Escape from air pollution: How does air quality in the place of residence shape tourism consumption?

Abstract

Air quality has been demonstrated to be an important determinant influencing tourist decision making. In this paper, we investigate the effect of air quality in the place of origin on a resident’s tourism consumption and the moderating effect of household income on this relationship. We develop a conceptual framework to rationalize the effect and empirically examine it at the household level using China Labor-force Dynamics Survey data from 2016. The results show that poor ambient air quality increases tourism consumption. Furthermore, this relationship becomes stronger for the higher-income group: One standard deviation increase in the Air Quality Index would increase tourism expenditure as much as a 22.56% increase in family income. The methods of instrumental variable and functional-coefficient regression were employed for robustness analysis. These findings contribute to the literature by providing a new perspective for tourism demand studies and direct implications for tourism management and policy making.

Keywords

push effect air quality residential place moderating effect tourism consumption

Introduction

The trade-off between economic development and the environment is always a central issue, especially for developing countries. While economic growth benefits the local country, the development model behind it might also lead to a rapid deterioration of the environment. China is a prime example of this paradox; although it has achieved great success in economic development, air pollution has become poorer (Chan and Yao, 2008). China’s air quality ranks second to last according to Yale University’s report (EPI, 2016). Almost half of the Chinese population lives in an area with polluted air beyond the highest hazard threshold proposed by the United States Environmental Protection Agency (Zhang et al., 2017b). Since severe air pollution damages residents’ physical and psychological health (Chen et al., 2018; Li et al., 2014; Zhang et al., 2017a, 2017b, 2018), the central government of China has decided to incorporate environmental performance into official assessments to improve air quality by launching a nationwide air quality monitoring program since 2013 (Zhang, 2021). Barwick et al. (2019) found that this disclosure of air quality significantly increased residents’ awareness of pollution and triggered household behavioral changes.

To avoid exposure to pollution, people could take reactions such as defense and evasion (Chen and Chen, 2020; Deschênes et al., 2017; Freeman et al., 2019; Liu and Yu, 2020; Xue et al., 2021). On the one hand, in the face of air pollution, people invest more in dust masks, air purifiers, and healthcare products (Chen and Chen, 2020; Deschênes et al., 2017; Rodrigues et al., 2021). On the other hand, people tend to stay far away from polluted areas. Xue et al. (2021) found that air pollution causes a local brain drain. Freeman et al. (2019) used a hedonic estimation model and found that the more severe the air pollution in one’s birthplace, the higher the proportion of the population leaving their hometown. Guo et al. (2022) and Liu and Yu (2020) further investigated this issue and added evidence to the strong relationship between air pollution and high eviction rates.

Interestingly, the above literature on the strategy of evasion has focused on relocation, while focusing less on temporary escaping behaviors. Travel provides an opportunity to temporarily escape the pollution of the residential area (Ma et al., 2020). Will travel become a means of escape? The existing literature provides evidence of escaping via tourism. Rodrigues et al. (2021) found that air pollution has already been considered by the majority of travelers and that the behavioral intention of escaping from pollution ranked second when pollution occurs. Dong et al. (2019) also found an interesting phenomenon: Travelers from polluted cities tended to spend more. This adjustment to consumption patterns may stem from the intention to avoid pollution exposure (Barwick et al., 2019).

The first attempt to test whether air pollution in the origin area encourages residents to travel was found in Wang et al.’s research (2018). They first investigated the push effect of origin air quality on China’s outbound tourism motivation, measured by the number of outbound travel products purchased. Their fundamental work provides a new perspective to understand the determinants of tourists’ decision making, while some limitations still remain to be addressed. First, Wang et al. (2018) used macro data on the number of outbound tourists aggregated to the city level, in which the aggregate data lost information about heterogeneities in tourist behavior. In addition, overseas travel is expensive and often considered luxury (Lim, 1997), so their conclusion only reflects the travel behavior of residents in the upper tail of income distribution. As a good supplement, the data we used come from a nationwide sample survey microdatabase, which is more representative of all income levels. Additionally, we used the tourism expenditure amount to reflect tourism demand from another perspective. It is worth mentioning that the information about moving contained in our data allows us to deal with possible endogenous problems such as “environmental migration” and adds credibility to the conclusions, which is another contribution to prior literature.

Furthermore, we employed novel methods to support theory with facts. A conceptual framework of the utility maximization analysis is established, which provides a solid theoretical basis for the “escape” behavior of tourists, compared to the prior research of Wang et al. (2018) that only provides empirical analysis. To reinforce the conceptual analysis, we empirically supplement an additional test in which perceived pollution is employed to explain travelers’ travel behaviors in the face of residential pollution. Moreover, a novel estimation method, functional-coefficient estimation, was introduced when analyzing the moderating effect of income. The method is a type of partial linear estimation that helps find the potential threshold of the push effect of ambient air pollution.

More specifically, we summarize our contribution as follows. First, a conceptual framework is established using the analysis tools of indifference curves and maximization of utility to describe how the push factors in the push and pull framework help make consumption decisions when tourists face air pollution in their residences. Next, we examine the push effect using the tourism demand model and find that income influences this relationship as a moderator. Finally, we propose a potential channel; that is, ambient air quality affects consumption through people’s perceptions of pollution.

The rest of this paper is organized as follows. The next section reviews the literature regarding the push and pull framework, the determinants of tourism consumption, the push effect of origin air quality, and the moderating effect of income. The third section presents a simple conceptual framework for developing our research hypotheses. The fourth section introduces data and descriptive statistics, followed by empirical analyses in the fifth and sixth sections. The seventh section provides additional tests to explore the channel of perceived pollution. Finally, this study concludes in the last section by revisiting the findings and elaborating on the managerial implications and theoretical contributions.

Literature review

Push and pull framework

Motivation has become the starting point for tourism studies since tourism motivation originates from certain needs and desires of travelers and guides the formation of tourism behavior (Vasant and Kalaivanthan, 2017). Push and pull theory provides a unified analysis framework for tourism motivation analysis (Bayih and Singh, 2020; Crompton, 1979; Dann, 1981). Crompton (1979) was the first to divide tourism motivation into push and pull factors. Push factors are generated by internal needs, such as anomie, self-enhancement, relaxation, and the promotion of social interaction. In contrast, the pulling factors come from the outside, such as novelty and education at the destination (Mihalič, 2002). Push and pull factors are two different decisions that focus on whether to travel and where to travel, respectively (Klenosky, 2002), and thus push factors are antecedents of pull factors (Dann, 1977).

Natural resources are an integral part of tourism, and are a key factor that directly affects the motivation of tourists (Beerli and Martın, 2004; Gómez Martín, 2005; Lancaster, 1966; Luo and Deng, 2008). As a typical natural resource throughout the entire journey, air quality has an obvious spatial distribution relationship with push and pull factors: the clean air in scenic spots has a pull effect to attract tourists (Becken et al., 2017; Dong et al., 2019; Qiao et al., 2021; Xu et al., 2020), while air pollution in residential areas pushes residents to escape to scenic spots (Wang et al., 2018). It should be noted that the influence of air quality may be the opposite after spatial segmentation, which suggests that it is necessary to distinguish the complex influences behind the same environmental factor.

Most scholars agree that clean air at attractions can draw tourists (Becken et al., 2017; Dong et al., 2019; Qiao et al., 2021; Xu et al., 2020). In other words, air pollution at attractions creates a negative pull and reduces the number of tourists. Xu et al. (2020) found that deteriorating air quality at destinations significantly reduced the number of inbound tourists, and Becken et al. (2017) drew similar conclusions through an online panel survey. In addition to the number of travelers, air quality at the destination also affects tourism consumption. Dong et al. (2019) found polluted air in scenic spots was linked with an increase in tourism consumption. By hedonic analysis, Qiao et al. (2021) noted that fresh air drove up house prices through a premium effect in the research on high-end rural homestay rooms.

Unlike the focus on air quality at the destination mentioned in the above paragraph, the impact of residential air quality has been ignored. At present, the first and only literature to fill this gap is Wang et al. (2018). We set out to expand their research in the aspects of research object, conceptual analysis, and rigorous empirical analysis.

