Why and how tourism affects green development: evidence for China

Abstract Green development, considering economic and environmental performance, has received increased attention globally. Moreover, the significant economic and environmental impacts of tourism are widely accepted by scholars. However, the nexus of tourism and green development is still unclear. To fill this gap, this study examines the effects of tourism on green development performance measured as green total factor productivity and the influence mechanisms. China’s 308 cities’ panel data between 2005 and 2019 provide empirical evidence. The results demonstrate an inverted U-shaped relationship between tourism and green total factor productivity, which is substantiated via various rigorous robustness tests. The key drivers of green total factor productivity change in cities include tourism-induced technological innovation, industrial structure optimisation, and environmental enhancement. Besides, the nexus of tourism and GTFP varies significantly across different socio-economic development regions. The positive relationship exists in relatively developed areas, while less-developed western China still keeps inverted U-shaped links between tourism and GTFP. The U-shaped relationship exists in central and northeastern China. The article is the first to systematically explore the impact of tourism on green development while revealing several potential mechanisms, thus significantly contributing to the literature on green development and sustainable tourism.


Introduction
Economic development has long been the most important goal pursued globally, especially in developing countries, even at the cost of environmental degradation.However, the effects of environmental pollution are cumulative and amplified and may become irreversibly worse over time, ultimately affecting the sustainability of the economy.Therefore, the tension between economic growth and environmental protection constrains sustainable global growth (Grossman & Krueger, 1995;Cordero et al., 2005).As a result, green total factor productivity (hereafter GTFP), which integrates economic and environmental performance, has received increasing attention in different countries and regions (e.g., Xie et al., 2021;Wu et al., 2020).Also, GTFP has become an essential indicator of green and sustainable development (Zhang et al., 2021).A GTFP-oriented development model can effectively reduce the impulse of local governments to pursue pure economic growth (often measured by GDP), increase the endogenous motivation to reduce energy and resource consumption, save energy and reduce emissions, and promote healthy competition between different countries and regions.
The significant economic impact of tourism has been well documented in the extensive literature (e.g., Nunkoo et al., 2020;Tang, 2021;Zhang & Zhang, 2021).Similarly, tourism-induced environmental change has also been supported (e.g., Lv & Xu, 2021;Sun et al., 2020;Peng et al., 2022).Since GTFP combines increased economic output and reduced environmental pollution, both influenced by tourism growth, tourism theoretically also significantly affects GTFP.However, to our best knowledge, few studies have focused on the nexus of tourism and GTFP and explained the underlying influence mechanism.In order to gain a deeper understanding of the relationship between tourism and green growth, two fundamental theoretical issues should be addressed.One is to clarify the effects of tourism on GTFP; the second is to explain how tourism affects GTFP.However, the two theoretical issues mentioned above have not been well explored.Not only is the impact of tourism on GTFP unknown, but how this impact occurs is also a black box.Therefore, as of now, we can still not link tourism and green development theoretically.
For a long time, China has led the world in economic growth.Since China's reform and opening up in 1978, China's GDP has grown by about 41.97 times by 2021, with an average annual growth rate of approximately 6.11%. 1 However, rapid economic growth has led to excessive consumption of natural resources and environmental pollution; a purely GDP-driven growth model has seriously undermined the basis for China's sustainable economic development (Song et al., 2018).In the 21st century, China's environmental pollution problems have attracted great attention from the government, which launched the 'Green GDP Accounting' project in 2004.Green GDP is obtained by deducting resource consumption and environmental losses from GDP; however, due to the high correlation between resources, energy and capital and the difficulty of measuring environmental costs, an accurate and authoritative technical accounting system for green GDP in China has not yet been established.
Another concept that has been gaining attention in recent years to integrate economic growth and environmental protection is the GTFP (Xia & Xu, 2020;Li & Chen, 2021;Lee & Lee, 2022).GTFP measures take into account not only the inputs of labour, capital and energy consumption in economic growth, but also ecological and environmental issues.It is estimated that China's GTFP fluctuates significantly, and after considering environmental factors, the GTFP growth was even negative in many cases (Xia & Xu, 2020).China's rapid economic growth is accompanied by fast-growing tourism industry.Before the COVID-19 pandemic, tourism's combined contribution to GDP exceeded 11% in 2019. 2 The environmental effects of increased tourism consumption are also considerable (Tang et al., 2022).For example, Tang et al. (2021) argued that China's tourism is emitting increasingly more carbon dioxide; Zhang and Zhang (2021) found that tourism Granger causes carbon emissions.
In summary, the development practice in China provides excellent empirical evidence to explore the link between tourism and GTFP.Consequently, we cite China as the case to empirically discuss the above-mentioned theoretical issues.Concretely, we develop the theoretical hypotheses of tourism and GTFP.Methodologically, we apply econometric methods to explore the nonlinear effect of tourism on GTFP and its influence mechanism as well as further explore the regional heterogeneity of such effects.
In contrast to prior studies, this article has the following unique contributions.This study establishes a theoretical association between tourism and GTFP and is the first to systematically explore the impact of tourism on regional GTFP focusing on both the economic and environmental aspects on the basis of the existing isolated studies of tourism's economic or environmental impact.Therefore, this study expands the scope of tourism effects and enriches the understanding of the relationship between tourism and sustainability.The article also provides a theoretical and decision-making reference for determining and regulating tourism development in the context of quality development.Conversely, the article offers a new perspective on tourism to enhance the regional GTFP.In addition, the study reveals several potential mechanisms by which tourism affects GTFP, contributing to the theoretical knowledge of GTFP and the influence of tourism.Examining the regional heterogeneity also contributes to deepening the understanding of the relationship between tourism and GTFP and helps build an organic and coordinated development mechanism that encompasses tourism growth and GTFP enhancement in a localised manner.
We organise this paper as follows.In Section 2 we provide the theoretical basis and summarise the research hypotheses.Then we introduce the research method, variable determination and data sources in Section 3. Next, we report the empirical results and perform a series of robustness tests in Section 4 as well as extend this study by discussing the influence mechanism and regional heterogeneity.Finally, we highlight the theoretical and practical implications of this study in Section 5.

