A Study On The Influencing Factors Of Tourism Demand From Mainland China To Hong Kong

Abstract

Tourism research increasingly uses search query data to forecast demand, but the literature rarely explores the mechanisms of the factors influencing demand. A time-varying parameter factor vector auto-regression model is constructed based on Baidu Index on six aspects (dining, shopping, transportation, tours, attractions, and lodging) of tourism demand from January 2011 to March 2019. The model can quantitatively and comprehensively analyze the mechanisms of tourism demand and its six important influencing factors, and can provide suggestions for subsequent planning, construction, and services in the tourism industry. The empirical results show that the relationship between the six factors and domestic tourism demand is time-varying. Dining, attractions, and shopping have a driving effect on tourism demand, and are thus stimulative factors; transportation, tours, and lodging hinder the growth of tourism demand, and are thus baffle factors.

Keywords

influencing factor tourism demand TVP-Fa-VAR generalized dynamic factor model search query data

Introduction

Tourism is one of the four pillar industries of the Hong Kong economy (G. Li et al., 2013). As reported in the latest publication of the Hong Kong Tourism Board (2019), tourist volume reached 65.1 million in 2018, an increase of 11.4% from the previous year. With arrivals amounting to 51 million, mainland China remains Hong Kong’s key source market, highlighting the importance of promoting the mainland’s tourism demand in the development of tourism in Hong Kong (Hong Kong Tourism Board, 2019). However, there are many factors that determine the growth of Hong Kong’s tourism demand (Ahn & McKercher, 2015; Suntikul et al., 2019; Wondirad & Agyeiwaah, 2016), and it would be of practical significance to explore the internal mechanisms of tourism demand and its influencing factors.

Many studies have been carried out on the influencing factors of tourism demand, such as Din et al. (2017) and Saayman and Saayman (2008). However, on the one hand, little research conducts comprehensive quantitative analysis of the six elements of dining, shopping, transportation, tours, attractions, and lodging simultaneously, due to a lack of data. On the other hand, advancements in information technology have inspired the development of search query data analytics, which can accurately capture the search behavior of Internet users. As tourists increasingly use search engines to get information about destinations, these data contain valuable information about their interests. It is suggested that search query data can offer significant benefits to forecasters in tourism (Bangwayo-Skeete & Skeete, 2015; Camacho & Pacce, 2018). But most of the existing literature uses search query data to predict demand, rarely to explore the relationship between tourism demand and its influencing factors. Efforts should be made to fill this gap.

This article makes contributions in three main aspects. First, different from previous research, we explore the internal mechanisms between tourism demand and its influencing factors, based on the search query data. More effective tourism facilities can be provided by tourism investors and local governments to increase tourism demand via this study. Second, methodologically, this article builds a time-varying parameter factor vector autoregressive (TVP-Fa-VAR) model based on a generalized dynamic factor model (GDFM) and a time-varying parameter stochastic volatility vector autoregressive model with Bayesian inference (TVP-SV-VAR), to analyze the influencing mechanism. In this model, GDFM is used to extract six factors, which allows factor sequences and special component sequences to be dynamic (Forni et al., 2000). Furthermore, the dynamic time-varying coefficients of the TVP-SV-VAR model can describe the time-varying relationship between tourism demand and its influence factors in different periods of the data sample from January 2011 to March 2019 (Nakajima et al., 2011). Thus, it will help us clarify the importance of various factors in different periods. Third, tourism demand has been divided into three regimes based on the estimation results of the Markov regime-switching model. Therefore, we can analyze the different effects of the six factors when tourism demand shows different trends.

In conclusion, the main objective of this article is to analyze the relationships between tourism demand and the aforementioned six factors, and provide suggestions for follow-up tourism planning, construction, and services, to enhance the quality and quantity of Hong Kong’s tourism demand. Therefore, in this study, we select keywords regarding the six aspects of dining, transportation, attractions, shopping, tours, and lodging, which are searched by people using the Baidu search engine, and most of them are potential tourists from Mainland China. In line with our research purposes, we built the TVP-Fa-VAR model to analyze the mechanism between tourism demand and the six important factors from a dynamic perspective. Specifically, the model combines the GDFM and TVP-SV-VAR models. The results are as follows: (a) in terms of 3D impulse response, the short-term effects of the six factors on tourist volume are different, and the long-term relationship tends to be stable. However, the time-varying patterns in the relationships between the six factors and domestic tourism demand differ. Specifically, tourism demand responds more intensely to shock in dining, tours, and lodging than other factors over the period; (b) in terms of regimes of tourism demand, the dining, attractions, and shopping factors have a driving effect on tourism demand, and so are stimulative factors, while the transportation, tours, and lodging factors hinder the growth of tourism demand, and so are baffling factors; (c) in terms of the degree of influence, the dining and transportation factors have the most intense effects.

The structure of this article is as follows. Section 2 provides a literature review. Section 3 explains the data and the econometric methodology. Section 4 presents the empirical results. Section 5 draws conclusions with recommendations for application. Finally, Section 6 discusses the limitations of this work and directions for future research.

Literature Review

Influencing Factors of Tourism Demand

Tourism demand is motivated by a range of factors, which have generated numerous studies in the field of tourism (D. C. Wu et al., 2017). The influencing factors of tourism demand commonly accepted in the literature can be summarized into several different types: infrastructural, economic, political, environmental, customer-related, seasonality, vacation policies, and random (J. L. Chen et al., 2019; Garcia et al., 2015; Habibi, 2017; Zhang et al., 2016). However, influencing factors between demand and supply are importantly reflected in the six aforementioned aspects: dining, shopping, transportation, tours, attractions, and lodging. The six factors are derived from the data of search queries related to the demand of tourists and the supply of the tourism industry, which can be used as an important representation of the tourism industries (Aratuo & Etienne, 2019).

