BACHELOR THESIS
Economy and Business Administration at the University of Amsterdam.
Olivia Strijd
University of Amsterdam 2018
Statement of Originality
This document is written by student named Olivia Strijd, who declares to take full responsibility for the contents of this document.
I declare that the text and the work presented in this document are original and that no sources other than those mentioned in the text and its references have been used in creating it.
The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents
I.
Abstract
This case study aims to investigate the effects of the Dutch consumer trust in the economy on the outbound tourism spending during and after the global crisis, which started in 2008. In the literature, there has not been a country-based research focusing specifically of the consumer trust on the holiday spending, while the role of this determinant has been mentioned to be important for the holiday spending analysis during the last global crisis. This paper will begin with motivation for doing this research, followed by the literature framework and hypotheses. In paragraph VI, the methodology of this research will be elaborated. Paragraph VII will show the empirical results, which at last will be followed by the conclusion in paragraph VIII. The findings of this paper could contribute in further expansion of empirical domains, which study tourism behavior.
II. Research question
How did the fall in consumer trust in economy, as a result of the last global crisis, affect the average outbound tourism spending of the Dutch households in the period of 2008 – 2014?
III.
Motivation
Many studies about the global recession and its impact on holiday demand have filled the scientific learning fields of both macroeconomics and microeconomics. To this date, most of the papers investigate this issue through the lens of the macroeconomics (Soria, Sintes, & Martin, 2014). The macro-economic data are supportive for general analysis of the effects of global economic crisis on holidays. However, there are a few limitations in using macro-economic data to model tourism consumption. For instance, the heterogeneity of consumer behavior is mostly cancelled out in aggregate data. For a case study that is focused on a specific region, applying micro-economic models to the analysis gives more advantages (Alegre & Pou, 2004). These models are more well-matched with economic consumer models (Gartner, 2009). They also include the diversity and heterogeneity of consumer behavior, as mentioned before. For these reasons, this paper has objectives to enlighten the impacts of the global recession and its effects on economic variables on the average holidays spending of Dutch population using micro-economic data, which will also be supported by econometric approaches in its estimation processes.
To discuss the importance of answering the research question above, it is necessary to determine the possible actors who would be able to use the findings of this paper to analyze and to manage similar crisis situations in the future. The tourism sector in the Netherlands, the Dutch government, the (potential) investors in the industry and the external supplier of touristic commodities are some that can be mentioned. It is generally known that economics of the tourism sector is influenced by socio-economic, demographic, geographical, and political variables. However, there is another multi-dimensional variable that has not been analyzed thoroughly as one of main determinants of tourism expenditure, namely the consumer trust in the economy1. This variable is referred as multi-dimensional because it is both stirred by and has effects on the variables above.
Many Western economies were severely hit by the outcomes of the recent crisis (Pappas, & Papatheodorou, 2017). The households’ purchasing power decreased considerably because of rising unemployment percentages. This initially caused the fall of consumers trust in economy. Moreover, lower capital returns, lower pension, rise in taxation, and savings insecurity as a result of share values’ depreciation only made the consumers even more pessimistic about the economy and their own financial situation. Therefore, it is necessary to inspect the possible changes in holidays expenditures as a result of the fall in consumers trust during and shortly after the recent global crisis. This is also important for better alignment between the tourism management and the developments in tourism demands. The findings of this case study could give more insights to strengthen value-for-money orientations, strategies of price sensitivity and psychological marketing methods. Bronner and Hoog (2011) characterized this understanding as management-oriented approach, which might give the ability to develop more effective strategies to stop or reduce the severity of the impacts of a systematic (economic) shock on business and society.
For the Dutch government and public policymakers, this case study might also be able to give direction on the make of policies to protect the tourism sector from potential negative effects of the crisis. Even also to create local touristy attractions for the potential vacationers that financially might be forced to cut back or to strategize their budget for holidays abroad. This to possibly increase the Dutch national income.
Lastly, the findings of this paper would also be important for investors in the industry and the (external) suppliers of touristic commodities such as foreign travel agencies, as it would give insights on how the Dutch population and their spending on holidays react to systematic shock comparable to this recent worldwide recession. The trust in economy and financial world plays a big role in investments decision (Pappas, & Papatheodorou, 2017). It is evident that there is a demand in deeper comprehension of consumption of discretionary goods and services such as tourism when a global systematic economic shock strikes, to estimate the holidays spending sensitivity to this shock for the sake of investment decisions.
On these grounds, the objective of this paper is to examine the effects of fall in consumer trust, due to global economic crisis, on the average holiday spending of Dutch vacationers during and after the economic downturn that started in 2008.
1 The consumers’ trust in the economy and their own financial crisis in the last 12 months and the next 12 months. For simplicity, this variable is referred as the
IV. Literature Review
Advance planning and coordination between the public and private agents are necessary to minimize the effects of the global recession on the tourism sector (Soria, Sintes & Martin, 2014). During the economic downturn that started in 2008, the effects and consequences of the crisis differed per country. However, some general effects remained: this economic downturn had massive effects on households’ individual disposable income, job security and trust in economy. According to some studies, which will be further described below, these effects made the total consumption of households decrease significantly. Yet, the impact of the recession on tourism was substantially softer when it was compared to foreign trade and industrial production (Smeral, 2010). According to Smeral (2010) the tourism demand elasticities are not symmetrical. This means that the relative fall in tourism demand during a severe economic downturn would be steeper than the relative increase in demand during economic upturn of a similar magnitude. This is why it is the objective of this case study to look on the micro-economic aspects to learn more about such developments during the last economic crisis.
