Meta-analysis of the climate change-tourism demand relationship

Abstract

As climate change urgency intensifies, understanding its impact on destinations and the development of adaptation strategies becomes critical for sustainable tourism. This article, through a meta-analysis of 290 climate change elasticity estimates in tourism from 34 studies, examines the influence of climate factors and tourists’ adaptive behaviours on tourism demand. It also projects the situations countries will face by 2050 under ongoing climate change scenarios. Findings suggest that low- and middle-income economies and Small Island Developing States, despite their minimal carbon contributions, will face significant tourism losses. In contrast, high-income countries might see increased travel demand and lead to more emissions. This will aggravate climatic conditions and inflict greater losses on less-developed communities. This finding exemplifies a vicious cycle of climate injustice, highlighting the disparities in climate-related social costs globally.

Keywords:

• Global warming
• climate change
• tourism demand
• meta-analysis
• social cost of carbon
• climate justice

Introduction

Climate change presents an urgent and complex challenge to humankind, narrowly defined as the long-term shifts in temperature and other weather patterns (NASA, 2023; UN, 2023). The problem has resulted in diverse climate concerns worldwide. Europe, for instance, has experienced record sunshine hours and rapid temperature increases, leading to more frequent heatwaves (Copernicus, 2020). In contrast, decreased atmospheric humidity and higher daytime radiative forcing are making the Mediterranean region and southern Africa increasingly arid, heightening the risk of extreme droughts. Meanwhile, regions like Southeast Asia and Eastern North America are witnessing notable increases in heavy precipitation, making them more susceptible to flooding and rising sea levels (IPCC, 2007).

With these shifts, policymakers in the tourism sector are increasingly encouraged to consider the impacts of changing climates in developing resilient and sustainable tourism systems. Noteworthy initiatives, such as the European Union’s roadmap for a twin (green and digital) transition (European Commission, 2022), the 2023 Tourism and Climate Change Stocktake report (TPCC, 2023) and WTTC connecting global climate actions (WTTC, 2015), stress the importance of incorporating climate measures into tourism strategies at various governance levels. However, only 18% of Nationally Determined Contributions (NDCs) include tourism specifically, suggesting tourism policy is not yet integrated with global and national climate change frameworks (TPCC, 2023). Considering that tourism makes considerable contributions to and is substantially impacted by climate change, the observed gaps in tourism policy need to be addressed urgently.

To improve the integration in policy documents, it is fundamental to understand the ramifications of climate change on the tourism industry. Much of our understanding has been generated through econometric analysis aimed at identifying the sensitivity of tourism demand to climate. However, as previous systematic reviews (Gössling et al., 2012; Hall & Higham, 2005; Scott et al., 2012) highlighted, there are notable knowledge gaps in this field. Specifically, we found (see Methods section) that a significant proportion of the literature focuses on one area (country, province or region). While these individual studies provide detailed insights into how climate fluctuation affects tourism in particular places, this approach is characterized by significant variations and fractional information, leaving a gap in providing a comprehensive understanding of global tourism dynamics. For instance, despite both having tropical climates, Mauritius has seen a decline in tourism due to rising temperatures, while Hainan Island, China, experienced increased arrivals (F. Chen et al., 2017; Fauzel, 2020). Moreover, the lack of integrated assessments of ‘multifarious impacts’ has been identified as a major gap in the tourism sector as far back as the publication of the Davos report 15 years ago (World Economic Forum, 2008). The general view that colder regions will benefit from global warming with an increased demand of tourism, while hotter regions face a reduced volume of visitors (Amelung et al., 2007; Bigano et al., 2005; Hamilton et al., 2005; Matei et al., 2023) generally overlooks the multiple climate risks and the diverse tourist responses.

In terms of the former, a paradigm example is found in studies where the previously confirmed positive correlations between temperature and tourism demand are reversed when factors such as precipitation and relative humidity are considered (Bae & Nam, 2020; Baig et al., 2021; Seetanah & Fauzel, 2019). Besides, demand factors, defined as how tourists respond and adapt to changing climates, further complicate the ‘climate change-tourism demand’ dynamics. A great variation in the adaptability and responsiveness of visitors to climate fluctuations was observed. Studies, such as those by Nunes (2013), Agnew and Palutikof (2006) and Rutty and Scott (2016), posit that domestic tourists exhibit greater sensitivity to climatic alterations, while Falk (2013) contends a more pronounced responsiveness among international visitors. Similar divergences also exist when contrasting visitor volumes against durations of stay. For instance, elevated temperatures may hinder visitor arrivals but paradoxically enhance the duration of overnight stays (Baig et al., 2021; Fauzel, 2020). Furthermore, demand responses may not be consistent between sub-national destinations, with implications on visitors’ spatial substitutions, leading to uncertainty about the overall impact on the country as a whole (Nunes, 2013; Vuković Darko, 2015).

It is clear that existing case studies have insufficiently investigated the tourism demand responses to climate change, an issue that has been prioritised in the most recent tourism and climate change synthesis (Scott & Gössling, 2022), the Australian intergenerational report (Australian Treasury, 2023) and the 2023 Tourism and Climate Change Stocktake report (TPCC, 2023). In response to the urgency, this study provides the first meta-analysis on how changes in climate and associated meteorological conditions impact global travel demand dynamics. By systematically analysing 34 studies with 290 estimates, we specifically aim to identify how a destination’s climate factors (climate zone and destination type) and demand factors (visitor segment, trip characteristics, temporal substitution, regional substitution and the time effect) influence the interplay of climate change and tourism. Grounded on the analysis, we further forecast the global tourism redistribution under future climate scenarios to 2050, revealing the worldwide tourism demand changes should the climate effect continue to evolve. By doing so, this study contributes to the literature by pinpointing varied destination typologies that are critical in determining the beneficiaries and victimizers in travel demand under the impact of climate change, signalling destinations that are relatively benefited/damaged and the extent to which communities are likely to experience. These findings advance knowledge of the climate change-travel demand nexus, uncovering the role of tourism as an impediment to climate justice, and providing a monetized basis for future work on estimating the social cost of carbon (SCC) in the tourism sector.

