Tourist Attractions and Economic Growth in China: A Difference-in-Differences Analysis

Abstract

The construction of tourist attractions has become an important manifestation of local performance and the image of tourist destinations, as well as an important means for local governments to promote economic development. However, the causal relationship between tourist attractions and economic growth remains unclear. The article’s main purpose is to explore the causal effect of tourist attractions on economic growth. To do so, a difference-in-differences model is employed based on China’s city-level panel data from 2001 to 2019 involving 313 cities and 5947 observations. The results demonstrate that tourist attractions have a significant positive causal effect on China’s economic growth. Such causality is significant only in the east and central regions. Highway density, urban disposable income per capita, and the share of the tertiary industry have significant moderating effects. The validity of the causal relationship is confirmed using various rigorous robustness tests.

1. Introduction

There is a strong link between both tourism and the Sustainable Development Goals [1]; therefore, the development of tourism is considered to be a major means of promoting sustainable development in the region. Consequently, among the many means of promoting economic growth, tourism is considered to be very efficient and effective [2,3]. During the development of tourism, tourist attractions constitute the core component of a destination’s attractiveness; therefore, it is arguably the tourist attractions that play a key role in economic growth to a large extent [4,5]. Thus, many tourist destinations are keen to regard the construction of tourist attractions, especially top tourist attractions, as a boost to their economic growth. However, the existing findings are limited to the correlation. In other words, the causal impact of tourist attractions on economic growth remains unknown. The existing correlations likely result from data correlation caused by coincidences, confounding factors, or reverse causality. In contrast, the current economics perspective places more emphasis on causal relationships, which contributes to a deeper understanding of economic phenomena and related decisions [6,7]. Causal inference based on experimental ideas has become the dominant paradigm for exploring economic phenomena. Taking this into account, this paper attempts to explore the causal relationship between tourist attractions and economic growth using a quasi-natural experimental method and the underlying mechanisms, thus deepening the knowledge of the link between tourism and sustainable development.

In China’s tourism administration, the quality rating and management of tourist attractions, especially top tourist attractions such as 5A National Tourist Attraction, has always been a central task. In China, the presence or absence of 5A National Tourist Attraction (5ANTA) has become an important indicator to measure local tourism development and the attractiveness of tourist destinations [8]. Therefore, the construction of the 5ANTA has become an important manifestation of local performance and the image of tourist destinations, as well as an important means for local governments to promote tourism development. In 2007, China launched the first selection of 5ANTA, with 66 selected that year including the most famous Palace Museum (Beijing), Badaling Great Wall (Beijing), Terracotta Army (Xi’an), the West Lake (Hangzhou), Lijiang River (Guilin), etc. By the end of 2019, the number had increased to 277, and the number of cities distributed also increased from 55 to 178. Whether the creation of top tourist attractions has contributed to local economic development is a very interesting question of both academic and theoretical significance.

The practice of 5ANTA in China provides a perfect quasi-natural experimental sample: a clear experimental group, i.e., areas with 5ANTA, and a control group, and also, two time periods, “pre” and “post”. Citing China as the case, this study addresses three theoretical questions to fill the existing gap: (1) the causal relationship between tourist attractions, (2) how tourist attractions promote economic development, and (3) the regional heterogeneity of such causality. To this end, the difference-in-differences (DID) model is employed in this study. This approach treats policy implementation as a ‘natural experiment’ exogenous to the economic system and analyses the impact of policy shocks on the economic system by comparing pre- and post-policy changes in the experimental and control groups. Thanks to its simplicity and rigorous economic arguments, the DID method has been widely used since its inception. The DID technique has also been applied in tourism research such as Zhang and Zhang [7] and Zhang and Zhang [9] who assessed the tourism impact of policies.

The potential scientific contributions of the article contain the following three facets. Firstly, the article explores for the first time the causal relationship between tourist attractions and economic development, which deepens the understanding of the link between the two in addition to the correlation. The theories of tourism motivation, economic accounting, and economic growth permit us to conceptualize such causality. This exploration of causality also provides new theoretical evidence for the role of tourism in enhancing regional sustainable development. Secondly, in terms of methodology, a time-varying DID model with a series of rigorous robustness tests is used to support the objectives of the article. More importantly, the article is based on city-level panel data rather than the provincial panel data used by Zhang and Zhang [7], which helps to identify more fine-grained evidence of tourism attractions’ causal effects on economic growth and is more in line with the large data size required by the DID model. Thirdly, this paper further analyzes the influence mechanism and regional heterogeneity of such a causal relationship. This improves the knowledge of the causal relationship between tourist attractions and economic growth and facilitates decision making.

The remainder of this paper is organized as follows: the Section 2 discusses the relevant literature; Section 3 introduces the DID model, variables selection, and data sources; Section 4 reports the empirical results and the last section provides the conclusions and discussion.

2. Literature Review

The relationship between tourism and economic growth is a traditional and topical agenda in tourism economics [10,11]. A growing body of literature focuses on this relationship and confirms the influential role of tourism in economic growth in different tourist destinations, such as Akadiri et al. [12], Gao et al. [13], Croes et al. [14], Dibeh et al. [15], Enilov and Wang [16], Jayaraman and Makun [17], Liu et al. [18], Paramati et al. [19], Pratt [20], Seetanah [21], Senbeto [22], and Sun and Drakeman [23].

Table 1. Summary of relevant literature on tourism and economic growth.