Theoretical background and hypothesis
Total factor productivity (TFP) is the output efficiency of the combined inputs of labour, capital, and other production factors (Van Beveren, 2012); GTFP, on the other hand, is an improvement of TFP by additionally introducing the factor of energy resources, taking into account both the production factor inputs and energy and resources consumption.Therefore, GTFP is a new sustainable development indicator that integrates factors of production such as labour, capital and consumption of energy and resources into a single analytical framework (Xia & Xu, 2020).Furthermore, the GTFP-oriented development model considers the rational consumption of energy and resources and the reduction of environmental pollution while guiding economic growth.Accordingly, GTFP has become an important concept in modern economics and an essential tool for analysing sustainable economic development, as well as an important basis for governments to formulate long-term growth policies.
Industrial structure optimisation and technological innovation are important drivers of GTFP improvement.As an important basis for economic growth, industrial structure optimisation is also considered a pollutant controller (Zhou et al., 2013), which largely determines the direction of economic development and the quality of environmental protection and thus constitutes a critical way to enhance GTFP (Li & Lin, 2017).Industrial structure optimisation is also often accompanied by the continuous release of the industrial structure dividend.Concretely, the continuous turnover between the old and new industries has led to the gradual emergence of clean and efficient sectors, which has led to the continuous optimisation of resource allocation and improvement of production efficiency of the whole economy, thus effectively improving GTFP (Sun et al., 2022).
Technological innovation as a means to enhance resource allocation efficiency can effectively reduce energy consumption per unit of output, thereby increasing GTFP in the long run (Liu & Dong, 2021;Wang et al., 2021).In particular, after limiting strict environmental red lines and emission standards, a more effective incentive mechanism for green technology innovation will be formed, further generating continuous positive support for GTFP (Ai et al., 2020).But, of course, the productivity paradox also argues that unsuitable technological innovation is difficult to match with the development stage and factor endowment structure of the economy, which will weaken the intrinsic motivation of technological innovation and thus is not conducive to improving GTFP (Jacobs & Nahuis, 2002).Also, the environmental paradox suggests that technological innovation oriented exclusively towards economic growth may negatively impact the environment and thus further hinder the improvement of GTFP.
With its traditional reputation as a green and smoke-free industry, tourism plays an important role in global sustainable development.Therefore, extensive prior studies have examined the effects of tourism on various aspects of sustainability, especially with a broader focus on the economic and environmental impacts.For example, Faber and Gaubert (2019), Li et al. (2018) and Zhang and Zhang (2021) found the positive effects of tourism on economic development; Balli et al. (2019) supported the positive environmental effects of tourism.In contrast, tourism also exerts adverse effects on the economy and environment in some cases, such as Antonakakis et al. (2015), Lee and Brahmasrene (2013), Yıldırım et al. (2021), andTian et al. (2021).Notably, these tourism impacts are often fragmented; scholars do not organically integrate economic and environmental impacts, even within the same tourism analytical framework.Given the integrated consideration of GTFP for economic and environmental development, exploring the impact of tourism on regional GTFP would undoubtedly lead to a better understanding of the role of tourism in regional sustainable development, yet this has not received sufficient scientific attention.Because of the significant impact of tourism on the economy and environment, we hypothesise that tourism also significantly impacts GTFP.
The continued growth of tourism may lead to a shift in its economic and environmental impact.For example, continued tourism growth will lead to a shift of resources to the non-tradable sector, which may jeopardise productivity and thus negatively affect economic development, namely the so-called Dutch Disease (Inchausti-Sintes, 2015).Similarly, the continued growth of tourism will also lead to a shift in the direction of its impact on the environment due to the increased demand for quality tourism products, the willingness of tourists to pay for quality eco-environment and the increasingly stringent environmental regulations.In other words, there may be nonlinear relationships between tourism and economic growth or environmental performance (e.g., Chiu & Yeh, 2017;Zuo & Huang, 2018;Ehigiamusoe, 2020;Bella, 2018).Hence, we also hypothesise that a nonlinear association exists between tourism and GTFP and obtain the first hypothesis of the article: Hypothesis 1 There is an inverted U-shaped relationship between tourism and GTFP.
