Escape or stay? Diverse effects of extreme weather risks on household tourism consumption

Abstract

A comprehensive assessment of extreme weather risks’ impact on tourism provides the scientific basis for climate-resilient tourism, yet research from the tourist-origin perspective remains limited. We establish a conceptual framework based on Protection Motivation Theory and present supporting empirical evidence. Using China Family Panel Studies data and extreme weather risk data, we demonstrate asymmetric effects: extreme high temperature elevates household tourism consumption, while extreme rainfall reduces it. Internet use significantly moderates the effects of extreme high temperature and extreme rainfall, while income moderates the extreme rainfall-consumption relationship. Married households and non-agricultural households demonstrate stronger negative reactions to extreme rainfall. Furthermore, the increase in extreme high temperature-driven consumption stems from additional consumption. This study offers specific tourism management implications in addition to enhancing the knowledge system regarding the extreme weather effects on tourism.

Keywords

extreme weather risks household tourism consumption tourist origins protection motivation theory

Introduction

Tourism is strongly subject to extreme weather (Lin et al., 2023), as its development is significantly determined by the natural environment and climatic conditions (IPCC, 2014). Arguably, weather is crucial for sustainable tourism (López-Dóriga et al., 2019). According to the World Economic Forum, numerous tourist destinations have seen substantial changes owing to extreme weather, and tourism in well-known destinations including Greece, Italy, and the Mediterranean and Caribbean have already been negatively impacted¹. The World Meteorological Organization predicts that extreme weather risks are expected to increase further throughout the next 5 years², placing a heavy burden on global tourism sustainability. Given that, investigating the relationship between extreme weather risks and tourism is vital for ensuring the long-term sustainable development of tourism.

The threat of extreme weather risks to tourism manifests in various direct and indirect ways (Susanto et al., 2020). Direct impacts include physical damage to tourism infrastructure from extreme events. Intense rainfall and flooding can wash out access roads and damage facilities, causing service disruptions and losses of cultural and natural heritage (Smith and Fitchett, 2020). Direct impacts also encompass ecological shocks to nature-based tourism resources, including national parks, coastal destinations, and ski or glacier resorts, where weather-related disturbances threaten wildlife (Dube and Nhamo, 2018), degrade marine biodiversity (Birchenough, 2017), and reduce snowfall and snow reliability (Steiger et al., 2020). Indirectly, heightened risk perceptions among tourists prompt behavioral adjustments that reduce destination attractiveness and effective hosting capacity, shorten length of stay, and lower visitor spending, thereby reshaping the spatial and temporal distribution of tourist flows (López-Dóriga et al., 2019).

Despite its importance, established studies on the effects of extreme weather risks on tourism are largely limited to the tourist destination perspective (Scott et al., 2019), lacking robust evidence from tourist origins. Exploring the tourist-origin perspective is crucial because a comprehensive assessment from both origins and destinations can strengthen the body of knowledge on the effects of extreme weather on tourism. In light of this, focusing on the tourist origins perspective, this study formulates a conceptual framework based on Protection Motivation Theory (PMT) to elucidate tourists’ travel decisions in the context of extreme weather risks, therefore advancing existing research. Second, we utilize econometric methods to examine the asymmetric effects of extreme weather risks on household tourism consumption. Further, we explore potential moderating mechanisms in two dimensions: Internet use and household income level, aiming to provide universal implications for developing climate-resilient tourism. We also discuss the heterogeneity of the effects and the incremental consumption effect.

This study contributes to the established research on the following three grounds. First, we add to the studies on the effects of extreme weather risks on tourism from a tourist origins perspective, and enrich the body of knowledge in the field. The significance of extreme weather risks to tourism has been examined from the destination perspective extensively (Smith and Fitchett, 2020). Nevertheless, the tourist origins perspective has received little consideration. Thus, this study intends to advance our knowledge of this under-researched area regarding the effects of extreme weather risks on household tourism consumption. We also believe that a comprehensive investigation is crucial for fostering resilient tourism systems, aligning with the urgent imperatives of the Global Goal on Adaptation³.

Second, this study develops a PMT-based conceptual framework that provides a systematic perspective for interdisciplinary research on extreme weather risk and household tourism consumption. Residential exposure to extreme weather risk influences households’ threat and coping appraisals, which together drive tourism consumption decisions under risk. We include income level as a moderator to address class-based resource inequities (Alegre et al., 2010). Furthermore, we conceptualize Internet use as an information conduit based on Negative Bias Theory that might influence risk perceptions, thereby affecting tourism consumption.

Third, systematic insights into the impacts of extreme weather risks on tourism from the tourist origins perspective are empirically gained, contributing to the scientific basis for sustainable development policies formulation in tourism that is oriented towards uncertain future weather scenarios. Most established research has been with macro data or selective surveys; we refer to a more representative sample covering 10,339 households from 6 waves of China Family Panel Studies (CFPS) data between 2012 and 2022. This unique, micro large-scale dataset enhances the robustness and validity of the results. We demonstrate the asymmetric effects of extreme weather risks, specifically, extreme high temperature promotes household tourism consumption, while extreme rainfall hinders it.

The framework is displayed in Figure 1.

Figure 1.

Research framework.

Literature review

Research on determinants of tourism consumption

Scholarly interest in tourism consumption has increased, as evidenced by a growing literature on its determinants. Drawing on the hierarchical leisure-constraints tradition (Crawford et al., 1991), the determinants of tourism consumption are shaped by both structural constraints (time, income, opportunities, destination attributes, weather conditions, family life cycle) and nonstructural constraints (intrapersonal and interpersonal factors such as health, abilities, interests, risk perceptions, and family relationships) (Crawford and Godbey, 1987; Godbey et al., 2017). On the structural side, sociodemographic and economic conditions systematically influence travel decisions and outlays. Income, assets, debt, and labor market and time constraints are especially salient (Jang and Ham, 2009). Indebted or non-home-owning households spend less on tourism (Hung et al., 2013; Lin et al., 2020), car ownership relaxes mobility and budget constraints (Lin et al., 2020), and better health of the household head increases both the likelihood and the level of tourism consumption (Cai and Zhang, 2023). On the nonstructural side, attitudinal and interpersonal factors operate above and beyond structural conditions (Hubbard and Mannell, 2001). For example, generalized social trust raises both the probability and the level of tourism consumption, highlighting the role of psychological dispositions in travel choices (Sun et al., 2022). Although prior research has identified the general determinants of tourism consumption, the role of weather conditions as a key structural determinant remains underexplored, particularly with respect to extreme weather risks, providing the theoretical motivation for our focus on weather risk.

Research on extreme weather and tourism

Given the foundational role of weather conditions in tourism, extreme weather risks have become a prominent focus of tourism research. In general, tourist destinations are exposed to extreme weather risks, such as extreme high temperature, extreme rainfall, and other significant climatic factors, which may diminish the appeal of tourist destinations (Scott et al., 2019). Specifically, sea-level rise may undermine essential sun-sand tourism assets (Karditsa et al., 2024), thus diminishing the economic viability and competitiveness of coastal beach destinations (Spencer et al., 2022). Other forms of nature-based tourism are similarly affected by extreme weather risks. For instance, the glacial melting degrades landscape aesthetics and cultural importance, thereby directly undermining destination value (Wang and Zhou, 2019), while snow scarcity compels market redistribution to higher altitudes, fragmenting the competitiveness of ski tourist destinations (Steiger and Stötter, 2013). Extreme weather risks can also disrupt the associated infrastructure (Smith and Fitchett, 2020), detrimentally affecting tourism (Gössling et al., 2012). The corresponding adaptation strategies primarily encompass technical modifications, including artificial snow production (Elsasser and Bürki, 2002), product diversification (Kaenzig et al., 2016), and spatial relocation (Nunes and Loureiro, 2016), designed to bolster the resilience of tourism on a macro scale.

Some scholars have found that extreme weather events may, under certain conditions, stimulate tourism development. It is important to clarify that this is not a direct benefit of the extreme events themselves. In the short term, such events typically have pronounced negative impacts, for example, heavy rainfall disrupting tourism activities. However, over the longer term, the broader trajectory of climate change (e.g., global warming) may indirectly generate positive effects in certain regions by reshaping the temporal and spatial distribution of tourism flows. For instance, gradual warming may enhance thermal comfort during shoulder seasons at certain destinations (Köberl et al., 2016), thereby improving suitability for outdoor recreation and raising demand and revenues. In addition, climate change is shifting comparative advantages toward higher-latitude and higher-altitude destinations (Hamilton et al., 2005). Locations that were previously less attractive because of low temperatures may become more competitive as warming ameliorates climatic conditions, drawing more visitors.