Scholars have carried out numerous studies and reviews on the six factors of tourism demand. (a) For example, food has become an indispensable part of travel because it is an important way to convey the cultural life of tourist destinations to tourists. Asero and Tomaselli (2015) highlight the importance of food to tourism demand in Sicily. Pérez Gálvez et al. (2017) show that food has a positive impact on consolidating tourist destinations. (b) In addition, transportation is an element that must be taken into consideration when tourists make travel plans (Cosma, 2017). Kulendran (1996) applies cointegration analysis to explore the long-running relationship between tourist volumes to Australia from the United States, Japan, the United Kingdom, and New Zealand and airfare, and the results show that tourists are responsive to changes in airfare. Divisekera (2016) indicates significant interdependencies between transport and tourism demand in Australia, New Zealand, the United States, and the United Kingdom. (c) Activities are also attractive to tourists, and tourists choose interesting destinations with more recreational activities (Heagney et al., 2018; Reintinger et al., 2016). (d) Moreover, some researchers consider that shopping is the crucial factor of tourism demand (Divisekera, 2009; Gunadhi & Boey, 1986) or at least identify it as an important part of tourism (Divisekera, 2010). Visitors often have a strong interest in the goods of destinations and traditional products with local characteristics, and some tourists’ main tourist motive is shopping. (e) However, Y. Chen et al. (2014), based on mainland China’s tourism demand in Hong Kong, show that package tours can lead to low levels of tourist satisfaction and deter tourists’ behavioral intention over time. Package tours or independent tours have an impact on the choice of a tourist destination. (f) Additionally, lodging is an essential demand for tourists. Narayan (2004) adopts cointegration techniques and error correction models to investigate the short-run and long-run relationships between tourist volumes and relative hotel prices in Fiji (hotel price index in Fiji relative to the tourist’s source country), and finds that relative hotel prices have a negative impact on tourism volumes. Therefore, lodging is also one of the factors affecting tourism demand.

However, in the literature, there is no comprehensive, simultaneous quantitative analysis of these six influencing factors.

Search Query Data Analytics in Tourism Research

Search engines have become a key way to find information online, with the accelerated adoption of the Internet. Google, the largest information provider, not only uses the information to improve its search algorithms but also sells it to other companies and to produce customized publicity for Internet users. Some private companies who exploit the information in search query data will be able to make decisions based on a deeper understanding of customers, which is a clear competitive advantage (Cabrera-Sánchez & Villarejo-Ramos, 2020). In addition, search query data are especially valuable to scholars, because they contain abundant and timely information. In recent years, search query data have been applied to many aspects. The original articles using search data are mainly found in the field of epidemiology (Ginsberg et al., 2009; Polgreen et al., 2008). Since then, search query data have been widely used in ranking universities (Vaughan & Romero-Frías, 2014), gathering opinions (Baram-Tsabari & Segev, 2011), and economics. For example, researchers apply search query data to forecast unemployment rates (H. Choi & Varian, 2012; Nikolaos Askitas, 2009), consumption (Vosen & Schmidt, 2011), and house prices (McLaren & Shanbhogue, 2011; L. Wu & Brynjolfsson, 2009).

Some research in the field of tourism has attempted to introduce search query data analytics (Fesenmaier et al., 2011; Volchek et al., 2018). Most researchers want to examine the accuracy of search query data in forecasting (Bokelmann & Lessmann, 2019; Wen et al., 2019). For instance, Pan et al. (2012) first investigate the performance of search query data in predicting demand for hotel rooms, and find that prediction accuracy has increased significantly. X. Li et al. (2017) adopt online data to construct indexes based on the GDFM and principal component analysis (PCA), and then forecast tourism demand, including tourist numbers and hotel occupancy. Their article shows the validity of the combination of search data and GDFM. Camacho and Pacce (2018) examine whether Google search data help forecast real-time check-in and overnight stays of tourists in Spain, and find conclusive evidence that tourism-related queries can improve tourism predictions. Sun et al. (2019) apply machine learning and Google and Baidu indices to predict tourist volumes and compare their predictive ability. The results demonstrate that using a composite index can significantly improve forecasting performance in accuracy and robustness.

However, according to our review, almost no research has used search query data to specifically analyze the factors influencing tourism demand. Doing so is worthwhile because search query data contain more information than traditional data used in tourism research; and tourists’ plans are intrinsically complex and multidimensional. With these new data, we can better study tourist behavior in many aspects, such as dining, shopping, transportation, tours, attractions, and lodging.

Generalized Dynamic Factor Model

A great deal of search query data faces the curse of dimensionality in VAR models. This can be countered by constructing indices, such as extracting common components from large search queries. Some articles apply PCA to create indices (S. W. Li et al., 2018; X. Li et al., 2015; Nilashi et al., 2015). In general, PCA is the first factor extracted from these data. The search query data are dynamically related to each other, but this method fails to comprehensively display the dynamics among all queries. In addition, researchers use a dynamic factor model (DFM) to deal with this large amount of information (Camacho & Pacce, 2018). This model allows factor sequences and special component sequences to be dynamic. However, it requires the spectral density matrix of a particular component term to be diagonal and the factor sequence to be unrelated to the sequence of the particular component. In most practical applications, the orthogonality assumption of special component terms is not valid. Forni et al. (2000) proposed the GDFM, which solved the drawbacks of the above two models. To conclude, this model allows for the dynamic representation of variables and special components, and also relaxes the requirements of the diagonal matrix.

The GDFM has been commonly adopted to analyze economic or financial activities (Barigozzi & Hallin, 2017; Triacca & Focker, 2014). For example, Altissimo et al. (2010) estimate in real time the current state of the economy based on the GDFM. Chang et al. (2017) investigate the application of the GDFM to identify market-wide liquidity across foreign exchange markets. Armeanu et al. (2017) apply it to forecast short-term economic growth in Romania. However, in the field of tourism, GDFM is rarely used. We found that X. Li et al. (2017) were the first to use the GDFM to construct indices on the basis of search query data. They compared a GDFM-based index with a PCA-based index and found the performance of the former to be superior. Therefore, the GDFM is adopted in this study to construct the six factors separately, since the model can relax the requirements for calculation and better display the dynamic characteristics among variables.

Methodology

Data

G. Li et al. (2005) and Song and Li (2008) have produced excellent reviews of tourism demand and found that the frequently used substitution indicators for demand are statistics related to tourist volume, tourist expenditure, and the number of nights spent in a destination. Tourist volume is found to be the most commonly used indicator among these, which is also the reason we choose tourist volume to represent tourism demand in our study.

The purpose of this study is to explore the influencing factors of tourism demand from mainland China to Hong Kong. Monthly data regarding the tourist volume from mainland China to Hong Kong were obtained from the Hong Kong Tourism Board. The tourism demand represented by the data of tourist volume is abbreviated as td, and ranges from January 2011 to March 2019. We adopt this time frame because we obtain search query data during the same period. For the sake of modeling convenience, the logs of tourist volumes are defined as logtd.

Baidu is now the most popular search engine and has the biggest market share in China since Google exited the mainland Chinese market in 2010 (Yang et al., 2015). Therefore, considering our area of interest, we choose the Baidu Index to describe tourist behavior in this study. First, we determine keywords. According to some existing articles (X. Li et al., 2017; Sun et al., 2019), we take the following steps to select keywords, as shown in Supplement Figure 1 (available in the online supplement).