Conceptual framework

The interplay between climate and travel demand can be understood through the combination of the Push and Pull factor theory (Lee, 1966) and the theory of demand (Hicks, 1986). The former states that pull factors are those present in the destinations, while push factors are from the origin countries (Lee, 1966). While it is acknowledged that origin’s climate has influenced travel demand (F. Chen et al., 2017), this push factor was rarely studied while the majority of empirical studies focus on destination climate in their modelling. We therefore construct the conceptual framework by primarily considering climate as a pull factor from the perspective of destinations. According to the law of demand theory, a change in demand refers to a shift in the demand curve caused by changes in non-price factors such as tastes, demographics and incomes (Hicks, 1986). In this context, climate, a non-price factor, can positively or negatively affect tourism demand by altering destinations’ natural attributes, weather events, thermal comfort and the feasible tourism activities allowed. Thus, a destination’s climate acts as either a negative pull factor in some places that hinders tourist numbers, shifting the demand curve to the left, or a positive pull factor that attracts tourists to certain areas, shifting the demand curve to the right (F. Chen et al., 2017; Scott & Lemieux, 2010).

Visitors are heterogeneous, with each coming with different motivations, perceptions and preferences towards a destination’s climate attributes (Scott & Lemieux, 2010). Under climate change, these inherent tourists’ psychological factors affect their willingness to travel, destination risk perception, satisfaction and ultimately their actual travel behaviours (Gössling & Hall, 2006; Scott et al., 2012). Translating this pattern into demand theory implies that different visitor groups may exhibit unique elasticity (responses) to climate change, with some being sensitive (elastic) and others not (inelastic). It is important to note that any changes in travel behaviours can be seen as adaptations, as tourists have the largest adaptive capacity to determine where to go, when to go, and how long to stay before travelling, en route, and at the destination (Scott et al., 2012). Each adaptation strategy is simultaneously moderated by temporal-spatial factors. For example, if the climate at destination A is unfavourable, will I substitute this with a trip to destination B? Similarly, if the climate at destination A is unfavourable this month, can I visit it at another time? Visitor adaptation strategies could create a ripple effect on demand for multiple regions at various time points. Considering these temporal-spatial factors helps to elucidate how the climate elasticity of tourism demand is being shaped.

Under climate change, we propose that tourists are strongly influenced by climate factors and demand factors. We expect that climate factors, including climate zones, destination types and warming indicators, will shift tourism demand curves (Figure 1A); Demand factors will then modify the slope of the demand curve (elasticity) (Figure 1B), encompassing visitor segments (domestic versus international tourists), trip behaviours (visits versus length of stay [LOS]), regional substitution, seasonal variations, and time trends. The interactions between climate and demand factors will collectively determine the new tourism demand (Figure 1C). For regions that benefit from climate change, tourism demand will increase (shift to the right) and visitors will be less sensitive to price changes, resulting in increased tourism revenue (D1, Figure 1C). In contrast, regions that suffer will experience a decrease in demand (shifting to the left) while visitors are price-sensitive and have low loyalty to the region. The likely outcome is decreased tourism revenue for the destination (D2, Figure 1C).

Figure 1. Conceptual framework.

Tourism climate factors

To systematically classify the tourism-demand pattern, we argue that climate zone and destination types play critical roles in shaping these dynamics.

Climate zone

Destinations in the same climate zone share similar climatic characteristics and are prone to similar climate exposures to some extent, empowering the climate zone as an appropriate reference for generalizing demand change patterns. According to the Köppen climate classification system, each region in the world can be assigned to one of the five climate groups: tropical, arid, temperate, continental, and polar zones¹ (Arnfield, 2023). While mid- and high-latitude regions are not immune to climate change, with climate risks including glacier retreat, reduced snowpack, increased solar radiation, etc., our literature review shows that tropical and arid regions are by far the most severely affected (IPCC, 2022; Meredith et al., 2019). The number of abnormally hot days in the tropics where interannual temperature variability is the lowest has significantly increased, making tourism activities challenging, and the rainfall shortage in arid regions further diminishes the appeal of local settings (e.g. reduced forest cover and shrub biomes), and exacerbates freshwater stress (Byrne, 2021; Mahlstein et al., 2011; Saverimuttu & Varua, 2014). Additionally, extreme weather events emerge at these lower latitudes earlier and more frequently compared to their high-latitude counterparts, such as longer-lasting droughts and larger-scale wildfires in arid climates, as well as more frequent heat waves, floods, and cyclones in tropical climates (Mahlstein et al., 2011). Overall, we surmise that the relative appeal of tropical and arid regions may decline due to deteriorating climatic suitability and rising travel risks, while temperate and continental regions are likely to benefit from taking on shifted tourism demand.

Destination type

Within each climate zone, different destination types face unique climate change patterns. Mountain and coastal areas are the commonly used classification that ties with the areas of highest demand (Gössling et al., 2012).