Author(s) (Year)	Independent Variable	Dependent Variable	Sample
Adedoyin et al. [25]	Tourist arrivals	GDP per capita	51 countries
Akadiri et al. [12]	International inbound tourists	GDP per capita	16 selected tourism island territories
Balsalobre-Lorente et al. [11]	Air transport	GDP per capita	Spain
Belgodere et al. [26]	Visitor arrivals per capita	GDP per capita	137 developing countries
Benkraiem et al. [27]	Tourism development index	GDP per capita	Top 10 tourist destinations
Bilen et al. [28]	International tourism receipts	GDP	12 Mediterranean countries
Brida et al. [29]	International tourist arrivals per inhabitant	GDP per capita growth	80 countries
Chiu and Yeh [30]	International tourism receipts growth rate	GDP per capita	84 countries
Comerio and Pacicco [31]	Tourist arrivals	GDP	Japanese prefectures
Danish and Wang [32]	Tourism receipts	GDP per Capita	BRICS economies
De Vita and Kyaw [33]	Tourism arrivals as a percentage of population	GDP per capita	62 countries
Dibeh et al. [15]	Tourist arrivals	GDP	Lebanon
Dogru and Bulut [34]	Growth in tourism receipts	GDP growth	Seven Mediterranean countries
Enilov and Wang [16]	International tourist arrivals	GDP per capita	23 developing and developed countries
Eyuboglu and Eyuboglu [35]	International tourism receipts	GDP per capita	9 emerging countries
Gao et al. [13]	Tourism receipts	GDP per capita	18 Mediterranean countries
Gounder [36]	Tourist arrivals	Industrial production growth	Mauritius
Harb and Bassil [37]	Tourist arrivals	GDP per capita	27 European countries
Jayaraman and Makun [17]	Tourism revenue	Output per capita	Pacific island countries
Kronenberg and Fuchs [10]	Tourist expenditures	Employment and income	Swedish region of Jämtland
Kumar et al. [38]	Visitor arrivals per capita	GDP per capita	Cook islands
Lin et al. [39]	Ratio of international tourism revenue to GDP	GDP per capita	China’s 29 provinces
Liu et al. [18]	Tourism revenue	GDP per capita	China’s prefecture-level cities
Liu et al. [40]	Tourism receipts as a percent of GDP	GDP per capita	31 Chinese provinces
Liu et al. [41]	Visitor arrivals	GDP	Hong Kong of China
Llorca-Rodríguez et al. [42]	Nights spent at tourist accommodation	GDP per capita growth	258 regions in the European Union
Paramati et al. [19]	International tourism receipts	GDP per capita	26 developed countries and 18 developing countries
Pérez-Rodríguez et al. [43]	Tourist arrivals	GDP	14 European countries
Pérez-Rodríguez et al. [44]	International tourist arrivals	GDP	Seven European countries
Pratt [20]	International tourism arrivals	GDP per Capita	Small island developing states
Pulido-Fernández and Cárdenas-García [45]	Six tourism variables	Nine economic variables	143 countries
Roudi et al. [46]	International tourist arrivals	GDP	Small island developing states
Sahni et al. [47]	Tourism receipts as a percent of GDP	GDP per capita growth	23 African countries
Seetanah [21]	Tourist arrivals	GDP per capita	19 island economies
Sun and Drakeman [23]	Tourists arrivals	GDP, government tax, income, gross profit, employment	Australia
Tugcu [48]	International tourism receipts and tourism expenditures	GDP per capita growth	22 Mediterranean countries
Wu et al. [49]	Per capita international tourism receipts growth rate	GDP per capita growth	22 OECD countries
Zuo and Huang [50]	Tourist arrivals as a percentage of host population (TA) and tourism receipts as a share of real GDP	GDP per capita growth	31 provinces in mainland China

summarizes the main empirical studies on the relationship between tourism and economic growth. In contrast to traditional economic growth theories, Inchausti-Sintes [24] argued that the contribution of tourism to economic growth is less dependent on productivity gains, which occur mainly in the industrial sector, and more on factor accumulation. Nevertheless, the author also believed that tourism has a positive long-term economic impact.

There is a basic consensus that tourism is difficult to incorporate into economic growth models as a factor of production, whether in the endogenous growth theory or the new economic growth theory. However, in terms of GDP accounting, tourism can be one of the key mechanisms by which tourism can effectively boost consumption, one of the main ways of growing GDP, to contribute to GDP growth. Specific economic phenomena can integrate tourism into mainstream economic growth theory. For example, Zuo and Huang [51] found that tourism boosts sectoral productivity through an industrial transfer from agriculture to tourism, i.e., emerging tourism-related industrialization, thus contributing to economic growth. Notably, this has occurred only in particular agricultural-dependent destinations. Furthermore, Liu and Wu [52] confirmed the economic spillover effects of tourism on other sectors through human capital, physical capital, and public services, based on the new economic growth theory.

Economic development provides a high-quality supporting environment for tourism and thus enhances the service capabilities for tourists. As a result, numerous scholars have found that economic growth also contributes to tourism development. In order to promote sustainable tourism development, tourism management should focus on investment in tourism resources, environmental protection and infrastructure [53]. Consequently, bidirectional Granger causality between tourism and economic growth is found by Bilen et al. [28], Comerio and Pacicco [31], Danish and Wang [32], Dogru and Bulut [34], Kumar et al. [38], Pulido-Fernández and Cárdenas-García [45], and Roudi et al. [46].

Unlike the relationship between tourism and economic growth, the economic growth effects of tourist attractions have received less attention. However, undoubtedly, tourist attractions constitute the core attractiveness of most tourist destinations and are the endogenous driving force of regional tourism development, as well as being the core production factor for the tourism economy. Therefore, it is necessary to understand the relationship between tourist attractions, the sources of tourism motivation, and economic growth. Bi et al. [54] and Guo et al. [55] suggested the importance of tourist attractions in regional tourism development. Further, McKercher and Koh [56] stated that tourist attractions have a crucial role in promoting tourism demand among tourists unaware of the destination but are relatively less attractive if these tourists are more knowledgeable about the destination.

In addition, the GDP accounting theory suggests that consumption is a major contributor to economic growth. Theoretically, tourist attractions can contribute to regional economic growth by increasing consumption. Generally, consumption in tourism destinations consists of two aspects: one is direct tourism consumption, known as tourism receipts; the other is comprehensive tourism-driven consumption, which refers to the other local consumption of tourists, reflecting the multiplier effect of tourism. Therefore, three influence chains exist between tourist attractions and economic growth: tourist attractions→increasing attractiveness of tourist destinations→attracting more tourists→increasing tourism revenue→promoting economic growth; tourist attractions→increasing prices of tourism products→increasing tourism revenue (because of the low price elasticity of top tourism products)→promoting economic growth; tourist attractions→increasing attractiveness of tourist destinations→attracting more tourists→creating more (tourism) consumption→promoting economic growth.

In China, most tourists are unwilling to learn about tourist attractions. These tourists travel for Pearce’s middle- and outer-layer motives and rather choose the destinations subject to their popularity. This is one of the main reasons why top tourist attractions are always overcrowded during the holidays in China. These popular and well-known tourist attractions are usually highly rated such as 5ANTA. In China, a 5ANTA is extremely attractive to most tourists. Therefore, Lin et al. [4] asserted that China’s 5ANTA positively affects tourist arrivals. Tian et al. [8] found that the effectiveness of tourism-led growth depends on the availability and number of World Heritage or 5ANTA in each city. This in effect recognizes the role of top tourist attractions in driving economic growth. Zhang et al. [5] supported that heritage (a type of tourist attraction) tourism contributes to local economic development. Therefore, this study proposes the following theoretical hypothesis.

 Hypothesis 1. 

Tourist attractions significantly contribute to economic growth.

