PLACE IN RETURN Box to remove this checkout from your record.
To AVOID FINES return on or before date due.
MAY BE RECALLED with earlier due date if requested.

 

DATE DUE

DATE DUE

DATE DUE

 

13511sz (2005

 

AUG 5 2005
L 1

 

92 ’05

 

 

 

 

 

 

 

 

 

 

 

 

6/01 cJCIRC/DanDuepes-p. 15

_..—_—.

VISITORS TO GUAM: MODELING SATISFACTION,
QUALITY AND INTENTIONS

By

Ramon Benedicto A. Alampay

 

A DISSERTATION

Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of

DOCTOR OF PHILOSOPHY IN PARK, RECREATION
AND TOURISM RESOURCES

Department of Park, Recreation and Tourism Resources

2003

 

ABSTRACT

VISITORS TO GUAM: MODELING SATISFACTION,
QUALITY AND INTENTIONS

By
Ramon Benedicto A. Alampay

Achieving customer satisfaction, it can be argued, is the core objective of
marketing. However, in the tourism field, academics and managers often use the terms
“satisfaction” and “perceived quality” interchangeably. While researchers generally agree
that the two are related, they have not yet agreed on the nature of the causal relationship
between them. Further, the research into how perceived destination quality and tourist
satisfaction influence fiiture purchase intentions remains limited.

Given these issues, a study of the relationships between tourist
satisfaction/dissatisfaction, perceived quality, and purchase intentions would contribute to
greater understanding of the marketing concept as it applies to tourism. This study
addresses the question of how tourists’ post-visit evaluations (i.e. satisfaction and
perceived quality) of their holiday experience influence their future intentions to return to
or recommend a destination to others.

Using survey data collected from 3,108 visitors to Guam between June 1998 to
November 1999, a conceptual model relating visitor satisfaction (with individual
components as well as with the destination as a whole), perceived quality of the
destination, post-visit image and future intentions was developed and tested using
structural equations modeling (SEM) techniques.

The model hypothesized that satisfaction with the various components of a

destination influenced evaluations of destination quality and overall satisfaction with the

destination. Consequently, these evaluations help form an image of the place as a
desirable vacation destination. This resulting image then influences tourists’ future
intentions regarding the place.

Multiple structural equations models (MSEM) were run to test for differences in
model fit between groups according to nationality, purpose of visit or previous visits to
Guam. Tourists from Japan, South Korea, Taiwan and Hong Kong were interviewed to
collect data for the analysis. By purpose of visit, the sample was divided into pleasure,
corporate, and honeymoon groupings. Potential differences between first-time and repeat
visitors were also assessed.

The SEM analysis suggested that the original hypothesized model - where image
fully mediated the effects of satisfaction and quality on intentions — could be improved by
adding a direct path from overall satisfaction to behavioral intentions. This alternative
model provided the better fit, indicating that post-visit image did not completely mediate
the effects of satisfaction on intentions as had been originally hypothesized. The model
also appeared to predict the probability of recommending the destination to others better
than did the one based on future intentions to return to Guam. Finally, the MSEM
indicated that the models for different groups — by nationality, purpose of visit or

previous experience — differed significantly along various paths within the models.

Copyright by
RAMON BENEDICTO ALDAY ALAMPAY
2003

 

This dissertation is dedicated to my wife and best friend, Gina,
and to our children: Kaipo, Laya and Andre

ACKNOWLEDGEMENTS

I must first give thanks to Dr. Thomas Iverson of the University of Guam and Ms. Ana
Cid from the Guam Visitors Bureau, who generously lent me all the survey data and
documentation I used for this study. I also wish to thank the members of my dissertation
committee: Dr. Donald Holecek, Dr. Nora Rifon, Dr. Bonnie Knutson, and especially, my
adviser, Dr. Joseph Fridgen, for their patience, direction and contribution to the final
form of this project. Finally, I wish to acknowledge Dr. Roger Posadas, Chancellor of the
University of the Philippines-Diliman, and Prof. Evangeline Ortiz, Dean of the UP Asian
Institute of Tourism, whose support made it possible for me to study at Michigan State

University.

vi

 

TABLE OF CONTENTS

LIST OF TABLES ....................................................................................................... x
LIST OF FIGURES ..................................................................................................... xi
CHAPTER I. INTRODUCTION .................................................................................. 1
Theoretical Background ................................................................................. 3
Statement of the Problem and Study Objectives ............................................. 6
Definitions of Key Constructs ........................................................................ 7
Proposed Model ............................................................................................. 8
Delimitations ................................................................................................. 9
Limitations of the study ............................................................................... 10
CHAPTER 11. REVIEW OF LITERATURE ............................................................. 11
Overview of Consumer Satisfaction Theory ................................................. 11

The Nature of Expectations .......................................................................... 13

Service Quality and Consumer Satisfaction. ................................................. 15

How do Satisfaction and Perceived Quality Differ? ...................................... 18

The Disconfirmation Approach Revisited .................................................... 19

The Relationship Between Perceived Quality and Satisfaction ..................... 21
Summary and Synthesis ............................................................................... 32
CHAPTER III. PROCEDURES ................................................................................. 36
Overview of Data Set ................................................................................... 37

Primary Data Source and Data Collection Methods ............................. 37

Sample Characteristics ........................................................................ 39

vii

Limitations of the Data ........................................................................ 43

Research Models and Hypotheses ................................................................ 45
Elements of the Model ......................................................................... 45
Relationships of the Elements .............................................................. 49
Hypotheses .......................................................................................... 51

Statistical Procedures ................................................................................... 54
Overview of SEM and the EQS program ............................................. 54
About EQS .......................................................................................... 56
SEM Procedures .................................................................................. 57

Analysis of the General Model ..................................................................... 58

Multiple Group Analysis of the Model ......................................................... 59

CHAPTER IV. FINDINGS AND DISCUSSION ....................................................... 61

Initial Assignment of Variables and Constructs ............................................ 61
The Working Data Set ......................................................................... 61
Data Reduction and Simplification. ..................................................... 62

The Measurement Model ............................................................................. 69

Assessment of Model F it. ............................................................................. 71

The Structural Model ................................................................................... 76

The Multi-Group Structural Equations Model (MSEM) ............................... 83

By Country of Origin ................................................................................... 84

By Purpose of Travel ................................................................................... 86

By Previous Experience ............................................................................... 89

viii

CHAPTER V. SUMMARY AND CONCLUSIONS .................................................. 93

Summary of Procedures ............................................................................... 94
Summary of Findings ................................................................................... 97
Structural Model ........................................................................................ 101
Implications ............................................................................................... 108
Implications for Destination Marketers ...................................................... 113
Recommendations for Further Study .......................................................... 116
Conclusion ................................................................................................. 120
APPENDICES ......................................................................................................... 124
BIBLIOGRAPHY .................................................................................................... 129

ix

LIST OF TABLES

Table 1. Distribution of Respondents by Nationality and by Gender. .......................... 40
Table 2. Distribution of Respondents by Nationality and by Age Group. .................... 41

Table 3. Distribution of Respondents by Nationality and by Travel
Companion First Mentioned. ..................................................................... 42

Table 4. Distribution of Respondents by Nationality and by Purpose of
Travel ....................................................................................................... 43

