High-speed railway and tourism development in China

Abstract

High-speed railway (HSR) is a new and increasingly popular transportation mode in China bringing about a significant impact on the economy, including tourism development. This article investigates the effect of HSR on tourism development in China based on a time-varying difference-in-differences model. Cities connected by HSR in 2013 and 2014 are regarded as the treatment group, while those without HSR services until 2017 are placed in the control group. The empirical analyses cover a large panel dataset comprising 163 cities in 2009–2017. The empirical results suggest that both domestic tourism revenue and tourist number are positively affected by HSR, and the effect is stronger for the undeveloped or geopolitically less important regions such as the inland or prefecture-level cities. Other relevant determinants of tourism include the availability of airports and the number of hotels in the cities. Our research findings have important policy implications for tourism development in China with respect to HSR.

Keywords

high-speed railways tourism development China difference-in-differences model

Introduction

In line with the increase in per capita income and the associated improvement in people’s living standards, people start to pay more attention to their leisure time planning. More and more people choose to travel as their leisure activities, leading to rapid and continuous development in the tourism industries. According to the United Nations World Tourism Organization (UNWTO), international tourist arrivals for the whole world increased from 952 million passengers in 2010 to 1407 million passengers in 2018, with international tourism receipts increasing from US$901 billion to US$1462 billion in the same period, representing an average annual growth rate of 5.31% and 6.92%, respectively. An increasing trend is also observed in China, especially for domestic tourism which is expanding more speedily. According to China’s Ministry of Culture and Tourism, the number of domestic tourists and tourism expenditure reached 5.54 billion passengers and 5.13 trillion RMB, respectively, in 2018, representing an annual average growth rate of 12.70% and 20.22%, respectively, during 2010–2018. Such growth was not only higher than the world average but also higher than that recorded in 2001–2010 when the average annual growth rates of the number of tourists and tourism revenues were, respectively, 11.15% and 15.40%. Fast expansion of the Chinese tourism industry has been coincided with the remarkable development of the country’s high-speed railway (HSR) services starting from August 1, 2008. This triggers us to inquire where and how HSR development may help to stimulate the steady expansion of China’s tourism industry over the sample period.

To further illustrate the rapid development of the tourism industry with HSR in China, we take Leshan and Meishan cities in Sichuan province which were connected by HSR in 2014 as two examples, the former is listed as one of the excellent tourist cities in China proposed by the country’s tourism authority, while the latter is not a famous tourist destination. The average number of domestic tourists in the pre-HSR period of 2009–2013 in Leshan and Meishan were 21.63 million and 12.9 million, respectively. In the post-HSR period of 2014–2017, the average number of tourists, respectively, rose to 41.77 million and 33.78 million. It demonstrates that HSR not only brings about more tourists to the famous tourism destinations but also contributes to tourism development in the less known localities in China.

Given its nonnecessity nature, the demand for tourism services is sensitive to the income level of tourists, which is regarded as an exogenous factor. As for the supply-side factors, accessibility is one of the three most important factors, even more important than tourism attraction (Albalate et al., 2017). Accessibility is determined by the availability and level of development of modern transportation infrastructures, such as HSR, motorways, and air transport. They work together to provide convenient, safe, and affordable transportation services to tourists to get to and return from tourist destinations. It is not surprising to find locations with better accessibility to be able to attract more tourists, with no exceptions in Spain and China, which, respectively, have the largest HSR network in Europe and in the World.

The first HSR line with a top speed over 250 km/h between Beijing and Tianjin was opened in August 2008 in China. Since then, HSR development has accelerated speedily in the following decade. By the end of 2019, 35,000 km of HSR lines were put into operation, accounting for over two-thirds of the global HSR mileage. The number of cities, which are connected by HSR increased continuously from 12 in 2009¹ to 101 in 2014² and 180 in 2019³, covering 32 of the 34 provincial-level administrative cities of the country as shown in Figure 1, with Macau and Tibet as the two exceptions. The total number of HSR passengers also skyrocketed from 46.51 million passenger trips in 2009 to 2.29 billion passenger trips in 2019 (China Daily, January 3, 2020).

Figure 1.

China’s “eight-vertical/eight-horizontal” HSR network in 2019.

A World Bank Report in 2014 suggests that transportation may lead to an induced agglomeration effect of productivity improvement due to market expansion and better matching between producers and consumers. Besides, an employment effect caused by changing business locations is observed. The investment in HSR is a positive stimulus to the economy, leading to demand and output expansion. It helps to lower the transportation costs to bring about an induced demand effect (Chen et al., 2016). HSR’s contribution to urban economic growth is found to be 0.11%, resulting mainly from the local effect (Chong et al., 2019). Ke et al. (2017) find the heterogeneous effects of HSR on the economic growth of targeted city nodes are economically important: ranging between a minimum of 5% increase for Zhenjiang on the Hu-Ning Segment and a maximum of 59% for Qingyuan on the Wu-Guang HSR. In addition, HSR contributes to regional convergence as it serves as a driver to sustainable and steady economic expansion of the Chinese regions (Yao et al., 2019, 2020).

From the tourism perspectives, the HSR-tourism nexus in China is firstly investigated in Chen and Haynes (2012, 2015) with respect to the effect on international tourism. Research evidence has shown that HSR has a significant and positive effect on both international tourism revenue and international tourist arrivals. As for the structural impact, the HSR network has enlarged and transformed the tourism market space in China. It intensifies competition and redistributes urban tourism centers (Wang et al., 2012). Connection with coastal cities to the hinterlands, HSR promotes international tourist arrivals in China (Li et al., 2019), with a more significant boost to tourist arrivals in the less-developed central and western regions than the more-developed eastern areas (Gao et al., 2019). Despite the evidence revealed on the contributions of HSR to tourism, the current findings are still far from complete when the asymmetric impact of HSR, which refers to the more favorable influence on tourism development in the less-developed localities connected by HSR, has not been comprehensively examined. To fulfill this gap, we apply three heterogeneity tests based on geographical locations and resource endowments to further study the asymmetric effect of HSR on tourism. First, our sample is divided into the eastern, central, and western cities to test the regional heterogeneity of the HSR effect. Second, a dummy variable distinguishing inland cities from coastal cities is used to test the heterogeneous effect at different economic development levels. Third, the differences between regional central cities and prefecture-level cities are used to test the heterogeneity effect based on different administrative levels.

Using data from 163 Chinese cities in 2009–2017, this article contributes to the literature through investigating the relationship between HSR and tourism development in China. A number of variables are considered to control for accessibility and the tourism supply-side factors in the designated areas to study their impact on domestic tourism in China. Deviated from the classic difference-in-differences (DID) model, this research utilizes the time-varying DID (Albalate et al., 2017; Beck et al., 2010) to deal with the differences in HSR opening years across different cities. Cities with HSR services in 2013 and 2014 are regarded as the treatment group in this study. This is because the number of HSR connected cities in the early stage of HSR development (2009–2012) was too small. There were only 21 cities connected with HSR by the end of 2011. The number of HSR connected cities rose significantly from 2012 onward, but most of the domestic tourism data for HSR connected cities in 2012 were not available. Thus, cities connected with HSR before 2012 were not included in our research due to data constraint and small sample. Moreover, to test the long-term effect of HSR as well as to facilitate the test of a common trend, cities which were connected with HSR after 2014, were not included in our sample either. This data treatment may have excluded some cities in our study, but the size of the sample is large and representative enough for the empirical analysis to have generated meaningful and robust results. Furthermore, Bertrand et al. (2004) argue that the estimation results of many DID studies are biased as the dependent variables are serially correlated. To properly resolve this problem, the autoregressive (AR) correction method is applied in our robustness tests. Apart from the aggregate analyses, regressions at the disaggregate levels by geographical location and city tier are also conducted to examine the asymmetric effect of HSR on tourism development.

The rest of the article is organized as follows. The second section provides a literature review on the HSR-tourism nexus. The third section introduces the empirical model, defines the main variables, and discusses data sources. The regression results of the time-varying DID model are summarized in the fourth section, whereas the robustness checks are shown in the fifth section. Finally, the conclusions with discussion are presented in the sixth section.

