key: cord-1056045-z6k3lek6
authors: Qiu, Richard T.R.; Park, Jinah; Li, Shina; Song, Haiyan
title: Social costs of tourism during the COVID-19 pandemic
date: 2020-07-07
journal: Ann Tour Res
DOI: 10.1016/j.annals.2020.102994
sha: 22afde6aae5f796b5eda9bf6cb9413303b52724c
doc_id: 1056045
cord_uid: z6k3lek6

The severe acute respiratory syndrome coronavirus 2 is currently spreading across the world at an alarming rate, resulting in the coronavirus disease 2019 pandemic. Amidst this crisis, tourism scholars are directing their attention to communities at tourist destinations, looking at their safety and well-being and the costs that they will bear due to the cessation of tourism activity. This article describes residents' perceptions of the risks posed by tourism activity, and estimates their willingness to pay to reduce public health risks based on hypothetical scenarios, using the triple-bounded dichotomous choice contingent valuation method. The social costs in three urban destinations are assessed and compared. Based on the findings, suggestions are made for appropriate post-pandemic recovery actions by local authorities and tourism organizations.

Close attention has been paid in tourism studies to the actual and perceived socio-economic contributions of tourism to destination communities (Lindberg & Johnson, 1997) . Inbound tourism has a profound impact on a society, and alongside its positive effects it can interfere with the social and economic well-being of residents in tourist destinations (Jordan, Moran, & Godwyll, 2019) . The negative impacts of inbound tourism may be exacerbated in times of crisis and disaster.

Currently, the world is facing such a crisis in the form of the novel coronavirus disease 2019 pandemic, which has spread to 206 countries or territories (WHO, April 3, 2020).

As COVID-19 swept across the world, travel restrictions and border shutdowns were introduced in many countries and regions to curb its spread (Al Jazeera, 2020). Richter (2003) suggested that the emergence or re-emergence of infectious diseases is one of the corollaries of global tourism and mobility. Urbanization and globalization drive the rapid spread of the virus (Hilsenrath, 2020) , but tourism clearly plays a role in exacerbating the resulting public health crises. It is thus critical to identify and quantify the perceived risks and social costs of tourism J o u r n a l P r e -p r o o f during the COVID-19 pandemic to minimize the negative effects of tourism on destination cities and regions.

While a number of studies have investigated the effects of crises (including natural disasters) on tourism (Aliperti et al., 2019; Cró & Martins, 2017; Kuo, Chen, Tseng, Ju, & Huang, 2008; Sio-Chong & So, 2020; Song, Livat, & Ye, 2019; Wang, 2009) , less attention has been paid to how tourism can amplify crisis events and have negative effects on the public interest and wellbeing of residents and stakeholder groups at tourist destinations (Ritchie, 2008) . As Scott and Laws (2006) indicated, the impact of a crisis should be viewed as occurring among the interconnected system of businesses and other stakeholders that make up a tourist destination.

Furthermore, the key stakeholders in responding to a crisis in a tourist destination belong to both the public and private sectors (Ritchie, 2008) . This network perspective is crucial, as a crisis may reshape the network at a destination site: in particular, new alliances may be forged between stakeholders seeking to minimize or avoid the negative effects of a crisis (Scott & Laws, 2006) . Some empirical studies have examined the social costs and benefits of tourism for destination communities (Lindberg, Andersson, & Dellaert, 2001; Torre & Scarborough, 2017) , and it has been found that the costs and benefits may be heterogeneous across stakeholders (Mayer, 2014) .

However, few studies have quantified the social costs of tourism for destination residents during a crisis. With the concepts of risk perception and crisis management having mainly grown out of from consumer behavior studies, tourism researchers have tended to approach the research problem from the tourists' perspective. Accordingly, they have analyzed the effects of crises on tourism businesses or destinations, while largely ignoring the perspectives of the residents of destination communities (Sharifpour, Walters, & Ritchie, 2014; Song et al., 2019) .

In the present study, the social costs borne by residents of tourist destinations amid the COVID-19 pandemic were estimated using the contingent valuation method (CVM). Specifically, local residents' willingness to pay (WTP) to mitigate the risk of COVID-19 was estimated, with a view to minimizing the negative effects or social costs of tourism during the pandemic. Given that inbound tourism increases the spread of COVID-19 and adds to the pressure on the local healthcare system, destination residents are the most affected. Therefore, the study was conducted based on residents' perceptions of risk. When a pandemic occurs, tourists tend to avoid traveling to affected destinations; meanwhile, destination residents try to minimize the risks associated with tourists who do choose to travel and thus potentially spread the virus. The J o u r n a l P r e -p r o o f latter scenario was examined in this study to determine how the residents of tourist destinations perceive the risk of tourism during a pandemic and measure how willing they are to sacrifice the economic benefits to avoid its social costs.

To achieve the above research objective, three Chinese urban destinations-Wuhan, Guangzhou, and Hong Kong-were selected to estimate the social costs of tourism during the COVID-19 pandemic. Wuhan (with 50,007 confirmed cases) was the city in which COVID-19 first emerged in China and was thus the most heavily affected. Guangzhou (451 confirmed cases)

is the capital of Guangdong province (the second most affected province in China) and, like Wuhan, is a major transportation hub. Hong Kong (803 confirmed cases) shares a border with mainland China, which is the source for more than 80% of its inbound tourism; Hong Kong thus needed to adopt strong measures to mitigate huge public health risks from inbound tourists during the COVID-19 outbreak.

