The declining dependence of ski lift operators on natural snow conditions

Abstract

This study tests for a structural shift in the relationship between revenues of ski lift operators and natural snow conditions. The analysis is based on time series data for the Swedish ski lift industry spanning from 1980 to 2017. Since 1970, snow depth in winter sport destinations has decreased markedly by about 5 cm per decade. Estimations based on the autoregressive distributed lag (ARDL) model show that revenues (in constant prices) of ski lift operators are significantly positively related to natural snow conditions, given the impact of relative prices and real GDP. However, ARDL estimations with rolling windows reveal that the sensitivity of revenues from ski lift ticket sales to variations in snow depth is declining over time. For the subsamples starting at the end of 1980s onward, revenues no longer significantly depend on natural snow depth. This is likely due to the implementation of adaptation measures such as investments in snowmaking facilities.

Keywords

autoregressive distributed lag (ARDL) model climate change skiing demand snow depth structural shift Sweden winter tourism

Introduction

The Intergovernmental Panel on Climate Change (IPCC) (2014) concludes that Northern Europe in the winter season is particularly affected by global warming. Regionalized climate models for the Scandinavian mountains show a temperature increase in the winter months (December to February) of about 5°C in the period 2071–2100 as compared to the period 1971–2000, based on the high emission future emissions scenarios (RCP 8.5) (EURO-CORDEX) (Strandberg et al., 2014; Figure 1A in the Online Appendix). For the Swedish provinces Dalarna and Jämtland, Tranos and Davoudi (2014) estimate a reduction in the number of days with snow coverage of between 55 days and 78 days until the period 2070–2100. According to the Swedish Meteorological and Hydrological Institute (SMHI), the maximum snow depth during the winter season declined between the period 1991 and 2014 as compared to 1961 and 1990 all across Sweden, except in northern Norrland (Wern, 2015). Despite the relatively high latitude, the location of ski areas at relatively low elevations up to 1200 metres above sea level makes the area vulnerable to climate variability.

Figure 1.

Loading and long-run coefficients from rolling subsample. Notes: t = 1 denotes subsample 1979/1980 to 2008/2009 and t = 9 denotes subsample 1987/1988 to 2016/2017.

The aim of this article is to investigate possible structural shifts in the relationship between natural snow depth and revenues of ski lift operators. An autoregressive distributed lag (ARDL) model combined with recursive estimation techniques is used to test whether the relationship between snow depth and revenues (in constant prices) changes over time. The main expectation is that the dependency decreases over time due to successful adaptation measures.

Many studies investigate the relationship between natural snow conditions and output of ski lift operators, measured as total skier visits or ski lift revenues. Studies using high-frequency data (daily or weekly) typically find a strong relationship between snow depth and skier visits (Damm et al., 2014; Demiroglu et al., 2015; Englin and Moeltner, 2004; Hamilton et al., 2007; Holmgren and McCracken, 2014; Pullman and Thompson, 2002; Shih et al., 2009). This is expected since good snow conditions usually bring more skiers and snowboarders to the ski areas. Other studies find a significant relationship between temperatures (measured as wind chill temperatures) and skier visits (Malasevska et al., 2017).

When analysing the climate change effects on skier demand or revenues of ski lift operating companies, data for the winter season over longer periods of time are appropriate. However, data for whole winter seasons imply that the results are averaged out as compared to studies based on daily or monthly data, if they are available. The findings based on annual data are not clear cut. While some studies find that snow depth accounts for the majority of the yearly variation of the visitation frequency over time (Bark et al., 2010; Pickering, 2011), other studies find that the magnitude of the relationship is quite modest (Dawson et al., 2009; Falk, 2015) or concentrated in the early season (Falk and Hagsten, 2016).

Few studies have investigated the link between year-to-year fluctuations in snow conditions and revenues of ski lift operators or skier visits using a sufficient long-time period to test structural shifts in the relationships. This is partly related to a lack of consistent data over time. For instance, recent studies including Falk (2015), Falk and Hagsten (2016), Falk and Vieru (2017), Gonseth (2013) and Pickering (2011), all use relatively short time periods – between 5 years and 20 years – that do not allow to test for the presence of a structural shift in the relationship between snow depth and outcome of ski lift operators. Since the beginning of the 1990s, ski lift operators have undertaken numerous adaptation strategies such as investments in snowmaking facilities, snow farming, expansion to higher areas and diversification (Source: SkiStar company annual reports). One implication might be that ski areas are becoming less dependent on natural snow supply.

A related literature uses overnight stays in winter sport destinations or regions to analyse the tourism effects of climate variability (Damm et al., 2017; Falk and Lin, 2018; Töglhofer et al., 2011). Overnight stays at ski resorts is an imperfect proxy of skiing demand but allows the study of the wider effects of climate variability. For instance, Töglhofer et al. (2011) find that the relationship between overnight stays and snow cover is decreasing over time, although still significant in the most recent period. Similarly, Falk and Lin (2018) show that the effect of temperature shocks on winter overnight stays declines in recent years. In this case, it is important to note that visitors come to winter sport destinations not only to ski but also for various other reasons (Bausch and Unseld, 2017). Using revenues of ski lift operating companies as an economic outcome measure makes it possible to study the direct effect of climate variability on the output of the possibly affected industry.

