Timing and Contextual Effects on Satisfaction Measurement

ABSTRACT - Does it matter when to ask customers how satisfied they are? When focusing on lasting services it is not evident when it is best to have consumers report upon their degree of satisfaction. There are various reasons for and against interviewing customers during or at the end of the service experience, or some elapse time thereafter. However, the direction of change in satisfaction judgements is contingent on several factors. This study investigates in a longitudinal approach satisfaction appraisals of leisure travellers’ holiday experiences using Oliver’s (1997) multi-domain satisfaction scale. Like in previous studies it could be demonstrated that satisfaction measures decline over time. Controlling for a one and a three months’ elapse time sub-sample no significant difference in their decline could be detected. As could be expected, complaining consumers showed a different pattern of satisfaction change; and so did customers of different travel arrangements. The incident of a new holiday experience or other strong emotional events between first and second appraisal had no impact on the satisfaction level. Implications of the various main and interaction effects are discussed.

Citation:

Andreas H. Zins (2001) ,"Timing and Contextual Effects on Satisfaction Measurement", in E - European Advances in Consumer Research Volume 5, eds. Andrea Groeppel-Klien and Frank-Rudolf Esch, Provo, UT : Association for Consumer Research, Pages: 37-45.

European Advances in Consumer Research Volume 5, 2001      Pages 37-45

TIMING AND CONTEXTUAL EFFECTS ON SATISFACTION MEASUREMENT

Andreas H. Zins, Vienna University of Economics and Business Administration, Austria

ABSTRACT -

Does it matter when to ask customers how satisfied they are? When focusing on lasting services it is not evident when it is best to have consumers report upon their degree of satisfaction. There are various reasons for and against interviewing customers during or at the end of the service experience, or some elapse time thereafter. However, the direction of change in satisfaction judgements is contingent on several factors. This study investigates in a longitudinal approach satisfaction appraisals of leisure travellers’ holiday experiences using Oliver’s (1997) multi-domain satisfaction scale. Like in previous studies it could be demonstrated that satisfaction measures decline over time. Controlling for a one and a three months’ elapse time sub-sample no significant difference in their decline could be detected. As could be expected, complaining consumers showed a different pattern of satisfaction change; and so did customers of different travel arrangements. The incident of a new holiday experience or other strong emotional events between first and second appraisal had no impact on the satisfaction level. Implications of the various main and interaction effects are discussed.

1. INTRODUCTION

Together with service quality the marketing construct of satisfaction is one of the most widely researched issue. Companies and non-profit organisations alike recognise customer satisfaction as one of the key objectives of their business activities. Apart from the conceptual variability of customer satisfaction it is questionable when customers should reveal their degree of satisfaction or framed differently: Does it matter when to ask customers how satisfied they are?

In the context of extended, lasting service experiences, such as stays at resorts, travel services, recreational outdoor activities, education, and hospital services, it could be observed that the consecutive collection of discrete encounters contributes to a varying degree to the overall post-purchase evaluation of satisfaction (DubT and Morgan 1998). The more recent encounters impact more on the final judgement than transactions and experiences in the initial phase of the service delivery process (Chadee und Mattsson 1995, Danahar und Mattsson 1994). DubT and Morga (1998) argued that satisfaction may be stable over time; despite the fact that emotional and cognitive aspects could change from one encounter to the next. They found empirically within the domain of health care services in a hospital that retrospective reports of global consumption satisfaction could be explained predominantly by the in-process dynamic of satisfaction (adj. R² = 0.446).

A related issue is addressed by Stewart and Hull (1992) when differentiating between real-time satisfaction (RTS) and post hoc satisfaction (PHS). Assessing satisfaction during the activity (referred to as the "recreation experience continuum"; adapted from Driver and Tocher 1974) is postulated to be a different concept compared to the postactivity assessment of satisfaction. The different timing is important but not the essence of distinction. The post hoc satisfaction concept captures not a recollection of on-site experiences but the appraisal of the current state of the respondent. Hence, the overall post hoc evaluation is influenced by introspection, memory recall and context effects (Stewart and Hull 1992, p. 197).

Consequently, the authors operationalized in their study both constructs differently: 1. RTS was captured by two statements (satisfaction with the experience right now, like to be at some place elsewhere right now). 2. PHS was covered by five items considering comparisons with ideal, cost, enjoyment, benefit and disappoint factors (scale development by Ditton, Graefe, and Fedler 1980; tested further by Schomaker and Knopf 1982). Hikers had to report on their RTS by means of a self-administering questionnaire ten times during the hike; then on their PHS at the end of the hike, three and nine months after the hiking experience. The PHS immediately at the end of the hike could be explained by the RTS measures only with an R² of 0.31. The model for the three month elapse PHS was even weaker at an R² of 0.18; altogether delivering support to accept that RTS and PHS are distinct constructs.