To assess the negative social impacts of tourism amidst a pandemic, three research questions were posed. First, how do residents perceive the negative impact of tourism during a pandemic?

Second, are residents willing to pay for the reduction of risk of COVID-19 increased by tourism (and, if so, by how much), and does the WTP vary by city and demographics? Third, what are the social costs of tourism due to a pandemic in the three destinations under study?

The study makes the following theoretical and practical contributions. At the conceptual and methodological levels, the findings contribute to the increasing body of literature on crisis management and the social impacts of tourism. Specifically, the important yet under-researched topic of the social costs borne by residents of tourist destinations during a pandemic is highlighted. A triple-bounded dichotomous choice contingent valuation method (DCCVM) was used to model and quantify residents' WTP to reduce the risks associated with tourism. Importantly, this model can minimize potential biases associated with CVMs. To the best of our knowledge, this is one of the first studies to use a triple-bounded DCCVM-based model, as a double-bounded model has been the standard approach in the previous literature. The social costs of tourism during the COVID-19 pandemic were also quantified and compared between the three study destinations. At a practical level, the study's findings can be used by local authorities and tourism organizations to develop policy responses to crises and design appropriate recovery strategies. The findings could also underpin regulatory and intervention measures for local authorities in destination cities and regions to prevent losses of residents during or after a crisis.

The remained of this paper is organized as follows. The second and third sections identify the research questions via a literature review. The fourth section describes the research methods and the research design of the WTP survey with hypothetical scenarios. The potential determinants of WTP are also discussed. Descriptive statistics of the three cities are also presented in this section.

The fifth section discusses the empirical results and the last section concludes the study.

Tourism has direct, indirect and induced impacts on a destination economy and society (Khan, Seng, & Cheong, 1990) , and the social and economic costs and benefits of these impacts vary between stakeholders (Mayer, 2014) . However, most research effort has been focused on evaluating the economic costs and benefits of tourism (Lindberg & Johnson, 1997) , with the societal costs and benefits having only been discussed at the conceptual level (Haralambopoulos & Pizam, 1996; Liu & Var, 1986) . Recently, some research has begun into various social costs and benefits related to tourism, focusing on public services and traffic congestion related to public infrastructure (Garau-Vadell, Gutierrez-Taño, & Diaz-Armas, 2018). A growing body of literature is documenting the "touristification" and gentrification risks induced by Airbnb-style rentals, and the discontent and resistance of urban residents who are concerned with the socioenvironmental costs of tourism (González-Pérez, 2020; Gurran, Zhang, & Shrestha, 2020; Smith, Sziva, & Olt, 2019) . In addition, researchers exploring the economic impact of tourism have recently begun to conduct analyses of the societal costs and benefits of tourism for destination communities (Torre & Scarborough, 2017) and environments (Bella, 2018) .

It is possible to measure the societal impacts of tourism, such as the increasing costs of real estate, crime (Smith & Krannich, 1998) , availability of housing and land, cultural and heritage representation, employment in the tourism and hospitality industry (Stephenson & Ali-Knight, 2010) , overcrowding and social carrying capacity beyond economic and ecological assessments (Gonzalez, Coromina, & Galí, 2018) , and harm to individual residents' well-being (Jordan et al., 2019) . However, obtaining such measurements is often difficult, as societal impacts are usually indirect (Milman & Pizam, 1988 ) and chronic (Liu & Var, 1986) , and perceptions of these impacts vary depending on the values of a community (Fredline & Faulkner, 2001 ).

In line with previous studies on the impact of crises on tourism from the perspective of residents, this study suggests that a crisis event may serve as a point of reference from which to evaluate the social costs of the event. The inherent complexity involved in assessing the social costs and benefits of tourism at the time of crises has previously been demonstrated (Weaver & Lawton, 2013) . A crisis may mitigate residents' negative perceptions of tourism. Garau-Vadell et al. (2018) found that an economic crisis increases residents' willingness to support tourism because it significantly decreases their perception of the costs of tourism. Residents become more tolerant of tourism and attach greater value to economic development during an economic crisis (Lindberg & Johnson, 1997) . The balance of costs and benefits may differ between groups of residents, and residents' individual levels of engagement with the tourism and hospitality industry can be a determinant of their attitudes (Andereck, Valentine, Knopf, & Vogt, 2005) .

Conversely, the negative impact of tourism can be amplified by a crisis. In particular, an influx of infected persons to a tourist destination during a pandemic (Epstein et al., 2007; Bajardi et al., 2011) might have serious consequences for public safety.

More recently, Weaver and Lawton (2013) examined the variations in the cost-benefit assessments performed by residents in response to contentious events, and suggested that mass media could amplify residents' perceptions of the negative impact of tourism, whilst the social dimensions of residents, such as their social circles and ethnic backgrounds, may lead to more considered attitudes to the crisis. In addition, Rittichainuwat and Chakraborty (2009) indicated the importance of measuring the perceived negative impact of tourism in a time of crisis to minimize the damage wrought by negative perceptions.

