Exploring the influence of economic policy uncertainty on the relationship between tourism and economic growth with an MF-VAR model

Abstract

Using a mixed-frequency vector autoregressive (MF-VAR) model, this article attempts to determine whether or not the relationship between tourism and economic growth changes in the presence or absence of economic policy uncertainty (EPU) shock. Moreover, we further our analysis by focusing on whether or not there is a significant difference in the distinct impact intensity of Hong Kong, Chinese, and global EPU. The study period spans April 1998 to March 2018. The results indicate the following. First, the existence of Hong Kong, Chinese, and global EPU does not affect the direction of the impulse response; rather, its primary influence is on the size of the impact. Second, the different ranges of EPU have different impact intensities. Third, compared to the MF-VAR model, the quarterly frequency vector autoregressive model does not fully capture the impact of EPU, especially the negative impact of global EPU on tourism. Therefore, policymakers and tourism stakeholders should develop targeted marketing plans to maintain expected tourism demand if economic uncertainty increases.

Keywords

economic growth economic policy uncertainty MF-VAR model tourism demand

Introduction

Tourism is one of the four pillar industries in Hong Kong, a coastal city in the southern part of China (Li et al., 2013). The economic implications of tourism play an important role in society, extending to government revenues, employment, consumption, and investment, all of which are inseparable from the appropriate economic policies proposed by the Hong Kong Government to promote the tourism industry.

The implementation of visa liberalization policies, such as the Individual Visit Scheme (IVS) in 2003 and the multiple-entry permits for Shenzhen residents in 2009, prompted a rapid and sustained growth, from 8.47 million in 2003 to 45.84 million in 2015, in the number of visitors to Hong Kong from mainland China (Hong Kong Tourism Board, 2016; Qiu et al., 2019). In 2015, the number of visitor arrivals (hereinafter, VA) from mainland China suffered a decline for the first time since the IVS was launched, declining for seven consecutive months (Hong Kong Tourism Board, 2016; Wong and Buckley, 2015). The reasons for this decline were the occurrence of adverse social events such as protests and strict restrictions on the purchase of infant formula milk (Liu and McKercher, 2016; Li et al., 2016; Qiu et al., 2019; Zhang et al., 2018). The implementation of strict policies is intended to lower the negative impact on the daily lives of residents, such as overcrowding and rising commodity prices, which are caused by the influx of consumers from mainland China. However, these incidents worsened the relationship between visitors from mainland China and Hong Kong residents, resulting in a reduction in VA. In 2017, the overall trend for Hong Kong’s tourism industry was encouraging, mainly due to the favorable economic situation in mainland China (which is the primary entry market for Hong Kong’s tourism industry) and the growth of short-haul markets.

We find that due to the nature of policy decision-making and implementation processes, economic policies typically generate a large amount of uncertainty (Zhang et al., 2015), which may (a) create difficulties for the economic growth of destinations and for decision-making and (b) affect economic agents’ behavior (Bernanke, 1983; Bloom et al., 2007; Brogaard and Detzel, 2015). Studies conducted by Giavazzi and McMahon (2012) and Baker et al. (2013) showed that businesses and households will postpone investment and consumption expenditure during times of increasing economic policy uncertainty (hereinafter, EPU), which may lead to a reduction and delay in tourism spending (Dragouni et al., 2016).

On the basis of the pillars of previous studies on EPU and the work of Liu and Song (2017), who conducted a detailed study to test the short- and long-term causal relationship between tourism and economic growth, we first attempt to extend existing research by using a new EPU index developed by Baker et al. (2016) to test whether or not the relationship between tourism and economic growth will change in the presence or absence of an EPU shock. Second, this study considered three different ranges of EPUs: Hong Kong, Chinese, and global. Demir and Ersan’s (2018) study showed that different ranges of EPU have distinct impact intensities. For example, domestic EPU involves a variety of region-specific uncertainties in a country, while global EPU corresponds to global economic and political uncertainty events (Singh et al., 2018). Therefore, another purpose of this study is to test whether three different ranges of EPU will have different impact intensity on the relationship between tourism and economic growth, and if so, why.

We contribute to the extant literature in several ways. First, this study is based on a new perspective on the mechanism of action between various factors, analyzing whether or not the relationship between tourism and economic growth will change in the presence or absence of an EPU shock.

Second, as the time-series data regarding Hong Kong’s tourism and economic growth and the three EPU shocks used in this article are based on different frequencies (e.g. quarterly VA, quarterly gross domestic product (hereinafter, GDP), and monthly EPU), and policy decisions based on the results of temporal aggregation effects often produce undesirable results (Zellner and Montmarquette, 1971), in this article, we use the mixed-frequency vector autoregressive (hereinafter, MF-VAR) model developed by Ghysels et al. (2016) which can exploit all available data whatever the sampling frequency. Moreover, we compare results from the MF-VAR model with those from the quarterly frequency vector autoregressive (hereinafter, QF-VAR) model to analyze whether or not the MF-VAR model captures results that the QF-VAR model does not find.

Third, the study results show that global EPU has a more significant negative impact on Hong Kong VA than Hong Kong or Chinese EPU. This finding can provide relevant advice for the development of tourism. As tourism is one of Hong Kong’s major industries, the revenue inflow of the tourism industry is a significant driver of economic growth. If economic uncertainty increases, especially a change in global EPU, the fall in tourism demand may impede economic growth in the long run. Therefore, policymakers and tourism stakeholders should develop targeted marketing plans to maintain expected tourism demand.

The rest of the article is organized as follows. In the next section, we provide a literature review and summarize some related work. Then, we describe the methodology and data and present the study’s empirical results. Finally, the last section presents the conclusions of the study.

Literature review

There are two main hypotheses about the relationship between tourism and economic growth: the tourism-led economic growth hypothesis (hereinafter, TLEGH) and the economy-driven tourism growth hypothesis (hereinafter, EDTGH). Most researchers commonly use the Granger causality test (Granger, 1969) or its extensions to study the relationship between the two variables. The results regarding the relationship between tourism and economic growth are not consistent. Some empirical studies of various countries and regions support the TLEGH (Balaguer and Cantavella-Jordá, 2002; Katircioglu, 2009) and others support the EDTGH (Lee, 2008; Oh, 2005; Payne and Mervar, 2010; Tang and Jang, 2009), while some support both (Dritsakis, 2012; Kim and Chen, 2006; Lee and Chang, 2008). One study proposes that neither hypothesis is correct and that there is no causal relationship between tourism and economic growth (Ozturk and Acaravci, 2009). However, is the relationship between tourism and economic growth affected by EPU? Responding to this question with fresh new insights is the primary purpose of this article.

Uncertainty drastically impacts tourism demand and can be evaluated in different forms. The effects of economic uncertainty in the form of financial and economic crisis on tourism activities have long been recognized (Alegre et al., 2013; Alonso-Almeida and Bremser, 2013; Frechtling, 1982; Hall, 2010; Li et al., 2010; Okumus et al., 2005; Papatheodorou et al., 2010; Perles-Ribes et al., 2016; Song et al., 2011; Song and Lin, 2009). Hall (2010) reviewed the literature on tourism and crisis, summarizing the various types of uncertainty, and suggested that economic and financial crisis receives the most research attention. Perles-Ribes et al. (2016) reviewed the literature on economic crisis and tourism and suggested that most of the literature tends to focus only on the impact of a particular crisis on one destination or region. In addition, some researchers have studied the impact of terrorism (Lepp and Gibson, 2003; Pizam and Fleischer, 2002), political instability (Fletcher and Hillingdon, 2008; Sonmez, 1998), health-related diseases (Cooper, 2005; Kuo et al., 2008; Kuo et al., 2009; McKercher and Chon, 2004), and natural disasters (Faulkner, 2001).