No significant relationship between the RTS and the PHS after nine months could be detected which led the authors to the conclusion that "sometime between 3 and 9 months after their hike, the subjects’ satisfaction appraisals of an image of the recreation experience became virtually independent of the actual on-site satisfactions." (Stewart and Hull 1992, p. 205). The comparisons between the three postactivity measurements showed that the immediate PHS scores were significantly greater than those in the later periods. However, during the three and nine months elapse time the satisfaction scores did not change significantly.

Corroborating findings on the effect of timing are reported by Peterson and Wilson (1992) from their own and other studies. Observing the satisfaction ratings from several thousand new car buyers in a longitudinal approach, the measures declined by 20 percent from 30 days to 90 days subsequent to the purchase. A less dramatic change was reported by McMahon and Forehand (1983) and by Fisk et al. (1990) in the context of health treatment. However, a significant drop in the satisfaction ratings traced longitudinally will be expected (->H1^-). The holiday experience varies in length considerably like other lasting services (e.g. hospital treatments, consulting services) do. If the response for service quality and satisfaction evaluations cannot be collected strictly at the end of the consumption experience for various reasons it is questionable if the time spot during this experience for measuring satisfaction may affect the outcome. Assuming a random occurrence of unforeseen positive and negative events during a holiday stay the time point is not expected to affect the satisfaction measurement (->H2). Yet, a minimum of the experiential variety in the course of a holiday stay should be expired. Adopting the Stewart and Hull (1992) findings it can be assumed that the decrease will be stronger in earlier periods than more distant to the initial consumption experience (->H3⁺). As more information is processed the further away from the consumption experience the satisfaction appraisal is prone to all sources f additional and comparative information (e.g. Sudman et al. 1996). It was argued that variation is due to the fact that different aspects of satisfaction are being measured at different points in time (Peterson and Wilson 1992). However, this criticism may apply where satisfaction is conceptualised in an unprecise manner.

This study investigates the timing effect of satisfaction measurement longitudinally in the context of leisure travelling. The influence of elapse time as well as of contextual factors of the consumption experience are controlled for. In this study other contextual factors than those of the measurement process are addressed and will be discussed in the next session In order to keep the conceptual variations as small as possible the identical multi-item satisfaction scale (derived from Oliver 1997) was applied two times consecutively.

2. CONCEPTUAL BACKGROUND

Satisfaction research has adopted the perspective that both emotional and cognitive responses are responsible for this kind of post-purchase evaluation. To find consistent structural relationships across different studies it would be necessary to compare identical or at least very similar concepts. However, this requirement cannot be observed throughout a number of published studies: neither on the side of the satisfaction concept (Bagozzi et al. 1999, Hunt 1977, Westbrook and Oliver 1991) nor on the antecedent side of cognitive and affective appraisals (Mano and Oliver 1993, Nyer 1997, Oliver 1993, Zajonc 1980).

The conceptualisation of the satisfaction construct seems to be an issue only for a very limited number of researchers (e.g. Giese and Cote 1999). It is rather exceptional to find a discussion about the content or to find more than a single-item measure to capture overall satisfaction with the exchange process. Bagozzi et al. (1999, p. 201) state that "satisfaction is neither a basic emotion nor a central emotional category in leading theories of emotions". However, the most frequently used measure is simply represented by a uni- or bi-polar scale for satisfaction. Nevertheless, there is great support to view satisfaction as an evaluative post-choice or post-experience judgement which has affinity to attitudes and yet is assumed to represent a hedonic continuum (Westbrook and Oliver 1991). Using Oliver’s 12-item scale of satisfaction (Oliver 1997) capturing the consumer’s fulfilment response it is assumed to encounter both a variety of emotional reactions which are presumably more important outcomes of purchase (Bagozzi et al. 1999) and the cognitive perspective of the consumption evaluation as well.

From focus group discussions Giese and Cote (1999) concluded that consumers assess their satisfaction at various points of time: about 40% before consumption, 48% during consumption, 12% after consumption. They found support for previous findings (Cote, Foxman, and Cutler 1989) "that satisfaction may not yet be determined when asked and that satisfaction may vary with time." (Giese and Cote 1999, p. 18). Consistent with Stewart and Hull (1992) it is contented that post-experience satisfaction is determined at the time of the assessment. It is therefore a changing phenomenon which is influenced by previous satisfaction responses (recall) as well as other cognitive processes driven by introspective and contextual sources. Hence, a limited number of influential factors will be incorporated into this study.

There is ample evidence that emotional reactions associated with the consumption experience is fundamental for the determination of satisfaction (Liljander and Strandvik 1997, Oliver 1993, Mano and Oliver 1993, Westbrook 1987, Westbrook and Oliver 1991). The affective perspective complements previous research findings confirming that satisfaction ascends when perceived performance is equal or greater than expected performance. Both perspectives, positive and negative affect, (dis-)confirmation as well as attribute satisfaction, were confirmed to impact satisfaction when applied to automobiles and to a university course (Oliver 1993). In this study, the former two aspects will be considered.

