The Impact of Holiday Trips on Life Satisfaction and Domains of Life Satisfaction

Data

To carry out this study we use five waves (1990, 1995, 1998, 2003, and 2008) taken from the German Socio-Economic Panel (GSOEP), a representative longitudinal data set of persons, households, and families in Germany. This survey covers a broad range of topics, including information at an individual and household level on living conditions, household composition, employment status, income, health, education, and satisfaction measures among others. However, only for those five waves does the GSOEP questionnaire include a set of questions concerning free time, which reveals how often individuals take part in some leisure activities, such as going out for a drink or a meal, visits to and from family members and relatives, attending religious events, watching TV/video, listening to the radio, and “go(ing) on a trip or short holidays.” In our case, we focus on the answers obtained for this latter activity, which we call “holiday trips” in this study. ² The possible responses to this question are the following: Daily, at least once a week, at least once a month, seldom, and never. Because of the low number of disabled individuals for the first response “daily” (only 98 individuals, of whom 12 are disabled), we have combined the first two responses into one, that is, “daily/weekly.”

Life satisfaction can be defined as the degree to which an individual favorably judges the overall quality of his life-as-a-whole (Veenhoven 1991). Most psychologists consider that the best way to study life satisfaction and/or other domains of satisfaction is by asking individuals how they feel. In our case, all satisfaction measures have been reported by the individuals themselves. For example, satisfaction with life is measured by participant responses to the following question and asked in each wave of the GSOEP: “How satisfied are you with life, all things considered?” This question is available in all waves and has been previously used, for example, by Lucas (2007), Oswald and Powdthavee (2008), and Pagan (2012a). The respondents of the GSOEP are also asked to report how satisfied they are at that moment with their health, job, housework, household income, place of dwelling, and free time. ³ The possible responses to all these satisfaction questions range from 0 (completely dissatisfied) to 10 (completely satisfied). Veenhoven (1996) points out that this single-item measure is generally as reliable and valid as multiitem measures. In addition, psychologists have usually considered these answers to satisfaction questions as cardinal (e.g., Schwartz 1995; Ng 1997), whereas some economists have assumed the ordinality of the answers (e.g., Van Praag 1991; Ferrer-i-Carbonell and Frijters 2004). Nevertheless, Ferrer-i-Carbonell and Frijters (2004) conclude that assuming cardinality or ordinality has little impact on estimation results.

As for our disability measure, this is based on whether the person reports a work-limiting health condition (e.g., Burkhauser and Daly 1998; Burchardt 2000; Hotchkiss 2004; Jenkins and Riggs 2004; Burkhauser, Houtenville, and Rovba 2006). Looking at the GSOEP questionnaires, we find that the work limitation included in the data set has been changed over time and it has not been asked in all its waves (from 1988 to 1991 and in 1993 and 1994, or at all from 2002 onwards). To bridge these gaps in the data, Burkhauser and Schroeder (2007), using the existence of this work limitation question as a benchmark when available, have created a work limitation–based measure of disability for all GSOEP years that is comparable to that work limitation question included in the Current Population Survey (CPS) of the United States. To do this, they combine two alternative survey questions in the GSOEP that have been asked each year. The first part of this combination definition is obtained from the current health satisfaction question, which ranges from 0 (completely dissatisfied) to 10 (completely satisfied) and is highly correlated with the self-perceived work limitation (Burkhauser and Schroeder 2007). The second part comes from the following question: “Are you officially registered as having a reduced capacity to work or as being severely disabled? If Yes, what is the degree of your disability?” The degree of disability ranges from 0 (not disabled) to 100 (severely disabled). They found that the best concordance of the combination definition with the severe limitation question is a combination of individuals with a health satisfaction level of at most 2 or a degree of disability of at least 53%. That is, a person is considered disabled if he/she has a health satisfaction level of at most 2 or a degree of disability of at least 53%. Namely, those individuals who fulfil at least one of these two conditions are considered disabled. ⁴ According to Burkhauser and Schroeder (2007), if we only use the degree of disability measure to identify the disabled population, we would miss individuals with a work limitation who were in the process of becoming registered that had short-term limitations or decided it was not worth registering.

Our sample consists of individuals aged 16 or over. Only those who answered the domains satisfaction questions, including the overall life satisfaction, are included in the analysis. It is worth noting that only a part of the disabled sample (around 21%) reports some values for satisfaction with job. For this reason, the final sample is not conditioned on whether the individual (disabled or not) also reports a level for job satisfaction. After dropping those individuals with missing information, the final samples used in the estimation process for people without and with disabilities are 39,486 (22,189 individuals) and 3,940 (2,855 individuals) observations, respectively. For the estimation of the job satisfaction equations, we have 21,964 observations (13,830 individuals) for the nondisabled, and 824 observations (727 individuals) for the disabled.

Method

Although subjective well-being can be analyzed through either a top–down or bottom–up approach (e.g., Diener 1984; Headey Holmstrom, and Wearing 1985; Lance et al. 1989; Sirgy 2002; Sirgy and Lee 2006; and Sirgy et al. 2011), we use a bottom–up approach wherein domains of satisfaction determine (and are the component of) satisfaction with life as a whole. According to Van Praag, Frijters, and Ferrer-i-Carbonell (2003), this yields the so-called two-layer satisfaction model. In particular, this bottom–up approach has been previously proposed and used by Sirgy et al. (2011) in the analysis of the effect of travel on tourists’ life satisfaction and domains satisfaction. These authors postulate that affect within a specific life domain accumulates and vertically spills over to superordinate domains (e.g., life in general), and thus we can make a case for the effects of satisfaction with certain tourist-related events (e.g., holiday trips) on satisfaction of various life domains and on life overall.