The regional heterogeneity of the impact of tourism on economic growth has also been widely acknowledged. This means that tourism’s impact on economic growth differs significantly across destinations. For example, the role of tourism in economic growth is higher in developing countries than in developed countries [16,47]. Furthermore, Seetanah [21] found that tourism in island economies may have relatively higher economic growth effects relative to developing and developed countries. Similarly, Benkraiem et al. [27], Harb and Bassil [37], Lin et al. [39], Liu et al. [41], Pérez-Rodríguez et al. [43], Tugcu [48], and Eyuboglu and Eyuboglu [35] also evidenced the regional heterogeneity of tourism’s relationship with economic growth. In conclusion, the tourism-led growth hypothesis has been widely accepted, but the strength and direction of this hypothesis vary significantly across empirical studies.

As indicated by Nunkoo et al. [3] and Fonseca and Sánchez-Rivero [57], regional heterogeneity essentially implies a moderating effect of socio-economic development on the relationship between tourism and economic growth. For example, Belgodere et al. [26] pointed out that inclusive institutions can increase the capacity of tourism to grow the economy, suggesting the importance of regional governance; Harb and Bassil [37] indicated that tourism has the greatest impact on economic growth in areas with a large number of highly educated populations, emphasizing the importance of human capital. Some other potential moderators include the financial technology index [49], economic and financial development [33], urban-rural disparity [40], information and communication technology [17], and income and infrastructure [25]. Analyzing these moderating factors allows for a more in-depth and specific discussion on the impact of tourism on economic growth and thus facilitates decision-making.

In addition to regional heterogeneity, other scholars have explored the temporal characteristics of the relationship between tourism and economic growth (e.g., Brida, et al. [29]; Wu et al. [49]). Different temporal stages often correspond to varying levels of economic development. These different economic characteristics, in a cross-sectional sense, correspond to various tourist destinations, such as developed and developing countries. Thus, temporal variability is essentially the same as regional heterogeneity. Another significant aspect of the temporal characteristics is the non-linear relationship between tourism and economic growth as explored by Chiu and Yeh [30] and Zuo and Huang [50]. The non-linear relationship actually still arises from the different economic characteristics of the tourist destinations at different periods.

Some scholars have discovered the potential negative effects of tourism on economic growth. For instance, Inchausti-Sintes [58] argued that although tourism boosts economic growth, it may lead to an over-reliance on the tourism sector, which could jeopardize long-term economic growth through undermining productivity gains, exchange rate appreciation and other Dutch Disease phenomena. Similarly, Llorca-Rodríguez et al. [42] identified inbound tourism as a drag on economic growth in the more developed neighborhoods. Pérez-Rodríguez et al. [44] denied the contribution of tourism to economic growth. Furthermore, other scholars have found that the relationship between tourism and the economy is influenced by contingent events, such as political and economic crises and natural disasters [36,41,43].

Based on the above discussion, this study summarizes the second hypothesis as follows:

 Hypothesis 2. 

Socio-economic factors significantly moderate the effects of tourist attractions on economic growth.

Despite such fruitful research on tourism (including tourist attractions) and economic growth, Kožić et al. [59] argued that the tourism-led growth hypothesis should be subject to rigorous scientific scrutiny. Existing studies focus mainly on the correlation between tourism and economic growth rather than causality. The correlation only proves a link between the two but does not explain why tourism causes economic growth. Moreover, current causality is more commonly known as Granger causality that, however, is essentially a time correlation. The Granger causality test only finds predictive causality but does not confirm true causality. Therefore, the existing conclusions on the relationship between tourism and economic growth should be treated with caution [60]. From this perspective, there is still a need for real causal inferences such as the natural experiment approach currently popular in economics to examine the impact of tourist attractions on economic growth.

In addition, there is a striking phenomenon in the available studies, namely the predominance of country-level panel data in the sample selection. These samples consist of global countries [30,33], European countries [34,37,44], OECD countries [49], island countries [11,17,46], BRICS countries [32], or the Mediterranean countries [28,48]. Several studies also involve provincial panel data, such as Lin et al. [39], Liu et al. [40], and Zuo and Huang [50]. However, the mainstream economics view believes that cross-country and provincial data are likely to produce pseudo-regressions and their conclusions are questionable due to the significant variability within provinces and countries and the small sample size. In fact, the use of country and provincial panel data has largely been abandoned by current mainstream economists. With this in mind, Tian et al.’s [8] empirical study based on urban panel data is a very good model.

A further shortcoming is that existing studies focus on the interpretation of relationships rather than on the underlying influence mechanisms, essentially due, of course, to the limitations of the Granger causality test. Prior literature has often neglected to explore in depth the mediating or moderating mechanisms in tourism and economic growth. Admittedly, some scholars have referred to various moderators, as indicated previously. However, these studies often have individual knowledge backgrounds, and most moderating effects are not general. Complete supporting facilities are more important for the economic role of tourist attractions. In China’s practice, not all destinations with top tourist attractions have a developed tourism business. Consequently, there is a need to systematically explore the role of regional socio-economic development levels in moderating the relationship between tourist attractions and economic growth.

Taking all the above discussion into account, this article employs a landmark method of quasi-natural experiments, namely the DID method, to explore the causal effects of tourist attractions on economic growth and the underlying mechanisms, based on Chinese city-level panel data, so as to significantly contribute to the existing body of knowledge.

3. Methodology

3.1. The DID Model

As mentioned above, from 2007 to now, the selection of China’s 5ANTA has been ongoing. In other words, the treating time varies significantly; therefore, the following time-varying fixed effects DID model is designed to estimate the treatment effects.

$$y_{i,t}=\alpha +{\mu }_i+{\lambda }_t+\theta tourist\_attractio{n}_i\times yea{r}_{i,t}+\beta \mathit{C}\mathit{O}{\mathit{N}}_{i,t}+{\epsilon }_{i,t}$$

(1)

where y_i,t represents the explained variable, namely economic growth, μ_i represents regional fixed effects, λ_t represents time fixed effects, i(i = 1,2,…N) represents the individuals, and t(t = 1,2,…,T) represents the time. CON represents the control variable that changes with time and individual, and ε_i,t are stochastic error terms. tourist_attraction_i denotes the policy dummy variable. If the individual i belongs to the experimental group, the value of touris_tattraction_i is 1; otherwise, its value is 0. Notably, the policy intensity in the model is also time-varying for the same region. Therefore, to precisely capture the effects of policy intensity, the dummy variable value in model (1) is also not limited to 0 or 1, but varies within the range 0, 1, 2, 3... depending on the actual number of 5ANTA in that year. year_i,t represents the dummy variable of the experimental period, indicating that experimental time varies across different individuals. When individuals in the experimental group enter the experimental period, the value of year_i,t is 1 or other natural number; otherwise, the value is 0. The coefficient θ represents the policy effect.