This study provides a first investigation into the stability of the relationship between natural snow conditions and skiing demand using time series of almost 40 years. While there are several studies investigating the relationship between skier visits or winter overnight stays and snow conditions for downhill skiing, no study has explicitly investigated the presence of a structural shift in this relationship. The article contributes to the growing literature on the relationship between climate variability and tourism demand (see Becken, 2013; Fang et al., 2018; Kaján and Saarinen, 2013; Rosselló-Nadal, 2014; Scott et al., 2012a, 2012b for surveys of the literature). Recent studies have employed either time series data (Goh, 2012; Zhang and Kulendran, 2017) or panel data on several origin or destination regions (Li et al., 2017, 2018; Töglhofer et al., 2011).

The article is structured as follows: next section presents the conceptual background and the empirical model, while third section contains the description of the data and descriptive statistics. Fourth section contains the main empirical results. The last section contains the conclusion and implications.

Conceptual background and empirical model

The empirical model can be derived from the theory of recreational or tourism demand (Englin and Moeltner, 2004; Morey, 1984; Peng et al., 2015; Song and Li, 2008; Song et al., 2009). Following the literature, skiing demand is specified as a function of economic variables and weather-related factors (King et al., 2014; Riddington, 1999, 2002; Shih et al., 2009). Economic factors include prices or relative prices and real income of the visitors, and thus are similar to those of related tourism activities (for instance visits of theme or national parks). In this study, revenues from ski lift tickets are specified as a function of (relative) prices of ski lift tickets, real income of the visitors and snow conditions. The key variable in this study is snow conditions measured as average snow depth for the winter season.

Previous studies using annual data for the entire winter season find a significant relationship between snow depth (also referred to as snow pack) in the ski area and the number of skier visits (Gonseth, 2013; Falk, 2015; Falk and Vieru, 2017; Pickering, 2011). A sufficient level of snow is a necessary factor for skiing. In the absence of powerful snowmaking facilities, a lack of snow can lead to delays in the opening of resorts, early closures in spring, shorter opening hours and closed lifts or terrain in the lower sections of the resorts. However, there is no consensus on the magnitude of the relationship between skiing demand and natural snow depth.

Adaptation measures can have a large impact in reducing negative impacts of climate change (Burke et al., 2016). Extreme weather seasons can lead the affected operators or destinations to intensify their adaptation measures, which makes the industry less vulnerable to extraordinary warm or snow poor winter seasons. This in turn can lead to a shift in the relationship between economic outcome and climate variability. After the extreme mild and snow poor winter season 1989/1990 in Europe, ski lift operators invested massively in snowmaking technologies. This is also the case for the ski areas under investigation that currently exhibit a share of slopes covered by snowmaking facilities of more than 60% (Source: SkiStar company annual report). Gonseth (2013) finds considerable heterogeneity in the relationship between skier visits and snow conditions depending on the coverage of slopes with snowmaking facilities using panel data for Swiss ski resorts. This means that the assumption of a stable relationship between output of ski lift operators and snow conditions over time can be questioned. Thus, the main hypothesis in this study is that these investments make the industry less vulnerable to variations in natural snow depth, and consequently that the relationship between snow depth and outcome of ski lift operators is declining in magnitude over time.

The static long-run equilibrium of revenues of ski lift operators at the aggregate level can be specified as follows:

\begin{matrix} ln {REVcp}_{t} = α_{0} + L_{GDPPC} ln {GDPPC}_{t} + L_{P CPI} ln (P_{t} / {CPI}_{t}) \\ + L_{SNOWDEPTH} ln {SNOWDEPTH}_{t} + v_{t} \end{matrix}

where t denotes time (winter season 1977/1978 to 2016/2017) and ln(·) represents the natural logarithm. The left-hand variable REVcp denotes ski lift revenues for the winter season, deflated by the price index of ski lift tickets for the corresponding period. GDPPC represents GDP per capita in constant prices, measured as the average of the last quarter and the subsequent first quarter. The Swedish domestic GDP per capita is used as a proxy of real income given the high share of domestic visitors (80%). P denotes the price index of ski lift tickets and CPI denotes the consumer price index for the winter season; both measured as the average from November to April. SNOWDEPTH refers to the average monthly snow depth of the nearest weather station in the provinces of Jämtland and Dalarna. Besides snow depth for the entire winter season (defined as December to April), snow depth of the early winter season (defined as December) is used as an alternative measure. This is particularly interesting because over the period 1978–2016, snow depth statistics for December show a stronger downward trend (source: SMHI). L_GDPPC, L_PCPI and L_SNOWDEPTH are the so-called long-run coefficients representing the elasticities of demand. We are particularly interested in the last one, which is elasticity of skiing demand with respect to snow depth. The long-run error term v_t represents the deviation of revenue from the long-run equilibrium at period t. The long-run model (1) can be estimated by an ARDL specification that can be rewritten further in an error correction mechanism form:

\begin{matrix} Δ ln {REVcp}_{t} = α_{1} [ln {REVcp}_{t - 1} - α_{0} - L_{GDPPC} ln {GDPPC}_{t - 1} - L_{P CPI} ln (P_{t} / {CPI}_{t}) \\ - L_{SNOWDEPTH} ln {SNOWDEPTH}_{t - 1}] + \sum_{i = 1}^{p} β_{i} Δ ln {REVcp}_{t - i} \\ + \sum_{i = 0}^{q 1} β_{i}^{GDPPC} Δ ln {GDPPC}_{t - i} + \sum_{i = 0}^{q 2} β_{i}^{P CPI} Δ ln (P_{t - i} / {CPI}_{t - i}) \\ + \sum_{i = 0}^{q 3} β_{i}^{SNOWDEPTH} {Δ SNOWDEPTH}_{t - i} + ε_{t} \end{matrix}

where α₁ denotes the adjustment coefficient and β’s are the short-run coefficients. The error-correction model can be estimated by ordinary least squares (OLS). To avoid the problem of spurious regression, the bounds test (Pesaran et al., 2001) is carried out here to test the cointegration relation among the variables in levels. The bounds test that has the null hypothesis of non-cointegration requires at least 30 observations according to the critical values in Narayan (2005).

