Winter tourism dependence: A cyclical and cointegration analysis. Case study for the Alps

Abstract

This work studies the evolution of winter tourism in the main European ski resorts in recent years, exploring the degree of dependency it presents on the gross domestic product gaps of those European countries with the main registered incoming tourists attending to these ski resorts. This study consists of two parts: first, a cyclical behavior analysis of the evolution of winter tourism demand in these regions and its level of external dependence. This is achieved by the application of decomposition techniques of the economic cycle to verify the influence of these variables on the degree of cyclical fluctuation of winter tourism. For the second half, cointegration techniques are applied to test the linear or nonlinear combinations of these variables. This dual analysis allows a wider perspective, considering short- and long-term fluctuations, when analyzing co-movements and dependence of the demand of tourism-based destinations with a proxy variable for the income.

Keywords

Cointegration economic cycle HP filter method region analysis winter tourism

Introduction

During the last decades, winter and snow tourism has become one of the main economic activities for most mountain areas (mainly in Europe), due to its key association to skiing and mountaineering activities (Abegg et al., 2007; Trawöger, 2014). This type of tourism has taken an important part on the regional development, being a substantial portion of its total income (Lasanta et al., 2007; WTO-UNEP, 2009). Following Butler and Mao (1997) seasonality classification, we can affirm that this type of tourism suffers from an important seasonality phenomenon, but it is understood as a “two-peak seasonality,” that occurs when there are two clear and different seasons, a winter and a summer one, showing two different types of touristic attractions in the destination. According to Hudson (2005), several impacts of seasonality that can be cited are low investment rates due to ski facilities and resorts operating only in winter, reduced tourism expenditures during the off-season months or crowding and saturation of destinations. Following a study carried out by Tuppen (2000), this massification of ski areas and resorts has evolved in a major constraint to skiing for French demand.

Visitors are estimated to be around 400 million skier visits worldwide (Vanat, 2019), not presenting any important increase for the last 10 years. In spite of a substantial absence of open data regarding ski-related tourism, altogether with weather conditions that may influence variations of the number of visitors between years, estimations have been stable for the last decade, since mature markets (as Europe or the Alps, where this work focuses) have reported reduced growth rates whilst other new markets were rising (China or Eastern Europe). Albeit on reduced growth rates, the Alps region holds a 44% of the total number of visitors, compared to a 21% of the following most important area, United States (Vanat, 2019).

The demand for overnight tourism stays is a function of various variables among which are income, price, and preferences, and in the case of ski winter tourism, the snow/weather conditions are also relevant factors (Falk and Hagsten, 2017; Falk and Lin, 2018; López Moreno et al., 2014; Rutty et al., 2017). Ski-related tourism has an important global presence despite its inherent climate dependency. In fact, this important global presence has been lately related to the overtourism phenomenon. Although winter tourism does not suffer so clearly from overtourism problems as other destinations that offer different touristic products, such as “sun and beach” or “cultural,” at certain times of the peak season, some ski areas and resorts are excessively overloaded. According to the Internationale Tourismous-Börse (ITB) Berlin World Travel Trends Report, ski resorts were the most crowded tourism locations in 2017 (ITB Berlin, 2018). Tourism sustainability and destinations’ carrying capacity are transversal concepts that affect many different aspects of the territory: ecological/environmental, physical, economic, socioperceptual, and policy/management (Alvarez-Sousa, 2018; García Hernández and de-la-Calle-Vaquero, 2012). In this article, an economic perspective is employed, taking into account that potential winter overtourism is, undoubtedly, a negative externality that must be taken into consideration both from the destination and business management perspective. Although the impact of climate change is causing financial problems to some ski resorts, with their closure due to reduced visitor frequency (Falk, 2013), it is undeniable that local communities and authorities are concerned about this overtourism problem but also businesses’ profitability could be disturbed if regulations are too strong on capacity limitations (Pearce, 2018). In short, overtourism is a potential problem for ski resorts that could strongly compromise ski and mountaineering tourism activities in the most overcrowded destinations. Currently, outdoor skiing is possible in 85 countries and other 15 offer indoor ski facilities. There are 2132 ski resorts worldwide, 84% of them located in the Alps. Indisputably, experiencing snow and snowy landscapes outdoors is one of nowadays main travelling purposes, linked to enjoying nature and health preserving, two important driving forces for nowadays way of life.

Nonetheless, receiving a great number of tourists is not always a positive fact and it, undoubtedly, needs a sustainable management of the destinations (Vera and Ivars, 2003). This phenomenon can have an impact on different factors that can be analyzed on the basis of Archer et al. (2004). From an economic point of view, overtourism has an important influence on price rising derived from increasing demand pressure, land conversion for tourism services, or employment temporality. It is also relevant to include a political perspective that analyzes if the direct beneficiaries of policy changes are tourists or tourism-related business, leaving out the residents’ needs and claims. Finally, an environmental standpoint highlights an overuse of natural resources and occupation of natural open spaces.

To minimize negative externalities and reach a compromise between the destination and its several dimensions (natural resources, population, and territory), sustainable development must be the main tourism policy instrument (Huete and Mantecón, 2018). According to Goodwin (2017), overtourism is the antithesis of responsible tourism and is defined as those destinations in which residents or guests, locals or visitors, affirm that there are too many visitors and living quality in the area or visitor experience has been unacceptably damaged. This means that destination has not been properly managed and quantity has exceeded quality, putting sustainability and durability of the associated tourist attractions in risk, as well as an important part of the local economy.

Following this sustainable perspective, climate change is another important factor that must be acknowledged as an important threat to the geographic tourism area, being the trigger of important changes in fragile eco-systems but also implying a reduction in the social and financial benefits of the activity. In particular, the presence or absence of snow and its durability are key elements when studying the sustainability of the sector. There is an important amount of literature evidencing how winter tourism is portraying bigger effects of this climate change than sun and beach destinations, probably because the adaptation process for winter tourism presents more problems, even considering the installation of artificial snow-making machines (López-Moreno et al, 2014; March et al., 2014).

In relation to winter-related activities, there are several studies addressing the importance of climate factors as demand determinants: a recent article analyzed snow depth as a leading indicator of skier days in French ski regions (Falk, 2015), while another one studied how vulnerable were Spanish ski resorts to climate change, trying to address the implications in several areas (Rodrigues et al., 2018). Climate change implies shorter seasons for winter and ski-based tourism while it enlarges sun and beach tourism ones. Moreover, the absence of snow precipitations involves significant artificial snow production costs as long as an important increase in the use of water (March et al., 2014).

