**4. Results**

This section reports the estimation results. First, based on the gravity model with province-level data, the important influence of air pollution on China's inbound tourism was identified. Then, a set of robustness checks were conducted. After that, the heterogeneities among different tourist groups were explored.

## *4.1. Main Result*

The regression results for Equation (1) are listed in Table 2. The table shows the impacts of different factors on China's inbound tourism. Column (i) shows the baseline regression result. Air pollution in local provinces substantially harmed inbound tourism. The estimated coefficient for the variable *AirPollutioni* is −0.017, statistically significant at the 1% level. The coefficient implies that, if PM2.5 concentration rose by 1 μg/m3, inbound tourism arrivals would decline by 1.7%. Thus, Hypothesis 1 in this study is supported. Given that the average scale of inbound tourist arrivals in China between 2010 and 2016 was 133 million person-times per year, this number indicates that, if the country could take effective actions to reduce PM2.5 by 1 μg/m3, annual inbound tourist arrivals would increase by 2.261 million person-times. This is indeed a large benefit.


**Table 2.** Impact of air pollution on China's inbound tourism estimated at the province level.

Statistical significance: \* *p* < 10%, \*\* *p* < 5%, \*\*\* *p* < 1%. Abbreviations: AQI, air quality index; IV, instrumental variable; 2SLS, two-stage least squares; GMM, general method of moments.

It is notable that the degree of air pollution in tourist origin countries also had an obvious impact on tourist arrivals, as suggested by the estimated coefficient of *AirPollutionj*, which is −0.038 and statistically significant. This finding is not consistent with the finding by Wang et al. [35] that local air pollution stimulated outbound tourism. They reported that Chinese residents tended to have an increased demand for outbound tourism when local air quality became worse. What they found was not detected for tourists from foreign countries to China. The estimate in this study indicates that, on average, if PM2.5 density in potential tourists' home countries increased by 1 μg/m3, the actual number of tourists visiting China would decline by 3.8%, equal to a decline of 5.054 million person-times per year. Thus, Hypothesis 2 in this study is confirmed.

The estimated coefficients of the control variables are also reported in the table. The coefficient of *ln*(*Population*)*i* is not statistically significant, indicating that inbound tourism is not sensitive to the local population size. The coefficient of *ln*(*GDPpc*)*i* is significantly positive, indicating that expansion of inbound tourism is accompanied by overall economic development. The variables *ln*(*Scenic*)*i* and *ln*(*Hotel*)*i* both have significant positive coefficients, reflecting the straightforward opinion that more scenic spots and hotels are associated with a larger scale of tourism. The variables used as proxies for the abundance of infrastructure, *Hospitali* and *Transporti*, both have positive coefficients, though the coefficient of *Hospitali* is not statistically significant. *Urbani* and *GDPgri* both have significantly negative coefficients, implying that rapid urbanization and GDP growth actually do not increase the attractiveness of China to foreign tourists. *Structurei*, the indicator for industrial structure, has a significantly positive coefficient, which was expected. The variables for weather, *Temperaturei* and *Raini*, do not demonstrate any statistically significant impact.

The variables describing the characteristics of origin countries generally do not have a significant impact on inbound tourism in China. An exception is the variable *Transportj*, whose coefficient is positive and significant at the 5% level. This indicates that international tourists' decision to visit China is affected by the convenience of cross-country transportation.

The signs of interaction variables are consistent with the economic intuition. *ln*(*ER*)*ij* has a statistically significant negative coefficient, indicating that an increase in the relative price of tourism in China would reduce the number of inbound tourist arrivals. The coefficient of *ln*(*Distance*)*ij* is significantly negative, indicating a strong negative impact of travel distance. *TradeOpenij* is positively correlated with tourist arrivals, revealing that the trade linkage between two regions is associated with people's mobility. As expected, *VisaFreeij*has a positive coefficient, though not statistically significant.