Classify keywords. This article aims to examine the mechanisms between influencing factors and tourism demand from the perspective of tourist demand and industry supply. Six sets of search query data are selected as the explanatory variables: dining, transportation, attractions, shopping, tours, and lodging. These six categories reflect what tourists are paying attention to when making tourism plans.

Select initial keywords. Several initial keywords are searched as seed keywords for each category in the Baidu Index, based on X. Li et al. (2017) and Wen et al. (2019). These seed keywords are adopted to retrieve more keywords in the following step.

Extend keywords. The aim of this step is to obtain more potential keywords to represent tourists’ interests. We iteratively obtain recommended keywords using a demand map interface provided by Baidu, including “Hong Kong food guidance,” “Hong Kong travel map,” “Hong Kong tourist attractions,” “Hong Kong shopping guide,” “Hong Kong one day trip,” “Hong Kong accommodation guide,” which is highly correlated to the abovementioned seed keywords.

Check keywords. Baidu does not provide how many times the keywords were searched. The abovementioned keywords must be manually checked in terms of availability for downloading. During this process, unavailable keywords are eliminated.

Finally, according to Wen et al. (2019), monthly data regarding 101 search query series are extracted from January 2011 to March 2019. The data frequency we determined is the same as the tourist volume. The sample is restricted to 8 years, since Baidu Index data are only available for this period.

TVP-Fa-VAR

We construct a TVP-Fa-VAR model based on the GDFM and TVP-SV-VAR model. The above process is divided into two steps. First, the GDFM is used to extract factors. Second, with these factors, the TVP-SV-VAR model is applied to estimate the results and obtain the impulse response function.

Generalized dynamic factor model. Search query data require the reduction of dimensionality, and this can be done by extracting common components using the GDFM, a method that is proposed by Forni et al. (2000). This model combines the advantages of the DFM and PCA, which allow for the dynamically representing parameters and also relaxes the operational requirements. According to Forni et al. (2000), the observed variables ${x_{i t}, i = 1, \dots, n, t = 1, \dots, T}$ represent six sets of search query data respectively and are extracted by Equations 1 and 2 six times:

x_{i t} = χ_{i t} + ε_{i t}

(1)

χ_{i t} = b_{i 1} (L) f_{1 t} + b_{i 2} (L) f_{2 t} + \dots + b_{i q} (L) f_{q t}

(2)

where $χ_{i t}$ and $ε_{i t}$ are the common component and idiosyncratic component, respectively, and L is the lag operator. The common components are driven by q common factors $f_{j t}, j = 1, \dots, q$ . In detail, we determined q is based on the information criterion proposed by Hallin and Liška (2007).

TVP-SV-VAR model. In this study, we adapt Nakajima et al. (2011) TVP-SV-VAR model to explore the impact of the six factors on tourism demand. This method is chosen because this study examines the time-varying relationship between six factors and tourism demand in different periods. Standard VAR analyses have led to doubts regarding its accuracy in estimating and predicting the relationship between variables because of the fixed coefficient hypothesis (Kim et al., 1998). Therefore, scholars have modified it to expand into nonlinear models, such as the Markov region transfer vector autoregressive model (MS-VAR) and the smooth transformation autoregressive model. Comparatively, Primiceri (2005) proposed the TVP-SV-VAR model to provide a new method with which to study the time-varying characteristics of the relationship between variables. On the one hand, the model can reflect the time variation of the simultaneous relationships among the variables of the model through time-varying coefficient estimation. On the other hand, it can solve the heteroscedasticity problem of the model through time-varying volatility (Nakajima et al., 2011).

Following Nakajima et al. (2011), the TVP-SV-VAR model is given by:

y_{t} = B_{1, t} y_{t - 1} + \dots + B_{s, t} y_{t - s} + A_{t}^{- 1} \sum_{t} ε_{t}, ε_{t} : N (0, I_{k}),

(3)

where $y_{t}$ is a $k \times 1$ column vector, which denotes tourism demand and six factors. $B_{i} (i = 1, \dots, k)$ are $k \times k$ matrices of coefficients, $A_{t}$ is a time-varying structure matrix, and ∑_t $\sum_{t}$ is a $k \times k$ diagonal matrix with the diagonal as $(σ_{1 t}, σ_{2 t}, \dots σ_{k t})$ . Stacking all the coefficients into $β_{t}$ ( $k^{2} s \times 1$ vector) and defining $X_{t} = I_{k} \otimes ({y^{'}}_{t - 1}, \dots, {y^{'}}_{t - k})$ , Equation 3 can be written as

y_{t} = X_{t} β_{t} + Α_{t}^{- 1} \sum_{t} ε_{t}

(4)

Following Nakajima et al. (2011), let $a_{t} = (a_{21}, a_{31}, a_{32}, \dots, a_{k, 1}, \dots, a_{k, k - 1})$ , which is the stacked vector of the lower triangular elements in $A_{t}$ . In addition, let $h_{t} = (h_{1 t}, h_{2 t}, \dots, h_{k t})$ , where $h_{i t} = \log σ_{i t}^{2}, i = 1, \dots k; t = s + 1, \dots, n$ . Therefore, $β_{t}$ is a time-varying coefficient, $a_{t}$ reflects simultaneous relations, and $h_{t}$ represents the volatility. We assume that all the parameters follow a random walk process as follows:

\begin{matrix} β_{t + 1} = β_{t} + μ_{β_{t}}, \\ a_{t + 1} = a_{t} + μ_{a_{t}}, \\ h_{t + 1} = h_{t} + μ_{h_{t}}, \end{matrix} (\begin{array}{l} ε_{t} \\ μ_{β_{t}} \\ μ_{a_{t}} \\ μ_{h_{t}} \end{array}) ~ N [0, (\begin{matrix} I & 0 & 0 & 0 \\ 0 & Σ_{β} & 0 & 0 \\ 0 & 0 & Σ_{α} & 0 \\ 0 & 0 & 0 & Σ_{h} \end{matrix})]

(5)

For $t = s + 1, \dots, n$ ,where $β_{s + 1} ~ N (μ_{β_{0}}, \sum_{β_{0}}), a_{s + 1} ~ N (μ_{a_{0}}, \sum_{a_{0}}), h_{s + 1} ~ N$ .

In order to reduce overparameterized problems, Markov chain Monte Carlo (MCMC) methods are required to perform Bayesian inference in evaluating the posterior distributions of the parameters and hyperparameters of the above models. Under certain prior probability distributions, the MCMC algorithm produces the sample drawn from a high-dimensional posterior distribution of parameters, including unobserved latent variables (Nakajima et al., 2011).