Destination type – Mountain

There is disagreement over whether mountain tourism would gain from climate change in travel demand. Some posit that mountains tend to be cooler than urban areas, and therefore warmer weather and longer summer seasons may significantly enhance the appeal of mountain summer activities. Examples such as higher temperatures were found to have an encouraging impact on Austrian summer glacier skier visits (Mayer et al., 2018), and extraordinary hot weather in urban heat islands led to an increase in overnight stays in Swiss Alpine resorts (Serquet & Rebetez, 2011). The opposing argument is that most mountain activities are highly impacted by unfavourable weather conditions. Mountain travellers normally prioritise the absence of rain (Steiger et al., 2022), whereas these regions are currently experiencing intense precipitation under climate change, with higher risks of mass movements (e.g. landslides and debris flows) (Baig et al., 2021). In addition, snow reliability remains a critical indicator of snow mountain activities. Without snowmaking, 53% and 98% of 2234 ski resorts in 28 European nations are estimated to be in very high danger of snow shortages under 2 °C and 4 °C global warming scenarios (François et al., 2023). While snowmaking is expected to alleviate tourists’ sensitivity to warming, it requires certain conditions (e.g. low temperature, sufficient water, and high energy consumption) that are associated with high operational difficulties and maintenance costs. In short, climate change can create either positive or negative effects on mountain tourism, making mixed observations and requiring meta-analysis to examine.

Destination type – Coasts

Global warming is anticipated to have positive implications for coastal tourism. Coastal locations warm slower than inland locations (IPCC, 2022), serving as a shelter from the extreme heat of the day. Notably, coastal travellers exhibit a greater tolerance for heat as seaside activities, like sunbathing, swimming and paddling, all require much warmer climates, and therefore coastal travellers have a higher tolerance for high temperature and ample sunshine (Coombes & Jones, 2010). According to Lise and Tol (2002), tourists generally prefer an average daily temperature of 21 °C, while beach vacationers show higher temperature acceptance, considering 26.8 °C (Scott et al., 2007), 27–30 °C (international segment) (Rutty & Scott, 2016) and 30 °C (Rutty & Scott, 2013) as the most favourable. Coastal tourism is also certainly not immune to climate change, while impacts (i.e. coral reef bleaching and beachline erosion) may not yet have significantly dampened demand. For instance, the state of coral reefs was found to be largely irrelevant to divers and snorkelers in Mauritius, as long as a certain threshold level (e.g. visibility and variety) was not exceeded (Gössling et al., 2008). This is consistent with the recent findings of Verkoeyen and Nepal (2019) on scuba divers’ responses to coral bleaching, which confirmed that changing diving activities was the least attractive adaptive option. Thus, the growing climatic suitability as well as the resilient demand of waterfront activities may moderate climate change impacts.

Here, we propose two hypotheses (H1–2) regarding how multiple climatic factors shift tourism demand.

H1. Climate zone: Tourism demand redistributes from tropical and arid zones (demand decreases) to temperate and continental zones (demand increases).

H2. Destination type: It will be determined how climate change will affect mountain tourism (2a), and demand for coastal tourism is expected to grow (2b).

Tourism demand factors

Five demand-side characteristics are proposed to explain the tourism redistributive pattern, including visitor segments, trip behaviour, regional substitution effect, seasonality and time trends.

Visitor segment

Domestic and international travellers exhibit notable disparities in their access to climate information, adaptability in altering trip plans, travel motivations and climate preferences. Domestic visitors are generally more aware of weather/natural variations and more adaptable given their easier access to local media, greater word-of-mouth from friends, and shorter travel planning periods (Cai et al., 2011; Falk, 2013; Nunes, 2013). Further, in terms of travel incentives, studies suggest domestic travellers were primarily driven by a desire for nature-based resorts, which are more dependent on the climate conditions, whereas foreign tourists were usually drawn to historical and cultural sites, which are more resilient to changing climates (Awaritefe, 2004; Cai et al., 2011). Even in severe weather, such as high temperatures, prolonged rainfall and strong winds, the international market demonstrates greater resilience that is markedly different from the local market (Rutty & Scott, 2016). We therefore assume that domestic visitors are more climate-sensitive than their foreign counterparts.

Trip behaviour

The number of tourists and the number of overnight stays are the two most frequently used demand measures in climate change-tourism demand studies (C.-M. Chen et al., 2017; Falk, 2014). In demand theory (Hicks, 1986), demand refers to the willingness of consumers to purchase a certain product/service, and in this context, whether demand is measured by visits or by nights spent should yield similar results. However, some studies reviewed show distinct responses, including an overall reduction in visits while observing an increase in the share of long-stay demand (Baig et al., 2021; F. Chen et al., 2017; Fauzel, 2020; Nunes, 2013). One explanation may be the complexity of the travel decision-making process, that is, the sequential consideration of the decision to travel, the choice of places and the decision on duration of stay (Dellaert et al., 1998). Whether to travel and where to travel to are mainly determined by income, prices, political stability, infrastructural (e.g. transport) and environmental (e.g. landscape, weather) factors (Bausch et al., 2021). Whereas the visitor nights reflect not only tourist volumes but the LOS, thus providing a nuanced picture of visitor engagement. As a subsequent stage, LOS is closely related to tourists’ socio-demographic characteristics (e.g. family travel status), holiday characteristics (e.g. time constraints) and destination activities (e.g. skiers usually stay for a shorter period of time) (Alegre & Pou, 2006).

Another explanation could be tourists’ varied reactions to weather risks. Good weather draw in a larger number of tourists and encourage them to extend their stays as a natural outcome (de Freitas, 2019). However, tourists’ responses to risks of bad weather may vary depending on the planned LOS. Short-term trips are more prone to be cancelled under severe weather while longer-term journeys are more resilient and tend to result in extended overnight stays (Bausch et al., 2021). Long-duration travellers are less likely to cancel trips or return early due to pre-arranged accommodations or transportation, as opposed to short-term or day visits. This implies destinations may see an overall reduction in visits while observing the share of long-stay demand increases.

