 deaths, while as of 26 May 2020, there were already 5,459,528 recorded cases (European Centre for Disease Prevention and Control, 2020). Table 1 presents the number of confirmed cases and deaths by region (WHO, 2020).

Table 1.

WHO COVID-19 Situation Reports of 28 April 2020.

Regions of the World	Confirmed Cases	Deaths in Numbers
European region	1386693	126429
Region of the Americas	1179607	60211
Eastern Mediterranean region	171238	7148
Western pacific region	145385	5998
South-east Asia region	48348	1917
African region	22239	881

Sources: Authors computation from WHO (2020).

For many developed and developing countries, the tourism sector is a major source of employment, government revenue, and foreign exchange earnings (Aziz et al., 2020; Zhu et al., 2021). Without this vital lifeline, many countries might experience a dramatic contraction in GDP and a rise in unemployment. However, according to the United Nations World Trade Organisation (2020), tourism was the worst hit of all major economic sectors globally. The industry is expected to have contracted by 20%–30% in 2020, which is equivalent to 300 to 450 US$ billion, about one-third of the total tourism receipt generated annually (US$ 1.5 trillion). Other sectors are expected to recover once restrictions are eased. Still, the effect of COVID-19 is expected to be long-lasting on international tourism due to reduced consumer confidence and the possibility of longer restrictions on travel (UNCTAD, 2020).

In the 10 most visited countries, such as Spain, Italy, China, and the United States (Koçak et al., 2020), the economic shock of tourism in these countries could be far-reaching. From the Spanish Costa pubs to the chic resorts of Italy and the Grand Canyon in the United States, no one knows how tourism will fare. The losses are already alarming. According to the European Commission (2020), restaurants and hotels will lose half their income in 2020. The banking group UBS has pointed out that tourism revenue fell by 95% and 77% in March 2020 in Italy and Spain, respectively. The economic policy uncertainty (EPU) also impacts tourism development (Lim & Won, 2020), causing delays or even cancellations of travel plans due to security, safety, and stability issues, and thus periods of increasing uncertainty can discourage or make individuals reluctant to travel overseas (Demir & Gözgör et al., 2018). If uncertainty and instability are evident in a country, a household’s consumption decreases, which could reduce and delay tourism expenditure, as tourism and travels expenditures are generally very sensitive to periods of uncertainty (Balsalobre-Lorente et al., 2022a, 2022b; Chaudhry et al., 2021; Doğan et al., 2022; Jahanger et al., 2022; Jiang et al., 2022; Sinha et al., 2022; Usman & Balsalobre-Lorente, 2022).

There is uncertainty surrounding almost every aspect of the COVID-19 crisis: On the epidemiological side, uncertainties include the infectiousness and lethality of the virus (Fauci et al., 2020), the time needed to develop and deploy vaccines (Koirala et al., 2020), and the duration and effectiveness of social distancing (Anderson et al., 2020). On the economic side, uncertainties include the near-term economic impact of the pandemic and policy responses (Baqaee et al., 2020), the speed of recovery as the pandemic recedes (Congressional Budget Office, 2020), and the extent to which pandemic-induced shifts in consumer spending patterns, business travel, and working from home will persist (Barrero et al., 2020). Hence, a major aspect for policy-makers of learning from this experience is the need to consider all the dimensions of uncertainty simultaneously or in sequence to ensure a more unified approach to dealing with such global disasters in the future. Therefore, the negative effect of COVID-19 on the economy makes it imperative to revisit the impact of pandemic uncertainty on tourism.

Moreover, many studies have explored the association between pandemics and tourism but ignored the non-linear trend of the variables. For instance, from the methodological perspective, most earlier studies adopted a linear approach (Sharif et al., 2021). Some merely presented information related to time while ignoring frequency-based information (Sharif et al., 2021). However, the cause of non-linearity in time series examination arises from ignoring the frequency information (Pal & Mitra, 2017). The previous studies thus yielded different results. Many other studies have employed conventional econometric approaches such as the OLS, ARDL, NARDL, VECM, pairwise granger causality. None of these traditional methods have comprehensively examined the relationships between tourism and its determinants. Furthermore, many studies are concerned with the impact of the crisis on airlines (Henderson, 2003), hotels (Chien & Law, 2003), restaurants, and travel agents (Alan et al., 2006); others have focused on the effects of pandemics like Severe Acute Respiratory Syndrome (McKercher & Chon, 2004), influenza (Monterrubio, 2010), and even bed bugs (Liu et al., 2015) but ignored the effect of economic and pandemic uncertainty on inbound international tourism in the most visited tourist destinations.

To fill this gap, the current study attempts to explore the effect of EPU and pandemic uncertainty (PU, specifically that associated with COVID-19) on tourism. The study contributes to the literature in the following ways. First, the current study is the first to estimate the relationship among the desired variables. Second, this study selects the USA, Italy, and Spain because these countries have been highly affected by COVID-19. Third, most researchers have investigated the nexus of tourism demand and its determinants in a linear framework. However, linear models cannot analyze structural changes and short-term volatilities (Po & Huang, 2008). As many macroeconomic variables exhibit non-linearities (Meo et al., 2018), the current study employs the QQ (Sim & Zhou, 2015) approach to estimate the effect of EPU and PU on tourism in quantiles following earlier researchers (Arain et al., 2020; Shahzad et al., 2017). Furthermore, the study employs a wavelet-based Granger causality test to consider the time-frequency-varying nature of the relationship between EPU, PU, and tourism, as this method aids in analyzing the association between variables within time scales and frequency bands and yields robust estimates for a small sample size (Sharif et al., 2021). Therefore, the current study explores the impact of EPU and PU on inbound tourism for the full sample ranging from 1995Q1 to 2021Q4 and then focuses on the time of COVID-19 ranging from 01-01-2020 to 30-05-2020 to ascertain whether the relationship changed during the COVID-19 period.

The remainder of this study is arranged as follows: Sec. 2 reviews the literature. Methodology and data source are addressed in Sec. 3. The results are reported and discussed in Sec. 4, while Sec. 5 concludes the study with relevant policy suggestions.

Literature Review and Theoretical Framework

The current literature provides an exposition of the effects of EPU and PU on tourism. These uncertainties are allied with unanticipated actions that influence the demand of tourism, for example, “conflicts, terrorism, natural catastrophes, epidemics, contagions, and financial crises” (Song & Lin, 2010). Tourism is more susceptible than most industries to disasters or crises (Cró & Martins, 2017). Most studies have explored tourism’s relationship with various forms of uncertainty, such as economic uncertainty (Zaman et al., 2016) and financial uncertainty. Such uncertainties caused significant damage to international tourism by a severe decline in tourist arrivals and are a culprit in the shrinking size of overall tourism activities (Webber et al., 2010).