## *4.2. Robustness Analysis*

This subsection describes several robustness checks that were conducted to inspect the robustness of the baseline estimation results. First, we inspected whether the estimate was robust to the selection of air pollution indicator. Previously, the level of PM2.5 concentration in ambient air was utilized to measure the degree of air pollution. Although PM2.5 is one of the most significant air pollutants in daily life, people sometimes check the AQI value rather than PM2.5 density to judge the severity of pollution. AQI has a nonlinear monotonic relationship with the physical density of air pollutants, and is also a widely used indicator of air pollution. Column (ii) of Table 2 reports the estimated coefficients if PM2.5 was replaced by the corresponding AQI value calculated based on the Chinese official standard. This time, the coefficient of *AirPollutioni* is −0.014, close to the value of −0.017 reported in column (i). The coefficient of *AirPollutionj* is −0.027, not far from the value of −0.038 reported in column (i). The results indicate that our estimation was not sensitive to the selection of index used to measure the degree of air pollution.

Next, the endogeneity issue in econometric regression was taken into account. Previous studies, such as Dong et al. [27], suggested paying attention to the possible endogeneity problem when estimating the impact of air pollution on tourism because environmental quality and tourism might have reciprocal interactions with each other [65]. The instrumental variable (IV) approach is an effective method to tackle the potential endogeneity problem. Two meteorological indicators, wind speed and

vapor pressure, were used as instrumental variables. Valid instrument variables should satisfy two conditions: they should be strongly correlated to the endogenous explanatory variable, and they should not directly affect the dependent variable, except through their link with the endogenous variable. The meteorological and environmental literature confirmed that meteorological conditions are strongly relevant to the degree of air pollution (e.g., [66–68]). There is no obvious reason to believe that tourists' visiting behaviors are sensitive to those meteorological indicators (as long as they are within normal ranges), except the relationship between air pollution and meteorological conditions. Thus, both conditions for the selection of IVs were satisfied.

Column (iii) reports the two-stage least squares (2SLS) IV estimation results. The estimated coefficient of *AirPollutioni* is −0.050, which is statistically significant and even larger than the coefficient in the baseline regression shown in column (i). The corresponding Cragg–Donald Wald F-statistic and Kleibergen–Paap rk Wald F-statistic are both statistically significant, indicating that the selected IVs were not "weak IVs". The Hansen J-statistic is not statistically significant, indicating that the regression model was not overidentified. Overall, these statistics imply that the IVs were properly used in the estimation. In order to further inspect the IV estimation results, the general method of moments (GMM) estimation was applied. The results, shown in column (iv), provide a statistically significant coefficient of −0.053 for *AirPollutioni*. Therefore, combining columns (iii) and (iv) together, the previous finding that air pollution harms China's inbound tourism is still supported after the potential endogeneity issue is explicitly addressed. In addition, it is easy to see that the finding about the negative effect of air pollution in tourist origin countries is robust. As demonstrated in columns (iii) and (iv), the estimated coefficients of *AirPollutionj* are −0.039 and −0.040, respectively, very close to the value of −0.038 reported in column (i).

#### *4.3. Heterogeneities among Different Tourist Groups*

Different types of inbound tourists may respond to air pollution in dissimilar ways. Considering this, tourists were classified into groups by the characteristics of their origin countries and destination regions, and their heterogeneous responses to air pollution were analyzed. Because the air pollution in tourists' origin countries is out of China's control, in this subsection, we focus on the heterogeneous responses of tourists to the air pollution within China. The analysis was based on Equation (2):

$$\begin{aligned} T\_{ijt} &= \eta\_1 A \text{irPollution}\_{it} + \eta\_2 A \text{irPollution}\_{it} \times D + \varrho \text{AirPollution}\_{jt} \\ &+ \text{Destination}\_{it}u + \text{Org} \dot{\eta}\_{jt}\mathbb{R} + \text{Interaction}\_{ijt} \gamma + s\_i + u\_j + v\_t + \varepsilon\_{ijt}. \end{aligned} \tag{2}$$

Equation (2) was revised from Equation (1) by adding the interactive term *AirPollutionit* × *D*. Here, *D* is a dummy variable equal to 1 or 0, depending on the classification of tourists. For the tourist group with *D* = 1, the impact of air pollution in China is measured by the coefficient *η*1 + *η*2; for the tourist group with *D* = 0, the impact is measured by the coefficient *η*1.