Following the method proposed by Van Praag, Frijters, and Ferrer-i-Carbonell (2003) and similar to Sirgy et al. (2011), we estimate a two-layer satisfaction model wherein overall life satisfaction (LS) is a function of satisfaction levels regarding various domains or facets of life (DS), such as satisfaction with health, job, housework, household income, place of dwelling, and free time (Figure 1). Satisfaction with life as a whole may be seen as an aggregate measure or as a weighted average of domain satisfaction, where the most important domains are given the most weight. The main advantage of this method, which these authors call the “Probit Adapted OLS (POLS),” is the possibility to measure the influence of some characteristic variables (e.g., gender, age, educational level, and “participation in holiday trips”) on LS through different domains of life satisfaction. This allows us to obtain a more detailed and more accurate description when we concentrate on the intermediate block in Figure 1 than when we try to cover the two stages in only one equation (Van Praag, Frijters, and Ferrer-i-Carbonell 2003).

OPEN IN VIEWER

Figure 1.

The two-layer model proposed by van Praag, Frijters, and Ferrer-i-Carbonell (2003).

POLS represents an alternative approach to traditional ordered response models and is based on the transformation of the data that allow discrete-choice variables as if they are distributed along the whole real line. One advantage of this method compared to ordered probit (OP) is that when estimating panel data, the random fixed effect error structure is much more easily implemented than with a similar model in the OP framework (and also substantially reduces computational time). In addition, POLS estimates (both coefficients and standard errors) are similar to OP estimates (except for a multiplicative factor) and it is especially useful in the presence of ordinal variables among regressors (see Terza 1987; Van Praag and Ferrer-i-Carbonell 2008). In addition, this method is very helpful when we employ Probit-type variables in longitudinal analysis (as in our case) with intertemporal correlations). The drawback is that a harsher normality assumption is needed. The reason why these authors prefer POLS to OP is that they can use simple OLS techniques without any loss of information (the resulting trade-off ratios are virtually identical, as are t ratios). Not only are the OLS variants computationally easier than the discrete methods that require the computation of many integrals and/or Monte Carlo simulations, but they also open the way to the application of linear classical methods to discrete response data.

POLS assumes that there is a set X of objectively measurable explanatory variables X₁, . . . , X_k that explain the various domains of life satisfaction, which can be denoted by DS₁, . . . , DS _j (in our case j = 6). According to Van Praag, Frijters, and Ferrer-i-Carbonell (2003), it is quite probable that the endogenous variables DS would influence each other. This being true, the intermediate block of the model in Figure 1 must be seen as a reduced model in which all cross-relations between domain satisfactions have been eliminated. As a result, we can assume a model equation of the type (Van Praag, Frijters, and Ferrer-i-Carbonell 2003):

LS = LS ( DS 1 , … , DS j ; Z )

and each domain of life satisfaction as follows:

DS j = DS j ( x j , Z ) j = 1 , … , 6

where x_j represents a subselection of x variables for the domain j, and the variable Z is, by definition, unobservable and, if no special treatment is given, becomes part of the error terms of the DS and LS equations. As a result, the explanatory variables DS in equation (1) are correlated among themselves and with the LS error term, which would cause an endogeneity bias. To avoid this problem, Van Praag, Frijters, and Ferrer-i-Carbonell (2003) propose constructing an instrumental variable for Z, which will be included in equation (1). This variable is obtained from the estimation of the six DS equations and the corresponding residuals in order to obtain the part of Z that is common to all the residuals. This is derived as the first principal component of the (j × j) error covariance matrix of the residuals obtained after the estimation of each DS equation (2). The inclusion of the variable Z as an explanatory variable in the overall life satisfaction equation (1) allows us to reasonably assume that the remaining LS error is no longer correlated with the DS errors and that the estimators of the coefficients in equation (1) do not suffer from endogeneity bias. ⁵

To apply this POLS method, we also have to take into consideration other important and relevant aspects. First, our dependent variables in equations (1) and (2) (i.e., LS and DS _j ) are categorical, naturally ordered and with i response categories (in our case, i is from 0 = completely dissatisfied to 10 = completely satisfied). To use POLS, it is necessary to transform the LS and DS _j variables into their conditional expectations and based on the properties of the normal distribution (Maddala 1983). For example, we define:

DS j = E ( DS j | µ j , i − 1 < DS j ≤ µ j , i ) = ( n ( µ j , i − 1 ) − n ( µ j , i ) ) / ( N ( µ j , i ) − N ( µ j , i − 1 ) )

where N is the normal distribution function, n is the normal density function, and μ_i are the values for which it holds that N(μ_j, _i ) − N(μ_j,_i–1) is the fraction of respondents belonging to response category i for each DS _j . This transformation has been applied to the overall life satisfaction (LS) and each domain of life satisfaction (DS _j ). Furthermore, and as noted earlier, use of this transformation, first advocated by Terza (1987), ensures that the dependent variables in equations (1) and (2) now vary over the whole real axis. ⁶

Secondly and following Van Praag and Ferrer-i-Carbonell (2008), we can assume that most variables are not fixed over life; that is, they fluctuate around an average or are perceived by the individual as doing so. In order to introduce some dynamic components in equation (1) by distinguishing between transitory and permanent effects for our key explanatory variables (i.e., DS _j ), we replace the contribution of each domain in the overall life satisfaction regression by their annual value and their mean over the period covered by the data. For example, the variable “satisfaction with health,” DS₁, is included in the overall life satisfaction equation as β DS 1 + γ DS ¯ 1 . This expression can be rewritten as β ( DS 1 – DS ¯ 1 ) + ( β + γ ) DS 1 . This decomposition allows us to look at two different effects: differences across individuals in the averages DS₁ measure between effects and the individual deviations from the averages per individual ( DS 1 – DS ¯ 1 ) measure within effects. The coefficient β reflects the short-term or transitory effect and (β + γ) measures the long-term or permanent effect of DS₁ on the overall life satisfaction (Friedman 1957).