In terms of the mechanisms by which tourism affects GTFP, we first discuss the possible effects of tourism on technological innovation and industrial structure and thus GTFP.Tourism exerts innovative effects in terms of talent clustering, investment factor clustering, and technology exchange or spillover.First, tourism development can enhance the brand image and destination's attractiveness and promote the development of destination services, especially high-end services and consumer centres.The resulting quality employment and living environment will further increase the concentration of talent, thus increasing regional innovation.Second, tourism growth increases the transparency of the destination, making it easier to attract investors' attention and reducing the cost of their visits, thus increasing the success rate of offsite investments and providing sufficient financial support for regional innovation.Third, tourism development strengthens the level of destinations' openness and communication and thus contributes to the exchange of scientific and technological innovations, promoting the overall level of regional innovation.Some studies provide empirical evidence for the above theoretical analysis.For instance, Richards (2020) found that cities increase their attractiveness to creative talent by developing 'creative cities' to leverage the positive role of creativity in placemaking in tourism destinations.Ong and Liu (2022) confirmed that urbanisation driven by large-scale tourism development enhances its attractiveness to capital and thus regional innovation capacity.Ruhanen et al. (2021) argued that new tourismrelated knowledge can be generated and shared through tourism talent assistance and that this function can be better enhanced to contribute to regional innovation.Weidenfeld (2013) suggested that intensive mobility of people promotes knowledge transfer and innovation diffusion between tourist destinations.Besides, as an important service industry, tourism has a great demand for technology development and applications in order to meet the high-quality experience of tourists, thus contributing to regional technological innovation to a certain extent (Aldebert et al., 2011).Consequently, we summarise the second hypothesis as follows: Hypothesis 2 Tourism affects GTFP through changing regional innovation capacity.
The tourism elements are involved in almost all service sectors.Therefore, tourism growth can enhance the development of the service sectors and give rise to new tourism services, thus optimising the industrial and factor structure.Specifically, the increase in tourist arrivals promotes the development of tourism-related services, reduces the proportion of traditional agriculture and secondary industries, and increases the proportion of tertiary industries such as services in the national economy, thus promoting industrial structure optimisation and green transformation.In addition, tourism and its related services are mostly labour-intensive and have a significant advantage in absorbing employment.The increase in tourism is bound to increase the employment rate and promote the transfer of production factors to services, thus optimising the factor structure.For example, Wang et al. (2022) found that tourism can effectively promote the optimisation of industrial structure.Furthermore, Inchausti-Sintes (2015) contended that tourism helps to reduce unemployment by improving terms of trade and enhancing capital accumulation.Thus, we sum up the third hypothesis: Hypothesis 3 Tourism affects GTFP through changing the share of the tertiary industry.
Apart from the technological innovation and industrial structure effects, we also consider tourism's possible environmental enhancement effects to influence GTFP.Since green development and low carbon are often accompanied by each other, the change in carbon intensity measures the environmental enhancement effect in this study.Carbon intensity is measured by carbon emissions per unit of economic output, which is consistent with the connotation of GTFP in accommodating economic growth and environmental protection.At the same time, energy input, a factor closely related to carbon intensity, is also considered in the GTFP calculation; therefore, changes in carbon intensity largely determine the level of green development.According to the definition of carbon intensity, the impact of tourism on carbon intensity includes two aspects, namely the effects on economic growth and carbon emissions, as indicated previously.As a result, theoretically, tourism also significantly affects carbon intensity.In sum, we hypothesise: Hypothesis 4 Tourism affects GTFP through changing regional carbon intensity.