Research gaps

However, important gaps remain. First, relative to the destination perspective that dominates impact assessment, the origin perspective is underdeveloped. Current research mostly investigates the impact of extreme weather risks on tourism from the perspective of the destination, including damage to tourist resources, diminished destination appeal, and disruption of infrastructure (Hamilton et al., 2005; Rosselló-Nadal et al., 2010; Wei and Huang, 2025). Nonetheless, there is limited knowledge regarding demand-side reactions to extreme weather risks at tourist origins. Tourism is affected by both the conditions at the destination and by environmental risks, perceived discomfort, and coping capacities prior to making travel decisions. Second, existing origin perspective evidence is largely macro aggregated and does not unpack the micro level decision mechanisms governing household tourism consumption. Third, most analyses consider a single risk, limiting comparative assessment of how distinct extremes differentially influence tourism (Sun et al., 2019; Wang et al., 2022).

As extreme weather events become more frequent, understanding demand-side responses is also essential for sustainable tourism planning and climate adaptation. Considering the destination alone cannot fully capture the decision-making mechanism underlying tourism consumption. The destination perspective emphasizes how supply-side risks affect tourists’ on-site experiences and destination operations, whereas the origin perspective highlights how local environmental risks at the tourist origins may shape households perceived risk, coping capacity, and tourism consumption decisions. Because decision-making on the demand side largely occurs before tourists reach the destination, it is difficult to observe from a destination-only viewpoint. Introducing the origin perspective is therefore essential for comprehensively assessing the impact of weather risks on tourism consumption. Given that, we pioneer an investigation into the effects of extreme weather risks on tourism from the tourist origins perspective. A theoretical analytical framework was established based on the PMT. Then, we empirically test the effects of multiple extreme weather risks on tourism consumption with a large-scale unbalanced dataset.

Theoretical analysis

Protection motivation theory

This study establishes a theoretical framework based on Protection Motivation Theory (PMT) to thoroughly investigate the influence of extreme weather risk on household tourism consumption. PMT, a widely utilized psychological model, effectively elucidates behavioral responses to perceived threats and is recognized as one of the most reliable theories for forecasting protective intentions and actions. It has been utilized across various fields, with tourism experts increasingly applying it to analyze the impact of risk on travel decision-making (Sönmez and Graefe, 1998; Zheng et al., 2021).

PMT asserts that humans interpret risk signals and prospective losses via defensive cognitive mechanisms influenced by personal traits and situational conditions, subsequently converting these evaluations into coping behaviors (Rogers, 1983). The framework consists of three components. First, information sources, including personal factors and external environmental conditions. Second, a cognitive mediation process involving threat appraisal and coping appraisal. Threat appraisal denotes an initial assessment of a threat, encompassing perceived severity (the seriousness of the repercussions) and perceived vulnerability (the possibility of being impacted). Coping appraisal encompasses self-efficacy (the conviction in one’s capacity to execute a protective response) and response efficacy (the belief that the response would effectively mitigate harm), together with an implicit evaluation of the necessary resources, including time, finances, and information. As risk appraisal increases and coping appraisal suggests the availability of viable and effective remedies, individuals are more inclined to engage in adaptive behaviors. In contrast, individuals are more prone to exhibit maladaptive behaviors, like avoidance, procrastination, or inaction. Third, coping modes, which can be classified as adaptive, which involves altering an unfavorable condition, or maladaptive, which entails preserving the current state.

In the present study, extreme weather risks at the tourist origin constitute external environmental threat cues. Household tourism consumption is treated as a discretionary consumption decision that can be maintained, increased, postponed, or reduced in response to perceived local weather risk. Threat appraisal captures households’ perceptions of health risk, physical discomfort, mobility constraints, schedule uncertainty, and potential consumption losses caused by local extreme weather. Coping appraisal captures households’ perceived ability to cope with these risks, including time flexibility, financial resources, information access, and the ability to adjust tourism consumption arrangements.

Direct effects of extreme weather risk on household tourism consumption

According to PMT, external environmental risks influence behavior through threat appraisal and coping appraisal, which jointly determine whether households adopt adaptive or avoidance-oriented responses (Rogers, 1975; Verkoeyen and Nepal, 2019). The relevant environmental risks are local extreme weather risks experienced at the household’s place of residence, including extreme high temperature, extreme low temperature, extreme rainfall, and extreme drought. Tourism consumption is a discretionary and optional form of household consumption; therefore, it is likely to be sensitive to perceived risk, expected utility, travel feasibility, time constraints, and coping costs (Juschten et al., 2019). Extreme weather risks may affect household tourism consumption through two competing channels. If households perceive tourism consumption as an effective way to sustain well-being or alleviate discomfort caused by local weather conditions, extreme weather may induce adaptive responses in tourist consumption. Second, if households perceive inclement weather as increasing health risks, mobility constraints, schedule uncertainty, or financial loss, they may reduce tourism consumption through protective avoidance. The direction of the impact of localized extreme weather risk on household tourism consumption is theoretically ambiguous. The direction is contingent upon the nature of weather threats and the equilibrium among perceived threat, coping capacity, and household resources.

Effect of extreme high temperature

Extreme high temperature can strengthen threat appraisal by increasing physical discomfort, heat-related health concerns, psychological stress, and constraints on outdoor activities (Fang et al., 2023). These risks may reduce household tourism consumption when households perceive heat exposure, departure inconvenience, or planning uncertainty as sufficiently severe. In this case, households may adopt protective avoidance by cancelling, postponing, or reducing tourism consumption. From a coping-appraisal perspective, extreme high temperature is typically seasonal and predictable, enabling households to assess their adequacy of time, information, income, and transportation resources for adjusting consumption plans. When households view tourism consumption as a viable strategy to alleviate local heat-induced discomfort, they may sustain or augment their consumption on transportation, accommodation, leisure services, and other tourism-related pursuits (Juschten et al., 2019). The impact of extreme high temperature on household tourism consumption is conceptually uncertain and contingent upon whether protective avoidance or adaptive consumption prevails.

Effect of extreme low temperature

Extreme low temperature may increase perceived severity and vulnerability by raising health concerns, physical discomfort, icy-road risk, and transport uncertainty (Chen et al., 2023). Under these conditions, households may reduce tourism consumption because the expected costs and risks of tourism activities become higher. Nonetheless, extreme low temperature may elevate the demand for comfort-enhancing leisure and tourist activities when households perceive that such consumption can enhance short-term well-being and possess the necessary coping resources. The relative strength of these two mechanisms may vary with infrastructure, household resources, local adaptation capacity, and weather severity (Liu, 2022). Consequently, the direct effect of extreme low temperature on household tourism consumption is theoretically indeterminate.

Effect of extreme rainfall

Extreme rainfall amplifies threat appraisal by creating immediate and localized risks, including flooding, road-safety risks, and transport disruption, which directly diminish the viability and dependability of tourism mobility (Hewer et al., 2015; Hübner and Gössling, 2012; Scott et al., 2012), thus reinforcing threat appraisal. Established studies have demonstrated that weather conditions influence tourism demand (Becken and Wilson, 2013; Wilkins et al., 2018). From the perspective of coping appraisal, tourism consumption is discretionary and more deferrable than essential consumption; households experiencing extreme rainfall are more inclined to engage in protective avoidance by postponing, canceling, or diminishing tourism consumption. Consequently, protective avoidance is expected to dominate the household response to extreme rainfall, suggesting a theoretically adverse direct impact on household tourism consumption.

Effect of extreme drought

Extreme drought is generally persistent and accumulative. From the perspective of PMT, prolonged drought may strengthen threat appraisal by heightening perceived resource scarcity, environmental strain, and uncertainty over local living conditions. These pressures may compel households to save resources, reallocate consumption toward essential coping requirements, and diminish discretionary tourism consumption (Alegre et al., 2010). Drought-induced discomfort, degradation of the local living environment, and compounded psychological stress may elevate the perceived value of tourism consumption as a short-term adaptive reaction (OBrien et al., 2014; Wang et al., 2022). The emergence of this response is contingent upon coping appraisal, particularly for households’ income adequacy, temporal flexibility, information accessibility, and confidence in their capacity to finance and organize tourism consumption. Consequently, the direct impact of severe drought on household tourism consumption is theoretically uncertain.

Moderating effect of internet use

Internet use may influence the correlation between extreme weather risk and household tourism consumption by altering households’ assessments of threat and coping appraisals, functioning as an information-processing and decision-support tool. Increased Internet use allows households to obtain prompt information regarding local weather warnings, transportation disruptions, and other risk-related notifications. This information may enhance risk awareness via the social amplification of risk and negativity bias, thereby reinforcing threat assessment and increasing the probability that households defer, diminish, or eliminate tourism consumption when local weather risks become more pertinent (Fuchs et al., 2024). Simultaneously, Internet use may improve coping evaluation by diminishing information asymmetry, search costs, and transaction frictions in tourism consumption (Xiang and Gretzel, 2010; Zarezadeh et al., 2023). If risk recognition prevails, Internet use may enhance protective avoidance; conversely, it may alleviate adverse weather impacts or facilitate adaptive consumption.