Satisfaction is not conceived to be a pure affective response to the consumption experience; however, satisfaction judgements may be biased by the current state of someone’s emotional condition. Peterson and Wilson (1992) reported significant positive correlations between satisfaction measures and mood from their own and other studies of the subjective well-being literature (cf. Sudman et al. 1996, p. 87ff ->H6⁺). On the other hand, service quality is found to be the key variable impacting on customer satisfaction (Brady and Robertson 2001, Szymanski and Henard 2001). Whether it be by directly measuring the performance or the disconfirmation of expectations (e.g. Oliver and DeSarbo 1988) or any combined measure of congruence (e.g. Spreng and Olshavsky 1993) there are strong conceptual as well as empirical commonalities with the evaluation of satisfaction (->H9⁺). Consequently, if tracking satisfaction over time the subjective’s assessment of service quality should be monitored simultaneously. A more product specific argument will be challenged as well. A holiday experience is based on a large bundle of different services and service episodes regularly provided by a number of independent suppliers. All-inclusive products emerged in the field of tourism by offering the complete service chain by applying the pay-once principle to the vacation traveller. In addition to the convenience aspect one strength of this product philosophy could be the homogeneous quality which should be easier to communicate and perceive. Therefore, it is hypothesised that all-inclusive travellers exhibit a higher satisfaction level than travellers with a regular arrangement (->H10⁺)

Apart from emotional states and cognitive appraisals of the consumption experience other factors may affect the satisfaction judgement. The decision making for vacation travel is in many cases a joint process where members of the travel party play different B and not necessarily balanced B roles. This means that the effort for information gathering, filtering, evaluation, selection and reservation activities is not equally distributed among the travel party. Therefore it is expected that decision and consumption involvement differs among travel party members. It will be assumed that a higher decision involvement leads to a higher satisfaction levels with more or less the same travel services experienced (->H8⁺).

Due to different expectations and/or personality characteristics service failures will be tolerated or not. The consequence of insufficient service delivery are manifold. One strong sign of disappointed service expectations is complaining (Dr÷ge and Halstead 1991, Nyer 1997). Sometimes complaining activities are difficult to set: time pressure, group restrictions, lack of available people to address or blame are some factors that restrict actual behaviour. However, the intention to complain about a bad service experience could act as a substitute for actual complaining behaviour and is thought to be strongly reflected in lower satisfaction scores (->H7^-).

So far, influential factors have been discussed which arise before or during the service experience. If we compare satisfaction judgements during or at the end of the service experience with satisfaction measured some time span after this experience many things could happen in between which change the retrospective and/or current perception (cf. Sudman et al. 1996, p. 111; ->H4). It is hypothesised that strong pleasant events will positively bias the satisfaction ratings in retrospect and vice versa. Among others it is important to observe whether customers repeated the same kind of service experience within the relevant time span. Confounding and distinguishing effects can be conceived (->H5). Due to recency effects the new experience leaves stronger traces in memory and masks the former impressions. In contrast, the previous experience could serve as a reference or anchor which helps the consumer to contrast two B ore or less B similar encounters. If the evaluation of such follow-up experiences would be known the direction of the satisfaction measurement bias could be investigated.

FIGURE 1

EXPLANATORY MODEL FOR TIME AND CONTEXTUAL EFFECTS ON SATISFACTION MEASUREMENT

2.1 Hypotheses

The following hypotheses summarize the introductory section and the previous discussion. They will be tested empirically in this study. Figure 1 gives a graphical representation of the proposed model. While antecedents such as mood, service quality and complaining issues can be found in traditional static satisfaction models as well this approach tries to incorporate additional factors which are presumably relevant for the dynamic perspective.

2.1.1 Time dependent effects:

H1: Satisfaction will decrease over time.

H2: The time point of the first satisfaction assessment relative to the duration of the consumption experience has no influence on the satisfaction.

H3: The decrease of satisfaction will be stronger the longer the elapse time between consumption experience and retrospection.

2.1.2 Contextual effects:

H4: Strong emotional B unplanned B events occurring between the consumption experience and the retrospection will influence the satisfaction judgement.

H5: A special case of such an event, though in a planned manner, is a repeated consumption experience. If occurring it will influence the satisfaction judgement.

H6: The current mood at the time of the satisfaction measurement will positively influence the satisfaction judgement.

H7: Consumers who actually complained or considered to complain exhibit a less favourable satisfaction than those who did not.

H8: Consumers with a higher decision involvement will be more satisfied than those with a lower involvement.

H9: Quality perceptions are expected to play the major influential role on satisfaction while positive perceptions reinforce satisfaction and less favourable perceptions reduce satisfaction.

H10: All-inclusive travellers are expected to be more satisfied with their holiday experience than regular travellers.