Spatial substitution effect

The spatial substitution effect is endowed mostly by tourists’ adaptative capacity to flexibly substitute destinations, even within a small region (Steiger et al., 2022). This flexibility is a key issue when considering consumers’ adaptation strategies to climate change. Countries usually have a variety of climate types and attractions, allowing travellers to switch between them. For example, 61% of Ontario, Canada skiers experiencing less-than-ideal skiing conditions (Rutty et al., 2015) and 77% of scuba divers’ confronting coral bleaching (Verkoeyen & Nepal, 2019) indicated that they would adopt inter-regional spatial adaptation by visiting other resorts in that region. Which means that even if climate change causes a reduction (or increase) in demand for a particular attraction/city/region, it may be compensated for (or offset) by other more (or less) attractive locations within the country. Consequently, climate change may have mixed impacts at the sub-national scale, making the national and subnational variations in travel demand inconsistent.

Seasonality

Research has confirmed that tourism seasonality in climates with pronounced annual variations (i.e. mid-to-high latitudes) is overwhelmingly influenced by climatic factors, whereas those in more stable climates (e.g. small intra-annual climate variations in low latitudes) are more influenced by institutional factors (Hadwen et al., 2011). However, higher latitudes are now facing diminishing distinctiveness of a four-season climate. With the minimum temperature rising faster than the average, warmer conditions extended from summer to other three seasons (Wang et al., 2021; Young & Young, 2021). A number of studies which employed the Tourism Climate Index (Grillakis et al., 2016; Matei et al., 2023; Scott et al., 2016) confirmed that tourist comfort is decreasing in summer but increasing in spring and fall, which leads to demand shift accordingly from the peak summer to traditionally shoulder seasons. We therefore expect a flattening of tourism demand seasonality.

Time trend

Climate change intensification, accompanied by wider and rapid media dissemination, has sensitized visitors’ awareness. Most of the warming has occurred in the past 40 years (IPCC, 2022). temperature has climbed more than twice as quickly and land precipitation has increased faster since 1981 (IPCC, 2022). Ice retreat has recently become apparent and the global sea level rise rate in the last two decades is accelerating every year (IPCC, 2007). Meanwhile, the rapid media dissemination nowadays provides tourists with more accurate weather forecasts, and its coverage of extreme weather may exacerbate tourists’ risk concerns, resulting in a more rapid and negative demand response (de Freitas, 2019; Scott & Lemieux, 2010). Thus, while there have been more people travelling since 1980 with the tourism industry development, there might be less than would otherwise occur due to a more negative tourism demand response.

Here, we expect to investigate how diverse demand factors modify the slope of demand curve (elasticity) by proposing hypotheses H3–7 as follows:

H3. Visitor segment: The absolute value of domestic visitor demand changes is greater than that of international visitors.

H4. Demand measures: The number of visits decreases while the number of overnight stays increases under the global warming phenomenon.

H5. Spatial substitution: The global warming elasticity is expected to differ at the subnational level (regions, cities, and attractions) and at the national level (countries).

H6. Seasonality. Intra-seasonality (monthly/quarterly/daily variations) of tourism demand under climate change is flatter.

H7. Time trends: The negative global warming elasticity of tourism demand strengthened after 1980.

Methods

Meta-analysis techniques have been widely used to identify the degree of consensus and thereby generalize overall conclusions through a systematic review of existing econometric studies of tourism demand under climate change. We followed the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) procedure (Liberati et al., 2009; Lipsey & Wilson, 2001) to first conduct a rigorous review.

PRISMA literature selection and screening

This study explores how climate change in the narrow sense – i.e. changes in meteorological conditions (NASA, 2023; UN, 2023) – affects travel demand by selecting climate proxies including temperature, relative humidity, rainfall and sunshine. These are the most frequently employed proxies within the climate change-tourism demand theme based on our review, maintaining a high level of granularity for our analysis. A PRISMA flow diagram (Figure 2) records the literature searching and screening process.

Figure 2. PRISMA process of literature search and screening.

We started a broad search of the English peer-review articles that analyse the relationship between climate change and tourism demand using econometric models through several major academic databases between November 2022 and April 2023. The initial search arrived at 768 studies. After removing duplicates and irrelevant cases through title screening (N = 454), we engaged in full-text screening and removed 173 papers, including 70 irrelevant studies and 103 non-empirical studies (qualitative studies and policy analyses). The final step was to ensure eligibility through two phases. Phase 1 excluded 93 cases that focused on business responses (e.g. ski resort operations), employed meteorological indices (e.g. tourism climate index) or covered time periods that spanned the pandemic period. After phase 1, a total of 48 papers were included in the systematic review database, all of which fit our theme but needed to be examined for feasibility and comparability of meta-analysis. Four cases were excluded because they did not model visits or LOS, three cases were removed because their analysis used climate factors from the departure countries (not destination countries) and seven cases were deemed ineligible because they did not provide elasticity values and we were unable to obtain raw data from the authors.

The two-phase screening returned 34 eligible studies, containing a total of 290 estimates that serve as the basis for the following meta-analysis. These collected case studies cover 23 countries across all continents and major destinations, characterized by a strong concentration on European (1/2) and Asian destinations (1/3), whereas the remaining regions are poorly studied, notably Africa (1%), South America (2%), small islands (4%) and North America (5%).

Pooled effect size

An essential feature of quantitative meta-analysis is its ability to compare effects across studies, which requires the use of a single metric. Elasticity is adopted as the most appropriate indicator to both harmonize results and reflect demand sensitivity, and the global warming elasticity of tourism demand is interpreted as ‘the percentage changes in tourism demand corresponding to a 1% change in respective global warming indicators’. We calculated elasticities, semi-elasticities and level changes extracted respectively from the three forms of econometric models: the double-log, log-linear and linear models, all of which were harmonized to elasticities (see Supplementary Appendix 3 for data sources and calculations).