Since Baker et al. (2013) proposed the EPU index, emerging researchers have begun to use the index to represent policy uncertainty to study its impact on tourism. Dragouni et al. (2016) studied the spillover effects of sentiment and mood shocks in the United States using the Consumer Sentiment Index and EPU as representatives of emotions. Some researchers have studied the impact of EPU on tourism receipts (Wu and Wu, 2018), VA (Balli et al., 2018; Gozgor and Ongan, 2017; Işık et al., 2019; Ongan and Gozgor, 2018; Singh et al., 2018; Tiwari et al., 2019), tourism spending (Gozgor and Ongan, 2017; Gozgor and Demir, 2018), hotel operating performance (Madanoglu and Ozdemir, 2019), business tourism (Tsui et al., 2018), and the stock prices of tourism companies (Demir and Ersan, 2018; Ersan et al., 2019). Most studies show that a higher level of EPU has a negative impact on tourism. Furthermore, some scholars have studied the impact of tourism on EPU: for example, Akadiri et al. (2019) examined whether tourists have the power to predict EPU in regions of America, Europe, and Asia Pacific.

In terms of study area distribution, we found that most of the research areas related to EPU are concentrated in popular tourism areas, such as Europe (Ghosh, 2019; Wu and Wu, 2018), the United States (Dragouni et al., 2016; Ghosh, 2019; Gozgor and Ongan, 2017; Singh et al., 2018), Turkey (Demir and Ersan, 2018), and New Zealand (Tsui et al., 2018). As Hong Kong is a major tourist destination, the impact of EPU on Hong Kong’s tourism industry cannot be ignored. In terms of the empirical methods used in the literature, some studies have used econometric models, including dynamic ordinary least squares (Işık et al., 2019), panel-related models (Akadiri et al., 2019; Saha et al., 2017; Tsui et al., 2018), and error correction models (Gozgor and Ongan, 2017; Işık et al., 2019; Kim et al., 2018; Ongan and Gozgor, 2018), to analyze the impact of EPU on tourism demand. Gozgor and Demir (2018) used fixed effects and least square dummy variable estimation techniques to study the effects of EPU on travel expenditures. In the latest research, novel wavelet analysis has been used to explore the real relationship between EPU and tourism (Balli et al., 2018; Singh et al., 2018; Tiwari et al., 2019; Wu and Wu, 2018).

Methodology and data description

Methodology

We achieved our objective using the MF-VAR model proposed by Ghysels et al. (2016) to examine the relationship between quarterly VA and quarterly GDP with and without EPU (monthly). A QF-VAR model was introduced to show that different sampling frequencies will have different effects on the empirical results. We began our empirical framework by specifying the following QF-VAR model with single-frequency data and the MF-VAR model.

QF-VAR model

As a benchmark for comparison with the MF-VAR model, a QF-VAR (4) model was specified as follows:

[\begin{matrix} V A_{t} \\ {GDP}_{t} \\ {EPU}_{t}^{Q} \end{matrix}] = \sum_{k = 1}^{4} [\begin{matrix} a_{11, k} & a_{12, k} & a_{13, k} \\ a_{21, k} & a_{22, k} & a_{23, k} \\ a_{31, k} & a_{32, k} & a_{33, k} \end{matrix}] [\begin{matrix} V A_{t - k} \\ {GDP}_{t - k} \\ {EPU}_{t - k}^{Q} \end{matrix}] + [\begin{matrix} ε_{1 t} \\ ε_{2 t} \\ ε_{3 t} \end{matrix}],

where $t \in \{1, \dots, n\}$ denotes each quarter, the superscript “Q” is used to explicitly distinguish a quarterly level from a monthly level, VA _t denotes VA in quarter t, GDP _t denotes real GDP in quarter t, and $E P U_{t}^{Q}$ denotes the EPU of quarter t, which can be interpreted as ${EPU}_{t}^{Q} = \frac{1}{3} \sum_{j = 1}^{3} {EPU}_{j t}$ (Ghysels et al., 2016; Motegi and Sadahiro, 2018).

In this study, five information criteria were used to select the optimal lag order. The results are shown in Table 1. When Hong Kong EPU was considered in constructing the QF-VAR and MF-VAR models, the optimal lag order chosen according to the Schwarz information criterion and the Hannan–Quinn information criterion (HQ) was always lag 1, and the remaining three criteria were chosen as lag 1 when the maximum lags were under four and as lag 4 otherwise. Moreover, we chose to set the lag length to four quarters to capture the potential seasonal impact.

Table 1.

Optimal VAR lag order selected by different criteria in different EPU cases.

Max Lag (HK)	1	2	3	4	5	6	7	8	9	10
LR	1	1	1	4	4	4	4	4	4	4
FPE	1	1	1	4	4	4	1	4	4	4
AIC	1	1	1	4	4	4	1	4	4	4
SC	1	1	1	1	1	1	1	1	1	1
HQ	1	1	1	1	1	1	1	1	1	1
Max Lag (Chinese)	1	2	3	4	5	6	7	8	9	10
LR	1	2	3	4	4	4	4	4	4	4
FPE	1	2	3	4	4	4	4	1	4	4
AIC	1	2	3	4	4	4	4	1	4	4
SIC	1	1	1	1	1	1	1	1	1	1
HQ	1	1	1	1	1	1	1	1	1	1
Max Lag (Global)	1	2	3	4	5	6	7	8	9	10
LR	1	1	3	4	4	4	4	4	4	4
FPE	1	1	3	4	4	4	4	4	4	4
AIC	1	1	3	4	4	4	4	4	4	4
SIC	1	1	1	1	1	1	1	1	1	1
HQ	1	1	1	1	1	1	1	1	1	1

Note: EPU: economic policy uncertainty; VAR: vector autoregression; LR: sequentially modified LR test statistic (each test at 5% level); FPE: final prediction error; AIC: Akaike information criterion; SIC: Schwarz information criterion; HQ: Hannan–Quinn information criterion.

Similar to the case of Hong Kong, lag 4 was selected as the optimal lag order when using Chinese EPU or global EPU to build a QF-VAR model or an MF-VAR model.