Micro-economic and econometric concepts are used frequently by many studies about holidays spending during the crisis. Soria, Sintes, and Martin (2014) used an adaption of the Heckman model to analyze tourism expenditure and its adjustment to the recession in EU-27 regions. Their model used a two-step approach: firstly, they analyzed the cutback of the vacationers and secondly, they examined the how-to-cut-back decision of these vacationers. According to their study, the how-to-cutback decision was affected by socioeconomic variables such as gender, employment, age and personal income. They found also that future income security played a big role both in the probability of cutting back on holidays spending and in how-to-cutback decision. Moreover, their findings showed that households located in regions with good climate are likely to cutback. The probability of cutback varies from a median of 38% for regions with bad climate to 66% for regions with good climate. 46.3% of Dutch interviewees cut back their holidays spending (Soria, Sintes & Martin, 2104). From these interviewees, 13.63% of them cut back their spending by having reduced stay length, 18.18% by having cheaper accommodations, 27.84% by going closer to home, 21.02% by having fewer holidays, 10.79% by going travel in cheaper period and 8.52% by using cheaper transport.
The study above is relevant to understand and to anticipate how vacationers react during economic crisis and it also gives an insight that the markets in EU-27 regions decreased their demand in tourism heterogeneously. Tourism, especially the international outbound vacation, is appreciated differently by households in different regions (Soria, Sintes & Martin, 2014). This is why it is essential to use a comprehensive dataset that combines microdata and regional attributes to study country-based effects of the recession on the holidays spending of households. Alegre et al (2013) studied the effects of labor market developments on holidays spending of Spanish households. This study showed that households with unemployed member(s) display a 15 % lower tourism participation rate than households with no unemployed member(s) during crisis. Conditional on participation, the households with unemployed member(s) have a 32% lower mean holidays spending than those without unemployed member(s). It was concluded that current job tenure, assessment of the future risk of a job loss and individual possibility of income variations are the major factors that influenced these findings.
For such similar researches, Bronner and Hoog (2011, 2012) are the only ones who have done a case study on Dutch households’ behavior towards holiday spending during the recession. According to Bronner and Hoog (2011) crisis hit the Dutch economy hardest in 2009 and 2010. With this study, they created a general framework to investigate the effects of the recession on the economizing strategies of Dutch households. Two major dimensions of crisis-like events are classified (Bronner & Hoog, 2011). These dimensions influence vacationers’ expenditure. The first dimension is range (or scope). Geographical aspects of an event are reflected in this dimension, in which economizing crisis is categorized as an event with no clear limitation of space and time. The second dimension is depth. In this dimension, the effects of such an event on the disposable income and economic confidence in the future are reflected. In their other study, Bronner and Hoog (2012) also tested the effects of unemployment, loss of income, insecurity of savings, decline of pensions, depreciation of shares and difficulties in obtaining mortgages on households’ consumer spending, especially on their holiday spending. These factors are logically fit to use as explanatory variables at first sight. However, in their empirical results Bronner and Hoog (2012) also revealed that the lack of clarity about when the economy would recover affected the holidays spending even more negatively. The latter is more related to consumers having much less confidence to consume during the recession. Yet, in the literature
there is still little known about this determinant which might affect the outbound tourism spending of the Dutch households.
The tourism industry needs more information, knowledge and better understanding of consumer behavior and attitudes to travel in times of economic recession (Sheldon & Dwyer, 2010). All the studies above have mentioned the possible effects of the consumers’ trust in economic situations and their own financial circumstances as one of determinants of their holiday spending (decision), but there is not a single research focusing specifically on this matter. The crisis caused an immense shock in the economy after all and it alarmed economist, politicians, and households in general (Arup, 2010). To most people this worldwide financial crisis was unexpected and neither financial experts nor consumers had a clear understanding of the crisis, its cause, the capabilities of economic and political stakeholders to take effective measures or what the future developments are to be expected (Arup, 2010).
As mentioned earlier, the future income insecurity (Soria, Sintes & Martin, 2014) and individual possibility of income variations (Alegre et al, 2013) played major roles on decreased consumer trust in economy in general, which had led to increased probability of cutback in holidays spending in particular. Bronner and Hoog (2012) have also shown the significantly negative effects of the lack of clarity of recovery on holiday spending of Dutch population. However, in their later study, Bronner and Hoog (2014) stated that economic developments during and shortly after the crisis had relatively minor effects on holidays plans and intensions of the Dutch households, especially on the outbound vacations. This gives an impression that outbound tourism has become less of a ‘luxury’ commodity for the Dutch households, compared to other European countries, in which the tourism sectors’ developments during the crisis have been analyzed.
Therefore, the general purpose of this case study is to further investigate the effects of economic confidence of the consumers on the outbound tourism spending, focusing on the consumer trust in economy as the main determinant. The research will be using a dataset of Dutch sample population and basic linear regression analysis to test the formed hypotheses, which are based on this theoretical framework. This paper could contribute in further expansion of empirical domains, which explore vacationers’ behavior.
V. Hypotheses
As the results of analyzing the literature framework given in the previous section, two hypotheses are formed: a. According to the theoretical frameworks of this case study, households’ consumption during the crisis is
strongly influenced by the future income insecurity and the lack of clarity about crisis recovery. Outbound tourism is specified as discretionary goods and therefore it is predicted that holiday spending of Dutch households might decrease during and post-crisis periods as a result of fall in the consumer trust, caused by greater future risk of job loss and income variations. The null hypotheses are that the estimated coefficient of the variable consumer trust is zero. The alternative hypotheses specify what are true if the null hypotheses are not. These hypotheses are denoted as follow:
i. H0 = βit Consumers trust variable = 0
H1 = βit Consumers trust variable ≠ 0; a
H1 = βit Consumers trust variable > 0; b
b. The average socio-economic level of the Dutch households and the climate in the Netherlands has made outbound tourism less of luxury goods and/or services for Dutch households. Therefore, it is predicted that the decrease in outbound vacation spending of Dutch households will be relatively on the lower side. This hypothesis will be tested by analyzing the percentage of the possible decrease of the average spending as a result of the consumers trust in economy.
i. Decrease of ≤ 5% of the average holidays spending on account of a decrease of 1 unit of balanced index-number of
consumers trust = the effect is low
Decrease of > 5% of the average holidays spending on account of a decrease of 1 unit of balanced index-number of consumers trust = the effect is on the high side
To support hypothesis b. the income elasticity of the holidays spending will also be estimated. A commodity is referred as a luxury good/service if its expenditure elasticity with regards to the income is higher than 1 (Stock & Watson, 2009). This will be tested based on the following hypotheses:
ii. H0 = βit Disposable income variable > 1
H1 = βit Disposable income variable < 1
In this paper, probabilistic calculations will be made to test the null hypotheses in a way that will account for sampling uncertainty by using the data to compute the p-value of the null hypotheses. These calculations and predictions of coefficients will be tested empirically using the statistical program fixed effects regression in STATA SE/15 and a dataset drawn from Continuous Vacation Panel conducted by the Dutch Central Bureau of Statistics, which will be more elaborated in the next section.