3. RESEARCH DESIGN AND METHODOLOGY

Data of the 1^st questionnaire were collected from 597 leisure travellers by face-to-face interviews during July and beginning of September 1999 in selected holiday regions of Austria. The sample was stratified by quotas for: gender (actually: 55% male), nationality (actually: 68% Germany, 32% Austria), comfort level of the accommodation (only 3- and 4-star or equivalent hotels), length of stay (minimum one week), and elapsed time after arrival (actually: 50% between at least 1/2 and 70% of their stay had expired, 50% later up to the end of their stay; cf. Figure 2). Contacted vacationers were promised to receive a modest present via mail as an incentive for participating in the study and releasing their postal address for a follow-up mail-survey. The oral interview covered socio-demographic characteristics, travel behaviour aspects in general and of the particular trip, a list of (dis-) confirmation elements of the entire vacation, holiday motivations, complaint behaviour or intentions, consumption experience emotions, satisfaction statements, finally repeat visit and recommendation intentions.

The follow-up mail questionnaire was sent out to all of the collected addresses with a pre-stamped return envelope. While the oral, face-to-face interview is a very efficient and effective way to collect complete and reliable data for quota samples the change frm an attended to an unattended mode has to be justified in view of the regional dispersal of the target audience. It is questionable whether a data collection mode bias automatically arises. Peterson and Wilson (1992) reported that satisfaction judgements from oral interviews are biased by +10% to +12% compared to those collected by a paper-and-pencil self-administering questionnaire. From Sudman and Bradburn’s (1974) huge meta-analysis about response effects for attitudinal themes just the opposite biases between face-to-face and self-administered interviews had been reported (p. 31; comparing more than 1.000 studies). Interestingly, for both modes positive deviations from mean ratings are documented on average while no negative individual results appear in the tables contrary to the distribution of deviations for behavioural questions. Of course, all these comparisons are based on cross-sectional studies. The only repeated-measure study cited by Peterson and Wilson (1992) compares telephone with mail interviews which is not strictly comparable to the longitudinal applied in this study.

The timing conditions A and B (cf. Figure 2) were combined resulting in four experimental conditions. A response rate of 50% was achieved. Data for 301 respondents for both surveys were collapsed. The non-response share among the first sample is completely recruited from the B1 sub-sample. All other characteristics are statistically not different from the response sample of the first survey. The questionnaire covered the same satisfaction and service quality statements as the 1^st interview together with additional items described below.

FIGURE 2

TIMING OF INITIAL AND FOLLOW-UP STUDY

TABLE 1

SATISFACTION SCALE (ADAPTED FROM OLIVER=S DOMAIN SPECIFIC INSTRUMENT, 1997, p.343)

3.1 Measurement of Satisfaction

Oliver’s 12-item Likert-type scale (Oliver 1997) was adapted and translated for the use in the particular tourism services context. This satisfaction scale covers the following domains: overall performance, evaluation and quality; need fulfilment; failed expectations; satisfaction anchor"; success attribution; regret; positive affect; failure attribution; negative affect; success attribution; purchase evaluation; and cognitive dissonance. The latter two have been dropped as the wording and content referred to the acquisition and possession of products which would not apply to service experiences. Each statement was measured on a 7-point scale ranging from 1 'entirely reject’ to 7 'fully accept’.

3.2 Measurement of Confirmation/Disconfirmation of travel attributes

The cognitive appraisal was operationalised by 23 items describing the vacation travel experience. Four domains can be distinguished a priori: 1. destination atmosphere (items e.g.: climate, safety, landscape, cleanness, friendliness), 2. infrastructure (e.g. public transportation, cultural offers, shopping facilities), 3. tourism related infrastructure (e.g. excursions, sports, entertainment, facilities for families), and 4. tourism services (e.g. accommodation, food and beverages, personnel). The scale was constructed following the (dis-)confirmation paradigm: '-3' marked the pole where expectations had not been fulfilled at all; the '0' marker signified met expectations while at the pole '+3' expectations had been more than fulfilled. In order to reduce the complexity of possible influences the quality perceptions were condensed by regressing the satisfaction scale against the 23 items of travel attributes. Quality assessment in time 1 explained 18% of the latent satisfaction construct. The following items were retained, summated and averaged for further analyses (cf. Table 2: items PQUAL1 and PQUAL2): landscape impression, climate and weather, comfort of the accommodation, empathy of service personnel, quality and variety of entertainment. The pre-processing for measurement in time 2 was applied similarly. The quality perceptions of five B partially different B components explained the variation in the satisfaction construct by 23%: climate and weather, landscape impression, service in the accommodation, quality and variety of catering, services for families with children.

TABLE 2

SUMMARY ON CONTROLLED VARIANCE FACTORS

3.3 Additional control variables

The initial sample had quotas for different types of vacation arrangements. One third of the respondents spent their holidays with an all-inclusive package covering accommodation, catering, leisure activities, child care and additional services. Another third of the sample represented vacationers who bought an inclusive-card granting free or a reduced-fee access to a large number of leisure attractions during their holiday stay. The last third covered the regular traveller without any special arrangement. Pre-processing of the data did not reveal clear differences between the latter two groups and have been collapsed for further analyses.