In order to track the policy effect and test the parallel trend hypothesis of the DID model, there is a need to analyze the dynamic effect of the policy [7], which can be formulated by the model (2) using the event-study approach proposed by Sun and Abraham [61].

$$y_{i,t}=\alpha +{\mu }_i+{\lambda }_t+{\displaystyle \sum _{k=-\zeta }^{\phi }({\theta }_k{(tourist\_attractio{n}_i\times yea{r}_{i,t})}^k)}+\beta \mathit{C}\mathit{O}{\mathit{N}}_{i,t}+{\epsilon }_{i,t},$$

(2)

where ζ denotes the ζth period before the experiment, and φ represents the φth period after the experiment.

Because the number of samples in the experimental group is larger than the control group in the model, it is no longer necessary to use the matching technique to select a matching control group but rather to use all the remaining samples as the control group. To verify the rationality of this approach, the article depicts the change in economic growth measured as GDP per capita in the experimental and control groups before and after the selection of 5ANTA, i.e., to test the prerequisites of the DID approach, the so-called parallel trend hypothesis [7]. If that change is significantly different, it indicates to some extent that 5ANTA selection has an impact on economic growth. Therefore, the choice of the control group is acceptable. Otherwise, the experimental group needs to be reselected and matched with the control group. Figure 1 represents that the change in the mean economic growth of the experimental group around 2007 is significantly different from the control group. Of course, there are many possible reasons for the change in economic development; the article will continue various tests to demonstrate the effect of 5ANTA on economic growth in Section 4 Results.

3.2. The Variable Selection

Typically, this study uses GDP per capita to indicate economic growth following Benkraiem et al. [27], Zuo and Huang [50], and Zhang and Zhang [62]. The policy dummy variable in the model is the number of 5ANTA in the region. As mentioned earlier, the 5ANTA construction has been a major tool for localities in China to enhance their tourist destination’s image and increase their tourism attractiveness. Theoretically, it has also become a meaningful way to promote economic growth.

Obviously, various factors are influencing economic growth. In economic accounting, there are three methods of calculating GDP: the production method, the income method, and the expenditure method. Each of the three methods reflects GDP and its components in a different way. The GDP calculated by the three methods represents the results of the productive activities of the same economy during the same period, and therefore, in theory, the results obtained by the three calculation methods should be consistent. In promoting economic growth in China, decisions are usually made from the expenditure approach, under which consumption, investment and exports basically constitute the GDP. Therefore, fixed asset investment and international trade are the key factors affecting GDP. This study identifies foreign direct investment, fixed asset investment, and trading openness as the control variables in the model.

In addition, endogenous growth theory considers technological progress to be a determinant of economic growth. Concretely, the acquisition of new knowledge such as innovation and technological progress, the accumulation of human capital, the stimulation of the application of new knowledge to production, and the resources to apply new knowledge such as human capital are the most central factors of economic growth. The new economic growth theory also emphasizes the importance of human capital. As a result, technological progress and investment in education are additional control variables in the model. In summary, this study selects technological progress, investment in education, foreign direct investment, fixed asset investment, and trading openness as control variables in the DID model [2,29,30], which are measured as the number of patents granted per 10,000 people, financial expenditure on education per capita, foreign direct investment per capita (FDI), fixed asset investment per capita (FAI), and import and export trade per capita (TRO), respectively.

3.3. Data Collection

Data on 5ANTA are collected from the Chinese Ministry of Culture and Tourism. Data on the number of patents granted and financial expenditure on education are collected from the China City Statistical Yearbook and city-level statistical bulletin of national economic and social development. Data on financial expenditure on education in individual regions come from public reports of local finance departments or estimates made from them. Data on foreign direct investment are obtained from the China City Statistical Yearbook, city-level statistical bulletin of national economic and social development, provincial statistical yearbooks, and public reports of local governments. Data on fixed asset investment come from the China City Statistical Yearbook and city-level statistical bulletin of national economic and social development. Data on import and export trade are collected from provincial statistical yearbooks and city-level statistical bulletin of national economic and social development.

In addition, this study cross-references data from the multiple statistical reports mentioned above to ensure their accuracy and consistency. In this case, the missing data for individual years for certain regions are supplemented by linear interpolation. Even so, there are some cities with significant missing data for several variables, which have to be removed from the sample. Finally, this study collects data from 313 cities (including prefecture, league, and municipalities) over the period 2001–2019, with a total of 5947 observations. Notably, this study uses constant 2001 prices to obtain real data for variables expressed in monetary terms. City-level price indices are obtained from provincial statistical yearbooks. Considering the significant differences in magnitudes of the different variables, the GDP per capita, financial expenditure on education per capita, FDI per capita, FAI per capita, and TRO per capita are converted into their natural logarithms.

Descriptive statistics for each variable are given in

Table 2. Descriptive statistics (313 city-level regions from 2001 to 2019).

	Maximum	Minimum	Mean	Median	Std. Dev.	Observations
GDP per capita (¥)	198,005.0	2730	30,841.06	23,932.00	23,160.03	5947
Dummy number of 5ANTA	1	0	0.296273	0	0.456661	5947
Number of 5ANTA	9	0	0.439284	0	0.86608	5947
Number of patents granted per 10,000 people	123.976	0	5.390516	1.7295	10.41141	5947
Financial expenditure on education per capita (¥)	5565	28	893.4753	827	567.8714	5947
FDI per capita (¥)	10,260	1	621.1157	224	986.5472	5947
FAI per capita(¥)	135,294	692	22,022.48	18,025.00	16,824.69	5947
TRO per capita (¥)	276,794	1	8824.836	1594	23,103.78	5947

. Table 2 indicates the significant differences in socio-economic development between different cities in China. Given this background, it is necessary to explore the regional heterogeneity of the impact of tourist attractions on economic growth and the possible moderating effects.

4. Results

4.1. The Benchmark Regression

Table 3. Benchmark regression results.