VI. Research Methodology
i. Data
As already mentioned before, the dataset used in this paper is drawn from Continuous Vacation Panel (CVO: Continue Vakantie Onderzoek in Dutch), conducted by Dutch Central Bureau of Statistics. The CVO is a yearly survey that was started in 1992 and is currently still on going. This survey has a sample of population of approximately six thousand respondents with Dutch nationalities, representative of the Dutch population (CBS, 2017). It is designed to gather detailed information on Dutch households’ income, their holiday spending and some socio-demographic and labor-related data concerning their households’ members.
ii. Materials
For this study, the dataset for the years of 2004-2014 are used, which facilitates for a broad time-analysis. This ten year is chosen specifically to make comparison between the period prior to (2004-2007), during (2008-2010) and after the crisis (2011-2014) possible. The year 2004 is chosen as the start of the 10-year period, because the consumer trust of Dutch population had increased again in most of regions2. The incident of 9/11 had had a massive effect in Western economies as well, in terms of the confidence in investing and consuming (Makinen, 2002). The span of 2008-2010 is chosen as crisis period because the crisis started being felt in the Dutch economy from the early 2008 (Bronner & Hoog, 2011). In the same year the outbound vacation spending started decreasing (extremely) as well3. Moreover, as previously mentioned, the crisis hit the Dutch economy hardest in 2009-2010. At last, the period of crisis is determined to end in 2010 with regards to the occurrence of the Greek fiscal crisis. The level of consumer trust in the globalized economy and perceived fairness of the tax system had been (once again) negatively affected by another single-country crisis which had its massive cross-border effects as well (Kaplanoglou & Rapanos, 2012).
The average spending of the outbound vacation in this research is defined as the specific costs consisting of travel and accommodation expenses plus all other non-durable costs such as insurances, entre prices, souvenirs and so forth. This vacation spending is defined as yearly expenditures on holiday abroad of Dutch households on average per vacationer (per family member going on holidays). The average number is calculated as follow:
Σ [(ū1 + ū2 + … + ūn) / n]
where ū is the average outbound vacation spending per Dutch vacationer for provincei, with i: 1,2,..., n4.
The graph on the next page depicts the yearly averaged outbound tourism spending of the Dutch sample population.
2 Referred to the data set of ‘consumentenvertrouwen’ period 2001-2005 on CBS Statline. 3 See graph I.
Graph I. The average outbound tourism spending of Dutch households per vacationer (per family member participating in the outbound vacation)
The consumer trust indicator is taken from the CBS’s Consumers Market Survey (het CCO: Consumenten Conjuctuur Onderzoek, in Dutch). This indicator provides reflections of the confidence and conceptions of Dutch households with regards to the developments in the Dutch economy and its effects on their own financial situation (Arends & Nieuweboer, 2017). These indicators are given as balanced average of the positive and negative answers regarding the research questionnaires consisted of these five main components5:
1. The general economic situation in the Netherlands for the last 12 months.
i. the developments in unemployment percentages. ii. the developments in the prices.
2. The expected general economic situation in the Netherlands in the next 12 months.
i. the developments in unemployment percentage. ii. the developments in the prices.
3. The financial situation of the household for the last 12 months.
i. the current financial situation the household. ii. the probability of saving (more).
4. The expected financial situation of the household in the next 12 months.
i. the probability of saving (more)
ii. which area of consuming would be cut back if the income decreases with at least 10%
5. Whether or not the current period is a favorable time for large purchases.
Some of the explicit questionnaires of this market survey could be found in table I. Respondents could indicate their answers in the questionnaires by choosing a number ranging from -100 (extremely worse) to +100 (extremely better). 0 stands for situation(s) remaining the same. The yearly average consumer trust in the Netherlands is illustrated in graph II, followed by the average consumers trust per province in graph III.
Besides the average consumer trust of Dutch households, this paper also takes into account the unemployment percentages and the disposable income per province. These variables are expected to be affected by the crisis and to influence the average holiday spending of the Dutch households as well. They will be used as control variables in the econometric models in this paper, which are more elaborated in the next section. Furthermore, the income elasticity of the outbound vacation spending will also be analyzed. Besides its importance to test hypothesis b. i., it is also analyzed because the tourism demand elasticities are found to be not symmetrical in earlier study.6
The correlations of these variables with the consumers trust will be estimated to see if inclusion of these variables into the regression models would make it more possible to avoid omitted variable bias. These are displayed in table II.
5 See ‘CCO 2017: Uitgebreide onderzoeksbeschrijving van Consumenten Conjuctuur Onderzoek’ in the references section. 6 Smeral, 2010. Described earlier in the literature framework section.
Table I. Some of the questionnaires out of CCO 2017
Subjects
Questions
The general economic situation in the Netherlands for the last 12 months
For the last 12 months, has the economy become better, worse or remained the same, in your opinion?
Follow up if answer is better/worse:
With this score, do you mean that it has become a little better/ worse or significantly better/ worse?
The general economic situation in the Netherlands for the next 12 months
In the next 12 months, do you think that the economy will get better, worse or will it remain relatively the same?
(‘I do not know’ is also a choice)
Follow up if answer is better/worse:
With this score, do you mean that it will get a little better/ worse or significantly better/ worse?