Moreover, regarding pandemic uncertainties, the worldwide outbreak of infectious diseases or pandemics can bring the world to a standstill. In the context of crises or pandemics, the tourism industry is predicted to be particularly vulnerable, and the UNWTO (2020), in its report, documented that pandemics caused a decline of 20%–30% in tourist arrivals. The overall evidence indicates that the pandemic led to a decrease in tourism-related expenditures due to reduced consumption by households on tourism, transportation, and leisure activities, and a higher demand for medical services. Joo et al. (2019) also found a correlation between 2.1 million drops in visitors and a US$2.6 billion loss in the 2015 MERS (Middle East respiratory syndrome) epidemic in the Republic of Korea. Likewise, in 2019, the world economy was again devastated by COVID-19 (coronavirus disease 2019), which greatly jeopardized the tourism industry. The tourism and transport-related industries were severely impacted by this pandemic. According to Baker (2020), the Chinese hotel market resulted in a decline of 71% in occupancy rate over just three days in January 2020. Yang et al. (2020) add that COVID-19 affects the health status and deteriorates the labor productivity of workers in tourism activities.

In other countries, Karim and Haque (2020) reported a negative impact of COVID-19 on the Malaysian tourism and hospitality industry and argued that as airline companies are continuing to stop all its operations to and from Malaysia, the tourism industry will continue to suffer negative effects. Wang et al. (2022) also found that pandemic risks decrease tourist arrivals. Likewise, Karabulut et al. (2020) reported that countries are economically affected by decreasing tourist arrivals. More recently, Yan et al. (2021) reported that air quality and COVID-19 have affected Hawaii’s tourism growth. A recent study by Agboola et al. (2021) also revealed the striking impacts of COVID-19 on the economy of the Kingdom of Saudi Arabia. At the micro-level, various other studies have explored this phenomenon and proposed various suggestions to counteract the aftermath of COVID-19 (Alhassan et al., 2021a, 2021b; Rahman et al., 2021; Hossain et al., 2021). Golets et al. (2021) indicated that perceived COVID-19 severity, the perceived probability of infection, and the expected duration of the pandemic are significant predictors of travel intentions.

Based on the above discussion, worldwide tourism is clearly hit hard by pandemics; virus disease negatively affects tourism sectors, and thus affects a country’s GDP, as the tourism sector is a major earning source. History has shown that epidemics and pandemics have an immediate impact on hotels and restaurants, airline industries, travel agencies, etc., due to international travel restrictions, media coverage, and government measures. However, to the best of the authors’ knowledge, this is the first study that empirically examined the impact of EPU and PU on tourism in selected economies based on QQ and wavelet causality. Figure 1(a) illustrates the theoretical framework of economic uncertainty, pandemic, and international tourism, showing the mechanism whereby economic activities were stopped due to pandemics and uncertainty, which further affected different economic sectors, including travelling and tourism.

Figure 1.

Theoretical framework: Uncertainties, pandemic, and international tourism.

Modelling, Data, and Methodology

Model Specification

In specifying the theoretical models, we emphasize the objectives of the study. The study’s first objective is to examine the impact of EPU and PU on tourism, while the second objective is to investigate the effect of EPU and the COVID-19 pandemic on tourism. Hence, the theoretical models are specified as:

Model 1

$$TOU=f\left(EPU\hspace{0.17em},PU\right)$$ (1)

When equation (1) is transformed into a multi-linear equation, the model becomes

$${TOU}_t={\beta }_0+{\beta }_1{EPU}_t+{\beta }_2{PU}_t+{\epsilon }_t$$ (2)

Model 2

$$TOU=f\left(EPU\hspace{0.17em},COVID-19\right)$$ (3)

From equation (3), we have

$${TOU}_t={\beta }_0+{\beta }_1{EPU}_t+{\beta }_2{COVID-19}_t+{\epsilon }_t,$$ (4)

where TOU denotes inbound tourism, EPU denotes economic policy uncertainty, PU denotes pandemic uncertainty, and COVID-19 denotes the number of infected cases of a novel COVID-19 in the countries of interest.

Data Specification

Before pre-estimation and post-estimation analysis can be performed in any time-series study, data for the variables under investigation must be sourced. This study examines the relationship between purposed variables using the two models specified in Equations (2) and (4). The first model is designed to explore EPU and PU’s impact on TOU using a dataset of 1995Q1 to 2021Q4. In contrast, the second model is designed to examine the effects of EPU and COVID-19 on TOU based on daily observation from 2020Q1 to 2021Q4 (a period associated with the coronavirus spread). Data on tourism, EPU, PU,¹ and COVID-19 were obtained from WDI and World Uncertainty Index. Table 2 shows the sources of the data.

Table 2.

Data Description.

Variables	Variables’ description	Measurement	Data sources
TOU	Tourism	Number of tourist arrivals	WDI and tradingeconomics.com/
EPU	Economic policy uncertainty index	Index	https://worlduncertaintyindex.com/
PU	Pandemic uncertainty index	Index	https://worlduncertaintyindex.com/
COVID-19	COVID-19	Number of infected cases	WHO

Wavelet Decomposition Method

The wavelet decomposition method helps analyze the causality between the dependent and independent variables (Sharif et al., 2017; Raza et al., 2017; Boubaker et al., 2017; Raza et al., 2017). Both the time and frequency domains are combined in wavelet analysis. Although the wavelet method allows the decomposition of time series data into different timescales, it can also preserve the time series data when the frequency of decomposition is carried out on the time data (Chu et al., 2016). Apart from these appealing merits of wavelet analysis, the primary reason for employing this econometric method is that the estimated results of the wavelet method remain unaffected by the size of the data sample. As a result, the challenge of limited data and conclusions drawn from spurious results is obviated by employing the wavelet method in this study. Following Chu et al. (2016), the wavelet decomposition method of a time series y(t) can be shown as follows

$$\mathrm{y}\left(\mathrm{t}\right)={\displaystyle \sum }_{\mathrm{k}}\hspace{0.17em}\hspace{0.17em}{\mathrm{d}}_{1,\mathrm{k}}{\mathrm{\Psi }}_{1,\mathrm{k}}\left(\mathrm{t}\right)+\dots .\hspace{0.17em}{\displaystyle \sum }_{\mathrm{k}}^{\mathrm{i}}\hspace{0.17em}\hspace{0.17em}{\mathrm{d}}_{\mathrm{j}-1,\mathrm{k}}{\mathrm{\Psi }}_{\mathrm{j}-1,\mathrm{k}}\left(\mathrm{t}\right)+{\displaystyle \sum }_{\mathrm{k}}^{\mathrm{i}}\hspace{0.17em}{\mathrm{d}}_{\mathrm{j},\mathrm{k}}{\mathrm{\Psi }}_{\mathrm{j},\mathrm{k}}\left(\mathrm{t}\right)+{\displaystyle \sum }_{\mathrm{k}}^{\mathrm{i}}\hspace{0.17em}{\mathrm{S}}_{\mathrm{j},\mathrm{k}}{\mathrm{\varphi }}_{\mathrm{j},\mathrm{k}}\left(\mathrm{t}\right)$$ (5)