Variable	log GDP per Capita
tourist_attraction × year	0.0029 *** (0.0009)	0.0030 *** (0.0007)	0.0098 * (0.0054)	0.0094 ** (0.0046)
patents		−0.0009 ** (0.0004)		−0.0011 *** (0.0004)
log education		0.2108 *** (0.0133)		0.2106 *** (0.0133)
log FDI		0.0151 *** (0.0021)		0.0150 *** (0.0020)
log FAI		0.2234 *** (0.0072)		0.2241 *** (0.0072)
log TRO		0.0331 *** (0.0038)		0.0330 *** (0.0038)
Constant	10.0923 *** (0.0037)	6.2216 *** (0.0924)	10.0958 *** (0.0034)	6.2141 *** (0.0924)
Time fixed effects	Yes	Yes	Yes	Yes
Regional fixed effects	Yes	Yes	Yes	Yes
Observations	5947	5947	5947	5947
Adjusted R-squared	0.9446	0.9639	0.9446	0.9639

Notes: Standard errors are in brackets; ***, **, and * indicate the significance level of 1%, 5%, and 10%, respectively.

reports the benchmark regression results. Columns 1 and 2 show the regression results for 0–1 policy dummy variables; columns 3 and 4 show the regression results for considering the policy intensity. Column 1 shows that tourist attractions have a positive effect on regional economic growth. After controlling the relevant variables, the positive impact is also significant at the level of 1%. Columns 3 and 4 demonstrate that the positive effects of tourist attractions on economic growth increase to 0.0098 and 0.0094. This indicates that the effect of tourist attractions on economic growth is more significant after considering the policy intensity. In sum, Hypothesis 1 is confirmed. An interesting finding is that the magnitude of the coefficient of the tourist attractions in the policy intensity treatment is greater than in the 0–1 dummy variable. This is because the dummy treatment assumes that each experimental individual has only one 5ANTA, while the policy intensity scenario takes into account that the number of 5ANTA per experimental individual may increase over time. The results in Table 3 show that as the number of 5ANTA increases, the contribution of tourist attractions to the economic growth of the experimental group becomes more significant, which is reasonable and confirms the positive effect of tourist attractions on regional economic growth.

Table 3 also shows that compared with cities without 5ANTAs, the GDP per capita in the treated cities will increase by approximately 1.0030% (equivalent to 0.0030% growth in log GDP per capita) in the dummy treatment scenario. This figure will increase to 1.0094% (equivalent to 0.0094% growth in log GDP per capita) in the policy intensity scenario. This indicates that the construction of 5ANTAs in China contributes to growing the regional economy, thus confirming hypothesis 1. This result is consistent with most existing studies such as Akadiri et al. [12], Gao et al. [13], and Dibeh et al. [15], except that this study confirms the causal effect of tourist attractions on economic growth, while prior literature shows a positive correlation between the two. Another significant difference is that the existing studies focus more on the tourism economy, while this study focuses on the construction of tourist attractions.

4.2. Robustness Check

A basic premise of the DID analysis is the parallel trend hypothesis, i.e., excluding exogenous policy shocks, the explained variable in the experimental and control groups should have similar trends before the policy implementation while differing significantly after the policy shock [7]. The article therefore first argues for the robustness of the DID results through the parallel trend hypothesis test. Second, given that the choice of variables and estimation method may also impact the regression results, this study re-estimates model (1) by changing the explained variable and estimation method. Third, this study also performs a placebo test by constructing a dummy experimental and control group, a standard robustness test for DID analysis [7]. Finally, this study employs the instrumental variables test, given the possible omitted variables and reverse causality.

4.2.1. Parallel Trend Hypothesis Test

The parallel trend hypothesis has been visually verified in Figure 1. This study follows Roth [63] to perform another parallel trend hypothesis test. Roth [63] suggested that “extrapolations of pre-treatment data can then be used to estimate the counterfactual post-treatment difference in trends”. Therefore, this study performs trend extrapolation (based on the TREND function included in Excel) based on the pre-treatment data of the experimental group and then compares the differences between the predicted data and the actual data after treatment. Notably, because different cities may have their 5ANTAs at different times, each city’s economic development projections rely on pre-treatment data up to the approval time of the city’s 5ANTAs. The results are shown in Figure 2. Figure 2 illustrates a significant difference between the trend change in economic growth in the experimental group without considering policy drivers and the actual growth trend after the construction of 5ANTA. In particular, the predicted GDP per capita of the experimental group is significantly smaller than the real GDP per capita. Hence, Figure 2 again visualizes the positive impact of 5ANTA on economic growth.

This study continues to test this hypothesis quantitatively according to model (2). Figure 3 illustrates that policy variable coefficients are not significantly different from zero in the six periods prior to treatment. However, in the post-treatment period from 2008 to 2019, the coefficients are greater than 0 and are significant at the 1% or 5% level, confirming the parallel trend hypothesis. Another significant finding is that the economic growth effect of tourist attractions gradually increases over time. In particular, there is a significant increase in the regression coefficient after 2011, reflecting the rapid increase in the number of 5ANTA in China: from 2007 to 2010, the figure increased slowly from 66 to 75, and rapidly to 119 in 2011.

4.2.2. Alternative Estimation Method and Explained Variable

This study re-estimates model (1) using a new approach. The Ordinary Least Squares (OLS) method and Generalized Method of Moments (GMM) are often used as alternatives to each other for robustness checks [64]; hence, this study re-estimates the DID equation using system-GMM to obtain the results in

Table 4. The effects of tourist attractions on economic growth: Alternative explained variable and estimation method.

Variable	log GDP per Capita–System GMM Estimation
tourist_attraction × year	0.0992 *** (0.0110)	0.1141 ** (0.0532)
patents		0.0048 *** (0.0009)
log education		0.0884 *** (0.0275)
log FDI		0.0138 (0.0098)
log FAI		0.1441 *** (0.0298)
log TRO		0.1212 *** (0.0252)
Constant
Time fixed effects
Regional fixed effects
Observations	5321	5321
Adjusted R-squared
AR(1) p-value	0.0000	0.0000
AR(2) p-value	0.7824	0.2941
Sargan test (p-value)	0.3270	0.5062

Notes: Standard errors are in brackets; *** and ** indicate the significance level of 1% and 5%, respectively.

. The system-GMM estimation results also support this conclusion. This study re-estimates the DID equation using nighttime light instead of GDP per capita. Nighttime light, a type of satellite remote sensing data that provides a more objective picture of regional production and life activities, has been widely used to measure economic activity [65]. Therefore, nighttime light can be an excellent alternative variable to GDP per capita for a robustness test. This study follows the study of Zhang and Zhang [65] to obtain Chinese city-level nighttime light data. The new results are reported in columns 3 and 4 of Table 4 and are consistent with Table 3. Therefore, the results in Table 3 are robust.

4.2.3. Placebo Test

Referring to Zhang and Zhang [7], this study uses two methods to construct a dummy experimental group for the placebo test. The first is to uniformly advance the time in which each region has a 5ANTA by two years. Second, economic development may be additionally influenced by the local locational environment rather than facilitated by tourist attractions. For example, eastern China has a superior geographical environment and has long been more progressive in various aspects than the central and western regions. Given this background, this study re-estimates the DID equation by setting all cities in the eastern provinces as the experimental group. Following the studies of Zhang [64] and Zhang and Zhang [66], eastern China includes Beijing, Tianjin, Hebei, Shanghai, Jiangsu, Zhejiang, Fujian, Shandong, Guangdong, and Hainan, and the other provincial regions constitute the central and western samples.