Literature review

Factors affecting tourism

Factors affecting tourism demand can be grouped into destination-specific factors or the supply-side factors and the origin-specific factors or the demand-side factors. In particular, the supply-side factors cover some monetary indicators such as domestic price and exchange rate, some transportation-related factors such as the distance to the main market, transportation capital stock, the number of airports, the length of roads or highways, the availability of HSR, and the number of HSR stations, together with tourism facilities and attractions including the number of hotel rooms, tourism infrastructure, the number of scenic spots, and the mega events organized. As for the demand-side factors, they normally refer to the income and price level of the origin, the price level of alternative destinations, as well as some other factors such as population, “word of mouth” or goodwill, trade/investment linkages, common language, cultural traditions, and the like. A number of researches have attempted to employ these supply-side and demand-side factors to estimate tourism demand in different countries and regions, including Australia, Europe, Hong Kong (China), the Philippines, South Korea, and international tourism as a whole (Culiue, 2014; Deluna and Jeon, 2014; Han et al., 2006; Kang, 2016; Shafiullah et al., 2019; Song et al., 2010).

In these studies, the role of the destination factors such as domestic price level (Culiue, 2014; Song et al., 2010), the number of migrants (Shafiullah et al., 2019), real exchange rate (Deluna and Jeon, 2014; Shafiullah et al., 2019; Song et al., 2010), and the distance to the main market (Culiue, 2014; Deluna and Jeon, 2014) have been addressed. In addition, the contributions of the origin factors including per capita gross domestic product (GDP) or income level (Deluna and Jeon, 2014; Kang, 2016), population of the origin (Deluna and Jeon, 2014; Kang, 2016), price level of the origin (Deluna and Jeon, 2014; Han et al., 2006; Kang, 2016; Song et al., 2010), world income level and transportation price (Shafiullah et al., 2019), the price level of alternative destinations (Shafiullah et al., 2019; Song et al., 2010), and tourism demand in the previous period or the word-of-mouth effect (Kang, 2016; Song et al., 2010) have also been discussed. It is found that the income levels of both origins and destinations, relative price, trade linkage, the distance between the origin and the destination, real exchange rate, and word of mouth are significant tourism determinants.

As for tourism demand in China, Tang et al. (2007) observe a unidirectional causality of foreign direct investment (FDI) led growth in international tourism arrivals. Similar conclusion is drawn in Sun et al. (2008), who consider openness as an important determinant of international tourism in China. Apart from openness, the other popular economic determinants of tourism demand in China include the income levels of the origin and the destination, relative prices in the origin and the destination, exchange rate, and the prices of various tourism services such as hotel rates and travel costs as specified in Li (2009). In Kang (2016), it observes that the word-of-mouth effect together with income, trade, and relative price are the main determinants for Chinese tourism demand to South Korea. In a more recent study, Xu et al. (2019) reveal that international tourism arrivals in China are somewhat income elastic but price inelastic. Trade and FDI are found to be significant factors affecting international tourism arrivals, signifying a complementary relationship between goods/capital flows and tourism flows.

As for domestic tourism in China, Cai et al. (2001) conclude that income is a significant factor affecting domestic demand, with an estimated income elasticity of tourism demand at 0.3. In Yang et al. (2014), the role of relative income is reviewed. It indicates that domestic tourism in China is dominated by the absolute income level with diversified impacts across provinces, while relative income is also found to be an important determinant.

Accessibility and tourism

Accessibility does not only allow tourists to arrive to their destinations but also provide them with transportation services at their destinations. As indicated in Page and Connell (2014) and Le-Klähn and Michael (2015), accessibility factors serve as important determinants in tourism development. The accessibility factors can be further divided into external connection and ground transportation. In Seetanah (2005), the role of transport accessibility at the destination, comprising air, land, and water transport is addressed for their contributions to tourism development in Mauritius. The estimation results suggest that aggregate transport capital stock contributes positively in attracting tourists to visit the country and the importance of transport accessibility is proven. At the disaggregate level, Khadaroo and Seetanah (2007) have attempted to estimate the role of transportation factors, namely paved roads (ground transportation), the number of terminals in international airports, and the number of ports (external connection) on tourism demand. The sum of the capital stock for air, land, and sea transport is composed to reflect the level of development of transportation infrastructure, which is a significant factor to determine the tourist arrivals of Mauritius. The length of highway is employed in Massidda and Etzo (2012) to assess its role of ground transportation in determining domestic tourism in Italy, and a significant and direct but secondary role is found. On the top of HSR development, Guirao et al. (2016) have considered the role of ordinary railway to the development of tourism in Spain. The estimation results reveal the insignificance of this ground transportation factor. In Khoshnevis Yazdi and Khanalizadeh (2017), the external connection of airport infrastructure is emphasized as a nonmonetary factor in determining tourism demand. It confirms that airport infrastructure is a significant factor, with a lower magnitude than per capita income, price, and real exchange rate in explaining tourism demand in the United States.

The effect of HSR on tourism

As for the contributions of HSR, Wang et al. (2012) have indicated that HSR has a significant impact on regional tourism transport accessibility. It reduces significantly the temporal and spatial distance between the origin and the destination. Gao et al. (2019) regard HSR as one of the modern transportation facilities that can improve the accessibility of tourism destinations in remote areas and an important transportation infrastructure for passenger movements. In spite of its contribution to improve accessibility, HSR is also expected to deliver positive and significant impacts on tourism, with supporting evidences provided in Masson and Petiot (2009), Kuriharaa and Wu (2016), Pagliara et al. (2017), Görçün (2018), and Campa et al. (2018).

In countries with well-developed HSR networks, such as Italy and Spain, the HSR-tourism nexus has already been thoughtfully discussed. For example, Masson and Petiot (2009) study the HSR between Perpignan (France) and Barcelona (Spain). They indicate the presence of agglomeration and dispersion forces in tourism development. The agglomeration force tends to bring about a favorable result to Barcelona at the expenses of Perpignan. Kuriharaa and Wu (2016) investigate the contributions of extended HSR services to tourism development in Japan. A significant increase in tourists is found in cities, which are newly connected by the extended HSR network. In Pagliara et al. (2017), the effect of HSR on tourism in Italy is reviewed, with strong evidence showing a potent effect of HSR on both the number of domestic visitors and the number of nights spent in the destinations. Görçün (2018) analyses the effect of HSR on regional and urban tourism development in Turkey and a positive and significant contribution is revealed.

Albalate and Fageda (2016) employ the DID panel data analyses to examine the effect of HSR services on tourism development in Spain. When HSR is more competitive in terms of frequency, travel time, and comfort, it posts a detrimental impact on air transport. Nevertheless, HSR has failed to promote tourism in the newly connected areas. Similar hypothesis is also raised in Albalate et al. (2017). The DID estimation results show that the newly constructed HSR corridors have limited effects on tourist number and the average length of stay at several end-line and intermediate cities in Spain. In Delaplace et al. (2016), a case study is conducted to assess the impact of HSR on two tourism destinations in France with significantly diverged results. The logistic regression shows that HSR is important to the choice of one destination (Disneyland). Its contribution to the tourist number for another destination (Futuroscope), however, is negligible. It is concluded that the connection between HSR and tourism development varies across destinations, while HSR does not always contribute to the tourism market. In contrast, Guirao et al. (2016) suggest a direct linkage between HSR and the development of Spanish tourism outputs in its gravity model analyses. So far mixed results are observed for the impacts of HSR on tourism in Spain, which has the longest HSR network in Europe and the second longest in the World.