In equation (5), the scaling parameter is denoted by j while scaling coefficients are denoted by S_j,k. k is the translation parameter ranging from 1 to the number of coefficients specified. d_j,k are the detail coefficients. The father wavelet and mother wavelet are represented by ϕ and ψ, respectively.

The basic functions that define the sequence of coefficients are set up by the father wavelets and mother wavelets. Consequently, we have

$${\mathrm{\varphi }}_{\mathrm{j},\mathrm{k}}\left(\mathrm{t}\right)=\frac{1}{2^{\mathrm{j}/2}}\hspace{0.17em}\mathrm{\varphi }\mathrm{j},\mathrm{k}\left(2^{-\mathrm{j}}\hspace{0.17em}\mathrm{t}-\mathrm{k}\right)\hspace{0.17em}\mathrm{w}\mathrm{i}\mathrm{t}\mathrm{h}\int \mathrm{\varphi }\left(\mathrm{t}\right)\mathrm{d}\mathrm{t}=1$$ (6)

The father wavelet is formally defined in equation (6). The father wavelet integrates into one, and it denotes the very long-scale smooth components of the signal

$${\mathrm{\Psi }}_{\mathrm{j},\mathrm{k}}\left(\mathrm{t}\right)=\frac{1}{2^{\mathrm{j}/2}}\hspace{0.17em}{\mathrm{\Psi }}_{\mathrm{j},\mathrm{k}}\left(2^{-\mathrm{j}}\hspace{0.17em}\mathrm{t}-\mathrm{k}\right)\hspace{0.17em}\mathrm{w}\mathrm{i}\mathrm{t}\mathrm{h}\int \mathrm{\Psi }\left(\mathrm{t}\right)\mathrm{d}\mathrm{t}=0$$ (7)

The mother wavelet is defined in equation (7). The deviations occurring in the smooth (scaling) components are represented by the mother wavelet. In addition, the mother wavelets integrate to zero.

$${\mathrm{s}}_{\mathrm{j},\mathrm{k}}=\int \mathrm{y}\left(\mathrm{t}\right){\mathrm{\varphi }}_{\mathrm{j},\mathrm{k}}\left(\mathrm{t}\right)\mathrm{d}\mathrm{t}$$ (8)

Equation (8) is used to approximate the scaling (s_j,k), referring to the scaling coefficients that were derived from the father wavelet. It has a maximum scale of 2^j

$${\mathrm{d}}_{\mathrm{j},\mathrm{k}}=\int \mathrm{y}\left(\mathrm{t}\right){\mathrm{\Psi }}_{\mathrm{j},\mathrm{k}}\left(\mathrm{t}\right)\mathrm{d}\mathrm{t},\hspace{0.17em}\mathrm{j}=1,\hspace{0.17em}2,\hspace{0.17em}.\hspace{0.17em}.\hspace{0.17em}.,\hspace{0.17em}\mathrm{J}$$ (9)

Equation (9) represents (d_j,k), the detail coefficients, which capture the higher-frequency oscillations. The detail coefficients are derived from the mother wavelet at all scales from 1 to j. In this study, the Maximum Overlap Discrete Wavelet Transform (MODWT) is used to obtain the scaling and detail coefficients because of its ability to keep the sample size values in all wavelet decomposition scales, which Discrete Wavelet Transform (DWT) would have discarded. Mishra et al. (2019) also pointed out that the detail and scaling (smooth) coefficients obtained from MODWT are related to zero phase filters. Thus, the MODWT estimator is also known to be asymptotically efficient. Besides, MODWT helps in the full utilization of information in any time series with respect to location in time and specific time horizon (Bouri et al., 2017). This implies that the application of MODWT helps to align the features of the original time series with the characteristics of Multi-Resolution Analysis (MRA). The decomposition of a time series to J levels with MODWT, however, requires the application of J pairs of filters, and the filtering operation at the jth level consists of applying a rescaled father wavelet to yield a set of detail coefficients. In contrast, the mother wavelet is rescaled to give a set of scaling coefficients (Chu et al., 2016)

$${\mathrm{S}}_{\mathrm{j},\mathrm{k}}={{\displaystyle \sum }}_{\mathrm{k}}\hspace{0.17em}{\mathrm{s}}_{\mathrm{j},\mathrm{k}}{\mathrm{\varphi }}_{\mathrm{j},\mathrm{k}}\left(\mathrm{t}\right)$$ (10)

$${\mathrm{D}}_{\mathrm{j},\mathrm{k}}={{\displaystyle \sum }}_{\mathrm{k}}\hspace{0.17em}{\mathrm{d}}_{\mathrm{j},\mathrm{k}}{\mathrm{\Psi }}_{\mathrm{j},\mathrm{k}}\left(\mathrm{t}\right),\mathrm{j}=1,\hspace{0.17em}2.\hspace{0.17em}.\hspace{0.17em}.\hspace{0.17em}.\hspace{0.17em}.,\hspace{0.17em}\mathrm{J}$$ (11)

Given these coefficients in equation (10) and (11), the wavelet series approximation of the original time series y(t) defined in equation (5) becomes

$$\mathrm{y}\left(\mathrm{t}\right)={\mathrm{D}}_1\left(\mathrm{t}\right)+.\hspace{0.17em}.\hspace{0.17em}.+{\mathrm{D}}_{\mathrm{j}-1}\left(\mathrm{t}\right)+{\mathrm{D}}_{\mathrm{j}}\left(\mathrm{t}\right)+\mathrm{S}\mathrm{j}\left(\mathrm{t}\right)$$ (12)

Equation (12) indicates that the recomposed series in the time domain of j is denoted by D_j and D_j. Equation (12) computes the jth level wavelet detail connected with the variations in the series at scale λ_j. S_j is defined as the cumulative sum of alterations at each level. As j increases, S_j becomes increasingly smooth (Gencay et al., 2002). Since wavelet decomposition requires the application of filtering for a robust and reliable outcome, the least asymmetric filter of length eight (LA8) is adopted because it is widely known to yield smoother coefficients with better uncorrelatedness across scales than the Haar filter.