First, tourists were classified according to the degree of air pollution in their origin countries. This classification makes sense because the basic logic behind this study is that people will compare the air quality in candidate destinations with that in their home country and, ceteris paribus, will be more willing to visit places with better air quality. Consistent with this logic, if the degree of air pollution in tourists' origin country is high, potential tourists may be highly concerned about the pollution problem, and be very responsive to the air pollution in China. To verify this viewpoint, we set up one "high-pollution" country group and one "low-pollution" country group. The dummy variable was defined such that *DHighPollution j* = 1 if the mean air pollution level in country *j* during the sample period was above the sample average, and *DHighPollution j* = 0, otherwise. The estimate results are reported in column (i) of Table 3. The estimated coefficients of *AirPollutioni* and *AirPollutioni* × *DHighPollution j* are −0.014 and −0.006, respectively. Both coefficients are statistically significant. Therefore, it was found that, if PM2.5 pollution increased by 1 μg/m3, the number of tourists from "low-pollution" and

"high-pollution" countries would decline by 1.4% and 2.0% (= 0.014 + 0.006), respectively. Indeed, as expected, tourists from more polluted countries were more responsive to China's air pollution.

Second, the influence of air pollution on tourists from Asian and non-Asian countries was examined. The majority of inbound tourists to China come from Asian regions. Asian and non-Asian tourists may react to air pollution differently. To investigate this, we set the dummy variable such that *DAsianCountries j* = 1 if tourist origin country *j* is in Asia, and *DAsianCountries j* = 0, otherwise. (We classified Russia as an Asian country because it shares very long common national boundaries with China.) As reported in column (ii) of Table 3, the estimated coefficient of *AirPollutioni* is −0.014, and the coefficient of *AirPollutioni* × *DAsianCountries j* is −0.005. Both are statistically significant. Thus, it was found that China's air pollution has a larger negative impact on Asian tourists than on non-Asian tourists.


**Table 3.** Heterogeneous impact of air pollution on China's inbound tourism estimated at the province level.

Statistical significance: \* *p* < 10%, \*\* *p* < 5%, \*\*\* *p* < 1%.

Third, tourists were classified according to the air pollution level in their visiting destinations. We set up one "high-pollution" province group and one "low-pollution" province group. The dummy variable was defined such that *DLowPollution i* = 1 if the mean air pollution level in Chinese province *i* during the sample period was below the sample average, and *DLowPollution i*= 0, otherwise.

The estimation results are reported in column (iii) of Table 3. The coefficient of *AirPollutioni* is −0.011, and the coefficient of *AirPollutioni* × *DLowPollution i* is −0.035. Both are statistically significant. Thus, we essentially detected a nonlinear effect of air pollution on China's inbound tourism. In "high-pollution" and "low-pollution" provinces, if PM2.5 concentration rose by 1 μg/m3, the number of inbound tourist arrivals would decline by 1.1% and 4.6% (= 0.011 + 0.035), respectively. In other words, tourists who decide to visit more polluted areas are less sensitive to the variations of air pollution, and those that choose to visit less polluted areas care much more about pollution.

Lastly, tourists were grouped according to the degree of popularity of their visiting destinations. We considered one "popular destinations" province group and one "less popular destinations" province group. We defined the dummy variable such that *DPopularDestiantions i* = 1 if the mean number of annual inbound visitors to province *i* during the sample period was above the sample average, and *DPopularDestinations i* = 0, otherwise. As demonstrated in column (iv) of Table 3, the estimated coefficient of *AirPollutioni* is not statistically significant, and the coefficient of *AirPollutioni* × *DPopularDestinations i* is −0.040 and significant at the 1% level. Thus, it was found that tourists visiting popular Chinese destinations are sensitive to air quality, but tourists who choose to visit less popular destinations are not responsive to air pollution. Indeed, if a foreign tourist decided to travel to a Chinese region that was not visited by many people, the tourist would probably have an extraordinary interest or reason, e.g., for the purpose of business or conference. In this circumstance, air quality may not be a major concern.

In summary, the heterogeneity analysis shows that the magnitude of tourists' responses to air pollution in China is dependent on the characteristics of their origin and destination regions. The impact of air pollution in China is larger for travelers coming from more polluted and Asian countries, and visiting less polluted and more popular destinations.

#### **5. Discussion and Implications**

This study emphasizes the importance of good air quality for inbound tourism development. While this point has been confirmed by previous literature, this study aimed to estimate the impact of air pollution on the basis of a wide sample by using a gravity model, in which the features of destinations, origin regions, and their interactive relationship were explicitly modelled. According to the estimation results, if PM2.5 concentration in China rose by 1 μg/m3, inbound tourist arrivals would decline by approximately 1.7%. Hypothesis 1 is strongly supported. This result confirms the importance of a clean environment as a favored characteristic of tourist attractions as argued in the literature, such as by Goodwin [4], Hu and Wall [5], Mihaliˇc [6], Zhang et al. [7], and Zhang et al. [8].