Table 5. Constructs of Interest and their Associated Measures ................................. 48

Table 6. Pattern Matrix from Factor Analysis of Hotel Satisfaction

Variables ................................................................................................... 64
Table 7. Pattern Matrix from Factor Analysis of Hotel, Dining and

Shopping Satisfaction ................................................................................ 65
Table 8. Factor Reliabilities of Hotel, Dining and Shopping Satisfaction .................. 66
Table 9. Goodness-of-Fit Measures for the Proposed Model ..................................... 72

Table 10. Assessment of Construct Reliabilities and Variances
Extracted ................................................................................................... 76

Table 11. Results of Cross-Validation Procedures; Comparison of
Goodness-of-Fit Measures for Hypothesized and Alternative

Models ...................................................................................................... 78
Table 12. Standardized Solutions for Dependent Constructs ..................................... 79
APPENDIX A. Means and Standard Deviations of Variables ................................... 124

APPENDIX B. Means and Standard Deviations of Variables by
Nationality .............................................................................................. 125

APPENDIX C. Means and Standard Deviations of Variables by Purpose
of Travel ................................................................................................. 126

APPENDIX D. Means and Standard Deviations of Variables by
Previous Experience with Guam .............................................................. 127

LIST OF FIGURES

Figure 1. Proposed Model of Post-Visit Evaluation and Attitude Formation ................. 8
Figure 2. Hypothesized Relationships between Constructs ......................................... 50
Figure 3. Scree Plot of Hotel Satisfaction Variables .................................................... 63
Figure 4. The Measurement Model ............................................................................. 70
Figure 5. The Structural Model ................................................................................... 77
Figure 6. Structural Model with Factor-to—Factor Path Coefficients ............................ 80
Figure 7. Multiple-Sample Structural Equations Model by Nationality ....................... 85
Figure 8. Multiple-Sample Structural Equations Model by Purpose of Visit ................ 88
Figure 9. Multiple-Sample Structural Equations Model Comparing First-Time vs.
Repeat Visitor ........................................................................................... 91
Figure 10. The Measurement Model ........................................................................... 98
Figure 11. The Structural Model ............................................................................... 102

xi

CHAPTER I

INTRODUCTION

Satisfaction, it can be argued, is the core objective of marketing. As defined in
many textbooks, the marketing concept holds that the function of business is to satisfy
consumers’ needs at a profit (Hunt, 1977). A typical textbook definition can be found in
Kotler, Bowen and Maken (1996) who define the marketing concept as follows:

“The company coordinates all the activities that will affect customer

satisfaction and makes its profits by creating and maintaining customer

satisfaction ...The marketing concept holds as a philosophy of business that it

is the fianction of business to satisfy consumers’ needs at a profit.” (p. 35)

Classic models of buyer behavior (Howard & Sheth, 1969) suggest that if a brand
or product proves satisfactory, it increases the probability of buying that brand, the next
time a similar buying situation occurs. In other words, the marketing concept implies that
customers are likely to re-purchase satisfying products or brands. Research also suggests
that consumer satisfaction/ dissatisfaction (CS/D) influences other post-purchase
behaviors such as word-of-mouth communication and complaining behavior.

In the tourism field, marketing theorists have developed consumer behavior
models that adhere to the general principles of the marketing concept. Typically, these
models incorporate CS/D as a feedback mechanism for future tourist’s decisions and
actions. According to these models, satisfaction or dissatisfaction with the tourist

experience can influence future intentions to return as they become inputs to the tourists’

post-visit beliefs (Um & Crompton, 1990), preferences (Woodside & Lysonski, 1989),

and image (Baloglu & McCleary, 1999; Chon, 1990) of the destination. Consequently,
all these new cognitions and affect arising from the tourists’ post-visit evaluation
reinforce their preference structures and influence future decision making (Teare, 1994).

However, the unique characteristics of the tourism product (destination) and
experience have important implications on the way that the tourist product is perceived
and evaluated. These, in turn, raise questions about the way that the general model of the
marketing concept would work in the tourist setting.

First, the substantial service component of the tourist product suggests that it will
also possess the distinctive qualities of services previously identified by authors such as
Gronroos (1984) and Parasuraman, Zeithaml and Berry (1988) —— intangibility,
heterogeneity, perishability, and simultaneous production and consumption. Furthermore,
the destination product is not one product, but a package of products, services and
activities — each capable of inducing feelings of satisfaction or dissatisfaction.

Another issue to consider is that many people have limited volitional control over
future decisions to travel or return to a destination. Perceived or actual constraints such as
cost, time and the availability of travel companions limit their opportunity and ability to
travel — even if they wished to do so.

Even if a tourist is satisfied with a visit and does not recognize any travel
constraints, other factors may inhibit a decision to return. For example, some tourists may
not want to return (in the near future) simply because they want to visit and experience
new places (Crompton, 1979; Dann, 1977).

Given these issues, a study of the relationships between tourist

satisfaction/dissatisfaction, attitudes and intentions would contribute to greater

understanding of the marketing concept as it applies to tourism. Possible modifications to
the general marketing concept would have important implications for both researchers
and managers.

For researchers, such a modification of the tourism marketing concept might also
imply reconceptualization of terms like satisfaction and dissatisfaction. At minimum,
refinement of what these constructs mean, and how they relate to each other, may
encourage more holistic approaches to tourist satisfaction research.

Destination marketers and managers, on the other hand, may be able to devise
more effective strategies by appreciating how concepts like satisfaction, quality and
image are distinct from, yet related to each other. Moreover, studies like these may help
them understand that the way that these constructs interact may be different in the case of
tourism destinations. In turn, managers might be encouraged to design creative programs
that harness the unique characteristics of the tourism product so that visitors are satisfied

and delighted.

Theoretical Background

A review of the literature revealed some key themes in consumer satisfaction
research — particularly, as these issues are applied to the tourism field. Researchers
generally agree that satisfaction involves a comparison between pre-experience and post-
experience cognitive states (Oliver, 1980), wherein the experience is usually defined in
terms of purchase or consumption. Based largely on Helson’s adaptation level theory (as
cited in Yokoyama, 1991), this line of thinking holds that the difference or “discrepancy”
between these cognitive states is the central concept to understanding consumer

satisfaction.

Typically, the post-experience component has been operationalized as the
consumer’s perception or evaluation of a product’s performance. However, researchers
have disagreed on just what the pre—experience standard —— the basis by which perceived
performance is judged — should be. In brief, the literature suggests that satisfaction is
determined by comparing perceived performance with consumers’ expectations of the
product based on their pre—purchase beliefs (Howard & Sheth, 1969), experience-based
norms (Cadotte, Woodruff & Jenkins, 1987), ideals (Boulding, Kalra, Staelin, &
Zeithaml, 1993) or sense of what is fair and equitable (Tse & Wilton, 1988). Which
standards are used depend on the type of product, the consumer and the environment in
which the consumer experience takes place.