Quantile-on-Quantile Approach (QQ)

Since Koenker and Bassett (1978) officially introduced quantile regression analysis (QRA) in their seminal paper “Regression quantile” in Econometrica in 1978 (Merilina & Domenico, 2018), it has become one of the most widely used econometric methods because of its ability to capture different time-varying of the dependent variable. Quantile regression analysis is also widely known to perform better than the classical linear correlation/regression counterpart in the generation of a more precise and accurate result of the impact of covariates on the regress and in time-varying degree and structure of dependence (Koenker, 2005). In addition to the merit of being a median analysis, QRA is useful in generating information on tail dependence, which aids the capturing of the normal phase of the dependent variable.

Despite the benefits of QRA as an econometric model, it cannot capture dependency in its entirety, especially in a situation of uncertainty (Sim & Zhou, 2015). Even though the heterogeneous relationship between covariates and dependent variables can be captured by quantile regression analysis, the use of the analysis is limited in this study because it neglects the possibility of uncertainty associated with the covariates on tourism. To overcome the challenges associated with the QRA and avoid making inferences from spurious estimation, QQ is employed in this study, as it enables the modelling of the quantile of tourism and its various frequencies to quantile the covariate uncertainty.

At each point of the distributions of tourism and distribution of its covariates, the relationship between the variables could vary across their respective distributions, allowing their dependence to be captured in its entirety. Thus, an absolute, clearer, and robust picture of dependency is obtainable with the QQ approach. Currently, there are two approaches to modelling QQ regression. The first is the triangular system equations, while the second approach is focused on a single equation regression approach. The first approach is based on Ma and Koenker’s (2005) model, the second on Sim et al. (2015). However, the single equation approach proposed by Sim et al. (2015) is adopted because it better aligns with the objectives of our study.

As previously mentioned, the specification of models for this study is purely based on its objectives. Since the primary focus is on model 1 and model 2 as stated in equations (1) and (3), respectively, “X” will be used to denote the covariate in the respective model. This will help to prevent awkwardness and unnecessary repetition in modelling the QQ approach. Given that the quantile of tourism is denoted by the superscript “θ” and the error term (with a zero θ-quantile) is denoted by ℰ_t^θ, the Quantile-on-Quantile equation can be formally expressed as follows

$${\mathrm{T}\mathrm{O}\mathrm{U}}_{\mathrm{t}}={\mathrm{\beta }}^{\mathrm{\theta }}{\mathrm{X}}_{\mathrm{t}}+{\mathcal{E}}_{\mathrm{t}}^{\mathrm{\theta }}$$ (13)

In equation (13), the relationship function (β^θ) remains unknown between the variables because we have neither perfect nor previous knowledge of the interlink between TOU (tourism) and changes in X. Thus, to examine the linkage between the θ-quantile of TOU and the θ-quantile of covariate X (i.e., X^τ), the relationship function (β^θ) has to be linearized by taking its first-order Taylor expansion around X^τ. As a result, we have

$${\mathrm{\beta }}^{\mathrm{\theta }}{\mathrm{X}}_{\mathrm{t}}\mathrm{\approx }{\mathrm{\beta }}^{\mathrm{\theta }}{\mathrm{X}}^{\mathrm{\tau }}\mathrm{+}{\mathrm{\beta }}^{\mathrm{\theta }\mathrm{ˊ}}\mathrm{\left(}{\mathrm{X}}^{\mathrm{\tau }}\mathrm{\right)}\mathrm{\left(}{\mathrm{X}}_{\mathrm{t}}\mathrm{-}{\mathrm{X}}^{\mathrm{\tau }}\mathrm{\right)}$$ (14)

In line with Sim and Zhou (2015), β^θX^τ and β^θˊ(X^τ) can be redefined as β₀(θX,τ) and β₁(θX,τ), respectively. Thus, equation (14) can be rewritten as

$${\mathrm{\beta }}^{\mathrm{\theta }}{\mathrm{X}}_{\mathrm{t}}\mathrm{\approx }{\mathrm{\beta }}_{\mathrm{0}}\mathrm{\left(}\mathrm{\theta }\mathrm{X}\mathrm{,}\mathrm{\tau }\mathrm{\right)}\mathrm{+}{\mathrm{\beta }}_{\mathrm{1}}\mathrm{\left(}\mathrm{\theta }\mathrm{X}\mathrm{,}\mathrm{\tau }\mathrm{\right)}\mathrm{\hspace{0.17em}}\left({\mathrm{X}}_{\mathrm{t}}\mathrm{-}{\mathrm{X}}^{\mathrm{\tau }}\mathrm{\right)}$$ (15)

When equation (15) is substituted into equation (13), it becomes

$${\mathrm{T}\mathrm{O}\mathrm{U}}_{\mathrm{t}}\mathrm{=}{\mathrm{\beta }}_{\mathrm{0}}\mathrm{\left(}\mathrm{\theta }\mathrm{X}\mathrm{,}\mathrm{\tau }\mathrm{\right)}\mathrm{+}{\mathrm{\beta }}_{\mathrm{1}}\mathrm{\left(}\mathrm{\theta }\mathrm{X}\mathrm{,}\mathrm{\tau }\mathrm{\right)}\mathrm{\hspace{0.17em}}\mathrm{\left(}{\mathrm{X}}_{\mathrm{t}}\mathrm{-}{\mathrm{X}}^{\mathrm{\tau }}\mathrm{\right)}\mathrm{+}{\mathcal{E}}_{\mathrm{t}}^{\mathrm{\theta }}$$ (16)

Equation (16) enables us to analyze the impact of various frequencies of the covariates (EPU, PU, and COVID-19) on TOU.

Empirical Results

Summary Statistics

Table 3 reports the summary statistics of tourist arrivals, EPU, and PU from 1995Q1 to 2021Q4. The USA has the highest value of maximum tourist arrivals while Spain has the lowest. Although the Jarque–Berra test shows non-normality in the data, the QQ approach nonetheless gives robust estimates.