As for China, which has the longest HSR network in the World, Chen and Haynes (2012) is one of the first studies to investigate the tourism impacts of China’s HSR. The gravity model and panel data analyses have confirmed the contribution of HSR to international tourism. Provinces connected by HSR can receive more tourists and tourism revenues. It also projects that the actual impact of HSR on tourism should be higher upon the completion of the network. Employing the HSR dummy, the number of HSR stations, and HSR density, Chen and Haynes (2015) apply the dynamic estimation technique of the generalized method of moments (GMM) to assess once again the impact of China’s HSR on international tourism. The constructive role of HSR consistent with Chen and Haynes (2012) is revealed and the explanatory power of the number of HSR stations is weaker than the HSR network. Yan et al. (2014) employ the AR and moving average model to examine the impact of the Wuhan–Guangzhou HSR on domestic tourism receipts for the three connected provinces. It exhibits that Guangdong and Hunan have benefited from HSR, with limited benefits enjoyed by Hubei due to the small number of HSR stations in the province. The unevenly distributed impact of HSR caused by provincial heterogeneity is revealed once again by this research. Li et al. (2019) apply the GMM method to conduct city-level analyses on tourism in China. The authors conclude that HSR connection has generated an increase in tourism flows with a stronger effect for international arrivals than domestic ones. Hou (2019) employs the generalized DID model to assess the impact of HSR on tourism development in China. Significant evidence is found to support the contributions of HSR connection to both domestic and international tourism in China.

Besides, national cores and hinterlands connected by HSR are found to receive more benefits than the regional cores. Based on the DID method and the city-level panel data in eastern and middle China over 2006–2013, Zhou and Li (2018) demonstrate a positive contribution of HSR to tourism revenue. Cities with lower overall accessibility tend to benefit more from HSR, while those with relatively lower level of tourism development can also derive more benefits from HSR. Applying the DID method, Gao et al. (2019) fail to find any benefits from HSR on tourism revenue, whereas positive effects are observed on tourism arrivals. In addition, an instrumental variable of potential HSR connection is developed to offer robustness checks. The article concludes that the less-developed central and western regions can benefit more than their more-developed eastern counterparts from HSR. In Yang et al. (2019), the gravity model is employed to compare the contributions of different transportation means, namely air, train, and HSR, and different modes of services (direct or intermodal) to the intercity tourist movements in China. Furthermore, the attractiveness of 5A scenic spots, national parks, and world heritage sites in China is also assessed. The estimation results show that air transport is a significant factor in long travel distances, while the effect of HSR is significant only for medium travel distances. In contrast, the traditional train has the strongest impact on intercity tourist movements given its well-developed network. The availability of 5A scenic spots, national parks, and world heritage sites in the destination are found to attract more incoming tourists.

So far, a series of studies have intended to estimate the impact of HSR on tourism in China across different provinces, regions, and cities for both international and domestic tourism. Evidences are observed to confirm the impact of HSR as a whole. Nevertheless, the heterogeneous HSR effect across connected regions or cities remains ambiguous and mixed, requiring further investigation and clarification, which is a key focus of this article.

Our work

This article contributes to the literature using a novel way to define HSR⁴ and a large panel dataset from 2009 to 2017 to cover the operational period of HSR. First, only services with a designed speed of 250 km/h or higher, that is the Inter-City Rail Service or G-Series High-Speed Train is defined as HSR. Besides, the HSR services in different cities do not open at the same time, a short operation history of HSR services may influence the efficiency of the network. Therefore, years 2013 and 2014 are selected as the cutoff point and for those without HSR connection until 2017 are defined as cities without HSR. In light of the endogeneity concern of HSR, as indicated in Shen et al. (2014), the location decision of HSR is based on land-planning criteria. The government plays a dominant role in determining the HSR route with little connection with the local tourism industry which will not cause the endogeneity problem. Given that, HSR is regarded as an exogenous variable in our article.

Although the gravity model is a popular technique in analyzing tourism demand (Chen and Haynes, 2012; Deluna and Jeon, 2014; Khadaroo and Seetanah, 2008; Morle et al., 2014; Xu et al., 2019; Yang et al., 2019), it is unable to capture the impact of HSR as a treatment on tourism demand when the opening time of the service is different across cities. In contrast, the DID model has been widely employed in many empirical studies (Albalate and Fageda, 2016; Albalate et al., 2017; Chen and Haynes, 2015; Gao et al., 2019; Zhou and Li, 2018) which is an accessible and well-defined model in estimating the effects of HSR. As the opening time of different HSR routes varies rather than using the standard DID technique, the time-varying DID model adopted in Beck et al. (2010) and Albalate and Fageda (2016) is employed in this article and the common trend test is also performed, which can help to overcome the bias problem faced by the standard DID analyses.

Empirical strategy

Methodology

The DID model is applied to address the HSR-tourism nexus of China. Owing to the fact that HSR connection times vary from city to city, the time-varying DID framework (Albalate and Fageda, 2016; Beck et al., 2010) is utilized in this study which is different from the assumption in the classic DID model with only one treatment time. The estimation is embedded in a two-way fixed-effect panel data model with the following specification:

ln t o u r_{i t} = β_{0} + β_{1} H S R_{i t} + β_{2} X_{i t} + η_{i} + λ_{t} + ε_{i t}

where tour represents the outcome of tourism which includes domestic tourist number and domestic tourism revenue, while HSR is the dummy variable for HSR connection, and $X_{i t}$ is a vector of control variables, which can help to explain the dependent variables. Subscripts i and t denote cities and years, respectively, $η_{i}$ and $λ_{t}$ represent the city fixed effect and time fixed effect, respectively. $ε_{i t}$ is the random disturbance term. Since HSR_it is a time-varying DID variable, it is equal to one in the years when city i is connected by HSR and zero otherwise. It is common in the literature to use a dummy variable to indicate if a city or region is connected by HSR, which can be found in Chen and Haynes (2012, 2015), Yan et al. (2014), Zhou and Li (2018), Gao et al. (2019), Li et al. (2019), and Hou (2019). β₁ in the model is our research interest which measures the effect of HSR connection on tourism outcomes. If it is significantly positive, then HSR-connected cities could enjoy stronger growth in tourism than those not connected by HSR, ceteris paribus.

Equation (1) should fulfill the common trend assumption given the time-varying feature of the treatment time, which means that the treatment group and the control group share the same trend in the years before HSR connection. The testing method proposed in Beck et al. (2010) and Gao et al. (2019) is then employed with the following specification:

\begin{array}{l} ln t o u r_{i t} = β_{0} + γ_{t - k} H S R_{i, t - k} + γ_{t - k + 1} H S R_{i, t - k + 1} + \dots + γ_{t - 1} H S R_{i, t - 1} + γ_{t + 1} H S R_{i, t + 1} \\ + \dots + γ_{t + m - 1} H S R_{i, t + m - 1} + γ_{t + m} H S R_{i, t + m} + β_{2} X_{i t} + η_{i} + λ_{t} + ε_{i t} \end{array}

where t is the year when city i is connected by HSR and the HSR dummy basically follows the previous definition with the following exceptions: HSR_i,t−k equals one for cities in the $k^{th}$ year before HSR connection, while HSR_i,t+m equals one for cities in the $m^{th}$ year after HSR connection, with k > 0 and m > 0. HSR_i,t is not included in the test to avoid the multicollinearity problem. The estimated coefficients $γ_{t - k}$ and $γ_{t + m}$ measure the effect of HSR on tourism outcomes. Thus, if $γ_{t - k}$ is statistically insignificant while $γ_{t + m}$ is statistically significant, it reveals that the common trend assumption cannot be rejected and HSR will boost tourism outcomes.

Additionally, considering the asymmetric effects of HSR on tourism development among different city classifications, we also introduce a set of dummy variables and their interaction terms with HSR _i,t to the baseline model to capture the variation in the effect of HSR on tourism development across different regions and cities. And the regression model is shown in the following equation:

ln t o u r_{i t} = β_{0} + β_{1} H S R_{i t} + β_{2} H S R_{i t} \times D_{i} + β_{3} X_{i t} + η_{i} + λ_{t} + ε_{i t}

where D_i is the dummy variable, which represents the category of city i. It takes the value of 1 if city i belongs to the category, otherwise 0. Thus, $(β_{1} + β_{2})$ measures the effect of HSR on tourism outcomes in cities when D_i = 1, while $β_{1}$ measures the effect of HSR connection on tourism outcomes of cities, which do not belong to the category, ceteris paribus.