Thirdly, the longitudinal dimension of our database allows us to estimate the equations (1) and (2) by individual random effects. However, this may be problematic since the individual random effect model may overlook the potential correlation between individual random effects (e.g., innate ability) with some explanatory variables such as age or income (Van Praag and Ferrer-i-Carbonell 2008; Skalli, Theodossiou, and Vasileiou 2008). To overcome this problem, Mundlak (1978) proposes that the error terms of the LS and DS equations be decomposed into two independent terms:

ε j n t = υ j n + η j n t ,

where n stands for the individual. The term υ _jn represents the individual random effect, that is, the unobservable individual characteristics, and the term η _jnt is the pure error term (which follows a normal distribution, N[0, σ²]). We assume that E(ε) = E(υ) = E(η) = 0, and E(η, x) = 0, namely, that the individual random effect is not correlated with the explanatory variables. Additionally, we introduce in equations (1) and (2) a fixed time and region effect through a set of year and region dummy variables in order to incorporate several effects (e.g., inflation rates, cultural aspects, and socioeconomic environment in each region) and any trend effects in satisfaction. All calculations were made using the statistical package Stata 11.

As for the explanatory variables included in equation (2) and following the standard literature on satisfaction literature (e.g., Zimmermann and Easterlin 2006; Oswald and Powdthavee 2008; Clark et al. 2008), we include, apart from our key dummy variables of participation in holiday trips (i.e., daily/weekly, monthly, seldom, and never), gender, age and its square, marital status, years of education, real household income (in logarithms), the number of children in the household, total number of persons living in the household, employment status, if the individual is a homeowner, year of interview, and region of residence.

Descriptive Analysis

Figure 2 shows the levels of life satisfaction as a whole and those for the different domains of satisfaction reported by nondisabled and disabled individuals according to their participation and intensity in holiday trips. ⁷ Overall, we find that the scores in all satisfaction domains are higher for nondisabled individuals as compared to disabled ones, except for satisfaction with leisure. Using a test of equality of means, we have found that the satisfaction gap between both groups is especially high in some cases, as for example in the domains of job and health satisfaction. For example, the differentials are 2.92 and 1.23 points in favor of nondisabled individuals if the participation in holiday trips is daily/weekly for the domains of health and job satisfaction, respectively. As noted earlier, social tourists (such as the disabled) differ from mainstream tourists (such as the nondisabled) because they are on low incomes and are affected by some specific issues and circumstances (e.g., barriers and obstacles that prevent them from full participation) that mainstream tourists are far less likely to encounter, which leads to having lower levels of satisfaction than the general population (McCabe 2009). In contrast, this differential is −0.25 and −0.09 points in favor of disabled individuals if we analyze the participation category daily/weekly and monthly for leisure satisfaction, respectively.

OPEN IN VIEWER

Figure 2.

Mean levels of overall life satisfaction scores and domains of satisfaction by participation in holiday trips and disability status in Germany.

According to Figure 2, we find a clear positive relationship between the intensity of the participation in holiday trips and all satisfaction domains. Namely, in all cases, the levels of satisfaction reported by nondisabled and disabled individuals increase as the participation in holiday trips is more intense. For example, the life satisfaction scores for the disabled sample in the category daily/weekly is 6.66 points, decreases slightly to 6.51 if the intensity is monthly, 5.67 points in the category “seldom,” and finally goes down to 4.67 if the person has never participated. Similar results are found for the remaining satisfaction domains. However, it is worth noting how taking part in holiday trips has a positive and significant effect on the levels of satisfaction with health and leisure among disabled individuals. These results are consistent with the spill-over effect pointed out by Sirgy et al. (2011), which leads to participation in holiday trips’ having an impact not only on the individual’s life satisfaction scores but also on the rest of the domains of life satisfaction. Although the levels of life satisfaction follow a similar pattern for the nondisabled sample, we observe that the distance in terms of satisfaction between the highest trip category (daily/weekly) and the lowest one (never) is greater for the disabled than the nondisabled (2.02 vs. 1.01 points). The finding is also made for the domains of satisfaction with health (0.7 vs. 1.74 points), job (0.6 vs. 1.18 points), housework (0.36 vs. 1.46 points), and leisure (0.72 vs. 1.11). However, for the disabled sample, this spread between these extreme categories is lower for satisfaction with place of dwelling (0.57 and 0.37 points for the nondisabled and disabled samples, respectively).

To shed further light on this preliminary analysis of life satisfaction and domains of life satisfaction by participation (and its intensity) in holiday trips for nondisabled and disabled individuals, Appendix Table A1 shows the sample distribution into each holiday trip category for the nondisabled and disabled samples used in the estimation process. In all cases, we find that disabled individuals are less likely to participate in holiday trips as compared to their nondisabled counterparts. For example, 27% of disabled individuals (included in the sample used to estimate all domains of satisfaction, except the job satisfaction equation, which has a lower number of observations) have never participated in holiday trips, whereas for the nondisabled sample, this percentage goes down to 12%. This initial finding is in line with the works of Smith (1987), Darcy (1998, 2002), Horner and Swarbrooke (2004), and Pagan (2012b). Following Smith (1987), disabled tourists face three types of barriers: (1) intrinsic barriers, resulting primarily from the tourist’s own levels of cognitive, physical, and psychological function; (2) environmental barriers, consisting of externally imposed limitations; and (3) interactive barriers, relating to skill-challenge incongruities and communication barriers. Many of these barriers are socially constructed (e.g., negative attitudes of staff members and passengers, user-unfriendly services, prejudice, employees’ negative attitudes toward people with disabilities, etc.), and information dissemination on accessible destinations becomes a key factor in increasing tourism opportunities among disabled individuals (Darcy 1998). In addition, those disabled people who are limited in their daily activities are even less likely to participate in tourism activities (Pagan 2012b). These considerations will help us explain and understand the econometric results obtained below.