Moderating effect of household income level

Household income may influence the impact of extreme weather risk on tourism consumption by determining the resources available for coping assessment. Tourism consumption is discretionary; thus, lower-income households face liquidity constraints and limited capacity for marginal adjustments, while higher-income households possess greater discretionary spending power and enhanced flexibility in managing consumption shocks (Alegre et al., 2010; Brida and Scuderi, 2013; Ma et al., 2024). This resource advantage can provide two contrasting moderating patterns. When tourism consumption is regarded as a viable adaptive response to local environmental stress, increased income may amplify beneficial impacts. Conversely, when local weather risk diminishes projected utility, comfort, or reliability, higher-income households may exhibit greater willingness and capacity to defer or diminish tourism consumption, so reinforcing protective avoidance. Consequently, income level may either enhance or diminish the consequences of weather, contingent upon whether adaptive consumption or protective avoidance prevails.

Method

Variables

Explained variable: Household tourism consumption

Household tourism consumption (travel) is quantified by total household consumption on tourism during the past year, derived from the CFPS questionnaire item, ‘What was your household’s consumption on tourism in the past year?’. The year 2012 serves as the base period and is deflated using the CPI index of the province in which the household is located before being processed using a logarithmic transformation.

Explaining variable: Extreme weather risks of cities where households reside

Extreme weather risks are measured using four indices: extreme low temperature (ExtrLowTemp), extreme high temperature (ExtrHighTemp), extreme rainfall (ExtrRain), and extreme drought (ExtrDrought). The indices are computed as follows (Guo et al., 2024).

Step 1: remove samples with substantial missing data.

Step 2: Compute the historical distributions of each indicator for the period from January 1, 1973 to December 31, 1992. The 90th and 10th percentiles define thresholds for extreme high and low temperatures, respectively, and the 95th and 5th percentiles define thresholds for extreme rainfall and extreme drought. Consistent with WCDMP guidance⁴, the 90th and 10th percentiles for temperature balance meaningful departures from normal with statistical reliability: they capture moderately rare warm and cold events without unduly shrinking the sample, thereby improving interannual stability and power for trend detection. By contrast, daily precipitation is intermittent and strongly right skewed; to isolate genuinely intense events, thresholds are set using the 95th percentile. Symmetrically, the 5th percentile is used to characterize extreme drought conditions.

Step 3: Compile the number of days with extreme weather events over the study period. The extreme high temperature days ${E x t r H i g h T e m p}_{i, n}$ , extreme low temperature days ${E x t r L o w T e m p}_{i, n}$ , extreme rainfall days ${E x t r R a i n f a l l}_{i, n}$ , and extreme drought days ${E x t r D r o u g h t}_{i, n}$ can be calculated.

{E x t r H i g h T e m p}_{i, n} = \sum_{d = 1}^{365} I (T_{i, n, t} > τ_{i}^{90})

(1)

{E x t r L o w T e m p}_{i, n} = \sum_{d = 1}^{365} I (T_{i, n, t} < τ_{i}^{10})

(2)

{E x t r R a i n}_{i, n} = \sum_{d = 1}^{365} I (R_{i, n, t} > R_{i}^{95})

(3)

{E x t r D r o u g h t}_{i, n} = \sum_{d = 1}^{365} I (H_{i, n, t} < H_{i}^{5})

(4)

Where

T_{i, n, t}

denotes the daily mean temperature at station i on day y of year n.

R_{i, n, t}

and

H_{i, n, t}

represent, respectively, rainfall and relative humidity at station i on day y of year n.

Step 4: To account for disparities in the characteristics of the diverse extreme weather observations, we standardized each index to construct the weather risk index. Taking the extreme high temperature days as an example:

{\bar{E x t r H i g h T e m p}}_{m, n} = \frac{{E x t r H i g h T e m p}_{m, n} - \min_{k = 1, . . ., K; l = 1, . . ., L} {{E x t r H i g h T e m p}_{k, l}}}{\max_{k = 1, . . ., K; l = 1, . . ., L} {{E x t r H i g h T e m p}_{k, l}} - \min_{k = 1, . . ., K; l = 1, . . ., L} {{E x t r H i g h T e m p}_{k, l}}} \times 100

(5)

Where K is the total number of cities, and L is the total number of years in the sample set. The same procedure applies to the other three indices. In this way, the four indices ExtrLowTemp, ExtrHighTemp, ExtrRain, and ExtrDrought can be obtained.

Moderating variables: Internet use and household income level

The theoretical analysis suggests that Internet use and household income level may moderate the effect of extreme weather risks. Utilizing the CFPS questionnaire item ‘Monthly Postal and Telecommunications Consumption (Yuan/Month). What is the average monthly consumption of your household on mailing and communication services, including telephones, cell phones, internet access, and postal services?’, we proxy Internet use by monthly household consumption on communication services (Yang et al., 2023). This statistic encompasses a wider range of communication expenditures beyond just Internet use, potentially including telephone, mobile communication, and postal service elements. Thus, it ought to be viewed as an indirect indicator of Internet-related communication access, rather than a direct assessment of online information retrieval or Internet search behavior. In addition, we utilize the logarithm of per capita household income to represent household income level (Liu et al., 2024).

Control variables

Other than extreme weather risk, many factors are demonstrated to influence household tourism consumption. To mitigate endogeneity problems that may result from omitted variables, inspired by Deng et al. (2022), Dong et al. (2022), Huang et al. (2022), Ma et al. (2022), Chen et al. (2024), Liu et al. (2024) and Wang et al. (2025), we include several control variables to reflect households’ characteristics (namely household income per capita, household indebtedness, household size, and total household property), householders’ characteristics (namely age, gender, education level, health status, working status, marital status, and life satisfaction), and cities’ characteristics (household registration category, GDP per capita, and transportation accessibility).

Descriptive statistics are displayed in Table 1.

Table 1.

Descriptive statistics.

Variable symbols	Calculations	Obs	Mean	Std. dev.	Min	Max
Explained variable
Household tourism consumption	Deflated using the province’s CPI index with 2012 as the base period	38,207	1.6229	3.1842	0	12.2010
Explaining variables
Extreme low temperature	Relative threshold-normalized method	38,547	2.9509	0.4492	0	3.7511
Extreme high temperature	Relative threshold-normalized method	38,547	3.7783	0.3460	0	4.4282
Extreme rainfall	Relative threshold-normalized method	38,547	2.5770	1.3013	0	4.9006
Extreme drought	Relative threshold-normalized method	38,547	2.8126	0.6129	0.3079	4.4350
Moderator variables
Internet use	The logarithm of monthly household consumption on communication services. Deflated using the province’s CPI index with 2012 as the base period.	29,858	4.8810	0.9613	−0.0071	8.4876
Household income level	The logarithm of per capita household income. Deflated using the province’s CPI index with 2012 as the base period	36,779	0.0424	1.3345	−10.6194	6.5305
Individual level control variables
Age	Age in year of survey	38,535	51.4976	14.3478	16	95
Gender	male = 1, female = 0	38,546	0.5268	0.4993	0	1
Education level	0 = illiterate/semi-illiterate; 3 = elementary school; 4 = junior high school; 5 = high school/middle school/technical school/vocational high school; 6 = junior college; 7 = bachelor’s degree; 8 = master’s degree; 9 = doctoral degree	38,275	2.2072	1.8729	0	9
Health status	Takes values of 1–5, the larger the value the healthier it is	38,547	2.7864	1.2193	1	5
Job status	With a job = 1, without a job = 0	38,547	0.7595	0.4274	0	1
Marital status	Having a spouse = 1, without spouse = 0	38,171	0.8541	0.3530	0	1
Life satisfaction	Takes values of 1–5, the larger the value the more satisfied one is	37,661	3.7210	1.0537	1	5
Household level control variables
Household income level	The logarithm of the per capita household income. Deflated using the province’s CPI index with 2012 as the base period	36,779	0.0424	1.3345	−10.6194	6.5305
Household indebtedness	With household indebtedness = 1, without household indebtedness = 0	38,547	0.3191	0.4661	0	1
Household size	Number of family members	38,547	2.6925	1.3068	1	11
Total house property	The logarithm of the total family property. Deflated using the province’s CPI index with 2012 as the base period	38,547	10.9669	3.9102	−0.0383	18.1856
City level control variables
Household registration category	Urban = 1, rural = 0	38,172	0.4749	0.4994	0	1
GDP per capita	The per capita GDP of the city where the household is located. Deflated using the province’s CPI index with 2012 as the base period	38,547	10.7896	0.6525	9.0066	12.1601
Transportation accessibility	The logarithm of per capita road areas	38,547	2.5798	0.5318	1.4061	3.5551

Model specification

Based on the following econometric specifications, we examine the effect of diverse extreme weather risks on household tourism consumption:

{l n T o u r C o n s u m p}_{i p t} = α + β_{1} {l n E x t r H i g h T e m p}_{p t} + β_{2} {l n E x t r L o w T e m p}_{p t} + β_{3} {l n E x t r R a i n}_{p t} + β_{4} {l n E x t r D r o u g h t}_{p t} + β_{5} {C o n t r o l s}_{i p t} + ρ_{p} + γ_{t} + ε_{i p t}

(6)

where subscripts

i

p

, and t denote household, prefecture-level city, and year, respectively;

{T o u r C o n s u m p}_{i p t}

denotes the tourism consumption of household

i

located in prefecture-level city

p

in year

t

;

{\ln E x t r L o w T e m p}_{p t}

{l n E x t r H i g h T e m p}_{p t}

{l n E x t r R a i n}_{p t}

, and

{l n E x t r D r o u g h t}_{p t}

denote the extreme low temperature, extreme high temperature, extreme rainfall, and extreme drought indexes of city

p

in year

t

, respectively;

{C o n t r o l s}_{i p t}

denotes control variables at the household level, individual level and city level, respectively.