The issue of complaining was addressed by asking for reasons for complaining. If respondents indicated some reason they could report on actual complaining behaviour or complaining intentions. Both reactions were treated within one category in this study. The occurrence of an additional holiday experience before interviewing the vacationers a second time was recorded in a dichotomous way. In contrast, other exceptional events could be reported as either negative or positive. If both circumstances happened the answer was coded as neutral. Single-item measures were introduced to capture the degree of decision involvement, the mood state at the first and the second interview.

4. RESULTS

Oliver’s satisfaction scale (1997) was subject to a factor and reliability analysis. The ten statements loaded as expected on one dimension and revealed a very good reliability score (Cronbach’s a = .90 for SATIS t₁ and a = .88 for SATIS t₂; comparable statistics in Oliver 1997, p. 344). Table 3 delivers further evidence from structural equation estimation (AMOS 4.0, Arbuckle and Wothke 1999) that the multi-domain concept of satisfaction can be consistently captured in one dimension. Model fit is reasonable, though a substantial number of correlated error terms (within and across measurement points) had to be accepted. Each indicator delivers a significant contribution (p < .05) to the measurement of satisfaction. However, the loading pattern changes over time. The right column of Table 3 documents the change of the chi²-value for the model fit if one factor loading is restricted to be equal for both measurements. The least reliable change in the loading structure can be observed for the statements addressing overall performance and the satisfaction "anchor". Not considering the absolute level of satisfaction the repeated measurement is significantly correlated with the first assessment in the amount of .54.

The following steps of analysis try to trace the impact on the differences in satisfaction scores during/at the end of the holiday experience versus one/three month later Model complexity is purposely reduced by calculating averaged scores for the uni-dimensional satisfaction construct. The average satisfaction rating diminishes from 6.09 to 5.27 in a statistically significant way; i.e. by 13% from one point in time to the next (H1 accepted; see discussion section). The questions postulated in the remaining hypotheses (H2 to H10) are now going to be investigated simultaneously. Analyses of variance with a repeated measures design, between- and within-subject effects are applied.

The statistical results from estimating time dependent, main and interaction effects are summarized in Tables 4 and 5. The first step of interpretation turns the attention to the non-significant effects. Most important is that time dependent satisfaction differences do not differ whether they are measured with a one ore a three months elapse time (H3 rejected). The same conclusion can be drawn for the different measurement points of time during the service experience (H2 accepted). In other words: the same reduction of satisfaction judgement appears whether vacationers are interviewed in the middle or rather at the end of their stay. No interaction effect between these two experimental conditions (A and B) could be detected.

24% of the respondents travelled for another holiday trip within the time of the first and the second interview. No direct influence on the satisfaction measurement could be observed. Other B positive (12% of the respondents) or negative (7%) B events did not reveal a significant effect either (H4 and H5 rejected).

The current mood state (at t₁ as well as at t₂) shows no main effect on changes in the satisfaction ratings over time or across subjects. However, mood coincides e.g. with the issue of complaining. Furthermore, this influence changes over time. If respondents had some reason to complain the current mood state while on holiday had a weaker correlation with the satisfaction measure. In contrast, when asked back home the new satisfaction measure was even more correlated with the "holiday" mood. This would imply that reasons for complaining act as a reinforcer on satisfaction in retrospect. This line of thought cannot be substantiated because the mood influence is even larger in the absence of complaining activities or intentions. A less reliable interaction effect between mood while on holiday and the retrospective quality perception (p= .045) appears when explaining the variation in the second satisfaction measurement. The negative regression parameter indicates a lower satisfaction if the quality perception is more mood-driven. This negative relationship can be observed with the current mood state at the time of evaluating the perceived quality and satisfaction a second time. A much stronger, though positive, mood effect is paired with the different travel arrangement types. The mood state (at t₂) improves the satisfaction judgement positively among all-inclusive travellers. Other vacationers give mood-biased satisfaction ratings either; however the impact is only half in strength (H6 accepted with conditions).

TABLE 3

MEASUREMENT AND STRUCTURAL MODEL FOR SATISFACTION SCALE IN TIME 1 AND 2

The type of travel arrangement seems to differentiate the satisfaction change over time. In general, satisfaction scores are diminishing. All-inclusive vacationers are less satisfied with their travel experience while still on holiday (H10 rejected). When asked back home this difference cannot be substantiated at an acceptable error level. Hence, the reduction in satisfaction scores is significantly different between all-inclusive and regular travellers.

Complaining intentions or behaviour (20% of the respondents) were expected to influence the satisfaction judgement. Indeed, a significant between-subject effect could be detected. However, in the present model of estimating several effects simultaneously these differences could not be disclosed for both satisfaction measurements in statistically significant manner. However, the influence goes in a plausible direction where complainers are less satisfied compared to non-complainers. The difference between both segments are even more obvious in the hindsight, considering both the direct and the indirect effect together with mood. (H7 accepted without statistical proof). This may be justified by the fact that almost 75% of the complainers were not satisfied how the complaint had been solved.