MF-VAR model

We then referred to articles written by Ghysels et al. (2016) and Motegi and Sadahiro (2018) to formulate an MF-VAR model consisting of quarterly VA, quarterly GDP, and monthly EPU. The MF-VAR (p) model was as follows:

\underset{= X_{t}}{\underset{︸}{[\begin{matrix} \begin{array}{l} V A_{t} \\ {GDP}_{t} \end{array} \\ {EPU}_{1 t} \\ ⋮ \\ {EPU}_{m t} \end{matrix}]}} = \sum_{k = 1}^{p} \underset{= A_{k}}{\underset{︸}{[\begin{matrix} A_{H H, k} & a_{H L, k} \\ a_{L H, k} & a_{L L, k} \end{matrix}]}} \underset{= X_{t - k}}{\underset{︸}{[\begin{matrix} \begin{array}{l} V A_{t - k} \\ {GDP}_{t - k} \end{array} \\ {EPU}_{1, t - k} \\ ⋮ \\ {EPU}_{m, t - k} \end{matrix}]}} + ε_{t},

where VA _t denotes VA in quarter t, GDP _t denotes real GDP in quarter t, and EPU _mt denotes the $j - t h$ month EPU of quarter t, where $j \in \{1, \dots, m\}$ ; m = 3 for the mixed-frequency data consisting of monthly and quarterly data, which means that one quarter has 3 months. The MF-VAR (p) model can also be written as

X_{t} = \sum_{k = 1}^{p} A_{k} X_{t - k} + ε_{t} .

On the basis of the MF-VAR (p) model, we formulated the MF-VAR (4) model. Furthermore, to make a fair comparison with the QF-VAR model of equation (1), we set the lag length of the MF-VAR (p) to four quarters as well. The MF-VAR (4) model was as follows:

\underset{= X_{t}}{\underset{︸}{[\begin{matrix} \begin{array}{l} V A_{t} \\ G D P_{t} \end{array} \\ E P U_{1 t} \\ E P U_{2 t} \\ E P U_{3 t} \end{matrix}]}} = \sum_{k = 1}^{4} \underset{= A_{k}}{\underset{︸}{[\begin{matrix} a_{11, k} & a_{12, k} & a_{13, k} & a_{14, k} & a_{15, k} \\ a_{21, k} & a_{22, k} & a_{23, k} & a_{24, k} & a_{25, k} \\ a_{31, k} & a_{32, k} & a_{33, k} & a_{34, k} & a_{35, k} \\ a_{41, k} & a_{42, k} & a_{43, k} & a_{44, k} & a_{45, k} \\ a_{51, k} & a_{52, k} & a_{53, k} & a_{54, k} & a_{55, k} \end{matrix}]}} \underset{= X_{t - k}}{\underset{︸}{[\begin{matrix} V A_{t - k} \\ G D P_{t - k} \\ E P U_{1, t - k} \\ E P U_{2, t - k} \\ E P U_{3, t - k} \end{matrix}]}} + \underset{= ε_{t}}{\underset{︸}{[\begin{matrix} ε_{1 t} \\ ε_{2 t} \\ ε_{3 t} \\ ε_{4 t} \\ ε_{5 t} \end{matrix}]}} .

The MF-VAR (4) model can then be written as

X_{t} = \sum_{k = 1}^{4} A_{k} X_{t - k} + ε_{t} .

It is worth noting that a key feature of equation (4) is the way in which the EPU indexes ( $E P U_{1 t}, E P U_{2 t}, E P U_{3 t}$ ) are stacked in a vector. For example, looking at the first row of equation (4), the link between VA and EPU can be implied as follows:

V A_{t} = \sum_{k = 1}^{4} [a_{11, k} V A_{t - k} + a_{12, k} G D P_{t - k} + \sum_{j = 1}^{3} a_{1 (j + 2), k} E P U_{j, t - k}] .

The same connection between VA and EPU can be written as the first row of equation (1) in the QF-VAR model:

V A_{t} = \sum_{k = 1}^{4} [a_{11, k} V A_{t - k} + a_{12, k} G D P_{t - k} + a_{13, k} (\frac{1}{3} \sum_{j = 1}^{3} E P U_{j, t - k})] .

Compared to equation (7), an advantage of the MF-VAR model is that the EPU from each month, $E P U_{i, t - k} (i = 1, 2, 3)$ , will have a heterogeneous effect $(a_{1 j, k}, j = 3, 4, 5)$ on VA _t . However, the impact of EPU produced in equation (7) implies that $E P U_{1, t - k}$ , $E P U_{2, t - k}$ , and $E P U_{3, t - k}$ have a homogeneous impact of $a_{13, k} /3$ on VA _t . Therefore, the QF-VAR model rules out the possibility of seasonal effects and lagged information transmission within each month.

Additionally, referring to other empirical studies that have used the MF-VAR model (Ghysels, 2016; Ghysels et al., 2016; Motegi and Sadahiro, 2018), the demeaned series has typically been used to avoid a constant to reduce the estimated parameters before fitting the model. The technical details of the symbols in the parsing specification of the MF-VAR model have been omitted for the sake of simplicity. More details can be found in Ghysels (2016) and Ghysels et al. (2016).

After the causality test of the variables, we also performed impulse response and variance decomposition analysis for each model to observe how long the effect of EPU shock would remain active through the system in relation to the variables. Please see the “Granger causality tests” subsection in the “Empirical results” section for details on the selection of the Cholesky order required in the impulse response analysis.

Data description

Data sources

Our study utilized data recorded in regard to the number of monthly Hong Kong, Chinese, and global EPU indexes from April 1998 to June 2018; quarterly real GDP; and quarterly Hong Kong VA for 1998Q2 to 2018Q1 (80 quarters).

We used a logarithmic form of the monthly EPU index. The raw data of Hong Kong, Chinese, and global EPU were obtained from the EPU website (www.policyuncertainty.com) designed by Baker et al. (2016). The EPU index is constructed on the basis of three components: (a) newspaper coverage of policy-related economic uncertainty, (b) future expiration of federal tax code provisions, and (c) future expectations of inflation and government spending in relation to the macroeconomic environment of the country (Dragouni et al., 2016) and current and future economic conditions (Baker et al., 2016). Economic growth was measured using real quarterly GDP data deflated by the price level in 2010 downloaded from the Hong Kong SAR Government’s Census and Statistics Department website. We used VA in Hong Kong as a proxy for inbound tourism; VA data were taken from the Hong Kong Tourism Board’s annual statistical reviews (Liu and Song, 2017). The variables of GDP and VA were calculated by 100 times year-to-year log differences to ensure stationarity.

Figure 1(a) shows the two 100 times annual log differences of quarterly series of VA and real GDP, and the three logarithmic forms of the Hong Kong, Chinese, and global monthly EPU indexes are shown in Figure 1(b). Several regularities were observed. First, the low point in the number of VA in Hong Kong was caused by the Asian financial crisis (1997–1998), and real GDP was at a low point during this period. Second, the most severe decline occurred during the SARS epidemic during the winter of 2002 and spring of 2003, but the decline in the real GDP growth rate was not as deep as that in tourism growth. Third, the global financial crisis caused VA to decrease during the 2008–2009 period. The effect of this decline on economic growth was almost the same as its effect on tourism growth. Fourth, the most recent significant decrease in VA occurred during 2015 and 2016. In 2015, the number of mainland Chinese tourists visiting Hong Kong fell for seven consecutive months, the first time this had happened since the IVS was launched (Qiu et al., 2019; Wong and Buckley, 2015). This decline may have been due to the fact that a large number of mainland visitors came to Hong Kong to buy daily necessities, but this led to rising prices and shortages of some goods (Liu and McKercher, 2016), which in turn had a negative effect on mainland visitor numbers. Such behavior also led to a deterioration in the relationship between mainland Chinese tourists and Hong Kong residents (Qiu et al., 2019). In addition, the relaxation of visas in neighboring countries such as Japan (Hong Kong Legislative Council, 2015) and the appreciation of the Hong Kong dollar also affected the competitiveness of Hong Kong’s tourism industry (Tolkach, 2018). Moreover, due to the tourism industry’s performance, economic growth was also negatively affected during this period.