Table 3.

Summary Statistics.

	USTOU	ITTOU	SPTOU	USEPU	ITEPU	SPEPU	USPU	ITPU	SPPU
Mean	591587	442878	573005	3.01	3.52	2.93	0.68	0.26	0.27
Median	553533	429622	559059	2.00	2.00	2.00	0.00	0.00	0.00
Maximum	850416	630752	840526	15.00	15.00	12.00	13.78	20.24	19.67
Minimum	408354	337693	312575	0.00	0.00	0.00	0.00	0.00	0.00
Std. Dev.	132979	805125	131046	2.78	3.18	2.55	2.46	2.12	2.10
Skewness	0.44	0.88	0.58	1.47	1.08	1.03	3.78	8.89	8.54
Kurtosis	1.72	3.01	2.76	5.71	3.86	3.87	17.02	83.04	77.49
Jarque–Bera	9.69	12.63	5.61	64.29	21.96	20.27	1015.99	26891.15	23364.41
Probability	0.00	0.00	0.06	0.00	0.00	0.00	0.00	0.00	0.00

Source: Authors Computation.

Further, the time paths of tourists’ arrivals, EPU, and PU for the USA, Italy, and Spain are portrayed in Figure 2. It is apparent that there is an upward trend of tourists’ arrivals in the late 2010s for all the economy, but the USA has the highest number of tourists’ arrivals. Similarly, the USA recorded the lowest tourist arrivals from 2001 to 2003. This corresponds with a period when the country experienced a recession. It was also observed that EPU is highly erratic, with periodic spikes for all the economies, but Italy records the highest spikes (in 2012) among the economies. Pandemic uncertainty plot reveals zero graphs with few plateaus. This indicates that COVID-19 is an unexpected outbreak, and the uncertainty regarding the COVID-19 epidemic is a great concern to tourists and policymakers worldwide.

Figure 2.

Time Series trend of tourists’ arrivals, EPU, and PU for the USA, Italy, and Spain.

Quantile Unit Roots

We began with a quantile unit root test before proceeding to the quantile-on-quantile regression analysis. The findings of the quantile unit root test are presented in Table 4. Traditional unit root tests yield average results, whereas the quantile unit root test yields more detailed information about the variables’ stationarity (Chowdhury et al., 2021). To overcome the issue of serial correlation, we used two lags in our analysis. The results show that at 5% level of significance, all of the variables are non-stationary.

Table 4.

Quantile Unit Root Results.

Quintiles	USTOU		ITTOU		SPTOU
	α(τ)	t-stats	α(τ)	t-stats	α(τ)	t-stats
0.05	0.55	−2.53	0.99	−2.70	0.78	−1.64
0.25	0.61	−1.87	0.90	−2.64	0.79	−1.78
0.50	0.61	−2.32	0.15	−2.08	0.77	−1.89
0.75	0.44	−2.53	0.07	−1.12	0.75	−1.69
0.95	0.25	2.22	0.05	−1.46	0.52	−3.15
	USEPU		ITEPU		SPEPU
	α(τ)	t-stats	α(τ)	t-stats	α(τ)	t-stats
0.05	0.50	−1.45	0.04	−2.86	0.10	−2.87
0.25	0.53	−1.84	0.03	−2.00	0.01	−1.53
0.50	0.50	−1.28	0.05	−2.06	0.03	−1.30
0.75	0.56	−1.69	0.06	−1.19	0.06	−1.69
0.95	0.53	−2.64	0.04	−2.28	0.01	−1.77
	USPU		ITPU		SPPU
	α(τ)	t-stats	α(τ)	t-stats	α(τ)	t-stats
0.05	0.69	−2.23	0.88	0.55	0.71	−1.77
0.25	0.71	−2.56	0.76	−1.33	0.69	−2.23
0.50	0.76	−2.33	0.78	−1.13	0.77	−1.89
0.75	0.77	−1.82	0.76	−1.32	0.75	−1.69
0.95	0.78	−1.64	0.71	−1.77	0.71	−1.77

Note: The table indicates the persistence values of the estimates α(τ) and t-statistics of the quantile unit root test proposed by Koenker and Xiao (2004) and Galvao (2009). The bold values of t-statistics show the rejection of the null hypothesis H0 = 1 at 5% level of significance.

Quantile-Based Granger Causality Test Results

The results of the Granger causality test in quantiles are presented in Table 5. The causality between EPU and tourist arrivals in Italy and Spain is bidirectional at 5% significant levels in all quantiles. Contrary to Italy and Spain’s economies, at the low quantiles (0.5–0.10), there is no causality between EPU and tourist arrivals in the United States of America because the economy is larger than those of its counterparts. Consequently, US policymakers’ decisions are usually delayed due to economic factors, which creates operational lags in the effect of EPU on tourism. The previous findings of Akadiri et al. (2020) also endorsed a two-way causality between EPU and tourism in the economies of France, Ireland, and the United States. Uni-directional causality runs from PU to tourist arrivals in the USA and Italy at all the quantiles (except the low quantiles 0.5 to 0.10). The uni-directional causality can be linked to a high level of information tourists acquire on the danger of the pandemic. Since health is wealth, some or most tourists minimize their travelling rate due to the rise in fear of PU. The results are in accord with the study of Nyarko et al. (2015), which found that the tourism sector is more susceptible to epidemics as tourists are unenthusiastic about traveling to countries having contagious diseases. Joo et al. (2019) and Baker (2020) report similar findings. Yang et al. (2020) infer that COVID-19 affects health status and impairs the labor productivity that drove tourism activities. In Spain, however, there is no causality in the low quantiles (0.5–0.35), while from the middle to high quantiles, uni-directional causality runs from PU to tourism.

Table 5.

Granger Causality Test in Quantiles.