The top four ski resorts worldwide, based on their average annual skier visits, are the following: La Plagne (region of Savoie, France), Skiwelt Wilder Kaiser-Brixental (region of Tirol, Austria), Les Arcs (region of Savoie, France), and Saalbach Hinterglemm Leogang Fieberbrunn (region of Salzburg, Austria). These resorts lie in explicit ski tourism clusters, gathering a significant portion of the winter tourism industry. In fact, these four major ski resorts and, moreover, France and Austria are in the first and third place, respectively, according to total skier visit figures, 5-year average (Vanat, 2019). Given the international importance of these French and Austrian regions, following the European Union NUTS criteria- Nomenclature of Territorial Units for Statistics-, a regional NUTS2 (Tirol and Salzburg) and NUTS3 (Savoie) level analysis is performed, with the purpose of identifying similar behavior patterns and long-term relations between the regions and their countries.

Regarding the proportion of domestic and incoming ski demand in these countries, foreign customers in Austria represent around a 66% of all skiers, while these visitors comprise around a 30% of the total demand in France, according to data offered by Tourismus Statistik and Eurostat. Due to the undeniable influence incoming tourism registers, this work presents an analysis of the cycle of winter visitors to the leading ski resorts of the European region and the gross domestic product (GDP) gaps of their main tourist sender countries (measured with the GDP). Cyclical fluctuations in the series can be considered as waves in a period, commonly not longer than 12 years, that have convergence patterns different from those derived from other macroeconomic variables in terms of persistence, synchronicity, or duration (Van Dujin, 1977). The fact of segregating the cyclical component of the series from the rest of it can deliver further evidence on the short-run relationship between the series. This approach allows to see how the condition and economic context of the major incoming countries to these regions could explain, in part, the development of overnight stays and, hence, the total tourism demand. The size difference between the French region (Savoie), defined as a NUTS3 region, and the Austrian ones (Tirol and Salzburg), identified as a NUTS2 classification level, must be considered as it affects the result, granting a major influence to the NUTS3 level, in terms of territory scale, as the region considered for analysis is much smaller than the ones under the NUTS2 classification.

Despite the fact that these regions are worldwide salient in ski-related tourism, holding the global top ski resorts, it must be taken into account that there is a share of winter tourism that is not involved in ski activities, such as culture-driven trips. This proportion of nonski-related tourism could not be estimated, due to an important lack of data, therefore, conclusions derived from this study cannot be fully ascribed to ski and mountaineering activities but they do acquire a relevant significance. Ski and mountaineering tourism magnitude within the studied regions is strongly higher in Savoie (France) and Tirol (Austria), whereas in Salzburg (Austria), it coexists with other types of winter tourism related to cultural issues (such as music) that also have an important impact. Anyhow, this work based on winter season data is worth considering, due to the high increase of ski-related visitors grasped when employing seasonal data.

This study is structured as follows: in the next section, methodology used in this work is explained. First, a Hodrick–Prescott filter method is used to extract the cyclical component of the series, being able then to compare their evolution and matching. Given that the presence of nonstationarity in the analyzed series could cause potential inconsistency in the estimations; second, cointegration techniques are applied to investigate the existence of linear combinations for these series. In such a case, linear combinations permit an estimation of error correction models (ECMs) on the long-term pathway and the short-term behavior of the variables.

Next, on the third section, results are displayed and analyzed: first, examining the relationship between chosen NUTS2/NUTS3 regions and its countries of demand reference, in terms of winter season overnights. Second, due to the importance of foreign demand in these ski areas, exploring the influence that the GDP gaps of the main incoming international tourism partners have on winter tourism evolution in these regions and countries.

Finally, this work is brought to a close with some final conclusions that aim to summarize the most relevant information. A bibliography listing is added with the references cited in the text.

Methodology

The aim of this study is to understand the different relationships winter tourism presents for the main European ski resorts, located in France and Austria, with other both domestic and foreign factors. First of all, a cyclical dependence relationship for these tourism overnight stays is studied. To do so, a cyclical component analysis is applied, with the purpose of identifying similar patterns between the cyclical behavior of the country, its regions and their main incoming countries’ GDPs. After obtaining results through this approach, a complementary cointegration analysis is performed to test the existence of certain dependence relationships between the series. This structure of analysis has been employed due to its complementarity: while the first one analyzes only the cyclical component behavior, the cointegration analysis is based on characterizing individual patterns of each series to find linear combinations between them, without decomposing the series and subtracting the trend component. This complementary analysis permits a wider approach in the process of studying tourism demand dependency. This combination of methodologies is not new, as it has been previously employed for different purposes (Cheng, 2005; Zhang and Chen, 2010), but this is the first attempt in using it in a demand comparison study for a specific touristic product.

Data collection

Annual data were obtained from different sources:

Winter overnight stays for Austria, Tirol (1994–2016), and Salzburg (2000–2016) from Eurostat, Tourismus Statistik, and Landesstatistik Tirol.

Winter overnight stays for France and Savoie (1994–2016) from Eurostat and Savoie Mont Blanc Tourisme.

GDP at constant 2010 prices for Austria, France, United Kingdom, Italy, Germany, Belgium, and Netherlands (1994–2016) from the Worldbank Databank.

These obtained series have no other data processing than an outlier detection and intervention analysis performed in France overnight stays. The original time series presents three-level shift outliers: 1996, 2000, and 2010. Data for this, and the following years, have been smoothed using dummy variables to obtain a new series more suitable for comparison.

Cyclical component analysis

Hodrick–Prescott filter (Hodrick and Prescott, 1997) is applied to remove the effect of the trend on the cyclical component and, therefore, to estimate a softened long-term trend of the series. This filter estimates the trend element, minimizing the diversion from the trend using the following function:

M i n \sum_{t = 1}^{N} {(Z_{t} - Z_{t}^{T})}^{2} + λ \sum_{t = 2}^{N - 1} {((Z_{t + 1}^{T} - Z_{t}^{T}) - (Z_{t - 1}^{T}))}^{2}

Z_t refers to the original series value in logarithms and $Z_{t}^{T}$ stands for its trend. The Z_t term explains the level of adjustment to the original series logarithmic trend and the second term shows the variability rate, expressed in second difference terms. The coefficient λ corrects the variation trend between successive periods. If λ = 0, the value of the trend corresponds to the original series (cyclical fluctuations are minimized without restrictions), while with a λ = ∞, the trend is linear by definition and does not present variability. To determine the parameter value of λ, the rating suggested by Hodrick and Prescott for annual series has been maintained λ = 100 (Falk, 2015; March et al., 2014).