Scholarly opinion on tourism and crises falls into two main contexts: risk perception at the individual level (on the demand side) and crisis management at the collective level (on the supply side). Research on the perceived risks associated with tourism has focused on tourists' perspectives rather than on the perspectives of destination communities, with the concept of perceived risk in tourism being associated primarily with studies of consumer behavior (Sharifpour et al., 2014) . Risk and safety issues have been generally examined from the tourists' J o u r n a l P r e -p r o o f perspective, seeking to determine why tourists perceive risks differently and what factors influence these perceptions (e.g., Gössling, Scott, Hall, Ceron, & Dubois, 2012; Lepp & Gibson, 2003) . Tourism-related risks may be those associated with terrorism, war, social instability (political or criminal), or health concerns. Perceived risks (or anxiety) may lead tourists to avoid a particular region, but this may be mitigated by tourists' past experiences, their levels of familiarity with similar events, their novelty-seeking behaviors (Sharifpour et al., 2014) , or their cultural orientations compared with those of the destination (Reisinger & Mavondo, 2006) . (Wang, 2009) , and of tourist boycotts (Yu, McManus, Yen, & Li, 2020) . Page, Yeoman, Munro, Connell, and Walker (2006) examined the effects of the swine flu pandemic on destination planning, in consideration of the risks presented to the public by the frenzied media coverage of this influenza outbreak.

Nonetheless, in the recent review of Ritchie and Jiang (2019) , covering 142 published studies on tourism crisis management, response and recovery strategies, and crisis prevention and planning practices, a lack of comprehensive theoretical and methodological assessments of the impacts of crises on the tourism industry was identified.

Some researchers have emphasized that tourism crisis management should consider the welfare of destination residents. The rapid growth of mass tourism, with the associated expansion of tourism infrastructure, has caused ecological crises in some destination regions (Lukashina, Amirkhanov, Anisimov, & Trunev, 1996) . It has been shown that the local communities of tourist destinations are aware of both the economic contributions of tourism and also the associated socio-environmental risks affecting their livelihoods (Schmidt, Gomes, Guerreiro, & O'Riordan, 2014) . Antunes, March, and Connolly (2020) found that the perceptions and responses to in situ risks, such as tourism-related gentrification, differ among different groups of destination residents. Easterling (2005) paid particular attention to the interests of residents and their representatives in managing a tourist destination, including crisis management. Eitzinger and Wiedemann (2007) stressed the importance of studying the risk perceptions of J o u r n a l P r e -p r o o f residents, because their views are shaped by their specific experiences, which differ from those of tourists.

The SARS epidemic of 2003 prompted many researchers to study the impacts of and responses to infectious respiratory diseases (Page et al., 2006) . The spread of infectious diseases, such as SARS, swine flu, and viral hemorrhagic fevers (such as that caused by the Ebola virus), via human travel has emerged as the main risk to tourism, as it leads to prohibitions on the international movement of people (Ala`a & Albattat, 2019). For instance, a high risk of influenza infection has been identified among people on cruises, on airplanes, or in tour groups (Freedman & Leder, 2005) . This was confirmed by the recent COVID-19 infections on cruise ships in Japan, the U.S., Australia, and France, which resulted in many countries banning cruise ships from docking in their ports (Al Jazeera, 2020). As COVID-19 has now spread to more than 200 countries or territories (WTO, 2020), many countries' governments have imposed unprecedented restrictions on the movements and behavior of their populations, and worldwide economic activity has decreased enormously. In addition to the health risks to tourists, there are obvious concerns that infected tourists may spread the disease to local residents (Gautret, Botelho-Nevers, Brouqui, & Parola, 2012) . Richter (2003) highlighted that during a global pandemic, large-scale international travel can spread infectious diseases and bring health threats to crowded urban areas. Moreover, there is a high risk of community transmission of COVID-19 (MacIntyre, 2020) and other respiratory diseases, such as seasonal influenza, being spread by travel within communities.

A few studies have examined residents' WTP for the costs of tourism (Andersson & Lundberg, 2013; Lindberg et al., 2001) but to the best of our knowledge no research has been conducted to quantify the social costs of pandemics, such as COVID-19, on tourist destinations.

To bridge this gap, a triple-bounded DCCVM was used to identify and quantify residents' WTP to mitigate the risk of the pandemic and the social costs it brings to destinations.

The target populations for this study were the residents of Hong Kong, Guangzhou, and Wuhan. These three cities were selected to reflect homogeneity and heterogeneity in urban characteristics, tourism development and its negative impact, and visibly negative social and economic effects and consequences of the COVID-19 pandemic.

As Table 1 shows, Hong Kong has an extremely high population density. This is the case because country parks and special areas cover 44,312 hectares, accounting for 40.9% of Hong Kong's total land area. Given that a relatively low population density is one of the indices of subjective well-being for urban residents (Cramer, Torgersen, & Kringlen, 2004) , the lower population densities of Guangzhou and Wuhan benefit these cities' residents. However, Hong Kong benefits economically from the enormous revenue generated by the tourism and hospitality industry, especially by visitors from mainland China, who account almost 80% of total arrivals.