Explanatory variables

The key research interest in this article is the role of HSR represented by the HSR dummy. The reason why we use a dummy variable to represent HSR owes to the formality of the DID model where a dummy variable takes the value of 1 after an exogenous shock and 0 otherwise. More accurate information about the operation of HSR can also be applied in the model, such as service frequencies, travel times, and ticket prices. However, due to a rather short period of time, the lack of information has restricted our ability to use any better alternative than a dummy variable to measure the effect of HSR on tourism development in this study.

Different from the existing studies which employ designed speed and operation speed as the benchmarks, we regard the operating line of Inter-City Rail Services or G-Series High-Speed Train as HSR. According to the National Railway Administration of China, the traditional HSR definition which requires a designed speed of 250 km/h and above and an initial operating speed of no less than 200 km/h consists of two kinds of railway lines, including the reconstruction of existing lines and the new lines. However, based on the construction standards from China Railway, the original line and the new passenger dedicated line carry completely different technical specifications and standards, which contrast markedly on both required technology and construction cost, with the latter one implying much higher investment and technology intensity. Thus, their implications should differ substantially and should be discussed separately. While some of the existing studies such as Chen and Haynes (2012, 2015) and Zhou and Li (2018) define the D-Series train which runs on the existing railway lines to a large extent as HSR, only the Inter-City Rail Services or G-Series High-Speed Train which operates on the newly built passenger dedicated line is regarded as HSR in this article.

Apart from the HSR dummy, other control variables are also considered to be potential determinants of tourism. As a popular and important explanatory variable, GDP per capita and population at the city level are employed which are widely used in the current literatures (Chen and Haynes, 2012, 2015; Gao et al., 2019; Li et al., 2018; Yang et al., 2014; Zhou and Li, 2018). Cities with higher GDP per capita and more people tend to have higher living standards and larger market size and thus would be equipped with better infrastructure and services to attract incoming tourists. For the other economic factors, such as public expenditure in Li et al. (2019) and Hu (2019), it is highly correlated with the income factor and is ignored in our model to avoid the multicollinearity problem.

The number of star-rated hotel⁵ is also considered as a control variable. Similar research design can be observed in Chen and Haynes (2012) and Zhou and Li (2018). Cities with more star-rated hotels have more capacity to serve tourists and thus may be able to receive more tourists and tourism revenue.

In addition, transport infrastructures other than HSR are also considered as explanatory variables. All of them strengthen the accessibility of a city and thus improve the attractiveness to tourists. Similar to Li et al. (2018), Zhou and Li (2018), Gao et al. (2019), and Li et al. (2019), both ground transportation of road and external connection factors of airport transport are considered in our model. In particular, the road length per unit area (km/km²) is employed to reflect the abundance of road infrastructure. Compared to highway length or passenger road ridership which is used in previous studies (Gao et al., 2019; Massidda and Etzo, 2012; Zhou and Li, 2018), road length per unit area is a density variable which can reflect more accurately the accessibility of road transport. To capture the impact of air transport, an airport dummy is introduced which is 1 for cities with an airport and scheduled flights and 0 otherwise. Similar setup can also be found in Li et al. (2018), Zhou and Li (2018), and Gao et al. (2019).

Finally, the supply-side factors provided in the destinations are important determinants to tourism development. These factors include the natural and man-made landscapes and heritage; in particular, the 4A or 5A scenic spots and UNSECO World Heritage Sites (Chen and Haynes, 2012; Gao et al., 2019; Yang et al., 2019; Zhou and Li, 2018 ) and special events offered by the destination (Li et al., 2019) which have been modelled in the existing studies. To address the contribution of attractions to tourism, a proxy variable is generated to show the number of 5A scenic spots⁶ or World Heritage (WH) Sites available in a city. Given the weak attractiveness of 4A scenic spots and the incomplete statistics, they are not considered in our estimation. Special events, such as the Beijing Olympic Games (Li et al., 2019), are not considered in our analyses since they are nonregular and may be correlated with the opening time of the HSR. An example is the Beijing Olympic Games which was organized in 2008 and the Beijing–Tianjin HSR route was opened in the year to support it.

Data

As China operated its first HSR line in the second half of 2008, the sampling period of this study starts from 2009 to 2017 to cover the operational period of HSR. HSR connection is extracted from National Railway Passenger Train Timetable of China. It equals to one when there is an Inter-City Rail Service or G-Series High-Speed Train station in the city. As explained previously, only cities which were HSR connected in 2013 and 2014 or had never been connected by HSR until 2017 are included in our data sample. Those cities which were newly connected by HSR after 2015 or had HSR stations before 2012 are excluded from this study. The domestic tourist number and domestic tourism revenue are extracted from multiple sources, including China Statistical Yearbook for Regional Economy 2010–2014 and China City Statistical Yearbook 2018. The city’s tourism data in 2014–2016 are missing in the national statistical yearbooks; we use the statistical yearbooks of the relevant provinces to supplement the missing data. However, not all statistical yearbooks of relevant provinces provide the number of domestic tourists’ number and domestic tourism revenue, which is another reason why we excluded the cities which were connected by HSR after 2015 or before 2012. As a result, our sample covers 163 cities over a 9-year period. The per capita GDP and annual average population statistics at the city level are collected from China City Statistical Yearbook 2010–2018. All the nominal monetary values have been converted to real values by the Consumer Price Indices and the GDP deflator to neutralize the impact of inflation with 2009 as the base year.

In light of the tourism supply-side factors, a number of 5A scenic spots⁷ and WH Sites⁸ are collected from the official websites. The star-rated hotel number is available in the Yearbook of China Tourism Statistics 2010–2018 at the provincial level. For the road length per unit area, provincial-level statistics from China Statistical Yearbook 2010–2018 are adopted as the proxies given that prefectural city-level statistics are not available. All of the Yearbooks are collected from China National Knowledge Infrastructure (https://www-cnki-net-443.web.bisu.edu.cn/). Finally, the airport data are collected from the Nationwide Airport Production Statistics Bulletin 2009–2017 by the Civil Aviation Administration of China.⁹ All the continuous variables above including domestic tourism revenue and tourist number are in their natural logarithms when running the regression. The summary statistics of the relevant variables are presented in Table 1.

Table 1.

Summary statistics of the relevant variables over the data period.

Variables	Description	Obs.	Mean	SD	Min.	Max.
TR	Domestic tourism revenue (100 million RMB yuan)	1337	134.486	180.727	0.542	3005.920
TN	Domestic tourist number (10,000 persons)	1299	1854.106	1799.799	5.200	15900
HSR	Dummy variable equals to 1 if city is HSR available, otherwise 0	1467	0.113	0.317	0	1
GDP	City level GDP per capita (RMB yuan)	1441	49346.780	50874.040	4600.880	428614
Pop	The annual average population (10,000 persons)	1463	366.512	227.578	19.900	1258
Hotel	The number of hotels (1000 hotels)	1456	0.436	0.228	0.057	1.088
Road	Road length per unit area (km/km²)	1467	0.818	0.413	0.061	1.702
Airport	Dummy variable equals to 1 if city has an airport, otherwise 0	1467	0.434	0.496	0	1
Aspot	Number of 5A scenic spots	1467	0.317	0.600	0	4
WH	Number of World Heritage Sites	1467	0.130	0.336	0	1

Note: SD, standard deviation; HSR: high-speed railway; GDP: gross domestic product. And the units of the variables are shown in the parentheses.

Estimation results

Common trend results

Figures 2 and 3 graph the trends of domestic tourism revenue and domestic tourist number for both the treatment and control groups. It is observed that before the inauguration of HSR, from 2009 to 2014, the treatment and control groups on tourism revenue and number share similar increasing trends expressed as two pairs of roughly parallel lines. The gap between the two trends started to widen after 2014; the starting point of HSR strictly defined in this article.

Figure 2.

Trends of domestic tourism revenues for HSR connected and nonconnected cities. HSR: high-speed railway.