The review of literature also suggests that there is a need to distinguish
satisfaction from other post-consumption constructs such as attitude, image, and,
especially, quality. Both academics and managers often end up using the terms

6

interchangeably. For example, ‘...the term satisfaction, when used by marketers,
frequently refers to an individual’s reactions to attributes of a vacation destination for
which leisure scientists would use the term quality” (Crompton & Love, 1995, p. 12).
Part of the difficulty in reaching consensus stems from the differences in
researchers’ use of the same terms for different contexts. For example, researchers and
managers have applied the terms to both global- (overall) and component-level
evaluations. Similarly, both satisfaction and quality have been used as cumulative
(enduring) as well as transaction-specific (temporary) concepts. Some authors have

argued that this is an important difl’erence between satisfaction and perceived quality. For

example, MacKay and Crompton (1990) have argued that satisfaction refers to a specific

transaction or experience, while quality evolves from the summation of past transactions
and experiences with a service.

They also suggest that quality is a predominantly cognitive evaluation based on
consumer expectations of what the product should be. Satisfaction, on the other hand, has
been described as a primarily affective evaluation based on consumer expectations as
predictions of what the product would be.

While researchers generally agree that the two constructs are related, they have
not yet agreed on the nature of the causal relationship between satisfaction and quality.
Some (Oliver, 1993a/1993b; Cronin & Taylor, 1994) take the position that service quality
leads to satisfaction. On the other hand, there remain others who see satisfaction as an
antecedent of service quality (Bitner, Booms, & Tetreault, 1990; Bolton & Drew, 1991).
Again, differences in the ways that these authors operationalized quality relative to
satisfaction could explain the different conclusions reached by each.

A final observation from the literature is that most authors generally agree that
higher levels of perceived quality and satisfaction should positively influence fixture
intentions toward a product. However, research into how this all works in the tourist
setting remains limited. Oh and Parks (1997) note that in the hospitality field, research
has been focused more on the antecedents of satisfaction rather than the consequences of
a satisfying vacation. The tourism marketing literature has also been characterized by an
almost complete lack of publications in the areas of destination loyalty and repeat
visitation (Oppermann, 1999). Furthermore, there remains a conspicuous lack of research
on the link between tourists’ satisfaction or dissatisfaction with a destination and their

future behavior (Hudson, 1999).

Statement of the Problem and Study Objectives

The main problem to be addressed in this study can be summarized in the

following manner:

How do tourists’ post-visit evaluations of their holiday experience
influence (if, at all) their future intentions regarding a destination?

Phrased in this manner, the research question implies an investigation of the
relationships between two main sets of consumer behavior constructs — one set
composed of post-visit evaluations, and another set of future behavioral intentions toward
the destination. This problem calls for better understanding of how these evaluative and
behavioral constructs relate to each other. Thus, it will be necessary to develop and test a
conceptual model that captures the interrelationships of these different constructs.

Thus, the main problem provides a research umbrella that covers several
objectives to be addressed in this study.

These objectives are:

1. T 0 develop a model of tourist behavior that defines and distinguishes
between the key post-experience constructs of satisfaction, quality and
image;

2. T 0 determine the relative contributions of the various components of
tourism (hotels, restaurants, shopping establishments, attractions, etc.) to
overall satisfaction with a destination;

3. T 0 identijjling dtflerences, if any, in the predictive power of the model
with regard to behaviors that people have varying degrees of control -
over (e. g., intentions to return and intentions to recommend to friends) ,'
and

4. T 0 test the eflect of individual diflerences in motivation and level of
information on the relationships between the various post-experience
constructs.

Definitions of Key Constmcts

The key constructs around which this dissertation revolves are satisfaction,
perceived destination quality, image and behavioral intentions.

Perceived destination quality is defined as the evaluation that the attributes
possessed by a product are consistent with the tourist’s belief-based expectations about
the characteristic attributes that superior members of that product class (in this context,
tourist destinations) should offer. These “characteristic” attributes are elements of the
destination over which the tourist-evaluators have very little or no control. Furthermore,
this definition suggests that the quality evaluation is predominantly cognitive in nature.

Satisfaction is defined as the evaluation of the tourist’s affective reaction to the
consumption experience. The tourist’s affect is based on a comparison of product
performance and the tourist’s desires, needs, motives and beliefs about the product’s
attributes and the benefits to be derived from them. The model acknowledges that tourists
realize satisfaction both with the individual components of their holiday experience as
well as with the destination as a whole. However, in this study, component-level and
overall satisfaction were measured differently.

11mg; refers to the mental picture that tourists hold about a destination or product.
In this study, image corresponds to an attitude in that it represents an enduring positive or
negative feeling about the destination (Petty & Cacioppo, 1981). Measured in terms of
desirability as compared to other international sites, image is an attitude in that it is also a
disposition to respond favorably or unfavorably toward the object (Ajzen, 1988).

Intentions are indications of the willingness and commitment of tourists to

perform particular behaviors or actions. For this particular study, the behavioral

intentions to be considered involve returning to the destination and recommending the

destination to friends and relatives back home.

Proposed Model

The model shown in Figure 1 is based on a review of the literature on customer

satisfaction fiom the marketing, recreation and tourism fields. It is a generic description

of relationships between constructs that this dissertation intends to test empirically. It

defines the destination product as a package of services and experiences that can be

evaluated independently. Thus, satisfaction can also be determined for each of the

destination’s individual components.

Lodging"

Shopping I

Attractions'

 
   

  

Perceived
Destination
Quality

    
 
   

Image Future

Intentions

   

Overall
Satisfaction

  

* - Satisfaction with Destination Component

Ej gure 1. Proposed Model of Post-Visit Evaluation and Attitude Formation

It is hypothesized that each of these component satisfactions will have direct

Effects on perceived quality of the destination, and on overall satisfaction with the

destination. Furthermore, destination quality is envisioned to be an antecedent of overall
satisfaction. In turn, both destination quality and overall satisfaction influence the
tourist’s post-visit image of the destination. Finally, image is posited to influence
behavioral intentions. In the process, it acts as a mediator of the effects of destination
quality and overall satisfaction on intentions.

The figure shows four different tour components, each with its own level of
satisfaction. These four components correspond to those elements evaluated for this
study. It should be stressed that the figure is not intended to depict a comprehensive list
of any destination’s attributes. The composition of these components will vary depending

on the geographic location and type of destination.

Delimitations

The study focused on the consequences, not the antecedents, of tourist
satisfaction. It was assumed that in forming satisfaction judgments on various
components of the destination product, tourists had consciously or unconsciously
processed their experiences in the manner prescribed by the disconfirmation of
expectations theory. Thus, the tourists’ expectations and performance evaluations were
assumed to be largely reflected in stated satisfaction levels.

Furthermore, because the vacation decision involves a much longer time frame
than typical consumer products, this project did not attempt to measure fixture behavior.
Rather, as suggested by the theory of planned behavior, the study instead used stated

behavioral intentions as a proxy for future behavior.

Limitations of the study

The dissertation was based on analysis of secondary data — data previously
collected by the Guam Visitors Bureau. In relying on secondary data, the researcher
recognizes that the original purposes for collecting this information were different from
those of this current study. As a result there were some inherent measurement problems
in the dataset. Because the focus of the original study was on tourist satisfaction, there
were more variables associated with component-level satisfaction than any other
construct in the proposed model.