Thereupon, once the $Z_{t}^{T}$ estimated trend of the series is obtained, the cyclical component ( $Z_{t}^{C}$ ) should be retrieved as the difference between the original series in logarithms and the estimated trend:

Z_{t}^{C} = Z_{t} - Z_{t}^{T}

This research methodology has been previously used in several articles concerning tourism such as López Morales and Such Devesa (2016), analyzing the business cycle and external dependence on tourism for Spain, or concerning similarities between countries as in Marcet and Ravn (2004), studying the HP-filter in cross-country comparisons.

Although the Hodrick–Prescott filter has been widely used for decomposing variables into trend and cycle, certain limitations to this approach have been stated (Cogley and Nason, 1995; De Jong and Sakarya, 2016; Phillips and Jin, 2015) such as the production of series presenting spurious dynamic relations or the significant difference of filtered values at the end of the series compared to those in the middle (Hamilton, 2018). Because of that and without the purpose of evaluating any statistical approach but being able to strengthen this study’s conclusions, an additional cointegration analysis is performed on the series, to find any possible linear combinations between the chosen variables.

Cointegration analysis

The economic variables used on a classical regression model are often required to be stationary. If nonstationary variables are given, the regression can be considered as spurious (Dickey and Fuller, 1979), that means they will present a high coefficient of determination, R², and significant t-statistics. Several answers have been given to avert spurious regressions such as cointegration techniques, which allow developing an ECM or a vector autoregressive (VAR) model.

The first step when testing for cointegration is performing the augmented Dickey–Fuller (ADF) Test to find a unit root on individual time series to test whether the variables are integrated (Dickey and Fuller, 1979). The ADF test permits a determination of the integration order of a series (that means, the number of times it must be differenced for it to be stationary). This test has to be performed, first, on the level series. If stationarity is rejected, it is then applied on the differenced series until estimators show the absence of a unit root (meaning series is now stationary). This test is based on an auxiliary regression with an intercept and trend when performed on level and only an intercept when performed on a first difference:

Δ y_{t} = α + δ_{t} + β y_{t - 1} + \sum_{t = 1}^{k} y_{t - 1} + ε_{t}

In our models, Y_t is the winter overnight stays variable of each country/region or the GDP variables (or each of them independent variables). To check whether it is stationary or not, the following procedure is performed ΔY_t = (Y_t − Y_t−1), where ΔY_t corresponds to the first difference operator and t refers to the time period. The series is considered stationary if the null hypothesis (H₀) is rejected when performed on level.

Once the order of integration of the time series has been defined, a cointegration test can be established to identify if there is a stable long-term relationship between the variables. There are several methods to conduct cointegration test. The two most widely used methods are the Engle–Granger cointegration test and the Johansen cointegration test.

The Engle–Granger cointegration test (Engle and Granger, 1987) is basically equivalent to a unit root contrast in the residuals of the cointegration equation. The Johansen cointegration test is based on the system of equation using VAR models suggested by Johansen (1988). A VAR approach is used to model each variable as a function of all the lagged endogenous variables in the system. Johansen considers a simple case where Y_t is integrated of order one, such that the first difference of Y_t (ΔY_t) is stationary. The procedure developed includes the identification of rank of the n × n matrix Π in the specification as given below:

Δ y_{t} = μ_{t} + Π Y_{t - k} + \sum_{i = 1}^{k - 1} Γ_{i} Δ Y_{t - 1} + u_{t}

Y_t conforms a column vector of the n variables, π and G are the coefficient matrices, Δ is the difference operator, K stands for the lag length and μ is the constant for the chosen time period. The long-term relationship between the Y_t variables is represented by the Π matrix; moreover, its rank is the number of linear independent and stationary linear combination of the chosen variable. This cointegration test means testing the rank of the Π matrix G and check whether the eigenvalues of Π are significantly different from zero. The maximum-likelihood approach permits the analysis of r cointegration relations among elements of Y_t. Therefore, the null hypothesis (H₀) of no cointegration relations (r = 0) implies Π = 0. In Johansen’s maximum-likelihood procedure, two types of tests are carried out, the first using a statistic of the trace and the second using a statistic of maximum own value, and the values obtained by MacKinnon et al. (1999) are used as critical values. To determine the number of cointegration relationships, the process consists on sequentially contrasting the cointegration range (r) from r = 0 to r = k − 1, where k is the number of endogenous variables analyzed.

Results

Case study: Savoie (NUTS3 region)

Savoie represents half of the demand of winter tourism nights for the NUTS1 region of Auvergne-Rhône-Alpes, where it is located. Savoie is one of the most important alpine regions in the European context, with more than 10 million winter overnight stays in for the 2017 period, according to Savoie Mont Blanc Tourisme. From the total of nights, around 15% corresponded to foreign tourists. Although it is a significant percentage, it is expected that the cyclical evolution of the total demand does not fully adjust to the movements of the GDPs, and it is more influenced by other internal variables of the country.

Figure 1 shows the cyclical evolution of winter overnight stays by applying the Hodrick–Prescott filter methodology, eliminating the long-term trend in the France and Savoie series. As stated in the graph, the overnight stays series for Savoie and France display a different pattern of evolution. The fluctuations experienced by the region are significantly more stable than the ones suffered by the country. In boom periods, the cyclical oscillations for France are more pronounced than in Savoie, but in the 2008 recession period, both series show a similar cyclical fluctuation that ranges near zero. There appears to be a noticeable difference between winter tourism performance in Savoie and France. Taking into account that Savoie does not hold an excessive weight over total French overnight stays despite being an important ski-related region and analyzing the original series and their growth in the chosen period, it is confirmed that they present different pathways. Whilst France more than duplicates its winter overnight stays (showing a 3.5% compound average annual growth rate), Savoie presents considerable low rates (only around a 0.6%). Therefore, this different pattern of cyclical evolution is somewhat logical.

Figure 2 demonstrates the connection between the economic cycles (using the logarithm of GDP in dollars at constant 2010 prices) of the main winter tourist incoming countries (Netherlands—12% of total international visitors, according to Savoie Mont Blanc 2019 press release, Belgium—9%, and Germany—7%) and the cyclical evolution of winter overnight stays in Savoie (previously stated as the logarithmic transformation of the original series). These partners in winter tourism have been obtained from the chosen ski resorts press publications or from regional authorities press dossiers, which give information about the nationalities of their most frequent visitors. The firsts years of the analysis (from 1994 to 2001) show a less similar pattern that the latest studied years (2012–2016), where it is possible to verify certain analogy between the variables. Consequently, the cyclical behavior of these series is somehow influenced by the foreign economic context despite the correlation constraints. Incoming Savoian ski-related tourism was estimated to be around 15%, as stated previously in the first section of this work; thus, it is consistent to say that the economic context of the major ski-related tourism partners shows some dependence with the cyclical evolution of winter tourism in the region.