Determinants of tourism consumption

The influencing factors of tourism consumption are closely related to the nature of tourism (Park et al., 2020). On the one hand, the determinants of tourism consumption should be classified data or micro-level data (Lin et al., 2015; Song et al., 2019), since macroscopic aggregated data will lose valid information related to tourism product types and personal traits (Laesser and Crouch, 2006; Wang and Davidson, 2010). On the other hand, just as travel behavior is complex (Dellaert et al., 2014), the motivation for travel is also complex, which requires a comprehensive analysis of geographical, social, political, and environmental factors.

First, sociodemographic factors are frequently mentioned. Based on consumer behavior theory, income is considered one of the most relevant determinants of tourism consumption (Marrocu et al., 2015). Yang et al. (2014) found that income was the main factor affecting Chinese tourism consumption, while relative income only affected tourism consumption in a few regions. Other scholars affirm the impact of income expectation (Kim et al., 2012) and illiquid wealth (Zhang and Feng, 2018). Life cycle theory supports the idea that travel needs are closely related to the age of travelers. Chen and Shoemaker (2014) pointed out that although changes in age are difficult to affect travel preferences, differences in physiological conditions can still lead to distinctions in travel behavior. Moreover, some scholars have found that other factors, such as gender, nationality, occupation, and education level, also affect tourism consumption (McGehee and Loker-Murphy, 1996; Park et al., 2020; Tan and Ooi, 2018; Thrane, 2014).

Some scholars dedicate themselves to finding the critical determinants of travel to broaden the study perspective of tourism consumption. Channels for booking travel products, number of destinations, travel purpose, travel size, accommodation type, travel duration, and other factors have been evaluated as demand determinants (Gómez-Déniz et al., 2019; Kim et al., 2012; Park et al., 2020; Thrane and Farstad, 2012; Yoon and Shafer, 1997). As a component of travel, air quality also plays an important role in tourism (Beerli and Martın, 2004; Gómez Martín, 2005; Luo and Deng, 2008). Air quality in scenic spots has been extensively studied as a tourism feature (Becken et al., 2017; Xu et al., 2020). On the contrary, the important role of residential air quality has just been noticed (Rodrigues et al., 2021; Wang et al., 2018), and our research in this area can help expand the factors of tourism consumption.

Since multidimensionality is the key proposition to comprehend the travel planning process (Jeng and Fesenmaier, 2002), taking into account the sociodemographic factors and travel-related factors comprehensively is helpful to better understand the motivation of demand (Choi et al., 2012).

The push effect of origin air quality on tourism consumption

As is well known, low air quality damages residents’ physical and psychological health (Chen et al., 2018; Li et al., 2014; Zhang et al., 2017a, 2017b, 2018). These mental and physical damages increase residents’ perception of risk (Li et al., 2014), and produce an escape motivation from annoying environments to seek self-recovery (Freeman et al., 2019; Ryan and Glendon, 1998; Trauer and Ryan, 2005). This need to “escape” is an essential source of push factors for travel (Dann, 1981). Travel provides an opportunity to move into a place far from home (Ma et al., 2020), so residents may choose to travel as one of the ways to “escape.” Rodrigues et al. (2021) found some evidence about escaping to travel: For travelers, the behavioral intention to escape from pollution ranked second among all reasons answered.

A possible influence channel on travel consumption is inspired by hedonic pricing. For hedonic tourism analyses, situational attributes such as climate have become necessary objects (White and Mulligan, 2002). Qiao et al. (2021) noted that fresher air drove up house prices through a premium effect in research on high-end rural homestay rooms. This premium is not isolated from the air quality of the origin area, since travel behavior is closely related to the difference in air quality between attractions and departure points (Chen et al., 2021). The fluctuation of origin air quality leads to perceived distinctions in intercity air, which helps tourists arrive at different consumption levels (Bayih and Singh, 2020; Becken et al., 2017; Li et al., 2016). When residential areas are polluted, clean air at the destination will bring more hedonic consumption, leading to high scenic pricing, which also implies a high willingness to pay on the demand side (Atkinson and Halvorsen, 1984). Additionally, Dong et al. (2019) found that travelers from polluted cities tended to spend more while on vacation. Overall, origin air quality influences travel decisions, such as whether to go or how much to spend.

The moderating effect of income

Shaw and Loomis (2008) determined that tourism consumption is decided by a series of allocation decisions within the income budget. In other words, income should be seen as an economic constraint on tourism consumption, clarifying the necessity of considering income while assessing other determinants of tourism consumption (Marrocu et al., 2015; Park et al., 2020). Naturally, we assume that the push effect generated by air quality in the place of origin varies among different income-tier families. It seems plausible that high-income families are likelier to be pushed to spend more on tourism.

First, tourism products are rich in income demand elasticity (Fuleky et al., 2014; Nelson et al., 2011), and their consumption depends on available time and income (Todd, 2001). Low-income families may have weaker responsiveness to low air quality or may even be unresponsive due to the constraints of low-income and limited leisure. This means that the marginal impact of the original air quality is small for them. In contrast, high-income people have almost no insurmountable travel thresholds (Wang et al., 2018). They tend to “escape” from their residence (Xue et al., 2021) and “compensate” for their health and emotional losses by tourism consumption more actively.

In addition, for purposes of self-reinforcement and defense, the high-income group increased their willingness to pay for travel. Tourism consumption is a symbol of social status (Dimanche and Samdahl, 1994). In the face of pollution, high-income people tend to increase their travel time and purchase luxury and other visible means for conspicuous consumption (Park et al., 2010). This kind of purchase behavior helps the rich enhance their ego, gain recognition, and demonstrate their high social status (Crompton, 1979). In addition, when faced with air pollution, the high-income tier invests more in air purifiers and healthcare products for defensive purposes (Chen and Chen, 2020; Deschênes et al., 2017).

In summary, high-income families may be more active in avoiding pollution and seeking self-reinforcement and defense, so they have a greater probability of being pushed by ambient air pollution. This viewpoint corresponds to the source of tourism motivation, “escape and seeking,” pointed out by Iso-Ahola (1982) in optimal arousal theory.

Conceptual framework and research hypotheses

Although we could propose the effect of air pollution on tourism consumption and the moderating effect of income according to the literature, we would like to introduce a conceptual framework of the push effect to develop our research hypotheses in this section.

Existing literature points out that air pollution reduces residents’ happiness (Li et al., 2014; Zhang et al., 2017a, 2017b, 2022), and deepens depression (Wei et al., 2020, 2022), anxiety (Vert et al., 2017), and mental health risks (Theron et al., 2022). These negative psychological effects imply a reduction in utility (Clark et al., 2008). Considering the heterogeneity of residents’ evaluation of air pollution (Sun et al., 2019; Zhang et al., 2017a), we add the perception of air pollution as a variable in the resident’s utility function. Specifically, we assume that the utility function of the represented family is a continuously differentiable function expressed as follows

U = U (P_{e}, T, C)

(1)

where

P_{e}

describes the perceived pollution of a family faced with ambient air pollution,

T

denotes the tourism consumption, and

C

represents the consumption of a numeraire good. We assume

\frac{\partial U}{\partial P_{e}} < 0

\frac{\partial U}{\partial T} > 0

, and

\frac{\partial U}{\partial C} > 0

; in other words, household perception of air pollution leads to negative utility, while the consumption of tourism and the numeraire good increases utility. In addition, we further assume diminishing marginal utility in tourism consumption and the numeraire good, that is,

\frac{\partial^{2} U}{\partial T^{2}} < 0

and

\frac{\partial^{2} U}{\partial C^{2}} < 0

The perceived pollution is assumed to be a continuously differentiable function of actual air pollution (P), tourism consumption, and family income (I)

P_{e} = f (P, I, T)

(2)

The high availability of pollution information renders residents to be easily aware of the air quality of their residence (Barwick et al., 2019). In addition to the rich influx of information, Zhang et al. (2018) observed that air pollution affects cognition through physical channels. Thus, we assume deteriorating air quality in residential areas will increase a household’s perception of pollution, that is, $\frac{\partial P_{e}}{\partial P} > 0$ . This inference is further tested in the seventh section.