Referring to Liu et al. (2024), we additionally control for prefecture-level city fixed effects ( $ρ_{p}$ ) and year fixed effects ( $γ_{t}$ ) in equation (6) to mitigate the potential interference of unobservable factors and the fact that controlling for individual fixed effects may lead to insufficient levels of change in the core explanatory variables. $ε_{i p t}$ is an idiosyncratic error term.

To elucidate the moderating effect of Internet use and household income level, we interact these two moderating variables with extreme weather risks in equation (7):

{l n T o u r C o n s u m p}_{i p t} = α_{0} + \sum_{k = 1}^{4} β_{k} l n E x t r W_{k p t} + {φ M}_{i t} + \sum_{k = 1}^{4} δ_{k} (l n E x t r W_{k p t} \times M_{i t}) + β_{5} {C o n t r o l s}_{i p t} + ρ_{p} + γ_{t} + ε_{i p t}

(7)

where

M_{i t}

denotes either Internet use or household income level, contingent upon the moderation specification. Subscripts i, p, and t represent household, prefecture-level city, and year, respectively;

E x t r W_{k p t}

refers to each of the four extreme weather risk indexes (namely, extreme high temperature, extreme low temperature, extreme rainfall, and extreme drought). The remaining symbols are identical to those in equation (6).

Data sources and descriptive statistics

This study constructs an unbalanced panel for 2012–2022 by matching CFPS household microdata with prefecture-level city data. The dataset covers 10,339 households in 72 prefecture-level cities across diverse climate zones in eastern, central, and western China. The matched panel integrates individual and household attributes from CFPS with prefecture-city macro indicators, including extreme weather risk indices and socioeconomic controls, enabling a systematic assessment of how extreme weather risks affect household tourism consumption. Specifically, we utilized three types of data: (1) Extreme weather risks data are drawn from https://figshare.com/articles/dataset/Climate_Physical_Risk_Index_CPRI_/25562229/1. (2) CFPS data, which originates from the China Family Tracking Survey Project, executed by the China Social Science Survey Centre at Peking University (https://www.isss.pku.edu.cn/cfps/). We selected the survey items from 2012, 2014, 2016, 2018, 2020, and 2022, excluding 2010 due to the absence of the tourism consumption question. Subsequently, we considered the influence of potential factors such as age, marital status, and education level and matched the household questionnaire with the adult questionnaire using household codes to obtain relevant information. Finally, we excluded samples with missing identifying variables and responses of ‘don’t know’ or ‘not applicable,’ and restricted the age of the head of household to 16 years or older to ensure the representativeness of the survey. (3) Statistical data. GDP per capita and road areas are derived from the China City Statistical Yearbooks.

Empirical results

Baseline regression

Table 2 presents the estimated effects of extreme weather risks on household tourism consumption across regressions with and without control variables based on equation (6). The results consistently display that the extreme high temperature significantly increases household tourism consumption, the extreme rainfall decreases it, while extreme low temperature and extreme drought have no significant effects. The specification, which includes all control variables in column (2), indicates that a 1% increase in the extreme high temperature index and extreme rainfall are associated with a 0.118% increase and a 0.048% decrease in household tourism consumption, respectively.

Table 2.

Baseline regression results.

	lnTourConsump
	(1)	(2)
lnExtrLowTemp	−0.051	−0.028
lnExtrLowTemp	(0.063)	(0.061)
lnExtrHighTemp	0.122^*	0.118^*
lnExtrHighTemp	(0.065)	(0.065)
lnExtrRain	−0.055^***	−0.048^**
lnExtrRain	(0.02)	(0.019)
lnExtrDrought	−0.037	−0.031
lnExtrDrought	(0.046)	(0.045)
Controls	No	Yes
Year FE	Yes	Yes
City FE	Yes	Yes
Cons	1.554^***	−0.03
Cons	(0.275)	(1.239)
Observations	38207	35475
R²	0.144	0.248

Note. ***: p < 0.01, **: p < 0.05, *: p < 0.1. S.E. are given in parentheses. The same significance notation applies to Tables 3–6.

Robustness tests

Endogeneity test

Two mechanisms may introduce endogeneity into our estimates. First, unobserved factors may be correlated with both local extreme weather risk and household tourism consumption, inducing omitted variable bias. Second, there is mutual influence among household tourism consumption and the control variables at the micro level of the household, which may also lead to endogeneity. To address these issues, we utilize an instrumental variable method and two-stage least squares approach. Referring to De Haas et al. (2025), the instrumental variable is derived by the arithmetic average of the climate risk index for all other prefecture-level cities in China in year t, excluding the focal city. Column (1) of Table 3 reports the 2SLS estimation results. The weak-instrument and under-identification test statistics reported support the relevance and validity of the instrumental variables. Importantly, the coefficients on lnExtrHighTemp and lnExtrRain remain significantly positive and negative, respectively, implying that the baseline regression results are robust.

Table 3.

Endogeneity test and other robustness test results.

	lnTourConsump
	(1)	(2)	(3)	(4)
lnExtrLowTemp	−0.038	−0.020	−0.053	−0.028
lnExtrLowTemp	(0.067)	(0.062)	(0.065)	(0.092)
lnExtrHighTemp	0.201^**	0.136^*	0.139^**	0.118^**
lnExtrHighTemp	(0.083)	(0.077)	(0.063)	(0.058)
lnExtrRain	−0.073^***	−0.045^**	−0.049^***	−0.048^**
lnExtrRain	(0.024)	(0.020)	(0.018)	(0.021)
lnExtrDrought	−0.043	−0.011	0.002	−0.031
lnExtrDrought	(0.043)	(0.047)	(0.043)	(0.063)
Controls	Yes	Yes	Yes	Yes
Year FE	Yes	Yes	Yes	Yes
City FE	Yes	Yes	Yes	Yes
Kleibergen-Paap rk LM statistic	7159.231^***
Cragg-Donald Wald F statistic	2.2e + 04
Cons		−0.099	0.933	−0.030
		(1.246)	(1.165)	(1.832)
Observations	35475	35475	30135	35475
R²	0.116	0.247	0.184	0.248

Other robustness tests

We also perform some other robustness tests. First, following Fiechter et al. (2024), we apply winsorization to the variables to attenuate the impact of outliers on the regression outcomes. The dependent variable and explanatory variables are winsorized at the 1st and 99th percentiles, respectively. The regression results are presented in Column (2) in Table 3. The baseline regression results are evidently valid.

Second, given that the unique characteristics of municipalities may bias the regression results, we optimize sample selection by excluding the samples of Beijing, Tianjin, Shanghai, and Chongqing, and subsequently re-estimate equation (6) for a robustness test. The results are shown in Column (3) of Table 3. The empirical results remain significant after eliminating the influence from the municipality sample, confirming the robustness of the baseline regression results.

Third, we further cluster at the city level to account for within-city correlation in unobserved shocks. Column (4) in Table 3 indicates that the coefficients of lnExtrHighTemp and lnExtrRain remain significantly positive and negative, respectively, therefore confirming the robustness of baseline regression results.

Moderating effects analysis

Table 4 presents coefficient estimates regarding the moderating effects of Internet use and income level within a linear interaction approach. Before interpreting these estimates, it is useful to clarify the meaning of the weather-risk coefficients in the interaction specifications. In the baseline model, these coefficients represent the average impact of extreme weather risks on household tourism consumption, accounting for controls and fixed effects. When weather-risk factors are interacted with Internet use or income level, their coefficients cease to reflect unconditional average effects for the entire sample. They denote conditional marginal effects assessed at the reference value of the pertinent moderator. Consequently, variations in the signs of the weather-risk coefficients across columns should not be construed as estimates of a uniform unconditional influence, nor should they be regarded rigidly as indicators of instability. Table 4 is utilized to ascertain the direction and statistical significance of the moderating relationships, whereas the substantive interpretation is primarily based on the conditional marginal effects assessed across the observed distribution of the moderators, as indicated in the subsequent binning estimates.

Table 4.