The random-effect meta-analytical model was selected, which assumes that the true effect could vary across studies due to the differences (heterogeneity) among them, and therefore the pooled estimate would be the weighted average of study-specific effect sizes. As a lower variance generally indicates higher precision, we employed an inverse-variance weighting method to down-weigh imprecisely estimates according to the variance $s_k^2$ of each effect size (Liberati et al., 2009), as (1). (1) $$w_k=\frac{1}{s_k^2}$$(1)

Therefore, for a set of 290 standardized effect size, the synthesized weighted global warming elasticity was obtained as (2). (2) $$\widehat{y}=\frac{{{\displaystyle \sum }}_{k=1}^{290}{\widehat{y}}_k{w}_k}{{{\displaystyle \sum }}_{k=1}^{290}{w}_k}$$(2)

The Paule-Mandel estimator was used to estimate the between-study variance. Heterogeneity was investigated using Cochrane Q (where p < .10 was deemed statistically significant) and Higgins $I^2$ statistics, where increasing values (ranging 0–100%) correspond to increasing heterogeneity. The heterogeneity was categorized as low ($I^{2\mathrm{ }}$≤ 25%), moderate (25% < $I^2$ < 75%) or high ($I^2$≥ 75%) (Hedges et al., 2010). To ascertain publication bias, Funnel plot and Egger’s test (Egger et al., 1997)were performed to assess whether missing studies threaten the validity of meta-analytic inferences and whether researchers tend to report a particular direction result (Lipsey & Wilson, 2001). In our study, the funnel plot of 34 studies suggested that publication bias is less likely (see Supplementary Appendix 1 – Figure 1), and the Egger’ test returned the p value = .88 of the intercept, verifying the findings of less evidence of bias.

Meta-regression

(3) $${\widehat{y}}_k={\beta }_0+\beta x_k+{ϵ}_k+{\mu }_k$$(3)

The random effects meta-regression model was developed as shown in (3). Where ${\widehat{y}}_k$ denoted the observed global warming elasticity of tourism demand. ${\beta }_0$ was the intercept; $\beta$ was a parameter matrix to be estimated through regression; $x$ was a matrix of explanatory variables that works as potential effect modifiers. Based on the eight research hypotheses, we operationalized all climate and demand variables as listed in Table 1. ${ϵ}_k$ was the sampling error through which the effect size of a study deviates from its true effect; and ${\mu }_k$ was the random effect term representing the between-study heterogeneity.

Table 1. Variable explanations.

Hypotheses	Categorical Variables	Definitions (0 = otherwise)
H1 Climate Zone
	Tropical	1 = Tropical climate (Af; Am; Aw)
	Arid	1 = Arid steppe climate (Bsh; Bsk)
	Temperate	1 = Temperate climate (Cfa; Cwa; Csa; Cfb)
	Subtropical	1 = Humid subtropical climate (Cfa; Cwa)
	Mediterranean	1 = Mediterranean climate (Csa)
	Marine west coast	1 = Marine west coast climate (Cfb)
	Continental	1 = Continental climate (Dfa; Dfb; Dwa)
	Not specified	1 = No identical climate types
H2 Destination type
	Mountain	1 = Destination dominated by mountain tourism
	Coast	1 = Destination dominated by coastal tourism
H3 Warming indicator
	Temperature	1 = Temperature in degree celsius
	Precipitation	1 = Precipitation in millimetres
	Sunshine duration	1 = Sunshine duration in hours
	Relative humidity	1 = Relative humidity in a percentage
H4 Visitor segment
	Domestic	1 = Domestic tourism demand
	Internal	1 = Both domestic and international tourism demand
	International	1 = International tourism demand
H5 Demand indicator
	Number of tourists	1 = Demand measured by visitor numbers
	Number of overnight stays	1 = Demand measured by visitor nights
H6 Spatial substitution
	National	1 = National data
	Sub-national	1 = Region/city/attraction data
H7 Seasonality
	Inter-annual	1 = Yearly data
	Intra-annual	1 = Monthly/quarterly data
H8 Time trend
	Before 1980	1 = Study includes data prior to 1980
	After 1980	1 = Study includes data from 1980 onwards

The classification of climate zones is based on the Köppen climate classification system (Arnfield, 2023), with multiple climate codes for each climate zone.

The modelling results will serve to (1) identify sources of heterogeneity across different studies. (2) Identify moderators: to explore how the relationship between the covariates and the effect sizes may change across different contexts or cohorts. The regression outcome (Supplementary Appendix 4 – Table 3) confirmed the presence of significant moderators in each hypothetical subgroup, which unearthed substantial heterogeneities and provided strong evidence for the following grouped analysis.

Results

Overall global warming elasticity

The overall climate change elasticity of tourism demand is −0.013%, which indicates that, on average, a 1% increase in climate change conditions leads to a decrease in tourism demand of −0.013%. This value needs to be interpreted carefully due to the high variability in demand changes evidenced by the large degree of heterogeneity observed (p value of Cochrane Q < 0.01, and $I^2$ = 99%), as well as the polarized and wide-ranging 95% confidence intervals [−0.08%, 0.06%]. Subgroup analysis must be carried out.

Grouped analysis

In Table 2, Model (I) records the pooled effect size for the overall sample, and Models (II) and (III) differentiate the warming elasticity by demand indicators.

Table 2. Pooled and grouped global warming elasticity.