Empirical Results

Six GDFM-Based Factors

We apply the GDFM to construct six factors (dining, transportation, attractions, shopping, tours, and lodging). The information criterion used in Hallin and Liška (2007) is chosen as the criterion for selecting the number of dynamic factors. The blue line represents the mean square error and the red line represents the number of factors. The number of factors is optional when the mean square error is in the stationary interval. Supplement Figure 2 (available in the online supplement) indicates that the mean square error has two stability intervals. Therefore, we set the number of dynamic factors to six sets of search query data based on the information criterion. Afterward, we generate six factors using Equation 1 and 2. The search query data are collected on a monthly basis, so the six factors have a monthly frequency, ranging from January 2011 to March 2019, noted as dini, tran, attr, shop, tour, lodg.

Cointegration and Granger Causality Test

It is necessary to demonstrate the feasibility of adopting the six factors series under consideration in the econometric models. This article adopts the Augmented Dickey–Fuller method to test the stationarity of the series. As shown in Supplement Table 1 (available in the online supplement), four groups of series reject the null hypothesis at the 0.01 level, so the above four time series of these data series are all stable. For the unstable sequences, we make a first order difference. The results show the null hypothesis can be rejected, indicating that the first order difference series do not have a unit root, and are stationary series. Therefore, we can perform the Johansen co-integration tests. The results of the test shown in Supplement Table 2 (available in the online supplement) evidence that the six factors and tourism demand are co-integrated, suggesting that a long-term relationship exists between the six factors and tourism demand.

Table 1

Granger Causality Tests Between the Six Factors and Tourist Volume

Null hypothesis	F statistic	Probability	Causality
logtd does not Granger Cause dini	6.8503	.00***	Yes
dini does not Granger Cause logtd	13.7906	.00***	Yes
logtd does not Granger Cause tran	32.7082	.00***	Yes
tran does not Granger Cause logtd	13.3859	.00***	Yes
logtd does not Granger Cause attr	7.4801	.00***	Yes
attr does not Granger Cause logtd	9.9234	.00***	Yes
logtd does not Granger Cause shop	8.8813	.00***	Yes
shop does not Granger Cause logtd	13.6136	.00***	Yes
logtd does not Granger Cause tour	0.0918	.91	No
tour does not Granger Cause logtd	4.5605	.01**	Yes
logtd does not Granger Cause lodg	19.0178	.00***	Yes
lodg does not Granger Cause logtd	9.5171	.00***	Yes

Note: ***, **, and * denote the rejection of the hypothesis at the .01 level, .05 level, and .10 level, respectively.

Table 2

Estimation Results for Selected Parameters

Parameter	M	SD	95% Confidence intervals	Convergence diagnostics	Inefficiency
${(\sum_{β})}_{1}$ (∑β)1	0.0023	0.0003	[0.0018, 0.0029]	0.106	6.75
${(\sum_{β})}_{2}$ (∑β)2	0.0023	0.0003	[0.0018, 0.0028]	0.749	4.49
${(\sum_{α})}_{1}$ (∑α)1	0.0056	0.0016	[0.0034, 0.0097]	0.287	29.34
${(\sum_{α})}_{2}$ (∑α)2	0.0057	0.0017	[0.0034, 0.0098]	0.077	39.41
${(\sum_{h})}_{1}$ (∑h)1	0.0056	0.0016	[0.0034, 0.0098]	0.291	23.29
${(\sum_{h})}_{2}$ (∑h)2	0.0055	0.0016	[0.0034, 0.0095]	0.669	28.65

We use the original series of the six factors to test causality and estimate a TVP-SV-VAR model. Many articles show that it is still desirable to construct a VAR model in levels, even if the variables have unit roots (S. Choi, 2017; Lin, 1996).

The purpose of the Granger Causality tests is to explore the causal relationship between six factors and tourist volume in the model. The test results of causality reported in Table 1 show the bidirectional Granger causal relationship between five factors (dini, tran, attr, shop, and lodg) and tourist volume. In addition, tour and tourist volume have unidirectional causality. These results suggest that some degrees of the causal relationship were detected among six factors and tourist volumes. Therefore, this article builds a TVP-SV-VAR model for further analysis and discussion.

TVP-SV-VAR

Estimation results in the TVP-SV-VAR model. The TVP-SV-VAR model we constructed includes seven variables (k = 7): six factors and tourist volume. Based on the Schwarz information criterion, the result is optimal when we use one lag (p = 1). The results based on 10000 MCMC draws are shown in Table 2, which gives the estimates for posterior means, standard deviations, the 95% credible intervals, the convergence diagnostics (CD), and inefficiency factors. The CD statistics show that the null hypothesis of convergence to the posterior distribution is not rejected for the parameters at the 5% significance level. The inefficiency factor measures the degree of mixing of the MCMC chain, which is low enough to posterior inference for the selected parameters. In this empirical result, the posterior mean of samples is in the 95% credible interval, so the null hypothesis that the parameters converge to the posterior distribution is accepted at the 5% significance level. The inefficiency factors of parameters are also all at reasonable levels, which indicates efficient sampling of the parameters in the TVP-SV-VAR model. Therefore, the estimation results of the parameters are generally robust and can be used for further analysis.

Impulse response analysis in the TVP-SV-VAR model. Impulse response analysis describes the application of one positive standard deviation shock on the basis of the error term, which will directly affect the current and future values of the model’s endogenous variables. It will also influence other endogenous variables through the model’s dynamic structure. Therefore, based on the TVP-SV-VAR model, this article applies a positive standard deviation to the six factors, and can obtain the impulse response path of tourism demand. For example, a positive shock hitting dining factor will affect tourism demand, so we analyze the change in the quantity and direction of tourism demand from horizons of impact and period for the entire sample. Because of the time-varying parameters, we can draw the 3D impulse responses of the six factors to the tourism demand at the past 12 periods over the whole sample, which is shown in Figure 1. The figure clearly reflects that the effect of tourist attention on tourist volume in Hong Kong is changing over time and indicates the importance of allowing for time variation in exploring the relationship between the six factors and tourism demand.

Figure 1

The 3D Impulse Responses of Tourism Demand to the Six Factors Shock

Impulse responses to a dining factor shock. The figure shows that, a shock hitting dining factor positively affects tourism demand over the entire 12-period horizon. In terms of the horizon, the impulse responses of tourism demand to dining factor shock remain positive and show a similar downward trend. Specifically, tourism demand drops significantly and quickly after the dining factor shock, and this downtrend moves downward after two months. In terms of the period, we observe that the response values of tourism demand to the dining factor shock from 2015 to 2017 are lower than those in other sample periods. (The average response values of tourism demand to the dining factor shock in 2011, 2016, and 2018 are 0.034, 0.025, and 0.027, respectively.) The results indicate that, during this period, tourist volumes were less sensitive to the attention of the dining factor. Looking back at the environment of Hong Kong tourism, we can see that some security incidents occurred during these periods. With the promotion of quality tourism services, the attraction of dining factor for tourists has improved.