Moderating effects analysis results.

	lnTourConsump
	(1)	(2)	(3)
lnExtrLowTemp × lnInternetUse	−0.003		0.006
lnExtrLowTemp × lnInternetUse	(0.048)		(0.051)
lnExtrHighTemp × lnInternetUse	0.095^*		0.095^*
lnExtrHighTemp × lnInternetUse	(0.055)		(0.057)
lnExtrRain × lnInternetUse	−0.104^***		−0.093^***
lnExtrRain × lnInternetUse	(0.016)		(0.016)
lnExtrDrought × lnInternetUse	0.033		0.021
lnExtrDrought × lnInternetUse	(0.035)		(0.036)
lnExtrLowTemp × lnincomelevel		−0.014	−0.018
lnExtrLowTemp × lnincomelevel		(0.037)	(0.039)
lnExtrHighTemp × lnincomelevel		0.003	−0.013
lnExtrHighTemp × lnincomelevel		(0.045)	(0.046)
lnExtrRain × lnincomelevel		−0.064^***	−0.044^***
lnExtrRain × lnincomelevel		(0.013)	(0.014)
lnExtrDrought × lnincomelevel		0.053^**	0.049*
lnExtrDrought × lnincomelevel		(0.027)	(0.028)
lnExtrLowTemp	0.120	0.123	0.100
lnExtrLowTemp	(0.254)	(0.078)	(0.260)
lnExtrHighTemp	−0.406	0.046	−0.427
lnExtrHighTemp	(0.284)	(0.076)	(0.290)
lnExtrRain	0.428^***	−0.075^***	0.374^***
lnExtrRain	(0.079)	(0.023)	(0.082)
lnExtrDrought	−0.130	0.031	−0.072
lnExtrDrought	(0.179)	(0.059)	(0.183)
lnInternetUse	0.401	0.577^***	0.379
lnInternetUse	(0.253)	(0.021)	(0.264)
lnincomelevel	0.314^***	0.368^*	0.399^*
lnincomelevel	(0.017)	(0.213)	(0.221)
Controls	Yes	Yes	Yes
Year FE	Yes	Yes	Yes
City FE	Yes	Yes	Yes
Cons	−2.966	−3.407^**	−2.532
Cons	(1.988)	(1.535)	(2.007)
Observations	28021	28021	28021
R²	0.268	0.267	0.271

With this interpretive caveat in mind, column (1) shows that the coefficient on extreme high temperature is negative, whereas its interaction with Internet use is positive, indicating that the marginal effect of extreme high temperature increases with Internet use and shifts from negative to positive. By contrast, the coefficient on extreme rainfall is positive, whereas its interaction with Internet use is strongly negative, indicating that the marginal effect of extreme rainfall declines as Internet use increases. Column (2) reports a significantly negative interaction between extreme rainfall and household income, suggesting that the adverse marginal effect of extreme rainfall is more pronounced among higher-income households. The positive drought–income interaction suggests that the conditional association between extreme drought and tourism consumption becomes more positive as income rises. Column (3) further includes all interaction terms simultaneously to assess the stability of these moderating patterns. The lnExtrHighTemp × lnincomelevel is statistically insignificant, even in the fully specified model. Consequently, we do not regard income as a robust moderator of the extreme high temperature effect and instead concentrate on the more consistent association between extreme rainfall and income.

Nonetheless, the interaction coefficients presented in Table 4 should be interpreted with caution, as traditional linear interaction models depend on the assumption of linear interaction effects and sufficient shared support throughout the moderator’s distribution. Hainmueller et al. (2018) underscore that an exclusive reliance on regression coefficients may mask varied moderation patterns and result in model-dependent findings. In accordance with Hainmueller et al. (2018), we utilize the binning estimator, which categorizes the moderator into low, medium, and high intervals to estimate the respective marginal effects within each interval. This method offers a more comprehensive diagnostic evaluation of the stability of the moderating effects presented in Table 4 across various ranges of the moderator and assesses the empirical plausibility of the linear interaction specification.

Figure 2 presents the binning estimates using Internet use as the moderator. At extreme high temperatures, the marginal effect of extreme high temperature increases with Internet use, transitioning from about zero at low Internet use to a significantly positive value at elevated Internet use. Conversely, the marginal effect of extreme rainfall diminishes with increased Internet use, transitioning from a somewhat favorable impact at low Internet use to a negative effect at mid and high levels.

Figure 2.

Binning estimation results for the moderating effect of Internet use.

Figure 3 displays the binning estimates with income level serving as the moderator. In accordance with the comprehensive interaction results, we concentrate on the extreme rainfall–income relationship rather than the extreme high temperature–income relationship, as the latter lacks statistical robustness. In the case of extreme rainfall, the marginal effect grows progressively negative with increasing income level, signifying that higher-income households diminish tourism consumption more significantly in the face of extreme rainfall.

Figure 3.

Binning estimation results for the moderating effect of income level.

Further analysis

Heterogeneity analysis

Considering significant differences in weather-risk adaptation and tourism consumption behaviors between married and unmarried households (Lin et al., 2015), Table 5 initially analyzes heterogeneity based on marital status. The findings suggest that the heterogeneity is primarily linked to extreme rainfall. In Column (1), the extreme rainfall coefficient for unmarried households is statistically insignificant, however, the interaction between extreme rainfall and marital status is strongly negative. This indicates that married households demonstrate a markedly greater negative reaction to extreme rainfall compared to unmarried households. Column (2) analyzes heterogeneity by employment sector, owing to substantial differences in weather-risk sensitivity, adaptation capacity, and consumption responses between agricultural and non-agricultural households (Aggarwal, 2021). Extreme rainfall markedly diminishes tourism consumption among non-agricultural households, but the positive interaction with agricultural employment indicates that this adverse impact is considerably less pronounced among agricultural households. Other interaction terms lack statistical significance.

Table 5.

Heterogeneity analysis results.

	lnTourConsump
	(1)	(2)
lnExtrLowTemp × mar	−0.084
lnExtrLowTemp × mar	(0.101)
lnExtrHighTemp × mar	0.202
lnExtrHighTemp × mar	(0.156)
lnExtrRain × mar	−0.150^**
lnExtrRain × mar	(0.061)
lnExtrDrought × mar	−0.142
lnExtrDrought × mar	(0.123)
lnExtrLowTemp × agri		−0.038
lnExtrLowTemp × agri		(0.073)
lnExtrHighTemp × agri		0.059
lnExtrHighTemp × agri		(0.093)
lnExtrRain × agri		0.084^***
lnExtrRain × agri		(0.030)
lnExtrDrought × agri		0.017
lnExtrDrought × agri		(0.058)
lnExtrLowTemp	0.045	−0.024
lnExtrLowTemp	(0.115)	(0.069)
lnExtrHighTemp	−0.056	0.087
lnExtrHighTemp	(0.149)	(0.083)
lnExtrRain	0.080	−0.095^***
lnExtrRain	(0.050)	(0.027)
lnExtrDrought	0.094	−0.036
lnExtrDrought	(0.129)	(0.057)
Controls ⁵	Yes	Yes
Year FE	Yes	Yes
City FE	Yes	Yes
Cons	−0.337	0.116
Cons	(2.059)	(1.257)
Observations	35475	35475
R²	0.249	0.252

Incremental consumption effect test

The preceding investigation has demonstrated that extreme high temperature positively affects household tourism consumption, while the specific characteristics of this effect regarding consumption structure remain unexamined. It is unclear if the promotion of extreme high temperature in household tourism consumption leads to an increase in tourism consumption (incremental consumption effect) or a substitution of non-tourism consumption (consumption substitution effect). Consequently, we investigate the effect of extreme high temperature on non-tourism consumption. Table 6 indicates that the regression coefficient of extreme high temperature on non-tourism consumption is insignificant, implying that the promotional effect is attributed to the tourism incremental consumption effect.

Table 6.

Consumption incremental effect test results.

	lnTourConsump	lnNonTourConsump
	(1)	(2)
lnExtrHighTemp	0.118^*	0.019
lnExtrHighTemp	(0.065)	(0.017)
Controls	Yes	Yes
Year FE	Yes	Yes
City FE	Yes	Yes
Cons	−0.030	9.862^***
Cons	(1.239)	(0.329)
Observations	35475	34612
R²	0.246	0.435

Note. Controls include lnExtrLowTemp, lnExtrRain, and lnExtrDrought.

Discussions

The asymmetric effect of extreme weather risks on household tourism consumption

Departing from established studies on tourist destinations, this paper investigates the impact of extreme weather risks on household tourism consumption from the underexplored perspective of the tourist origins. Our theoretical analysis posits that extreme weather risks influence household tourism consumption in the tourist origins. The baseline regressions empirically demonstrate a significant positive effect of extreme high temperature on household tourism consumption. This finding starkly contrasts with destination-focused studies (Atstāja and Cakrani, 2024), which indicate that extreme high temperature deters visitation by directly diminishing tourist experiences and posing health threats. This disparity highlights the importance of analytical perspective.

Destination-focused research investigates the immediate repercussions of extreme weather risks on tourist experiences. In contrast, this study reveals the decision-making logic of households may maintain or increase tourism consumption as an adaptive consumption response to local heat-related discomfort. Moreover, our results indicate that extreme rainfall inhibits household tourism consumption, as it obstructs transportation, rendering travel physically unfeasible. This corroborates the conclusions of Chen et al. (2024), affirming the detrimental impact of extreme rainfall on tourism.