The prices in the last 12 months
Do you think that the prices have gone up/ down/ remained relatively the same?
Follow up if the answer is that prices have gone up:
With this score, do you mean that the prices have gone up weakly, reasonably of significantly high?
The prices in the next 12 months
Do you expect that the prices will go up/ down/ will it remain relatively the same in the next twelve months?
Follow up if the answer is that prices are expected to go up:
With this score, do you mean that the prices will increase weakly, reasonably or significantly?
The unemployment in the next 12 months
Do you expect the unemployment percentages to go up/ go down/ remain relatively the same in the next 12 months?
Follow up if the answer is go up/ go down:
With this score, do you mean that it will go up/ go down weakly, reasonably, or significantly?
Financial situation of the households in the last 12 months
Has your financial situation gotten better, worse or has it remained the same the last 12 months?
Follow up question if the answer is better/worse:
With this score, do you mean to indicate that it has gone a little better/worse? Or that it has gone significantly better/worse?
Current financial situation of the households
How is your financial situation currently? Is your total income higher than your total spending? Or is it the other way around, that you are in short of money? Or can you get by just enough?
Follow up question if the answer is short of money:
Does it mean that you would have to have debt? Or does it mean that you have to use your savings?
Follow up question if the answer is that the total income is higher than the total spending:
Is it a little higher? Or is it significantly higher?
Financial situation of the households in the last 12 months
How do you expect your financial situation will be in the next 12 months? Would it get better/ worse/ or will it remain relatively the same?
Follow up question if the answer is better/worse:
With this score, do you mean to indicate that it will get a little better/worse or significantly better/worse?
Big expenses in the 12 months
Considering big purchases, such as furniture, wash machine, a television and other durable products. Do you expect that, in your households, more/ less/ relative the same will be spent for these products in the next 12 months?
Follow up question if the answer is more/less:
With this score, do you mean to indicate a little more/less or significantly more/less?
Money put aside
Do you think you would be able to put aside some money in the next 12 months?
Follow up question if the answer is yes/no:
With this score, do you mean that you think that you are/ are not too sure about it? Or do you think that you most probably might/ might not be able to put aside some money?
Cut-back
If your income is about to decrease by 10% for instance, in which area would you cutback your spending?
*The spontaneous answers were coded. However not all are coded, only the ones that were mostly mentioned by interviewee. Vacation is one of the most mentioned area.
**The follow-up questions are tools to weigh the total scores to be more fairly and balanced.
Graph II. The Dutch consumers’ trust in the economy and their finances according to the CCO.
iii. Econometric modelling
The effects of the fall of the Dutch consumer trust in economy and their own financial situations and all other control variables (mentioned above) on the average outbound tourism spending per Dutch vacationer will be analyzed with the fixed effects regression models. This model is an estimation method that could control for omitted variables in panel data when the omitted variables vary across entities but do or might not change over time (Stock & Watson, 2012). In this paper, we use provinces as the entities. The general concept of the fixed effects regression is shown below.
Y = β0 + β1 X + β2 Z+ u,
where X is a variable that changes per province over time and Z is a variable unobserved that varies from one province to the next but is time-invariant. With the data from Continuous Vacation Panel more possible control variables could be added to avoid omitted variable bias and hence this model could be extended as following:
Yit = β0 + β1 Xit + ….. + βk Xk,it + βk-1 Zi + uit
with i = 1,2,..., N; t = 1, 2,..., T.
Since Zi differ per province and are assumed to be relatively constant over time, then the following could be
specified:
αi = β0 + βk-1 Zi,
in which Zi could be anything from cultural attitudes towards vacation: average housing prices and expenses that might affect households’ disposable income, vacation destination trends, average income, and so forth. Furthermore, αi,....,αn are considered as unknown intercepts that are to be estimated, one for each province. This interpretation of province-specific intercepts comes from considering that in the (multiple) regression line, β1, is the same for all
provinces, yet the intercept of the regression line varies for every province (Stock & Watson, 2012). Hence, the intercepts can be viewed as the effect of being in province i. The regression model is therefore defined as below:
Yit = β1 Xit + ….. + βk Xk,it + uit
with i = 1,2,..., N; t = 1, 2,..., T. 7
As mentioned earlier, the goal of this study is to investigate the effects of fall in consumer trust of Dutch households on the average outbound tourism spending during and shortly after the crisis. The consumer trust is also correlated on some variables that are affected directly by the recession. Unemployment percentages started to arise in 2008 and since this could affect the future income insecurity and individual possibility of income variation, this variable will be added to the regressions as another control variable. The disposable incomes and income elasticities of holidays spending will also be found in the equations, since these variables are expected to (negatively) changed during and perhaps also after the recession.
7
Model i
The average outbound tourism spending per vacationer*8
= β1 Consumers trust + β2 Dummy for period of crisis + β3 Dummy for period after crisis + αi + uit
with i = 1,2,..., 12; t = 2004, 2005, …, 2014; Dummy for period of crisis = 1 | year = 2008,2009,2010 and 0 otherwise; and Dummy for period after crisis = 1 | 2011, 2012, 2013 or 0 otherwise
Model ii
The average outbound tourism spending per vacationer*9
= β1 Consumers trust + β2 Consumers trust * Dummy of period of crisis+ β3 Consumer trust * Dummy of period after crisis + β4 Dummy for period of crisis +
β5 Dummy for period after crisis + αi + uit
with i = 1,2,..., 12; t = 2004, 2005, …, 2014; Dummy for period of crisis = 1 | year = 2008,2009,2010 and 0 otherwise; and Dummy for period after crisis = 1 | 2011, 2012, 2013 or 0 otherwise
Correlation between the X’s.
sxi,yi = (1/(n-1)) ∑ (X – Xaverage) (Y – Yaverage)
with i: the average consumer trust, the unemployment percentage, the average disposable income, the income elasticity of the outbound vacation spending, and all the possible interaction term variables.