Impulse responses to a transportation factor shock. The results show that the patterns in the responses of the transportation factor shocks to tourism demand are very similar across the sample period. In terms of the horizon, the response of the transportation factor shocks to tourism demand shows that it has a positive effect in the first month and then tilts downtrend after the initial summit. Furthermore, the response presents a downward tendency, then reaches its lowest point in the third month, after which it begins to rally and gradually tends to zero. (The average response value of tourism demand to the transportation factor shock in the third month is −0.036.) This finding indicates that the impulse responses of the transportation factor shocks to tourism demand will gradually diminish in the future, implying that more attention paid to transportation might not have a strong impact on tourism demand in the long term.

Impulse responses to an attraction factor shock. The responses to the attraction shock are presented in the Figure 1 (c), where they are shown to be immediately positive in terms of the horizon. Nevertheless, as is clearly shown by the figure, the impact response falls over time. Starting from the sixth month, there is a stable weaker negative impact. (The average response value of tourism demand to the attraction factor shock in the sixth month is −0.001.) In terms of period, the figure shows a similar tendency in regard to impulse response functions across the entire period. Thus, the results imply that positive shocks to the attractions factor enhance tourism demand in the short run and the effects of the shocks will become negative in the long run.

Impulse responses to a shopping factor shock. In terms of the horizon, the result plotted in Figure 1 (d) shows that a standard deviation shock in the shopping factor causes tourism demand to increase slightly. It peaks in the second month only to converge back to zero in the long run. (The average response value of tourism demand to the shopping factor shock in the second month is 0.023.) In terms of the period, the tendency of impulse response functions in the sample is a little bit different. The upward impact of the shopping factor on tourist volume is lowest around 2011 and the lowest point of the 3D impulse responses is seen in February 2011 in the sixth period. (The response value of tourism demand to the shopping factor shock at that point is −0.0008.)

Impulse responses to a tours factor shock. In terms of the horizon, an upward tour shock has positive effects on tourism demand in the short term over the whole period (most of them are less than three months). Specifically, the result of an impulse of the tours factor on tourism demand is initially positive, with a definitive negative trend that persists up to the 12th period. The negative impact did not improve significantly until 2018 in terms of period. (The average response value of tourism demand to the tours factor shock in 2011 is −0.0078, and in 2018 it is −0.0001) Incidentally, this result may be related to the opening of the Hong Kong–Zhuhai–Macao Bridge, which has caused more tourists to choose self-driving tours. Therefore, compared with the past, more tourists who search for tours will take action and come to Hong Kong.

Impulse responses to a lodging factor shock. In terms of horizon, the impact of lodging on tourism demand is initially slightly negative and then starts to fluctuate, after which it has a value close to 0. In terms of period, the range of fluctuation gradually decreases. And there has been no positive influence on tourism demand from 2015 to now, only a negative influence.

The effects of six factors on tourism demand need further empirical investigation. We wish to compare and analyze the different amplitudes of the impulse responses when tourism demand shows different trends. In this article, the Markov regime-switching model with three regimes is used to divide the regimes of the year-to-year growth of tourist volume. The estimation results indicate that the means of tourism demand growth are different in three regimes. As shown in Supplement Figure 3 (available in the online supplement), we divide the change in tourism demand into three regimes using the shadow: 2011M01-2015M03 is the first regime, which indicates that tourism demand is in the stage of rapid growth; 2015M04-2018M02 is the second regime, which shows that tourism demand is in the stage of slow growth; and 2018M03-2019M03 is the third regime, which denotes that tourism demand recovers to a stage of more rapid. Figure 2 depicts the average of the impact of the six influencing factors in each regime.

Figure 2

Mean Values of Impulse Response of Factors Affecting Tourism Demand in Different Periods

From Figure 2, it can be seen that the effects of dini, attr, and shop on tourism demand are positive in each regime. In addition, the shock impacts of tran, tour, and lodg on tourism demand have a negative effect. Among them, dini has always been the main factor driving tourism demand growth. In fact, Hong Kong has a reputation as “the capital of Asian cuisine.” In addition to its traditional local cuisine, Hong Kong also combines Japanese, Korean, Southeast Asian, Indian, European, and American cuisine, since influenced by Western culture. Therefore, culinary travel is one of the key purposes for tourists choosing Hong Kong (Nok et al., 2017; Suntikul et al., 2019). In contrast, tran has the greatest negative impact on tourism demand. When making travel plans, using keywords related to transportation such as “route,” “map,” and “fare,” tourists may find that their route is not suitable or that the fare is too expensive, in which case, it is likely that their travel plans will be canceled. Therefore, tran is the least peasant factor of travel planning.

Considering the three regimes, the dini trend is consistent with the tourism demand. In the second regime, the positive impact of dini on tourism demand weakens. Similarly, there was no significant increase in tourism demand in this regime. However, with the positive impact of dini strengthening, tourism demand also appears to be following an increasing trend in the third regime. Thus, this also confirms that dini is the main factor driving growth in regard to tourism demand. The positive influence of shop has also been increasing. Most things in Hong Kong are taxfree and so are much cheaper than on the mainland, especially items such as cosmetics, clothes, and watches, so Hong Kong has become something of a shopping paradise for mainland Chinese tourists (Qiu et al., 2019). Therefore, traveling for shopping is a trend. In addition, the negative impacts of tour have gradually reduced, but the negative impact of lodg is increasing. Similar to tran, these two factors belong to elements of the planning category and thus have a negative influence. As more and more tourists post online about their own travel strategies, subsequent potential tourists will be able to access more strategies that fit their own travel intentions when searching for a tour; and thus, they will be more willing to travel. However, the high cost of accommodation has become a barrier to tourists visiting Hong Kong.

Conclusion

Limitations

The main limitations of this article are as follows: (a) Although the Baidu Index is the largest search engine in China, it does not include all possible search information. Therefore, subsequent research will attempt to more accurately reflect the influencing mechanism of tourism demand on data across more search engine. (b) This article only analyzes the influencing factors of tourism demand from China to Hong Kong—the next step will be to broaden the applicability of the model to other tourism contexts. (c) Recently political is unrest in Hong Kong, which largely lowers the number of tourist arrivals in general. Our sample is not included in this period, so that has not been affected by this incident. Even if the sample length gets longer, the effects of negative event can also be solved by introducing dummy variables. Which means the qualitative factors about the political unrest can be transformed into dummy variables, since the qualitative factor cannot be directly measured.