Hypotheses	Variables	(I) Pooled effect size		Subgroup heterogeneity	(II) Tourists		Subgroup heterogeneity	(III) Overnight stays		Subgroup heterogeneity
		Counts	Elasticity	p Value	Counts	Elasticity	p Value	Counts	Elasticity	p Value
H1 Climate Zone				.001			.002			.369
	Tropical	39	−0.22 ·		39	−0.22*		N	N
	Arid	25	−0.12		25	−0.12		N	N
	Temperate	183	0.02		82	−0.08		101	0.09**
	Subtropical	60	0.13**		60	0.13**		N	N
	Mediterranean	46	−0.26*		18	−0.69*		28	0.02
	Marine west coast	77	0.11**		–	–		73	0.12***
	Continental	35	0.14**		23	0.16		12	0.06**
H2 Destination type				.437			.954			.071
	Mountain	43	−0.13		27	−0.18		16	0.02*
	Coast	85	0.04		65	−0.06		20	0.29*
H3 Visitor segment				.152			.163			.488
	International	147	−0.04		106	−0.08		41	0.04
	Domestic	68	−0.07		22	−0.41		46	0.05***
	Internal	75	0.11		49	0.06		28	0.19
H4 Demand indicator				.006
	Tourists	177	−0.01·
	Overnight stays	113	0.09***
H5 Spatial substitution				.068			.111			.842
	National	71	0.07·		71	0.04		32	0.07***
	Sub-national	219	−0.04		219	−0.13 ·		81	0.08*
H6 Seasonality				.097			.239			.288
	Inter-annual	67	−0.18·		67	−0.25		20	0.00
	Intra-annual	34	0.05		34	−0.01		93	0.11***
H7 Time trend				.1			.383			.299
	Before 1980	103	0.05*		103	−0.01		78	0.05***
	After 1980	187	−0.05		182	0.01		35	0.15
Overall		290	−0.013 [–0.08, 0.06]		177	−0.09		113	0.09**

We omitted observations with fewer than 10 and replaced them with ‘–’.

p < .1^·, .05 *, .01 **, .001 ***.

From Table 2 Model(I), most stratifications display statistically within-group heterogeneities, with variations reported across climate zone (p = .001), warming indicator (p = .056), demand indicator (p = .006), spatial substitution (p = .068), seasonality (p = .097) and time trend (p = .01) groups – supporting the proposed hypotheses in Conceptual framework Section. This finding also verifies that the interrelationship between climate change and tourism demand is too intricate to be represented as one-size-fits-all generalizations.

Climate factors

The shift of tourists across different climatic zones is the first indication of how tourism is being redistributed. As a consequence of climate change, Table 2 Model(I) reveals that the climate change elasticity of tropical (−0.22%) and arid zone (−0.12%) is negative while positive in temperate (0.02%) and continental zones (0.14%), supporting H1. Within the temperate zone, we observe positive demand change in the subtropical (0.13%) and marine west coast regions (0.11%), while there is a huge loss in the Mediterranean (−0.26%). The overall pattern provides a fine-grained picture of shifts in demand by climatic zones, specifically the redistribution of tourism demand from tropical, arid and Mediterranean regions to subtropical, marine west coast and continental regions. Furthermore, the violin plot (Figure 3) visualizes the kernel density distribution of estimates for each climate group. Scattered distributions in marine west coast and continental regions display concentrated clusters around the median, suggesting that most regions in these two climate types are being observed to constantly experience an increase in tourism. While the long tails in tropical, arid and Mediterranean regions present a large proportion of regions that have suffered a significant reduction in tourism demand.

Figure 3. Violin plot of global warming elasticity by climate types².

In addition, a cross-sectional comparison of the parameters corresponding to the destination types listed in Table 2(I), (II) and (III) reveals that both destination types of tourism exhibit a decline in visitor numbers (coastal: −0.06%; mountainous: −0.18%) but increases in overnight stays (coastal: 0.29%; mountainous: 0.02%), while the coasts benefit from a more prominent increase in visitor nights and a smaller decrease in visits compared to the mountains. Overall, that leads to coastal tourism (0.04%) benefit from climate change, while mountain tourism (−0.13%) suffer, supporting H2. More interestingly, mountain and coastal tourism are displaying opposing patterns across different climate zones. In lower latitudes, warming seems to attract tourists to mountainous regions (arid: 1.27%) but reduces demand for coastal tourism (arid: −0.03%); this pattern is reversed in high latitudes, with mountainous tourism being undermined (temperate: −0.25%, continental: −0.05%) but coastal destinations benefiting (temperate: 0.34%, continental: 0.06%) (Figure 4).

Figure 4. Split violin plots of global warming elasticity of demand for mountains and tourism by climate zone.

Demand factors

Despite the between-group differences not being statistically significant, H3 is weakly supported by a stronger negative response (−0.07%) of local visitors than inbound visitors (−0.04%), which is consistent with the observations of Cai et al. (2011) and Nunes (2013) that domestic tourists are more sensitive to climate change. In addition, opposite results are observed for tourist volume (−0.09%) and overnight stays (0.09%), supporting H4 that visitor numbers are negatively impacted by climate change, but the share of long-stay visitors has increased so that the average LOS has prolonged. Furthermore, visitors spatial (H5) and temporal (H6) adaptation strategies have been observed. The analysis conducted at the national level indicates a 0.07% increase in demand, whereas the sub-national level study (i.e. regions, cities and attractions) reveals a 0.04% decrease. This finding justifies sizable positive spatial substitution effects (H5) that maintains overall national demand, that is, tourism losses in certain regions were compensated by increases in others. The seasonal pattern corresponds to a very small variation (0.05%), signalling that the prominent climatic characteristics of the four seasons are weakening and being replaced by longer warm seasons, which drives the intra-annual seasonality of travel demand to flatten out (Hewer et al., 2016; Young & Young, 2021), supporting H6. While the annual demand variation recorded by yearly data is −0.18%, revealing the yearly gradual reduction in demand. Lastly, the elasticity of global warming is positive until the 1980s, at 0.05%, while the effect is reversed after 1980, at −0.05%, confirming the depressing effect of global warming on tourism demand seems to intensify over time (H7).