The income elasticities of the outbound vacation spending
The average spending of outbound tourism per Dutch vacationer = = β1 Disposable incomeit + αi + uit
with i = 1,2,..., 12; t = 2005, 2006, 2007 Average holidays abroad spending* = β1 Disposable incomeit + αi + uit
with i = 1,2,..., 12; t = 2008, 2009, 2010 Average holidays abroad spending* = β1 Disposable incomeit + αi + uit
with i = 1,2,..., 12; t = 2011, 2012, 2013
Model iii
The average outbound tourism spending per vacationer*10
= β1 Consumers trustit + β2 Unemployment percentagesi t+ β3 Disposable incomeit + β4 Consumers trustit * Unemployment percentageit + β5 Consumers trustit *
Disposable incomeit + β6 Unemployment percentageit * Dummy for period of crisis + β7 Disposable incomeit * Dummy for period of crisis+ β8 Unemployment
percentageit * Dummy for period after crisis + β9 Disposable incomeit * Dummy for period after crisis + β10 Consumers trustit * Dummy for period of crisis + β11
Consumers trustit * Dummy for period after crisis + β12 Dummy for period of crisis + β13 Dummy for period after crisis + αi + uit
with i = 1,2,..., 12; t = 2004, 2005, …, 2014; Dummy for period of crisis = 1 | year = 2008,2009,2010 and 0 otherwise; and Dummy for period after crisis = 1 | 2011, 2012, 2013 or 0 otherwise
8 Predicted dependent variable. See models’ limitation section. 9 Predicted dependent variable. See models’ limitation section. 10 Predicted dependent variable. See models’ limitation section.
Model iv
The average outbound tourism spending per vacationer*11
= β1 Consumers trustit + β2 Dummy variable of the income elasticity during the crisis + β3 Dummy variable of the income elasticity after the crisis + αi + uit
with i = 1,2,..., 12; t = 2005, 2006, …, 2013; Dummy variable of the income elasticity during the crisis = -22.59 | year = 2008, 2009, 2010 and 0 otherwise; Dummy variable of the income elasticity after the crisis = 0.97 | year = 2011, 2012, 2013 and 0 otherwise.
Model v
The average outbound tourism spending per vacationer*12
= β1 Consumers trustit + β2 Unemployment percentagesi t + β3 Consumers trustit * Unemployment percentageit + β4 Unemployment
percentageit * Dummy for period of crisis + β5 Unemployment percentageit * Dummy for period after crisis + β6 Consumers trustit *
Dummy variable of the income elasticity during the crisis + β7 Consumers trustit * Dummy variable of the income elasticity after the crisis
+ β8 Dummy for period of crisis + β8 Dummy for period after crisis + αi + uit
with i = 1,2,..., 12; t = 2005, 2006, …, 2013; Dummy for period of crisis = 1 | year = 2008,2009,2010 and 0 otherwise; and Dummy for period after crisis = 1 | 2011, 2012, 2013 or 0 otherwise; Dummy variable of the income elasticity during the crisis = -22.59 | year = 2008, 2009, 2010 and 0 otherwise; Dummy variable of the income elasticity after the crisis = 0.97 | year = 2011, 2012, 2013 and 0 otherwise.
Model vi
The average outbound tourism spending per vacationer*13
= β1 Consumers trustit + β2 Unemployment percentagesi t + β3Unemployment percentageit * Dummy for period of crisis + β4
Unemployment percentageit * Dummy for period after crisis + β5 Consumers trustit * Dummy variable of the income elasticity during the
crisis + β6 Consumers trustit * Dummy variable of the income elasticity after the crisis + β7 Dummy for period of crisis + β8 Dummy for
period after crisis + αi + uit
with i = 1,2,..., 12; t = 2005, 2006, …, 2013; Dummy for period of crisis = 1 | year = 2008,2009,2010 and 0 otherwise; and Dummy for period after crisis = 1 | 2011, 2012, 2013 or 0 otherwise; Dummy variable of the income elasticity during the crisis = -22.59 | year = 2008, 2009, 2010 and 0 otherwise; Dummy variable of the income elasticity after the crisis = 0.97 | year = 2011, 2012, 2013 and 0 otherwise.
11 Predicted dependent variable. See models’ limitation section. 12 Predicted dependent variable. See models’ limitation section. 13 Predicted dependent variable. See models’ limitation section.
iv. Models limitation
In the fixed effects model, the provincial-specific effect is a random variable that is allowed to be correlated with the explanatory. This combined with the assumption that the time-varying explanatory variables have non-zero within-variance (are not perfectly collinear), makes it impossible to include a constant (αi) or any time-invariant variables (Zi)
in the model. This means that the model only estimates the parameters β1,..,k,it. The fixed effects estimation subtracts
the time averages from the initial model. The time averages are calculated as follow: Yaverage = 1/T Σt Yit
Subtracting this from the initial model as given earlier, Yit = β1 Xit + ….. + βk Xk,it + uit
with i = 1,2,..., 12; t = 2004, 2005, …, 2014. 14
yields the so-called within model (or in this paper called the predicted model Y*): Y* = β1 X*it + ….. + βk X*k,it + uit
with i = 1,2,..., 12; t = 2004, 2005, …, 2014. 15
Since this model estimates only the within-effects, there is no access to the effects of variables that do not vary (much) within the provinces. The average housing prices or rent expenses are examples that might be useful to consider in the analysis of the average holiday spending. However, these variables are ones of these so-called intragroup variables, which must be foregone when the fixed effects estimation is used.
Another limitation of the fixed effects models is that they do not completely eliminate the threat of omitted variable bias by controlling for all time-invariant differences in both observed and unobserved variables. If the unobserved variables vary over time and within each province (e.g., rent expenses) and these changes are also correlated with the changes in the (predicted) average holidays spending, then the regression will still suffer from omitted variable bias.