Figure 3.

Trends of domestic tourist numbers for HSR connected and nonconnected cities. HSR: high-speed railway.

To conduct the time-varying DID analyses, the common trend assumption is firstly examined. The test method proposed in Beck et al. (2010) and Gao et al. (2019) with specification stated in Equation (2) is applied. It is conducted by including all the periods before and after HSR connection. At the end points, HSR_i,−4 measures the lead effect which equals to one for cities which will be connected by HSR for 4 years or longer, while HSR_i,+3 measures the lag effect which equals to one for cities which had HSR services for 3 years or longer. The results are reported in Table 2 and Figures 4 and 5 showing the 95% confidence intervals of the HSR dummies with different lags.

Table 2.

The result of common trend test.

Variables	Dependent variables
Variables	Domestic tourism revenues (in logs)	Domestic tourist numbers (in logs)
4+ year before	−0.004 (0.045)	−0.022 (0.032)
3 year before	−0.018 (0.045)	0.006 (0.030)
2 year before	0.012 (0.043)	0.044 (0.034)
1 year before	0.018 (0.041)	0.064 (0.039)
1 year after	0.008 (0.092)	0.098*** (0.029)
2 year after	0.043 (0.106)	0.129*** (0.037)
3+ year after	0.199** (0.077)	0.164*** (0.060)
Constant	−0.012 (1.318)	2.529* (1.392)
No. of observations	1299	1259
No. of cities	161	161
R ²	0.8230	0.8340

Note: GDP: gross domestic product. Robust standard errors clustered at the city level are in parentheses. Both city and year fixed effects are embedded in two-way fixed-effects panel data model. Variables including domestic tourism revenue, domestic tourist number, GDP per capita, hotel number, and road length per unit area are in natural logarithm.

*, **, *** indicate significance level at 10%, 5%, and 1%, respectively.

Figure 4.

The 95% CI of HSR on domestic tourism revenue. CI: confidence interval; HSR: high-speed railway.

Figure 5.

The 95% CI of HSR on domestic tourist number. CI: confidence interval; HSR: high-speed railway.

The common trend test results show that despite there is a 1-year lead effect of HSR on domestic tourist number at the 1% significance level, yet all the HSR dummies in the preconnection period are insignificant. As a whole, it reveals that there is basically no lead effect for both tourism revenues and numbers. In sharp contrast, HSR connection has brought about positive and statistically significant impacts on both tourism revenues and numbers in the postconnection periods, with a time lag for the impacts on tourism revenues which is only significant three periods after the HSR connection. Meanwhile, a long-lasting shock is observed for tourist numbers with an increasing significance level in the post-HSR connection period. As a whole, the common trend estimation illustrates that the effect of HSR on domestic tourism is remarkable.

Given the test results, there is no violation of the common trend assumption between the treatment and control groups for both tourist numbers and revenues. HSR connection is found to have an impact on both tourist numbers and revenues. Consequently, the time-varying DID technique could be employed to examine the impact of HSR connection on domestic tourism development.

DID estimation results

The estimation results of the time-varying DID based on Equation (1) are presented in Table 3, showing that HSR has a significantly positive effect on both domestic tourism revenues and domestic tourist numbers. The result is consistent with Chen and Haynes (2015), Li et al. (2018), Zhou and Li (2018), and Gao et al. (2019). Given that a city is regarded as HSR connected only after the HSR services is available for a period of time, the estimated coefficient has a stronger magnitude than those in the previous studies. Besides, as a transportation infrastructure which brings people from one place to another, HSR inauguration brings about a more potent impact on tourism development. As presented in Table 3, cities with HSR connection attract, respectively, 9.1% and 10.5% more tourism revenues and tourist arrivals than those without HSR connection.

Table 3.

The results of the time-varying DID estimation.

Variables	Domestic tourism revenues (in logs)	Domestic tourist numbers (in logs)
HSR	0.091* (0.056)	0.105** (0.041)
lnGDP	0.120* (0.069)	0.083 (0.077)
lnPop	0.426** (0.192)	0.528*** (0.200)
lnHotel	0.172** (0.081)	0.134* (0.075)
lnRoad	−0.021 (0.020)	−0.024 (0.015)
Airport	0.209*** (0.073)	0.192** (0.075)
Aspot	−0.024 (0.044)	0.019 (0.032)
Constant	0.078 (1.322)	2.542* (1.391)
City fixed effect	YES	YES
Time fixed effect	YES	YES
No. of observations	1299	1259
No. of cities	161	161
R ²	0.8220	0.8330

Note: DID: difference-in-differences; HSR: high-speed railway; GDP: gross domestic product. Cities connected with HSR in 2013 and 2014 are the treatment group, and cities which had not been connected by HSR until 2017 are the control group. Both city and year fixed effects are embedded in the two-way fixed-effect panel data model. Robust standard errors clustered at the city level are in parentheses. Variables including domestic tourism revenue, domestic tourist number, GDP per capita, hotel number, and road length per unit area are in natural logarithm.

*, **, *** indicate significance level at 10%, 5%, and 1%, respectively.

The city-level GDP per capita is found to have little impact on tourist numbers, but for 1% increase in GDP per capita, the tourism revenue will increase by 0.12% at the 10% significance level. This finding is not dissimilar with Chen and Haynes (2015), implying that the level of economic development of a particular city is not necessarily an important determinant of tourist arrivals in the local areas although higher GDP per capita may raise the price level, leading to more tourism revenue.

Population shows a significantly positive effect on tourism. For 1% increase in population, the tourism revenue and the tourist numbers will increase by 0.426% and 0.528%, which is consistent with the result of the gravity model. Hotels and Airport, however, are found to play an important role in stimulating tourism development, as their estimated coefficients are statistically significant at the 10% level. For 1% increase in the number of hotels, the tourism revenue and the tourist numbers will increase by 0.172% and 0.134%. Cities with airport attract, respectively, 20.9% and 19.2% more tourism revenues and tourist arrivals than those without airport. The significant and positive role played by Airport in tourism development is consistent with the finding of Gao et al. (2019) but different from the findings of Li et al. (2018) and Hu (2019). The highly significant airport dummy observed here can be explained by the strong connection function of air transport which carries passengers from the origin directly to the destination. Additionally, among the 81 cities which had their own airports in our data sample, less than half of them are connected by HSR which may intensify the importance of airports in attracting tourists. In contrast, even HSR services are available, a city, which is not a popular tourism destination, can hardly attract tourists to stopover and the contribution of HSR could be limited, resulting in a less potent contribution of HSR relative to the airport factor.

A number of 5A scenic spots and road density are found to have an insignificant impact on tourism development in our data sample, which is similar to the insignificant effect of 5A scenic spots observed in Gao et al. (2019) but different from the findings of Li et al. (2019) and Yang et al. (2019). The insignificance of these factors is likely attributed to the relatively short sampling period in our study during which both 5A scenic spots and road density are relatively stable and hence weakly correlated with the rapidly expanding tourism industry which is significantly driven by HSR development. Besides, as explained in Gao et al. (2019), the insignificant effect of 5A spots may be due to their high admission ticket prices. It may also suggest that visiting 5A scenic spots is not the main purpose of tourism decision.