Suppose the coefficient of the policy dummy variable remains significantly positive under the above virtual operations. In that case, it suggests that economic development is likely caused by other policy changes or stochastic factors rather than tourist attractions. The placebo test results are shown in

Table 5. The effects of tourist attractions on economic growth: Placebo test.

Variable	log GDP per Capita
tourist_attraction × year-two-years-in-advance	0.0075 (0.0060)	0.0024 (0.0055)
tourist_attraction × year-new-experimental-group			0.0856 *** (0.0109)	0.0175 (0.0127)
patents		−0.0012 *** (0.0004)		−0.0006 ** (0.0003)
log education		0.3125 *** (0.0141)		0.2102 *** (0.0133)
log FDI		0.0256 *** (0.0022)		0.0151 *** (0.0020)
log FAI		0.0682 *** (0.0072)		0.2223 *** (0.0072)
log TRO		0.0439 *** (0.0041)		0.0330 *** (0.0038)
Constant	10.0953 *** (0.0039)	7.0432 *** (0.0995)	10.1207 *** (0.0044)	6.2412 *** (0.0930)
Time fixed effects	Yes	Yes	Yes	Yes
Regional fixed effects	Yes	Yes	Yes	Yes
Observations	5947	5947	5947	5947
Adjusted R-squared	0.9446	0.9570	0.9454	0.9639

Notes: Standard errors are in brackets; *** and ** indicate the significance level of 1% and 5%, respectively.

. The results show that the effect of tourist attractions on economic growth is no longer statistically significant, either by advancing the policy time or by creating new fictitious experimental and control groups. This suggests that the causality between tourist attractions and economic growth does not come from other policies or random factors, nor the locational context, thereby confirming the robustness of the results in Table 3.

4.2.4. Instrumental Variable Test

This paper further adopts the instrumental variable method to test the robustness of the results. Limited by the availability of data and the complexity of economic growth, the model may have omitted variables. If the omitted variables are correlated with the error term, the OLS estimates will be biased. Although GMM estimates can overcome the endogeneity problem to some extent, the GMM is primarily used to address the endogeneity of lagged dependent variables. If the independent variable also has endogeneity problems, new instrumental variables are still theoretically needed to solve them. The regions with top tourist attractions are probably those with more developed economies, such as the three developed provinces of Jiangsu, Zhejiang, and Guangdong, which account for approximately 20% of the number of 5ANTA in 2019, such that there may be an inverse causal relationship between tourist attractions and economic development. Therefore, to avoid potential endogeneity problems caused by omitted variables and inverse causality, this study constructs an instrumental variable for tourist attractions and uses the two-stage least squares (TSLS) method to perform the robustness test.

Typically, a good instrumental variable should be correlated with the endogenous variables but not with the error term [67]. This study chooses the share of the provincial number of 5ANTA in that year as the instrumental variable for tourist attractions. Obviously, this proportion is highly related to the number of 5ANTA in each city, but this proportion does not directly affect the city’s economic development. The instrumental variable is regressed into model (1) to examine if it is appropriate. The results show that the coefficient of the instrumental variable is not significant and that the coefficient of the tourist attractions does not change significantly (this regression is not presented in this study due to the limited space). Therefore, there is no direct relationship between the instrumental variable and economic development. This study then performs the instrumental variable test using TSLS, and the results are shown in

Table 6. The effects of tourist attractions on economic growth: Instrumental variable test.

Variable	log GDP per Capita
tourist_attraction × year	0.0084 (0.0055)	0.0100 ** (0.0046)
patents		−0.0009 ** (0.0004)
log education		0.2117 *** (0.0133)
log FDI		0.0149 *** (0.0021)
log FAI		0.2240 *** (0.0073)
log TRO		0.0333 *** (0.0038)
Constant	10.1084 *** (0.0097)	6.2164 *** (0.0924)
Time fixed effects	Yes	Yes
Regional fixed effects	Yes	Yes
Observations	5947	5947
Adjusted R-squared	0.9446	0.9639
First stage regression
Instrumental variable	0.4181 *** (0.3014)	0.3691 *** (0.2524)
F-statistic	246.1501	378.3858

Notes: Standard errors are in brackets; *** and ** indicate the significance level of 1% and 5%, respectively.

. The results show that the impact of tourist attractions on economic growth remains significantly positive, and its coefficient does not change significantly (0.0094 versus 0.0100). Therefore, the concerns about endogeneity problems due to omitted variables and bi-directional causality do not exist, and therefore, the conclusions of Table 3 are robust.

4.3. Heterogeneity Analysis

Theoretically, designing a universal policy in a destination as vast as China is challenging, where the natural and economic environment varies considerably. Moreover, the attractiveness of a tourist attraction depends not only on its own construction but also on the level of socio-economic development of the tourist destination. Tourism attractiveness is made up of multiple factors other than the tourism product itself, such as the natural environment, transport infrastructure, supporting industrial development, and population size. Therefore, in theory, the effect of tourist attractions policies also depends on other conditions in tourist destinations. To test this argument, this study proceeds to explore the regional heterogeneity of the impact of tourist attractions on economic growth via group regression.

The full sample is divided into three groups, the east, central and west according to Zhang and Zhang [66] and Zhang [64]. Generally speaking, these three regional divisions broadly indicate the different levels of socio-economic development in China, with the eastern region having the highest level of socio-economic development, the central region the second highest, and the western region the lowest. The results of group regression are shown in

Table 7. The effects of tourist attractions on economic growth: Regional heterogeneity test.

Variable	log GDP per Capita
	Eastern China	Middle China	Western China
tourist_attraction × year	0.0034 *** (0.0008)	0.0182 ** (0.0069)	0.0097 (0.0092)
patents	0.0002 (0.0004)	0.0056 *** (0.0010)	−0.0058 *** (0.0020)
log education	0.1569 *** (0.0248)	0.1817 *** (0.0172)	0.2239 *** (0.0259)
log FDI	0.0047 (0.0065)	0.0142 *** (0.0028)	0.0170 *** (0.0033)
log FAI	0.2071 *** (0.0166)	0.2371 *** (0.0084)	0.2089 *** (0.0145)
log TRO	0.1068 *** (0.0109)	0.0320 *** (0.0053)	0.0189 *** (0.0059)
Constant	6.3885 *** (0.2137)	6.2435 *** (0.1044)	6.1799 *** (0.1971)
Time fixed effects	Yes	Yes	Yes
Regional fixed effects	Yes	Yes	Yes
Observations	1634	2166	2147
Adjusted R-squared	0.9676	0.9719	0.9489

Notes: Standard errors are in brackets; *** and ** indicate the significance level of 1% and 5%, respectively.