Econometric Analysis

Appendix Table A1 also shows the mean and standard deviation of the explanatory variables used in the estimation of equation (2) for each domain of life satisfaction. As noted earlier, we have to take into account that the final samples used in the estimation process were not conditioned on whether the individual (disabled or not) also reported a score for job satisfaction. For this reason, Appendix Table A1 shows the mean and standard deviation of the regressors distinguishing the job satisfaction equation (with 21,964 observations) from the rest of domains of satisfaction (all with 39,486 observations). Apart from the lower participation in holiday trips of disabled individuals (as indicated above), we find that these individuals are older, nonmarried, with lower levels of education and household incomes, living in households with a lower number of children and members, out of the labor market, and less likely to be homeowners, as compared to their nondisabled counterparts.

Turning to the econometric analysis, the estimates of the six DS equations (2) corresponding to satisfaction with health, job, housework, household income, place of dwelling, and leisure are reported by disability status in Tables 1 and 2. Initially, we have tested the existence of a possible problem of selectivity bias in all our samples. For example and according to Clark (1997), if dissatisfied disabled individuals can leave the labor market more easily than equally dissatisfied nondisabled individuals, the remaining disabled workers will have higher average levels of job satisfaction because the sample is biased. The underlying assumption is that the potential job satisfaction is related to the probability of being an employee. Furthermore, and as we have panel data, there is a potential selection problem related to nonrandom sample attrition. If disability is linked to a higher (or lower) propensity to remain in the sample, the estimated coefficients will be biased (see Baltagi and Song 2006, for a survey of this type of bias). To test the existence of this problem, we have applied the method proposed by Verbeek and Nijman (1992) for random effects models. ⁸ This test rejected the relevance of a selectivity bias in all samples. ⁹

OPEN IN VIEWER

Table 1.

Domains of Life Satisfaction for NONDISABLED (POLS Individual Random Effects).

	Health		Job		Housework		Household Income		Dwelling		Leisure
	Coefficient	z	Coefficient	z	Coefficient	z	Coefficient	z	Coefficient	z	Coefficient	z
Participation in holiday trips:
Daily/weekly	0.211	8.31 ***	0.105	2.56 ***	0.189	6.76 ***	0.351	14.16 ***	0.170	6.53 ***	0.341	13.16 ***
Monthly	0.207	11.79 ***	0.114	4.00 ***	0.186	9.86 ***	0.343	19.89 ***	0.168	9.32 ***	0.308	17.14 ***
Seldom	0.106	6.67 ***	0.028	1.04	0.109	6.38 ***	0.208	13.36 ***	0.112	6.86 ***	0.173	10.60 ***
Never (reference)	–	–	–	–	–	–	–	–	–	–	–	–
Male	0.050	4.09 ***	−0.034	−2.06 **	−0.065	−5.01 ***	−0.120	−10.52 ***	−0.053	−4.46 ***	0.115	9.79***
Age	−0.044	−20.48 ***	−0.047	−9.88 ***	−0.006	−2.73 ***	−0.035	−17.12 ***	−0.010	−4.82 ***	−0.014	−6.54 ***
Age2	0.027	12.94 ***	0.050	8.97 ***	0.006	2.56 ***	0.043	21.33 ***	0.016	7.57 ***	0.023	10.87***
Marital status
Married (reference)	–	–	–	–	–	–	–	–	–	–	–	–
Single	−0.015	−0.79	−0.057	−2.30 **	−0.153	−7.57 ***	−0.087	−4.74 ***	−0.072	−3.60 ***	0.074	3.84 ***
Widowed	−0.039	−1.43	0.018	0.31	0.005	0.18	−0.013	−0.50	0.003	0.10	0.102	3.89 ***
Divorced	0.036	1.55	0.022	0.75	−0.143	−6.00 ***	−0.233	−10.92 ***	−0.093	−4.12 ***	0.005	0.21
Separated	−0.036	−0.93	0.102	2.12 **	−0.069	−1.76 *	−0.281	−7.43 ***	−0.126	−3.23 ***	−0.012	−0.30
Years of education	0.010	6.12 ***	0.005	2.29 **	−0.004	−2.47 **	0.007	4.61 ***	−0.001	−0.54	−0.009	−5.82 ***
Log (real household income)	0.122	10.99 ***	0.158	8.97 ***	0.021	1.79 *	0.521	35.18 ***	0.183	16.18 ***	0.029	2.73 ***
Number of children
0 (reference)	–	–	–	–	–	–	–	–	–	–	–	–
1	0.034	2.11 **	0.023	1.10	−0.063	−3.67 ***	0.038	2.43 ***	0.001	0.08	−0.072	−4.43 ***
2	0.047	2.21 **	0.078	2.79 ***	−0.102	−4.42 ***	0.138	6.51 ***	0.036	1.64 *	−0.185	−8.36 ***
3+	0.070	2.39 ***	0.115	2.75 ***	−0.075	−2.36 ***	0.147	5.08 ***	−0.002	−0.05	−0.293	−9.75 ***
Household size
1 (reference)	–	–	–	–	–	–	–	–	–	–	–	–
2	−0.049	−2.48 ***	−0.013	−0.48	0.044	2.09 **	0.014	0.74	0.034	1.69 *	0.017	0.86
3	−0.040	−1.73 *	0.032	1.06	−0.032	−1.29	−0.048	−2.15 **	−0.039	−1.68 *	−0.073	−3.25 ***
4+	−0.034	−1.30	0.035	1.05	−0.069	−2.46 ***	−0.068	−2.69 ***	−0.073	−2.82 ***	−0.044	−1.71 *
Employment status
Full-time (reference)	–	–	–	–	–	–	–	–	–	–	–	–
Part-time	0.002	0.16	−0.040	−2.22 **	−0.030	−1.85 *	−0.069	−4.76 ***	0.041	2.75 ***	0.271	18.36 ***
Not working	0.010	0.71	–	–	0.007	0.45	−0.137	−10.11 ***	0.014	1.00	0.415	29.95 ***
Home owner	0.018	1.64 *	0.006	0.37	0.011	0.99	0.117	10.99 ***	0.370	33.09 ***	0.044	4.14 ***
Year
1990	0.075	4.18 ***	0.160	5.99 ***	0.025	1.28	0.266	15.70 ***	−0.026	−1.43	−0.087	−4.83 ***
1995	0.055	3.78 ***	0.067	3.15 ***	−0.011	−0.74	0.163	11.79 ***	−0.071	−4.78 ***	0.019	1.34
1998	0.029	2.16 **	0.083	4.30 ***	−0.036	−2.56 ***	0.131	10.26 ***	−0.032	−2.39 ***	0.013	0.98
2003	0.063	6.20 ***	0.075	5.00 ***	−0.021	−1.94 **	0.085	8.90 ***	−0.036	−3.62 ***	0.001	0.13
2008 (reference)	–	–	–	–	–	–	–	–	–	–	–	–
Constant	0.549	6.40 ***	0.084	0.60	0.149	1.65 *	−2.367	−25.24 ***	−1.035	−11.91 ***	−0.370	−4.40 ***
σ u	0.549	6.40	0.084	0.60	0.149	1.65	−2.367	−25.24 ***	−1.035	−11.91	−0.370	−4.40 ***
σ e	0.596		0.594		0.627		0.572		0.570		0.576
ρ	0.715		0.782		0.760		0.668		0.725		0.705
R 2	0.410		0.366		0.405		0.423		0.382		0.401
Number of observations	0.116		0.039		0.022		0.217		0.109		0.147
Number of individuals	39,486		21,964		39,486		39,486		39,486		39,486