The optimal lag lengths for p, q1, q2 and q3 can be determined by R ², and/or standard Akaike, and/or Schwarz and/or Hannan–Quinn information criteria. We also carry out various diagnostic tests for the short-run residuals ε_t: the Lagrange multiplier (LM) autocorrelation, the Jarque–Bera normality, Breusch–Pagan’s heteroscedasticity and the Ramsey functional form (FF) tests.

The main research question is to test the presence of a structural break in the relationship between snow depth and skiing demand. Time-varying models are commonly employed to investigate the stability of the parameters over time (Pérez-Rodríguez et al., 2015; Song and Wong, 2003). Popular techniques are rolling regression technique (Tang and Abosedra, 2016) or recursive regression analysis where one new observation is added to the end of each sample (Tang and Tan, 2013). Another popular technique to detect a structural shift is the Kalman filter method applied to the error-correction model (Li et al., 2006; Song and Wong, 2003). Note that these approaches require a sufficiently large initial subsample of T observations. Given the relatively small number of observations, recursive estimations are used where parameters are allowed to vary over time. In addition, the stability of parameters is tested by the cumulative sum of the recursive residual (CUSUM) and CUSUM squares (CUSUMSQ). The two tests are based on the analysis of the 95% confidence band. The main hypothesis is that the sensitivity of revenues of ski lift operators to variations of natural snow conditions declines over time. This is mainly related to extensive investments in snowmaking facilities and other adaptation measures. In particular, we employ the rolling estimation technique to estimate (2) recursively based on subsamples with a fixed rolling window size of 30 observations. The procedure is starting from the sample 1979/1980 to 2008/2009, and then the sample will be extended by 1 year each until the sample 1987/1988 to 2016/2017 is reached.

Specification (2) is re-estimated for several sub-periods and bounds, and other diagnostic tests are carried out in each of the rolling subsamples. This is based on the fact that the presence of cointegration in the full sample may not imply similar conditions in all subsamples. At the same time, the best fitted models might also be different, in terms of determinations of p, q1, q2 and q3. Thus, the approach is to repeat the whole process for each subsample in order to guarantee the best fitted models. We then plot the point estimates of each parameter and associated 95% confidence intervals for all rolling windows. The goal is to capture possible changes in parameters over time.

Data and descriptive statistics

The Swedish mountain region is one of few areas in Europe that still shows a slight rise in revenues from ski lift tickets. At the same time, this region is more affected by the increase in global warming than the European Alps, according to climate change scenarios (EURO-CORDEX). SMHI (2017) expects an increase in winter (minimum) temperatures by 4–7° until the end of the millennium (2070–2100), given that greenhouse emissions continue to expand in line with the medium or high forecasts (RCP 4.5 or RCP 8.5).

Snow-based sport activities such as downhill skiing and cross-country skiing are the main attraction for visitors to the Swedish mountains in the winter season. Ski lift operators generally provide the whole infrastructure for skiers and snowboarders, including transportation, snowmaking and preparation of slopes.

Demand for downhill skiing can be measured in several ways: skier visits, revenues or persons transported uphill. Ski lift revenues are often regarded as a better measure of output of ski lift operators than skier visits because the number of skier visits often also include those who do not pay (small children) or those who have received a discount (older people and seasonal card holders). Revenues may also capture better the trend towards part-day tickets.

In this study, demand for downhill skiing is measured as total lift ticket revenues for the winter seasons. Revenues from ski rentals, accommodation and restaurants, which often belong to the ski lift operators, are not included. The advantage of the use of lift ticket sales as compared to overnight stays is that day trippers are included. It may well be the case that skiers or snowboarders visit the area only for day trip and do not stay overnight in a hotel, holiday home or in a second home.

Total lift ticket revenues for the winter seasons 1977/1978 to 2016/2017 are provided by the Swedish ski areas industry association (Svenska Liftanläggningars Organisation, SLAO, http://www.slao.se). Revenues are concentrated to two provinces, Jämtland and Dalarna, accounting for about 85% of total ski lift revenues in the country. Lift ticket revenues REVcp are measured exclusive of VAT and are deflated by the price index of ski lift tickets, provided by Statistics Sweden. Since a price index for ski lift tickets is available only from the mid of 1990s onward, the price of transport services is used for the earlier years of the time series. We use the transport price, P, as proxy for ski lift tickets. The remaining data, domestic GDP per capita in constant prices at the quarterly level (for the last quarter and the first subsequent quarter) (GDPPC), and the consumer price index (CPI) are found in the Statistics Sweden databases (http://www.scb.se).