An interesting finding of this study is that, if PM2.5 in tourist origin countries increased by 1 μg/m3, tourist arrivals in China would drop by 3.8%. This large impact has not been noted in previous studies. Environmental studies, such as those by Atari et al. [37], Dong et al. [40], and Moffatt et al. [38], reported that the existence of local air pollution would raise residents' awareness of and concern about the pollution problem. Therefore, if air quality in their home countries go<sup>t</sup> worse, potential tourists living in foreign countries would be less willing to choose China as a tourism destination. The estimation in this study confirms Hypothesis 2. Although this finding has no direct practical implication for China's tourism development because China cannot change the level of air pollution in foreign countries, it reminds researchers that air pollution in tourist origin regions is an explanatory variable in tourism demand analysis that cannot be ignored.

It is notable that the estimated magnitude of the impact of air pollution in China is different from that reported in previous studies. Table 4 briefly summarizes several previous studies on the impact of air pollution in China on inbound tourist arrivals. There are five columns in the table. The first column, "Literature", lists the authors' names. The second column, "Area Studied", and third column, "Period Covered" provide information about the sample regions and periods studied. Different air pollutants were used in previous studies to represent the degree of air pollution. The fourth column of the table, "Air Pollutant", lists the names of air pollution indicators. The last column, "Estimated Effect", reports the estimated response of inbound tourist arrivals to a 1 μg/m<sup>3</sup> increase in air pollutant concentration.


**Table 4.** Estimated response of inbound tourist arrivals to a 1 μg/m<sup>3</sup> increase in air pollutant concentration in China, as reported in previous studies.

For instance, according to the study by Dong et al. [27], who investigated 274 cities in China for the period 2009–2012, if PM10 concentration in ambient air increased by 1 μg/m3, inbound tourist arrivals would decline by 0.56%. Zhou et al. [31] focused on Beijing City, and reported that, if PM10 density increased by 1 μg/m3, inbound tourist arrivals in Beijing would decline by 0.33%. Differently from these two studies, Xu et al. [29] reported that the response of inbound tourist arrivals to PM10 pollution was not significant statistically, based on a sample covering 337 cities. However, they found that inbound tourists were sensitive to PM2.5 pollution. The estimated effect of a 1 μg/m<sup>3</sup> rise in PM2.5 concentration on tourist arrivals was −1.23%. But this finding was not supported by Liu et al. [16], whose study did not report a statistically significant impact of PM2.5 pollution on tourism. Overall, these previous studies have not provided a consensus on the magnitude of the impact of air pollution on inbound tourism in China.

Although the core finding of this study on the harmful influence of air pollution in China is qualitatively consistent with the findings of Dong et al. [27], Xu et al. [29], and Zhou et al. [31], the estimated effect of air pollution in this study is quantitatively different from that reported in previous studies. Particularly, the estimated negative impact is much stronger than that reported by Liu et al. [16] and Xu et al. [29], who also took PM2.5 as the indicator of air pollution. Since this study utilized a sample covering almost all Chinese regions and focused on recent years, the results may better reflect the general situation in China in the recent period.

The estimation results in this study enables a quantitative evaluation of the potential of promoting inbound tourism by improving air quality. According to the estimation, if PM2.5 density could be reduced by 1 μg/m3, inbound tourist arrivals would rise by 1.7%. This impact is substantial. For instance, if PM2.5 density can be reduced by 10 μg/m3, which is not an unrealistic target, it is expected that inbound tourist arrivals would rise by around 17%. In 2016, the number of annual inbound tourist arrivals was 138 million person-times. An increase of 17% represents 23.46 million person-times. Obviously, improving air quality should be considered as a practical and effective way to promote China's inbound tourism. If air quality in China can be substantially improved in the future, inbound tourist arrivals could potentially increase by at least tens of millions of person-times.