The intensity of decision involvement plays an ambiguous role. A between-subject effect appears to be relevant (p = .033). However, the direction of this influence changes over time. While on holiday the relationship is inverse: The less involved the respondent had been in the decision making process the more satisfied he or she is with the travel experience. Back home, this contrast turns into the opposite direction. For the all-inclusive traveller this involvement effect does not seem to operate: neither during or at the end of the holiday nor when return home. However, the significance level for the latter contrast is not sufficient (p = .065; H8 rejected).

Finally, the covariate of the parallel quality evaluation has to be considered. Quality perceptions contribute second best to the explanation of satisfaction change over time (partial eta² = .03) and are most responsible for between-subject variations in the satisfaction judgement (partial eta² = .16). This effect is significant at both points of measurement time (H9 accepted). However, the quality perceptions while on holiday do not appear to correlate in a very strong and reliable way. The relationship between perceived quality and satisfaction B assessed with a substantial distance to the real experience B is much stronger and highly significant (p < .001). Among all-inclusive vacationers the qualityBsatisfaction relationship is even stronger.

5. DISCUSSION

Measuring satisfaction at two consecutive points of time should reveal insights into the antecedents and sources of change of the satisfaction construct. For this purpose the multi-domain concept of satisfaction (developed by Oliver 1997) was adapted to be applicable in a service context like tourism. The empirical data substantiate that satisfaction can be conceptualized in a multi-faceted way, yet, representing a uni-dimensional latent construct. From the structural equation analysis it can be derived that the importance of the ten different indicators changes over time. What appeared to be relatively stable are the two facets called "overall performance" and "satisfaction anchor" (Oliver 1997, p. 343). The largest differences emerged with statements capturing "failed expectations", "failure attribution" and "negative affect", altogether negative aspects of the satisfaction formation. This shift in the loading pattern has its counterpart in the absolute differences of scores of the single indicators. Overall, the negative indicators deteriorated much more than the positive ones. While at the single item level the correlations ranged between .15 (failure attribution) and .48 (overall performance) the correlation between the latent satisfaction variables amounted to .54.

TABLE 4

MAIN AND INTERACTION EFFECTS FOR EXPLAINING SATISFACTION AT TIME POINT t¹ AND t²

TABLE 5

PARAMETER ESTIMATES FOR SEPARATE SATISFACTION MODELS FOR t¹ AND t²

From previous research findings it was deduced that satisfaction will decrease over time. The within-subject reduction was determined with 13%. Other studies reported much larger shifts. However, no difference between the one month and three months elapse time measurement could be detected, though Stewart and Hull (1992) reported significant differences within a three-months elapse time. Comparing between-subject contrasts (this study) with within-subject contrasts (Stewart and Hull study) may not be convincing; however sub-samples (experimental condition B) have been carefully constructed and checked for homogeneity. Stewart and Hull (1992) concluded from their results that the decline in satisfaction measures is significant during the first months but did not continue in later periods. Though without precise evidence, it could be expected that if satisfaction decreases over time it will decrease much more immediately after the service experience than in later time periods. Therefor, a substantial difference between the one month and the three months sub-samples could have been expected. Yet, no significant effect turned out to exist, not even if the elapse time (calculated in weeks after the first interview) were introduced as a continuous variable. This robustness against further decrease in later periods may be ascribed to the use of a multi-item satisfaction measurement instrument. Two arguments can be discussed: 1. Introspection of a number of arguments helps to activate more appropriate memory traces than referring to only one emotional stimulus. 2. Even if the perspective of ad-hod constructed evaluations were more appropriate the stability of satisfaction would mean that the cognitive processes underlying the appraisal does not change in this consumption context (Sudman and Bradburn 1974).

There are strong arguments that customer evaluations and attitudes are constructed at the time when being asked and not recalled (Bettman, Luce and Payne 1998; Stewart and Hull 1992). If appraisals are more governed by constructive processes they are more prone to cognitive changes which occurred more recently. Despite some strong emotional events and/or similar consumption experiences respondents in this study did not change their second satisfaction judgement. This finding can be interpreted in favour of more retrospection and less ad-hoc construction of appraisals or in favour of a satisfaction concept which builds on enduring more cognitive and less emotional components.

Various studies have demonstrated that the overall evaluation of a lasting service experience is influenced by every service episode. The more recent encounters were shown to have a larger impact on satisfaction judgements than earlier ones. As satisfaction is conceived overwhelmingly as a post-consumption construct it should be best measured not before the end of the total consumption experience. However, this imperative cannot be obeyed in any cases for various pragmatic reasons of commercial market research. Therefore, this study investigated the effect of asking travellers some days before the end of their holiday stay. Again, no within-subject design was applied. Though from comparing two very similar sub-samples (experimental condition A: relative point of interview during the holiday stay), no significant differences could be detected. It has to be mentioned that the majority of the respondents in the "early" sub-sample had already passed 50% up to 70%) of their stay. This means that results would be probably different and/or invalid if travellers would be asked about their overall satisfaction in the early stages of their consumption experience.