Figure 1.

Mixed-frequency time series of VA, real GDP, and three different EPU indexes. (a) Quarterly data and (b) monthly data.

Through visual inspection, we observed that during the study period, (a) the EPU index was sensitive to political and economic shocks and (b) three series of the EPU index seemed to follow the changes in the number of VA and economic growth. The occurrence of the Asian financial crisis (1997–1998), the SARS epidemic (2002–2003), and the global financial crisis (2008–2009) led Hong Kong, Chinese, and global EPU reaching high points to varying degrees due to the instability of the economic environment.

Stationarity tests

We first employed the most commonly augmented Dickey–Fuller unit root test (Dickey and Fuller, 1979, 1981) for the variables of $\{V A_{t}, G D P_{t}, E P U_{1 t}, E P U_{2 t}, E P U_{3 t}\}$ . In the time-series graph in Figure 1, only an intercept is included for each series; a trend is not included for each series. Therefore, we only considered the case when an intercept was included in the specific unit root test settings. The results in Table 2 show that in the case of including an intercept, the null hypothesis that a “unit root is present in the time series” was not rejected at the 10% significance level for the global EPU of the third month of each quarter, and all the remaining series rejected the null hypothesis at the 10% level. To ensure the stability of the data, we also employed Phillips and Perron’s (PP, 1988) and Ng and Perron’s (Ng–Perron, 2001) unit root tests to test the variables for stationarity. The null hypothesis of both the PP and Ng–Perron unit root tests is that a “unit root is present in the time series.” The results showed that the unit root null hypothesis was rejected at the 10% significant level for all series. In addition, as seasonal data were used in this study, the existence of seasonal unit roots was also checked using the Hylleberg, Engle, Granger, and Yoo’s (hereinafter, HEGY) seasonal unit root test proposed by Hylleberg et al. (1990). It should be noted that according to equation (2), when the data are brought into the MF-VAR model, the monthly data have been transformed into quarterly frequency data according to the different months of each quarter, so seasonal unit root tests are performed two or four quarters per time-series cycle. The null hypothesis that a “seasonal unit root is present in 2 or 4 quarters per time series cycle” was rejected at the 5% significant level for all series. This suggests that the variables were stationary in level form. On the basis of the results of the unit root test excluding spurious relationships among all the variables, we proceeded to conduct the Granger causality test via the MF-VAR model.

Table 2.

Four different unit root tests in levels when only the intercept is considered.

Variables	ADF test statistics	PP test statistics	Ng–Perron test statistics				HEGY test statistics
Variables	p-value	p-value	MZa	MZt	MSB	MPT	2 quarters	4 quarters
$Δ ln (V A)$	0.0000***	0.0000***	−17.2822***	−2.9614***	0.1665***	1.4534***	0.0056***	0.0000***
$Δ ln (G D P)$	0.0212**	0.0084***	−23.3421**	−3.4128**	0.1462**	3.9251**	0.0000***	0.0000***
$ln (E P U_{1}^{H K})$	0.0000***	0.0000***	−6.0121*	−1.7126*	0.2849*	4.1441*	0.0196**	0.0114**
$ln (E P U_{2}^{H K})$	0.0002***	0.0004***	−8.0465*	−1.9902*	0.2473*	3.1048**	0.0088***	0.0000***
$ln (E P U_{3}^{H K})$	0.0000***	0.0000***	−18.8760**	−3.0703**	0.1627**	4.8390**	0.0102**	0.0000***
$ln (E P U_{Q}^{H K})$	0.0016***	0.0019***	−8.6019**	−2.0493**	0.2382*	2.9441**	0.0271**	0.0000***
$ln (E P U_{1}^{C h i n a})$	0.0004***	0.0003***	−19.8921**	−3.0919**	0.1554**	4.9594**	0.0056***	0.0000***
$ln (E P U_{2}^{C h i n a})$	0.0002***	0.0005***	−35.0886***	−4.1883***	0.1194***	2.5987***	0.0056***	0.0005***
$ln (E P U_{3}^{C h i n a})$	0.0035***	0.0054***	−22.5149**	−3.3544**	0.1490**	4.0523**	0.0319**	0.0000***
$ln (E P U_{Q}^{C h i n a})$	0.0214**	0.0276**	−169.5510***	−23.8000***	0.0543***	0.5503***	0.0073***	0.0010***
$ln (E P U_{1}^{G l o b a l})$	0.0974*	0.0089***	−19.3124**	−3.0830**	0.1596**	4.8688**	0.0028***	0.0001***
$ln (E P U_{2}^{G l o b a l})$	0.0076***	0.0076***	−26.0172***	−3.6053***	0.1386***	3.5114***	0.0036***	0.0000***
$ln (E P U_{3}^{G l o b a l})$	0.1419	0.0044***	−20.5633***	−3.2055**	0.1559**	4.4380**	0.0056***	0.0000***
$ln (E P U_{Q}^{G l o b a l})$	0.0742*	0.0724*	−18.6825***	−3.0493**	0.1632**	4.9211**	0.0028***	0.0003***

Note: ADF: augmented Dickey–Fuller; PP: Phillips and Perron; AIC: Akaike information criterion; HEGY: the Hylleberg, Engle, Granger, and Yoo’s. The null hypothesis is that a unit root exists, and the lag length of the Ng–Perron and HEGY unit root tests are both selected on the basis of the AIC.

***,**, and * denote significance at the 1%, 5%, and 10% significance levels.

Empirical results

This study examined (1) whether or not the relationship between tourism and economic growth will change in the presence or absence of EPU shocks and (2) whether or not there is a significant difference in the impact intensity of Hong Kong, Chinese, and global EPU.

Granger causality tests

The Granger causality test aims to study the past influence of conditional variables against other current variables in the present (Mohd Yusof et al., 2018) and to explore the existence of causal relationships between variables in a specific direction from a statistical perspective. This study adopted the mixed-frequency Granger causality test (MFGCT) proposed by Ghysels et al. (2016) to examine whether or not the participation of three EPUs—monthly Hong Kong, Chinese, and global EPU—has an impact on the relationship between quarterly VA and GDP (Liu and Song, 2017). We compared these results with those of the low-frequency Granger causality test (LFGCT).

The Granger causality test results in Table 3 show that, first of all, the bidirectional Granger causal relationship between economic growth and tourism was not established, and only the null hypotheses of non-causality running from economic growth to tourism at the 5% significance level in the absence of EPU were rejected. Significant causality was still only established when Hong Kong EPU was taken into account.

Table 3.

Bootstrapped p-values of granger causality tests (quarterly VAR vs. MF-VAR).