Following the methodology previously explained in the “Cointegration analysis” section, the ADF unit root test has been performed on the series, testing the cointegration degree of the winter overnight stays series for Savoie and its main regional winter tourism incoming countries’ GDPs for the considered time period.

ADF unit root test results for the logarithm (LOG) of the level variables; the first (D) and second (D2) differenced series are displayed in Table 1. Intercept and trend, and intercept and none deterministic specifications given by the test are considered in the level series. When a level series is studied, three possibilities are valid: having a constant term (intercept), presenting growth (trend), and not presenting any of them (none). Nonetheless, when a differenced series is considered, usually to make the variable stationary, growth is not commonly present. Thus, only the intercept option is considered for the differenced series. It must be considered that none of the estimated test of the series presents residual correlation according to values of the Durbin–Watson estimators

Table 1.

Dickey–Fuller tests for Savoie ADF output for Savoie-related series (winter overnight stays and GDPs), 1994–2016.

Level series	SAVOIE (LOG)		NETHERLANDS (LOG)		GERMANY (LOG)			BELGIUM (LOG)
Level series	t Stat	p Value	t Stat	p Value	t Stat		p Value	t Stat	p Value
Intercept and trend	−4.29*	0.0143	−2.36	0.3854	−3.93*		0.0289	−1.61	0.8893
Intercept	−2.09	0.2477	−2.87	0.0666	−0.57		0.8569	−2.32	0.1723
None	1.72	0.9751	1.68	0.9733	2.61		0.9963	5.84	1.0000
	D SAVOIE (LOG)		D NETHERLANDS (LOG)		D GERMAN (LOG)			D BELGIUM (LOG)
First differenced series	t Stat	p Value	t Stat	p Value	t Stat		p Value	t -Stat	p Value
Intercept	−6.89*	0.0000	−2.55	0.1100	−4.83*		0.0011	−3.99*	0.0064
	D²LSAVOIE		D²LNETHERLANDS		D²LGERMANY			D²LBELGIUM
Second differenced series	t Stat	p Value	t Stat	p Value	t Stat	p Value		t Stat	p Value
Intercept	−11.00*	0.0000	−4.67*	0.0017	−5.29*	0.0005		−5.76*	0.0002

ADF: augmented Dickey–Fuller; GDP: gross domestic product; LOG: logarithm.

Though for most of series, the ADF tests do not reject the null hypothesis of being I(1); there are some cases where results are not conclusive under the different deterministic specifications included (Table 1). The series that, according to the test, cannot be considered I(1) is Netherlands’ GDP series. Nonetheless, according to the existent literature in the field, having a time series whose order of integration is 2, as results are suggesting, is extremely weird. Following the proposal by Nelson and Plosser (1982) that real output levels are nonstationary, recent studies have applied different unit root tests to panel data from several countries’ GDP time series (Rapach, 2002), to gain robustness in their analyses, concluding that there is a nonstationary real output behavior in Organisation for Economic Co-operation and Development -OECD- countries. As mentioned before, GDP series have been thoroughly studied and statistics and inference are much affected by the presence of outliers in the estimated period. To show an example of this, further on in Table 3, we study the unit root presence for this time series in a wider period, concluding it can be considered as I(1)

Table 2.

Dickey–Fuller tests for Tirol ADF output for Austria-related series (winter overnight stays and GDPs), 1994–2016.

Level series	TYROL (LOG)		NETHERLANDS (LOG)		BELGIUM (LOG)		GERMANY (LOG)
Level series	t Stat	p Value	t Stat	p Value	t Stat	p Value	t Stat	p Value
Intercept and trend	−2.20	0.4500	−2.36	0.3854	−1.18	0.8893	−3.93	0.028
Intercept	−0.40	0.8900	−2.87	0.0666	−2.32	0.1723	−0.57	0.856
None	1.30	0.9500	1.68	0.9733	5.84	1.0000	2.61	0.99
	D TYROL (LOG)		D NETHERLANDS (LOG)		D BELGIUM (LOG)		DGERMANY (LOG)
First differenced series	t Stat	p Value	t Stat	p Value	t Stat	p Value	t Stat	p Value
Intercept	−3.79*	0.0098	−2.55	0.1100	−3.99*	0.0060	−4.83*	0.0011
	D²LTYROL		D²LNETHERLANDS		D²LBELGIUM		D²LGERMANY
Second differenced series	t Stat	p Value	t Stat	p Value	t Stat	p Value	t Stat	p Value
Intercept	−3.89*	0.0092	−4.67*	0.0017	−5.67*	0.0020	−5.29*	0.0005

GDP: gross domestic product; LOG: logarithm.

Table 3.

ADF output for Netherlands’ GDP for the period 1968–2017

Level series	t Stat	p Value
Intercept and trend	−2.21	0.47
First differenced series
First differenced series	t Stat	p Value
Intercept	−4.23*	0.001

GDP: gross domestic product.

Because of all these reasons, all of them have been considered as I(1). This fact allows a consistent cointegration analysis through the Johansen cointegration test.

As explained before, an analysis of the behavior and relationship between the regions and their main winter tourism partners can also be tested with cointegration techniques. The employment of this type of technique is notably important when analyzed series present stationarity problems in levels and, therefore, previously obtained results derived from a decomposition of the trend could be nonconsistent on a statistical basis. In this section, the purpose is to find linear combinations between the series. These linear combinations are organized in bivariate models, to test the influence that presents either their own countries’ winter overnight stays or their main incoming partners’ GDPs on each one of the chosen NUTS2/NUTS3 region.

Given the Engle and Granger cointegration test requires restrictive conditions, considering only one possible cointegration relation between the variables, it has been additionally performed the Johansen cointegration test that allows to test all possible long-term relations between the chosen series. Johansen cointegration test determines the number of cointegration equations between variables, named as cointegration range (r). This analysis proves the null hypothesis of no cointegration. If this null hypothesis is not rejected (p-value higher than 0.05), series are not cointegrated and, thus, have no long-term relationship. If the null hypothesis is rejected, the cointegration range stated will exist.