It is worth noting that adding tourism consumption $(T)$ into function (2) is a key assumption for the push effect of air pollution. When residents travel, they spend less time in the residence, which means they will be less exposed to the air pollution in the origin area and thus have a lower perception level of pollution $(P_{e})$ . We express this assumption as $\frac{\partial P_{e}}{\partial T} < 0$ .

In addition, Maslow’s hierarchy of needs theory states that people will only develop higher-level needs when lower-level needs are met. Sirgy (1986) pointed out that people in less-developed countries care less about quality of life, implying that the low-income tier may be less sensitive to air quality. Thus, we assume that given the actual air pollution, the richer the family is, the more pollution they perceive, that is, $\frac{\partial P_{e}}{\partial I} > 0$ .

Based on the utility function (1), we obtain the marginal utility of tourism consumption as follows

M U_{T} = \frac{\partial U}{\partial T} + \frac{\partial U}{\partial P} \frac{\partial P_{e}}{\partial T}

(3)

The first term on the right-hand side of the above equation is the direct effect of tourism consumption. The second term is the indirect effect generated by perceived pollution, which is positive since $\frac{\partial U}{\partial P_{e}} < 0$ and $\frac{\partial P_{e}}{\partial T} < 0$ . It is easy to derive that $M U_{T}$ increases when the perceived pollution $(P_{e})$ rises with the air pollution (P). We depict this idea in Figure 1. When the actual air pollution increases from $P_{1}$ to $P_{2}$ , the indifference curve would rotate upward since the marginal effect of tourism consumption would rise relatively.

Figure 1.

An illustration of how air quality shapes the utility of family.

To further analyze consumption decisions, we first consider the following optimization problem

\begin{array}{l} \max U (P_{e}, T, C) \\ s . t . ξ T + C \leq I \end{array}

(4)

where

ξ

represents the price of tourism. The first-order condition for a family’s utility maximization problem is

\frac{M U_{T}}{ξ} = M U_{c}

,¹ which can be easily rearranged into the form of the marginal rate of substitution as follows

| M R S | = \frac{M U_{T}}{M U_{c}} = ξ

(5)

where

M R S

is the marginal rate of substitution between tourism and the numeraire good and

| \cdot |

means to take the absolute value. Suppose

(T_{1}, C)

is the optimal combination given the air pollution

P_{1}

and household income

I_{1}

. When the ambient pollution becomes poorer from

P_{1}

P_{2}

, we would have the relationship

| M R S (T_{1}, C_{1}) | > ξ

since

M U_{T}

increases and

M U_{C}

remained unchanged. The broken equilibrium

(| M R S (T_{1}, C_{1}) | \neq ξ)

implies the bundle

(T_{1}, C_{1})

is no longer the optimal decision. Thus, the family has to change consumption behavior to reach a new equilibrium. To be specific, the family should increase the tourism consumption to lower

M U_{T}

while cutting the consumption of the numeraire good to improve

M U_{c}

. Then, a new optimal combination

(T_{2}, C_{2})

meets the first-order condition and we have the relationship

T_{2} > T_{1}

and

C_{2} < C_{1}

. The analysis of the change in tourism consumption when the air pollution in the residence occurs is pictured in Figure 2 and gives rise to the first research hypothesis.

Figure 2.

An illustration of the push effect of air pollution.

H1: Air pollution in the place of origin increases the household’s tourism consumption.

To further explore the moderating effect of income, we turn to how tourism consumption would change by comparing income level $I_{1}$ with $I_{2}$ under a fixed ambient air quality.

Even if air pollution remains unchanged, the rise in income would increase tourism consumption, as it enlarges the budget set. Because this direct impact is not relevant to this study, we removed it from the analyses. We use Figure 3 for the decomposition. For a given air pollution level $P_{2}$ , we hold the perception of air pollution constant $P_{e} = {\bar{P}}_{e}$ , which will be relaxed in subsequent models. In this case, the shape of the indifference curve of utility does not change, and $| M R S (T_{3}, C_{3} | I_{2,} P_{2}) | = ξ$ . When the budget line shifts from $I_{1}$ to $I_{2}$ , the tourism consumption would increase from $T_{2}$ to $T_{3}$ . Thus, the direct effect is formulated as $T_{3} - T_{2}$ .

Figure 3.

An illustration of the moderating effect of income.

In the analysis of H1, we assume the inequality $\frac{\partial P_{e}}{\partial I} > 0$ . Thus, we let the perceived pollution ( $P_{e}$ ) increase with household income (I). Given the pollution level $P_{2}$ , we can obtain $| M R S (T_{3}, C_{3} | I_{2,} P_{2}) | > ξ$ as $\frac{\partial U}{\partial P_{e}} \frac{\partial P_{e}}{\partial T} (T, C) | I = I_{1} < \frac{\partial U}{\partial P_{e}} \frac{\partial P_{e}}{\partial T} (T, C) | I = I_{2}$ for any consumption bundle. The indifference curve of utility rotates upward, as Figure 3 shows, and tourism consumption rises from $T_{3}$ to $T_{4}$ . It can be concluded that when compared to the family with an income level $I_{1}$ , the family with an income level $I_{2}$ would spend $T_{4} - T_{3}$ more on travel in response to air pollution. Thus, we propose our second research hypothesis.

H2: Household income has a significantly positive moderating effect on the relationship between tourism consumption and air pollution in the place of origin, that is, the relationship will be stronger for households with higher income.

Data

Tourism expenditure and sociodemographic characteristics

We use data from the China Labor-force Dynamic Survey (CLDS), which is a comprehensive database conducted by Sun Yat-sen University in 2011, 2012, 2014, and 2016. The questionnaire records details at the individual, family, and community levels and asks respondents questions using a time frame of 1 year prior to the survey year. For example, the 2011 survey looked at the information of respondents in 2010.

The China National Environmental Monitoring Centre has published data on the AQI index since the end of October 2013. We matched the 2016 CLDS survey data with the air quality data in 2015, specifically examining the answers from the 2016 CLDS family questionnaire F4.25.11 (“Let’s ask about the consumption on travel vacation of your whole family in 2015?”) as the data source of family tourism expenditure. Among the individual, family, and community datasets, F4.25.11 is a unique variable describing travel behavior. In addition, all the householder and household information examined in the descriptive statistics (see Table 1) was also derived from the 2016 CLDS family questionnaire. After the data cleaning process described in Appendix A, we obtained 13,312 observations (93.58% of the raw data).

Table 1.

The descriptive statistics.

Variable	Sample size	Max	Min	Mean	Standard deviation
Household travel expenditure	13,312	200,000	0	1098	5703
AQI	13,312	195.8	35.71	80.87	24.97
PM2.5	13,312	116.4	16.01	52.41	19.11
PM10	13,312	305.4	32.65	88.25	35.97
Family income	13,312	3,600,000	0	53,761	90,770
Householder age	13,312	107	18	53.84	13.73
Family size	13,312	19	1	4.222	2.169
Retirement status	13,312	1 (retire)	0 (others)	0.136	0.342
Householder gender	13,312	1 (male)	0 (female)	0.844	0.363
Householder registration	13,312	1 (rural)	0 (others)	0.693	0.461

Air quality

Air quality is the independent variable of interest in this study. It comes from the China National Environmental Monitoring Centre. The primary air quality variables considered in this study included AQI, PM2.5, and PM10. The Air Quality Index (AQI) is a comprehensive index calculated based on the concentration of air pollutants, such as SO₂, NO₂, particulate matter (including PM2.5 and PM10), CO, and O₃. The larger the AQI values, the poorer the air quality. PM2.5 and PM10 report the concentrations of each particulate. Prior research has shown that AQI, PM2.5, and PM10 collectively lead to a large number of social welfare losses through both physiological and psychological channels (Chen et al., 2018; Li et al., 2014; Zhang et al., 2017a, 2017b, 2018).