At last, the fixed effects regression has four following assumptions: 1. The results from the E(u )|Xi1 , Xi2 ,…, XiT , αi) = 0
2. (Xi1 , Xi2 ,…, XiT, ui1, ui2,…, uiT), I = 1,…., n are identically and independently drawn from their joint distribution 3. (Xit, uit) have nonzero finite fourth moments
4. There is no perfect multicollinearity.
These assumptions are complicated to tested when the dataset used is drawn from final results of a panel research. The outcomes of the CVO are put on Statline of the CBS website and spread unto specific and fitting themes, which means that the initial data collection is not easily obtainable. For this reason, the assumptions above are difficult to be tested. Therefore, heteroscedasticity-consistent estimators are used in this model. These so-called robust estimators get less biased as the time-series dimension increases (Kezdi, 2004).
14
Note that αi is excluded from the equation now because it is subsumed under the error components.
15
VII. Empirical Results
i. Overview in tables
Table II. Correlations between the determinants of the average holiday spending of Dutch vacationers. Consumer
trust Unemployment percentage Disposable Income Income elasticity of
the holiday spending before the crisis Income elasticity of the holiday spending during the crisis Income elasticity of the holiday spending after the crisis Consumer trust * Income elasticity of the holiday spending before the crisis Consumer trust * Income elasticity of the holiday spending during the crisis Consumer trust * Income elasticity of the holiday spending during the crisis Consumer trust 1.0000 Unemployment percentage -0.2976 1.000 Disposable Income 0.0841 -0.0972 1.0000 Income elasticity of the holiday spending before the crisis 0.3990 -0.0911 -.5783 1.0000 Income elasticity of the holiday spending during the crisis 0.0348 0.4900 -0.1739 -0.2391 1.0000 Income elasticity of the holiday spending after the crisis -0.3668 0.5448 0.4173 0.2951 0.4629 1.0000 Consumer trust * Income elasticity of the holiday spending before the crisis 0.3855 -0.2209 0.5810 -0.5165 -0.2391 0.2951 1.0000 Consumer trust * Income elasticity of the holiday spending during the crisis -0.080 -0.5072 0.1465 -0.4437 -0.9686 0.4437 0.2292 1.0000 Consumer trust * Income elasticity of the holiday spending after the crisis 0.5963 -0.4368 -0.2707 0.4891 -0.3962 -0.8560 -0.2526 0.3798 1.000
Table III. The income elasticity of the average outbound vacation spending per vacationer
Notes:
The interpretation of the coefficients:
1. An increase of the yearly disposable income by 1000 euros would increase the average outbound vacation spending per vacationer by 9.86 euros in the period before the crisis.
2. An increase of the yearly disposable income by 1000 euros would decrease the average outbound vacation spending per vacationer by 22.59 euros during the crisis.
3. An increase of the yearly disposable income by 1000 euros would increase the average outbound vacation spending per vacationer by 0.971 euro shortly after the crisis.
Table IV. The averages of outbound vacation spending per vacationer per period.
Graph V. The average disposable income (x 1000 euros) per province.
Before the crisis
(2005-2007) During the crisis (2008-2010) After the crisis (2011-2013)
Disposable income 9.852**1 -22.588** 2 0.971*** 3
Before the crisis
(2005-2007) During the crisis (2008-2010) After the crisis (2011-2013) The average spending on outbound vacation
Table V. The results of fixed effect regression models with the predicted average of outbound vacation spending per Dutch vacationer as the dependent variables
Models Independent variables i ii iii iv v° vi° Consumer trust 0.6377*** 1.1458*** -2.1600** 0.6377*** 0.3180 1.2970*** Unemployment percentage 7.2270** 5.9176*** 3.2855** Disposable income (per 1000 euros) 6.3610 Consumer trust * Unemployment percentage 0.2956*** 0.1960*** Consumer trust * Disposable income 0.0485**
Consumer trust * Dummy
of crisis period -1.9493*** -0.9877***
Consumer trust * Dummy of after crisis period
-1.1622*** -1.8782***
Unemployment percentage *
Dummy of crisis period
-9.8571*** -7.4854*** -8.1910***
Unemployment percentage *
Dummy of after crisis period
6.0692** 7.1776*** 5.5328
Disposable income (per 1000
euros) * Dummy of crisis
period
-1.4551*
Disposable income (per 1000
euros) * Dummy of after
crisis period
0.6795
Income elasticity of the average holiday spending * dummy of crisis period
-2.8520***
Income elasticity of the average holiday spending * dummy after crisis period
Notes:
1. Std. Err.
Clustered robust estimators are calculated in these models.
2. p-values
Testing: H0 = βit Consumers trust variable = 0
H1 = βit Consumers trust variable ≠ 0
*
The results are statistically significant because the confidence interval of 90% does not contain the null hypothesis values. **
The results are statistically significant because the confidence interval of 95% does not contain the null hypothesis values. ***
The results are statistically significant because the confidence interval of 99% does not contain the null hypothesis values.
3. °
For these regressions, the dataset from 2005-2013 are used. This is because the income elasticity of the average outbound vacation spending is calculated for period 2005-2007, 2008-2010, and 2011-2013, so that every period has the same amount of years.
Consumer trust * (Income elasticity of the average holiday spending * dummy of crisis period)
0.0726*** 0.0760***
Consumer trust * (Income elasticity of the average holiday spending *
dummy after crisis period)
-1.8125*** -1.4617***
Dummy of crisis period 64.4272*** 32.1027*** 123.2383*** 72.3920*** 76.7281***
Dummy of after crisis
ii. Analysis
The results of the fixed effects estimations proposed in section VI. iii are shown in table V. The estimated effects of the independent variables on the predicted average outbound tourism per vacationer are given by the coefficients in the models. It should be noted that in any model estimations with dummy variables, the omitted alternative should be taken into account in the analysis of the predicted results. This means that every coefficient has to be interpreted in relation to the base category. Regarding to every model used in this research, the base category is the period before the crisis (2004-2007).