Methods for this study
We develop the following econometric model with two-way fixed effects to estimate the nonlinear effects of tourism on GTFP.Usually, control variables are observable and time-varying and regionally heterogeneous, so to consider unobservables and time-fixed factors, we control time fixed effects and region fixed effects to increase the robustness of the model.
where GTFP represents green total factor productivity, tourism is indicated by tourism specialisation.The coefficients b 1 and b 2 are the effects of tourism on GTFP to be examined.con denotes control variables, c i denotes city fixed effects, k t denotes time fixed effects.Referring to the model of Song et al. (2018), the control variables contain foreign investment (FOI), industrial structure (IND), population density (POPD), and population quality (POPQ).To prevent possible omitted variables, we add economic development as an additional control variable.
We measure GTFP referring to Oh and Heshmati (2010).The conventional Malmquist-Luenberger Productivity Index cannot address the bias in productivity calculations caused by technological progress; for this reason, Oh and Heshmati (2010) proposed the Global Malmquist-Luenberger Productivity Index calculated using the SBM directional distance function to yield a more accurate productivity index.Output indicators for measuring GTFP include desired and undesired ones.The desired output indicator is GDP (Unit: ¥10,000), obtained by using the constant price of the year 2005.The undesired output indicators include the volume of industrial waste water discharged (Unit: ton), the volume of industrial soot(dust) emission (Unit: ton), and the volume of industrial sulphur dioxide emission (Unit: ton).A composite pollution index is calculated using the entropy method based on the above three pollutants as the undesired output.Input indicators include capital, labour, and energy.The capital input is expressed as fixed capital stock (Unit: ¥10,000), the annual capital increment is defined as fixed asset investment, and the capital depreciation rate is set at 9.6%.The initial capital stock is determined by dividing the actual total fixed asset investment in the city in 2005 by 10%.The fixed capital stock is also converted to constant price based on the GDP price index by using the constant price of the year 2005.Labour input is expressed as total employment (Unit: 10,000 persons), equal to the sum of persons employed in urban units and private enterprises and self-employed individuals at year-end.The energy input is expressed as electricity consumption, equal to annual electricity consumption (Unit: 10,000 kWh).
Tourism specialisation reflects the relative economic position of tourism and highlights the willingness of local governments to use tourism as a key decision-making tool to achieve green growth objectives, often measured by the ratio of tourism revenue to GDP (Croes et al., 2021;Zhao, 2021), compared to per capita or overall tourism revenue, which is the proxy for tourism in the existing literature.Foreign investment is measured as the ratio of total foreign investment to GDP.According to the Petty-Clark law, industrial optimisation involves gradually reducing the proportion of primary industries while increasing the proportion of tertiary sectors; therefore, the industrial structure is measured as the share of tertiary industries' value added to GDP.Population density is calculated as the number of people per square kilometre.Population quality is measured as the ratio of financial expenditure on education to GDP.Finally, economic development is measured as the real GDP per capita by using the constant price of the year 2005.
Data on GDP and GDP per capita were collected from the China City Statistical Yearbook or provincial statistical yearbooks.Data on industrial waste water discharged, industrial soot(dust) emission, industrial sulphur dioxide emission, fixed asset investment, total employment, the added value of the tertiary industry, and electricity consumption were derived from China City Statistical Yearbook, provincial statistical yearbooks, and city-level statistical yearbooks.Data on tourism revenue were obtained from the city's annual national economic and social development statistical bulletins.Data on foreign investment were obtained from the China City Statistical Yearbook, and the missing data were supplemented by collecting the city's annual national economic and social development statistical bulletins, provincial statistical yearbooks, as well as government reports.Data on financial expenditures on education were obtained from the China City Statistical Yearbook and the city's annual national economic and social development statistical bulletins, and individual regions' data were obtained from public reports of local finance departments or estimates made from them.
After the above data collection, there were still some missing data in some years supplemented by the linear interpolation method.Even so, there were some cities with serious missing data for several variables.Therefore, we removed these cities from the study sample.Thus we obtained panel data for 308 prefecture-level administrative regions from 2005 to 2019.Notably, our sample does not include the four provincial-level municipalities, namely Beijing, Tianjin, Shanghai, and Chongqing.We obtained the GTFP of each city from 2005 to 2019 by cumulative multiplication year by year using the GTFP of 2004 as the base period value of 1.The descriptive statistics for the data of all variables are shown in Table 1.