Lastly, in the sensitivity test (Supplementary Appendix 5), leave-one-out analysis (Hedges et al., 2010) was first performed to confirm that there are no statistical outliers, suggesting the robustness of our findings. Further examinations were carried out pertaining to model specifications, that is , squared terms, and an inverted U-shaped association between temperature and tourism demand at high latitudes was discovered in a few studies (N = 3) that included squared factors.

Tourism demand forecast by climate types

We used estimates of global warming elasticities corresponding to each climate type to illustrate how changes in tourism demand are expected, holding all other variables constant. The percentage change in future temperatures was calculated based on median (50th percentile) temperature projections³ from 2000⁴ to 2050 for 193 countries, with the raw data sourced from the World Bank’s Climate Change Knowledge Portal (World Bank Group, 2023). This temperature projection was modelled from the multi-model ensemble compilations of the CMIP6 (IPCC, 2022) under Shared Socioeconomic Pathway 2-4.5 (SSP2-4.5), which represents a ‘Middle of the Road’ scenario depicting that CO2 emissions hover around current levels with social, economic, and technological trends that do not shift markedly (IPCC, 2022). Hence, the prediction could be regarded as conservative. The forecast was made based on the climate type of their capital city for 193 countries by multiplying the corresponding warming elasticities (Table 2 Model(I), H1) with the projected temperature change to depict a global tourism demand shift map (Figure 5).

Figure 5. Global forecast of tourism demand to 2050.

Based on our dataset, from 2000 to 2050, the global average temperature will increase by 1.54$°\mathrm{C}$. The temperature surge at high latitudes is about twice as large as at low latitudes, with the largest increase in the continental regions in the range [1.46$°\mathrm{C}$, 2.66$°\mathrm{C}$] and the smallest change in temperature amplitude in the tropics, which ranged between [0.17$°\mathrm{C}$, 1.76$°\mathrm{C}$]. The demand for travel is expected to increase mostly in middle- and high-latitude countries; however, the benefits do not spread equally among nations with different levels of prosperity. Three-quarters (74%) of the countries experiencing growth in tourism demand are high-income (51%) and upper-middle-income (23%) countries. Among them, 84% may experience demand growth rates exceeding 1.5% solely based on the climate effect, with Canada, Russia, Finland, Norway and Sweden being the top beneficiaries. On the contrary, it is seen that the demand for tourism is diminishing in the lower latitudes and Mediterranean regions. Among these negatively impacted countries, the share of low-income (16%) and lower-middle-income (30%) countries rises to half. Notably, all 37 Small Island Developing States (SIDS) that have been modelled are estimated to experience declining tourism demand due to climate change, with a likely range of [−0.45%, −1.03%] such as Guyana, Suriname and the Dominican Republic. It should be recognized that climate is only one of the influences affecting tourists and not a determining factor in terms of causality. This predicted geographical distribution map should be interpreted further with caution by simultaneously considering destination’s climate vulnerability, adaptive capacity and tourism resources.

The analysis explicitly exposes three groups of relative beneficiaries and impaired countries behind this implicit information. Group 1, high-income countries that are located in the most tourism-favourable climate zones, especially OECD members, could be regarded as the most benefit group that enjoy both excellent internal tourism resources and adaptation capabilities. While the most disadvantaged group (Group 2) will be low- and middle-income countries and SIDS located in Africa, the Middle East, South Asia and in the Caribbean, Indian and Pacific Oceans that battle scorching heat, humidity and rainfall along with their low capacity to manage climate vulnerability. Exceptionally, there also countries that falls between these two groups (Group 3a,b). The first case in this category includes high-income nations located in low latitudes that are subject to increasingly unsuitable weather patterns (e.g. Brazil and Chile) and in Mediterranean climate regions (e.g. Italy) (Group 3a). While these stand to lose out on demand, their high adaptative capacities (indoor attractions, fountains, green areas, projects of fresh air recirculation in buildings, etc.) will be expected to mitigate tourism losses to some extent (Agulles et al., 2022; Rastegar et al., 2023; Scott et al., 2019). The other case is the opposite, that is, including several low- and middle-income nations that are seeing increases in tourism demand from a climatic viewpoint, but their attractiveness of the climate alone is not supported by their underdevelopment in tourism (Group 3b). Examples such as Tajikistan and the Kyrgyz Republic face insufficient infrastructure, resources and marketing resources to keep up with a rising tourism demand (Higgins-Desbiolles, 2018; WTO, 2017).

The win-lose scenarios discussed above align with the findings from the latest global simulations by Scott et al. (2019). These simulations evaluated the Climate Change Vulnerability Index for Tourism across 181 countries using 27 indicators, encompassing aspects of tourism demand, supply and adaptation. The results highlighted that OECD countries at higher latitudes exhibit the lowest vulnerability. In contrast, SIDS and countries where tourism significantly contributes to the GDP showed the highest vulnerability. However, our results differ from the projection of Hamilton et al. (2005). Particularly, countries such as Guinea-Bissau, Benin and Sudan are estimated by Hamilton et al. to be among the top 20 countries to benefit from climate change. In contrast, our projections indicate a decline in demand for these countries. The reality aligns with this forecast, showing that these low-income or lower-middle-income communities are both underdeveloped in terms of tourism and lacking in climate resilience. Possible reasons for discrepancies with the earlier study could include their outdated dataset (1995) or the reliance on single-year tourism data to make global projection, which restrict the capacity to perceive the spatial-temporal intricacies of the climate change-tourism demand linkage. In contrast, our meta-analysis synthesized the last two decades of research with a fine-grained categorization of countries by climate type and the utilization of multiple climate proxies that compensate to some extent for Hamilton et al.’s limitations.