Concluding Summary

Based on the monthly data regarding tourist volume from mainland China to Hong Kong and the Baidu Index from January 2011 to March 2019, this article takes Hong Kong as an example and performs an impulse response analysis with the TVP-Fa-VAR model to explore the relationship between tourist volume and the tourist attention represented by Baidu keywords. The interrelationships found lead to the following main conclusions.

First, the results show that, on average, dining, attractions, and shopping present positive effects on tourism demand in each regime, and the shock impacts of transportation, tours, and lodging on tourism demand have negative effects. Therefore, we divide these six factors into stimulative factors and baffling factors, based on both these results and our own experience. The dining, attractions, and shopping factors belong to the stimulative category because, when visitors search for these factors, the information they get is likely to be pictures of delicious food and interesting places, which prompt a sense of anticipation in potential visitors. Among them, dining is the most important stimulative factor. However, transportation, tours, and lodging belong to the baffling category. Tourists perhaps get information that is inconsistent with their plans when searching with keywords relating to these factors, such as routes and fares. The direct flight to Hong Kong is expensive, especially for mainland tourists who are not from the adjacent province, and transfers from other places take much longer. Therefore, high-cost air tickets or inconvenient routes will increase the money and time costs for tourists from mainland China seeking to travel to Hong Kong, and will ultimately affect tourist volume. So, tourists are more likely to abandon their travel plans. Transportation is the most baffling factor.

Second, according to the 3D impulse response analysis, we conclude that the short-term effects between the six factors and tourist volume are different, and the long-term relationship tends to be stable, which is reflected in the following aspects: (a) the dining, tours, and lodging factors’ trend regarding shock response changes most intensely with time in the whole sample; (b) the impact of dining and shopping on tourism demand shows a positive impact in the early stage, but eventually becomes insignificant; (c) the response of transportation and attraction shocks to tourism demand shows that they have positive effects in the early month, after which, negative influences occur, and the final effect will gradually stabilize; (d) the impulse result of tours and lodging on tourism demand gradually stabilizes in positive and negative influence fluctuations.

In conclusion, the empirical results provide important policy implications for Hong Kong, which can be utilized by tourism investors and local governments when making development decisions. First, our findings can help tourism services priorities to be adjusted in a timely manner. During the whole sample period, the focus of potential tourists’ attention on Hong Kong is not completely consistent. This is reflected by the Baidu keywords used. For example, dining has a significant positive impact on tourism demand in the short term. Managers can adjust their own focus on services by providing the relevant information. The effect of transportation shock, however, is negative in the long run. This reflects that tourists are not satisfied with some aspects of transportation. This requires management to make timely adjustments. Second, investors can strengthen tourism marketing. Furthermore, they can improve marketing according to the interest factors of tourists, catering to the needs of tourism, and strive to attract more tourists. For instance, we can release shopping maps within a month after the number of searches regarding the shopping factor increases. This will make it easier and more memorable for tourists who visit Hong Kong for shopping, and those tourists will make up a larger proportion of repeat visitors. Third, it is convenient for local governments to perfect tourism information. Tourism investors can improve the relevant tourism information through the contribution of various factors to the changes in tourist volume, enrich the content of websites, and enhance the accuracy and accessibility of information. For example, with the increase in search volume regarding the lodging factor, managers should promptly release relevant tourist information, tourist volumes and other information regarding accommodation, and inform tourists of the latest situation regarding scenic spots.

Supplemental Material

Supplemental_material – Supplemental material for A Study On The Influencing Factors Of Tourism Demand From Mainland China To Hong Kong

Supplemental material, Supplemental_material for A Study On The Influencing Factors Of Tourism Demand From Mainland China To Hong Kong by Han Liu, Wei Liu and Yonglian Wang in Journal of Hospitality & Tourism Research

Footnotes

Authors’ Note:

The authors would like to acknowledge the Natural Science Foundation of China (Grant No. NSFC71673233), the Humanities and Social Sciences Youth Foundation of Ministry of Education of China (Grant No. 20YJC79007), the 2017 National Statistical Science Research Project (Grant No. 2017LD01), the Scientific Research Project of Jilin Provincial Department of Education (Grant No. JJKH20190730SK), and the Fundamental Research Funds for the Central Universities for the financial support of the study.

ORCID iDs

Han Liu

Wei Liu

Yonglian Wang

Supplemental Material

Supplemental material for this article is available online.

Han Liu (e-mail: hanliu@jlu.edu.cn) is associate professor in the Center for Quantitative Economics and Business School, Jilin University, Changchun, Jilin, China. Wei Liu (e-mail: liuwwweispace@sina.com) is a graduate student in the Business School, Jilin University, Changchun, Jilin, China. Yonglian Wang (corresponding author; e-mail: sophia_ylwang@sina.com) is assistant professor in the School of Statistics, Jilin University of Finance and Economics, Changchun, Jilin, China.

References

Ahn

M. J.

McKercher

(2015). The effect of cultural distance on tourism: A study of international visitors to Hong Kong. Asia Pacific Journal of Tourism Research, 20(1), 94-113. https://doi.org/10.1080/10941665.2013.866586

Altissimo

Cristadoro

Forni

Lippi

Veronese

(2010). New Eurocoin: Tracking economic growth in real time. Review of Economics and Statistics, 92(4), 1024-1034. https://doi.org/10.1162/REST_a_00045

Aratuo

D. N.

Etienne

X. L.

(2019). Industry level analysis of tourism-economic growth in the United States. Tourism Management, 70(February), 333-340. https://doi.org/10.1016/j.tourman.2018.09.004

Armeanu

Andrei

J. V.

Lache

Panait

(2017). A multifactor approach to forecasting Romanian gross domestic product (GDP) in the short run. PLOS ONE, 12(7), Article e0181379. https://doi.org/10.1371/journal.pone.0181379

Asero

Tomaselli

(2015). Research note: Analysing tourism demand in tourist districts: The case of Sicily. Tourism Economics, 21(5), 1111-1119. https://doi.org/10.5367/te.2014.0392

Bangwayo-Skeete

P. F.

Skeete

R. W.

(2015). Can Google data improve the forecasting performance of tourist arrivals? Mixed-data sampling approach. Tourism Management, 46(February), 454-464. https://doi.org/10.1016/j.tourman.2014.07.014

Baram-Tsabari

Segev

(2011). Exploring new web-based tools to identify public interest in science. Public Understanding of Science, 20(1), 130-143. https://doi.org/10.1177/0963662509346496

Barigozzi

Hallin

(2017). Generalized dynamic factor models and volatilities: Estimation and forecasting. Journal of Econometrics, 201(2), 307-321. https://doi.org/10.1016/j.jeconom.2017.08.010

Bokelmann

Lessmann

(2019). Spurious patterns in Google Trends data: An analysis of the effects on tourism demand forecasting in Germany. Tourism Management, 75(December), 1-12. https://doi.org/10.1016/j.tourman.2019.04.015

10.