The effects of extreme low temperature and extreme drought were statistically insignificant. In northern China, central heating significantly mitigates the threat of extreme low temperature to individuals’ bodily and mental well-being (Hu et al., 2022), thereby diminishing the motivation for tourism driven by extreme weather risks. Meanwhile, the persistent nature of extreme drought creates temporal delays that our models do not consider. Moreover, the tourism consumption dataset employed fails to differentiate among indoor and outdoor activities, thus obscuring subtle behavioral responses. Although extreme low temperature may inhibit general tourism, it could promote particular indoor tourism activities, such as hot springs.

The moderating effect of internet use and household income level

The Internet use-extreme high temperature interaction reflects an adaptive-capacity mechanism. At minimal levels of Internet use, extreme high temperature may diminish tourism consumption as households encounter heightened health risks, thermal discomfort, and trip unpredictability, in accordance with PMT (Rogers, 1975). As Internet use increases, households may obtain real-time data on meteorological conditions, transportation, lodging, indoor options, and off-peak arrangements, thereby enhancing response efficacy. Recent evidence indicates that digital infrastructure and Internet-based communication channels can enhance household tourism consumption by mitigating information asymmetry and transaction frictions (Yu et al., 2025).

The extreme rainfall effect turns negative as Internet use increases, reflecting an information-induced risk-recognition mechanism. Households with limited internet connectivity are likely to possess diminished access to prompt and comprehensive risk information, resulting in their tourism decisions being less influenced by real-time alerts regarding traffic disruptions, flooding, site closures, and cancellations of outdoor activities. As Internet use increases, extreme rainfall emerges as a more prominent and actionable risk signal, enhancing households’ risk assessment and prompting them to defer, cancel, or diminish tourism consumption, consistent with PMT and with evidence from Baños-Pino et al. (2023).

The income-based moderation of extreme rainfall can be explained by the transition from quantity-constrained to quality-sensitive tourism consumption. Lower-income households generally have more constrained tourism budgets, so rainfall may not substantially alter already limited tourism consumption. As income rises, however, tourism becomes more discretionary, experience-oriented and sensitive to service quality, comfort and travel reliability. Evidence on Chinese households shows that tourism consumption responds heterogeneously to income (Ma et al., 2024). Higher-income households are more likely to postpone, cancel or substitute tourism consumption when rainfall reduces expected travel utility, consistent with climate-sensitive tourism demand evidence (Zhou et al., 2024).

Drought is a gradual and cumulative environmental stressor. It may diminish local comfort, heighten perceived environmental stress, and induce psychological tension. Previous studies indicate that drought can function as a cumulative stressor affecting mental health and well-being, especially when aridity endures over an extended period (O'Brien et al., 2014). Simultaneously, involvement in tourism is significantly limited by household financial resources, with budgetary constraints serving as a primary obstacle to tourism engagement (Alegre et al., 2010). Consequently, lower-income households may react to drought-related uncertainty by conserving resources and prioritizing essential consumption, while higher-income households possess a stronger discretionary ability to transform drought-induced discomfort into tourism consumption.

Heterogeneous effect of extreme weather risks on household tourism consumption

The civil-status heterogeneity suggests that married households respond more adversely to extreme rainfall than unmarried households. Tourism consumption is closely shaped by household demographic structure and life-cycle stage (Lin et al., 2015), while household holidays often involve joint decision-making and coordination within couples (Mottiar and Quinn, 2004). Extreme rainfall may increase perceived travel risks, mobility constraints, and itinerary uncertainty. According to PMT, stronger threat appraisal and higher response costs can encourage protective avoidance. Thus, married households may be more inclined to postpone or reduce tourism consumption when extreme rainfall rises, which is also consistent with Jeuring (2017).

The employment-sector heterogeneity indicates that the negative tourism-consumption response to extreme rainfall is more evident among non-agricultural households, which is consistent with Aggarwal (2021). Agricultural households are routinely exposed to rainfall variability and often incorporate weather-related adjustment into livelihood management, production timing, and resource allocation. Therefore, extreme rainfall may be partly absorbed through existing coping and adaptation practices (Girard et al., 2021). By contrast, non-agricultural households’ tourism consumption is more likely to depend on scheduled leisure time, transport accessibility, and short-term mobility conditions. Extreme rainfall can therefore operate as a more direct constraint on their tourism consumption (Jeuring, 2017).

Consumption incremental effect

When faced with extreme high temperature, residents tend to spend more on tourism rather than converting their non-tourism consumption to tourism, according to the incremental effect test of consumption, which shows that extreme high temperature increases household tourism consumption. Non-tourism consumption is less vulnerable to extreme high temperature, especially when it comes to necessities (like food and clothes) and long-term investments (like housing, education, and transportation). There is an inherent connection between this varied response and the underlying socio-economic differences. As a result, wealthy households might use travel as a voluntary adaptive strategy to lessen the effects of extreme high temperature. Low-income households, on the other hand, constrained by discretionary cash and essential consumption, are unable to take advantage of this adaptive pathway through tourism consumption. This dynamic effectively redistributes consumption opportunities, disproportionately eroding the adaptive capacity of low-income households through the attrition of resilience (Hallegatte and Rozenberg, 2017).

Conclusions

Main findings

By providing insightful information for sustainable tourism and encouraging interdisciplinary discussion between tourism research and climate change economics, this study contributes to our understanding of temporary mobility under extreme weather risks from the tourist origins perspective. Theoretically, we establish a systematic analytical framework for weather risk-tourism consumption research based on the PMT. Utilizing CFPS and weather risk data from 2012 to 2022, we empirically confirm that extreme high temperature elevates household tourism consumption, extreme rainfall reduces it, while extreme low temperature and drought exert no significant influence. The robustness tests support the baseline findings. These divergent findings provide scientific underpinnings for the development of sustainable tourism in the context of climate uncertainty by exposing the significant asymmetric effects of extreme weather risks. We investigate moderating effects and determine that Internet use significantly moderates the impacts of extreme high temperature and extreme rainfall, while household income significantly moderates the effects of extreme rainfall. Further heterogeneity analyses show that extreme rainfall has a stronger adverse effect on married households and significantly reduces tourism consumption among non-agricultural households. Furthermore, the incremental-consumption test suggests that the positive effect of extreme high temperature on household tourism consumption stems from the incremental effect of tourism consumption.

Implications

In the context of extreme weather risks, the results of this study inform the optimization of resilience management and policy for tourism consumption in tourist origins. Specifically, the unequal impact of extreme weather risks, in particular, emphasizes the significance of developing a resilient tourism consumption system that prioritizes demand-side management at tourist origins in the context of growing climate uncertainty. National and local governments should integrate considerations of extreme weather risks into guidance on tourism consumption and safeguarding policies.

In regions with frequent extreme high-temperature events, strengthening the dissemination of early warnings about the carrying capacity of peak-season destinations is required. This entails enhancing safety and health precautions during periods of high temperature and urging residents to make sensible decisions about when and where to travel. Crucially, governments must address the related climatic inequality in light of the counterintuitive stimulation of tourism consumption by extreme high temperature. Low-income households may be unable to take advantage of these possibilities due to budgetary constraints, whereas wealthy households may freely do so. To guarantee fair participation in this consumption shift, targeted interventions are required, such as programs that provide affordable access to cooler areas or subsidized cooling-oriented tourism vouchers.

In regions with extreme rainfall frequencies, the focus should be on developing and promoting alternative tourism products to stabilize the tourism market and associated employment. Simultaneously, governments ought to emphasize married and non-agricultural households by enhancing their ability to withstand excessive rainfall via weather-related tourism insurance and adaptable leave provisions. Furthermore, it is crucial to leverage Internet platforms for the targeted dissemination of extreme weather risk warnings, alongside sustainable tourism advice and information on alternative products, to foster climate-adaptive consumption habits among residents.

In addition, in order to follow the effects of extreme weather risks and make appropriate adjustments to support policies and marketing plans, relevant tourism authorities and industry associations should set up dynamic monitoring and assessment procedures. As a way to help stakeholders adjust business operations and safeguard livelihoods in the context of extreme weather risks, it can be beneficial to foster interdepartmental collaboration across pertinent ministries, including civil affairs, culture and tourism, and meteorology. The markets for the consumption of origin tourism will remain stable and sustainable as a result.