Discussion and implications

Global warming clearly introduces relative beneficiaries and victims in travel demand. Here, we summarized meta-analysis results into three typologies in Figure 6: (A) vulnerability spectrum categorizes climate factors based on their influence on tourism, serving destinations in self-assessing their climate vulnerability to changes in tourism demand. (B) adaptation spectrum identifies the heterogeneity and adaptive behaviour of tourists, empowering destinations to effectively foster adaptation strategies by intervening in tourism demand accordingly. And (C) redistribution spectrum discusses the likely beneficiaries and victimizers that may rise should the redistribution of tourists continue to intensify.

Figure 6. Three benefit-harm spectrums.

Specifically, from Figure 6(A), a changing climate will contribute to more tourists for those located in humid subtropical, marine west coast, and continental zones, especially for high-latitude coastal areas and low-latitude mountain communities. In contrast, the climate zones that are likely to see a decrease in tourism demand are: tropical, arid, and Mediterranean regions, especially low-latitude coastal regions or high-latitude mountainous areas. While regulating these climate factors is beyond the capacity of destinations, we explored how demand factors, including tourist heterogeneity and spatial-temporal adaptation, shape the climate elasticity of tourism demand in Figure 6(B). For example, enhancing a destination’s capacity to host long-stay and international segments and coordinating resources to increase carrying capacity towards shoulder seasons are instrumental in maintaining a good share of less climate-sensitive tourists. Lastly, the global forecast (Figure 6C) of tourism demand redistribution by 2050 illustrates three country classifications, visually illuminating the salient challenges of unequal climate burdens across countries’ income status. Knowledge of these global win-lose situations will have far-reaching implications for estimating the SCC and determining climate justice processes in the tourism industry.

As one of the main financial instruments for reaching the climate goal of the Paris Agreement, the SCC is defined as the present value of all net damages resulting from an additional metric ton of CO2 emissions (Nordhaus, 2017). It has been widely introduced by governments as the basis for rationalizing the costs of emissions reduction projects. Despite tourism being highly susceptible to climate change, the current SCC calculations do not account for the economic gains/losses of this sector, preventing tourism from being incorporated into national climate adaptation and mitigation efforts. This specific omission problem was especially emphasized by IPCC AR6 (IPCC, 2018), Tol (2011), and Roson and Sartori (2016). And the primary cause of this issue is the lack of a comprehensive understanding, quantification, and monetization of the impacts of climate change on tourism demand. By quantifying changes in tourism demand, our study takes a necessary first step towards estimating the value of the neglected SCC in tourism in a disaggregated manner, which will inform the future construction of tourism damage equations and serve as a warning and rationale for countries (especially those tourism demand-hit countries) to incorporate tourism into the global and national climate adaptation/mitigation regulatory framework.

In addition, climate justice refers to the unequal distribution of risks and responsibilities among vulnerable economies/groups in addressing climate change (Alves & Mariano, 2018; Camargo et al., 2007). Tourism has the potential to impede the climate justice process, as it remains an overall luxury enjoyed by a small percentage of the population and significantly increases emissions in selected major destinations (Higgins-Desbiolles, 2018). Notably, approximately 1% of the world’s population that enjoys the most frequent flights accounted for more than half of the total emissions from passenger air transport (Gössling & Humpe, 2020). And half of the tourism carbon footprint growth occurred in high-income countries and due to high-income visitors (Lenzen et al., 2018). Thus, identifying the benefit-victimization scenario in diverse climate change-affected destinations is a crucial step in locating tourism’s role in climate justice. Our findings unveiled that the favourite climate zones for tourism are mainly located in high-income countries, including the major tourism destinations - USA, China, and northern and central Europe. These are the places that will reap the benefits of climate change while leading to more tourism-related emissions being produced. Meanwhile, most low- and middle-income economies and SIDs will bear the cost of tourism losses even though their relatively small tourism sectors are not big emitters. This is precisely the vicious cycle of expanding tourism emissions in developed countries to unfairly induce climate risks to less developed nations, and the magnitude of these influences is expected to intensify over time. Therefore, the second implication of this study is exposing the role of tourism as an impediment to climate justice under climate change, which urged global actions.

While our research has contributed to the clarification of the climate change-tourism demand relationship, there are several directions that could be improved by future research. Most of the individual studies focus on one region, and the only two global studies we included did not address domestic demand for tourism. Therefore, the need for a comprehensive global model remains urgent. At the same time, regional knowledge gaps still remain in Africa, America, and small islands. These areas are considered to be relatively less resilient to climate change but are projected to experience the fastest growth in tourism over the next 30 years (Scott et al., 2019), calling for increased understanding and planning for climate change. Also, we focused on the impact of meteorological indicators on tourism demand, but it is also valuable to expand the focus on other climate change impacts (e.g., extreme weather) to other tourism assets like urban destinations, parks, cultural heritage, and biodiversity to provide a more comprehensive picture. Finally, temperature effects may be nonlinear. Although data is currently limited, a threshold relationship between tourism and temperature is emerging, and whether such a relationship exists for global tourism needs to be further determined.

Clearly, mainstreaming climate change considerations into tourism planning and management is imperative. As emphasized by the World Bank (2020) and TPCC (2023), the tourism sector over the next 20 years will be characterized by the full integration of climate change and related issues into operational strategies, which will require practical guidance and capacity building for tourism stakeholders on climate change adaptation and mitigation. In these regards, our study can serve as a reference for subsequent investigations of the climate change-tourism demand relationship.

Disclosure statement

No potential conflict of interest was reported by the authors.

Additional information

Funding

This work was funded by Australian Research Council Discovery Projects (DP210103352).

Notes

1 Not included because there are no studies specific to these areas.

2 The red line represents the overall climate change elasticity of tourism demand.

3 We only look at temperature changes to minimize the impact of multicollinearity on the forecasts. Although most of the literature uses multiple climate change indicators, nearly one-third of the articles choose temperature as the only representative climate indicator.

4 2000 is the median of the distribution of years of data cited in the 290 estimates and is therefore considered a base year.