Information on weather indicators is drawn from the SMHI (http://www.shmi.se). SMHI provides daily figures for all of its weather variables, including temperature and snow depth (http://www.smhi.se). Data on different weather variables for about 200 weather stations can be downloaded free of charge. Since ski areas in the two provinces Jämtland and Dalarna account for 85% of total ski lift revenues, we only select weather stations located in those two provinces. The criterion for selection of the station is its closeness to the main ski areas. We first download daily snow depth data for four weather stations in each of the two provinces (Dalarna: Idre, Särna, and Trängslet; and Jämtland: Storlien, Mörsil, Myskelåsen, and Höglekardalen). Daily snow depth data are first aggregated at the monthly level, and then for each province, average snow depth for the months December to April and December is calculated. Finally, snow depth is weighted by the share of winter overnight stays of the two provinces (weights for Dalarna and Jämtland are 0.72 and 0.38, respectively). Two different measures of daily snow depth data are aggregated to the monthly level: One refers to the December average and the other is based on the average snow depth from December to April.

The sample period covers the winter seasons 1979/1980 to 2016/2017 with 38 annual observations. For rolling estimations, each subsample contains 30 observations to fulfil the minimum requirement for bounds test according to Narayan (2005). Thus, there are a total of nine rolling subsamples.

Lift ticket revenues in constant prices increase steadily until the end of the 1980s and then decline considerably during the recession period 1992/1993 (Figure 1B in the Online Appendix). In the following years, sales recovered with a strong upward trend until 2010. Most recently, there is first a slight rise, followed by a strong growth in real terms during the last two seasons, despite the lower than average snow depth. Snow depth of the eight weather stations for the entire winter season declines significantly over the study period; from an average of 58 cm during the period 1979/1980 to 1988/1989 to 38 cm on average in the years 2010/2011 to 2016/2017. The downward trend is significant at the 1% level with a coefficient of −0.53, indicating that average snow depth is declining by 5 cm per decade on average during the sample period.

Empirical results

First, unit-root tests for all variables are carried out. We employ the augmented Dickey–Fuller (ADF) (Dickey and Fuller, 1979), PP (Phillips and Perron, 1988), KPSS (Kwiatkowski et al., 1992) and ADF tests allowing for a break. The ADF, PP and ADF with a break tests have the unit root as the null. The KPSS has a null of stationarity. The ADF and PP tests cannot reject the null of unit root for all variables except snow depth (Table 1).

Table 1.

Descriptive statistics and unit-root tests.

	lnREVcp	lnGDPPC	ln(P/CPI)	lnSNOWDEPTH
Descriptive statistics
Mean	5.15	14.22	0.38	−0.78
Standard deviation	0.18	0.24	0.21	0.28
Unit-root tests^a
Levels
ADF	−2.58	−2.93	−1.92	−4.29***
PP	−2.75	−2.39	−1.72	−4.31**
KPSS	0.13*	0.08	0.18**	0.10
Break	−3.93	−4.03	−4.51
Differences
ADF	−5.64***	−5.33***	−6.64***
PP	−5.99***	−5.10***	−7.17***
KPSS	0.08	0.09	0.36*

Note: The t-statistics are reported for ADF, PP unit-root tests and the unit-root test with one break; the LM-statistic for KPSS unit-root test. ADF: augmented Dickey–Fuller; PP: Phillips and Perron; KPSS: Kwiatkowski, Phillips, Schmidt and Shin.

***Significant at 1%.

**Significant at 5%.

*Significant at 10%.

However, KPSS tests reject the null hypothesis that the series are I(0) at 10% level for real revenues (lnREVcp) and at 5% level for relative prices, ln(P/CPI). The KPSS cannot reject the null for real GDP per capita (lnGDPPC) and snow depth (lnSNOWDEPTH). Thus, snow depth is stationary across the different unit-root tests, and relative ski lift ticket prices are non-stationary in all cases. The mixed results for real revenues and real GDP per capita might be due to structural breaks. We, therefore, carry out the unit-root tests with a break. The result shows that real revenues, GDP per capita and relative ski lift ticket prices are all non-stationary. The unit-root tests for the first-order differences of lnREVcp, lnGDPPC and ln(P/CPI) show that they are stationary. To conclude, the unit-root tests indicate that all variables are integrated of order 1 except snow depth, which is stationary. Given that the dependent variable and two explanatory variables are integrated of order one and snow depth is stationary, ARDL model seems the best choice for estimating long-run parameters in equation (1).

Table 2 shows the OLS estimations of the ARDL models for the whole sample as well as for each rolling window. The bounds tests show that the null hypothesis of non-cointegration can be rejected in all cases, at least at the 5% level. This indicates a long-run relationship among lift ticket revenues, real GDP, relative prices and snow conditions for the whole period 1979/1980 to 2016/2017, as well as for the nine rolling windows. At the same time, all samples pass all diagnostic and stability tests. Note that for one subsample, a dummy variable has to be added so that both CUSUM and CUSUMSQ tests can be passed.

Table 2.

ARDL results of the determinants of ski lift revenues in constant prices.