	GDP $↛$ VA	VA $↛$ GDP	EPU $↛$ VA	VA $↛$ EPU	EPU $↛$ GDP	GDP $↛$ EPU
${QF - VAR}_{without EPU}$	0.026**	0.138	—	—	—	—
${QF - VAR}_{with HK EPU}$	0.309	0.078*	0.381	0.293	0.072*	0.358
${MF - VAR}_{with HK EPU}$	0.088*	0.411	0.031**	0.675	0.015**	0.086*
${QF - VAR}_{with Chinese EPU}$	0.677	0.107	0.446	0.112	0.080*	0.026**
${MF - VAR}_{with Chinese EPU}$	0.870	0.171	0.498	0.010*	0.404	0.193
${QF - VAR}_{with Global EPU}$	0.529	0.145	0.405	0.373	0.027**	0.379
${MF - VAR}_{with Global EPU}$	0.777	0.149	0.0004***	0.481	0.0009***	0.660

Note: VA: visitor arrivals; EPU: economic policy uncertainty; MF-VAR: mixed frequency vector autoregressive; QF-VAR: quarterly frequency vector autoregressive. “***”, “**”, and “*” denote significance at the 1%, 5% and 10% levels, respectively. The p-values are computed with the heteroscedasticity robust parametric bootstrap of (Gonçalves and Kilian, 2004). EPU signifies joint non-causality from/to {EPU₁,EPU₂, EPU₃}.

Second, three forms of unidirectional causality between the three EPUs and tourism were found using the MFGCT, including the significant causality running from Hong Kong EPU and global EPU to tourism at the 5% significance level and the significant causality running from tourism to Chinese EPU at the 5% significance level. However, the LFGCT results showed that there was no Granger causality between the three EPUs and tourism regardless of whether the 5% or 10% level was used.

Third, the MFGCT results showed there was a unidirectional causality running from Hong Kong EPU and global EPU to economic growth at the 5% significance level. There was significant causality running from economic growth to Hong Kong EPU at the 10% significance level. The LFGCT results showed that there was a unidirectional causality running from the three EPUs to economic growth, but only the significant causality running from economic growth to Chinese EPU was supported.

Our results suggest that the MFGCT produces more intuitive results; this may be due to the way in which the MF-VAR model has higher asymptotic power than the LF-VAR model and is better able to uncover causal relationships in an underlying high-frequency data generation process (Ghysels et al., 2016). However, why did we have so many pairs of non-rejected results? Although the Granger causality test can examine the existence of causal relationships between variables in a specific direction from a statistical perspective, a critical analysis of the significant relationships and exactly how long the effect of an EPU shock on macroeconomic variables remains active through the system was warranted. Variance decomposition and impulse response functions are suitable methods to use to observe these relationships (Chiweza and Aye, 2018).

Referring to Baker et al. (2013) allowed us to identify the dynamic relationship among the variables through differences in the timing of movements in the variables. The Cholesky decomposition applied in this article was used to calculate the impulse response functions with the following ordering of variables: $V A \to G D P \to E P U$ . This ordering presumed that the response of GDP is due to the VA shock, which means that GDP does not have an instantaneous influence on VA, while VA exerts an impact on GDP. Similarly, it presumed that VA and GDP responses due to an EPU shock do not have an instantaneous influence, while the structural shock corresponding to the first two variables exerts an impact on the EPU index.

Impulse response and variance decomposition analyses between VA and GDP in the presence of three EPUs

Impulse response analyses

Figure 2(a) and 2(c) show the impulse responses of economic growth resulting from a one-standard-deviation change in VA in the presence and absence of three EPUs using the QF-VAR and MF-VAR models, respectively. First, we found that a positive VA shock had an initial strong positive influence on economic growth, but this soon dissipated, and in the fourth quarter, a maximum negative response was observed. In the long run, the impact of the shock disappeared. Second, the existence of the three EPUs did not affect the direction of the impulse response; rather, the primary influence of EPU was on the size of the impact.

Figure 2.

Impulse response function between VA and GDP variables in the presence and absence of three EPUs, based on QF-VAR and MF-VAR models. (a) VA→GDP (QF-VAR); (b) GDP→VA (QF-VAR); (c) VA→GDP (MF-VAR); and (d) GDP→VA (MF-VAR). MF-VAR: mixed frequency vector autoregressive; QF-VAR: quarterly frequency vector autoregressive.

We also found that the intensity of the impact of EPU varied according to the range of the EPU. Compared with the existence of Hong Kong EPU and Chinese EPU, the existence of global EPU had an intervention effect on the impulse responses of economic growth to a one-standard-deviation change in VA, and the effects were more severe in the MF-VAR model. Specifically, the effect of an EPU shock showed a subsequent strong negative impact in the fourth quarter; it then increased steadily up to the eighth quarter, when the response reached another small peak. The results of the MF-VAR model imply that the existence of global EPU causes the positive shock of VA on economic growth to have a more delayed impact.

We then turned to investigating the impulse response of VA to a one-standard-deviation shock in economic growth in the presence and absence of three EPUs using the QF-VAR model and the MF-VAR model, as reported in Figure 2(b) and (d). We found that the impulse response of VA was not monotonic; a positive economic growth shock showed a robust initial influence on VA in the short term, regardless of the impact of the three EPUs. The effects of shocks disappeared in the long run.

Specifically, considering the effect of an economic growth shock on VA, as shown in the QF-VAR model in Figure 2(b), compared to the absence of EPU, there was no significant difference in the impulse response route except that the existence of the three EPUs weakened the short-term positive impact strength. The impulse response functions of the MF-VAR model in Figure 2(d) look different from those of the QF-VAR model. The impulse response of VA to a one-standard-deviation shock in economic growth did not show such a strong initial response in the presence of global EPU.

Variance decomposition analyses

The impulse responses denote how the series reacts to exogenous shocks over a period of time, while the variance decomposition shows how much of the forecast error variance of the dependent variable can be explained by exogenous shocks to the independent variables (Katircioglu, 2014). Therefore, forecast error variance shows how much of the variation in VA and economic growth changes are due to the variations in the variables included in the model. Table 4 shows the results of the forecast error variance decomposition for tourist and economic growth in the presence and absence of the three EPUs using the QF-VAR model and the MF-VAR model, respectively.

Table 4.

Comparative analysis of the forecast error variance decomposition of VA and GDP in the absence and presence of the three EPUs.