Heterogeneity across cities

Given their diversified geographical locations and resource endowments, the impact of HSR on different Chinese cities varies significantly. Three sources of heterogeneity are addressed in this study. The first is regional heterogeneity across the eastern, central, and western areas in China¹⁰. The 163 sampling cities are grouped into three subsamples based on their geographical locations and two dummy variables, namely E_i and M_i, are set in the regional heterogeneity model. E_i takes a value of 1 if a city is in the eastern region and 0 otherwise, and M_i takes a value of 1 if a city is in the central region and 0 otherwise. The western region dummy variable is omitted from the regression for comparison purposes. The specific regression model with regional dummies is given in the following equation:

ln t o u r_{i t} = β_{0} + β_{1} H S R_{i t} + β_{2} H S R_{i t} \times E_{i} + β_{3} H S R_{i t} \times M_{i} + β_{4} X_{i t} + η_{i} + λ_{t} + ε_{i t}

The estimated results are reported in Table 4. Similar with the baseline results presented in Table 3, HSR plays a positive role once all the sample cities are included in the regression. All the three regions take advantage of HSR for tourism revenues as the estimated coefficients of the regional dummies are statistically insignificant. However, the impact of HSR on tourism arrivals varies between regions. Compared to cities located in the east, tourism development benefited more significantly in the cities located in the inland areas (central and western regions). Cities with HSR connection in the central and western areas attract 19.3% more tourist arrivals than those without HSR connection, and cities with HSR connection in the eastern areas attract 1.3% less tourist arrivals than those without HSR connection. This is because the estimated coefficient of the cross-term of the dummy variable for the eastern region with HSR is statistically negative, implying that HSR has a less potent impact on tourism arrivals in the eastern region compared with the western or the central region. As the respective coefficient for the central region is statistically insignificant, it implies that the effect of HSR on tourism arrivals is indifferent between the central and western regions.

Table 4.

Estimated results of HSR connection on tourism outcome by regions.

Variables	Domestic tourism revenues (in logs)	Domestic tourist numbers (in logs)
HSR	0.175** (0.076)	0.193*** (0.066)
HSR×E_i	−0.196 (0.121)	−0.206** (0.082)
HSR×M_i	0.003 (0.105)	0.007 (0.070)
lnGDP	0.098 (0.066)	0.059 (0.074)
lnPop	0.425** (0.194)	0.530*** (0.201)
lnHotel	0.181** (0.082)	0.145* (0.076)
lnRoad	−0.017 (0.021)	−0.020 (0.016)
Airport	0.201*** (0.071)	0.183** (0.074)
Aspot	−0.018 (0.043)	0.023 (0.031)
Constant	0.313 (1.312)	2.796** (1.385)
City fixed effect	YES	YES
Time fixed effect	YES	YES
No. of observations	1299	1259
No. of cities	161	161
R ²	0.8230	0.8350

Note: HSR: high-speed railway; GDP: gross domestic product. Cities connected with HSR in 2013 and 2014 are the treatment group, and cities which had not been connected by HSR until 2017 are the control group. Robust standard errors clustered at the city level are in parentheses. Variables including domestic tourism revenue, domestic tourist number, GDP per capita, hotel number, and road length per unit area are in natural logarithm.

*, **, *** indicate significance level at 10%, 5%, and 1%, respectively.

Both per capita GDP and road length per unit area are insignificant given the presence of HSR and airport factors, irrespective of the geographical regions. It reflects the inelastic demand for tourism, especially when the destination is connected by HSR and the overwhelming contribution of HSR and airport which has surpassed that of road length to tourism development.

To conduct a robustness test on the above results, we further divide the sample cities into two subsamples, one consisting of cities located in the coastal region (27 cities) and the other consisting of cities located in the inland areas (136 cities). In general, the economies in the coastal cities are more prosperous and better developed than their inland counterparts. The different levels of economic development may be associated with a heterogeneous effect of HSR on tourism development in different localities.

Apart from the division of geographic locations, we also test the heterogeneity effect of HSR on tourism development by grouping cities into different tiers based on their administrative status. According to the National Urban System Planning (2005–2020), cities in China are classified into national central cities, regional central cities, and prefecture-level cities based on their significance in finance, transportation, culture, and other aspects¹¹. Based on this classification, the 163 sample cities are divided into regional central and prefecture-level cities given none of them belonging to the category of national central city.

Since both classifications above have two groups of cities, we only need to set one dummy variable in the heterogeneity regressions which follow the below format:

ln t o u r_{i t} = β_{0} + β_{1} H S R_{i t} + β_{2} H S R_{i t} \times C o a s t a l_{i} + β_{3} X_{i t} + η_{i} + λ_{t} + ε_{i t}

ln t o u r_{i t} = β_{0} + β_{1} H S R_{i t} + β_{2} H S R_{i t} \times R e g i o n a l c e n t r a l_{i} + β_{3} X_{i t} + η_{i} + λ_{t} + ε_{i t}

The estimated results are presented in Table 5. The results show that HSR has a significant and positive impact on tourism development measured in both tourist arrivals and revenues because the estimated coefficients of HSR are all significantly positive in all the model specifications. However, the results also indicate that the tourism outcomes of the inland and/or prefecture-level cities are more boosted by HSR compared to their coastal or regional central counterparts. This is implied by the significantly negative coefficients of the cross-terms of coastal cities dummy variable and the regional central cities dummy variable with HSR. For example, cities with HSR connection in the inland areas attract 14.2% more tourist arrivals than those without HSR connection, and cities with HSR connection in the coastal region attract 12% less tourist arrivals than those without HSR connection.

Table 5.

Estimated results of HSR connection on tourism outcomes by city groups.

Variables	Domestic tourism revenues (in logs)		Domestic tourist numbers (in logs)
HSR	0.158*** (0.049)	0.149*** (0.049)	0.142*** (0.041)	0.122*** (0.041)
HSR×Coastal	−0.515** (0.224)		−0.262*** (0.053)
HSR×Regionalcentral		−0.386* (0.210)		−0.134 (0.095)
lnGDP	0.072 (0.068)	0.128* (0.072)	0.054 (0.078)	0.080 (0.079)
lnPop	0.419** (0.187)	0.459** (0.197)	0.526*** (0.197)	0.536*** (0.201)
lnHotel	0.166** (0.079)	0.183** (0.081)	0.274* (0.173)	0.141* (0.074)
lnRoad	−0.015 (0.021)	−0.011 (0.021)	−0.020 (0.015)	−0.020 (0.016)
Airport	0.201*** (0.072)	0.203*** (0.072)	0.183** (0.074)	0.186** (0.074)
Aspot	−0.020 (0.042)	−0.015 (0.042)	0.021 (0.030)	0.022 (0.032)
Constant	3.002*** (0.690)	2.443*** (0.732)	5.863*** (0.796)	5.598*** (0.798)
City fixed effect	YES	YES	YES	YES
Time fixed effect	YES	YES	YES	YES
No. of observations	1303	1303	1262	1262
No. of cities	161	161	161	161
R ²	0.8257	0.8241	0.8322	0.8310

*, **, *** indicate significance level at 10%, 5%, and 1%, respectively.

Robustness checks

Robustness tests are also conducted to cross-check the findings from the baseline analyses and the results are reported in Table 6. First, the municipalities and provincial capitals are removed from our sample and the results are reported in columns (1) and (5). Comparing with the previous findings, the values of the estimated HSR coefficients are larger for both tourism revenues and numbers by approximately 13.9% and 11.7%, respectively, with improved statistical significance. It indicates that HSR contributes more to the peripheral cities than the municipalities and provincial capitals. This finding is consistent with the results from the disaggregate analyses by city groups as presented in Table 5.

Table 6.

Robustness checks.