Johansen test estimations analyzing any cointegration relationship existing within the bivariate specified models for Savoie are collected in Online Appendix A (Tables 5 and 8–10). One single lag is employed for all the estimations, following the result given by optimizing the VAR model for the series.

Following the same structure employed in the cyclical component analysis, the long-run relation between French and Savoian overnight stays is going to be studied. Results for these tests, both from the trace test and the maximum eigenvalue, show that no cointegration vector exists at a critical value of 0.05 nor 0.1 (r = 0). Thus, Savoie and France (Online Appendix A, Table 5) show no long-term relationship. This outcome implies a careful interpretation of the cyclical evolution previously displayed for these series, since they could be nonconsistent on a statistical basis due to the presence of stationarity problems when studied in levels.

The cointegration tests are organized as bivariate models consisting of the overnight stays series for the region (Savoie, Tirol, and Salzburg) and the GDPs series of the major incoming countries travelling toward the chosen territory, one at a time. Since results could be misleading if several GDPs series are put together in the same model, as long-run relations are frequent between them, cointegration models have been articulated in pairs, following the scheme of a regional overnight stays series and a GDP.

French region of Savoie is confirmed to have three cointegration relationships: Germany, Netherlands, and Belgium. For the first country, the analysis presents (Online Appendix A, Table 8), at least, one equation between the two series at 0.05 critical value under both trace test and the maximum eigenvalue, presenting an associated probability of 0.03 for the first one and 0.027 for the latter (r = 1).

For the Netherlands analysis (Online Appendix A, Table 9), there are, at least, two feasible equations (r = 2) between the series at a 0.1 critical value under the trace test (probability of 0.0595 and 0.0123, respectively) confronting with the maximum eigenvalue showing no equations and, therefore, no cointegration patterns. Finally, Belgium and Savoie groups (Online Appendix A, Table 10) report two potential equations at a 0.05 critical value (r = 2) testing with trace criteria (probability of 0.019 and 0.034), yet Maximum eigenvalue test shows no potential cointegration between the series. In conclusion, Savoie’s long-term evolution of winter tourism overnight stays is dependent on the GDP gaps of Germany, Netherlands, and Belgium, among other factors, and this relation between them can be modeled.

Case study 2: Tirol

The Tirolean region of Austria has experienced an increase of a 14.8% in winter tourism overnight stays in the last 10 years, according to the Tourism Department of the region. The tourism GDP contribution is estimated to be around 17% for all the regions and it is substantially increased when analyzing points devoted to the touristic economy, as alpine and ski resorts. For the 2017 winter season, the NUTS2 region registered a total amount of 27.6 million of overnight stays.

Figure 3 demonstrates the evolution of the winter tourism overnight stays for Austria and Tirol (NUTS2 region). It is feasible to visually verify a strong similar evolution path between Tirol and Austria, which seems consistent since the Tirolean region represents around 60% of total Austrian winter tourism. Cyclical dependency for winter tourism in Tirol with the total amount of tourism received by Austria can be visually verified just by observing the series.

Figure 4 shows the cyclical comparison between Tirol and its main incoming tourism countries, Germany—50,6%, data obtained from Tirol Werbung, Belgium—4%, and Netherlands—11%. The series present a clear contracyclical but irregular movement. It is possible to distinguish two different sections in the graph; the first one, a precrisis scenery (1994–2008) versus the most recent years (2008–2016). While the first period apparently shows an opposite performance between Tirolean overnight stays and the economic cycle of Belgium, Germany, and the Netherlands’, a certain mild similarity can be observed in the last subset of data, although it fairly enables any conclusions to be drawn. Due to this fact, there is a clear opposite relationship between the series for the first 10 years of the period, but the pattern approaches from the 2008 crisis.

ADF unit root test has been performed on the series, testing the cointegration degree of the winter overnight stays series for Tirol and its main regional winter tourism incoming countries’ GDPs (Netherlands, Germany, and Belgium) for the considered time period.

Table 2 presents that all series are suitable to be categorized as integrated series of first order I(1) except for the Netherlands’ GDP that seems to have to be considered under the second differenced time series, as explained in the previous case study.

Due to influences of the chosen years where winter overnight data are available and with a strong literature backup, a complementary ADF unit root test is performed with the same time series for the period 1968–2017, presented in Table 3. Netherlands’ GDP, after this extended time analysis, is confirmed to be integrated series with an I(1) order:

Following Johansen (1988, 1991), it is possible to perform a cointegration test on these time series since the integration order of all of them is the same. As stated before, statistics and inference are much affected by the presence of outliers in the estimated period.

The long-run relations between Austrian and Tirolean overnight stays are going to be studied. Results for these tests, both from the trace test and the maximum eigenvalue, show that no cointegration vector exists at a critical value of 0.05 nor 0.1 (r = 0). Thus, Austria and Tirol (Online Appendix A, Table 6) show no long-term relationship. This outcome, as stated for the French case of study, implies a careful interpretation of the cyclical evolution previously displayed for these series, since they could be nonconsistent on a statistical basis due to the presence of stationarity problems when studied in levels.

The Tirolean overnight stays permit cointegration combinations with three countries’ economic contexts. The first cointegration match is reached with Belgium (Online Appendix A, Table 11), obtaining equations at 0.1 critical value with associated probabilities of 0.055 and 0.044 under the Trace criteria but no relationship if the employed test is maximum eigenvalue. In the analysis performed with the Netherlands’ GDP series (Online Appendix A, Table 12), both trace and maximum eigenvalue tests display, at least, one equation at a 0.05 significance value (with probabilities of 0.018 and 0.061, respectively). Finally, there is also a long-term relationship with Germany’s GDP series, both by the trace and maximum eigenvalue procedures, presenting probabilities of 0.0012 and 0.003.

Case study 3: Salzburg

Salzburg NUTS2 region had over 10 million of winter overnight stays for the last two seasons, corresponding to 2017 and 2018. Following an estimation made by the government of the region, Land Salzburg, around an 8% of all the tourism activity generated is due to the Saalbach Hinterglemm Leogang Fieberbrun ski resort. Given that Salzburg is an important European destination among cultural and weekend trips from all over the European region, the percentage of the total tourism economy that can be assigned to the ski resorts is substantial.

Figure 5 analyzes the cyclical dependency winter tourism arrivals to Salzburg present with those received by the whole country. Patterns for the series can be easily understood as different, showing different impulses and evolution. Diversification of the type of tourism offered by Salzburg can be a feasible explanation on the reasons why country-level and regional-level behavior is so remarkably different. Salzburg attractions are not only related to ski and mountaineering activities, given that cultural events or Christmas attractions are also important appeals of the region.