Hong Kong in response to the COVID-19 crisis, but there is also the possibility of a "second wave" of infections due to incoming travelers, including people returning home from the U.S. and Europe. Hong Kong's painful experience of the outbreak of SARS in 2003, and its subsequent recovery (Hung, 2003) , served as a profound lesson for the Hong Kong government (Lee, 2009; Wan, 2013) and its people on how to respond to a health crisis (Lau, Yang, Tsui, & Kim, 2003) . Analogous to the economic effects of the SARS outbreak, the COVID-19 pandemic has already caused a year-on-year 96.4% decline in Hong Kong tourism, equating to a loss of HK$24.437 billion (US$3.15 billion) (see Table 1 ). Given the multiplier impact of tourism, the overall losses due to the COVID-19 crisis are likely to be even greater.

Hubei and Guangdong are the Chinese provinces hardest hit by the pandemic, with 67,802 1 and 1,514 2 confirmed cases of COVID-19, respectively, as of April 3, 2020. The capital cities of these two provinces, Wuhan and Guangzhou, accounted for 50,007 and 451 of these cases, respectively. Importantly, Wuhan was the city in which the first cases of COVID-19 were [Insert Table 1 (Aizaki, Nakatani, & Sato, 2014, p. 21) . The triple-bounded DCCVM extends the double-bounded DCCVM by offering respondents one additional set of choices, which results in more specific range of WTP that provides greater certainty. With an appropriate sampling design, a triple-bounded DCCVM can generate results that are statistically superior to those generated by other DCCVMs (Langford, Bateman, & Langford, 1996) . Demand curves and social costs were generated at the city level by aggregating the elicited WTP values.

Analyzing the responses to WTP questions in a contingent valuation survey constitutes an empirical approach to measuring economic concepts by quantifying the value that individuals place on the investigated attribute/activity (Hanemann, 1999) . A CVM based on individuals' WTP for public goods (resources or activities) evaluates the benefit (or cost) contingent on hypothetical changes (Lee & Han, 2002) . CVMs have been used to ascertain monetary measures of welfare changes, where these result from changes in the availability of public goods or amenities rather than from price changes (Hanemann, 1991) . WTP has been widely applied in consumer behavior studies, as it is a robust gauge of people's values or concerns, and can be regarded as a direct antecedent of consumers' purchasing intentions and behaviors (Ayadi & Lapeyre, 2016; Li, Li, & Kambele, 2012) . In tourism and hospitality studies, WTP is used to J o u r n a l P r e -p r o o f qualify and quantify the concept of "value" in a non-market product or service, including attractions in the natural environment and outdoor recreation (Asafu-Adjaye & Tapsuwan, 2008; Reynisdottir, Song, & Agrusa, 2008) , cultural festivals (Herrero, Sanz, & Devesa, 2011) , sustainability in luxury hotel brands (Kim, Barber, & Kim, 2019) , carbon emission reduction (Seetaram, Song, Ye, & Page, 2018) , and green events (Saayman, Krugell, & Saayman, 2016) .

In terms of the determinants of WTP, income is widely acknowledged as a positive influence (Broberg, 2010; Hanemann, 1991; Pollicino & Maddison, 2001; Reynisdottir et al., 2008) , although the effect remains controversial (Lee & Han, 2002; Seetaram et al., 2018) . Other demographic features, such as sex, age, and education level, may also affect an individual's WTP (Kim, Wong, & Cho, 2007; Seetaram et al., 2018) . The dependence on tourism of the household income of a tourist destination's residents affects their WTP, although the magnitude of this effect is also debated (Lindberg et al., 2001; Liu & Var, 1986) .

The triple-bounded DCCVM estimation of WTP in this study extended the double-bounded dichotomous choice (DBDC) model used by Aizaki et al. (2014) . Thus, a monetary value ( * ) was associated with the acceptance of a given option, such that the utility of realizing the option and paying * was equivalent to the utility of not realizing the option but retaining * . In addition, * was assumed to be individually specific and associated with a linear combination of individual characteristics and the presented bid:

where represents a matrix of the characteristics of individual n; is the presented bid; is the error term; and , , are model parameters. Natural logarithm is employed on the variable so that the coefficients can be directly referenced as elasticities.

Each respondent was presented with at most three dichotomous choices at sequential bid levels. It was assumed that a respondent would answer "yes" if and only if the presented bid was lower than * . Hence, there exist six possibilities, as illustrated by the decision tree in Figure 1, where bid 0 is the initial bid level in the first choice, whereas bid H1 and bid H2 (bid L1 and bid L2 ) are the subsequent bid levels in the following choices if the respondent answers "yes" ("no"). The axes show the ascending or descending direction of the bid level, and the labels on the right describe a respondent's answer to each choice (for example, NNY means answering "no" to the first two choices but "yes" to the third choice). It should be noted that in the cases of YN ("yes" J o u r n a l P r e -p r o o f to the first choice but "no" to the second choice) and NY ("no" to the first choice but "yes" to the second choice), only two choices are presented.