Since the travel expenditure reported in the CLDS is the total value in 2015, this study calculates the mean value of the air quality indicators in 2015 to match the CLDS data. Due to the lack of air quality data in Ulanqab City, we substituted the missing data with the reading of its nearest city, Datong. Ultimately, we included 155 cities (including all cities in the raw data) located in the 29 provinces.

Table 1 shows the descriptive statistics for the continuous variables and categorical variables. There is a large difference in tourism expenditure between the samples, and 13.72% of the samples have no tourism expenditure. For AQI, PM2.5, and PM10, the average values of air pollution are all above 50, failing to meet the pollution-free standard (according to the Ambient Air Quality Standard²). In addition, the samples have a wide range of incomes, covering both rural and urban areas. Householders are mostly male, over 50 years old, and not retired. The size of the typical family is small, usually consisting of 4–5 members. Table B1 in Appendix B presents the correlation coefficients between the variables, implying no need to be overly concerned about multicollinearity.

Table B1.

The correlation coefficients of variables

	IHS(T)	AQI	PM2.5	PM10	ln (income)	Age	Famsize	Gender	Retirement	Registration
IHS(T)	1
AQI	0.0100	1
PM2.5	0.0120	0.975	1
PM10	−0.0130	0.974	0.922	1
ln (income)	0.212	−0.039	−0.038	−0.049	1
Age	−0.123	0.021	0.033	0.00700	−0.209	1
Famsize	−0.080	−0.124	−0.096	−0.109	0.073	0.071	1
Retire	−0.052	−0.00700	−0.00300	0.00500	0.045	−0.092	0.153	1
Gender	0.110	0.018	0.0110	−0.00900	0.116	0.392	−0.112	−0.153	1
Residence	−0.292	−0.044	−0.029	−0.0130	−0.219	−0.051	0.212	0.203	−0.458	1

Empirical analysis of the main effect

Methodology of the main effect

To examine Hypothesis 1 on the main effect for the push effect of air pollution on tourism consumption, we apply the following linear Model I as equation (6)

I H S (T_{i j}) = α + β P_{j} + H^{'} γ + F^{'} λ + δ_{p r} + ε_{i j}

(6)

where

T_{i j}

is the tourism expenditures of family i in city j, and

I H S (T_{i j})

represents the inverse hyperbolic sine form of

T_{i j}

during 2015. The inverse hyperbolic sine function

I H S (T_{i j}) = I n (T_{i j} + {(1 + T_{i j}^{2})}^{\frac{1}{2}})

is approximate to the logarithmic form but with good adaptability to zero values (Barreca et al., 2017), and thus is more applicable to our sample. After this formal transformation, the marginal effects can be interpreted as percentage changes (Fang et al., 2021). The key independent variable

P_{j}

is the air quality in city j, which is represented by three indicators: AQI, PM2.5, and PM10. Considering the CLDS family survey questionnaires, we selected the following control variables: householder characteristics and family characteristics.

H_{i j}^{'}

represents the householder characteristics variables, including gender, age, and whether retired.

F_{i j}^{'}

represents the family characteristics, including family size, householder registration type (rural or urban), and the logarithmic form of total family income.

H_{i j}^{'}

and

F_{i j}^{'}

represent the sociodemographic characteristics of families surveyed.

δ_{p r}

denotes the province-level fixed effects.

ε_{i j}

is a random error term.

Results for the push effect of air pollution

The regression results are shown in Table 2. The coefficients of AQI, PM2.5, and PM10 are estimated to be 0.00860, 0.00966, and 0.00304, respectively, and all are statistically significant. This implies that the increase in all three indicators contributes positively to family tourism expenditures, initially confirming Hypothesis 1. In other words, as the severity of ambient air pollution increases, families tend to increase tourism expenditures. Take the AQI indicator as an example. If the AQI increases by one standard deviation (SD) (i.e., 24.97), family tourism expenditure will increase by 21.47%. Although the average increase in tourism expenditure for a household is only about 234.74 CNY, its economic impact size should not be underestimated. Take Guangzhou, for example: it has about seven million households, resulting in a roughly 1.65 billion CNY (≈254 million USD) increase in tourism market size. Air pollution in the origin area can drive such large consumption to shift into tourism, which is enough to raise the attention of tourism managers.

Table 2.

Results for the main effect.

	(1)	(2)	(3)
AQI	0.00860^*** (0.00216)
PM2.5		0.00966^*** (0.00260)
PM10			0.00304^** (0.00136)
Householder characteristics	Yes	Yes	Yes
Family characteristics	Yes	Yes	Yes
Province fixed effect	Yes	Yes	Yes
Number of observations	13,312	13,312	13,312

Notes: ***p < 0.01, **p < 0.05, *p < 0.1, Robust standard errors are reported in parentheses.

Interestingly, Column (3) shows that the coefficients of PM10 are the smallest among the three pollution indexes. In comparison, households are more sensitive to the concentration of PM2.5 but relatively insensitive to PM10. These results echo the findings of Zhang et al. (2017b), which show that the economic and statistical significance of PM10 on mental health is smaller than that of PM2.5. This may result from residents’ insensitivity to PM10, which was examined in additional analyses described in the Additional Tests section. PM10 contains all particles with a diameter of 10 microns or less. In medical terms, the larger the particle size, the lower the proportion of carcinogenic compounds it has (Sosa et al., 2017). PM10 is much less likely to invade the most sensitive area of the respiratory system when compared with PM2.5. The larger diameter particles contained in PM10 are less toxic to humans than PM2.5, so residents are less aware of its danger and thus are less pushed. As for the public’s perception, the media, which is a vital way to influence residents’ perceptions, also pays less attention to PM10 than PM2.5 (Zhai and Cheng, 2020).

Robustness checks for the push effect

According to Ambient Air Quality Standards, we count the days with an AQI > 100 (PM2.5 > 100; PM10 > 100) in 2015 as polluted days, which is another measure of origin air quality. The air quality indicators in Model I are replaced with the number of polluted days, and the results are shown in Table 3. The same conclusion on the push effect as with the baseline models can be drawn. All coefficients of polluted days were positive and significant.

Table 3.

Robustness check with polluted days.

	(1)	(2)	(3)
Polluted days of AQI	0.00308^*** (0.000866)
Polluted days of PM2.5		0.00557^*** (0.00137)
Polluted days of PM10			0.00130^* (0.000782)
Householder characteristics	Yes	Yes	Yes
Family characteristics	Yes	Yes	Yes
Province fixed effect	Yes	Yes	Yes
Number of observations	13,312	13,312	13,312

Notes: ***p < 0.01, **p < 0.05, *p < 0.1; Robust standard errors are reported in parentheses.

A potential estimation issue is also considered. When faced with pollution, residents will take reactions such as defense and evasion to avoid exposure to contaminants (Chen and Chen, 2020; Deschênes et al., 2017; Freeman et al., 2019; Liu and Yu, 2020; Xue et al., 2021). “Environmental migration” is one type of escape behavior that means residents choose to move because of the pollution of the place where they used to live (Chen et al., 2017; Warner, 2010). Freeman et al. (2019) found evidence that the more severe the air pollution in the birthplace, the higher the share of the population leaving their birthplace. This “location sorting” based on the preference for clean air may directly affect tourism consumption and air quality at the current residence, which could be the most vital endogenous source in this study (Zhang et al., 2017b). To the best of our knowledge, previous push effect studies have not considered this issue.

We addressed the concern about this possible endogeneity issue in three ways. First, we eliminated the floating population from our sample to ensure correspondence between the interviewed place and the permanent residence area and circumvent unexpected migration. Liu and Yu (2020) found that residential air pollution leads to the loss of a floating population, and our estimate excludes the potential confusion about this effect. We follow the CLDS questionnaire to define the floating population, which is to screen the population through question F1.5.1 (“Where is the household address of householder?”). We consider the observations with the answer “outside the neighborhood committee of the interviewed town or districts” as the floating population (which is a method officially proposed by the CLDS to distinguish the floating population) and eliminate these families from our sample. Table 4 reports the results after eliminating the floating population. In Columns (1) to (3), the coefficients of air pollution are all positive and statistically significant. Consequently, the push effect stays robust even considering the issue of “environmental migration.”