	Whole sample 1980–2017	Subsample 1980–2009	Subsample 1981–2010	Subsample 1982–2011	Subsample 1983–2012	Subsample 1984–2013	Subsample 1985–2014	Subsample 1986–2015	Subsample 1987–2016	Subsample 1988–2017
Loading and long-run coefficients and standard errors
α₁	−0.4331*** (0.11)	−0.5206*** (0.12)	−0.5890*** (0.16)	−0.5752*** (0.16)	−0.6420*** (0.14)	−0.6966*** (0.13)	−0.7450*** (0.14)	−0.8666*** (0.22)	−0.8764*** (0.23)	−1.2369*** (0.28)
L _GDPPC	1.3438*** (0.46)	0.8219 (0.48)	0.4148 (0.49)	0.2415 (0.53)	0.7670 (0.47)	0.8682* (0.43)	0.8093** (0.36)	2.2272*** (0.60)	2.5184*** (0.68)	2.3870*** (0.29)
L_PCPI	−0.6242 (0.52)	0.1386 (0.52)	0.4030 (0.55)	0.6129 (0.63)	−0.0442 (0.54)	−0.1625 (0.49)	−0.1276 (0.42)	−1.8317** (0.76)	−2.2005*** (0.68)	−2.0013*** (0.37)
L_SNOWDEPTH	0.5018*** (0.16)	0.5880*** (0.14)	0.7168*** (0.16)	0.7393*** (0.16)	0.7081*** (0.13)	0.6989*** (0.12)	0.7130*** (0.11)	0.1043 (0.25)	−0.0231 (0.19)	0.0876 (0.10)
α₀	−13.31** (6.36)	−6.04 (6.73)	−0.36 (6.81)	−2.04 (7.44)	−5.16 (6.50)	−6.55** (5.91)	−5.79 (5.01)	−57.70*** (8.45)	−29.80*** (7.49)	−21.92*** (4.06)
dummy2006							0.1193* (0.06)
Tests statistics
F _PSS	7.95^†	6.78^††	4.53^†††	5.35^†††	6.82^††	7.05^††	8.57^††	4.07^‡	5.00^†††	5.73^†††
R²/adj R²	0.88/0.86	0.85/0.82	0.87/0.83	0.88/0.85	0.91/0.88	0.92/0.88	0.93/0.90	0.94/0.89	0.94/0.90	0.97/0.92
AIC/SIC	−2.39/−2.17	−2.46/−2.18	−2.44/−2.11	−2.46/−2.13	−2.63/−2.21	−2.64/−2.22	−2.75/−2.28	−2.56/−1.91	−2.53/−1.87	−2.76/−1.96
$χ_{H}^{2}$	10.00	10.19	5.71	6.22	7.89	7.36	11.48	7.60	8.28	5.93
$χ_{SIC}^{2} (6)$	12.82**	12.53*	7.97	6.76	3.63	3.44	3.52	11.42*	8.90	6.41
$χ_{FF}^{2}$	0.09	1.06	1.04	0.19	0.89	1.77	0.99	0.61	0.05	1.75
$χ_{normal}^{2}$	3.14	1.88	1.53	1.02	0.62	0.23	0.64	1.18	1.68	1.45
CUSUM	Stable	Stable	Stable	Stable	Stable	Stable	Stable	Stable	Stable	Stable
CUSUMSQ	Stable	Stable	Stable	Stable	Stable	Stable	Stable	Stable	Stable	Stable

Note: The standard errors are presented in the parentheses. GDPPC: GDP per capita; PCPI: consumer price index adjusted ticket price; SNOWDEPTH: snow depth; CUSUM: cumulative sum; CUSUMSQ: CUSUM squares; PSS: Pesaran, Shin and Smith; AIC: Akaike information criterion; SIC: Schwarz information criterion; FF: function form.

***Significant at 1%.

**Significant at 5%.

*Significant at 10%.

^†Significant at 1% for the bounds test. The 1% critical value of upper bound, case 2, is 5.816 for a sample size of 35 according to Narayan (2005).

^††Significant at 1% for the bounds test. The 1% critical value of upper bound, case 2, is 5.966 for a sample size of 30 according to Narayan (2005).

^†††Significant at 5% for the bounds test. The 5% critical value of upper bound, case 2, is 4.306 for a sample size of 30 according to Narayan (2005).

^‡Significant at 10% for the bounds test. The 10% critical value of upper bound, case 2, is 3.586 for a sample size of 30 according to Narayan (2005).

The long-run coefficients can be directly interpreted as long-run elasticities. For the whole sample, the long-run elasticity of snow depth is 0.50, which is highly significant. This implies that an increase in average snow depth by 10% compared to the previous winter season will lead lift ticket revenues to rise by 5%.

Economic factors such as GDP and relative prices also have a significant impact on ski lift revenues. While the long-run GDP coefficient measures income elasticity, the long-run relative price coefficient captures the elasticity of relative ski lift ticket prices in a given ski area. Income elasticity is about 1.34, and the price elasticity is about −0.62. This means that a 1% increase in real GDP would lead to a 1.3% increase in revenues. Income elasticities greater than one are frequently interpreted to be associated with superior or luxury goods, and income elasticities lower than one with necessity goods. Meanwhile, the price elasticity is not significantly different from zero.

Figure 1 reports the results from the rolling estimations, including the individual long-run elasticities as well as their 95% confidence intervals over time. Note that parameters as well as their significance levels vary considerably over time. In particular, the rolling estimations reveal that income and price elasticities are becoming temporarily instable during, and the period after, the economic and financial crisis period.

The key result is that the long-run elasticity of revenues of ski lift operators with respect to natural snow depth declines dramatically over time and is no longer significant for the three most recent subsamples (subsamples 1985/1986–2014/2015 to 1987/1988–2016/2017). This clearly indicates that the output of ski lift operators is currently independent of variations in snow depth, in contrast to the earlier periods (1979/1980–2008/2009 to 1984/1985–2013/2014). For these subsamples, the long-run elasticity of snow depth ranges between 0.59 and 0.74 and are significant at the 1% level in all cases.