Variables	Domestic tourism revenues (in logs)				Domestic tourist numbers (in logs)
Variables	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
HSR variables
HSR	0.139*** (0.049)		0.090 (0.056)		0.117*** (0.042)		0.107*** (0.041)
HSR_new		0.117*** (0.041)				0.099*** (0.036)
F_HSR				0.080*** (0.025)				0.108*** (0.025)
Controlled variables
lnGDP	0.120* (0.069)	0.117* (0.065)	0.121* (0.069)	0.286*** (0.023)	0.083 (0.077)	0.079 (0.072)	0.083 (0.077)	0.211*** (0.023)
lnPop	0.473** (0.198)	0.386** (0.183)	0.433** (0.191)	0.578*** (0.028)	0.569*** (0.211)	0.497*** (0.189)	0.523** (0.201)	0.518*** (0.025)
lnHotel	0.177** (0.086)	0.182** (0.081)	0.176** (0.083)	0.153*** (0.028)	0.170** (0.078)	0.145* (0.075)	0.131* (0.075)	0.113*** (0.027)
lnRoad	−0.010 (0.020)	−0.022 (0.020)	−0.020 (0.020)	0.021 (0.017)	−0.022 (0.015)	−0.027* (0.015)	−0.024 (0.015)	0.070*** (0.020)
Airport	0.202*** (0.072)	0.212*** (0.072)	0.209*** (0.072)	0.177*** (0.024)	0.189** (0.074)	0.193*** (0.074)	0.193** (0.075)	0.102*** (0.024)
Aspot	0.008 (0.043)	−0.026 (0.045)		0.107*** (0.016)	0.035 (0.032)	0.018 (0.031)		0.125*** (0.017)
WH			−0.106 (0.177)				0.095 (0.104)
Constant	−0.227 (1.356)	0.340 (1.258)	0.034 (1.315)	−2.501*** (0.317)	2.314 (1.452)	2.762** (1.326)	2.568* (1.401)	1.367*** (0.312)
City fixed effect	YES	YES	YES	YES	YES	YES	YES	YES
Time fixed effect	YES	YES	YES	YES	YES	YES	YES	YES
No. of observations	1246	1299	1299	1299	1212	1259	1259	1259
No. of cities	154	161	161	161	154	161	161	161
$\hat{ρ}$				0.8316				0.8240
R ²	0.8380	0.8230	0.8220		0.8320	0.8330	0.8330

Note: HSR: high-speed railway; GDP: gross domestic product; AR: autoregressive. HSR, HSR_new, F_HSR stand for original HSR definition, HSR measurement using National Development and Reform Commission of China (NDRC) definition, and HSR generalized variable, respectively. Robust standard errors clustered at the city level are in parentheses. represents the estimator of autocorrelation parameter in AR (1) model. Variables including domestic tourism revenue, domestic tourist number, GDP per capita, hotel number, and road length per unit area are in natural logarithm.

*, **, *** indicate significance level at 10%, 5%, and 1%, respectively.

Second, an alternative definition of HSR is adopted and cities, which are connected by D-Series train, are also regarded as HSR connected. A new dummy variable namely HSR_new is introduced and the results are reported in columns (2) and (6) in Table 6. As expected, HSR_new has a stronger magnitude than the original HSR given that the contributions of the D-Series train are also reflected and as a whole, this finding is consistent with those associated with the baseline model.

Third, a more demanding international standard of WH Site is utilized to substitute for the national standard of 5A scenic spot. Nevertheless, the tourism attraction factor remains to be statistically insignificant. Despite the presence of internationally recognized tourism attractions, a city is still unable to attract more tourism revenues or arrivals, indicating that visiting famous tourism attractions may not be the main purpose of tourism for domestic clients.

Finally, Bertrand et al. (2004) argue that the OLS estimations may generate biased results due to the incorrect estimation of standard errors caused by serial correlation. Thus, an AR process with a first-order lag is considered when extra lags are impossible to apply given the short sampling period. The results are shown in columns (4) and (8), which are consistent with those from the baseline model. The impacts of HSR on tourism revenues and numbers are significant and positive and do not decline after one period, reflecting the long-lasting contribution of HSR to tourism.

Conclusion

This article sheds light on the effects of HSR on tourism development with respect to domestic tourism revenues and numbers in China at the city level. HSR refers to the Inter-City Rail Service or G-Series High-Speed Train. Cities are defined as HSR connected if HSR service was available in 2013–2014. Then the domestic tourism revenues and numbers for a total of 163 cities, with 38 and 125 in the treatment and control groups over the period 2009–2017, are analyzed. The time-varying DID model is employed and the empirical results show that HSR contributes to 9.1% more domestic tourism revenues and 10.5% more domestic tourist arrivals to the treatment group compared with the control group. For example, Heze in Shandong Province is well-known for its peony flower and will be connected to the HSR network in 2021. The domestic tourism revenues and tourist arrivals of Heze are 20.8 billion yuan and 23.65 million passengers, respectively, in 2019 based on official statistics. The local government have constant targets of GDP and population growth. Therefore, without HSR, the domestic tourism revenues and tourist arrivals are growing 2.5 billion yuan and 2 million passengers a year for the time trend. Whereas, considering our empirical results, the event of HSR operating in Heza is expected to increase the tourism revenues and tourist arrivals to 28.14 billion yuan and 30.55 million passengers, respectively.

The effect of HSR varies between different groups of cities. It is found that cities located in the inland areas benefited more for tourism development induced by HSR compared to their coastal counterparts. In addition, prefecture-level cities are found to be more boosted by HSR for tourism development compared to the regional central cities. Similar heterogeneous effects of HSR on tourism development across different types of cities are also found in other empirical studies for China. Take Sanming in Fujian Province and Lanzhou in Gansu Province as two contrasting examples. The former is located in the eastern region and the latter in the western area. However, after the opening of HSR, the number of domestic tourist arrivals in Sanming even dropped to 27.49 million, which was less than the level in 2013 when the city was connected with HSR. In contrast, the number of Lanzhou’s domestic tourist arrivals rose from 25.03 million in 2012 to 65.87 million in 2017.

Apart from the positively significant role of HSR, our empirical analyses reveal that per capita GDP and the availability of famous tourism attractions in destinations play a limited role in boosting tourism development in our sample. The finding that tourist numbers are not significantly affected by per capita GDP and the number of 5A scenic spots is somewhat surprising but it may due to the strong effect of HSR in diverting or attracting tourists in favor of the relatively less-developed regions in China where the lack of accessibility had prohibited tourism development before the HSR era. However, it is found that tourism revenues are significantly affected by per capita GDP, population, and the price level of tourism destinations. In other words, domestic tourists do not view living standards and infrastructures as well as visiting attractions as important factors in making their travel decision. In contrast, accessibility including HSR and airport facilities are significant factors with a potent impact on tourism outcomes. Cities with HSR and airport are found to have attracted 10.5% and 19.2%, respectively, more tourist arrivals than those without HSR and airport. The other supply-side factors such as the number of star-rated hotels also exhibit a significant effect on tourism revenues and tourist numbers.

Compared to many existing studies, our research not only adopts a more restrictive and appropriate classification of HSR but also defines the connection years in a more stringent way to take into account the scale effect of the available HSR network. Our refined setup yields findings consistent with the previous work, demonstrating that HSR connection can boost tourism revenues and tourist arrivals. However, our article is still suffering from the imperfection of an over-simplified HSR definition in the form of a dummy variable, given that service frequencies, travel times, and ticket prices are not reflected in the model. This is, however, due to a rather short period of time, and the lack of information has restricted our ability to obtain more useful data relating to HSR development in China.

Irrespective of this caveat, all the estimated results are robust, consistently supporting the positive and significant role of HSR in tourism development for the sample cities. One clear and important contribution of this article to the literature is that it also studies the heterogeneous effect of HSR on tourism development between the coastal and inland cities as well as between the prefecture-level and regional central cities. The fact that the inland and prefecture-level cities have benefited more for their tourism development as a result of HSR implies that HSR development must have been amicable for reducing China’ regional economic development imbalances, which is of significant importance for the country to sustain its long-term and high-quality economic growth through reducing regional inequality and maximizing the utilization of resources throughout the country. The research finding in this article, therefore, strongly supports China’s effort in developing the country’s HSR network which is connecting the vast majority of the prefecture-level cities, or even some county-level cities, as far as tourism development is concerned.

China is the fastest growing and potentially one of largest tourism markets in the World. It has1.4 billion people with per capita GDP exceeding US$10,000 in 2019 and a middle-class population of some 400 million people. Our empirical study shows that travel accessibility has a potent impact on domestic tourism decision, but the popularity of tourism attractions and the richness of tourism destinations are not so important, suggesting that the demand for leisure and vacation is gradually replacing the demand for sightseeing in the relatively more remote areas thanks to the time-space compression effect of HSR. To attract more tourists, policy makers and tourism authorities in China and the rest of the World should pay more attention on improving the accessibility of tourism destinations, such as connecting cities to the HSR network, opening more direct HSR trains and flights, and providing free visas for Chinese tourists in the international tourism markets and the like.