However, mood as an immediate feeling state influences interest, perception and memory (Schwarz and Clore 1983). The mood bias on satisfaction measurement could be demonstrated in this study as hypothesized. The effect, however, was mediated by reasons for complaining and other contextual variables. In addition, complaining was supposed to play a direct negative role on satisfaction. While a negative effect could be identified together with the current mood state during the holiday stay differences between complaining and non-complaining vacationers intensified after the return back home. Without having identified exactly the reasons and the precise consequences of complaints (relevant for almost 20% of the respondents) it can be interpreted that complaints are often based on very unpleasant irregularities with an enduring impact. Complaining activities or complaining intentions may act as mood regulating but do affect quality and satisfaction evaluations persistently.

The concept of decision involvement refers to the stage of preparing the destination selection and the trip itself. Though relevant at different points of time it can be argued that it induces consumers to develop different expectations. In this study it was hypothesized that involvement will encourage higher satisfaction levels. Indeed, this positive correlation holds for the second satisfaction measurement after one or three months, while the opposite result was derived for the immediate satisfaction appraisal. In addition, this relationship holds only or non-all-inclusive travellers. This is probably due to the options all-inclusive travellers acquire when booking an all-inclusive package. Expectations are directed towards a flexible and convenient pool of activities during the stay. The degree of involvement of preparing the vacation stay in advance does not matter.

It can be imagined that the higher involvement with the preparation of the consumption experience the more precise are the expectations and the less open the individual will be to accept deviations from these. Failure attribution, responsibility and regret are more likely to be incorporated by consumers with a higher decision involvement. A lower involvement helps the traveller to be more relaxed and more tolerant to experience things B in a way B he or she had no clear imagination about. In the retrospection this influence turns to the inverse situation. Low-involvement travellers are somewhat less satisfied than highly involved ones. Maybe that more involved people are more in favour of their holiday decisions whereas less involved travellers stress more their unfulfilled desires blaming others, not themselves.

Perceived quality evaluation turned out to be correlated with satisfaction measures as expected. The retrospective judgements revealed a strong and consistent relationship. While on holiday the influence emerged to be by far less intense. In any case, it has to be mentioned that the measurement of perceived quality was reduced from 23 to 5 indicators (taking the significant regression variables only). This procedure facilitated the comprehensive analysis and could be criticized due to an over-simplification. Insights from other industries would suggest to investigate the influence of each of the quality indicators for linear and non-linear relationships.

6. CONCLUSION

This study tried to test various influential factors on satisfaction measures under partially experimental conditions. The hypothesized within-subject decrease of satisfaction with a vacation travel experience substantiated. The entire effect appears to be the same asking travellers one or three months after their return back home. This robustness may be due to the complex ten-item satisfaction instrument which is relatively balanced between emotional and cognitive elements. For practitioners, it could be useful to operate with multi-item measures especially when comparing satisfaction ratings for experiences which date back to different time horizons.

However, only part of the tested effects stems from within-subject contrasts. Much more reliability could be achieved if more factors would be traced longitudinally. A very much appreciated extension of the study design would be additional time points, e.g. one, three and six months for each respondent. Initial sample volumes would have to be adapted accordingly. The remaining 300 respondents in the last round, which were rather too few for so many effects, would afford an initial sample size of at least 2,500 respondents considering a constant drop-out ratio of 50%.

Further research should be done especially in the field of complaining. This study reported a reinforcing effect of complaining travellers when asked to evaluate their experience in the hindsight. Though the effect coefficients were not significant it is interesting to investigate this phenomenon in more detail: e.g. with larger sample sizes, with different services, with more insights into the complaining and responding processes.

Against the overall trend that satisfaction ratings decrease over time customers of an all-inclusive travel arrangement did not show the same amount of decay. This may point to the advantage of all-inclusive products in such a way that they support a holistic perception and experience which is more stable and enduring in memory than a number of independent vacation elements which are offered and consumed without conscious interrelationship.

7. REFERENCES

Arbuckle, J. L. and Wothke, W. (1999). Amos 4.0 user’s guide. Chicago: Smallwaters Corporation.

Bagozzi, R. P., Gopinath, M. and Nyer, P. U. (1999). The role of emotions in marketing. Journal of the Academy of Marketing Science, 27 (2), 184-206.

Bettman, J. R., Luce, M. F. and Payne, J. W. (1998). Constructive consumer choice processes. Journal of Consumer Research, 25 (December), 187-217.

Brady, M. K. and Robertson, C. J. (2001). Searching for a consensus on the antecedent role of service quality and satisfaction: an exploratory cross-national study. Journal of Business Research, 51, 53-60.

Chadee, D. and Mattsson, J. (1995). Measuring customer satisfaction in tourist service encounters. Journal of Travel & Tourism Marketing, 4 (4), 97-107.