The results are consistent with Töglhofer et al. (2011) and Falk and Lin (2018) who suggest that the demand effects due to variations in snow depth or temperatures declined over time. This might indicate that the ski industry becomes more and more independent of climate variability. However, the results stand in contrast to earlier studies that show that revenues or number of skier visits significantly depend on snow depth for the total season (Falk and Vieru, 2017; Gonseth, 2013). For instance, based on Finnish ski resorts, Falk and Vieru (2017) find snow depth elasticities ranging between 0.1 and 0.3 depending on the latitude of the ski area.

One explanation that revenues are no longer dependent on snow depth is the widespread use of snowmaking facilities in Swedish ski areas. According to SkiStar – the largest ski lift conglomerate in Sweden – the share of slopes equipped with snowmaking facilities ranges between 60% and 80%. (Evidence is based on the ski areas Sälen, Åre and Vemdalen; Source: annual report 2014/2015.) However, given that the coverage of slopes with snowmaking facilities has already reached a high level, extension plans should be carefully reconsidered. Certainly, extensions of snowmaking facilities have implications broader than for the ski lift operator itself, such as environmental aspects. The large amounts of water needed for snow production are subject of criticism from environmental perspectives (De Jong, 2015; Rixen et al., 2011; Vanham et al., 2008). In the early winter season and in periods of intensive snowmaking, conflicts over water demand between local residents and ski lift operators are not uncommon (De Jong, 2015; Magnier and Reynard, 2012). Even if the proportion of energy needed for snowmaking is not large, possible rising costs affect the profit margin of the ski lift operators (Damm et al., 2014). In addition, snowmaking can have negative ecological impacts on biodiversity, vegetation and soil (Zu Schlochtern et al., 2014).

Several robustness checks are undertaken to estimate the sensitivity of results. First, an alternative measure of snow depth is used that captures the possible effect of a lack of snow in the early season. Lift ticket revenues for the entire season are likely to depend more on a lack of snow in the early part than in the main winter season. Results show that snow depth is positive and highly significant, indicating that revenues for an overall winter season are higher when winter arrives early. Again, the magnitude of the relationship becomes insignificant in recent years. Second, a quadratic term is included to test if a possible nonlinear relationship exists between lift revenues and snow depth. It is likely that the long-run elasticity of the relationship between snow depth and ski lift revenues declines in magnitude as snow depth increases. Unreported results show that the squared term is not significantly different from zero (results are available upon request). Finally, overnight stays in the winter months for the two provinces Jämtland and Dalarna is used as an alternative measure of skiing demand. Findings show that the role of snow depth declined over time and is no longer significant at the 1% level in the most recent subsample.

Conclusion and implications

This study investigates structural shifts in the relationship between revenues of ski lift operators and natural snow conditions, controlling for economic factors such as real GDP and ticket prices. Data used for the analysis are based on total revenues of ski lift operators in the Swedish mountains for the winter seasons 1979/1980 to 2016/2017. The results of the ARDL model with rolling windows show that the revenues of the ski lift operators are becoming less sensitive to variations in snow depth over time. From the subsamples 1985/1986 to 2014/2015 onward, revenues (in constant prices) are independent of variations in natural snow conditions. For the most recent subsamples, revenues only depend on real income and relative ski lift ticket prices.

The results indicate that large investments in snowmaking facilities and other adaptation measures pay off. However, adaptation measures might be a short- or medium-term solution (Pickering, 2011). With expected increases in temperatures of between 4°C and 7°C in the Swedish mountains in the next 50 to 80 years (SMHI, 2017), optimal conditions to produce snow will deteriorate, so that the early season and late season might be at risk. Another explanation could be that skiers have adjusted to worse natural snow conditions. The results imply that snow poor winter seasons, such as in the last few years, are no longer (or temporarily not) a threat for the performance of ski lift operators.

The results do not necessarily mean that climate change is no longer a threat for ski operators. Given the expected drastic increase in temperatures of between 5°C and 7°C in the next 80 years depending on the assumption of emission scenario (RCP 4.5 or 8.5), particularly the start of winter season may be at risk (depending on region). Such a temperature increase leads to a strong reduction of the number of days with optimal conditions to produce snow. Given the increase in summer temperatures, some advisable strategies include developing non-snow-based leisure activities for the non-snow season.

The results also indicate that the Swedish ski industry has developed better recently than its counterpart in the European Alps. The reason for this is not yet completely clear, but there are several possible explanations. One is the heavy investments in new lifts and efficient snowmaking systems, which makes it more comfortable for skiers and also allows a longer season. In the winter season 2016/2017, Swedish ski lift companies invested SEK 455 million in new ski lifts, ski runs and snowmaking facilities, where the latter accounts for 40% of total investment (Swedish ski area industry association, SLAO, 2017). This means that the level of snowmaking is relatively high compared with ski resorts in France and Switzerland (Spandre et al., 2015; Seilbahnen Schweiz, 2017). Investments have also been made to enhance service facilities, accommodations and other infrastructures including roads. These measures have made the winter holidays more attractive. Furthermore, lower flight prices and the supply of charter flights have attracted significantly more foreign visitors. The proportion of foreign investors is currently relatively low at 11% and thus there is real potential for more foreign guests. Another reason might be that Swedish residents have changed recently their travel behaviour due to the difficult security situation in some countries and prefer safer rural areas in their home country. Finally, less snow in south of Sweden in recent years makes spontaneous skiing day trips from home difficult and might lead to more organized trips to the Swedish mountains.