Benchmark regressions
Classical Ordinary Least Squares (OLS) is used to estimate model (1).Table 2 reports the estimations results.Column 1 reports the regression results when only controlling city and time fixed effects.The coefficients of tourism and its quadratic term are positive and negative, respectively, and both are statistically significant.This suggests that tourism will raise regional GTFP and then decrease it, thus supporting Hypothesis 1. Columns 2 to 5 introduce POPQ and other control variables successively, whilst the regression results do not change essentially compared with column 1.Therefore, there exists an inverted U-shaped relationship between tourism and GTFP.Taking column 6 as an example, the inflection point of tourism specialisation is 0.0717/(0.0252Ã 2)¼1.4226.The economic implication of this result is that tourism specialisation starts to have a negative impact on regional GTFP when it deviates upward from the actual tourism growth by 42.26%, while it has a positive effect on regional GTFP until tourism specialisation approaches 1.4226 times the actual tourism growth from below.

Robustness test
To verify the reliability of the benchmark regression results, we perform the following robustness tests.First, we change the explanatory variable by replacing the ratio of total tourism revenue to GDP with the ratio of total tourist arrivals to the resident population to indicate tourism specialisation (De Vita & Kyaw, 2017).Table 3 reports the regression results for replacing the explanatory variable.Although there is a significant change in the magnitude of the regression coefficients (mainly due to the different magnitudes of the explanatory variables), there is no essential change in the direction of influence.Here, the tourism and its quadratic term coefficients are still significantly positive and negative, respectively.Hence, although the explanatory variable tourism revenue/GDP is replaced with tourist arrivals/population, there is still a significant inverted U-shaped association between tourism and GTFP.Second, we examine the sensitivity of the relationship between tourism and GTFP to the sample.For this purpose, we set different samples to re-estimate the model (1), and the results are shown in Table 4.The new samples cover 2005-2018 and 2006-2019, respectively.When the 2005 or 2019 data are removed, the coefficients of tourism and its quadratic term remain significantly positive and negative, with only a slight change in their magnitude relative to the whole sample.Accordingly, the inverted U-shaped relationship between tourism and GTFP still exists.Furthermore, Table 4 indicates that the change in the sample does not essentially change the nonlinear association between tourism and GTFP.
Finally, we use the instrumental variables approach to address the possible endogeneity of the explanatory variable.In general, the endogeneity of the explanatory variable arises from two aspects.One is the possible omitted variables in the model because it is impossible to control all the factors potentially affecting GTFP.The other is a potential reverse causality, i.e., GTFP may conversely affect tourism.For this reason, we seek suitable instrumental variables for tourism specialisation.Following the principle that the instrumental variables should be correlated with the explanatory variables but not with the residuals, we use provincial tourism and its squared term as the instrumental variables for the city's tourism and its squared term.Generally speaking, local tourism and its related policies are influenced by the development of tourism at the higher level (e.g., the provincial level), while provincial tourism does not have a direct impact on the city's GTFP.Also, we choose the mean of tourism in other cities in the province and its squared term as the instrumental variable for tourism in that city and its squared term.Because there is often horizontal competition among local governments, when the level of tourism development in other cities in the province increases, the local government also tends to develop higher tourism.Moreover, no direct relationship exists between tourism in other cities in the province and the GTFP of the city.We employ the Two-Stage Least Squares for the instrumental variable test, and the results are shown in Table 5. Columns 2 and 3 indicate the effect of provincial tourism specialisation and its squared term on the city's tourism specialisation and its squared term.The results show the positive effect of the former on the latter, which validates the above theoretical analysis.Columns 5 and 6 also indicate that tourism in the city is positively influenced by the tourism in other cities in the province.Therefore, our choice of instrumental variables is reasonable.Furthermore, column 1 shows that when provincial tourism and its squared term are chosen as the instrumental variables, there is also an inverted U-shaped relationship between tourism and GTFP.Such an inverted U-shaped relationship also exists when the mean of tourism in other cities in the province and its squared term are chosen as the instrumental variables (see column 4).