[Insert Figure 1 about here]

The probabilities of the six events are given by = ( 2 ); = ( 1 ) − ( 2 ); = ( 0 ) − ( 1 ); = ( 1 ) − ( 0 ); = ( 2 ) − ( 1 ); = 1 − ( 2 ), with () being the cumulative density function of error term .

The log-likelihood function can be formulated as

where * are binary-valued indicators of the associated event ( * = 1 if the event occurs). In this study, was assumed to follow a logistic distribution and the model estimation was performed in R using author-generated code based on the package DCchoice (Nakatani, Aizaki, & Sato, 2020) .

The survey population comprised the residents of Hong Kong, Guangzhou, and Wuhan. Two professional data companies were appointed to administer the survey electronically to a random sample of this survey population. A pilot survey was first conducted with 305 respondents to test the validity of the questionnaire and initial data. Subsequently, the main survey was conducted during February and March 2020, with 1,627 valid samples being collected for the main study, comprising n = 520 residents of Hong Kong, n = 503 of Guangzhou, and n = 604 of Wuhan. The respondents' demographic characteristics are shown in Table 3 , and consistent scales were J o u r n a l P r e -p r o o f applied across the three cities for all variables except for annual income, which is substantially different between Hong Kong and the two cities in mainland China. Different income ranges were designed to reflect this disparity in annual wages (Table 1) , and the data were later recoded into lower and higher groups, based on the average wage of each city (see Table 1 ), for further analysis.

The survey questionnaire comprised three parts. In the first part, questions covered the risk perceptions of residents toward the negative impact of tourism on their city amid the COVID-19

pandemic. Based on the pilot survey, nine items were selected, covering three elements of health risk (i.e., risk of cross-infection, shortage of medical supplies, the difficulty of prevention and patient tracing), three elements of the negative effects on social life (i.e., social panic and instability, commodity shortage, and environmental degradation), and three elements of the negative effects on tourism (i.e., reputational crisis in tourism, host-guest conflicts, and xenophobia). The results from this part of the survey measuring residents' perceived risks of tourism in the three cities-that is, the negative social impacts of the pandemic on these cities insofar as they are amplified by tourism-are presented in Figure 2 .

[Insert Figure 2 about here]

In the second part, the respondents' WTP was determined via their response to dichotomous choices, as described in the previous section. The monetary values presented in Table 2 were generated based on a pilot test and previous applications of CVM in the literature (e.g. Asafu-Adjaye & Tapsuwan, 2008; Bateman, Langford, Jones & Kerr, 2001) . Every respondent was randomly allocated to one of the three scenarios and answered at least two bids, and up to a maximum of three bids. This process enabled identification of the respondents' estimated WTP to reduce tourism's negative impact amid the pandemic crisis. To minimize biases in this CVMbased data collection, three scenarios with three biddings (either upper or lower) each were presented for acceptance or rejection by the respondents, as shown in Table 2 .

[Insert Table 2 about here]

The third part of the survey comprised questions regarding the demographic characteristics of the respondents (sex, age, and education level), as well as two questions on factors potentially influencing WTP, such as tourism employment and COVID-19 infection of the respondent or his/her acquaintances. We assumed that infected individuals and those who were acquainted with an infected person would generally have more proactive attitudes toward compensating for the damage of the pandemic, and would be thus willing to pay more to reduce the risks of tourism activity.

[Insert Table 3 about here]

Using the triple-bounded DCCVM, multiple models with different combinations of influencing factors were estimated and the best models were chosen according to the Akaike information criterion (AIC). The model selection process largely followed the general-to-specific procedure (Song, Witt, & Li, 2008, p. 46-69) , in which the potential influencing factors (i.e. demographic variables and bids) were included in the initial model and the improvements of the model estimates were made based on AIC by recursively eliminating the insignificant and irrelevant factors from the model. In this process, various combinations of demographic groups were considered. For example, the 11 income groups (seven in the case of Hong Kong) were combined into a "high income" group and a "low income" group according to the average income of each city, and the six age groups were restructured into "young", "middle aged", and "senior" groups. To facilitate comparisons of the influential factors among the three cities under consideration, a factor (or group) was included in the model estimation for all three cities if it was found to be significant/relevant in one of the models. The estimation of WTP and the coefficients of its influencing factors in the chosen model are summarized in Table 4 . The estimation of the mean of WTP followed the normalized truncation of WTP by Boyle, Welsh, and Bishop (1988) , while the confidence intervals of the mean of WTP and the median of WTP were constructed using the method proposed by Krinsky and Robb (1990) .

[Insert Table 4 about here]

The mean and median of WTP in each of the three investigated cities were approximately 300 and 200, respectively, in the local currency, with Guangzhou's median slightly higher than those of the other two cities. This indicates that the residents of these three cities were willing to pay an average of 300 in local currency to reduce the risk of negative tourism-generated pandemic effects, and that half of the population were willing to pay at least 200 in local currency for such preventative measures. Although the intensity of the pandemic differed between the three cities, no significant difference was observed in the basic WTP of residents between cities. This can be attributed to the extensive media coverage of COVID-19, which meant that all residents were aware of the severity of the pandemic, even in a city with fewer confirmed cases.