Fewer variables were associated with constructs such as destination quality,
image and behavioral intentions. In the absence of variables designed to explicitly
measure this study’s constructs of interest, other variables were utilized as proxies for the
missing variables. For example, there was no variable in the survey that directly asked
respondents about their overall satisfaction with Guam. There was, however, an item
which asked whether “Guam was better than expected.” This statement appears to be
reflective of disconflrmation, a recognized antecedent of satisfaction. Thus, this specific
variable was used as an indicator of overall satisfaction.

The original study also did not include measures of tourist expectations and
motivations. However, through the use of simultaneous multi-group modeling, this
dissertation explored the use of various categorical variables as possible proxy variables

for these absent measures.

10

CHAPTER II

REVIEW OF LITERATURE

Overview of Consumer Satisfaction Theory

Consumer satisfaction has long been a pillar of the marketing concept. Through
the years, marketing researchers have generally defined consumer satisfaction in a
relatively consistent fashion (e. g., Howard & Sheth, 1969; Hunt, 1977; Martilla & James,
1977; Swan & Combs, 1976). In its broadest sense, consumer satisfaction involves some
form of comparison between pre-experience and post-experience cognitive states
(Oliver, 1980), wherein the experience is usually defined in terms of purchase or
consumption. Based largely on Helson’s (as cited in Yokoyama, 1991) adaptation level
theory], this line of thinking holds that the difference or “discrepancy” between these
cognitive states is the central concept to understanding consumer satisfaction.

In applying Helson’s theory to consumer behavior, the consumers’ prior
expectations represent the standard to which they compare their post-consumption
perceptions of a product’s performance (Howard & Sheth, 1969). The consumer’s
expectations are said to be confirmed when the product performs as expected.
Expectations are positively disconfirmed when the brand performs better than expected.
On the other hand, expectations are said to be negatively disconfirmed when the service

or product is perceived to perform worse than expected. Thus, confirmed or positively

 

' Oliver (1980) warns that this was not originally proposed as a satisfaction theory. However, Oliver writes
that the theory “squares” with the satisfaction data known to him, and that the theory is “sufficiently
general” as to accommodate the consumer satisfaction theories.

11

disconfirmed expectations lead to satisfaction, while negatively disconfirmed
expectations result in dissatisfaction.

Other authors have attempted to refine the model by proposing more nuanced
alternatives to the basic confrrmation-disconfirmation framework. For example,
assimilation-contrast theory (Sherif & Hovland, 1961) — as in classic discrepancy theory
— involves a comparison between pre-purchase expectations and post-consumption
perceptions. However, assimilation-contrast adds that satisfaction or dissatisfaction is, in
the end, a function of the degree of difference between expectations and perceived
performance.

If the differences are “not too large” (moderate disconfirmation), consumers are
likely to assimilate their evaluation toward their earlier expectations. Thus, high
expectations lead to more favorable evaluations; low expectations lead to less favorable
perceptions under these conditions. If the disconfirmation is perceived to be “very
large”, consumers are likely to magnify this contrast in the direction of the discrepancy.
Large differences between perceived performance and prior expectations should result in
greater satisfaction or dissatisfaction, depending on the direction of the evaluation.

Sirgy’s (1984) evaluative congruity model holds that the greatest satisfaction is
possible under conditions of positive incongruity, or when the consumer’s expectations
are negative and perceived performance is positive. On the other hand, negative
incongruity (positive expectations vs. negative perceived performance outcome) leads to
the least amount of satisfaction. Thus, the model suggests that surprise is an important

influence on consumer satisfaction or dissatisfaction.

12

The Nature of Expectations

In spite of these alternative models, it is nevertheless fair to say that the basic
disconfirrnation of expectations theory continues to be the dominant model of consumer
satisfaction (Spreng & Olshavsky, 1993). However, the popularity of the discrepancy
fi'amework has not curtailed debate on what prior expectations really mean and how the
evaluation process unfolds.

Early studies (Martilla & James, 1977; Swan & Combs, 1976) tended to flame
consumer satisfaction in terms of the predicted quality of the product and its attributes.
This means that expectations refer to the consumer’s beliefs about what product
performance would be.

Cadotte, Woodruff and Jenkins (1987) argued that the comparison standard
should reflect what consumers believe a brand should, rather than would, provide to meet
their needs and wants. They proposed that consumers hold “experience-based norms” that
reflect desired performance in meeting their needs and wants. However, the consumers’
experiences with real brands work to establish constraints on what the level of
performance can be. These norms may be developed in two ways. The product type norm
is an average of one’s past experiences with the class of products to which the focal brand
belongs. On the other hand, best brand norms derive from a person’s favorite brand.

Boulding, Kalra, Staelin and Zeithaml (1993) identified two classes of
expectations - normative and ideal expectations. Normative expectations involve
customer expectations about what will and what should happen in their next encounter
with a firm. Ideal expectations represent the optimal product performance that a

customer wants in an ideal sense. In addition to ideal and expected (normative) product

13

performance, Tse and Wilton (1988) suggested that consumers also held notions of
equitable product pejormance. This represents the level of performance the customer
ought to or deserves to receive based on a perceived set of costs. By comparing
cost/reward ratios to those of others in their environment (e. g., merchants, other users of
the product, or non-consumers), the consumer is able to evaluate the fairness or
unfairness of the transaction.

This confusion over the ideal comparison standard has not been limited to the
marketing field. In his treatise on job satisfaction, Locke (1969) lamented the failure to
distinguish between expectations, needs, wishes and values. These terms, he noted, were
being used interchangeably to establish the standard for evaluating job satisfaction.
Whereas expectations denote one’s beliefs about what will occur in the future, Locke
argued that needs should refer to objects and conditions that people require to maintain
their physical and psychological well-being. Furthermore, needs should be distinguished
from wishes or values — which are what people actually seek to gain or consider
beneficial.

Similar notions have been raised in the recreation, leisure and tourism literature.
The particularly strong marketing orientation of the tourism field (Ritchie, 1996) has
tended to encourage attribute-oriented satisfaction research (e.g., Mazursky, 1989;
Pizam & Milman, 1993; Teye & Leclerc, 1998). However, the influence of the recreation

and leisure tradition2 has also encouraged the perspective that the evaluation is not

 

2 The exchange between Dann (1978) and Pizam, Neumann, & Reichel (197 8) is a good illustration of
the resulting tension from the multiplicity of research traditions in the tourism field. From Dann’s
leisure-oriented perspective, tourist satisfaction should be prOperly studied as a domain of overall
satisfaction with quality of life. He held that the tourist satisfaction was much too complex to be
subject to the transactional framework of consumer satisfaction. On the other hand, Pizam et a1. argued
that, indeed, tourist satisfaction could and should be analyzed as a form of consumer satisfaction, albeit
a more complex one.

14

limited to the quality of product attributes per se. Leisure researchers have long
recognized that recreationists and tourists also hold expectations about a product’s
potential to firlfill needs or desires and to provide fiinctional as well as psychological
benefits (e.g., Bultena & Klessig, 1969; Floyd, 1997).

More recently, other consumer researchers have moved toward integrative rather
than exclusive models of satisfaction. These models recognize that attribute-based and
need-based expectations both (often times, jointly) influence the consumers’ evaluation
of and subsequent satisfaction with products and services. It is notable that much of the
progress in this area has been achieved in the services marketing arena, where satisfaction

research has often been conducted in step with service quality research.