Influence mechanism analysis
We continue to explore the previous hypothetical paths by which tourism affects regional GTFP: the technological innovation effect, the industrial structure optimisation effect, and the environmental enhancement effect.We respectively introduce technological innovation and carbon intensity as control variables in model ( 1), where technological innovation is measured as the number of patents granted, and carbon intensity is expressed as the carbon emissions per 10,000 Yuan of GDP.Data on the number of patents granted were collected from China City Statistical Yearbook and the city's annual national economic and social development statistical bulletins.Data on carbon emissions were derived from Chen et al. (2020).Since the carbon emission data provided by Chen et al. (2020) are only available up to 2017, our sample data only between 2005 and 2017 are used to examine the environmental enhancement effect.The test results for Hypotheses 2 to 4 are shown in Table 6.
Column 1 shows that the coefficient of technological innovation is significantly positive, so technological innovation effectively increases GTFP.Meanwhile, tourism is significantly positive, and its squared term is significantly negative, while the inflection point of tourism specialisation is 1.4302.In other words, tourism contributes to GTFP when the ratio of tourism revenue to GDP does not exceed 1.4302 times the actual tourism specialisation; otherwise, tourism reduces GTFP., When considering technological innovation, the inflection point of tourism specialisation shifts to the right somewhat, relative to the baseline inflection point of 1.4226.We further examine whether regional innovation capacity is affected by tourism, and the results are shown in column 2 in Table 6.
The results show that the coefficients of tourism and its squared term are significantly positive and negative, respectively.That is, there is also an inverted U-shaped relationship between tourism and technological innovation, thus confirming Hypothesis 2. Consequently, the impact of tourism on technological innovation constitutes a channel through which tourism ultimately influences GTFP.As mentioned earlier, the increase in tourism contributes to the concentration of regional talent and investment as well as the technology exchange, thus contributing to improving the regional technological innovation and hence the GTFP.However, when tourism specialisation is too high, various resources flow to the tourism industry with relatively weak technical barriers, which in turn reduces regional attractiveness to talent, capital, and technology, thus reducing innovation performance and ultimately negatively affecting GTFP.
Column 3 in Table 6 replicates the results of column 6 in Table 2 and shows that tertiary sector development can significantly increase GTFP.We further examine whether tourism affects the tertiary industry, and the results are shown in column 4 in Table 6.We found that there is also an inverted U-shaped relationship between tourism and the tertiary industry, thereby substantiating Hypothesis 3. Thus, the nexus of tourism and industrial structure constitutes another influence mechanism by which tourism affects GTFP.We have previously explained that the increase in tourism specialisation helps promote the development of tourism-related services and attract more employment to these industries, thus optimising the industrial structure and ultimately increasing GTFP.However, when tourism specialisation is too high, resources are overly concentrated in tourism and its related industries, leading to an increasingly distorted industrial structure, making the whole economy less productive, which eventually negatively impacts GTFP.
Column 5 indicates that the coefficient of carbon intensity is significantly negative, suggesting that the increase in carbon intensity reduces GTFP.At the same time, tourism and its squared term are significantly positive and negative, respectively, and the inflection point of tourism specialisation is 1.4195.Therefore, when considering the carbon intensity, the inflection point of tourism specialisation will shift to the left to some extent, relative to the baseline case.We also analyse the effects of tourism on carbon intensity.Column 6 shows that the coefficients of tourism and its quadratic term are significantly negative and positive.That is, there exists a U-shaped relationship between tourism and carbon intensity.Therefore, our results support Hypothesis 4. The impact of tourism on carbon intensity also constitutes an intermediate channel by which tourism ultimately influences GTFP.As widely recognised relatively low-carbon sectors, the enhancement of tourism and its related services can effectively reduce the regional carbon intensity, thus improving the regional environmental quality and GTFP.However, as the tourism economy continues to grow, the energy consumption of the tourism industry is increasing, but its contribution to economic development is gradually bottlenecked, which in turn increases the carbon intensity, thus reducing the GTFP.