Limitations

Our study also inevitably bears several limitations. We focus on Chinese households. The empirical results are grounded in China’s traditional social context. In subsequent research, future research could compare China with other countries and offer novel views on advancing the sustainable growth of tourism. In addition, data limitations confine this study to examining household tourism consumption responses to extreme weather risks. The CFPS data do not provide trip-level information on travel destinations, travel frequency, duration of stays, travel distance, or the climate-risk conditions of destinations. Therefore, we cannot identify whether changes in tourism consumption are associated with specific destination choices or spatial adjustments in travel behavior. Future research should link household survey data with trip-level data, such as mobile location traces or booking records, to provide a more detailed understanding of how extreme weather risks affect household tourism consumption. A further limitation concerns the measurement of Internet use. Due to data limitations, we approximate Internet use by monthly household communication expenditures, encompassing costs for telephone services, mobile communication, internet access, and postal services. This metric cannot accurately reflect online information-seeking behavior or the intensity of Internet usage. Subsequent study employing more direct markers, such as Internet accessibility or online search behavior, may yield a more accurate evaluation of the proposed mechanism.

Footnotes

ORCID iDs

Rui Zhang

Yun Tong

Ya-Fei Liu

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: The Philosophy and Social Sciences Research Project of the Education Department of Hubei Province (24Q088), The National Social Science Fund of China (25BJY153), The Philosophical and Social Sciences Planning Project of Hainan Province (HNSK(ZC)25-186), The Natural Science Foundation of Hainan Province (725RC741), and the Project of Higher Education Scientific Research of Hainan Province (No. Hnky2024-5).

Declaration of conflicting interests

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

Notes

Author biographies

Rui Zhang, Ph.D., Lecturer, Hubei University of Chinese Medicine. Her research focuses on tourism economics and climate change. She has published over ten papers on reputable journals.

Yun Tong, Ph.D., Professor, Hainan University. His research focuses on tourism economics and sustainable tourism. He has published over twenty papers on reputable journals.

Ya-Fei Liu, Ph.D., Lecturer, Hainan University. Her research focuses tourism economics. She has published over ten papers on leading journals like TM, ATR.

References

Aggarwal

(2021) Impacts of climate shocks on household consumption and inequality in India. Environment and Development Economics 26: 488–511. https://doi.org/10.1017/s1355770x20000388

Alegre

Mateo

Pou

(2010) An analysis of households’ appraisal of their budget constraints for potential participation in tourism. Tourism Management 31: 45–56. https://doi.org/10.1016/j.tourman.2009.02.004

Atstāja

Cakrani

(2024) Impact of climate change on international tourism evidence from Baltic sea countries. Sustainability 16(12): 5203. https://doi.org/10.3390/su16125203

Baños-Pino

Boto-García

Del Valle

, et al. (2023) Is visitors’ expenditure at destination influenced by weather conditions? Current Issues in Tourism 26(10): 1554–1572. https://doi.org/10.1080/13683500.2022.2058468

Becken

Wilson

(2013) The impacts of weather on tourist travel. Tourism Geographies 15(4): 620–639. https://doi.org/10.1080/14616688.2012.762541

Birchenough

(2017) Impacts of climate change on biodiversity in the coastal and marine environments of Caribbean small island developing states (SIDS). Caribbean Marine Climate Change Report Card: Science Review 40–51.

Brida

Scuderi

(2013) Determinants of tourist expenditure: a review of micro econometric models. Tourism Management Perspectives 6: 28–40. https://doi.org/10.1016/j.tmp.2012.10.006

Cai

Zhang

(2023) Population aging and household tourism consumption—an empirical study based on china family panel studies (CFPS) data. Sustainability 15: 9989. https://doi.org/10.3390/su15139989

Chen

Yang

, et al. (2023) Association between ambient temperatures and injuries: a time series analysis using emergency ambulance dispatches in Chongqing, China. Environmental Health and Preventive Medicine 28: 28. https://doi.org/10.1265/ehpm.22-00224

10.

Chen

Mao

Huang

(2024) Age structure of the population and household consumption expenditure on tourism. Finance Research Letters 60: 104896. https://doi.org/10.1016/j.frl.2023.104896

11.

Crawford

Godbey

(1987) Reconceptualizing barriers to family leisure. Leisure Sciences 9: 119–127. https://doi.org/10.1080/01490408709512151

12.

Crawford

Jackson

Godbey

(1991) A hierarchical model of leisure constraints. Leisure Sciences 13: 309–320. https://doi.org/10.1080/01490409109513147

13.

De Haas

Martin

Muûls

, et al. (2025) Managerial and financial barriers to the green transition. Management Science 71: 2890–2921. https://doi.org/10.1287/mnsc.2023.00772

14.

Deng

Zhao

(2022) Retirement and household tourism consumption—A case study in China. Tourism Economics 29: 1055–1073. https://doi.org/10.1177/13548166221090170

15.

Dong

Quan

Song

, et al. (2022) Influencing factors of empty nest family tourism consumption. Journal of Travel & Tourism Marketing 39: 290–304. https://doi.org/10.1080/10548408.2022.2089951

16.

Dube

Nhamo

(2018) Climate variability, change and potential impacts on tourism: evidence from the Zambian side of the Victoria falls. Environmental Science & Policy 84: 113–123. https://doi.org/10.1016/j.envsci.2018.03.009

17.

Elsasser

Bürki

(2002) Climate change as a threat to tourism in the alps. Climate Research 20: 253–257. https://doi.org/10.3354/cr020253

18.

Fang

Liu

Yin

, et al. (2023) Heat exposure intervention, anxiety level, and multi-omic profiles: a randomized crossover study. Environment International 181: 108247. https://doi.org/10.1016/j.envint.2023.108247

19.

Fiechter

Landsman

Peasnell

, et al. (2024) Do industry-specific accounting standards matter for capital allocation decisions? Journal of Accounting and Economics 77(2-3): 101670. https://doi.org/10.1016/j.jacceco.2023.101670

20.

Fuchs

Efrat-Treister

Westphal

(2024) When, where, and with whom during crisis: the effect of risk perceptions and psychological distance on travel intentions. Tourism Management 100: 104809. https://doi.org/10.1016/j.tourman.2023.104809

21.

Girard

Delacote

Leblois

(2021) Agricultural households’ adaptation to weather shocks in Sub-Saharan Africa: implications for land-use change and deforestation. Environment and Development Economics 26(5–6): 538–560. https://doi.org/10.1017/s1355770x2000056x

22.

Godbey

Crawford

Shen

(2017) Assessing hierarchical leisure constraints theory after two decades. Journal of Leisure Research 42: 111–134. https://doi.org/10.1080/00222216.2010.11950197

23.

Gössling

Scott

Hall

, et al. (2012) Consumer behaviour and demand response of tourists to climate change. Annals of Tourism Research 39: 36–58. https://doi.org/10.1016/j.annals.2011.11.002

24.

Guo

Zhang

(2024) A dataset to measure global climate physical risk. Data in Brief 54: 110502. https://doi.org/10.1016/j.dib.2024.110502

25.

Hainmueller

Mummolo

(2018) How much should we trust estimates from multiplicative interaction models? Simple tools to improve empirical practice. Political Analysis 27(2): 163–192. https://doi.org/10.1017/pan.2018.46

26.

Hallegatte

Rozenberg

(2017) Climate change through a poverty lens. Nature climate Change 7(4): 250–256. https://doi.org/10.1038/nclimate3253

27.

Hamilton

Maddison

Tol

RSJ

(2005) Effects of climate change on international tourism. Climate Research 29: 245–254. https://doi.org/10.3354/cr029245

28.

Hewer

Scott

Gough

(2015) Differential weather preferences for tourist and recreationists. Tourism Management 46: 266–283.

29.

Gong

Yin

, et al. (2022) Central heating and winter mortality in China: a national study based on 364 Chinese locations. Urban Climate 41: 101045. https://doi.org/10.1016/j.uclim.2021.101045

30.

Huang

Zhan

Huang

, et al. (2022) How commuting time influences hedonic consumption: the role of perceived stress. Journal of Consumer Behaviour 22: 439–454. https://doi.org/10.1002/cb.2091

31.

Hubbard

Mannell

(2001) Testing competing models of the leisure constraint negotiation process in a corporate employee recreation setting. Leisure Sciences 23: 145–163. https://doi.org/10.1080/014904001316896846

32.

Hübner

Gössling

(2012) Tourist perceptions of extreme weather events in Martinique. Journal of Destination Marketing & Management 1(1-2): 47–55. https://doi.org/10.1016/j.jdmm.2012.09.003

33.

Hung

W-T

Shang

J-K

Wang

F-C

(2013) A multilevel analysis on the determinants of household tourism expenditure. Current Issues in Tourism 16: 612–617. https://doi.org/10.1080/13683500.2012.725714

34.

IPCC (2014) Summary for policymakers. In: B

Barros

Dokken

, et al. (eds) Climate Change 2014: Impacts, Adaptation, and Vulnerability. Part A: Global and Sectoral Aspects. Contribution of Working Group II to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change. Cambridge University Press, 1132.

35.

Jang

Ham

(2009) A double-hurdle analysis of travel expenditure: baby boomer seniors versus older seniors. Tourism Management 30: 372–380. https://doi.org/10.1016/j.tourman.2008.08.005

36.

Jeuring

JHG

(2017) Weather perceptions, holiday satisfaction and perceived attractiveness of domestic vacationing in the Netherlands. Tourism Management 61: 70–81. https://doi.org/10.1016/j.tourman.2017.01.018

37.