As mentioned earlier, in these linear regression models, the consumer trust of the Dutch households is the key determinant in the models’ equations. Table V.i shows that the consumer trust has a positive effect on the predicted average outbound tourism spending per vacationer. The results of model i is given in this table. A 1-unit rise of balanced average index-number of the consumer trust would increase the predicted average spending by 0.64 euro in the period before the crisis. During the crisis, this average spending would increase by 64.43 euros more compared to the increase during the period before the crisis. For the period after the crisis, it is 87 euros more when it is compared to the period before the crisis. Note that it is assumed here that all other possible factors of the average spending are kept constant.
In the reality, the effect on the dependent variable of a change in one independent variable could depend on the value of another possible independent variable (Stock & Watson, 2012). Interaction terms are created to estimate this effect. On that ground, model ii is proposed, in which the consumer trust is being interacted with the dummy variables of period during and after the crisis. The estimated coefficients in table V. ii show that before the crisis the effect of a 1-unit increase of balanced average index-number of the consumer trust is predicted to be an increase of 1.15 euros in the average outbound vacation spending per vacationer. During the crisis, the effect is predicted to be 31.30 euros (32.10 + 1.15 – 1.95). Note that, according to this model, the predicted effect of the consumer trust after the crisis is even larger than during the crisis itself, amounting to an estimated effect of 65.67 euros (65.68 + 1.15 – 1.16).
The unemployment percentages and the disposable income are the other observed determinants of the average outbound tourism spending per vacationer, which are correlated with the consumers trust and which also change over time. The correlation between these variables are depicted in table II. These variables should also be included in the regression models to avoid omitted variable bias (Stock & Watson, 2012). Table V. iii shows the estimated results of model iii. This table displays that the dummy variable for the period after the recession do not have statistically significant effect on the average spending per vacationer. This would make the comparison of the effects before, during and after the crisis difficult to obtain. Note that in this model the effect of the dependent variables become extremely higher during the crisis. It is 123.24 euros more than the effect before the crisis. This difference is also higher than the effect after the crisis, compared to the effect before the crisis (even though the effect of dummy variable for the period after the crisis is insignificant). Note that in model i and ii, when it is compared to the effect before the crisis, the difference in the estimated effect in the period after the crisis has always been higher than the estimated effect during the crisis (model i: +87.00 euros > +64.43euros; model ii: +65.68 euros > + 32.10 euros).
Furthermore, it is seen in table V. iii that the disposable income does not have a significant effect on the predicted average outbound vacation spending per Dutch vacationer. This can be caused by the fact that the average disposable income relatively did not decrease as a result of the crisis, see graph V. The Dutch social security system might be one of the reasons of this non-decreasing disposable income of Dutch household during the crisis.
Another variable related to the disposable income could be created with the available dataset in the CVO, namely the income elasticity of the average outbound vacation spending. Table III depicts the effects of the disposable income on the average spending of outbound tourism per vacationer in the period before, during and after the crisis. These estimated coefficients give insights about the income elasticity of the average spending. The disposable income has a negative effect on the average spending during the crisis, with a notable decrease of spending by 22.59 euros by an increase of 1000 euros of the yearly average income, holding all other possible variables constant. In contrast, the effects are positive before and after the crisis. Although the effect after the crisis of 0.97 euro could not come any
near the positive effect of 9.58 euros before the period of crisis. With these estimations, it is shown that the income elasticity of the average outbound tourism per vacationer becomes larger during the systematic shock of the global crisis. At the first sight, the significant decrease of 22.59 might seem like a very small effect on the averaged outbound tourism spending of 670 euros per vacationer. It is only a 3.37 % decrease of the spending. However, an increase of 1000 euros is equal to an increase of 3.06 % in the yearly averaged disposable income in period 2008-2010. This means that an increase of 1% of the yearly averaged disposable income of the Dutch households would decrease the averaged outbound tourism spending per vacationer by 1.10%. The outbound tourism had, in theory, become a luxury commodity during the crisis.
This insight forms a good reason to calculate the correlation between these elasticities and the consumer trust, to see if these elasticities could be valuable to be added into the regression models. The correlations are found in table II, which displays reasonably correlations between the consumer trust and the elasticities before and after the crisis. Therefore, to make sure that the possible effects of the disposable income-related variable are still taken into account, the interaction terms of the consumers trust and the income elasticity of the outbound tourism spending are created. Note here that the income elasticity is taken as a dummy period-variable instead of a yearly elasticity dummy variable (βi,period income elasticity on the average spending ≠ βi,t income elasticity on the average spending). Since the elasticity
variables are continuous period-variables (d1 = -22.59 | d1 = 1 if year = 2008, 2009, 2010 and d1 = 0 | d1 = 0 if year ≠ 2008, 2009, 2010), the interaction terms can be again interacted with the dummy variables of the crisis and after-crisis period. These interaction terms replace the interaction terms of the consumers trust and the dummy variables of the crisis and after-crisis period. The estimated results of these terms might be complicated to interpret, but this would give a valuable prediction of the effects of the consumer trust on the average spending of vacation abroad.
Table V. vi shows the estimated effects of these interaction terms. Apart from the interacted term of the unemployment percentage and the dummy variable of period after the crisis, all variables have statistically significant effects on the average spending of outbound tourism per vacationer. The effect of the dummy variable of period during the crisis is once again greater than the effect of the dummy variable of period after the crisis, having the effect of the dummy variable before the crisis as the base category. This confirms that, when the model is also controlled by other relevant determinants (and the model becomes more fit for the regression), the effects of the consumers trust on the average spending is greater during the crisis than after the crisis, having the period before the crisis as the base category. Moreover, the consumer trust has a positive effect on the average spending during the crisis. This means that, holding all other variables constant, every 1-unit increase/decrease of the balanced average index-number of the consumer trust would increase/reduce the average spending by 76.22 euros {1.30 + [0.08 (1* -22.59)] + 76.73}16, 74.92 euros [0.08 (1*- 22.58) + 76.73] more/less spent compared to 1-unit decrease of the balanced average index-number of consumers trust before the crisis. At last, holding all other variables constant, the effects of every 1-unit increase/decrease in the balanced average of index-number of the consumer trust after the crisis would increase/reduce the average spending by 22.57 euros, and thus 21.27 euros more/less spent, compared to the average expenditure before the crisis. This means that a decrease of every 1-unit of the balanced index-number of the Dutch consumer trust would reduce the average spending of outbound tourism per vacationer by approximately 11.38% during the crisis and by 3.28% after the crisis.