Cabrera-Sánchez

J. P.

Villarejo-Ramos

Á. F.

(2020). Acceptance and use of big data techniques in services companies. Journal of Retailing and Consumer Services, 52(January), Article 101888. https://doi.org/10.1016/j.jretconser.2019.101888

11.

Camacho

Pacce

M. J.

(2018). Forecasting travellers in Spain with Google’s search volume indices. Tourism Economics, 24(4), 434-448. https://doi.org/10.1177/1354816617737227

12.

Chang

Y. T.

Gau

Y. F.

Hsu

C. C.

(2017). Liquidity commonality in foreign exchange markets during the global financial crisis and the sovereign debt crisis: Effects of macroeconomic and quantitative easing announcements. North American Journal of Economics and Finance, 42(November), 172-192. https://doi.org/10.1016/j.najef.2017.06.004

13.

Chen

J. L.

D. C.

Shen

(2019). Forecasting seasonal tourism demand using a multiseries structural time series method. Journal of Travel Research, 58(1), 92-103. https://doi.org/10.1177/0047287517737191

14.

Chen

Schuckert

Song

Chon

(2014). Why can package tours hurt tourists? Evidence from China’s tourism demand in Hong Kong. Journal of Travel Research, 55(4), 427-439. https://doi.org/10.1177/0047287515612597

15.

Choi

Varian

(2012). Predicting the present with Google Trends. Economic Record, 88(Suppl. 1), 2-9. https://doi.org/10.1111/j.1475-4932.2012.00809.x

16.

Choi

(2017). Variability in the effects of uncertainty shocks: New stylized facts from OECD countries. Journal of Macroeconomics, 53(September), 1339-1351. https://doi.org/10.1016/j.jmacro.2017.06.006

17.

Cosma

R. M.

(2017). The analyse of tourism demand in the Dobrogea region. Quality: Access to Success, 18(Suppl. 2), 141-144.

18.

Din

B. H.

Habibullah

M. S.

Tan

S. H.

(2017). The effects of world heritage sites and governance on tourist arrivals: Worldwide evidence. International Journal of Economics and Management, 11(2), 437-448. https://pdfs.semanticscholar.org/0e78/617d22a82f4e34b4bbd493c8f4c2ab8edc41.pdf

19.

Divisekera

(2009). Ex post demand for Australian tourism goods and services. Tourism Economics, 15(1), 153-180. https://doi.org/10.5367/000000009787536735

20.

Divisekera

(2010). Economics of leisure and non-leisure tourist demand: A study of domestic demand for Australian tourism. Tourism Economics, 16(1), 117-136. https://doi.org/10.5367/000000010790872132

21.

Divisekera

(2016). Interdependencies of demand for international air transportation and international tourism. Tourism Economics, 22(6), 1191-1206. https://doi.org/10.1177/1354816616669007

22.

Fesenmaier

D. R.

Xiang

Pan

Law

(2011). A framework of search engine use for travel planning. Journal of Travel Research, 50(6), 587-601. https://doi.org/10.1177/0047287510385466

23.

Forni

Hallin

Lippi

Reichlin

(2000). The generalized dynamic-factor model: Identification and estimation. Review of Economics and Statistics, 82(4), 540-554. https://doi.org/10.1162/003465300559037

24.

Garcia

M. J.

Sanchez

A. R. P.

Chico

J. R.

(2015). Factores incidentes en la recuperación turística en la actual crisis en las diferentes regiones europeas, un análisis estático [Incidental factors on tourism recovery in different European tourist regions, an approach]. Revista De Estudios Regionales, 104, 33-57.

25.

Ginsberg

Mohebbi

M. H.

Patel

R. S.

Brammer

Smolinski

M. S.

Brilliant

(2009). Detecting influenza epidemics using search engine query data. Nature, 457(7232), 1012-1014. https://doi.org/10.1038/nature07634

26.

Gunadhi

Boey

C. K.

(1986). Demand elasticities of tourism in Singapore. Tourism Management, 7(4), 239-253. https://doi.org/10.1016/0261-5177(86)90034-8

27.

Habibi

(2017). The determinants of inbound tourism to Malaysia: A panel data analysis. Current Issues in Tourism, 20(9), 909-930. https://doi.org/10.1080/13683500.2016.1145630

28.

Hallin

Liška

(2007). Determining the number of factors in the general dynamic factor model. Journal of the American Statistical Association, 102(478), 603-617. https://doi.org/10.1198/016214506000001275

29.

Heagney

E. C.

Rose

J. M.

Ardeshiri

Kovac

(2018). Optimising recreation services from protected areas: Understanding the role of natural values, built infrastructure and contextual factors. Ecosystem Services, 31(Part C), 358-370. https://doi.org/10.1016/j.ecoser.2017.10.007

30.

Hong Kong Tourism Board. (2019). Visitor arrival statistics: Dec 2018. https://partnernet.hktb.com/filemanager/intranet/pm/VisitorArrivalStatistics/ViS_Stat_C/ViS_C_2018/Tourism%20Statistics%2012%202018_R1.pdf

31.

Kim

Shephard

Chib

(1998). Stochastic volatility: Likelihood inference and comparison with ARCH models. Review of Economic Studies, 65(3), 361-393. https://doi.org/10.1111/1467-937X.00050

32.

Kulendran

(1996). Modelling quarterly tourist flows to Australia using cointegration analysis. Tourism Economics, 2(3), 203-222. https://doi.org/10.1177/135481669600200301

33.

Song

Cao

D. C.

(2013). How competitive is Hong Kong against its competitors? An econometric study. Tourism Management, 36(June), 247-256. https://doi.org/10.1016/j.tourman.2012.11.019

34.

Song

Witt

S. F.

(2005). Recent developments in econometric modeling and forecasting. Journal of Travel Research, 44(1), 82-99. https://doi.org/10.1177/0047287505276594

35.

S. W.

Chen

Wang

Ming

C. H.

(2018). Effective tourist volume forecasting supported by PCA and improved BPNN using Baidu index. Tourism Management, 68(October), 116-126. https://doi.org/10.1016/j.tourman.2018.03.006

36.

Pan

Law

Huang

(2017). Forecasting tourism demand with composite search index. Tourism Management, 59(April), 57-66. https://doi.org/10.1016/j.tourman.2016.07.005

37.