Cote, J. A., Foxman, E. R. and Cutler, B. D. (1989). Selecting an appropriate standard of comparison for post purchase evaluations. In T. K. Srull (ed.), Advances in Consumer Research 16. pp. 502-506. Ann Arbor, MI: Association for Consumer Research.

Danahar, P. J. and Mattsson, J. (1994). Customer satisfaction during the service delivery process. European Journal of Marketing, 28 (5), 42491.

Ditton, R. B., Graefe, A. R. and Fedler, A. J. (1980). Recreational satisfaction at Buffalo National River: Some measurement concerns. In D. Lime (ed.), Some recent products of river recreation research (USDA Forest Service General Technical Report No. NC-63. pp. 9-17. Washington, D.C.: U.S. Government Printing Office.

Driver, B. L. and Tocher, S. R. (1974). Toward a behavioral interpretation of recreational engagements, with implications for planning. In B. L. Driver (ed.), Elements in outdoor recreation planning. pp. 9-31. Ann Arbor: University of Michigan.

Dr÷ge, C. and Halstead, D. (1991). Postpurchase hierarchies of effects: The antecedents and consequences of satisfaction for complainers versus non-complainers. International Journal of Research in Marketing, 8, 315-328.

DubT, L. and Morgan, M. S. (1998). Capturing the dynamics of in-process consumption emotions and satisfaction in extended service transactions. International Journal of Research in Marketing, 15, 309-320.

Fisk, T. A., Brown, C. J., Cannizzaro, K. and Naftal, B. (1990). Creating patient satisfaction and loyalty. Journal of Health Care Marketing, 10 (June), 5-15.

Giese, J. L. and Cote, J. A. (1999). Defining consumer satisfaction. Academy of Marketing Science Review [Online], 00 (1).

Hunt, H. K. (1977). CS/D-overview and future research directions. In H. K. Hunt (ed.), Conceptualization and measurement of consumer satisfaction and dissatisfaction. pp. 455-488. Cambridge, MA: Marketing Science Institute.

Liljander, V. and Strandvik, T. (1997). Emotions in service satisfaction. International Journal of Service Industry Management, 8 (2), 148-169.

Mano, H. and Oliver, R. (1993). Assessing the dimensionality and structure of the consumption experience: evaluation, feeling, and satisfaction. Journal of Consumer Research, 20, 451-466.

McMahon, R. J. and Forhand, R. L. (1983). Consumer satisfaction in behavioral treatments of children: Types, issues, and recommendations. Behavior Therapy, 14 (March), 209-225.

Nyer, P. U. (1997). A study of the relationships between cognitive appraisals and consumption emotions. Journal of the Academy of Marketing Science, 25 (4), 296-304.

Oliver, R. L. (1993). A conceptual model of service quality and service satisfaction: compatible goals, different concepts. In T. A. Swartz, D. E. Bowen and S. W. Brown (eds.), Advances in services marketing and management, Vol. 2. pp. 65-85. Greenwich, CT: JAI.

Oliver, R. L. (1997). Satisfaction: A behavioral perspective on the consumer. New York: McGraw-Hill.

Oliver, R. L. and DeSarbo, W. S. (1988). Response determinants in satisfaction judgements. Journal of Consumer Research, 14 (March), 495-507.

Peterson, R. A. and Wilson, W. R. (1992). Measuring customer satisfaction: fact and artifact. Journal of the Academy of Marketing Science, 20 (1), 61-71.

Schomaker, J. H. and Knopf, R. C. (1982). Generalizability of a measure of visitor satisfaction with outdoor recreation. Applied Psychological Measurement, 6, 173-183.

Schwarz, N. and Clore, G. L. (1983). Mood, misattribution, and judgments of well-being: Informative and directive functions of affective states. Journal of Personality and Social Psychology, 45 (September), 513-523.

Spreng, R. A. and Olshavsky, R. W. (1993). A desires congruency model of consumer satisfaction. Journal of the Academy of Marketing Science, 21 (3), 169-177.

Stewart, W. P. and Hull IV, R. B. (1992). Satisfaction of what? Post hoc versus real-time construct validity. Leisure Sciences, 14, 195-209.

Sudman, S., Bradburn, N. M and Schwarz, N. (1996). Thinking about answers. The application of cognitive processes to survey methodology. San Francisco: Jossey-Bass.

Sudman, S. and Bradburn, N. M (1974). Response effects in surveys. Chicago: Aldine.

Szymanski, D. M. and Henard, D. H. (2001). Customer satisfaction: A meta-analysis of the empirical evidence. Journal of the Academy of Marketing Science, 29 (1),16-35.

Westbrook, R. A. (1987). Product/consumption based affective responses and post purchase processes. Journal of Marketing Research, 24 (August), 258-270.

Westbrook, R. A. and Oliver, R. L. (1991). The dimensionality of consumption emotion patterns and consumer satisfaction. Journal of Consumer Research, 18 (June), 84-91.

Zajonc, R. B. (1980). Feeling and thinking: Preferences need no inferences. American Psychologist, 35, 151-175.

----------------------------------------