Regional heterogeneity analysis
Up to now, we have confirmed an inverted U-shaped relationship between tourism and GTFP.We further investigate whether this association differs across different regions.We customarily divide the overall Chinese sample into four subsamples: eastern, central, western, and northeastern.Table 7 reports the results of the regional heterogeneity analysis.Column 1 shows that the coefficients of tourism and its squared term are both significantly positive in eastern China, implying that tourism significantly increases GTFP.This result differs considerably from the inverted U-shaped relationship for the whole sample.Column 2 indicates that in middle China, the coefficients of tourism and its squared term are significantly negative and positive, respectively, suggesting the Ushaped relationship between tourism and GTFP.This finding is also inconsistent with the results for the national sample.Column 3 shows that in western China, there is an inverted U-shaped relationship between tourism and GTFP, which is similar to the conclusion for the whole sample.The difference is that the inflection point of tourism specialisation has shifted to the left compared to the national sample (1.3091 versus 1.4226).In the Northeast, tourism and GTFP are linked in a similar way compared to the Middle.
The results demonstrate the positive linear impact of tourism on GTFP in eastern China which is relatively socio-economically developed with outstanding advantages over the other regions in terms of economic development level, technology level, and ability to attract foreign investment.These benefits are more conducive to exerting the positive role of tourism in talent pooling, innovation, and industrial structure optimisation, thus contributing to green development.In addition, the advantages can effectively offset the positive or negative impact of tourism on the three influence mechanisms, thereby maintaining the positive effect of tourism on GTFP.In general, China's central and northeastern regions have relatively similar socioeconomic development levels and are therefore sometimes classified as one region, such as Zhang and Zhang (2022) and Zhang (2023).Thus our results show similar linkages between tourism and GTFP in these two regions.In these regions, as tourism specialisation increases, its role in technological innovation, industrial structure optimisation, and environmental enhancement gradually emerges and then begins to affect GTEP positively.The curve for the western region is similar to that of the whole country.Tourism initially also contributes to technological innovation, industrial structure optimisation, and environmental enhancement.However, the negative impact of tourism on GTFP in Western China is presented earlier than the whole sample.