Shang

Wang

(2015). A MIDAS modelling framework for Chinese inflation index forecast incorporating Google search data. Electronic Commerce Research and Applications, 14(2), 112-125. https://doi.org/10.1016/j.elerap.2015.01.001

38.

Lin

J. L.

(1996). Co-integration constraint and forecasting: An empirical examination. Journal of Applied Econometrics, 11(5), 519-538. https://doi.org/10.1002/(SICI)1099-1255(199609)11:5%3C519::AID-JAE410%3E3.0.CO;2-Q

39.

McLaren

Shanbhogue

(2011). Using Internet search data as economic indicators. Bank of England Quarterly Bulletin, 51(2), 134-140. https://doi.org/10.2139/ssrn.1865276

40.

Nakajima

Kasuya

Watanabe

(2011). Bayesian analysis of time-varying parameter vector autoregressive model for the Japanese economy and monetary policy. Journal of the Japanese and International Economies, 25(3), 225-245. https://doi.org/10.1016/j.jjie.2011.07.004

41.

Narayan

P. K.

(2004). Fiji’s tourism demand: The ARDL approach to cointegration. Tourism Economics, 10(2), 193-206. https://doi.org/10.5367/000000004323142425

42.

Nikolaos Askitas

K. F. Z

. (2009). Google econometrics and unemployment forecasting. Applied Economics Quarterly, 55(2), 107-120. https://doi.org/10.3790/aeq.55.2.107

43.

Nilashi

bin Ibrahim

Ithnin

Sarmin

N. H.

(2015). A multi-criteria collaborative filtering recommender system for the tourism domain using Expectation Maximization (EM) and PCA-ANFIS. Electronic Commerce Research and Applications, 14(6), 542-562. https://doi.org/10.1016/j.elerap.2015.08.004

44.

Nok

L. C.

Suntikul

Agyeiwaah

Tolkach

(2017). Backpackers in Hong Kong–motivations, preferences and contribution to sustainable tourism. Journal of Travel and Tourism Marketing, 34(8), 1058-1070.

45.

Pan

D. C.

Song

(2012). Forecasting hotel room demand using search engine data. Journal of Hospitality and Tourism Technology, 3(3), 196-210. https://doi.org/10.1108/17579881211264486

46.

Pérez Gálvez

J. C.

Torres-Naranjo

Lopez-Guzman

Carvache Franco

. (2017). Tourism demand of a WHS destination: An analysis from the viewpoint of gastronomy. International Journal of Tourism Cities, 3(1), 1-16. https://doi.org/10.1108/IJTC-10-2016-0039

47.

Polgreen

P. M.

Chen

Pennock

D. M.

Nelson

F. D.

(2008). Using Internet searches for influenza surveillance. Clinical Infectious Diseases, 47(11), 1443-1448. https://doi.org/10.1086/593098

48.

Primiceri

G. E.

(2005). Time varying structural vector autoregressions and monetary policy. Review of Economic Studies, 72(3), 821-852. https://doi.org/10.1111/j.1467-937X.2005.00353.x

49.

Qiu

Fan

D. X. F.

Lyu

Lin

P. M. C.

Jenkins

C. L.

(2019). Analyzing the economic sustainability of tourism development: Evidence from Hong Kong. Journal of Hospitality & Tourism Research, 43(2), 226-248. https://doi.org/10.1177/1096348018777046

50.

Reintinger

Berghammer

Schmude

(2016). Simulating changes in tourism demand: A case study of two German regions. Tourism Geographies, 18(3), 233-257. https://doi.org/10.1080/14616688.2016.1169312

51.

Saayman

(2008). Determinants of inbound tourism to South Africa. Tourism Economics, 14(1), 81-96. https://doi.org/10.5367/000000008783554893

52.

Song

(2008). Tourism demand modelling and forecasting-A review of recent research. Tourism Management, 29(2), 203-220. https://doi.org/10.1016/j.tourman.2007.07.016

53.

Sun

Wei

Tsui

K. L.

Wang

(2019). Forecasting tourist arrivals with machine learning and internet search index. Tourism Management, 70(February), 1-10. https://doi.org/10.1016/j.tourman.2018.07.010

54.

Suntikul

Pratt

Chong

Y. W. J.

(2019). Factors that influence Chinese outbound tourists’ intention to consume local food. Journal of China Tourism Research, 16(2), 230-247. https://doi.org/10.1080/19388160.2019.1625841

55.

Triacca

Focker

(2014). Estimating overnight volatility of asset returns by using the generalized dynamic factor model approach. Decisions in Economics and Finance, 37(2), 235-254. https://doi.org/10.1007/s10203-012-0130-x

56.

Vaughan

Romero-Frías

(2014). Web search volume as a predictor of academic fame: An exploration of Google trends. Journal of the Association for Information Science and Technology, 65(4), 707-720. https://doi.org/10.1002/asi.23016

57.

Volchek

Liu

Song

Buhalis

(2018). Forecasting tourist arrivals at attractions: Search engine empowered methodologies. Tourism Economics, 25(3), 425-447. https://doi.org/10.1177/1354816618811558

58.

Vosen

Schmidt

(2011). Forecasting private consumption: Survey-based indicators vs. Google trends. Journal of Forecasting, 30(6), 565-578. https://doi.org/10.1002/for.1213

59.

Wen

Liu

Song

(2019). Forecasting tourism demand using search query data: A hybrid modelling approach. Tourism Economics, 25(3), 309-329. https://doi.org/10.1177/1354816618768317

60.

Wondirad

Agyeiwaah

(2016). What matters to Hong Kong’s inbound market? Modeling Hong Kong’s tourism demand. Tourism Planning and Development, 13(4), 396-408. https://doi.org/10.1080/21568316.2015.1136837

61.

D. C.

Song

Shen

(2017). New developments in tourism and hotel demand modeling and forecasting. International Journal of Contemporary Hospitality Management, 29(1), 507-529. https://doi.org/10.1108/IJCHM-05-2015-0249

62.

Brynjolfsson

(2009, December). The future of prediction: How Google searches foreshadow housing prices and sales [Paper presentation]. 30th International Conference on Information Systems, ICIS 2009, Phoenix, AZ. https://doi.org/10.2139/ssrn.2022293

63.

Yang

Pan

Evans

J. A.

(2015, August 30). Forecasting Chinese tourist volume with search engine data. Tourism Management, 46, 386-397. https://doi.org/10.2139/ssrn.2022293

64.

Zhang

Singh

A. J.

(2016). Does it matter? Examining the impact of China’s vacation policies on domestic tourism demand. Journal of Hospitality & Tourism Research, 40(5), 527-556. https://doi.org/10.1177/1096348013503993

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.55 MB