This study uses aggregate data from ski lift operators, which limits the interpretation of the results. For instance, it is likely that the relationship between natural snow cover and revenues varies across ski areas depending on elevation, latitude and snowmaking capacity. Disaggregated data for revenues and snow depth are needed to understand how and to what extent the relationships differ across ski resorts. This is left for future work. Further, the findings of the study can only be generalized for ski industries operating under similar conditions.

Supplemental Material

Supplemental Material, Supplemental_material - The declining dependence of ski lift operators on natural snow conditions

Supplemental Material, Supplemental_material for The declining dependence of ski lift operators on natural snow conditions by Martin Falk, and Xiang Lin in Tourism Economics

Footnotes

Acknowledgements

The authors would like to thank Noelle Crist-See and Caroline Wigerstad for careful proofreading of the article. Special thanks go to Eva Hagsten for helpful comments on an earlier version of the article.

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) received no financial support for the research, authorship and/or publication of this article.

Supplemental Material

Supplemental Material for this article is available online.

References

Bark

Colby

Dominguez

(2010) Snow days? Snowmaking adaptation and the future of low latitude, high elevation skiing in Arizona, USA. Climatic Change 102: 467–491.

Bausch

Unseld

(2017) Winter tourism in Germany is much more than skiing! Consumer motives and implications to Alpine destination marketing. Journal of Vacation Marketing, early view.

Becken

(2013) A review of tourism and climate change as an evolving knowledge domain. Tourism Management Perspectives 6: 53–62.

Burke

Craxton

Kolstad

. (2016) Some research challenges in the economics of climate change. Climate Change Economics 7(02): 1650002–01.

Damm

Köberl

Prettenthaler

(2014) Does artificial snow production pay under future climate conditions? A case study for a vulnerable ski area in Austria. Tourism Management 43: 8–21.

Damm

Greuell

Landgren

. (2017) Impacts of +2 C global warming on winter tourism demand in Europe. Climate Services 7: 31–46.

Dawson

Scott

McBoyle

(2009) Analogue analysis of climate change vulnerability in the US Northeast ski tourism. Climate Research 39: 1–9.

De Jong

(2015) Challenges for mountain hydrology in the third millennium. Frontiers in Environmental Science 3(38): 1–13.

Demiroglu

Kučerová

Ozcelebi

(2015) Snow reliability and climate elasticity: case of a Slovak ski resort. Tourism Review 70(1): 1–12.

10.

Dickey

Fuller

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

11.

Englin

Moeltner

(2004) The value of snowfall to skiers and boarders. Environmental and Resource Economics 29(1): 123–136.

12.

Falk

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

13.

Falk

Hagsten

(2016) Importance of early snowfall for Swedish ski resorts: evidence based on monthly data. Tourism Management 53: 61–73.

14.

Falk

Lin

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

15.

Falk

Vieru

(2017) Demand for downhill skiing in subarctic climates. Scandinavian Journal of Hospitality and Tourism 17(4): 388–405.

16.

Fang

Yin

(2018) Climate change and tourism: a scientometric analysis using CiteSpace. Journal of Sustainable Tourism 26(1): 108–126.

17.

Goh

(2012) Exploring impact of climate on tourism demand. Annals of Tourism Research 39(4): 1859–1883.

18.

Gonseth

(2013) Impact of snow variability on the Swiss winter tourism sector: implications in an era of climate change. Climatic Change 119(2): 307–320.

19.

Hamilton

Brown

Keim

(2007) Ski areas, weather and climate: time series models for New England case studies. International Journal of Climatology 27(15): 2113–2124.

20.

Holmgren

McCracken

(2014) What affects demand for the greatest snow on earth? Journal of Hospitality Marketing & Management 23: 1–20.

21.

Intergovernmental Panel on Climate Change (IPCC) (2014) Climate change, 2014: impacts, adaptation and vulnerability, Geneva. Available at: http://wwwipccch/pdf/assessment-report/ar5/wg2/WGIIAR5-Chap23_FINAL.pdf (accessed 22 March 2017).

22.

Kaján

Saarinen

(2013) Tourism, climate change and adaptation: a review. Current Issues in Tourism 16(2): 167–195.

23.

King

Abrahams

Ragsdale

(2014) Ensemble methods for advanced skier days prediction. Expert Systems with Applications 41(4): 1176–1188.

24.

Kwiatkowski

Phillips

Schmidt

. (1992) Testing the null hypothesis of stationary against the alternative of a unit root. Journal of Econometrics 54: 159–178.

25.

Goh

Hung

. (2018) Relative climate index and its effect on seasonal tourism demand. Journal of Travel Research 57(2): 178–192.

26.

Song

(2017) A dynamic panel data analysis of climate and tourism demand: Additional evidence. Journal of Travel Research 56(2): 158–171.

27.

Wong

Song

. (2006) Tourism demand forecasting: a time varying parameter error correction model. Journal of Travel Research 45(2): 175–185.

28.

Magnier

Reynard

(2012) Artificial snowmaking and potential water conflicts in mountain resorts. The case of Avoriaz (Haute-Savoie, France). In: EGU General Assembly Conference Abstracts, Vienna, Austria, 22–27 April 2012. vol. 14, p. 3761.

29.

Malasevska

Haugom

Lien

(2017) Research note: modelling and forecasting alpine skier visits. Tourism Economics 23(3): 669–679.

30.