		Variance decomposition of VA		Variance decomposition of GDP
Period		VA	GDP	VA	GDP
h = 4	${QF - VAR}_{without EPU}$	0.9427	0.0573	0.0919	0.9081
h = 4	${MF - VAR}_{without EPU}$	—	—	—	—
h = 8	${QF - VAR}_{without EPU}$	0.9355	0.0645	0.0876	0.9124
h = 8	${MF - VAR}_{without EPU}$	—	—	—	—
h = 12	${QF - VAR}_{without EPU}$	0.9341	0.0659	0.0842	0.9158
h = 12	${MF - VAR}_{without EPU}$	—	—	—	—
h = 4	${QF - VAR}_{with HK EPU}$	0.8795	0.0385	0.0723	0.8156
h = 4	${MF - VAR}_{with HK EPU}$	0.6995	0.0445	0.0775	0.7700
h = 8	${QF - VAR}_{with HK EPU}$	0.8668	0.0508	0.0702	0.8234
h = 8	${MF - VAR}_{with HK EPU}$	0.6644	0.0594	0.0752	0.7650
h = 12	${QF - VAR}_{with HK EPU}$	0.8610	0.0575	0.0664	0.8322
h = 12	${MF - VAR}_{with HK EPU}$	0.6540	0.0655	0.0700	0.7659
h = 4	${QF - VAR}_{with Chinese EPU}$	0.8811	0.0302	0.0740	0.8115
h = 4	${MF - VAR}_{with Chinese EPU}$	0.7853	0.0342	0.0691	0.7569
h = 8	${QF - VAR}_{with Chinese EPU}$	0.8741	0.0373	0.0771	0.8100
h = 8	${MF - VAR}_{with Chinese EPU}$	0.7483	0.0352	0.0670	0.7467
h = 12	${QF - VAR}_{with Chinese EPU}$	0.8713	0.0393	0.0757	0.8164
h = 12	${MF - VAR}_{with Chinese EPU}$	0.7417	0.0370	0.0650	0.7522
h = 4	${QF - VAR}_{with Global EPU}$	0.7801	0.0324	0.0638	0.7049
h = 4	${MF - VAR}_{with Global EPU}$	0.7277	0.0065	0.1144	0.5215
h = 8	${QF - VAR}_{with Global EPU}$	0.7612	0.0336	0.0664	0.7031
h = 8	${MF - VAR}_{with Global EPU}$	0.7106	0.0078	0.1187	0.5086
h = 12	${QF - VAR}_{with Global EPU}$	0.7599	0.0347	0.0660	0.7099
h = 12	${MF - VAR}_{with Global EPU}$	0.7067	0.0096	0.1184	0.5108

Note: VA: visitor arrivals; EPU: economic policy uncertainty; QF-VAR: quarterly frequency vector autoregressive; MF-VAR: MF-VAR: mixed frequency vector autoregressive. In this table, we perform the forecast error variance decomposition of VA and GDP series of prediction horizons h = 4, 8, and 12 quarters. Critical comparisons are (1) whether the forecast error variance decomposition of VA and GDP explained by itself and each other will change after adding different EPUs and (2) whether the results of the QF-VAR and MF-VAR models are different.

The boldface values in the table indicate the period when the variance decomposition value changes significantly.

Considering the long-run forecast error variance of h = 12, we saw first that the forecast error variance of VA was explained 6.59% by economic growth and 93.41% by itself when not considering the three EPUs. In the presence of the three EPUs, we found that the existence of the three EPUs weakened the explanatory power of economic growth and VA both in the QF-VAR model and the MF-VAR model. Compared to the presence of the three EPUs, the explanatory power of VA was lower in the MF-VAR model than in the QF-VAR model. However, the explanatory power of economic growth in regard to VA was similar in the two models, except that the existence of global EPU severely weakened the explanatory power of economic growth in regard to VA in the MF-VAR model. The results of the variance decomposition analysis confirmed the findings from the impulse responses. The impulse response of VA to a one-standard-deviation shock in economic growth was not strong in the presence of global EPU.

Second, 8.42% of the forecast error variance of economic growth was explained by VA and 91.58% by economic growth itself in the absence of the three EPUs. Consistent with the forecast error variance results related to VA, the existence of the three EPUs also weakened the explanatory power of economic growth and VA in both the QF-VAR model and the MF-VAR model. Compared to the presence of the three EPUs, the explanatory power of economic growth on itself was lower in the MF-VAR model than in the QF-VAR model. The existence of global EPU severely weakened the explanatory power of economic growth on itself. Correspondingly, the existence of global EPU significantly enhanced the explanatory power of VA in regard to economic growth by 11.84% in the MF-VAR model. The results of the variance decomposition analysis also confirmed the findings regarding the impulse response of economic growth to a one-standard-deviation shock in VA and the way in which this had a more delayed impact in the presence of global EPU.

Conclusion

This article has studied the changing relationship between tourism and economic growth in the presence or absence of EPU shock using a new tool: the MF-VAR model. The study period spanned April 1998 to March 2018. An advantage of the MF-VAR model is that monthly EPU can be allowed to have heterogeneous impacts on the other quarterly series, unlike the QF-VAR model, which aggregates monthly time series into quarterly series (Motegi and Sadahiro, 2018). This study added three EPUs (Hong Kong, Chinese, and global EPU) to ensure that the analysis was comprehensive. The model consisted of monthly EPU (one of the three mentioned above), quarterly GDP, and quarterly VA.

Our findings are as follows. First, the positive shocks of VA and economic growth have an initial positive influence on both parties in the short term, but then this becomes a negative influence. In the long run, the influence disappears. Second, the existence of the three EPUs does not affect the direction of the impulse response; rather, it primarily influences the size of the impact. Moreover, the QF-VAR model does not fully capture the impact of EPU on the relationship between Hong Kong’s tourism industry and economic growth. In the MF-VAR model, the existence of the three EPUs makes the shocks have a more delayed impact. In particular, the existence of global EPU makes the shock from VA have a more delayed impact on GDP.

Further analysis of the results reveals that tourism demand is highly sensitive to EPU in any direction and that different ranges of EPU have different impact intensities. Specifically, in recent years, mainland China’s economy has maintained stable development; modernization and the development of a consumer economy have also increased people’s demand for tourism (Zhang et al., 2018). Hong Kong is popular with mainland Chinese tourists due to its convenient visa system and lower taxes, with shopping being considered a prominent push factor (Choi et al., 2008; Huang and Hsu, 2005; Qiu Zhang et al., 2017). Indeed, the growth in mainland VA has also maintained a steady trend in recent years. Although EPU in China has a short-term negative impact on the relationship between Hong Kong’s tourism and economic growth, compared with the impact of Hong Kong and global EPU, its impact intensity is the smallest. In addition, we found that global EPU has a significant impact on the relationship between tourism and economic growth. It seems that tourists’ travel plans are more vulnerable to the impact of the global economic environment. Tourists will more easily postpone travel plans to avoid possible future risks when there is EPU (Gozgor and Ongan, 2017).

As tourism is one of Hong Kong’s major industries (Li et al., 2013), its revenue inflow is a major driver of economic growth. If economic uncertainty increases, especially changes in global EPU, then, compared with basic necessities such as food and housing, consumers may give priority to canceling or postponing their travel or vacation plans to hedge future economic risks (Gozgor and Ongan, 2017; Madanoglu and Ozdemir, 2019). So, first of all, Hong Kong policymakers should be encouraged to strive to reduce Hong Kong’s economic uncertainties while being alert to the impact of Chinese and global EPU. To create a stable economic environment for Hong Kong’s tourism industry, Hong Kong policymakers should help tourism professionals to formulate targeted marketing plans by providing useful information. Second, the managers of Hong Kong’s tourism-related industries, such as hotel practitioners, need to be aware of the long-term impact of all EPUs, especially global EPU, on their businesses. To maintain expected future demand in Hong Kong, effective room capacity, cost-reducing labor allocation plans, and cost-effective marketing campaigns should be designed to cope with possible decline in demand and room prices based on EPU information (Madanoglu and Ozdemir, 2019).