Table 4.

Robustness check with the non-floating family sample.

	(1)	(2)	(3)
AQI	0.00674^*** (0.00222)
PM2.5		0.00735^*** (0.00266)
PM10			0.00262^* (0.00140)
Householder characteristics	Yes	Yes	Yes
Family characteristics	Yes	Yes	Yes
Province fixed effect	Yes	Yes	Yes
Number of observations	11,777	11,777	11,777

Notes: ***p < 0.01, **p < 0.05, *p < 0.1, Robust standard errors are reported in parentheses.

Another method follows the data processing method proposed by Deschenes et al. (2020). The literature indicates that air pollution in China has become poorer since 1990, following economic growth (Chan and Yao, 2008). Thus, the migration caused by dirty air should happen mainly after 1990. We retained families with non-migrated householders since 1990 in the sample. In addition, we match the 2016 CLDS household dataset with the individual dataset since only the individual dataset includes the accurate year of residence migration. The results shown in Table 5 are consistent with our main results.

Table 5.

Robustness check with the non-migrated family sample.

	(1)	(2)	(3)
AQI	0.00873^*** (0.00289)
PM2.5		0.00961^*** (0.00348)
PM10			0.00382^** (0.00179)
Householder characteristics	Yes	Yes	Yes
Family characteristics	Yes	Yes	Yes
Province fixed effect	Yes	Yes	Yes
Number of observations	6971	6971	6971

Notes: ***p < 0.01, **p < 0.05, *p < 0.1, Robust standard errors are reported in parentheses.

We also introduce instrumental variable methods to address the endogeneity issue. Following the work of Zhang et al. (2020), annual average wind speed was employed as an instrumental variable (IV). Chen et al. (2020) explained that a higher wind speed would lower PM2.5 levels through the blowing-off effect. Meanwhile, Becken (2013) found that wind speed has no direct impact on scenic flights. Therefore, Chen et al. (2020) and Becken (2013) together implied that wind speed satisfies the condition of IV. The two-stage least square (2SLS) method was employed, and the results displayed in Table 6 further support our previous findings.

Table 6.

Results of IV estimation.

	(1)	(2)	(3)
Panel A: First-stage estimation (relation between wind speed and air pollution)
Windspeed	−0.249^*** (0.0120)	−0.226^*** (0.0106)	−0.287^*** (0.0168)
F-statistics	430.34	451.15	290.45
Panel B: Second-stage estimation (effects of air pollution on tourism consumption)
AQI	0.0419^*** (0.0134)
PM2.5		0.0462^*** (0.0147)
PM10			0.0364^*** (0.0118)
Householder characteristics	Yes	Yes	Yes
Family characteristics	Yes	Yes	Yes
Province fixed effect	Yes	Yes	Yes
Number of observations	13,312	13,312	13,312

Notes: ***p < 0.01, **p < 0.05, *p < 0.1, Robust standard errors are reported in parentheses.

Empirical analysis of the moderating effect

The above results show the vital role of ambient air quality in tourism consumption. This section considers whether families with different incomes will adopt different travel behaviors under the push of ambient air quality, that is, Hypothesis 2 on the moderating effect of household income level.

Regression analysis with different income groups

To examine whether income has a moderating effect, we first considered a simple method. More specifically, we divide the sample into five groups equidistantly at 20% intervals according to household income: the lower-income group, low-income group, middle-income group, high-income group, and higher-income group. This classification is consistent with the Chinese Bureau of Statistics’ classification of five income categories.

We implemented the estimation of Model I for each group, respectively. Table 7 reports the group regression results. None of the air pollution variables were significant across all groups except for the high-income group. Hence, there is no evidence that air quality deterioration statistically significantly impacts the low-income tier. However, when the income level increases, the push effect becomes statistically significant. Consequently, Hypothesis 2 is supported. In addition, the relatively insignificant and small income coefficient of low-income groups indicates that their tourism decisions are subject to economic constraints.

Table 7.

Results of regression analysis by different income groups.

	(1)	(2)	(3)	(4)	(5)
	Lower-income	Low-income	Middle-income	High-income	Higher-income
AQI	0.00223 (0.00232)	0.00137 (0.00275)	0.00532 (0.00525)	0.00941 (0.0100)	0.0325^*** (0.0109)
PM2.5	0.00302 (0.00267)	0.00150 (0.00307)	0.00465 (0.00603)	0.00816 (0.0111)	0.0332^*** (0.0126)
PM10	0.000738 (0.00152)	0.000629 (0.00185)	0.00256 (0.00349)	0.00456 (0.00740)	0.0154^** (0.00756)
Householder characteristics	Yes	Yes	Yes	Yes	Yes
Family characteristics	Yes	Yes	Yes	Yes	Yes
Province fixed effect	Yes	Yes	Yes	Yes	Yes
Number of observations	2929	3414	2424	2004	2541

Notes: ***p < 0.01, **p < 0.05, *p < 0.1, Robust standard errors are reported in parentheses.

The moderating effect has implications for tourism practitioners and socio-economic development. On the one hand, the push effect suggests that it is worth advocating for tourism practitioners to strengthen tourism promotion in tourism-generating regions with high levels of pollution. However, after accounting for moderating effects, this form of marketing seems to be profitable only when targeted at high-income groups, and a crude, widespread advertisement may not be economical or effective. On the other hand, when considering the high-income group as the main force of tourism, the tourism spending saved by the high-income group after the regulation of polluted air may be large. Once redistributed to the local economy, the money saved will contribute to the prosperity of the local economy.

These results initially assert the existence of the moderating effect of income. However, the regression analysis by subdivided groups also significantly reduces the sample size, and thus, loses the common information across groups. In addition, the coefficients of the different groups cannot be directly compared. Therefore, we included the interaction terms of income group dummies and air quality in Model I. The results are reported in Table 8. To intuitively reflect the magnitude of the moderating effects, we drew a coefficient graph in Figure 4.

Table 8.

Results of regression analysis with interaction terms.

	(1) AQI	(2) PM2.5	(3) PM10
Lower-income × pollutant	0.00615^*** (0.00205)	0.00530^** (0.00246)	0.00243^* (0.00128)
Low-income × pollutant	0.00390^* (0.00210)	0.00138 (0.00252)	0.000220 (0.00138)
Middle-income × pollutant	0.00674^*** (0.00223)	0.00551^** (0.00275)	0.00236 (0.00153)
High-income × pollutant	0.0132^*** (0.00251)	0.0151^*** (0.00323)	0.00806^*** (0.00186)
Higher-income × pollutant	0.0325^*** (0.00263)	0.0442^*** (0.00344)	0.0250^*** (0.00199)
Householder characteristics	Yes	Yes	Yes
Family characteristics	Yes	Yes	Yes
Province fixed effect	Yes	Yes	Yes
Number of observations	13,312	13,312	13,312

Notes: ***p < 0.01, **p < 0.05, ^*p < 0.1, Robust standard errors are reported in parentheses.

Figure 4.

The coefficient graph of interaction terms regression.

According to the results, the coefficient remains statistically insignificant initially in the low-income tier. This becomes statistically significant when the income level increases. In addition, the income change equivalent to the one SD pollutant increase is calculated as follows:

Δ I n I n c o m e = \frac{β_{p o l l u t a n t}}{β_{I n (I n c o m e)} \times S D_{p o l l u \tan t}}

. Taking AQI as an example, Table 9 gives an estimation of the economic effect. As income increases, the economic significance of the marginal impact of push increases. For the higher-income group, one SD increase in the AQI would increase tourism expenditure as much as a 22.56% increase in family income.

Table 9.

The economic significance in different income groups.

	(1) lower-income	(2) low-income	(3) middle-income	(4) high-income	(5) higher-income
$β_{A Q I}$	0.00615^*** (0.00205)	0.00390^* (0.00210)	0.00674^*** (0.00223)	0.0132^*** (0.00251)	0.0325^*** (0.00263)
$β_{I n (I n c o m e)}$			0.0360^*** (0.0128)
$Δ I n c o m e$	4.272^** (2.093)	2.706 (1.851)	4.681^* (2.463)	9.157^** (4.006)	22.56^*** (8.617)

Notes: According to Table 1, the standard deviation of AQI is 24.97. ^***p < 0.01, ^**p < 0.05, ^*p < 0.1, the stand errors of $Δ I n I n c o m e$ is recalculated using delta method.