Economic Policy Uncertainty and Tourist Arrivals Across Quantiles
USA	0.05	0.10	0.15	0.20	0.25	0.30	0.35	0.40	0.45	0.50	0.55	0.60	0.65	0.70	0.75	0.80	0.85	0.90	0.95
$∆{\mathrm{E}\mathrm{P}\mathrm{U}}_{\mathrm{t}}\hspace{0.17em}\mathrm{t}\mathrm{o}{∆\mathrm{T}\mathrm{O}\mathrm{U}}_{\mathrm{t}}$	0.21	0.31	0.01	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.02	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00
$∆{\mathrm{T}\mathrm{O}\mathrm{U}}_{\mathrm{t}}\hspace{0.17em}\mathrm{t}\mathrm{o}{∆\mathrm{E}\mathrm{P}\mathrm{U}}_{\mathrm{t}}$	0.43	0.54	0.00	0.02	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00
Italy	0.05	0.10	0.15	0.20	0.25	0.30	0.35	0.40	0.45	0.50	0.55	0.60	0.65	0.70	0.75	0.80	0.85	0.90	0.95
$∆{\mathrm{E}\mathrm{P}\mathrm{U}}_{\mathrm{t}}\hspace{0.17em}\mathrm{t}\mathrm{o}{∆\mathrm{T}\mathrm{O}\mathrm{U}}_{\mathrm{t}}$	0.00	0.09	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00
$∆{\mathrm{T}\mathrm{O}\mathrm{U}}_{\mathrm{t}}\hspace{0.17em}\mathrm{t}\mathrm{o}{∆\mathrm{E}\mathrm{P}\mathrm{U}}_{\mathrm{t}}$	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00
Spain	0.05	0.10	0.15	0.20	0.25	0.30	0.35	0.40	0.45	0.50	0.55	0.60	0.65	0.70	0.75	0.80	0.85	0.90	0.95
$∆{\mathrm{E}\mathrm{P}\mathrm{U}}_{\mathrm{t}}\hspace{0.17em}\mathrm{t}\mathrm{o}{∆\mathrm{T}\mathrm{O}\mathrm{U}}_{\mathrm{t}}$	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00
$∆{\mathrm{T}\mathrm{O}\mathrm{U}}_{\mathrm{t}}\hspace{0.17em}\mathrm{t}\mathrm{o}{∆\mathrm{E}\mathrm{P}\mathrm{U}}_{\mathrm{t}}$	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00
Pandemic uncertainty and inbound tourism across quantiles
USA	0.05	0.10	0.15	0.20	0.25	0.30	0.35	0.40	0.45	0.50	0.55	0.60	0.65	0.70	0.75	0.80	0.85	0.90	0.95
$∆{\mathrm{P}\mathrm{U}}_{\mathrm{t}}\hspace{0.17em}\mathrm{t}\mathrm{o}{∆\mathrm{T}\mathrm{O}\mathrm{U}}_{\mathrm{t}}$	0.14	0.09	0.00	0.00	0.00	0.00	0.01	0.01	0.01	0.04	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00
$∆{\mathrm{T}\mathrm{O}\mathrm{U}}_{\mathrm{t}}\hspace{0.17em}\mathrm{t}\mathrm{o}{∆\mathrm{P}\mathrm{U}}_{\mathrm{t}}$	0.76	0.21	0.12	0.97	0.83	0.81	0.93	0.42	0.72	0.67	0.48	0.18	0.71	0.83	0.71	0.82	0.92	0.53	0.45
Italy	0.05	0.10	0.15	0.20	0.25	0.30	0.35	0.40	0.45	0.50	0.55	0.60	0.65	0.70	0.75	0.80	0.85	0.90	0.95
$∆{\mathrm{P}\mathrm{U}}_{\mathrm{t}}\hspace{0.17em}\mathrm{t}\mathrm{o}{∆\mathrm{T}\mathrm{O}\mathrm{U}}_{\mathrm{t}}$	0.45	0.34	0.00	0.00	0.00	0.00	0.00	0.02	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00
$∆{\mathrm{T}\mathrm{O}\mathrm{U}}_{\mathrm{t}}\hspace{0.17em}\mathrm{t}\mathrm{o}{∆\mathrm{P}\mathrm{U}}_{\mathrm{t}}$	0.45	0.23	0.12	0.12	0.91	0.50	0.12	0.62	0.81	0.70	0.51	0.33	0.69	0.12	0.98	0.11	0.67	0.64	0.45
Spain	0.05	0.10	0.15	0.20	0.25	0.30	0.35	0.40	0.45	0.50	0.55	0.60	0.65	0.70	0.75	0.80	0.85	0.90	0.95
$∆{\mathrm{P}\mathrm{U}}_{\mathrm{t}}\hspace{0.17em}\mathrm{t}\mathrm{o}{∆\mathrm{T}\mathrm{O}\mathrm{U}}_{\mathrm{t}}$	0.21	0.01	0.02	0.41	0.34	0.19	0.52	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00
$∆{\mathrm{T}\mathrm{O}\mathrm{U}}_{\mathrm{t}}\hspace{0.17em}\mathrm{t}\mathrm{o}{∆\mathrm{P}\mathrm{U}}_{\mathrm{t}}$	0.32	0.34	0.20	0.18	0.95	0.78	0.55	0.17	0.58	0.56	0.27	0.56	0.57	0.12	0.23	0.56	0.34	0.12	0.12

Note:Table 5 presents the subsampling p-values of the S_T test. $∆{\mathrm{E}\mathrm{P}\mathrm{U}}_{\mathrm{t}}\hspace{0.17em},\hspace{0.17em}{∆\mathrm{T}\mathrm{O}\mathrm{U}}_{\mathrm{t}\hspace{0.17em}},\hspace{0.17em}\mathrm{a}\mathrm{n}\mathrm{d}\hspace{0.17em}{∆\mathrm{P}\mathrm{U}}_{\mathrm{t},}$ is economic policy uncertainty index, inbound tourism, and pandemic uncertainty index respectively.

Wavelet-Based Granger Causality Test Results

The traditional causality and wavelet-based Granger causality are reported in Table 6. In ensuring robust estimates and unbiased inferences, the Granger causality based on wavelet analysis is also used in this study for seven frequency domains (D1 to D7). The result shows no directional causal linkage between tourist arrivals and EPU in all the economies for domains D1 and D2. The absence of directional causal linkage between tourist arrivals and EPU implies no significant fall in tourists’ confidence in the economy; uncertainty remains low, and there are operational lags in macroeconomic policy relating to tourism in the low domain (D1 and D2). Similarly, there is no directional causality between PU and tourist arrival in domains D1 and D2 for all the economies except for Spain, where a significant effect of PU on tourism is detected in domain D2. In the third domain (D3), wavelet causality runs from PU to tourist arrivals in Italy and Spain. However, there is evidence of bidirectional causality between the United States’ variables. From the fourth domain (D4) to the seventh domain (D7), PU strongly influences tourist arrivals in the economy. At the domains D4 to D7, there is a significant rise in PU in all economies, which prompted legal restrictions on people’s movement. Moreover, travelers from highly developed countries are more sensitive to diseases and have a greater risk of PU, which pushes the tourism sector into an abyss. Thus, tourism becomes highly responsive to PU. The results are analogous to the findings for previous epidemic outbreaks, such as the avian flu epidemic in Asian and Pacific countries and the UK’s foot and mouth disease (Blake et al., 2003), resulting in a sharp decline in tourism expenditures. The outcome is also supported by Joo et al.’s (2019) findings on the decline in tourist arrivals during the MERS (Middle East respiratory syndrome) outbreak, showing that infectious disease outbreak notably impacts the tourism sector. The findings also show bidirectional wavelet causality between EPU and tourist arrivals from the third frequency (D3) domain to the last frequency domain (D4). This finding corroborates the quantile cointegration results. Besides, the traditional cointegration approach shows that EPU and PU Granger cause tourist arrival, and the causality is uni-directional.