Overall, from the perspective of theoretical contributions, this study is based on a new perspective on the mechanism of action between various factors, to analyze the impact of different ranges of EPU on the relationship between Hong Kong’s tourism and economic growth. Second, this research is one of few studies to apply the mixed-frequency method to a study of the impact of EPU on Hong Kong’s tourism industry; the mixed-frequency approach yields more valuable economic insights than the single-frequency approach.

However, this study has some limitations. First, we only focus on Hong Kong due to tourism being one of Hong Kong’s four pillar industries that play a vital role in Hong Kong’s economy. However, focusing on one specific economy raises the issue of the generalizability of the results. We will continue to expand this study to other regions and countries to promote this line of research. Second, we have only considered the impact of the representative EPU indexes developed by Baker et al. (2016). Future studies could further study the impact of uncertainty on tourism in various situations by considering various indicators of uncertainty.

Footnotes

Acknowledgements

The authors would like to acknowledge the Editor Professor Raffaele Scuderi and two anonymous reviewers, whose comments significantly improved our paper.

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: This research was supported by the “Natural Science Foundation of China” (Grant No. NSFC71673233), “Humanities and Social Science Fund of the Ministry of Education” (Grant No. 20YJC79007), “2017 National Statistical Science Research Project” (Grant No. 2017LD01), and “Fundamental Research Funds for the Central Universities.”

ORCID iDs

Han Liu

Ying Liu

Yonglian Wang

References

Akadiri

Alola

Uzuner

(2019) Economic policy uncertainty and tourism: evidence from the heterogeneous panel. Current Issues in Tourism. Epub ahead of print 11 November 2019. DOI: 10.1080/13683500.2019.1687662.

Alegre

Mateo

Pou

(2013) Tourism participation and expenditure by Spanish households: the effects of the economic crisis and unemployment. Tourism Management 39: 37–49.

Alonso-Almeida

MdM

Bremser

(2013) Strategic responses of the Spanish hospitality sector to the financial crisis. International Journal of Hospitality Management 32: 141–148.

Baker

Bloom

Davis

(2013) Measuring Economic Policy Uncertainty. Chicago: Chicago Booth Research Paper, pp. 13–20.

Baker

Bloom

Davis

(2016) Measuring economic policy uncertainty. Quarterly Journal of Economics 131: 1593–1636.

Balaguer

Cantavella-Jordá

(2002) Tourism as a long-run economic growth factor: the Spanish case. Applied Economics 34: 877–884.

Balli

Shahzad

SJH

Salah Uddin

(2018) A tale of two shocks: What do we learn from the impacts of economic policy uncertainties on tourism? Tourism Management 68: 470–475.

Bernanke

(1983) Irreversibility, uncertainty, and cyclical investment. The Quarterly Journal of Economics 98: 85–106.

Bloom

Bond

Van Reenen

(2007) Uncertainty and investment dynamics. The Review of Economic Studies 74: 391–415.

10.

Brogaard

Detzel

(2015) The asset-pricing implications of government economic policy uncertainty. Management Science 61: 3–18.

11.

Chiweza

Aye

(2018) The effects of oil price uncertainty on economic activities in South Africa. Cogent Economics and Finance 6: 1–17.

12.

Choi

T-M

Liu

S-C

Pang

K-M

, et al. (2008) Shopping behaviors of individual tourists from the Chinese Mainland to Hong Kong. Tourism Management 29: 811–820.

13.

Cooper

(2005) Japanese tourism and the SARS epidemic of 2003. Journal of Travel & Tourism Marketing 19: 117–131.

14.

Demir

Ersan

(2018) The impact of economic policy uncertainty on stock returns of Turkish tourism companies. Current Issues in Tourism 21: 847–855.

15.

Dickey

Fuller

(1979) Distribution of the estimators for autoregressive time series with a unit root. Journal of the American Statistical Association 74: 427–431.

16.

Dickey

Fuller

(1981) Likelihood ratio statistics for autoregressive time series with a unit root. Econometrica 49: 1057–1072.

17.

Dragouni

Filis

Gavriilidis

, et al. (2016) Sentiment, mood and outbound tourism demand. Annals of Tourism Research 60: 80–96.

18.

Dritsakis

(2012) Tourism development and economic growth in seven Mediterranean countries: a panel data approach. Tourism Economics 18: 801–816.

19.

Ersan

Akron

Demir

(2019) The effect of European and global uncertainty on stock returns of travel and leisure companies. Tourism Economics 25: 51–66.

20.

Faulkner

(2001) Towards a framework for tourism disaster management. Tourism Management 22: 135–147.

21.

Fletcher

Hillingdon

(2008) Tourism activity, terrorism and political instability within the Commonwealth: the cases of Fiji and Kenya. International Journal of Tourism Research 10: 537–556.

22.

Frechtling

(1982) Tourism trends and the business cycle: tourism in recession. Tourism Management 3: 285–290.

23.

Ghosh

(2019) Uncertainty, economic growth its impact on tourism, some country experiences. Asia Pacific Journal of Tourism Research 24: 83–107.

24.

Ghysels

(2016) Macroeconomics and the reality of mixed frequency data. Journal of Econometrics 193: 294–314.

25.

Ghysels

Hill

Motegi

(2016) Testing for Granger causality with mixed frequency data. Journal of Econometrics 192: 207–230.

26.

Giavazzi

McMahon

(2012) Policy uncertainty and household savings. The Review of Economics and Statistics 94: 517–531.

27.

Gonçalves

Kilian

(2004) Bootstrapping autoregressions with conditional heteroskedasticity of unknown form. Journal of Econometrics 123: 89–120.

28.

Gozgor

Demir

(2018) The effects of economic policy uncertainty on outbound travel expenditures. Journal of Competitiveness 10: 5–15.

29.

Gozgor

Ongan

(2017) Economic policy uncertainty and tourism demand: empirical evidence from the USA. International Journal of Tourism Research 19: 99–106.

30.

Granger

CWJ

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

31.

Hall

(2010) Crisis events in tourism: subjects of crisis in tourism. Current Issues in Tourism 13: 401–417.

32.

Hong Kong Legislative Council (2015) Hong Kong’s tourism industry. Research Brief, Issue No. 6, 2014-2015. Available at: http://www.legco.gov.hk/research-publications/english/1415rb06-hong-kongs-tourism-industry-20150805-e.pdf (accessed 20 April 2019).

33.

Hong Kong Tourism Board (2016) Visitor arrival statistics-Jan 2016. Available at: http://securepartnernet.hktb.com/filemanager/intranet/dept_info/private_20/paper/VAS2002/VAS2016/VAS_01_2016_0.pdf (accessed 20 April 2019).

34.

Huang

Hsu

CHC

(2005) Mainland Chinese residents’ perceptions and motivations of visiting Hong Kong: evidence from focus group interviews. Asia Pacific Journal of Tourism Research 10: 191–205.

35.

Hylleberg

Engle

Granger

CWJ

, et al. (1990) Seasonal integration and cointegration. Journal of Econometrics 44: 215–238.

36.

Işık

Sirakaya-Turk

Ongan

(2019) Testing the efficacy of the economic policy uncertainty index on tourism demand in USMCA: theory and evidence. Tourism Economics. Epub ahead of print 22 November 2019. DOI: 10.1177/1354816619888346.

37.