The Austrian series of Salzburg are studied in the same way as their forerunners, observing a similar behavior for the overnight series and the GDPs of its main partners: Germany—34%, data obtained from Tourismusstatistik, Land Salzburg, Netherlands—7,8%, and Italy—2% (Figure 6). Salzburg data are only available from 2000, thus this graph and the following correlation analysis present a difference in length from the other performed analysis, whose periods are 1994–2016. The growth path of the series shows an irregular behavior pattern, presenting the GDP series, lower growth rates in boom periods, and higher values when in recession. This behavior in the recession periods could be explained because the customers substitute more expensive overseas trips with relatively cheaper trips to the Alps. From 2008 crisis, this phenomenon has been transforming toward a more equal behavior among the four series, but it can also be confirmed that tourism recovers quicker from the recession period, although its first impact may be sharper.

After these countries regions analyses have been performed and following the structure, long-term-dependent relations between regional overnight stays and their main incoming countries’ GDPs are examined. These cointegration tests (Table 4), as explained before, are organized as bivariate models composed, in this case, of the overnight stays series for Salzburg and the GDPs series of the major incoming countries travelling toward the chosen territory (Germany, Netherlands, and Italy), one at a time. Results could be misleading if several GDPs series are put together in the same model, as long-run relations are frequent between them. To avoid this effect, models have been articulated in pairs following the scheme of a regional overnight stays series and a GDP one.

Table 4.

Dickey–Fuller tests for Salzburg ADF output for Austria-related series (winter overnight stays and GDPs), 1994–2016.

Level series	SALZBURG (LOG)		NETHERLANDS (LOG)		GERMANY (LOG)		ITALY (LOG)
Level series	t Stat	t Stat	p Value	p Value	t Stat	p Value	t Stat	p Value
Intercept and trend	0.38	−4.21	0.016	0.3854	−3.93	0.028	−1.52	0.7905
Intercept	0.06	−0.30	0.908	0.0666	−0.57	0.856	−2.80	0.0732
None	0.97	2.54	0.995	0.9733	2.61	0.99	1.45	0.9591
	D SALZBURG (LOG)		D NETHERLANDS (LOG)		DGERMANY (LOG)		D ITALY (LOG)
First differenced series	t Stat	t Stat	p Value	p Value	t Stat	p Value	t Stat	p Value
Intercept	−3.50*	−4.97 *	0.0008	0.1100	−4.83*	0.0011	−3.37*	0.0239
	D²LSALZBURG		D²LNETHERLANDS		D²GERMANY (LOG)		D²LITALY
2nd differenced series	t Stat	t Stat	p Value	p Value	t Stat	p Value	t Stat	p Value
Intercept	−5.3*	−5.39 *	0.0004*	0.0017	−5.29*	0.0005	−5.70*	0.0020

GDP: gross domestic product; LOG: logarithm.

For the Austrian region of Salzburg, no cointegrating equations were found under any of the two possible criteria (Online Appendix A, Tables 13–15). On this basis, no extended analysis is elaborated for these series.

Once cointegration has been confirmed for some of the chosen series, they are suitable to be modeled through an ECM (Engle and Granger, 1987). This ECM can be formulated when series present a cointegration relationship and it could be extended toward any dynamic ADL model (p, q1, q2,…, qk). Despite the fact that the existence of long-run relations is tested in this work through this cointegration analysis, a possible extension of it could be the modelling of these relations, which is beyond the purpose of the current article.

Conclusions

This work has made an attempt at examining winter tourism evolution, specifically ski and mountaineering-related tourism, a phenomenon that takes place in the three French and Austrian alpine regions (Savoie, Tyrol, and Salzburg), among others. These regions are home to four of the most important ski resorts and stations worldwide.

In the first place, a cycle decomposition procedure has been employed to do so, subtracting the trend component from the series (using a Hodrick–Prescott filter). This method permits a comparison between the regional behavior of winter tourists and the total country winter visitors, trying to identify a long-term relationship. Furthermore and given that a significant portion of total tourism overnights stays in ski resorts in the selected regions are incoming visitors, the aim of this study was to check the dependency grade of the tourist cycle component in relation to the GDP gaps of the main incoming tourism European countries regarding these three regions.

Cyclical component investigation results prove that, for the case of the French Alpine region of Savoie, no important correlation seems to exist between regional and total winter tourism in France. This fact must be an answer to the low weight this region holds concerning total winter tourism demand in France. In contrast, an important relationship between the evolution of the winter tourism overnight stays for the Tirolean region and those for its country, Austria, is acknowledged by data. This region congregates more than half of total tourists visiting Austria in winter season. Despite this fact, results for the Salzburg NUTS2 region cannot confirm such a clear common pathway, due to winter tourism in this region being also related to other activities on the outside of skiing or mountaineering.

A complementary cointegration analysis has been performed on the series, searching for a long-term linear or nonlinear relationship between them. To identify the presence of this long-sighted relation Johansen test has been performed on the group of series, elaborating bivariate models with them. External dependence is proven after this test for Savoie’s winter tourism demand with Germany, Netherlands, and Belgium’s economic contexts (the main winter incoming tourist countries to this region) and also for the Tirolean winter demand, which cointegrates with Belgium, Germany, and Netherlands’s GDPs.

To sum up, there is no dependence between France and Savoie or between Austria and Salzburg attending to neither of the criteria employed. A weak cyclical dependence is shown between Savoie and its main incoming tourism partners (Germany, Belgium, and Netherlands). Attending to a long-run dependence analysis, carried out with the cointegration analysis, overnight stays for Savoie shows cointegration with all the three GDP series. The existence of both a cyclical and a cointegration dependence shows an important relation between these series. As a reminder, it is important to take into account that models were designed in peers, so the outcome relations of the procedure are always between the regional overnight stays and a GDP series and not between GDP series one another.

The Tirolean region shows an important cyclical dependence with Austria, attending to the correspondent winter overnight stays series. According to a long-run tendency correlation, overnight stays present cointegration equations with all GDP series. This means that, in a short-term, evolution of the tourism demand is related to the evolution of this same concept in Austria meanwhile, in a long-term analysis, tourism demand growth will be linked to the GDP gaps of its main tourism incoming partners. Finally, Salzburg region only has a cyclical correspondence with its main incoming tourism partners (Germany, Netherlands, and Italy), allowing a conclusion of absence of a common trend between tourism demand and GDP gaps. This region contains an important cultural destination within its limits, and this fact must be taken into account when drawing conclusions. The city of Salzburg is currently experiencing an important demand massification process, with a ratio between visitors and inhabitants around 12%, according to WIFO (2019). This makes it similar to cities as Amsterdam or Lisbon. Therefore, it is somewhat logical that tourism demand responds to other different drivers in addition to GDP, so results obtained from this analysis are aligned with the reality of the destination.