Morey

(1984) The choice of ski areas: estimation of a generalized CES preference ordering with characteristics. The Review of Economics and Statistics 66(4): 584–590.

31.

Narayan

(2005) The saving and investment nexus for China: evidence from cointegration tests. Applied Economics 37(17): 1979–1990.

32.

Peng

Song

Crouch

. (2015) A meta-analysis of international tourism demand elasticities. Journal of Travel Research 54(5): 611–633.

33.

Pérez-Rodríguez

Ledesma-Rodríguez

Santana-Gallego

(2015) Testing dependence between GDP and tourism’s growth rates. Tourism Management 48: 268–282.

34.

Pesaran

Shin

Smith

(2001) Bounds testing approaches to the analysis of level relationships. Journal of Applied Econometrics 16(3): 289–326.

35.

Phillips

Perron

(1988) Testing for a unit root in time series regression. Biometrika 75: 335–346.

36.

Pickering

(2011) Changes in demand for tourism with climate change: a case study of visitation patterns to six ski resorts in Australia. Journal of Sustainable Tourism 19(6): 767–781.

37.

Pullman

Thompson

(2002) Evaluating capacity and demand management decisions at a ski resort. Cornell Hotel & Restaurant Administration Quarterly 43(6): 25–36.

38.

Riddington

(1999) Forecasting ski demand: comparing learning curve and time varying parameter approaches. Journal of Forecasting 18(3): 205–214.

39.

Riddington

(2002) Learning and ability to pay: developing a model to forecast ski tourism. Journal of Travel & Tourism Marketing 13(1–2): 109–124.

40.

Rixen

Teich

Lardelli

. (2011) Winter tourism and climate change in the Alps: an assessment of resource consumption, snow reliability, and future snowmaking potential. Mountain Research and Development 31(3): 229–236.

41.

Rosselló-Nadal

(2014) How to evaluate the effects of climate change on tourism. Tourism Management 42: 334–340.

42.

Scott

Hall

Gössling

(2012a) Tourism and Climate change Impacts, Adaptation & Mitigation. London: Routledge.

43.

Scott

Gössling

Hall

(2012b) International tourism and climate change. Wiley Interdisciplinary Reviews – Climate Change 3(3): 213–232.

44.

Seilbahnen Schweiz (2017) Fakten und Zahlen. Available at: https://www.seilbahnen.org (accessed 21 February 2018).

45.

Shih

Nicholls

Holecek

(2009) Impact of weather on downhill ski lift ticket sales. Journal of Travel Research 47(3): 359–372.

46.

Svenska Liftanläggningars Organisation (SLAO) (2017) Branschrapport 2016/2017. Available at: http://www.slao.se (accessed 21 February 2018).

47.

Swedish Meteorological and Hydrological Institute (SMHI) (2017) Climate projections cover different geographical areas, scenarios, time periods and climate indices. Available at: http://www.smhi.se/klimat/framtidens-klimat/ladda-ner-scenariodata?area=lan&sc=rcp85&var=t&seas=vin&sp=en (accessed 18 October 2017).

48.

Song

(2008) Tourism demand modelling and forecasting: a review of recent research. Tourism Management 29(2): 203–220.

49.

Song

Wong

(2003) Tourism demand modeling: a time-varying parameter approach. Journal of Travel Research 42(1): 57–64.

50.

Song

Witt

(2009) The Advanced Econometrics of Tourism Demand. New York; London: Routledge.

51.

Spandre

François

Morin

. (2015) Snowmaking in the French Alps. Climatic context, existing facilities and outlook. Journal of Alpine Research | Revue de géographie alpine 103(2), DOI: 10.4000/rga.2913

52.

Strandberg

Bärring

Hansson

. (2014) CORDEX Scenarios for Europe from the Rossby Centre Regional Climate Model RCA4. Reports Meteorology and Climatology, 116. Norrköping, Sverige: SMHI.

53.

Tang

Abosedra

(2016) Tourism and growth in Lebanon: new evidence from bootstrap simulation and rolling causality approaches. Empirical Economics 50(2): 679–696.

54.

Tang

Tan

(2013) How stable is the tourism-led growth hypothesis in Malaysia? Evidence from disaggregated tourism markets. Tourism Management 37: 52–57.

55.

Töglhofer

Eigner

Prettenthaler

(2011) Impacts of snow conditions on tourism demand in Austrian ski areas. Climate Research 46(1): 1–14.

56.

Tranos

Davoudi

(2014) The regional impact of climate change on winter tourism in Europe. Tourism Planning & Development 11(2): 163–178.

57.

Vanham

Fleischhacker

Rauch

(2008) Technical note: seasonality in alpine water resources management? A regional assessment. Hydrology and Earth System Sciences Discussions 12(1): 91–100.

58.

Wern

(2015) Snödjup i Sverige 1904/05–2013/14. Rapport Serie: Meteorologi 158 SMHI. NORRKÖPING. Available at: https://www.smhi.se/en/publications/snodjup-i-sverige-1904-05-2013-14-1.84652- (accessed 6 December 2016).

59.

Zhang

Kulendran

(2017) The impact of climate variables on seasonal variation in Hong Kong inbound tourism demand. Journal of Travel Research 56(1): 94–107.

60.

Zu Schlochtern

MPM

Rixen

Wipf

. (2014) Management, winter climate and plant–soil feedbacks on ski slopes: a synthesis. Ecological Research 29(4): 583–592.

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

0.09 MB