Regarding influencing factors, the estimated coefficient associated with each factor describes the marginal change in the probability of the respondents' answering "yes" to the dichotomous choice. The probability of a respondent answering "yes" was defined by the ratio between the probability of a "yes" answer and the probability of a "no" answer. In the current context, the higher probability of a "yes" answer generally implied a higher WTP to reduce the risk generated by tourism activity. A typical law of demand applied in the WTP for risk reduction: the probability of residents from all three cities answering "yes" decreased when the presented bid increased. In particular, there was a larger coefficient in Guangzhou (2.0101) than Hong Kong (1.6946) or Wuhan (1.6175), indicating a more elastic demand for risk reduction in Guangzhou. Given the extreme severity of the pandemic in Wuhan and the painful memory of SARS in Hong Kong, it is reasonable for the demands for risk reduction made by residents of these two cities to be much less elastic than that of residents of other cities.

There is an interesting relationship between age and WTP for risk reduction: younger residents were willing to pay more for risk reduction than were their parents and grandparents.

Specifically, the estimated coefficients for middle-aged and senior respondents were either negative or insignificant, indicating that these groups had a lower WTP than the younger group.

This observation is well supported by some anecdotal evidence during the COVID-19 outbreak of younger people needing to persuade their parents and grandparents to stay at home to minimize exposure to the virus. This phenomenon may be linked with the connectivity of the J o u r n a l P r e -p r o o f younger generation via the Internet and social media, which have provided extensive information on the pandemic. An alternative explanation of this research finding is that older people may be more philosophical and accepting about threats to their health, and thus prefer to take precautions while preserving their freedom in their daily lives.

The investigation of income groups points to the limits of the expected relationship between income and WTP amid the COVID-19 pandemic. In Hong Kong and Guangzhou, residents with higher incomes were willing to pay more to reduce the risks, but this was not the case in Wuhan.

This could reflect the severity of the negative impact of COVID-19 in Wuhan overpowering the income effect.

Of the 1,627 respondents from the 3 cities, 215 (13.2%) were employed in the tourism industry. Complexities in their WTP for risk reduction are to be expected: on the one hand, they should certainly be willing to pay to reduce the risk brought about by the COVID-19 pandemic;

on the other hand, the realization of this risk reduction would entail reducing the tourism activities that provide their income. Nonetheless, in Guangzhou and Wuhan, tourism practitioners were found to be willing to pay more than other residents to reduce the risks 

We approximated the demand curves from the WTP data to assess the costs of perceived risks due to tourism activities in the three cities. The demand curves were shaped by plotting the aggregate number of respondents who were willing to pay each assigned monetary amount (Seetaram et al., 2018) . Random sampling allowed us to extrapolate WTP to the general population of the city. The population for given levels was calculated from the respondents' aggregate number and was representative of the labor force, as the questionnaires were J o u r n a l P r e -p r o o f distributed to residents above 18 years of age. Based on the demand curve, we analyzed the social cost by aggregating the changes in individual welfare, which was approximated by the individual surplus from the WTP analysis. A numerical measure of social cost was generated by the integration of the areas of all trapezoids under the empirical demand curves (shaded areas in Figure 3 ). It should be acknowledged that the demand curve was approximated by using the mean value of respondents' WTP, as the dichotomous choice approach obliged respondents to reply "yes" or "no" to certain fixed amounts. 

This study estimated residents' WTP to reduce the risk associated with tourism activities in three Chinese cities amid a pandemic crisis. By applying the DCCVM approach and the elicited WTP values, the demand curves and social costs of tourism were compared in the three cities.

The findings of this study contribute to tourism research and have the following theoretical and practical implications. Our analysis shows that most respondents were willing to pay for risk reduction and action in responding to the pandemic crisis, and that there was no significant difference between residents' WTP in the three cities. At a time of crisis or disaster, the stakeholders in a tourist destination become even more (Scott & Laws, 2006) , and it appears that residents perceive their individual WTP and the costs borne by the community as a whole as inseparably linked, which makes them more willing to individually pay more to maintain public health and reduce the infection risk from tourism activity. This echoes Lindberg and Johnson (1997) , who reported the enlightened attitudes of residents toward tourism during a crisis, and suggested that a crisis invariably brings out the best in residents in terms of values and responsible behavior. Given that the majority of the respondents in the present study were willing to pay for reducing the social costs of tourism when confronted with a pandemic, charities could benefit from setting up specific funds to receive donations for risk reduction.

Second, residents' WTP was significantly affected by age, income, and tourism employment.

One specific finding was that younger residents were willing to pay more for risk reduction. This may be attributable to the fact that younger generations are more digitally savvy and more often connected to the Internet than older residents, which allows younger generations to access the most up-to-date information about the pandemic crisis in real time. This implies that local authorities and tourism organizations should involve younger generations, who are likely to be more motivated and knowledgeable, in crisis recovery actions in the aftermath of the pandemic.