Source: Author’s calculations using the German Socio-Economic Panel (GSOEP) for the years 1990, 1995, 1998, 2003 and 2008.

Note: All equations include regional dummies. The standard errors are robust. Individuals aged 16 or over. We use overall R².

*

p < 0.1, **p < .05, ***p < .01.

OPEN IN VIEWER

Table 2.

Domains of Life Satisfaction for the Disabled (POLS Individual Random Effects).

	Health		Job		Housework		Household Income		Dwelling		Leisure
	Coefficient	z	Coefficient	z	Coefficient	z	Coefficient	z	Coefficient	z	Coefficient	z
Participation in holiday trips:
Daily/Weekly	0.476	5.90***	0.244	1.22	0.439	5.14***	0.202	2.59***	0.016	0.21	0.256	3.45***
Monthly	0.450	9.41***	0.332	2.54***	0.416	8.33***	0.318	7.03***	0.111	2.34**	0.301	6.50***
Seldom	0.275	7.44***	0.236	2.10**	0.299	7.46***	0.139	3.88***	0.062	1.67*	0.126	3.34***
Never (reference)	–	–	–	–	–	–	–	–	–	–	–	–
Male	0.261	7.20***	−0.007	−0.08	0.025	0.69	−0.043	−1.29	0.022	0.63	0.129	3.77***
Age	0.004	0.47	−0.021	−0.93	0.022	2.92***	−0.014	−1.90*	0.015	1.90*	0.015	2.05**
Age 2	0.000	0.00	0.026	1.04	−0.020	−3.17***	0.027	4.34***	−0.006	−0.99	−0.005	−0.82
Marital status:
Married (reference)	–	–	–	–	–	–	–	–	–	–	–	–
Single	0.171	2.28**	−0.082	−0.58	−0.018	−0.25	−0.065	−0.95	−0.099	−1.32	0.119	1.72*
Widowed	0.043	0.68	0.120	0.72	−0.065	−1.07	−0.083	−1.42	0.027	0.44	0.043	0.71
Divorced	−0.039	−0.61	−0.117	−0.81	0.054	0.79	−0.407	−6.25***	−0.149	−2.15**	−0.151	−2.41***
Separated	−0.096	−0.78	−0.193	−1.10	−0.063	−0.50	−0.384	−3.47***	−0.195	−1.45	−0.187	−1.57
Years of education	0.002	0.41	−0.004	−0.33	−0.006	−1.01	0.009	1.43	0.008	1.23	−0.008	−1.31
Log (real household income)	0.035	1.08	0.002	0.02	−0.017	−0.52	0.394	8.77***	0.038	0.94	−0.072	−2.01**
Number of children:
0 (reference)	–	–	–	–	–	–	–	–	–	–	–	–
1	−0.082	−1.20	−0.060	−0.52	0.031	0.44	0.008	0.13	−0.157	−2.16**	−0.166	−2.48***
2	−0.064	−0.66	0.179	1.01	−0.015	−0.13	0.054	0.59	−0.134	−1.24	−0.134	−1.33
3+	−0.293	−2.37**	0.211	0.89	−0.141	−1.03	−0.037	−0.30	−0.377	−2.16**	−0.452	−3.67 ***
Household size:
1 (reference)	–	–	–	–	–	–	–	–	–	–	–	–
2	0.038	0.65	0.002	0.01	−0.009	−0.16	0.021	0.37	0.000	−0.01	−0.044	−0.77
3	−0.003	−0.03	−0.023	−0.14	−0.129	−1.75*	−0.077	−1.07	0.016	0.21	−0.110	−1.50
4+	−0.105	−1.07	−0.310	−1.57	−0.113	−1.17	−0.184	−2.07**	0.010	0.10	−0.146	−1.59
Employment status:
Full-time (reference)	–	–	–	–	–	–	–	–	–	–	–	–
Part-time	0.174	2.53***	0.207	2.34**	0.101	1.36	−0.044	−0.62	0.141	1.86*	0.295	4.33***
Not working	0.213	4.06***	−	−	0.004	0.07	−0.060	−1.25	0.078	1.38	0.444	8.41***
Homeowner	0.064	1.84**	0.120	1.55	0.032	0.89	0.185	5.71***	0.405	11.83***	0.095	2.80***
Year
1990	−0.060	−1.09	0.253	2.09**	0.175	3.20***	0.247	4.93***	0.069	1.31	0.178	3.42***
1995	−0.084	−1.78*	−0.269	−2.41***	−0.041	−0.85	0.258	5.97***	−0.009	−0.19	0.093	2.06**
1998	−0.058	−1.37	−0.028	−0.30	−0.062	−1.42	0.224	5.68***	−0.020	−0.47	0.029	0.69
2003	−0.047	−1.43	0.033	0.38	−0.080	−2.15 **	0.098	3.11***	−0.072	−2.17	0.004	0.12
2008 (reference)	–	–	–	–	–	–	–	–	–	–	–	–
Constant	−0.766	−2.48***	0.374	0.50	−0.544	−1.82*	−2.308	−7.03***	−1.128	−3.45***	−0.738	−2.56***
σ u	0.648		0.811		0.640		0.625		0.643		0.631
σ e	0.657		0.537		0.724		0.608		0.680		0.661
ρ	0.493		0.696		0.439		0.514		0.472		0.477
R 2	0.101		0.089		0.069		0.254		0.118		0.149
Number of observations	3,940		824		3,940		3,940		3,940		3,940
Number of individuals	2,855		727		2,855		2,855		2,855		2,855