. In eastern and central China, there exist significant positive effects of tourist attractions on economic growth. However, the regression coefficients differ significantly: 0.0034 versus 0.0182. On the contrary, although the regression coefficient is also positive in the western region, it is not statistically significant.

The rough regional heterogeneity does not allow for a clearer understanding of the relationship between tourist attractions and economic growth, so this study further analyzes moderating effects. This study determines the moderating variables according to the theory of tourism motivation. First, the population of a tourist destination has a significant impact on local tourism demand [68]. The more populated a place is, the more likely it is to generate more tourists. Especially in China, short-haul tourism activities make up the bulk of tourism activities [66]. Therefore, tourist attractions are more likely to motivate local people to visit. Thus, the population is identified as the first moderating variable.

The second one is transport. Transport infrastructure is the primary vehicle for the movement of tourists. The article uses land transport miles to represent transport infrastructure [66]. However, it is important to note that there are no official statistics on rail transport at the prefecture level in China and that many cities do not have significant annual changes in rail mileage. Therefore, considering that tourism in China is still predominantly short-haul and relies heavily on road transport [66], only highway mileage is used to represent transport infrastructure, measured in highway miles per square kilometer. Third, this study estimates the moderating effect of per capita disposable income. Residents’ income is also the main factor generating tourism motivation. Given that urban residents currently constitute the majority of tourists in China, urban disposable income per capita is chosen as the moderating variable. Fourth, this study estimates the moderating effect of the industrial structure measured as the share of the tertiary industry. Tourism itself belongs to the tertiary industry and, to a large extent, the development level of the tertiary industry reflects the overall hospitality level of a tourist destination, as well as being an important indicator of economic development.

In sum, four moderating variables are identified: population, highway mileage per square kilometer, urban disposable income per capita, and the share of tertiary industry. The corresponding moderating effects are modeled as follows:

$$\begin{array}{l}{y}_{i,t}=\alpha +{\mu }_i+{\lambda }_t+\theta tourist\_attractio{n}_i\times yea{r}_{i,t}+\beta \mathit{C}\mathit{O}{\mathit{N}}_{i,t}+{\beta }_{1}populatio{n}_{i,t.}+\\ {\beta }_{2}tourist\_attractio{n}_i\times yea{r}_{i,t}\times populatio{n}_{i,t}+{\epsilon }_{i,t}\end{array}$$

(3)

$$\begin{array}{l}{y}_{i,t}=\alpha +{\mu }_i+{\lambda }_t+\theta tourist\_attractio{n}_i\times yea{r}_{i,t}+\beta \mathit{C}\mathit{O}{\mathit{N}}_{i,t}+{\beta }_{1}highwa{y}_{i,t.}+\\ {\beta }_{2}tourist\_attractio{n}_i\times yea{r}_{i,t}\times highwa{y}_{i,t}+{\epsilon }_{i,t}\end{array}$$

(4)

$$\begin{array}{l}{y}_{i,t}=\alpha +{\mu }_i+{\lambda }_t+\theta tourist\_attractio{n}_i\times yea{r}_{i,t}+\beta \mathit{C}\mathit{O}{\mathit{N}}_{i,t}+{\beta }_{1}incom{e}_{i,t.}+\\ {\beta }_{2}tourist\_attractio{n}_i\times yea{r}_{i,t}\times incom{e}_{i,t}+{\epsilon }_{i,t}\end{array}$$

(5)

$$\begin{array}{l}{y}_{i,t}=\alpha +{\mu }_i+{\lambda }_t+\theta tourist\_attractio{n}_i\times yea{r}_{i,t}+\beta \mathit{C}\mathit{O}{\mathit{N}}_{i,t}+{\beta }_{1}tertiar{y}_{i,t.}+\\ {\beta }_{2}tourist\_attractio{n}_i\times yea{r}_{i,t}\times tertiar{y}_{i,t}+{\epsilon }_{i,t}\end{array}$$

(6)

Population refers to the permanent population (unit: ten thousand people). Data on population and urban disposable income per capita (unit: ten thousand yuan) come from provincial statistical yearbooks and city-level statistical bulletin of national economic and social development. Data on the proportion of tertiary industry (unit: %) are from the China City Statistical Yearbook or the city-level statistical bulletin of national economic and social development. Data on highway mileage (unit: highway miles per square kilometer) are collected from provincial statistical yearbooks, city-level statistical bulletin of national economic and social development, or the relevant city-level statistical yearbooks. Missing data for individual years are also supplemented by linear interpolation. Urban disposable income per capita is also the real value in constant 2001 prices. Likewise, population and urban disposable income per capita are converted into their natural logarithms.

Table 8. The effects of tourist attractions on economic growth: Moderating effects.

Variable	log GDP per Capita
	Model (7)	Model (8)	Model (9)	Model (10)
tourist_attraction × year	0.0127 ** (0.0058)	0.0006 *** (0.0002)	0.0062 *** (0.0019)	0.0672 *** (0.0204)
log population	−0.2612 *** (0.0377)
tourist_attraction × year × log population	−0.0003 (0.0041)
Highway		0.0278 ** (0.0124)
tourist_attraction × year × highway		0.0076 *** (0.0029)
log income			0.0780 *** (0.0146)
tourist_attraction × year × log income			0.0348 *** (0.0093)
Tertiary				0.2883 *** (0.0177)
tourist_attraction × year × tertiary				0.0191 *** (0.0031)
patents	−0.0004 (0.0004)	−0.0013 *** (0.0004)	−0.0010 ** (0.0004)	−0.0014 *** (0.0003)
log education	0.2036 *** (0.0133)	0.2093 *** (0.0133)	0.2072 *** (0.0133)	0.2018 *** (0.0129)
log FDI	0.0146 *** (0.0021)	0.0145 *** (0.0020)	0.0147 *** (0.0021)	0.0158 *** (0.0020)
log FAI	0.2187 *** (0.0073)	0.2233 *** (0.0072)	0.2223 *** (0.0073)	0.2024 *** (0.0072)
log TRO	0.0319 *** (0.0038)	0.0328 *** (0.0038)	0.0336 *** (0.0038)	0.0297 *** (0.0037)
Constant	7.8191 *** (0.2495)	6.2095 *** (0.0925)	5.4951 *** (0.1627)	7.5521 *** (0.1217)
Time fixed effects	Yes	Yes	Yes	Yes
Regional fixed effects	Yes	Yes	Yes	Yes
Observations	5947	5947	5947	5947
Adjusted R-squared	0.9643	0.9640	0.9668	0.9660

Notes: Standard errors are in brackets; *** and ** indicate the significance level of 1% and 5%, respectively.

shows the results of the moderating effect test, where columns 2–5 correspond to the moderating effect of population, highway mileage, urban disposable income per capita, and the share of the tertiary industry, respectively. The results show that highway mileage, urban disposable income per capita, and the share of the tertiary industry have significant positive moderating effects. In other words, these three variables have a synergistic effect with tourist attractions on economic growth and can enhance the causal effects of tourist attractions. On the contrary, the moderating effect of the population is not significant. Conclusively, hypothesis 2 is also confirmed.