Third, the social costs were compared between Hong Kong, Guangzhou, and Wuhan, which differ in the characteristics if their urban and tourism development, and have experiences different COVID-19 infection trajectories. At a conceptual level, it has been widely demonstrated that "the residents' attitudes toward tourism are directly related to the amount of its presence in the destination community" (Smith & Krannich, 1998, p. 784) . However, "the amount of its presence" in the local population, especially under the threat posed by a pandemic crisis, has yet to be quantified. In this study, tourism demand curves were created with data obtained by estimating and aggregating the number of respondents at given WTP values and quantifying the social costs of inbound tourism during the pandemic crisis. The demand curves based on WTP for three cities show a relatively similar pattern, which is logical given that J o u r n a l P r e -p r o o f COVID-19 has rapidly spread worldwide and adversely affected global public health. This indicates that the response plans of tourism destinations for the pandemic crisis can be conducted within the broader global framework of the destination system. Furthermore, these findings indicate that tourism may generate great social costs to the local community during the COVID-19 pandemic, and residents' involvement in tourism recovery strategies is therefore critical.

Furthermore, the findings of this study provide useful suggestions for the introduction of recovery and stimulus measures during and after the COVID-19 pandemic. In light of the approximation of the social costs, relief packages should be designed to benefit the society at large in tourist destinations suffering significant negative socio-economic impacts due to the pandemic. Conventional policy measures may not be able to overcome this crisis, as it has profoundly changed peoples' perception of the public health risks associated with tourism. The case of China shows that the tourism industry has been gradually and partially regaining its momentum in closer-to-home leisure destinations during weekends, but this is not applicable to major cities (STR, 2020). To satisfy peoples' needs for safety and desires for travel, recovery strategies should be formulated with a holistic and innovative mindset instead of focusing narrowly and directly on tourism recovery, which was the approach taken by many destinations after SARS (EDB, 2015). As Yang, Zhang, and Chen (2020) argued, welfare policies responding to the COVID-19 pandemic should be designed to allocate financial support across all sectors, including tourism, health, and other areas of the public and private sectors, to ensure the balanced recovery of the cities and regions affected.

There are two main limitations to this study that should be considered. First, the survey was conducted at the peak of the COVID-19 pandemic in China to elicit a timely response from the respondents. However, the respondents may have been influenced by a range of strong emotions at the time, meaning that the WTP decisions may have been biased by their affective states. For a similar reason, the survey was only available online during the peak of the pandemic, and this might have generated a sample distribution that was skewed towards individuals with higher education levels. Second, the social costs of tourism amid the COVID-19 crisis were explored rather than the impacts of COVID-19 on tourism. However, this crisis is still unfolding and its impacts on tourism are worthy of examination in future studies. The effects of the pandemic are likely to decrease over time, and there may be some positive impact on the tourism industry and/or on certain tourist destinations, such as ecological improvements from the dramatic drop in J o u r n a l P r e -p r o o f carbon emissions during the crisis Chen, Wang, Huang, Kinney, & Paul, 2020) .

Future studies are warranted to comprehensively evaluate these impacts.

J o u r n a l P r e -p r o o f Notes: 95% confidence interval in brackets; Standard error in parentheses; * p < 0.1, ** p < 0.05, *** p < 0.01

Stated preference methods using R

Current issue in tourism: Disease transmission as a potential risk for travellers

Tourism, crisis, disaster: An interdisciplinary approach

Coronavirus: Travel restrictions, border shutdowns by country. Al Jazeera and News Agencies

The paradox of the 'blue sky' created by corona 19: A greater environmental crisis is coming. The Seoul Shinmun

Residents' perceptions of community tourism impacts

Commensurability and sustainability: Triple impact assessments of a tourism event

Spatializing gentrification in situ: A critical cartography of resident perceptions of neighbourhood change in Vallcarca

A contingent valuation study of scuba diving benefits: Case study in Mu Ko Similan marine national park

Consumer purchase intentions for green products: Mediating role of WTP and moderating impacts of framing

Human mobility networks, travel restrictions, and the global spread of 2009 H1N1 pandemic

Bound and path effects in double and triple bounded dichotomous choice contingent valuation

Estimating the tourism induced environmental Kuznets curve in France

Validation of empirical measures of welfare change: Comment

Income treatment impacts in contingent valuation: The case of the Swedish predator policy

Air pollution reduction and mortality benefit during the COVID-19 outbreak in China. medRxiv

Quality of life in a city: The impact of population density

Structural breaks in international tourism demand: Are they caused by crises or disasters? Tourism Management

The residents' perspective in tourism research: A review and synthesis

Economic analysis and evaluation of government policies-Individual Visit Scheme in Hong Kong

Risk perceptions in the alpine tourist destination Tyrol-Aan exploratory analysis of residents' views

Controlling pandemic flu: The value of international air travel restrictions

Variations in residents' reactions to major motorsport events: Why residents perceive the impacts of events differently

Influenza: Changing approaches to prevention and treatment in travelers

Economic crisis and residents' perception of the impacts of tourism in mass tourism destinations

The spread of vaccinepreventable diseases by international travellers: A public -health concern

The dispute over tourist cities. Tourism gentrification in the historic Centre of Palma

Pop-up' tourism or 'invasion'? Airbnb in coastal Australia

Overtourism: Rresidents' perceptions of tourism impact as an indicator of resident social carrying capacity -case study of a Spanish heritage town

Consumer behaviour and demand response of tourists to climate change

Willingness to pay and willingness to accept: How much can they differ?