Source: Author’s calculations using the German Socio-Economic Panel (GSOEP) for the years 1990, 1995, 1998, 2003and 2008.

Note: All equations include regional dummies. The standard errors are robust. Individuals aged 16 or over. We use overall R².

*

p < 0.1, **p < .05, ***p < .01.

According to Tables 1 and 2, we find that participation and intensity in holiday trips have a positive and significant effect on each of our domains of life satisfaction both for the nondisabled and for the disabled sample. These results confirm the need to introduce in these studies the spill-over effect of holiday trips on all domains of life satisfaction, as Sirgy et al. (2011) concluded. Although this general result is observed for nondisabled and disabled individuals, we observe significant differences in the magnitude or size of the coefficients estimated in almost all satisfaction domains by disability status. Although society assumes that tourism (and holiday trips, in particular) is healthy for all people (Hobson and Dietrich 1994), the contribution of holiday trips to the levels of satisfaction with health is especially higher (more than double all trip categories) for the disabled sample as compared to the nondisabled one. This finding shows the important impact of holiday trips on health satisfaction scores reported by disabled travelers. The benefits of holiday trips in relation to health include enjoyment, personal growth and development, well-being, and freedom from stress (Clark and Clift 1996). Travelers feel relaxed, rested, renewed, and recharged after the trip (Sirgy et al. 2011). These positive implications of holiday trips may be even higher for disabled people because of their isolation, discrimination practices, and the substantial barriers to participating in tourism activities. For example, for cancer patients, feelings of isolation, being ostracised, and fear can be reduced throughout holiday trips because they provide a range of therapeutic opportunities and a necessary means of escapism (Hunter-Jones 2003). In the same line, Mackie and Brattle (1999) point out the significant positive effects of holiday trips on increasing energy, doing more of the things one wishes, and achieving independence for people living with multiple sclerosis. In addition, Pols and Kroon (2007) conclude that holiday trips contribute to psychiatric rehabilitation and better health by promoting and supporting, among other things, community participation, social relations, and a feeling of freedom.

Looking at the results for the job satisfaction domain, once again we find that the contribution of taking part in holiday trips is higher for disabled individuals than their nondisabled counterparts. For the disabled sample, the results show that participating in holiday trips daily/weekly increases levels of job satisfaction by 0.244 points as compared to the reference category (never). On the contrary, the size of this coefficient for the nondisabled is only 0.105 points, that is, 0.139 points lower. Following the Effort-Recovery Theory (e.g., Meijman and Mulder 1998; Geurts and Sonnentag 2006), this premium in job satisfaction that is obtained by disabled people who participate in holiday trips daily/weekly can be due to the fact that these people often recover insufficiently during time off work, and thus they see holiday trips as a vehicle to recover energy and reduce prolonged fatigue, workload, and stress. Furthermore, we have to take into consideration that disability “steals” time, because persons with disabilities need more leisure time to rest, obtain medical care, and accomplish everyday activities (Oi 1991). In addition, time consumed by transportation to and from work may be longer. For example, Oi (1991) mentions that the disabled make nearly three times as many physician visits a year and purchase more than four times as many prescriptions as do individuals with no activity limitations. Therefore, holiday trips boost job satisfaction for disabled individuals because they allow them to break away from the work routine as well as from the demands and constraints of the workplace, aspects especially important for this group (Sirgy et al. 2011).

Similar results are found for satisfaction with housework. Once again, holiday trips contribute to increasing the levels of satisfaction for this domain reported by all individuals (disabled or not). However, the levels of satisfaction are higher for disabled individuals as compared to the nondisabled. The differential between both groups is even higher than that observed for the first two domains analyzed (i.e., health and job satisfaction). According to the existing literature (e.g., Pagan 2013), disabled people (especially females) are more likely than nondisabled ones to spend more time on housework activities (and household production). This fact leads to higher levels of tiredness, fatigue, and stress of the disabled individual as a result of their health concerns, time constraints, activity patterns, and lifestyle requirements. In this sense, vacations can increase the levels of satisfaction with housework observed by disabled individuals (especially among females, as the econometric results confirm) because they allow them to recover from housework duties and spend their time in a different and more relaxing environment that may allow individuals to detach physiologically from housework and from other daily demand and routines (De Bloom et al. 2009).