Functional-coefficient regression

Although we have employed two methods to examine the heterogeneous impact of ambient air pollution on family travel consumption at different income levels, this method still has shortcomings. For example, grouping is subjective, and the coefficient of the interaction term jumps at the threshold. It is difficult to explain the jump due to the lack of a theoretical foundation. Besides, Thrane (2014) emphasized that it is necessary to examine nonlinear tourism expenditure models when taking economic theory and econometric practices into consideration. Prior literature has tried to employ a quantile regression methodology in tourism demand studies to improve estimated efficiency (Almeida and Garrod, 2017; Santos and Vieira, 2012; Sharma et al., 2020). However, the quantile regression could only provide discontinuity estimation. Therefore, we further adopted the functional-coefficient regression model to obtain a continuous nonlinear estimation for a robustness check.

The functional-coefficient regression model carries out a nonparametric estimation, and regards the coefficient of the air quality variable as an unknown function of income, which better solves the problems of sample splitting. The model treats all samples as a whole and allows heterogeneity among the samples. Therefore, its estimation has the characteristics of objectivity and continuity, with no jumping, which overcomes the shortcomings of grouping methods.

We built a functional-coefficient model to estimate the nonlinear relationship of the moderating effect. A slight change for Model I is implemented to achieve the functional-coefficient form transformation, substituting the coefficients of the key independent variables with a function of income $G (r i n c)$ . The model is expressed as follows

I H S (T_{i j}) = α + G (r i n c) P_{j} + H_{i j}^{'} γ + F_{i j}^{'} λ + δ_{p r} + ε_{i j}

(7)

where

r i n c

is defined as the ratio of household income to the highest household income in the sample:

r i n c = \frac{i n c o m e}{\max (i n c o m e)}

The results of the functional-coefficient regression are shown in Figure 5.³ The solid line in the figure represents the marginal thrust for households with different incomes, and the shaded part represents the 95% confidence interval.

Figure 5.

Functional-coefficient regression.

First, the upward trend of the curve shows that the push effect increases rapidly with household income. Moreover, the functional-coefficient regression model allows us to find the insensitive income threshold of the push effect. The results of all three pollution indicators show that families with $r i n c$ of 0.123 (equivalent to an annual income of 49,200 CNY) or more will be likelier to increase tourism consumption.

The slow decline at the end of the curve may be due to the large base of tourism budgets of the extra-high-income group. These families tend to purchase high-priced, high-quality tourism products, so there is little room for further expenditure. At the same time, the extra-high-income group has the ability to create situations with high indoor air quality, which may also weaken the push effect of outdoor air quality. The specific reasons remain to be further analyzed. Interestingly, this part of the change is consistent with the findings of Wang et al. (2018) using overseas travel data.

Another point of concern in the functional-coefficient model is that the lower bound of the confidence interval for low-income people is below zero, which means that the increase in surrounding air pollution could reduce the low-income tier group’s spending on travel. This anomaly may be explained that incentives to escape pollution may be satisfied at a relatively inferior destination.

What needs to be emphasized is that the marginal moderating effect of the extremely high-income group is still more statistically significant than that of the low-income group, which is consistent with the group regression results. Accordingly, the results of the functional-coefficient regression are consistent with our previous results and confirm the robust moderating effect of income. We also implemented a robustness analysis using the non-migrated family sample and the instrumental variable method, obtaining similar results which are shown in Figure C1 and Figure C2 of Appendix C, respectively.

Figure C1.

Results of the functional-coefficient models with the non-migrated family sample.

Figure C2.

Results of the functional-coefficient models with the instrumental variable.

Additional Tests

In the conceptual framework, we introduce the idea that families realize the change in ambient air quality, and thus, families are pushed to consume more travel. In this section, we collect data for a family’s perceived pollution of the origin area to test the channel of perceived pollution $(P_{e})$ . The 2016 CLDS household survey asked each family to answer the question, “How serious do you think the air pollution problems are in the area where you live?” on a scale from one (very serious) to four (nothing serious). We replaced the explained variable in Model I with the perceived pollution $(P_{e})$ . Then, we use ordered logit regression and OLS regression to test whether the family is sensitive to air pollution in the living area. The results of the grouping regression are shown in Figure 6.

Figure 6.

The coefficient graph of household sensitivity to air pollution.

Figure 6 not only demonstrates that a family’s perception of air pollution is significantly related to pollution indicators, but also implies that the relationship is moderated by income. Figure 6 shows that all the residents were aware of the pollution, and the high-income tiers perceived air pollution at higher levels compared to the perception for low-income tiers. The channels of perceived pollution proposed in the conceptual framework have been supported.

Discussion and conclusions

It is of central interest to the tourism industry to understand the determinants of tourism consumption. Based on the push and pull framework, we collected household-level data from the CLDS conducted in 2016 and air quality data from the China National Environmental Monitoring Centre and performed empirical analyses in this study. The results yielded two main findings. First, air pollution, measured by the AQI, PM2.5, and PM10, exerts a push effect on residents’ tourism consumption; that is, tourists tend to spend more on tourism when air pollution in the place of origin gets poorer. Second, this relationship is positively moderated by household income. The findings are robust across different model specifications.

New findings were revealed in this study. Wang et al. (2018) argued that as disposable income increases, the thrust of residential air pollution weakens, while this conclusion could only be applied to a sample with relatively high incomes. Using a sample with a wide distribution of income, we find that as household income increases, residential air pollution drives households to spend more on travel. More specifically, when air pollution occurs, high-income households are likelier to increase their travel budgets, but this push effect becomes insignificant in low-income families.

This study provides several theoretical contributions. First, while prior literature mainly pays attention to the relationship between air quality at the destination and tourism consumption, few studies have explored the effect of air quality in the place of origin on tourism consumption. Considering that tourists’ intentions to escape pollution and residents’ awareness of pollution are becoming stronger, the push effect of residential air pollution should draw wider attention from researchers and policymakers. Wang et al.’s (2018) first attempt to answer the importance of this push was limited in scope, which has been improved in this study. Our dataset comprises micro-level data from national surveys, which means that our conclusions can be generalized to a wider range of customers. Moreover, our study not only identifies the push effect but also explores the underlying channel. Although the existing literature on the push effect of air quality in the place of origin infers the channel according to related literature, the channel has not yet been examined to the best of the authors’ knowledge. Employing CLDS data, this study is the first to explore the proposed underlying channel.

Our results yield direct managerial implications for travel agents and the government. First, marketing should be carried out to target high pollution source areas since the thrust generated by air pollution in tourist sources has increased residents’ consumption in tourism. Our preliminary calculations found that the increased currency value for Guangzhou alone was as high as 254 million USD. This is very encouraging for the tourism industry. It seems profitable to strengthen advertising or set more branches wherever air quality is inferior to route consumers to scenic spots. Our results also provide guidance for seasonal marketing. Considering seasonal air pollution, it would be better to strengthen the publicity of clean air in scenic spots in autumn and winter when air pollution is serious. Moreover, differentiated marketing could be implemented based on the sensitive nature of the push effect. According to the moderating effect of income, high-income earners living in the same level of air pollution are more willing to pay for clean air. Furthermore, since air-contaminated areas are widespread, especially in developing countries (Greenstone and Hanna, 2014; Tanaka, 2015), our conclusions could be generalized to the inhabitants in these places, which should have a considerable impact.

In addition, even for non-scenic areas, our conclusions show benefits to the economy. Because local clean air may reduce the residents’ travel expenditures to other places, the subsequent saved funds would be likely to flow to the local area, which will undoubtedly promote the development of the local economy. This channel has not been considered in previous environmental governance decisions. With reference to the economic scale of the impact on the tourism industry mentioned earlier, once this positive effect is included in the cost efficiency analysis, environmental governance will see more benefits in implementation.