Table 6.

Wavelets Based Causality (Granger Test in Decomposed Components).

Economic Policy Uncertainty, Pandemic Uncertainty, and Inbound Tourism
USA								Traditional Causality
Quantiles	DI	D2	D3	D4	D5	D6	D7
$∆{\mathrm{E}\mathrm{P}\mathrm{U}}_{\mathrm{t}}\hspace{0.17em}\nLeftrightarrow {∆\mathrm{T}\mathrm{O}\mathrm{U}}_{\mathrm{t}}$	0.32	0.34	0.00	0.00	0.00	0.00	0.00	0.02
$∆{\mathrm{T}\mathrm{O}\mathrm{U}}_{\mathrm{t}}\hspace{0.17em}\nLeftrightarrow {∆\mathrm{E}\mathrm{P}\mathrm{U}}_{\mathrm{t}}$	0.75	0.52	0.00	0.00	0.00	0.00	0.00	0.25
$∆{\mathrm{P}\mathrm{U}}_{\mathrm{t}}\hspace{0.17em}\nLeftrightarrow {∆\mathrm{T}\mathrm{O}\mathrm{U}}_{\mathrm{t}}$	0.04	0.09	0.00	0.00	0.00	0.00	0.00	0.00
$∆{\mathrm{T}\mathrm{O}\mathrm{U}}_{\mathrm{t}}\hspace{0.17em}\nLeftrightarrow {∆\mathrm{P}\mathrm{U}}_{\mathrm{t}}$	0.04	0.11	0.05	0.11	0.41	0.41	0.71	0.62
Italy
$∆{\mathrm{E}\mathrm{P}\mathrm{U}}_{\mathrm{t}}\hspace{0.17em}\nLeftrightarrow {∆\mathrm{T}\mathrm{O}\mathrm{U}}_{\mathrm{t}}$	0.53	0.45	0.00	0.00	0.00	0.00	0.00	0.00
$∆{\mathrm{T}\mathrm{O}\mathrm{U}}_{\mathrm{t}}\hspace{0.17em}\nLeftrightarrow {∆\mathrm{E}\mathrm{P}\mathrm{U}}_{\mathrm{t}}$	0.53	0.61	0.00	0.00	0.00	0.00	0.00	0.54
$∆{\mathrm{P}\mathrm{U}}_{\mathrm{t}}\hspace{0.17em}\nLeftrightarrow {∆\mathrm{T}\mathrm{O}\mathrm{U}}_{\mathrm{t}}$	0.14	0.09	0.00	0.00	0.00	0.00	0.01	0.01
$∆{\mathrm{T}\mathrm{O}\mathrm{U}}_{\mathrm{t}}\hspace{0.17em}\nLeftrightarrow {∆\mathrm{P}\mathrm{U}}_{\mathrm{t}}$	0.76	0.21	0.12	0.97	0.83	0.81	0.93	0.42
Spain
$∆{\mathrm{E}\mathrm{P}\mathrm{U}}_{\mathrm{t}}\hspace{0.17em}\nLeftrightarrow {∆\mathrm{T}\mathrm{O}\mathrm{U}}_{\mathrm{t}}$	0.67	0.54	0.00	0.00	0.00	0.00	0.00	0.00
$∆{\mathrm{T}\mathrm{O}\mathrm{U}}_{\mathrm{t}}\hspace{0.17em}\nLeftrightarrow {∆\mathrm{E}\mathrm{P}\mathrm{U}}_{\mathrm{t}}$	0.65	0.67	0.00	0.00	0.00	0.00	0.00	0.62
$∆{\mathrm{P}\mathrm{U}}_{\mathrm{t}}\hspace{0.17em}\nLeftrightarrow {∆\mathrm{T}\mathrm{O}\mathrm{U}}_{\mathrm{t}}$	0.12	0.00	0.00	0.00	0.00	0.00	0.00	0.00
$∆{\mathrm{T}\mathrm{O}\mathrm{U}}_{\mathrm{t}}\hspace{0.17em}\nLeftrightarrow {∆\mathrm{P}\mathrm{U}}_{\mathrm{t}}$	0.32	0.34	0.20	0.18	0.95	0.78	0.55	0.17

Note:Table 6 presents the subsampling p-values of the S_T test. $∆{\mathrm{E}\mathrm{P}\mathrm{U}}_{\mathrm{t}}\hspace{0.17em},\hspace{0.17em}{∆\mathrm{T}\mathrm{O}\mathrm{U}}_{\mathrm{t}\hspace{0.17em}},\hspace{0.17em}\mathrm{a}\mathrm{n}\mathrm{d}\hspace{0.17em}{∆\mathrm{P}\mathrm{U}}_{\mathrm{t},}$ is the log-difference of EPU, tourism, and PU index, respectively.

DI –D7 indicates the oscillations of periods 2–4, 4–8, 8–16, 16–32, 32–64, 64–128, and 128–256 quarters. $\nLeftrightarrow$ refers to null hypothesis of no-causality.