Both Netherlands and Belgium’s GDPs present linear combinations with two of our three studied European regions. Thus, the economic contexts of these countries are key for the development and growth of winter tourism activities in these regions and seem to have a strong influence on ski-related tourism.

It has been possible to demonstrate several links between main incoming winter partners’ GDP gaps and winter tourism demand evolution. Having left behind the important 2008 global recession, genuinely shocking for the European region, economic growth in the European area is expected to be present for the next years. If this GDP trend is related to winter tourism demand for our ski regions, overtourism problem in the studied areas is expected to be compounded. With an increasing problematic in tourism overcrowding, policymakers and the different stakeholders of these ski resorts will have to take action to dampen these negative externalities. It is essential for them to make an effort to preserve this special tourist attractions and diminish potential negative externalities affecting natural environment and tourists’ experiences.

The studied alpine destinations are, as has been previously mentioned, severely affected by this overtourism problematic. Despite the existence of underutilization problems on some ski resorts, especially smaller ones (Falk and Steiger, 2018), most of the stakeholders related to this tourist attractions have previously demonstrated their competitiveness up to this moment, when their ability to compete is jeopardized in a short/medium term by this overtourism phenomenon. Sustainability of tourism must be thought about in the long-run, and all the efforts and policy decisions associated with it should be geared toward lifelong structural changes in the sector. Several studies have found that Research, Development, and Innovation are the fundamental pillars for it. In lockstep, encouragement and attraction of human talent, progression, and understanding of digital economy (Fernández-Alcantud et al., 2017; Moreno-Izquierdo et al., 2016). It is also crucial to establish a cooperation and partnership culture, between public and private actors, to generate added value through innovation and knowledge transference ecosystems, conforming true enterprise networks that will surely help destination management and performance.

Some limitations of this study comprise that data series are not very extensive and this fact could have affected both cyclical and cointegration analysis, although they include enough observations to allow the performance of this analysis. Aside from this, overnight stays data employed refer to winter seasons in the regions. Despite the fact that this winter seasons overnight present a clear and direct relationship with ski and mountaineering activities; it is not possible to completely attribute the full range of winter overnight stays to these activities. As has been explained before in the results section, Salzburg region is especially sensitive to this due to having a wider touristic offer and, thus, this attribution is more compromised in this particular case. This study could become more accurate if ski resorts’ data were available. As an addition to this, accurateness of results should be enhanced if decomposed overnight stays according to visiting countries data were available since both regional and national overnight stays are an aggregated series of domestic and foreign demand. In short, limitations of the study mainly concern an important lack of public data on this ski and mountaineering-related tourism in the alpine region, which hinders substantially the study of such an important source of development, employment, and economic growth for these regions.

A future extension of this work could be the specification of the model containing the cointegrated series trough the formulation of its ECM. A dynamic or time-varying ECM should be considered if a potential future study focuses its purpose on the evolution of this dependence relation. This will allow an accurately measure for the weight of each economic series on the tourism winter demand for these regions.

Supplemental material

Supplemental Material, sj-pdf-1-teu-10.1177_1354816620932007 - Winter tourism dependence: A cyclical and cointegration analysis. Case study for the Alps

Supplemental Material, sj-pdf-1-teu-10.1177_1354816620932007 for Winter tourism dependence: A cyclical and cointegration analysis. Case study for the Alps by Patricia Aranda-Cuéllar, José María López-Morales and María Jesús Such-Devesa in Tourism Economics

Footnotes

Declaration of conflicting interests

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

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: Ayudas de Iniciación en la Actividad Investigadora [Initiation on Research Grants] by Universidad de Alcala, under the Young Researchers funding, Programa Propio. ID: 347275.

ORCID iD

Patricia Aranda-Cuéllar

Supplemental material

Supplemental material for this article is available online.

Author biographies

Patricia Aranda-Cuéllar is a PhD student and researcher at the Economics Department of Universidad de Alcalá, Spain. Master in Applied Economic Analysis by Universidad de Alcalá and Master in Strategies and Technology for Development by Complutense University of Madrid. In 2019, she obtained a predoctoral scholarship for research training awarded by Universidad de Alcalá.

José María López-Morales is a professor at the Department of Economics, University of Alcala, Spain. PhD in Economics from the Complutense University of Madrid and Master in Applied Economics from University Research Institute Ortega y Gasset. Researcher in the Tourism Sector Analysis and Innovation Group and in the Complex Systems in Social Sciences Group at the University of Alcala. His main research areas, the last few years, include sectoral studies of the Spanish economy, especially in the field of tourism economy.

María Jesús Such-Devesa is a tenured professor at the Economics Department of Universidad de Alcalá, Spain. She is a recognized researcher in tourism economics, with specialized research in tourism competitiveness, hotel industry management, internationalization, innovation, and digitalization and its impact on the tourism sector. She is editorial board member of Tourism Economics, Cuadernos de Turismo, Investigaciones Turísticas, among others.

References

Abegg

Agrawala

Crick

, et al. (2007) Climate change impacts and adaptation in winter tourism. In: Agrawala

(ed), Climate Change in the European Alps. Paris: OECD Publishing, pp, 25–60.

Alvarez-Sousa

(2018) The problems of tourist sustainability in cultural cities: socio-political perceptions and interests management. Sustainability 10(2): 503.

Archer

Cooper

Ruhanen

(2004) The positive and negative impacts of tourism. In: Theobald

(ed), Global Tourism Third Edition. London: Routledge, pp. 79–102.

Butler

Mao

(1997) Seasonality in tourism: problems and measurement. In: Murphy

(ed), Quality Management in Urban Tourism. Chichester, UK: Wiley.

Cheng

(2005) The Effects of Exchange Rate Misalignment on Producer Support Estimates: Empirical Evidence From India and China. Dissertation. Chennai: Virginia Tech

Cogley

Nason

(1995) Efects of the Hodrick-Prescott filter on trend and diference stationary time series: implications for business cycle research. Journal of Economic Dynamics and Control 19(1–2): 253–278.

De Jong

Sakarya

(2016) The econometrics of the Hodrick-Prescott filter. Review of Economics and Statistics 98(2): 310–317.