Footnotes

Acknowledgements

We gratefully thank anonymous referees for their valuable comments but take full responsibility for any error or omission herein.

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 the following financial support for the research, authorship, and/or publication of this article: the National Natural Science Foundation of China (71673033, 71973019), the National Social Science Foundation of China (18ZDA005, 19ZDA082), Chongqing Social Science Planning Project (2017YBJJ024), and Fundamental Research Funds for the Central Universities (2020CDJSK02PT26, 2019CDSKXYJG0037).

ORCID iD

Feng Wang

Notes

References

Albalate

Campos

Jiménez

(2017) Tourism and high speed rail in Spain: does the AVE increase local visitors? Annals of Tourism Research 65: 71–82.

Albalate

Fageda

(2016) High speed rail and tourism: empirical evidence from Spain. Transportation Research Part A: Policy and Practice 85: 174–185.

Beck

Levine

Levkov

(2010) Big bad banks? The winners and losers from bank deregulation in the United States. Journal of Finance 65(5): 1637–1667.

Bertrand

Duflo

Mullainathan

(2004) How much should we trust differences-in-differences estimates? The Quarterly Journal of Economics 119(1): 249–275.

Cai

Feng

(2001) Domestic tourism demand in China’s urban centers: empirical analyses and marketing implications. Journal of Vacation Marketing 8(1): 64–78.

Campa

Arce

López-Lambas

(2018) Can HSR improve the mobility of international tourists visiting Spain? Territorial evidence derived from the Spanish experience. Journal of Transport Geography 73: 94–107.

Chen

Haynes

(2012) Tourism industry and high speed rail – Is there a linkage? Evidence from China’s high speed rail development. George Mason University School of Public Policy, Research Paper: 2012-14.

Chen

Haynes

(2015) Impact of high-speed rail on international tourism demand in China. Applied Economics Letters 22: 57–60.

Chen

Xue

Rose

, et al. (2016) The impact of high-speed rail investment on economic and environmental change in China: a dynamic CGE analysis. Transportation Research Part A 92: 232–245.

10.

Chong

Qian

Chen

(2019) Estimating the economic benefits of high-speed rail in China: a new perspective from the connectivity improvement. The Journal of Transport and Land Use 12(1): 287–302.

11.

Culiuc

(2014). Determinants of international tourism. International Monetary Fund Working Papers:WP/14/82.

12.

Delaplace

Pagliara

Pietra

(2016) Does high-speed rail affect destination choice for tourism purpose? Disneyland Paris and Futuroscope Case Studies, Belgeo: 3.

13.

Deluna

Jeon

(2014) Determinants of international tourism demand for the Philippines: an augmented gravity model approach. Munich Personal RePEc Archive Paper: No. 55294.

14.

Gao

Wang

(2019) Does high-speed rail boost tourism growth? New evidence from China. Tourism Management 72: 220–231.

15.

Görçün

(2018) Impacts of high speed trains on tourism development: a case study of Ankara-Konya high speed rail lines. Journal of Tourism Theory and Research 4(1): 21–28.

16.

Guirao

Campa

López-Lambas

(2016) The assessment of the HSR impacts on Spanish tourism: an approach based on multivariate panel data analysis. Transportation Research Procedia 18: 197–204.

17.

Han

Durbarry

Sinclair

(2006) Modelling US tourism demand for European destinations. Tourism Management 27: 1–10.

18.

Hou

(2019) High-speed railway and city tourism in China: a quasi-experimental study on HSR operation. Sustainability 11: 1512.

19.

Kang

(2016) Determinants of Chinese tourism demand to South Korea. International Journal of Tourism and Hospitality Research 30(4): 69–83.

20.

Chen

Hong

, et al. (2017) Do China’s high-speed-rail projects promote local economy? — New evidence from a panel data approach. China Economic Review 44: 203–226.

21.

Khadaroo

Seetanah

(2007) Transport infrastructure and tourism development. Annals of Tourism Research 34(4): 1021–1032.

22.

Khadaroo

Seetanah

(2008) The role of transport infrastructure in international tourism development: a gravity model approach. Tourism Management 29: 831–840.

23.

Khoshnevis Yazdi

Khanalizadeh

(2017) Tourism demand: a panel data approach. Current Issues in Tourism 20(8): 787–800.

24.

Kuriharaa

(2016) The impact of high speed rail on tourism development: a case study of Japan. The Open Transportation Journal 10: 35–44.

25.

Le-Klähn

D-T

Michael

(2015) Tourist use of public transport at destinations – a review. Current Issues in Tourism 18(8): 785–803.

26.

(2009) Tourism demand modelling and forecasting: a review of literature related to Greater China. Journal of China Tourism Research 5: 2–40.

27.

Yang

Cui

(2019) High-speed rail and tourism in China: an urban agglomeration perspective. International Journal of Tourism Research 21: 45–60.

28.

Massidda

Etzo

(2012) The determinants of Italian domestic tourism: a panel data analysis. Tourism Management 33: 603–610.

29.

Masson

Petiot

(2009) Can the high speed rail reinforce tourism attractiveness? The case of the high speed rail between Perpignan (France) and Barcelona (Spain). Technovation 29: 0–617.

30.

Morley

Rosselló

Santana-Gallego

(2014) Gravity models for tourism demand: theory and use. Annals of Tourism Research 48: 1–10.

31.

Page

Connell

(2014) Transport and tourism. In: Alan

Michael

Allan

(eds) The Wiley Blackwell Companion to Tourism. Hoboken, NY: John Wiley & Sons, pp. 155–167.

32.

Pagliara

Mauriello

Garofalo

(2017) Exploring the interdependences between high speed Rail systems and tourism: some evidence from Italy. Transportation Research Part A Policy & Practice 106: 300–308.

33.

Seetanah

(2005) Transport capital as a determinant of tourism development: a time series approach. Tourismos: An International Multidisciplinary Journal of Tourism 1(1): 55–73.

34.

Shafiullah

Okafor

Khalid

(2019) Determinants of international tourism demand: evidence from Australiam states and territories. Tourism Economics 25(2): 274–296.

35.

Shen

Silva

Martínez

(2014) HSR station location choice and its local land use impacts on small cities: a case study of Aveiro, Portugal, Procedia. Social and Behavioral Sciences 111: 470–479.

36.

Song

Witt

, et al. (2010) Tourism demand modelling and forecasting: how should demand be measured? Tourism Economics 16(1): 63–81.

37.

Sun

Han

(2008) A study on relationship between foreign openness degree and inbound tourism development in China. International Journal of Business and Management 3(1): 62–69.

38.

Tang

Selvanathan

(2007) The relationship between foreign direct investment and tourism: empirical evidence from China. Tourism Economics 13(1): 25–39.

39.

The World Bank (2014) Regional Economic Impact Analysis of High Speed Rail in China. China and Mongolia Sustainable Development Sector Unit, East Asia and Pacific Region, 25 June 2014.

40.

Wang

Huang

Zou

, et al. (2012) Effects of the high speed rail network on China’s regional tourism development. Tourism Management Perspectives 1: 34–38.

41.

Wang

, et al. (2019) Modelling international tourism flows to China: a panel data analysis with gravity model. Tourism Economics 25(7): 1047–1069.

42.

Yan

Zhang

(2014) Assessing the impacts of the high-speed train on tourism demand in China. Tourism Economics 20(1): 157–169.

43.

Yang

(2019) Public transport connectivity and intercity tourists flows. Journal of Travel Research 58(1): 25–41.

44.

Yang

Liu

(2014) Domestic tourism demand of urban and rural residents in China: does relative income matter? Tourism Management 40: 193–202.

45.

Yao

Fang

(2020) Can time-space compression promote urban economic growth? Evidence from China’s high-speed rail projects. China & World Economy 28(5): 90–117.

46.

Yao

Zhang

Wang

, et al. (2019) Regional economic growth and the role of high-speed rail in China. Applied Economics 51: 3465–3479.

47.

Zhou

(2018) The impact of high-speed trains on regional tourism economies: empirical evidence from China. Tourism Economics 24(2): 187–203.