Service Quality and Consumer Satisfaction.

While researchers in the manufacturing sector have traditionally relied on a
supply-based definition of product quality — one based on minimizing manufacturing
defects (Parasuraman, Zeithaml, & Berry, 1985) — service marketers have leaned toward
a more consumer-oriented definition of quality. One reason for this is that consumption
and production are said to take place simultaneously in a service. More than with tangible
goods, clients play a large part in determining the quality of the service experience.
Furthermore, the quality of a service is often more difficult to evaluate than the quality of
a typical good. In contrast to goods, services offer few attributes that can be determined
prior to purchase and more attributes which can only be discerned after purchase or

consumption (Parasuraman et al., 1985).

15

Thus, service quality, like satisfaction, has often been characterized as a post-
experience evaluation. And as with satisfaction, service quality has largely been defined
in terms of a disconfirmation of expectations framework. For example, Parasuraman et
al. (1985) offered this definition:

“The quality that a consumer perceives in a service is a function of the

magnitude and direction of the gap between expected service and

perceived service” (p.46).

The application of discrepancy models to describe both service quality and
consumer satisfaction has led to a situation wherein the two constructs have been used
interchangeably by both academic researchers and professional marketers. In addition,
discrepancy models in studies of perceived quality or satisfaction have tended to focus on
relatively similar facets of the product or service.

A review of selected tourism satisfaction studies indicates that researchers often
choose to focus on a familiar list of destination and tourist service attributes. For
example, various studies (Danaher & Arweiler, 1996; Pizam et al., 1978; Yokoyama,
1991) have identified the following as key dimensions of tourist satisfaction with a
destination: lodging facilities and services, transportation, tourist activities and
attractions, and the natural environment or physical setting. Studies of specific tourism or
recreation services (Crompton, Mackay, & Fesenmaier, 1991; Heung & Cheng, 2000)
have similarly focused on specific attributes of the focal service. However, in keeping
with the service marketing literature, they recognize that these service attributes may be
tangible as well as intangible in character.

Satisfaction studies dealing with specific recreation or tourism activities have also

recognized the role of attribute evaluations in determining overall satisfaction with the

16

experience. However, these studies have acknowledged that individual motivations and
objectives also play roles in the final evaluation of the experience.

A study by Mayer, Johnson, Hu, and Chen (1998) identified atmosphere,
customer service and process as dimensions of satisfaction with casino gaming.
Atmosphere refers to tangible elements of the casino such as interior design and casino
size. On the other hand, customer service and process covered the intangible elements of
the gaming services. However, in addition to these three dimensions, Mayer et al. also
identified a “chance of winning” dimension which reflected visitor expectations about the
fairness of the games and their expectations about achieving individual objectives for
participation.

Lounsbury and Hoopes (1985) investigated tourist satisfaction with the vacation
experience and identified five key factors. Two of these —— food and lodging, and natural
environment —- are associated with specific elements of the destination. Two other
factors appear to be related to tourist motivations for going on vacation. A factor labeled
“escape” refers to opportunities for getting away from it all. Another factor called
“relaxation and leisure” includes tourist perceptions of the way their plans worked out,
the amount of fun they had, as well as the amount of relaxation they achieved. Finally,
Lounsbury and Hoopes identified a fifth dimension of satisfaction “marriage and family”
that corresponded to items pertaining to the respondent’s relationship with spouse and
family, including the behavior of children during the trip. This suggests that aside fi'om
their evaluations of the destination itself, tourists also recognize and evaluate elements
that they (as individuals and as members of a travel party) bring to the overall vacation

experience.

17

How do Satisfaction and Perceived Quality Differ?

Some researchers now argue that these personal contributions to the experience
— motivations, objectives, desires, constraints, etc. — represent the core issue that
differentiates perceived quality from consumer satisfaction. Oh and Parks ( 1997), for
example, have argued that satisfaction is a consumer’s subjective comparison of
expectation and performance. In contrast, service quality represents the researcher’s
objective comparison of expectation and performance.

Oliver (1993a) argued that satisfaction is a primarily affective construct even as it
also reflects cognitive evaluations of a product’s attributes. He noted that attribute
satisfaction was distinct fiom attribute performance. Performance was assessed in the
context of its variance and the strength of consumer beliefs. Although, not explicitly
labeled as such, this sounds very much like a notion of perceived quality. He also argued
that satisfaction could be considered a psychological response to the performance
judgment.

Oliver (1993b) expanded on this theme in another paper that proposed a model
integrating the satisfaction and service quality literatures. Parasuraman, Zeithaml and
Berry (1988) had suggested that service quality is based on expectations as consumer
desires or wants, i.e. what the consumer feels the service should be. On the other hand,
satisfaction researchers saw expectations as predictions of what the service would be. In
his model, Oliver recognized that both desires and expectations were used as standards
for evaluating product or service performance. This meant that two forms of
disconfirmation occurred — one based on ideals or desires, and one based on predictions

or anticipations of the product’s performance.

18

Spreng and Mackoy (1996) proposed and tested a modified version of this model
wherein they identified desires congruency (the disconfirmation of desires) as influencing
both perceived service quality and overall satisfaction. In contrast, expectations
disconfirmation directly affected overall satisfaction but did not have an influence on the
overall perception of service quality.

Since expectations disconfirmation suggests an element of surprise (i.e., actual
difi’ering fi'om predicted outcomes), the conclusion is that satisfaction is a more affective
evaluation. On the other hand, service quality appears to be a primarily cognitive
evaluation based on consumer expectations of what the product or service should be.

Sirrrilar insights have been advanced in the tourism and recreation literature.
Crompton and Love (1995) argued that service quality refers to the attributes of a service
primarily controlled by the supplier. On the other hand, satisfaction (or quality of
experience) refers to an emotional state of mind after exposure to the opportunity or
service. This framework recognizes that satisfaction is influenced by. the tourist’s socio-
psychological state (mood, needs, disposition) as well as by the specific attributes

controlled by suppliers (e. g., facilities, services and attractions).

The Disconfirmation Approach Revisited

The disconfirmation model enjoys strong intuitive appeal as an explanation of
how evaluations of satisfaction and quality are formed. However, there is a growing body
of research that suggests that direct measures of performance are better predictors of
satisfaction, quality and intentions than measures derived through disconfirmation

equations.

19

Peter, Churchill and Brown (1992) have pointed out psychometric problems
arising from the use of difference scores, i.e. subtracting one measure from another. They
applied these arguments specifically to the SERVQUAL scale (Brown, Churchill, &
Peter, 1992) and concluded that there were serious problems in conceptualizing service
quality as a difference score. Carman (1990) opined that expectations not based on
experience would be likely to be poor standards against which to measure quality of
performance.

Cronin and Taylor (1992) found that a directly measured scale (SERVPERF) of
the same items covered by SERVQUAL was more efficient than the latter scale. Similar
tests in the tourism field have tended to support this conclusion.

Fick and Ritchie (1991) measured service quality in several types of tourism
businesses and found that mean perception of performance scores provided better
evaluations of service quality than a computed (perceptions-minns-expectations) score. In
his study of tourists to Michigan’s northwestern coastal region, Yokoyama (1991) found
that performance-based model was a better predictor of tourist satisfaction than a
discrepancy-based model. Because there was a length of time between the measurement
of expectations and performance, he surmised that respondents “could not keep their
expectations current and available as a basis to form the comparison which would predict
their satisfaction level”(p. 12). Furthermore, he suggested that the comparison. process
was already included — probably unconsciously and unintentionally — in the
respondents’ assessment of performance.