This research also inevitably bears several limitations. First, due to the limited questionnaire setting, our study could not obtain more travel and destination information. A remaining topic for future research is to examine whether the push effect of the origin air has an impact on the choice of destination type. Second, household-level unobservable heterogeneity might exist since we cannot incorporate household-level fixed effects due to the data limits. However, we have attempted to address this issue by including province fixed effects, householder characteristics, and family characteristics in the model. We believe that the majority of household-level heterogeneity has been addressed by adding these controls.

Footnotes

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by the National Natural Science Foundation of China (Grant Nos. 72074184, 72171200, 71701139).

ORCID iDs

Kerui Du

Bin Fang

Rob Law

Notes

Author biographies

Luojia Wang is a graduate student at School of Management, Xiamen University. Her research interest is environmental economics.

Kerui Du, PhD, is an associate professor at School of Management, Xiamen University. His research interest is applied econometrics.

Bin Fang, PhD, is an associate professor in the School of Management, Xiamen University.

Rob Law, PhD, is a Professor in the School of Hotel & Tourism Management at The Hong Kong Polytechnic University.

Appendix A. Data cleaning

First, we deleted 347 observations with missing travel expenditure and dropped 14 observations with travel expenditure larger than total household expenditure, since these observations may suffer from the possible reporting error issue. Second, we deleted 118 observations with missing household income and dropped seven observations with a total household income (after cleaning) less than zero. Third, the family with the highest income is considered a possible outlier since its income in 2015 is 144 million yuan CNY, while the family’s annual expenditure is 30,000 yuan CNY. We believe this observation is incorrect and consequently deleted it. Finally, we deleted 141 observations with missing city information.

Appendix B. The correlation coefficients of variables

Appendix C. Robustness analysis of the functional-coefficient models

References

Almeida

Garrod

(2017) Insights from analysing tourist expenditure using quantile regression. Tourism Economics 23(5): 1138–1145.

Atkinson

Halvorsen

(1984) A new hedonic technique for estimating attribute demand: an application to the demand for automobile fuel efficiency. The Review of Economics and Statistics 66(3): 417.

Barreca

Neidell

Sanders

(2017) Long-Run Pollution Exposure and Adult Mortality: Evidence from the Acid Rain Program (No. W23524). Cambridge, MA: National Bureau of Economic Research.

Barwick

Lin

, et al. (2019) From Fog to Smog: The Value of Pollution Information (No. W26541). Cambridge, MA: National Bureau of Economic Research.

Bayih

Singh

(2020) Modeling domestic tourism: motivations, satisfaction and tourist behavioral intentions. Heliyon 6(9): e04839.

Becken

(2013) Measuring the effect of weather on tourism. Journal of Travel Research 52(2): 156–167.

Becken

Jin

Zhang

, et al. (2017) Urban air pollution in China: destination image and risk perceptions. Journal of Sustainable Tourism 25(1): 130–147.

Beerli

Martın

(2004) Tourists’ characteristics and the perceived image of tourist destinations: a quantitative analysis-a case study of Lanzarote, Spain. Tourism Management 25(5): 623–636.

Chan

Yao

(2008) Air pollution in mega cities in China. Atmospheric Environment 42(1): 1–42.

10.

Chen

(2020) Air pollution and avoidance behavior: a perspective from the demand for medical insurance. Journal of Cleaner Production 259: 120970.

11.

Chen

Shoemaker

(2014) Age and cohort effects: the American senior tourism market. Annals of Tourism Research 48: 58–75.

12.

Chen

Lei

, et al. (2021) Chasing clean air: pollution-induced travels in China. Journal of the Association of Environmental and Resource Economists 8(1): 59–89.

13.

Chen

Guo

Huang

(2018) Air pollution, student health, and school absences: evidence from China. Journal of Environmental Economics and Management 92: 465–497.

14.

Chen

Oliva

Zhang

(2017) The Effect of Air Pollution on Migration: Evidence from China (No. W24036). Cambridge, MA: National Bureau of Economic Research.

15.

Chen

Zhao

, et al. (2020) Influence of meteorological conditions on PM2.5 concentrations across China: a review of methodology and mechanism. Environment International 139: 105558.

16.

Choi

Lehto

Morrison

, et al. (2012) Structure of travel planning processes and information use patterns. Journal of Travel Research 51(1): 26–40.

17.

Clark

Frijters

Shields

(2008) Relative income, happiness, and utility: an explanation for the easterlin paradox and other puzzles. Journal of Economic Literature 46(1): 95–144.

18.

Crompton

(1979) Motivations for pleasure vacation. Annals of Tourism Research 6(4): 408–424.

19.

Dann

GMS

(1977) Anomie, ego-enhancement and tourism. Annals of Tourism Research 4(4): 184–194.

20.

Dann

GMS

(1981) Tourist motivation an appraisal. Annals of Tourism Research 8(2): 187–219.

21.

Dellaert

BGC

Arentze

Horeni

(2014) Tourists’ mental representations of complex travel decision problems. Journal of Travel Research 53(1): 3–11.

22.

Deschênes

Greenstone

Shapiro

(2017) Defensive investments and the demand for air quality: evidence from the NOx budget program. American Economic Review 107(10): 2958–2989.

23.

Deschenes

Wang

, et al. (2020) The effect of air pollution on body weight and obesity: evidence from China. Journal of Development Economics 145: 102461.

24.

Dimanche

Samdahl

(1994) Leisure as symbolic consumption: a conceptualization and prospectus for future research. Leisure Sciences 16(2): 119–129.

25.

Dong

Wong

(2019) Estimating the impact of air pollution on inbound tourism in China: an analysis based on regression discontinuity design. Sustainability 11(6): 1682.

26.

Zhang

Zhou

(2020) Fitting partially linear functional-coefficient panel-data models with stata. The Stata Journal: Promoting Communications on Statistics and Stata 20(4): 976–998.

27.

Fang

Lei

Lin

, et al. (2021) Family Companionship and Elderly Suicide: Evidence from the Chinese Lunar New Year (No. W28566). Cambridge, MA: National Bureau of Economic Research.

28.

Freeman

Liang

Song

, et al. (2019) Willingness to pay for clean air in China. Journal of Environmental Economics and Management 94: 188–216.

29.

Fuleky

Zhao

Bonham

(2014) Estimating demand elasticities in non-stationary panels: the case of Hawaii tourism. Annals of Tourism Research 44: 131–142.

30.

Gómez Martín

(2005) Weather, climate and tourism a geographical perspective. Annals of Tourism Research 32(3): 571–591.

31.

Gómez-Déniz

Pérez-Rodríguez

Boza-Chirino

(2019) Modelling tourist expenditure at origin and destination. Tourism Economics 26(3): 437–460.

32.

Greenstone

Hanna

(2014) Environmental regulations, air and water pollution, and infant mortality in india. American Economic Review 104(10): 3038–3072.

33.

Guo

Wang

Zhang

, et al. (2022) Environmental migration effects of air pollution: micro-level evidence from China. Environmental Pollution 292: 118263.

34.

EPI (2016) Environmental Performance Index. New Haven, CT: Yale University. Available: www.epi.yale.edu

35.

Iso-Ahola

(1982) Toward a social psychological theory of tourism motivation: a rejoinder. Annals of Tourism Research 9(2): 256–262.

36.

Jeng

Fesenmaier

(2002) Conceptualizing the travel decision-making hierarchy: a review of recent developments. Tourism Analysis 7(1): 15–32.

37.

Kim

H-B

Park

J-H

Lee

, et al. (2012) Do expectations of future wealth increase outbound tourism? evidence from Korea. Tourism Management 33(5): 1141–1147.

38.

Klenosky

(2002) The “pull” of tourism destinations: a means-end investigation. Journal of Travel Research 40(4): 396–403.

39.

Laesser

Crouch

(2006) Segmenting markets by travel expenditure patterns: the case of international visitors to Australia. Journal of Travel Research 44(4): 397–406.

40.

Lancaster

(1966) A new approach to consumer theory. Journal of Political Economy 74(2): 132–157.

41.