At last, it should also be noted that in comparison to each other, the average spending per vacationer in a particular province may be more or less sensitive to the confidence shock caused by the global crisis, that started 2008. This is because every province has its own specific αi, which is absorbed in uit and not possible to obtain when
the fixed effects models are used. The graph below depicts the average holidays spending abroad per vacationer for every province.
16Y
Graph VI. The yearly averaged outbound tourism spending per vacationer per province in the Netherlands.
VIII. Conclusion
This case study investigated the effects of the consumer trust on the average outbound tourism spending per vacationer. Taking the estimation results in the previous section into consideration, it can be concluded that the Dutch consumers trust on the economy and their own finances does have significant effects on this average spending. Thus, the null hypothesis in a.i. is rejected based on the estimated results in table V.vi. With a confidence interval of 95%, it can be concluded that the outbound tourism spending per Dutch vacationer in period 2008-2014 could decrease due to fall in consumer trust, which was caused by the global economic downturn.
Moreover, the estimated results also confirmed that before the crisis started, the outbound tourism was not a luxury commodity for the Dutch households, as it was statistically proven that the expenditure elasticity with regards to the income is lower than 1. However, this elasticity changed to 1.1 during the crisis, which made the outbound tourism, in theory, a luxury commodity after all. After the crisis, this elasticity became lower than 1 again. Smeral (2010) stated that the tourism demand elasticities are not symmetrical, which means that the relative fall in tourism demand during a severe economic downturn would be steeper than the relative increase in demand during the economic upturn of a similar magnitude. The results in this paper could test this statement with respect to the spending elasticities. With the results above, it can be concluded that the spending elasticities are not symmetrical either. The relative fall in tourism demand during the last global recession was much steeper than the relative increase in spending during the economic upturn in period before 2007. For the Dutch households, the outbound tourism even became a luxury commodity, while it normally is not one in the usual economic conditions. During the crisis, the null hypothesis b. ii. cannot be rejected with a confidence interval of 95%.
Additionally, it is shown above that the decrease in outbound vacation spending of the Dutch households, calculated per vacationer, was on the high side, up to 11.38% during the last global crisis per a decrease of 1-unit balanced average index-number of the consumer trust. After 2010, the average spending of the outbound tourism per Dutch vacationer started going up again, despite the rising of the European debt-crisis. The consumer trust, on the other hand, did fall again due to this debt-crisis, caused by Greek government’s financial instability. As shown in table V.vi the decrease in the average spending of outbound tourism per vacationer after the global crisis was still greater than before the crisis, as a result of a decrease per 1-unit balanced average index-number of the consumer trust, holding all the other variables constant. This means that after the global economic shock, the once again fallen consumer trust of the Dutch households had less intense effect on the average spending of the outbound tourism per vacationer, seen in the increasing average spending in the periods after 2010 (see graph I). This effect should be further researched in the future.
Thus, it can be concluded that the Dutch consumers’ confidence in the economy and their own financial situation is a statistically significant determinant of the average spending of the outbound tourism. Also, compared to the periods before and after the crisis, this consumer trust has significant greater effects on the average spending of the outbound tourism of Dutch households during the crisis. Based on the literature framework, it could be caused by greater future risk of job loss and income variation which strongly influenced the consumer trust in economy of the Dutch households.
IX. References
Alegre, J., & Pou, L. (2014). Micro-economic determinants of the probability of tourism consumption. Tourism Economics, 10(2), pp. 124-144.
Alegre, J., Mateo, S., & Pou, L. (2013). Tourism participation and expenditure by Spanish households: the effects of the economic crisis and unemployment. Tourism Management, 39, 37e49.
Arends, J., & Nieuweboer J. (2017) Uitgebreide onderzoeksbeschrijving van het CCO (CCO2017). Heerlen: CBS. Arup, C. (2010). The Global Financial Crisis: Learning from Regulatory and Governance Studies.
Law &Policy, 32(3), pp. 363-381.
Bronner, F., & Hoog, R. (2011). Economizing strategies during an economic crisis. Annals of Tourism Research, 39(2), pp. 1048-1069.
Bronner, F., & Hoog, R. (2014). Vacationers and the economic “double dip” in Europe. Tourism Management,
40, pp. 330-337.
CBS (2017). Continu Vakantie Onderzoek. Retrieved from
https://www.cbs.nl/nl-nl/onze-diensten/methoden/onderzoeksomschrijvingen/korte-onderzoeksbeschrijvingen/continu-vakantie-onderzoek--cvo--
Gartner, M. (2013). Macroeconomics. London: Pearson Education
Makinen, G. (2002). The Economic Effects of 9/11: A Retrospective Assesment. (RL31617). Washington DC: The Library of Congress.
Kaplanoglou, G., & Rapanos, V. (2012). Tax and Trust: The Fiscal Crisis in Greece. Journal of South European Society and Politics, 3, pp. 283-304.
Papatheodorou, A., & Pappas, N. (2017). Economic Recession, Job Vulnerability, and Tourism Decision Making: A Qualitative Comparative Analysis. Journal of Travel Research, 56(5), pp. 663-677. Sheldon, P., & Dwyer, L. (2010). The global financial crisis and tourism: perspectives of the academy.
Journal of Travel Research, 49(1), 3e4.
Smeral, E. (2009). The impact of the financial and economic crisis on European tourism. Journal of Travel Research, 48(1), pp. 3-13.
Smeral, E. (2010). Impacts of the World recession and economic crisis on tourism: forecasts and potential risks. Journal of Travel Research, 49(1), pp. 31-38.
Soria, J., Sintes, F. & Martin, J. (2014). Understanding tourists’ economizing strategies during the global economic crisis. Tourism Management, 2015, 48, pp. 164-173.