Juschten

Brandenburg

Hössinger

, et al. (2019) Out of the city heat-way to less or more sustainable futures? Sustainability 11(1): 214. https://doi.org/10.3390/su11010214

38.

Kaenzig

Rebetez

Serquet

(2016) Climate change adaptation of the tourism sector in the Bolivian Andes. Tourism Geographies 18: 111–128. https://doi.org/10.1080/14616688.2016.1144642

39.

Karditsa

Niavis

Paramana

, et al. (2024) Is the insular coastal tourism of Western Greece at risk due to climate induced sea level rise? Ocean & Coastal Management 251: 107088. https://doi.org/10.1016/j.ocecoaman.2024.107088

40.

Köberl

Prettenthaler

Bird

(2016) Modelling climate change impacts on tourism demand: a comparative study from Sardinia (Italy) and cap bon (Tunisia). Science of The Total Environment 543: 1039–1053. https://doi.org/10.1016/j.scitotenv.2015.03.099

41.

Lin

Mao

Song

(2015) Tourism expenditure patterns in China. Annals of Tourism Research 54: 100–117. https://doi.org/10.1016/j.annals.2015.07.001

42.

Lin

Qin

, et al. (2020) Determinants of Chinese households' tourism consumption: evidence from China family panel studies. International Journal of Tourism Research 23: 542–554. https://doi.org/10.1002/jtr.2425

43.

Lin

Jiang

, et al. (2023) Impacts of risk aversion on tourism consumption: a hierarchical age-period-cohort analysis. Annals of Tourism Research 101: 103607. https://doi.org/10.1016/j.annals.2023.103607

44.

Liu

(2022) The effect of temperature on outdoor recreation activities: evidence from visits to federal recreation sites. Environmental Research Letters 17(4): 044037. https://doi.org/10.1088/1748-9326/ac5693

45.

Liu

Shi

Wang

, et al. (2024) Social networks and household consumption. Finance Research Letters 60: 104958. https://doi.org/10.1016/j.frl.2023.104958

46.

López-Dóriga

Jiménez

Valdemoro

, et al. (2019) Impact of sea-level rise on the tourist-carrying capacity of Catalan beaches. Ocean & Coastal Management 170: 40–50. https://doi.org/10.1016/j.ocecoaman.2018.12.028

47.

Qiao

Han

(2022) Interpreting the humanistic space of urban-rural interface using consumption behaviors. Journal of Rural Studies 93: 513–521. https://doi.org/10.1016/j.jrurstud.2019.12.014

48.

Zhou

Abdulai

, et al. (2024) Household income and tourism expenditure: an unconditional quantile regression approach. Applied Economics 56(50): 6144–6157. https://doi.org/10.1080/00036846.2023.2267823

49.

Mottiar

Quinn

(2004) Couple dynamics in household tourism decision making: women as the gatekeepers? Journal of Vacation Marketing 10(2): 149–160. https://doi.org/10.1177/135676670401000205

50.

Nunes

PALD

Loureiro

(2016) Economic valuation of climate‐change‐induced vinery landscape impacts on tourism flows in Tuscany. Agricultural Economics 47: 365–374. https://doi.org/10.1111/agec.12236

51.

OBrien

Berry

Coleman

, et al. (2014) Drought as a mental health exposure. Environmental Research 131: 181–187. https://doi.org/10.1016/j.envres.2014.03.014

52.

Rogers

(1975) A protection motivation theory of fear appeals and attitude Change1. J Psychol 91: 93–114. https://doi.org/10.1080/00223980.1975.9915803

53.

Rogers

(1983) Cognitive and physiological processes in fear appeals and attitude change: a revised theory of protection motivation. In: Cacioppo

Petty

(eds) Social Psychophysiology: A Sourcebook. Guilford Press, pp. 153–176.

54.

Rosselló-Nadal

Riera-Font

Cárdenas

(2010) The impact of weather variability on British outbound flows. Climatic Change 105: 281–292. https://doi.org/10.1007/s10584-010-9873-y

55.

Scott

Gössling

Hall

(2012) International tourism and climate change. Wiley Interdisciplinary Reviews: Climate Change 3(3): 213–232. https://doi.org/10.1002/wcc.165

56.

Scott

Steiger

Knowles

, et al. (2019) Regional ski tourism risk to climate change: an inter-comparison of Eastern Canada and US Northeast markets. Journal of Sustainable Tourism 28: 568–586. https://doi.org/10.1080/09669582.2019.1684932

57.

Smith

Fitchett

(2020) Drought challenges for nature tourism in the sabi sands game reserve in the eastern region of South Africa. African Journal of Range & Forage Science 37: 107–117. https://doi.org/10.2989/10220119.2019.1700162

58.

Sönmez

Graefe

(1998) Determining future travel behavior from past travel experience and perceptions of risk and safety. Journal of Travel Research 37(2): 171–177. https://doi.org/10.1177/004728759803700209

59.

Spencer

Strobl

Campbell

(2022) Sea level rise under climate change: implications for beach tourism in the Caribbean. Ocean & Coastal Management 225: 106207. https://doi.org/10.1016/j.ocecoaman.2022.106207

60.

Steiger

Stötter

(2013) Climate change impact assessment of ski tourism in Tyrol. Tourism Geographies 15: 577–600. https://doi.org/10.1080/14616688.2012.762539

61.

Steiger

Posch

Tappeiner

, et al. (2020) The impact of climate change on demand of ski tourism - a simulation study based on stated preferences. Ecological Economics 170: 106589. https://doi.org/10.1016/j.ecolecon.2019.106589

62.

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. https://doi.org/10.1016/j.scitotenv.2019.06.415

63.

Sun

Qian

Liu

, et al. (2022) Social network and tourism consumption by households: evidence from China. PLoS One 17: e0275418. https://doi.org/10.1371/journal.pone.0275418

64.

Susanto

Zheng

Liu

, et al. (2020) The impacts of climate variables and climate-related extreme events on island country’s tourism: evidence from Indonesia. Journal of Cleaner Production 276: 124204. https://doi.org/10.1016/j.jclepro.2020.124204

65.

Verkoeyen

Nepal

(2019) Understanding scuba divers’ response to coral bleaching: an application of protection motivation theory. Journal of Environmental Management 231: 869–877. https://doi.org/10.1016/j.jenvman.2018.10.030

66.

Wang

Zhou

(2019) Integrated impacts of climate change on glacier tourism. Advances in Climate Change Research 10: 71–79. https://doi.org/10.1016/j.accre.2019.06.006

67.

Wang

Fang

, et al. (2022a) Escape from air pollution: how does air quality in the place of residence shape tourism consumption? Tourism Economics 29: 1074–1099. https://doi.org/10.1177/13548166221091749

68.

Wang

Xie

, et al. (2022b) A review of the impacts of climate change on tourism in the arid areas: a case study of Xinjiang Uygur autonomous region in China. Advances in Meteorology 2022: 1–11.

69.

Wang

Cheng

Chen

(2025) Family digital competence, residents' subjective well-being, and tourism consumption levels—A study based on CFPS data. International Review of Financial Analysis 102: 104114. https://doi.org/10.1016/j.irfa.2025.104114

70.

Wei

Huang

(2025) Impact of high temperatures on tourist flows in urban and rural areas: climate adaptation strategies in China. Agriculture 15: 980. https://doi.org/10.3390/agriculture15090980

71.

Wilkins

de Urioste-Stone

Weiskittel

, et al. (2018) Effects of weather conditions on tourism spending: implications for future trends under climate change. Journal of Travel Research 57(8): 1042–1053. https://doi.org/10.1177/0047287517728591

72.

Xiang

Gretzel

(2010) Role of social media in online travel information search. Tourism Management 31(2): 179–188. https://doi.org/10.1016/j.tourman.2009.02.016

73.

Yang

Zhang

Wang

(2023) The impact of the internet on household consumption expenditure: an empirical study based on China family panel studies data. Economic Research-Ekonomska Istraživanja 36(3): 2150255.

74.

Fan

Chen

, et al. (2025) An analysis of the impact and mechanism of new digital infrastructures on travel consumption behaviours of middle-aged and elderly households in China. Current Issues in Tourism 28(9): 1513–1530. https://doi.org/10.1080/13683500.2024.2343085

75.

Zarezadeh

Benckendorff

Gretzel

(2023) Online tourist information search strategies. Tourism Management Perspectives 48: 101140. https://doi.org/10.1016/j.tmp.2023.101140

76.

Zheng

Luo

Ritchie

(2021) Afraid to travel after COVID-19? Self-protection, coping and resilience against pandemic ‘travel fear. Tourism Management 83: 104261. https://doi.org/10.1016/j.tourman.2020.104261

77.

Zhou

Faturay

Driml

, et al. (2024) Meta-analysis of the climate change-tourism demand relationship. Journal of Sustainable Tourism 32(9): 1762–1783. https://doi.org/10.1080/09669582.2024.2354882