Crompton and Love (1995) evaluated seven alternative operationalizations of

quality (E, I-P, I x E, I x P, P-E, I x (P-E), and P); where E = expectations; I =

20

importance; P = performance) and concluded that performance-based operationalizations
were the best predictors while the least accurate predictors of quality were the
disconfirmation-based measures. Similar findings were reached by Childress and
Crompton (1997). However, they noted that the direct performance measure offers little
diagnostic potential despite its superior predictive power. This, they argued, could lead to
inappropriate priorities being set. Thus, disconfirmation-based formats might still be

more useful to managers.

The Relationship Between Perceived Quality and Satisfaction

Another issue that needs to be clarified is the nature of the relationship between
perceived quality and satisfaction. While researchers agree that the two are related, there
has been some debate regarding the direction of this relationship. Spreng and Mackoy’s
(1996) model advocated a causal link from service quality to satisfaction. In other words,
they supported Oliver’s proposition that perceived service quality (PSQ) subsequently
resulted in recognition of consumer satisfaction or dissatisfaction. While others
(Cronin & Taylor, 1992; Oliver, 1993b) have also adopted this position, another school of
thought has advocated the opposite relationship — that satisfaction influences or
mediates overall service quality (Bitner et al., 1990; Bolton & Drew, 1991). They hold
that the accumulation of satisfaction over time subsequently results in a change in the
consumer’s assessment of service quality for a specific organization.

Gotlieb, Grewal and Brown (1994) tested these competing models and concluded

that the perceived quality —> satisfaction model provided a better fit for their data than

21

did a model showing the opposite relationship. Furthermore the quality-to-satisfaction
model explained more of the variance in behavioral intentions than did the alternative
model.

The early confusion regarding the relationship between service quality and
satisfaction may be explained in part by the differing notions of specificity associated
with each construct. Depending on their research objectives, researchers tended to vary in
whether they saw either construct as an enduring evaluation or as a transaction-specific
one. Current thinking on satisfaction and service quality now appears to hold that both
these constructs can be defined on a transaction-specific (component) level as well as a
long-term relationship (or global) level (Teas, 1993).

In the leisure field, Mannell and Kleiber (1997) wrote that satisfaction studies
could be distinguished by their level of specificity. For example, some researchers have
investigated satisfaction at a transaction-specific level — or what Mannell and Kleiber
call the molecular level of specificity. These include studies on specific activities such as
camping (Bultena & Klessig, 1969), fishing (Graefe & Fedler, 1986) and hunting
(Floyd, 1997). Other researchers have had more global or “molar” concerns related to the
issue of how people met all of their leisure needs. In other words, these studies were
interested in determining how satisfied people were with their leisure style.

Parasuraman, Zeithaml, and Berry’s original work (1985) suggested that service
quality was a global and enduring judgment of a service while satisfaction related to a
specific transaction. Consequently, they held that service quality resulted from an
accumulation of incidents of satisfaction. In a later article (Parasuraman, Zeithaml, &

Berry, 1994), the authors reversed themselves and offered a model that held a customer’s

22

overall satisfaction to be a fiinction of his or her assessment of service quality, product
quality, and price. However, in this model, both service quality and satisfaction were
recognized as transaction-specific evaluations.

Thus, when both constructs are held to be transaction-specific, the contemporary
literature now appears to favor the quality-leads-to-satisfaction perspective. The
“appraisal -—) emotional response —> coping” framework applied by Gotlieb, Grewal and
Brown (1994) is typical of this school of thought. It holds that an individual’s appraisal of
the quality of a product is followed by an affective response, satisfaction. Ultimately, this
affective response produces coping responses such as behavioral intentions toward the
product.

Boulding et al. (1993) explained the link between transaction-specific and global
evaluations using a dynamic process model of service quality. The model showed that
transaction-specific evaluations of a service contributed, on a cumulative basis, to the
overall assessment of service quality.3 Considering that satisfaction is also transaction-
specific, the argument could be made that CS/D adds to a cumulative evaluation of the
product. While it can be argued that the term quality still applies, it may be more
appropriate to use a different label — if only, to distinguish between the immediate
evaluation of quality and its enduring counterpart.

There is general agreement that higher levels of perceived quality and satisfaction
should positively influence future intentions toward a product. How this process actually

works has yet to be fully explained. Compared to the body of work on the antecedents of

 

3 Boulding et al. (1993) also noted that the customer satisfaction as understood in the popular media, as
well as by many corporations, often connoted a cumulative concept. In contrast, the satisfaction measure
favored in most academic studies tended to be transaction specific.

23

both consumer satisfaction and perceived quality, research on the consequences of
satisfaction has been relatively limited.

Barsky (1992) conducted a survey of 284 guests of a large San Francisco hotel.
Customer satisfaction scores were computed from their responses to questions about their
expectations of nine classes of hotel attributes as well as the relative importance of each.
These scores were then transformed into ordinal data by establishing three categories of
satisfaction scores: highly satisfied, satisfied, and not satisfied. Chi-squared analysis
indicated that satisfaction is significantly related to repeat purchase intention (x2 = 7.850,
df= 2; p = .020), i.e. satisfied and highly satisfied guests also indicated a willingness to
return to the hotel.

However, he stressed that this test indicated a relation to repeat purchase
intention, not actual purchase. He recommended that the effect of time on purchase
behavior be investigated further, suggesting that expectations may change over time.

Boulding et al. (1993) tested their model of service quality using data from two
different studies of service quality. The first study involved simulated visits to a hotel by
107 business professionals. The other study was based on a service quality survey of an
educational institution’s current customers. Given the different service settings, the two
studies had different measures for quality and behavioral intentions. In the hotel
experiment, the subjects were asked about their likelihood to stay at the hotel again, and
to recommend the hotel to friends. The school study, on the other hand, had six items of
intended behaviors. These included intentions to say positive things about the school to
other people, to contribute money after graduation, and to recommend the school as a

place to recruit. Despite the different data collection methods and service settings, their

24

analysis showed that overall perceived quality significantly related to the index of
behavioral intentions in both cases.

Danaher and Arweiler (1996) interviewed 189 tourists about their New Zealand
vacation at the airport, just before they departed. Each respondent was asked about four
primary components of their vacation: transportation, accommodation, activities and
attractions. Each component, in turn, was further divided into subcomponents. For
example, the subcomponents of accommodations were hotel, motel, backpacker/YMCA,
and relatives/fiiends home.

Satisfaction with the subcomponents was measured with a three-point
disconfirmation scale (worse than expected, about as expected, better than expected).
After the subcomponents were evaluated, satisfaction with their related component was
gauged on a “very dissatisfied-to-very satisfied” Il-point scale. After all the components
were rated in this fashion, respondents were asked about their overall satisfaction with the
entire vacation and the likelihood of recommending New Zealand as a holiday destination
to family and friends.