As for satisfaction with household income, we also find a significant effect of the participation in holiday trips on this domain of satisfaction. For the nondisabled sample, the coefficient on daily/weekly is positive and significant (0.351) at conventional levels with respect to the reference category (never). For the disabled sample, this coefficient is also positive and significant but lower than that obtained for the nondisabled sample (0.202). This lower contribution of holiday trips among disabled individuals may reflect the difficulty of accessing certain tourism activities because of lack of financial resources to fully enjoy the trip (Smith 1987; Sirgy et al. 2011; Pagan 2012b). We have to bear in mind the fact that the personal incomes (and also household incomes) of disabled people depend primarily on their work status. As noted earlier, the labor market integration of the disabled people is not satisfactory in terms of employment, which thus reduces their likelihood to participate in holiday trips. Disabled people have two other important sources of income, disability-related benefits and, to a lesser extent, early or regular retirement benefits (OECD 2003). In any case and as noted earlier, disabled people live in low-income households and this fact may limit their possibilities to request travel and tourism services and products, which have a significant impact on life satisfaction in general and household income satisfaction in particular.

Looking at the satisfaction with place of dwelling, we also find positive and significant effects of holiday trips on the satisfaction scores. However, these effects are lower for the disabled sample, and even the coefficient on daily/weekly is no longer significant at conventional levels. According to Eastman (1978), there are significant differences between households with and without a disabled member regarding the need for additions and alterations to the dwelling, with households with a disabled member having a greater need for these modifications and alterations. In many cases, these needs are not met because of the lack of financial resources, which become the main determinants for the levels of housing satisfaction. In addition, there are other factors or constraints that prevent housing modifications such as, for example, the lack of time, energy, ability, and skills to make needed alterations. These facts may explain the low contribution of holiday trips to housing satisfaction for the disabled sample as compared to the nondisabled one.

Finally and as we expected, holiday trips contribute to increasing the levels of leisure satisfaction reported by nondisabled and disabled individuals. This result is consistent with the existing literature (e.g., Lazarus, Kanner, and Folkman 1980; Neal, Sirgy, and Uysal 2004; Nawijn and Veenhoven 2011; and Dolnicar, Yanamandram, and Cliff 2012), which concludes that leisure activities, including holiday trips, are salutary and may serve as “breathers” that provide a chance to take a break, engage in a pleasurable diversionary activity, and consequently induce positive emotions and feelings that contribute to increasing individuals’ life satisfaction (and leisure satisfaction). According to Tables 1 and 2, the contribution of holiday trips to leisure satisfaction is quite similar for the nondisabled and disabled samples, except for the coefficient on daily/weekly, which is slightly higher for the nondisabled sample (0.341 vs. 0.256 points). This outcome supports the idea that taking a vacation is important for disabled individuals, not only because it offers a means of increasing their leisure and life satisfaction but also because it is a means to an end for them in becoming (or being) self-reliant, independent, confident, and “able” persons (Blichfeldt and Nicolaisen 2011). This notion of being “able” persons through vacations as pointed out by these authors is extremely important. According to Smears (1996), disability relates to how one is defined by “the surrounding society” and, henceforth, being a tourist helps change the social construction as an “unable” person. Tourism enables people with disabilities to develop more positive and empowering feelings and experiences. Hence, holiday trips relate to overcoming self-doubt and redefining oneself and seem to be the means to the end of being an “able” person (Blichfeldt and Nicolaisen 2011).

As for the remaining explanatory variables, it is worth noting some results. Nondisabled males report lower levels of satisfaction with job, housework, household income, and dwelling than do nondisabled females, whereas the opposite result is found for satisfaction with health and leisure. For the disabled, we only find significant differences by gender for the domains of health and leisure. For the nondisabled sample, there is a U-shaped relationship between age and all domains of satisfaction. On the contrary, this U-shaped relationship is only found for satisfaction with household income for the disabled sample. For the nondisabled sample, the variable years of education has a significant effect on almost all domains of satisfaction (except for satisfaction with dwelling), whereas for the disabled, none of the coefficients are significant at conventional levels. The variables number of children in the household and household size have a significant effect for the nondisabled on all domains of satisfaction. For the disabled sample, the effects of these two variables on domains of satisfaction are more limited. For the nondisabled, being a part-time worker reduces the levels of satisfaction with job, housework, and household income and increases the levels of satisfaction with dwelling and leisure. For the disabled, being out of the labor force increases the levels of satisfaction with health and leisure. Finally, we find important and significant differences by year of the interview, especially in the nondisabled sample.

Table 3 reports the results from the estimation of equation (1), which reflect the relationship between overall life satisfaction and each domain of life satisfaction. As noted earlier, we have a reduced number of disabled workers who respond to the job satisfaction question (only 824 observations, which correspond to 727 individuals). For this reason, initially we have estimated equation (1) without the domain of job satisfaction, hence avoiding the loss of a large number of observations (and individuals). In any case, we have also included a table in the Appendix showing the results obtained if we add the domain of job satisfaction as an additional explanatory variable in equation (2). These results will be commented on later. According to Table 3, we find first that the additional variable Z in equation (1) used to correct the possible endogeneity bias is negative in both samples (nondisabled and disabled) but only significant at the 1% level for the nondisabled sample. This result corroborates that endogeneity is indeed a crucial issue that must be taken into account in these kinds of studies, where life satisfaction is considered as a multidimensional variable derived from various life satisfaction domains. Second, we observe that the contribution of the individual random effect to the total variance is relatively large for the disabled sample (0.433). Third, we detect a lower number of significant coefficients in the disabled sample (at the 10%, 5%, and 1% level, respectively) as compared to the nondisabled sample.

OPEN IN VIEWER

Table 3.

Overall Life Satisfaction (POLS Individual Random Effects).