Valuing environmental preferences: Theory and practice of the contingent valuation method in the US, EU, and developing countries

Perceived impacts of tourism: The case of Samos

Measuring the economic value and social viability of a cultural festival as a tourism prototype

Global viral outbreaks like coronavirus, once rare, will become more common

The SARS epidemic in Hong Kong: What lessons have we learned

Does tourism really cause stress? A natural experiment utilizing ArcGIS survey123

Tourism multiplier impacts on Singapore

Assessing the economic value of a world heritage site and willingness-to-pay determinants: A case of Changdeok Palace

Sustainability research in the hotel industry: Past, present, and future

On approximating the statistical properties of elasticities: A correction

Assessing impacts of SARS and avian flu on international tourism demand to Asia

A multilevel modelling approach to triple-bounded dichotomous choice contingent valuation

Monitoring community responses to the SARS epidemic in Hong Kong: From day 10 to day 62

Estimating the use and preservation values of national parks' tourism resources using a contingent valuation method

How the Hong Kong government lost the public trust in SARS: Insights for government communication in a health crisis

Tourist roles, perceived risk and international tourism

Luxury fashion brand consumers in China: Perceived value, fashion lifestyle, and willingness to pay

Tourism development: Assessing social gains and losses

The economic values of tourism's social impacts

Resident attitudes toward tourism impacts in Hawaii

Tourism and environmental degradation in Sochi

Global spread of COVID-19 and pandemic potential

Can nature-based tourism benefits compensate for the costs of national parks? A study of the Bavarian Forest National Park

Social impacts of tourism on central Florida

DCchoice: An R package for analyzing dichotomous choice contingent valuation data

Assessing the impacts of the global economic crisis and swine flu on inbound tourism demand in the United Kingdom

A case study of best practice-Visit Scotland's prepared response to an influenza pandemic

Valuing the benefits of cleaning Lincoln Cathedral

Cultural differences in travel risk perception

Willingness to pay entrance fees to natural attractions: An Icelandic case study

International tourism and its global public health consequences

Tourism disaster planning and management: From response and recovery to reduction and readiness

A review of research on tourism risk, crisis and disaster management: Launching the Annals of Tourism Research curated collection on tourism risk, crisis and disaster management

Perceived travel risks regarding terrorism and disease: The case of Thailand

Willingness to pay: Who are the cheap talkers?

Are we all on the same boat? The challenge of adaptation facing Portuguese coastal communities: Risk perception, trust-building and genuine participation

Tourism crises and disasters: Enhancing understanding of system impacts

The influence of terrorism in tourism arrivals: A longitudinal approach in a Mediterranean country

Estimating willingness to pay air passenger duty

Risk perception, prior knowledge, and willingness to travel: Investigating the Australian tourist market's risk perceptions towards the Middle East

The impacts of financial and non-financial crises on tourism: Evidence from Macao and Hong Kong. Tourism Management Perspectives

Tourism dependence and resident attitudes

Overtourism and resident resistance in Budapest

The Advanced Econometrics of Tourism Demand

Effects of terrorist attacks on tourist flows to France: Is wine tourism a substitute for urban tourism

Dubai's tourism industry and its societal impact: Social implications and sustainable challenges

COVID-19 webinar summary: 5 key points on Asia Pacific hotel performance, 28 May

Reconsidering the estimation of the economic impact of cultural tourism

A comparison of the governance of tourism planning in the two special administrative regions (SARs) of China-Hong Kong and Macao

The impact of crisis events and macroeconomic activity on Taiwan's international inbound tourism demand. Tourism Management

Resident perceptions of a contentious tourism event

Coronavirus disease (COVID-19) outbreak situation. World Health Organization (WHO)

Megacities and more: A guide to China's most impressive urban centers. CNN travel

Coronavirus pandemic and tourism: Dynamic stochastic general equilibrium modeling of infectious disease outbreak

Tourism boycotts and animosity: A study of seven events

Qiu is an assistant professor in the Department of Integrated Resort and Tourism Management

Jinah Park is a postdoctoral fellow in the School of Hotel and Tourism Management

Shina Li is a professor in the School of Tourism Management

Haiyan Song is Mr and Mrs Chan Chak Fu Professor in International Tourism at the School of Hotel and Tourism Management

The authors would like to acknowledge the financial support of Mr and Mrs Chan Chak Fu Endowed Professorship Fund.

J o u r n a l P r e -p r o o f 

The main contribution of this study is the identification and quantification of the social costs of tourism during the COVID-19 pandemic in three urban destinations in China. Methodologically, hypothetical scenarios and triple-bounded dichotomous choice contingent valuation method are used to quantify the residents' risk perception of the COVID-19 pandemic and their willingness to pay for reducing such risk. The empirical results of the study contributes to the understanding of the attitudes of destination residents toward the mitigation and management of the pandemic crises, which sheds light on the post-pandemic recovery and planning strategies in destinations that suffer from the pandemic.

The study's social science perspective is reflected across the research design and discussions of the research findings. Studies of willingness to pay using contingent valuation method has been widely applied in social science research to assess the costs of the public goods or environmental amenities. This study extends the scope to assess the possible social costs of tourism during an unfolding pandemic across the world with far reaching social and economic implications.