Regression analysis indicated that tourist activities (e.g., hiking, white water
rafting, etc.) had the strongest impact on overall satisfaction, accounting for almost one-
third of the explainable variation. Accommodation and attractions each accounted for a
little over one-fourth of the variation. As for the likelihood of recommending, activities
was the only component that had a significant impact, accounting for almost eighty
percent of the variation.

The R2-values for both regressions were relatively low (R2 = 0.10 for overall

satisfaction and R2 = 0.065 for recommending). They speculated that this was likely due

25

to lack of variation in both dependent and independent variables, noting that satisfaction
was very high at both the component and overall levels. Arguing that the four
components had been identified factors in previous research, they believed it unlikely that
an omitted critical component was the reason for the low RZ-values.

Although, he did not publish any subsequent follow-up, Mazursky (1989) wrote a
very interesting article on the relationships between several post-experience variables:
perceived performance, disconfirmation of expectations, satisfaction and future
intentions. It was interesting not only for its findings but also because the mode of
analysis allowed him to illustrate the relations-hips between these constructs in a relatively
novel fashion. By plotting their relative positions on a two-dimensional space, Mazursky
submitted very intriguing propositions about the way that these constructs did or did not
interact with each other.

His article was based on an exploratory survey of two hundred (200) adult tourists
visiting a stalactite cave. A two-page questionnaire administered at arrival recorded their
expectations about the tour, as well as their past experience and norms regarding visits to
stalactite and other types of caves, and attitudes toward nature reserves. Another
questionnaire, given upon completion of the tour, addressed the post-exposure measures
(perceived performance, disconfirmation of expectations, satisfaction, and future
intentions).

The pre- and post-exposure data were then subjected to a Weighted Smallest
Space Analysis (WSSAI) which allowed the variables to be represented on a two-
dimensional plot. It was observed that the variables associated with the traditional

disconfirmation paradigm (expectations, perceived performance, disconfirmation and

26

satisfaction) were located along a horizontal line within a fairly narrow range. Mazursky
described this line as a simplex structure, which means that the variables in it have a
degree of similarity such that each is more like its neighbor than those further down the
line.

Another set of variables formed a vertical structure, also plotting within a
relatively narrow range. These other variables included measures of norms, past
experience, and future intentions (to visit and to recommend). Perceived performance and
satisfaction variables also fell within this vertical structure. In other words, these two
dimensions represented the intersection of the horizontal and vertical simplex structures.
One important conclusion from this was that behavioral intentions were not fully
mediated by satisfaction derived from a single experience. The vertical structure
indicated that intentions were also related directly to past behavior and norms.

A hierarchical regression analysis on the pre- and post-exposure measures
provided firrther insight into these mechanisms. It indicated that satisfaction was solely
determined by expectations, performance, and disconfirmation (R2 = 0.48). None of the
past-experience and norms measures significantly affected satisfaction. This appears to
provide support for the conception of consumer satisfaction as a transaction-specific
construct.

The analysis also indicated that previous experience and normative standards
about caves — along with expectations, perceived performance and satisfaction (but not
disconfirmation) -— significantly influenced future intentions to visit and to recommend

(R2 = 0.51). This further supports Mazursky’s thesis that the experience-based approach

27

enriches the understanding of satisfaction beyond that which would be developed fiom a
strictly disconfirmation-based approach.

One noticeable limitation of the current satisfaction-quality literature has been the
lack of studies that have used both perceived quality and satisfaction to predict future
intentions. Cronin and Taylor (1992) were among the first to incorporate both constructs
in a single research framework. To summarize their model briefly, they proposed that
customer satisfaction was an antecedent of perceived service quality and that both
constructs would have significant impacts on purchase intentions. They then tried to
apply this model to four types of service firms: banks, pest control, dry cleaning and fast
food.

They found that service quality had a significant (p = .05) effect on consumer
satisfaction in all four samples. However, the direction of this effect was the opposite of
what they initially proposed. Furthermore, the results from all four samples confirmed
that consumer satisfaction had a significant effect on purchase intentions. However, the
effect of service quality on intentions was statistically significant in only three of the four
industries (banking, pest control and fast food). They concluded that satisfaction had a
stronger and more consistent effect on purchase intentions than did service quality.

Using data collected fiom visitors to a festival, Baker and Crompton (2000)
developed a structural equations model using operationalized constructs similar to those
from the Cronin and Taylor study. Quality was operationalized using a perceptions
measure while satisfaction was measured with a single item (very unsatisfied - very

satisfied) scale. Behavioral intentions were measured with seven items which were

28

assigned a priori to two domains: loyalty (five items) and willingness-to-pay more (two
items).

As with Cronin and Taylor, they found quality to have a significant direct effect
on satisfaction. They found that satisfaction (.60, t=9.95, p<.01) had a stronger direct
efl’ect on visitors’ behavioral intentions than quality (.41, t=6.37, p<.01). However, the
total effect of quality on intentions was greater than the total effect from satisfaction.
Furthermore, the indirect effect of quality was not fully mediated by their level of
satisfaction.

Murphy, Pritchard and Smith (2000) developed a conceptual model of a
destination’s environment and infrastructure relationships to perceived quality, price and
intent. The model was tested using a secondary dataset consisting of 3,088 surveys
collected fi'om visitors to Victoria, British Columbia in 1994. Sixteen items fiom the
survey were used as multiple indicators for the five key constructs of their model, with
each item being evaluated on a five point, strongly-agree—to-strongly-disagree Likert
scale.

They analyzed their data using partial least squares (PLS) analysis. The model
explained twelve percent (12%) of the variance in “intention to return”, as well as
twenty-five (25%) and thirty-seven percent (3 7%) of the variance in quality and value
respectively. They also found that quality ([3 = 0.30), but not value (B = 0.08), had a
significant effect on intention to return. As the objective of PLS is prediction in a
regression sense, they did not report an overall goodness-of-fit index. Nevertheless, the
results of their study provided additional evidence of the relationship between quality and

intention to return.

29

One notable aspect of their model was that their quality construct combined
assessments of both quality and satisfaction. The construct had two indicators. One was a
statement regarding Victoria’s “quality relative to the United States”. The second quality
indicator was a statement regarding overall satisfaction with Victoria as a destination. In
addition, two constructs — environment and infrastructure — also appear to correspond
to perceived quality, albeit at the component level (e. g., pleasant climate, attractive
scenery, good hotels, etc).

Ifone accepts the constructs as labeled, the model appears to hold that destination
quality is evaluated at both the component and global levels. However, the inclusion of
overall satisfaction as an indicator of global quality suggests that this particular construct
can be likened to an attitude in that apparently cognitive (quality) and affective

(satisfaction) elements are both present.

Another issue that has not been explored extensively has to do with the effects of
nationality and similar variables on satisfaction and behavioral intentions. The
investigator found only one study that specifically addressed this issue. Komba (1997)
investigated the effects of national characteristics on the satisfaction of tourists with
hotels in Tanzania by comparing two international groups (American and German) with
domestic tourists from Tanzania. A total of 224 tourists (70 American, 83 German, and
71 Tanzanian) were surveyed at Tanzanian hotels during two separate survey periods in
1994 and 1995. Using a log linear model, he sought to determine whether, given the
satisfaction levels with various aspects of a hotel’s service, the likelihood of certain

behavioral intentions varied from one national group to the other.