Pearce

Morrison

, et al. (2016) Up in smoke? the impact of smog on risk perception and satisfaction of international tourists in Beijing. International Journal of Tourism Research 18(4): 373–386.

42.

Folmer

Xue

(2014) To what extent does air pollution affect happiness? the case of the jinchuan mining area, China. Ecological Economics 99: 88–99.

43.

Lim

(1997) Review of international tourism demand models. Annals of Tourism Research 24(4): 835–849.

44.

Lin

Mao

Song

(2015) Tourism expenditure patterns in China. Annals of Tourism Research 54: 100–117.

45.

Liu

(2020) Stay or leave? the role of air pollution in Urban migration choices. Ecological Economics 177: 106780.

46.

Luo

Deng

(2008) The new environmental paradigm and nature-based tourism motivation. Journal of Travel Research 46(4): 392–402.

47.

Craig

Feng

(2020) The camping climate index (CCI): the development, validation, and application of a camping-sector tourism climate index. Tourism Management 80: 104105.

48.

Marrocu

Paci

Zara

(2015) Micro-economic determinants of tourist expenditure: a quantile regression approach. Tourism Management 50: 13–30.

49.

McGehee

Loker-Murphy

(1996) The Australian international pleasure travel market: motivations from a gendered perspective. The Journal of Tourism Studies 7(1): 45–57.

50.

Mihalič

(2002) Tourism and Economic Development Issues. Clevedon: Channel View Publications, pp. 81–111.

51.

Nelson

Dickey

Smith

(2011) Estimating time series and cross section tourism demand models: mainland United States to Hawaii data. Tourism Management 32(1): 28–38.

52.

Park

K-S

Reisinger

Noh

E-H

(2010) Luxury shopping in tourism. International Journal of Tourism Research 12(2): 164–178.

53.

Park

Woo

Nicolau

(2020) Determinant factors of tourist expenses. Journal of Travel Research 59(2): 267–280.

54.

Qiao

H-H

Wang

C-H

Chen

M-H

, et al. (2021) Hedonic price analysis for high-end rural homestay room rates. Journal of Hospitality and Tourism Management 49: 1–11.

55.

Rodrigues

Carneiro

Eusébio

, et al. (2021) How important is air quality in travel decision-making? Journal of Outdoor Recreation and Tourism 35: 100380.

56.

Ryan

Glendon

(1998) Application of leisure motivation scale to tourism. Annals of Tourism Research 25(1): 169–184.

57.

Santos

Vieira

(2012) An analysis of visitors’ expenditures in a tourist destination: OLS, quantile regression and instrumental variable estimators. Tourism Economics 18(3): 555–576.

58.

Sharma

Woodward

Grillini

(2020) Unconditional quantile regression analysis of UK inbound tourist expenditures. Economics Letters 186: 108857.

59.

Shaw

Loomis

(2008) Frameworks for analyzing the economic effects of climate change on outdoor recreation. Climate Research 36: 259–269.

60.

Sirgy

(1986) A quality-of-life theory derived from maslow’s developmental perspective. American Journal of Economics and Sociology 45(3): 329–342.

61.

Song

Qiu

RTR

Park

(2019) A review of research on tourism demand forecasting: launching the annals of tourism research curated collection on tourism demand forecasting. Annals of Tourism Research 75: 338–362.

62.

Sosa

Porta

Colman Lerner

, et al. (2017) Human health risk due to variations in PM 10 -PM 2.5 and associated PAHs levels. Atmospheric Environment 160: 27–35.

63.

Sun

Zhang

J-H

Wang

, et al. (2019) Escape or stay? effects of haze pollution on domestic travel: comparative analysis of different regions in China. Science of The Total Environment 690: 151–157.

64.

Tan

GW-H

Ooi

K-B

(2018) Gender and age: do they really moderate mobile tourism shopping behavior Telematics and Informatics 35(6): 1617–1642.

65.

Tanaka

(2015) Environmental regulations on air pollution in China and their impact on infant mortality. Journal of Health Economics 42: 90–103.

66.

Theron

Abreu-Villaça

Augusto-Oliveira

, et al. (2022) A systematic review of the mental health risks and resilience among pollution-exposed adolescents. Journal of Psychiatric Research 146: 55–66.

67.

Thrane

(2014) Modelling micro-level tourism expenditure: recommendations on the choice of independent variables, functional form and estimation technique. Tourism Economics 20(1): 51–60.

68.

Thrane

Farstad

(2012) Tourists’ Length of Stay: The Case of International Summer Visitors to Norway. Tourism Economics 18(5): 1069–1082.

69.

Todd

(2001) Self-concept: a tourism application. Journal of Consumer Behaviour 1(2): 184–196.

70.

Trauer

Ryan

(2005) Destination image, romance and place experience-an application of intimacy theory in tourism. Tourism Management 26(4): 481–491.

71.

Vasant

Kalaivanthan

(2017) Handbook of Research on Holistic Optimization Techniques in the Hospitality, Tourism, and Travel Industry. Hershey, PA: IGI Global.

72.

Vert

Sánchez-Benavides

Martínez

, et al. (2017) Effect of long-term exposure to air pollution on anxiety and depression in adults: a cross-sectional study. International Journal of Hygiene and Environmental Health 220(6): 1074–1080.

73.

Wang

Fang

Law

(2018) Effect of air quality in the place of origin on outbound tourism demand: disposable income as a moderator. Tourism Management 68: 152–161.

74.

Wang

Davidson

MCG

(2010) Chinese holiday makers’ expenditure: implications for marketing and management. Journal of Hospitality Marketing & Management 19(4): 373–396.

75.

Warner

(2010) Global environmental change and migration: governance challenges. Global Environmental Change 20(3): 402–413.

76.

Wei

Qian

, et al. (2020) Association between short-term exposure to ambient air pollution and hospital visits for depression in China. Science of The Total Environment 724: 138207.

77.

Wei

Zhang

, et al. (2022) Long-term exposure to ambient air pollution and incidence of depression: a population-based cohort study in China. Science of The Total Environment 804: 149986.

78.

White

Mulligan

(2002) Hedonic estimates of lodging rates in the four corners region. The Professional Geographer 54(4): 533–543.

79.

Huang

Hou

, et al. (2020) The spatial spillover effects of haze pollution on inbound tourism: evidence from mid-eastern China. Tourism Geographies 22(1): 83–104.

80.

Xue

Zhang

Zhao

(2021) Brain drain: the impact of air pollution on firm performance. Journal of Environmental Economics and Management 110: 102546.

81.

Yang

Liu

Z-H

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

82.

Yoon

Shafer

(1997) Research: an analysis of sun-spot destination resort market segments: all-inclusive package versus independent travel arrangements. Journal of Hospitality & Tourism Research 21(1): 141–159.

83.

Zhai

Cheng

(2020) A long short-term memory approach to predicting air quality based on social media data. Atmospheric Environment 237: 117411.

84.

Zhang

Feng

(2018) More wealth, less leisure? effect of housing wealth on tourism expenditure in China. Tourism Economics 24(5): 526–540.

85.

Zhang

Ren

, et al. (2022) The impact of air pollution on individual subjective well-being: evidence from China. Journal of Cleaner Production 336: 130413.

86.

Zhang

Ren

Zhang

, et al. (2020) Air pollution and tourism development: an interplay. Annals of Tourism Research 85: 103032.

87.

Zhang

(2021) Target interactions and target aspiration level adaptation: how do government leaders tackle the “environment‐economy” nexus? Public Administration Review 81(2): 220–230.

88.

Zhang

Chen

Zhang

(2018) The impact of exposure to air pollution on cognitive performance. Proceedings of the National Academy of Sciences 115(37): 9193–9197.

89.

Zhang

Chen

(2017a) Happiness in the air: how does a dirty sky affect mental health and subjective well-being? Journal of Environmental Economics and Management 85: 81–94.

90.

Zhang

Chen

(2017b) Valuing air quality using happiness data: the case of China. Ecological Economics 137: 29–36.

91.

Zhang

Zhou

(2021) Partially linear functional-coefficient dynamic panel data models: sieve estimation and specification testing. Econometric Reviews 40(10): 983–1006.