Dickey

Fuller

(1979) Distribution of the estimators for autoregressive time series with a unit root. Journal of the American Statistical Association 74(366): 427–431.

Engle

Granger

CWJ

(1987) Co-integration and error correction: representation, estimation, and testing. Econometrica 55(2): 251–276.

10.

Falk

(2013) A survival analysis of ski lift companies. Tourism Management 36: 377–390.

11.

Falk

(2015) The demand for winter sports: empirical evidence for the largest French ski-lift operator. Tourism Economics 21(3): 561–580.

12.

Falk

Hagsten

(2017) Climate change threats to one of the world’s largest cross-country skiing races. Climatic Change 143(1–2): 59–71.

13.

Falk

Lin

(2018) Sensitivity of winter tourism to temperature increases over the last decades. Economic Modelling 71: 174–183.

14.

Falk

Steiger

(2018) An exploration of the debt ratio of ski lift operators. Sustainability 10(9): 2985.

15.

Fernández-Alcantud

López Morales

Moreno Izquierdo

, et al. (2017) Innovación y destinos inteligentes: oportunidad para el know-how turístico español. Información Comercial Española 894: 137–150.

16.

García-Hernández

de-la-Calle-Vaquero

(2012) Capacidad de carga en grandes recursos turístico-culturales [Load capacity in large cultural tourism resources]. Anales De Geografía De La Universidad Complutense 32(2): 253–274.

17.

Goodwin

(2017) The challenge of overtourism. Responsible Tourism Partnership Working Paper 4, October 2017.

18.

Hamilton

(2018) Why you should never use the hodrick-prescott filter. The Review of Economics and Statistics 100(5): 831–843.

19.

Hodrick

Prescott

(1997) Postwar U.S. Business cycles: an empirical investigation. Journal of Money, Credit, and Banking 29(1): 1–16.

20.

Hudson

(2005) Winter sports destinations: dealing with seasonality. Sport Tourism Destinations: 188–204. DOI: 10.1016/B978-0-7506-5937-6.50021-3.

21.

Huete

Mantecón

(2018) El auge de la Turismofobia, ¿hipótesis de investigación o ruido ideológico? Pasos, Revista de Turismo y Patrimonio Cultural 16: 9–19.

22.

ITB Berlin (2018) Boom or Bust? Where is Tourism Heading. IPK International and ITB Berlin World Travel Trends Report. Berlin (2018).

23.

Johansen

(1988) Statistical analysis of cointegration vectors. Journal of Economics Dynamics and Control 12(2–3): 231–254.

24.

Johansen

(1991) Estimation and hypothesis testing of cointegration vectors in Gaussian vector autorregressive models. Econometrica 59(6): 1551–1580.

25.

Lasanta

Laguna

Vicente-Serrano

(2007) Do tourism-based ski resorts contribute to the homogeneous development of the Mediterranean mountains? A case study in the central Spanish Pyrenees. Tourism Management 28(5): 1326–1339.

26.

López-Morales

Such-Devesa

(2016) Economic cycle and external dependence on tourism: evidence for Spain. Tourism Economics 23(1): 187–199.

27.

López Moreno

Goyette

Beniston

(2014) Climate change prediction over complex areas: spatial variability of uncertainties and predictions over the Pyrenees from a set of regional climate models. International Journal of Climatology 28(11): 1535–1550.

28.

MacKinnon

Haug

Michelis

(1999) Numerical distribution functions of likelihood ratio tests for cointegration. Journal of Applied Econometrics 14(5): 563–577.

29.

Marcet

Ravn

(2004) The HP-filter in cross-country comparisons. CEPR Discussion Papers, 4244.

30.

March

Saurí

Llurdés

(2014) Perception of the effects of climate change in winter and summer tourist areas: the Pyrenees and the Catalan and Balearic coasts, Spain. Regional Environmental Change 14(3): 1189–1201.

31.

Moreno-Izquierdo

Ramón-Rodriguez

Such-Devesa

(2016) Turismo colaborativo: ¿Está AirBnB transformando el sector del alojamiento? [Collaborative tourism: Is AirBnB transforming the accommodation sector?]. Economistas 107–119.

32.

Nelson

Plosser

(1982) Trends and random walks in macroeconomic time series: Some evidence and implications. Journal of Monetary Economics 10(2): 139–162.

33.

Pearce

(2018) Limiting overtourism; the desirable new behaviours of the smart tourist. In Proceedings of the tourism intelligence forum (t-Forum) global exchange conference, Palma de Mallorca, Spain, 11–13 March 2018.

34.

Phillips

Jin

(2015) Business cycles, trend elimination, and the HP filter. Cowles Foundation Discussion Paper 2005. Yale University.

35.

Rapach

(2002) The long-run relationship between inflation and real stock prices. Journal of Macroeconomics 24(3): 331–351.

36.

Rodrigues

Freire-González

González Puig

, et al. (2018) Climate change adaptation of alpine ski tourism in Spain. Climate 6(2): 29.

37.

Rutty

Scott

Johnson

, et al. (2017) Using ski industry response to climatic variability to assess climate change risk: an analogue study in Eastern Canada. Tourism Management 58: 196–204.

38.

Trawöger

(2014) Convinced, ambivalent or annoyed: Tyrolean ski tourism stakeholders and their perceptions of climate change. Tourism Management 40: 338–351.

39.

Tuppen

(2000) The restructuring of winter sports resorts in the French alps: problems, processes and policies. International Journal of Tourism Research 2: 327–344.

40.

Van Duijn

(1977) The long wave in economic life. De Economist 125(4): 544–576.

41.

Vanat

(2019) 2019 International report on snow & mountain tourism. (Overview of the key industry figures for ski resorts). Geneva.

42.

Vera

Ivars

(2003) Measuring sustainability in a mass tourist destination: pressures, perceptions and policy responses in Torrevieja, Spain. Journal of Sustainable Tourism 11(2–3): 181–203.

43.

WIFO (10-12-2019) Auf Einen Salzburger Kommen 12 Touristen. Available at: https://www.salzburg24.at/news/salzburg/overtourism-11-9-touristen-in-salzburg-pro-einwohner-80460244 (accessed 27 June 2020).

44.

WTO-UNEP (2009) Trade and Climate Change. World Trade Organization and United Nations Environment Programme Report. Geneva.

45.

Zhang

Chen

(2010) HP filter analysis and cointegration test of import and export trade driving GDP growth—based on the analysis of china’s data from 1978 to 2007. International Trade Issues 2: 10–14.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.13 MB