	Nondisabled		Disabled
	Coefficient	z	Coefficient	z
Health satisfaction	0.245	25.31***	0.309	8.92***
Housework satisfaction	0.137	13.87***	0.191	5.31***
Household income satisfaction	0.279	26.09***	0.222	5.36***
Dwelling satisfaction	0.143	14.47***	0.077	2.12***
Leisure satisfaction	0.118	11.79***	0.128	3.80***
Mean health satisfaction	0.099	9.74***	0.011	0.29
Mean housework satisfaction	0.005	0.47	−0.008	−0.21
Mean household income satisfaction	0.052	4.57***	0.012	0.28
Mean dwelling satisfaction	0.032	3.05***	0.028	0.73
Mean leisure satisfaction	0.066	6.27***	−0.029	−0.80
Z	−0.097	−8.96***	−0.038	−1.43
Year
1990	0.161	10.67***	0.190	4.10***
1995	0.071	5.87***	0.101	2.59***
1998	0.075	6.74***	0.096	2.70***
2003	−0.024	−2.71***	0.054	1.72**
2008 (reference)	–	–	–	–
Constant	−0.011	−0.45	−0.038	−0.46
σ u	0.379		0.526
σ e	0.659		0.602
ρ	0.247		0.433
R 2	0.413		0.389
Number of observations	39,486		3,940
Number of individuals	22,189		2,855

Source: Author’s calculations using the German Socio-Economic Panel (GSOEP) for the years 1990, 1995, 1998, 2003, and 2008.

Note: All equations include regional dummies. The standard errors are robust. Individuals aged 16 or over. We use overall R².

*

p < 0.1, **p < .05, ***p < .01.

As noted earlier in the Data and Method section, we can distinguish between transitory (i.e., coefficient on each domain of life satisfaction) and permanent effects (i.e., coefficient on the mean satisfaction of the corresponding domain). For the nondisabled sample, we find that almost all these transitory and permanent effects on life satisfaction domains have a positive and significant effect on life satisfaction. The only exception is the coefficient on mean housework satisfaction. However, we observe that the transitory effects seem to be different and higher than the permanents effects. For example, an increase by 1 percentage point in the transitory satisfaction with health increases the overall life satisfaction of the nondisabled by 24.5%, whereas a 1 percentage point increase in permanent satisfaction with health leads to an additional 9.9% increase in the nondisabled worker’s overall life satisfaction. In addition, we find a significant transitory effect of satisfaction with household income on life satisfaction (27.9%), whereas the transitory effect of satisfaction with leisure is the domain with the lowest contribution to life satisfaction (11.8%). For the disabled sample, we find that only the transitory effects are significant at the 1% level. Namely, holiday trips have no long-term or permanent effects on the levels of life satisfaction of disabled individuals. The highest transitory effect is found for the domain of satisfaction with health (30.9%), followed by household income (22.2%). In contrast, the lowest transitory effect is found for satisfaction with dwelling (7.7%). The nonsignificance of the permanent effects for the disabled sample is striking and is in line with the results obtained by Nawijn and Veenhoven (2011), who concluded, using data from the GSOEP, that holiday trips have a short-lived effect on individuals’ life satisfaction. De Bloom, Guerts, and Kompier (2012) also found that health and well-being levels improved early in the holiday but rapidly returned to base levels on return to work. However, in our case, this outcome is only supported by the disabled sample. This finding may be explained by the set point theory, which argues that individuals return to a life satisfaction baseline after painful (such as divorce, unemployment, etc.) or happy (marriage, the birth of a child, going on holiday, etc.) life events. In this view, particular experiences as, for example, holiday trips, can at best provide a temporary lift, after which we return to our set point (Cummins 2005; Veenhoven 2006).

Additionally, we have reestimated equation (1) but now including as an additional explanatory variable the domain of satisfaction with job. According to Appendix Table A2, we find that the transitory and permanent effects of satisfaction with job are positive and significant at conventional levels for the nondisabled. An increase by 1 percentage point in the transitory satisfaction with job increases the overall life satisfaction by 13.2%, whereas a 1 percentage point increase in permanent satisfaction with job provokes an additional 2.4% increase in the nondisabled worker’s overall life satisfaction. The coefficients of the remaining satisfaction domains are similar to those included in Table 3. For the disabled sample, the number of significant coefficients is even lower than that shown in Table 3 (mainly because of the considerable reduction in the number of observations). In addition, we find that the coefficients of job satisfaction and mean job satisfaction are not significant and thus they do not have an effect on disabled individuals’ overall life satisfaction.

Finally and similar to Van Praag, Frijters, and Ferrer-i-Carbonell (2003), Table 4 shows the cumulative transitory and permanent effects of satisfaction with each domain on the overall life satisfaction. To calculate these cumulative effects, we have summed up the transitory and permanent effects for each domain as reported in Table 3. We also include in parentheses a ranking of domains of satisfaction from the most to the least positive impact on the overall life satisfaction. This ranking shows the importance attached by individuals to each domain in terms of the overall satisfaction effect. For nondisabled individuals, the domains with the highest positive effects on life satisfaction are health (0.344), household income (0.331), and dwelling (0.175). On the other hand, the domains of satisfaction with leisure and housework are located in the lowest positions, with 0.184 and 0.142 points, respectively. As for disabled individuals, once again the domains of health and household income are, in term of satisfaction, the main components that most affect the levels of overall life satisfaction for this group (0.32 and 0.234, respectively), whereas the domains of dwelling and leisure are the components with the least contribution to overall life satisfaction (0.105 and 0.099, respectively). However, we have to bear in mind that all these positive values for the disabled sample have been obtained by adding up the permanent effects which are very low (and with no significant coefficients). Therefore, all these level effects of satisfaction with life domains on overall life satisfaction for disabled people are mainly reflecting transitory effects. ¹⁰

OPEN IN VIEWER

Table 4.

Level Effects of Each Domain of Satisfaction on Life Satisfaction.

	Nondisabled		Nonlimited Disabled
		Rank		Rank
Health satisfaction	0.344	(1)	0.320	(1)
Housework satisfaction	0.142	(5)	0.183	(3)
Household income satisfaction	0.331	(2)	0.234	(2)
Dwelling satisfaction	0.175	(3)	0.105	(4)
Leisure satisfaction	0.184	(4)	0.099	(5)

Source: Results shown in Table 3.