Quantile-on-Quantile Results

Impact of economic policy uncertainty and inbound tourism

The quantile-on-quantile results are shown in Figure 3, which portrays the estimates of the slope coefficient, β₁(τ, θ), and captures the effect of the τth quantile of EPU on the θth quantile of TOU at different values of τ and θ. The QQ plots reveal interesting relations between TOU and its covariates. The results show that the co-movement between TOU and its covariate (EPU) changes under different economic conditions. Overall, a persistent negative correlation is observed between EPU and tourism in the US, Spain, and Italy at all levels of EPU. This implies that the impact of EPU on tourism demand remains negative irrespective of the US and Italy’s economic conditions. Thus, it is expected that a rise in EPU will lead to a fall in tourism demand in both the US and Italy. However, in Italy, a positive correlation is observed between the high quantiles of EPU and low quantiles of tourism. Overall, the findings imply that a high level of EPU is associated with low tourist arrivals for the economy. When EPU rises, tourists raise their precautionary demand for money and reduce their tourism expenditures because a rise in EPU reduces their trust in their economy’s condition. Other studies report the same phenomenon, such as Drakos and Kutan’s (2003) finding that EPU harms tourism in Greece. Recently, Alawin and Lila (2016) validate the economic-political uncertainty as a contributing factor leading to the drop in tourist numbers. Singh et al. (2019) also confirmed the time-frequency varying nature of the relationship between economic policy uncertainty and tourism.

Figure 3.

Quantile on Quantile estimates of economic policy uncertainty and inbound tourism. Note: This graph depicts the estimates of the slope coefficient, β1 $\left(\tau ,\hspace{0.17em}\theta \right)$, which is placed on the z-axis against the quantiles of the EPU $\left(\tau \right)$x-axis and quantiles of the tourism (TOU) $\left(\theta \right)$ on the y-axis. The color bar shows the degree of co-movement between proposed variables, dark red color indicates a strong positive relationship, and dark negative shows a strong negative relationship between proposed variables.

Impact of pandemic uncertainty (PU) and inbound tourism

The study also examined the impact of PU on tourism for all the economies and presented in Figure 4. The results show that PU harms tourism demand for all the economies (US, Spain, and Italy). This is because a rise in PU usually leads to stricter restrictions on the inflow and outflow of tourists to curtail the spread of infectious diseases. Hence, a rise in PU will lead to a fall in tourism demand. Our results resonate with the previous findings of Kongoley-MIH (2015) and Novelli et al. (2018), who also evidenced the suspension of all air flights in the era of an epidemic outbreak. The outcome is also supported by Joo et al.’s (2019) findings of a decline in figures of tourist arrivals in case of MERS (Middle East respiratory syndrome) outbreak epidemic and envisages that infectious disease outbreak notably impacts the tourism sector. Wang et al. (2022) also revealed that pandemic risks decrease tourist arrivals.

Figure 4.

QQ estimates of Pandemic Uncertainty and inbound tourism.

Finally, the study precisely investigated the effect of EPU and COVID-19 on tourism as set out in Model 2. The results are presented in Figure 5, which shows that the effects of EPU and COVID-19 on tourism are negative in all quantiles and do not differ across the economies. This means that an increase in the EPU would lead to a decline in demand for tourism. Similarly, a decrease in tourist arrivals is expected as the number of people infected with COVID-19 increases in sampled countries. This negative relationship between COVID-19 and tourism is due to the fact that as the number of people infected increases, stricter steps are taken to limit people’s movement from one country to another. In addition, tourists become reluctant to spend on tourism activities, resulting in declining demand. Balsalobre-Lorente (2020), found that China’s isolation is less responsive to its economic growth while the country’s political willpower is elastic, as demonstrated by a government commitment to dampening the effect of the COVID-19 pandemic. This confinement is marked by the aggressive response by the government officials determined to flatten the exponential impact of the pandemic.

Figure 5.

QQ estimates of economic policy uncertainty and COVID-19 on inbound tourism (during COVID-19).

Conclusion and Policy Implications

This study is a timely addition to increasingly important strands of the tourism literature. Though tourism is a stable economic growth source (Husein et al., 2020; Wu et al., 2020), COVID-19 has drastically affected the tourism sector. Social distancing policies designed to prevent virus transmission imply a particularly negative supply shock for the tourism industry. In addition, the fear of infection significantly affects all elements of the tourism product value chain, drastically reducing its demand. Therefore, many occupations linked to tourism show significant risks of being negatively affected by the COVID-19 crisis due to both supply and demand shocks. This study explores the effects of EPU and PU on inbound tourism in the United States, Italy, and Spain. The quantile causality results indicate that EPU and inbound tourism have overall bi-directional causality across all quantiles and uni-directional causality running from PU to inbound tourism in the USA, Italy, and Spain. Moreover, Granger’s wavelet-based causality confirms bi-directional causality between EPU and inbound tourism in the long run and uni-directional causality from PU to inbound tourism. It is notable that the effects of EPU and PU have little immediate impact on tourism, while medium to long-term shocks continue due to major undesirable EPU and PU events. Finally, the QQ approach confirms that overall EPU and PU negatively influence inbound tourism in the USA, Italy, and Spain, though the strength of the relationship differs with economic conditions. In addition, the findings further indicate that during COVID-19, EPU and COVID-19 both had a strong negative effect on inbound tourism in the USA, Italy, and Spain. The association did not differ across time.

Based on the empirical findings, it is concluded that tourism responds to health and economic policy-related uncertainties. The study’s results (based on the wavelet analysis) indicate that policymakers should concentrate on long-term economic and pandemic uncertainty policies since pandemic uncertainty has a long-term impact on tourism, which requires long-term policy. In addition, it is not wise to equate COVID-19 with previous pandemics, as it is new, so policymakers should concentrate on lifting travel restrictions (applying new health protocols for safe travel), restoring traveler trust (visitor information apps and domestic tourism promotion campaigns), and rethinking (preparing detailed tourism recovery plans, encouraging innovation and investment). Another way to promote safety and health care in the studied bloc must involve smart processes. Smart tourism, related to assistive technology and augmented reality (artificial intelligence, virtual reality, speech recognition, and standardized responses), should be presented in all tourism environments to provide quality, service, and satisfaction, thereby exceeding expectations, improving accessibility promoting heritage and culture.

The limitations of this research are mostly related to the unprecedented character of the present crisis. The COVID-19 virus is slightly less lethal but more contagious than SARS. This implies that its fear effect could be more pronounced and last longer among tourists. Therefore, future research needs to consider the latest data and could include additional measures of COVID-19. Future research should also expand the sample countries. Moreover, COVID-19 is a unique case due to the speed with which it has spread worldwide and wreaked havoc in the global economy. The evidence available indicates that most countries have taken the right steps to restrict its further spread by restricting human movement, which has had a debilitating impact on economic activity. As many of these countries are now moving toward the gradual easing of these restrictions, it would be interesting to see how quickly the tourism industry and broader economies have rebounded by sampling different countries.

ORCID iDs

Xin Zhao https://orcid.org/0000-0002-6232-9835

Muhammad Saeed Meo https://orcid.org/0000-0002-8340-0442