Conclusions and discussion
This paper theoretically links tourism and regional green total factor productivity, for which we propose four theoretical hypotheses and test them to address the two basic questions posed in the paper.We theorised that the impact of tourism may not be linear but rather that a decline follows an increase.We tested this theoretical hypothesis based on panel data for 308 prefecture-level cities in China from 2005 to 2019.Consistent with theoretical expectations, the results demonstrate an inverted U-shaped relationship between tourism (measured as the ratio of tourism revenue to GDP) and GTFP.We confirmed the reliability of this linkage through various rigorous robustness tests.This addresses the first theoretical question.Our findings also suggest that tourism affects GTFP through the technological innovation effect, the industrial structure optimisation effect, and the environmental enhancement effect, which reveals the mechanism of the effect of tourism on green total factor productivity and provides a theoretical basis for promoting green growth through tourism development.Besides, the nexus of tourism and GTFP varies significantly across China's regions.The inverted U-shaped relationship exists only in western China, while the U-shaped relationship exists in central and northeastern China and a positive relationship exists in eastern China.These results indicate that economically developed regions with higher-level economic development, technology and ability to attract foreign investment are more conducive to the positive effects of tourism on green development.Therefore, the second theoretical question is addressed.
The research design and findings of the article contribute to the existing literature and suggest the following theoretical implications.Firstly, for the first time, the article examines the impact of tourism economic activities on regional green development performance.GTFP is a good indicator of regional green development performance.Although tourism is difficult to be included as a production factor in general economic growth models, the article confirms the objective role of tourism activities for sustainable regional growth (e.g., green growth).We found the nonlinear effect of tourism on regional green development performance.In the moderate interval, tourism specialisation can increase GTFP, but after exceeding a certain threshold, it has a suppressive effect on GTFP.Therefore, tourism growth is not always green, and dependence on tourism must be kept within a reasonable range.We also encourage more researchers to explore the nonlinear role of tourism on GTFP, thus providing more substantial evidence on the link between the two.
Secondly, the mechanism of tourism's impact on GTFP needs to be widely discussed.We validate several potential channels through which tourism affects GTFP.Tourism growth is conducive to the regional concentration of talent, investment, and technology exchange, as well as the development of the tertiary industries, thus enhancing innovation capacity and optimising industrial structure, but such effects are also nonlinear.Excessive tourism growth can in turn reduce innovation dynamics and capacity.Meanwhile, the overconcentration of resources in tourism and related industries will lead to an increasingly distorted industrial structure, and the crude growth of the relatively low-end tourism service industry will cause a decrease in productivity and the increase in environmental pressure.Accordingly, the tourism-only development concept is not always conducive to the sustainable development of the economy and the environment, and a better coordination between tourism and other industries is needed.Furthermore, in addition to the innovation, structural optimisation, and environmental effects, given the comprehensive nature of tourism and the wide range of its impacts (e.g., the relationship between tourism and the 17 Sustainable Development Goals), more impact mechanisms could be further explored.
Thirdly, our study deepens the understanding of the link between tourism and green development in a developing country with significant internal differences like China.We compared the nexus of tourism with GTFP in the more socio-economically developed eastern region, the less developed western region, and the moderately developed central and northeastern regions.Extensive research has confirmed that tourism impacts are related to the socio-economic characteristics of the destinations (e.g., Zhang & Zhang, 2022;Ehigiamusoe, 2020).We also demonstrate the regional heterogeneity of the impact of tourism on GTFP.In socio-economically developed regions, we only found a linear positive effect of tourism on GTFP, suggesting that tourism increases regional green development performance almost unrestrictedly in a relatively developed socio-economic system.This result is logical because these regions often have their own traditional industrial structures and technological endowments, and tourism development is more of an icing on the cake than a significant driver for the overall regional socio-economic system.However, the inverted U-shaped association between tourism and GTFP still exists in less socio-economically developed regions, and the turning point of tourism specialisation is more to the left than in the overall sample.This suggests that establishing a less tourism-dependent economic system is even more important in these regions.
Finally, an important concern is a precipitous decline in tourism during the COVID-19 pandemic.Although our sample pre-dates the COVID-19 pandemic, our findings suggest that a significant reduction in regional green development performance is implied in COVID-19.During the pandemic or post-pandemic period, in the face of greater economic recovery and growth pressures, it is highly likely that tourism will not be expected for some time due to travel restrictions, but rather that some carbon-intensive manufacturing or other high-emission sectors (even those that were once restricted) will re-emerge.In the foreseeable short term, tourism specialisation will remain at a low level, and its positive impact on GTFP may be significantly reduced, warranting future attention.
Our findings also highlight some key policy implications.At the present stage, the proportion of tourism in the national economy could be continuously increased, so as to give more play to the positive role of tourism in green development, which is particularly important in developed areas.In the central and western areas, there is a need to consider the inflection point of tourism specialisation in promoting green urban development in a context-specific manner.It is recommended to facilitate or delay the inflection point as much as possible through technological innovation, the development of high-end services, and environmental management.In addition, tourism should play a positive role in the flow of talent, capital and technology, and the transfer of innovation factors from developed to less developed regions.At the same time, the quality of tourism should be improved, the traditional model of large-scale growth should be changed, the coordination between tourism and high-end services should be strengthened, the proportion of clean energy in tourism should be increased, and the low-carbon transition of tourism should be promoted.Conclusively, it is proposed to establish an economic and management system that can organically link high-quality tourism, innovation, industrial structure optimisation, and environmental protection.
The article can be enhanced in the following ways.Due to limited data availability, we apply the ratio of tourism revenue to GDP to indicate tourism specialisation.This is also a common current practice, such as Croes et al. (2021) and Zhao (2021); however, strictly speaking, such a comparison is statistically incorrect.In the future, more artificial calculations could be made to obtain data on the value-added of tourism to better capture tourism specialisation in order to assess its impact on GTFP more accurately.Also, other complex theoretical mechanisms may affect the nexus of tourism and GTFP, such as regional development policies and resource endowments.Thus, an integrated analysis of impact mechanisms will yield richer theoretical and practical insights that deserve future attention.Finally, the huge differences in socio-economic development within China mean that a rough division of China into four regions may not be a good way of highlighting regional differences.Therefore, future enhancements could focus on forming more divisions based on strict criteria to enrich the understanding of the regional heterogeneity of tourism's association with GTFP.

Table 1 .
Descriptive statistics for the whole sample.

Table 2 .
The effects of tourism on green total factor productivity.

Table 3 .
The effects of tourism on green total factor productivity: Alternative independent variable.

Table 4 .
The effects of tourism on green total factor productivity: Alternative samples.

Table 6 .
Influence mechanism of the nexus of tourism and green total factor productivity.Robust standard errors are in brackets.Ã , ÃÃ , ÃÃÃ respectively indicate the 10%, 5%, and 1% significance level.
Source: Authors own calculations.

Table 7 .
Heterogeneity of the effects of tourism on green total factor productivity.Robust standard errors are in brackets.Ã , ÃÃ , ÃÃÃ respectively indicate the 10%, 5%, and 1% significance level.
Source: Authors own calculations.