Katircioglu

(2009) Revisiting the tourism-led-growth hypothesis for Turkey using the bounds test and Johansen approach for cointegration. Tourism Management 30: 17–20.

38.

Katircioglu

(2014) International tourism, energy consumption, and environmental pollution: the case of Turkey. Renewable & Sustainable Energy Reviews 36: 180–187.

39.

Kim

Chen

M-H

(2006) Tourism expansion and economic development: the case of Taiwan. Tourism Management 27: 925–933.

40.

Kim

Lee

C-K

Mjelde

(2018) Impact of economic policy on international tourism demand: the case of Abenomics. Current Issues in Tourism 21: 1912–1929.

41.

Kuo

H-I

Chang

C-L

Huang

B-W

, et al. (2009) Estimating the impact of avian flu on international tourism demand using panel data. Tourism Economics 15: 501–511.

42.

Kuo

H-I

Chen

C-C

Tseng

W-C

, et al. (2008) Assessing impacts of SARS and avian flu on international tourism demand to Asia. Tourism Management 29: 917–928.

43.

Lee

Chang

(2008) Tourism development and economic growth: a closer look at panels. Tourism Management 29: 180–192.

44.

Lee

(2008) Tourism and economic growth: the case of Singapore. Regional and Sectoral Economic Studies 8: 89–98.

45.

Lepp

Gibson

(2003) Tourist roles, perceived risk and international tourism. Annals of Tourism Research 30: 606–624.

46.

Song

Cao

, et al. (2013) How competitive is Hong Kong against its competitors? An econometric study. Tourism Management 36: 247–256.

47.

Blake

Cooper

(2010) China’s tourism in a global financial crisis: a computable general equilibrium approach. Current Issues in Tourism 13: 435–453.

48.

Turner

Cui

(2016) Confrontations and concessions: an everyday politics of tourism in three ethnic minority villages, Guizhou Province, China. Journal of Tourism and Cultural Change 14: 45–61.

49.

Liu

McKercher

(2016) The impact of visa liberalization on tourist behaviors—the case of China outbound market visiting Hong Kong. Journal of Travel Research 55: 603–611.

50.

Liu

Song

(2017) New evidence of dynamic links between tourism and economic growth based on mixed-frequency granger causality tests. Journal of Travel Research 57: 899–907.

51.

Madanoglu

Ozdemir

(2019) Economic policy uncertainty and hotel operating performance. Tourism Management 71: 443–452.

52.

McKercher

Chon

(2004) The over-reaction to SARS and the collapse of Asian tourism. Annals of Tourism Research 31: 716–719.

53.

Mohd Yusof

Usman

Mahfudz

, et al. (2018) Macroeconomic shocks, fragility and home financing in Malaysia: Can rental index be the answer? Journal of Islamic Accounting and Business Research 9: 17–44.

54.

Motegi

Sadahiro

(2018) Sluggish private investment in Japan’s lost decade: mixed frequency vector autoregression approach. The North American Journal of Economics and Finance 43: 118–128.

55.

Perron

(2001) Lag length selection and the construction of unit root tests with good size and power. Econometrica 69: 1519–1554.

56.

C-O

(2005) The contribution of tourism development to economic growth in the Korean economy. Tourism Management 26: 39–44.

57.

Okumus

Altinay

Arasli

(2005) The impact of Turkey’s economic crisis of February 2001 on the tourism industry in Northern Cyprus. Tourism Management 26: 95–104.

58.

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: 308–316.

59.

Ozturk

Acaravci

(2009) On the causality between tourism growth and economic growth: empirical evidence from Turkey. Transylvanian Review of Administrative Sciences 5: 73–81.

60.

Papatheodorou

Rosselló

Xiao

(2010) Global economic crisis and tourism: consequences and perspectives. Journal of Travel Research 49: 39–45.

61.

Payne

Mervar

(2010) Research note: the tourism–growth nexus in Croatia. Tourism Economics 16: 1089–1094.

62.

Perles-Ribes

Ramón-Rodríguez

Rubia-Serrano

, et al. (2016) Economic crisis and tourism competitiveness in Spain: Permanent effects or transitory shocks? Current Issues in Tourism 19: 1210–1234.

63.

Phillips

PCB

Perron

(1988) Testing for a unit root in time series regression. Biometrika 75: 335–346.

64.

Pizam

Fleischer

(2002) Severity versus frequency of acts of terrorism: Which has a larger impact on tourism demand? Journal of Travel Research 40: 337–339.

65.

Qiu

Fan

DXF

Lyu

, et al. (2019) Analyzing the economic sustainability of tourism development: evidence from Hong Kong. Journal of Hospitality & Tourism Research 43: 226–248.

66.

Qiu Zhang

Fan

DXF

Tse

TSM

, et al. (2017) Creating a scale for assessing socially sustainable tourism. Journal of Sustainable Tourism 25: 61–78.

67.

Saha

Campbell

(2017) Does political and economic freedom matter for inbound tourism? A cross-national panel data estimation. Journal of Travel Research 56: 221–234.

68.

Singh

Das

Jana

, et al. (2018) A wavelet analysis for exploring the relationship between economic policy uncertainty and tourist footfalls in the USA. Current Issues in Tourism 22: 1–8.

69.

Song

Lin

(2009) Impacts of the financial and economic crisis on tourism in Asia. Journal of Travel Research 49: 16–30.

70.

Song

Lin

Witt

, et al. (2011) Impact of financial/economic crisis on demand for hotel rooms in Hong Kong. Tourism Management 32: 172–186.

71.

Sonmez

(1998) Tourism, terrorism, and political instability. Annals of Tourism Research 25: 416–456.

72.

Tang

C-H

Jang

(2009) The tourism–economy causality in the United States: a sub-industry level examination. Tourism Management 30: 553–558.

73.

Tiwari

Das

Dutta

(2019) Geopolitical risk, economic policy uncertainty and tourist arrivals: evidence from a developing country. Tourism Management 75: 323–327.

74.

Tolkach

(2018) A view of occupy central impacts on Hong Kong tourism from the other side of the great firewall: a rejoinder to Luo & Zhai. Tourism Management 67: 307–311.

75.

Tsui

WHK

Balli

Tan

DTW

, et al. (2018) New Zealand business tourism: exploring the impact of economic policy uncertainties. Tourism Economics 24: 386–417.

76.

Wong

Buckley

(2015) Protesters confront mainland shoppers in Hong Kong. New York Times. Available at: https://sinosphere.blogs.nytimes.com/2015/02/08/protesters-confront-mainland-shoppers-in-hong-kong (accessed 20 April 2019).

77.

(2018) Causality between European economic policy uncertainty and tourism using wavelet-based approaches. Journal of Travel Research 58: 1347–1356.

78.

Zellner

Montmarquette

(1971) A study of some aspects of temporal aggregation problems in econometric analyses. The Review of Economics and Statistics 53: 335–342.

79.

Zhang

Han

Pan

, et al. (2015) Economic policy uncertainty and capital structure choice: evidence from China. Economic Systems 39: 439–457.

80.

Zhang

Wong

PPY

Lai

(2018) A geographic analysis of hosts’ irritation levels towards mainland Chinese cross-border day-trippers. Tourism Management 68: 367–374.