4.4. Discussion

This study is an important attempt to explore the causal mechanisms between tourist attractions and economic growth. Conclusively, this study makes the following major theoretical contributions to tourism economics. This study is the first to establish a causal relationship between tourist attractions and economic growth and to explore the mediating mechanisms of tourism revenue and tourist arrivals, as well as the moderating mechanisms of population, highway mileage, urban disposable income per capita, and tertiary industry development. This study breaks away from the limitations of existing studies that have focused on the correlation between tourism and economic growth and confirms the causal impact of tourist attractions on economic growth and the underlying influence mechanisms with city-level data. From a more general social science perspective, the article can enlighten more causality studies in the field of tourism. Methodologically, this study contributes to the knowledge of robustness tests in the tourism literature by proposing the substitution of explained variables, the placebo test, and the instrumental variables test.

The findings of this article advance the knowledge of the relationship between tourism (tourist attractions) and economic growth as evidenced by existing research. The identification of causal relationships provides solid theoretical support for the construction of top tourist attractions to promote economic growth. Regional heterogeneity analysis shows that the causal effect of tourist attractions on economic growth is more significant in east and central China. This is due to the more advanced and superior socio-economic development of the east and central regions compared to the western region. Good socio-economic development can provide various supports for tourism growth. These results are inconsistent with Enilov and Wang [16] and Sahni et al. [47] who evidenced that tourism contributes more to economic development in developing countries than in developed countries. However, our findings support to some extent Belgodere et al. [26] and Harb and Bassil [37] who found that better socio-economic development contributes to the role of tourism in economic growth. For example, several moderating variables identified in the article, such as highway mileage, urban disposable income per capita and tertiary industry development, are much better in the east and central regions than in the west. However, these economic endowments are relatively underdeveloped in western China; therefore, it is difficult to turn tourist attractions into an engine of economic growth. This partly explains why western China has many first-class tourism resources, but (tourism) economic development lags far behind that of the east and central regions.

The most pronounced causal effect in the central region suggests that, in addition to the contribution of the socio-economic environment to the causality of tourist attractions on economic growth, the central region is more dependent on tourism development than the eastern region. Western China has many excellent tourism resources that could create more top tourist attractions. However, significant investment in infrastructure and human resources is required to fulfill the economic growth function of tourist attractions effectively. Unfortunately, such investment is currently very limited. The limited financial, material, and human resources of the western region are often tilted towards people’s livelihoods or environmental protection, preventing more investment in the construction of top tourist attractions and related projects, thus preventing the promotion of tourism for the economy. The emphasis on economic development through developing tourism in the west, especially creating top tourist attractions, has not been achieved.

5. Conclusions, Policy Implications and Future Improvements

In contrast to existing studies, this article explores the causal impact of 5A National Tourist Attraction on economic growth using a difference-in-differences model based on Chinese 313 cities’ panel data from 2001 to 2019. The results show that tourist attractions significantly contribute to economic growth for the whole sample, and a series of robustness tests confirm the validity of this finding. The causality between tourist attractions and economic growth is only significant in China’s eastern and central regions. Road transport, urban disposable income per capita, and the share of the tertiary industry have a significant positive moderating effect on such causality. We conclude this study by making policy recommendations and suggesting current research limitations and future directions.

5.1. Policy Implications

This study proposes the following policy implications based on the above findings and discussion. First, it is feasible to promote economic growth by constructing top tourist attractions in many regions of China. Moreover, it is essential to pay attention to the multiplier effect of tourist attractions on economic development, as tourist arrivals do not only increase tourism consumption but also their comprehensive consumption, and their contribution to economic growth is significant. Consequently, the operating costs of the top tourist attractions can be somewhat ignored and compensated for by the total economic contribution to the local area, suggesting that low admission prices for the top tourist attractions are desirable.

Second, especially in the eastern and central regions of China, economic growth can be promoted by increasing the construction of top tourist attractions and the development of various supporting facilities. As tourist attractions promote economic growth mainly through tourist arrivals, the destinations could attract more tourists through the brand influence of top tourist attractions and boost tourism consumption and various investments. It is also vital to take full advantage of the good socio-economic development of the east-central regions to improve the level of services provided to tourists by supporting the development of various aspects of tourist hospitality such as transportation, accommodation, catering, and shopping, in addition to the construction of tourist attractions.

Third, developing top tourist attractions in the West must be cautiously approached. Even if top tourist attractions are developed in the western region with good supporting facilities, economic growth may not be necessarily promoted, as the lower disposable income of residents determines its limited domestic tourism market. Moreover, constructing top tourist attractions in the west would take up limited economic resources and lead to under-investment in other essential businesses, hindering regional sustainable development. In addition, the under-utilization of top tourist attractions in the west will result in unused facilities and wasted resources. This is especially true in the case of emergencies, such as COVID-19 that had a fatal impact on tourist attractions in the west. On the contrary, in the context of restricting the widespread movement of tourists, eastern regions are able to fully exploit their local source markets, mitigate the negative impact of unexpected events to some extent and recover more quickly from the shock of the crisis.

Fourth, our findings also have some implications for investment in tourist attractions. From an economic point of view, investing in the east and central regions of China is a wise move. The population should not be a key variable in investment decisions. Instead, investors should be concerned about whether local incomes, transport infrastructure, and the tertiary industry’s development level can provide sustainable support for the development of tourist attractions.

5.2. Limitations and Future Improvements

This study also could be improved by addressing the following issues in the future. While the causal impacts explored in this article are limited to the economy, the social impacts of tourist attractions are also significant. From a sustainable development perspective, the one-dimensional analysis of the economic effects at the expense of social effects should be criticized [10,22]. In the context of climate change, environmental impacts have become more prominent [23]. Thus, the role of tourist attractions in local visibility, image, cultural exchange, and environment also needs to be further explored. Furthermore, as with Liu and Wu [52] and Zhang [64], the distinction between domestic and inbound tourism may also lead to meaningful findings that deserve further exploration. Finally, it is worth noting that there exist significant differences between rural and urban tourism in China; therefore, it is important to distinguish between urban and rural tourism in exploring the impact of tourism.