As reported in the section on heterogeneity analysis of different tourist groups, inbound tourists are more sensitive to air pollution in less polluted and more popular destinations. This finding indicates that air quality is a critically important factor of sustainable development in tourism-dependent areas. Because many scenic spots in less polluted areas are famous for their natural landscape, air pollution will substantially reduce the attractiveness or even destroy the beauty of these spots. Since popular destinations are representative of China and preferred options for most international travelers, their ability to attract tourists largely determines China's position in the world tourism market. Thus, particular emphasis should be placed on pollution control in currently less polluted and more popular tourist destinations, in order to improve the attractiveness and competitiveness of China's tourism.

Based on the results of this research, two practical implications can be drawn for China's inbound tourism. First, from the perspective of industrial policy, it is urgen<sup>t</sup> to implement environmental regulations effectively to ensure that the environment can be improved in the future. This is especially crucial in districts whose economies are largely dependent on tourism. The local governments in those districts should prioritize the issue of air quality improvement. Second, from the perspective of tourism marketing, in order to attract more international tourists, China's destination image should be repaired from the negative influence of air pollution. It is valuable to inform potential foreign tourists that there are attractive places with good air quality and the air quality is getting better. The tourism sector needs to have supportive policies and conduct strong tourism destination marketing campaigns, such as participating in international tourism exhibitions and forums.

In fact, in the past few years, China has implemented a comprehensive and complicated set of policies to reduce air pollution, including many environmental laws and standards, environmental action plans proposed by the central and local governments, and specific and detailed regulatory measures on the production and economic activities [69,70]. China has achieved some substantial success in air pollution reduction, especially concerning SO2 and NO*x* emissions. However, so far, PM concentrations remain high, and haze remains a severe problem in many areas [69,71]. It is necessary to promote research on the sources of PM pollutants, and apply new technologies and methods to further reduce emissions. Improving air quality will generate grea<sup>t</sup> benefits. More tourists will be attracted to boost the economy, and local residents' life and public health will also be ameliorated.

#### **6. Conclusions and Directions for Future Studies**

In this study, we explored the negative effects of air pollution on inbound tourist arrivals in China, based on a gravity model using data of province-level inbound tourist arrivals from 13 origin countries during the period 2010–2016. The estimation results show that, on average, if PM2.5 concentration in China increased by 1 μg/m3, inbound tourist arrivals would decline by approximately 1.7%. This verifies Hypothesis 1, confirming that clean air is an important element of attractive tourist destinations. This finding generates a clear policy implication: China's inbound tourism can be substantially expanded by implementing environmental protection policies. In addition, it was found that, if PM2.5 concentration in tourist origin countries rose by 1 μg/m3, inbound tourist arrivals in China would decline by roughly 3.8%. This supports Hypothesis 2, which can be explained by potential tourists' increased perception of and concern about air pollution in response to the pollution problems in their home countries. This finding indicates that an accurate modelling of tourism demand should also take into account the influence of pollution in regions where tourists come from.

This study was restricted by several limitations that could be addressed in the future. First, this study focused on air pollution and neglected other types of pollution such as water pollution and solid waste. Although air pollution has a severe impact on tourism, as reported by this study, other kinds of pollution might also have a substantial influence (e.g., [72–74]). Taking into account multiple pollution categories would help provide a more comprehensive understanding of the environment–tourism nexus. In the future, researchers can consider different pollution types as independent variables in one regression model, and compare their estimated impacts. This will help identify the relative importance of different pollutants and facilitate the design of efficient policies. Second, this study evaluated the benefits of improving air quality to boost tourism, but did not assess the costs of air pollution reduction. Obviously, actions to improve air quality are not free. For instance, some environment-friendly production processes should be adopted and some air-cleaning equipment needs to be installed. In the future, a detailed cost–benefit analysis would provide valuable suggestions for tourism policy-makers. This requires the researchers to collect detailed information about the costs and benefits of pollution reduction. As it is difficult to obtain sufficient data for a wide geographic area, researchers may start from the analysis on a specific small region, such as one scenic spot.

**Author Contributions:** Conceptualization, data curation, formal analysis, funding acquisition, methodology, and original draft preparation, D.D.; literature review, review and editing, software, and validation, B.X. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by the Fundamental Research Funds for the Central Universities (Grant No. JBK1809054).

**Acknowledgments:** The authors are grateful to the editors and three anonymous referees for their comments and suggestions.

**Conflicts of Interest:** The authors declare no conflict of interest.