30

The three groups were fairly similar in terms of their criteria for selecting and
evaluating hotels. For example, all three groups said that they considered the prices,
fiiendliness of tourists, safety of property and cleanliness of the hotel in evaluating a
hotel. However, a pairwise comparison of the groups showed that the American and
German markets tended to behave similarly compared to the Tanzanian market. That is,
the most differences were observed between the American and Tanzanian segments,
while the least differences were detected between the American and German groups. This
was consistent with Komba’s expectation that the economic, political and cultural
similarities between the two international groups would translate to relatively similar
consumer behavior.

With regard to future intentions, the results generally supported the thesis that
greater satisfaction leads to a greater likelihood of complimenting behavior such as
recommending to friends and repeat visitation. However, Komba also found that
dissatisfaction was not often followed by a greater likelihood of complaining behavior. In
other words, the impact of satisfaction on future intentions was greater than the
corresponding impact of dissatisfaction.

Komba explained that cultural and practical considerations could affect the
intentions to engage in complaining behavior. Dissatisfied Tanzanians, for example, may
complain less because they attribute certain failures to external conditions in the country
that are beyond control of the hotels. On the other hand, international tourists may believe
that the brevity of their stay in the country lowers the probability of a successful
complaint. Thus, they may choose not to complain unless the dissatisfaction involves a

factor deemed extremely important such as safety.

31

Summary and Synthesis

This review has highlighted at least three consistent themes in the consumer
satisfaction - service quality literature. First, it is apparent that the disconfirrnation of
expectations mode is still the dominant paradigm for examining either consumer
satisfaction or perceptions of product quality. While there has been some discussion as to
the nature of the comparison standard, i.e. expectations, it now appears that researchers
are now more open to the idea of multiple standards. In other words, not only are belief
expectations used to evaluate performance, but norms, desires and other potential bases
for comparison may also be used in conjunction with prior beliefs.

Despite the dominance of the discrepancy framework, comparative studies of
alternative measures have consistently shown stronger predictive validity for direct
measures of satisfaction and quality. Nevertheless, the disconfirmation model still holds a
lot of appeal among researchers and managers.

One could probably explain this as simply a disciplinary resistance to change,
given how entrenched the model and its associated instruments (especially, SERVQUAL)
are in research and trade circles. Yokoyama (1991), for example, appears hesitant to
totally forego disconfirmation despite findings that a direct performance model proved
better than a discrepancy-based alternative. He rationalizes his findings by surmising that
perhaps disconfirmation of expectations may already be incorporated into the responses
to the direct questions about satisfaction and quality.

Crompton and Love (1995) acknowledged the superior predictive validity of

direct measures of quality. However, they held that the disconfirmation model offered

32

managers greater diagnostic utility. Thus, the relative inefficiency of the framework is but
a small price to pay for the richness of information that it offers.

Another key element of the literature is that satisfaction and perceived quality are
related but distinct constructs. However, in various trade and research publications, it is
not uncommon for writers to confirse one construct for the other, freely using satisfaction
for quality, and vice versa. This confusion stems in part from the use of virtually similar
instruments to measure both constructs. In response, researchers have meticulously
argued that satisfaction and quality can be differentiated in terms of “would”, “should”
and “could” expectations. Unfortunately, the semantic differences do not appear to be
clear or obvious enough for people to immediately agree on which construct means what.

Another possible explanation for the frequent misapplication of the terms, is that
“satisfaction” and “quality” have been used to refer to transaction-level evaluations as
well as to enduring, overall assessments of products and services. This results in
different, sometimes contradictory, theories of the causal relationships between CS and
SQ. However, in many cases, a more careful reading would reveal that these theoretical
differences are likely to be due to differences in level of specificity.

Finally, another consistent thread in the literature has been the proposition that
satisfaction and perceived quality affect future intentions, as suggested by the marketing
concept. For many managers, satisfaction studies should ultimately be designed to help
them identify future courses of action that would increase the probability of repeat
purchases. Much of the research has so far been focused on the independent effects of
satisfaction or quality on future intentions. Only recently, have researchers moved toward

models involving both constructs. There still is no consensus as to which construct has a

33

stronger effect on intentions, thus there still is some debate on whether managers will be
better served by focusing on satisfaction over quality, or vice versa.

Looking specifically at tourist satisfaction research, this area appears to be ripe
for more in-depth study. To date, the tourism literature has been built largely on studies
involving the elements that make up or lead to satisfaction or perceptions of quality.
There still appears to be a lot of area to cover in terms of furthering our understanding of
the consequences of this satisfaction.

Although researchers have acknowledged the relationship between tourist
satisfaction and future behavior, there has not been a lot of attention given to the unique
characteristics of the tourist purchase. Researchers should consider that these purchases,
especially those involving long-haul holidays, are highly episodic and infrequent. Thus,
the satisfaction-to-behavior link may not be as clear as it might be for an ordinary
consumer transaction that involves frequent purchases of products.

The way that most researchers have diagrammed the process, satisfaction or
perceived quality directly affects intention. Given the length of time between purchases
or visits, it may be appropriate to incorporate an attitudinal change component into the
equation. As tourists learn from their experience with a product or destination, their
attitudes are inevitably modified (Yi, 1990). It is reasonable to suggest that this new,
modified attitude will influence the causal relationship between tourists’ immediate
satisfaction and their longer term intentions.

Another gap in the literature appears to be that related to group comparisons. As
stated earlier, this investigator could only find one study that specifically addressed issues

of nationality as related to visitor satisfaction and behavioral intentions. Studies such as

34

that by Mazursky (1989) also suggest that visitor experience is a key mechanism in
decisions to return or to recommend to friends. Thus, experience may be another viable
grouping variable in future comparative studies. Another form of group comparison that
has not received much attention in the tourist satisfaction literature has been that based on
purpose of travel. Given the documented importance of motivation in tourist decisions, it
is striking that this aspect of tourist satisfaction research has not been explored

extensively.

3S

CHAPTER IH

PROCEDURES

This chapter is divided into three parts: (1) a description of the data set to be used,
(2) the research models and hypotheses tested, and (3) the statistical procedures and
techniques applied to the data set. This dissertation relies on secondary data — data
previously collected for another study. Thus, the first part of this chapter covers the
source of this data, including an overview of the methods used to collect the data
originally. The specific variables from the dataset to be utilized in the study will be
identified here. A profile of the subjects included in the study sample will also be
included in this section. The second major section of the chapter then discusses the
research models and hypotheses to be tested. Definitions of the various elements
(constructs and measures) of the models will be presented by relating these elements to
specific variables from the data set. Finally, the statistical procedures for testing the
hypotheses will be outlined. Because structural equations modeling (SEM) is the major
procedural component of the study, a short primer on SEM and the EQS software

package will be included in this section.

36

Overview of Data Set

Primag Data Source and Data Collection Methods

This dissertation was based on secondary data — data previously collected by the
Guam Visitors Bureau (GVB). GVB has been conducting these surveys on a quarterly
basis since December 1991. Initially, the survey was administered only to visitors fiom
Japan, Guam’s most important source of international traffic. Thus, these first surveys
were written in the Japanese language. In 1994, recognizing the growing importance of
other markets, the GVB began a similar series of Korean and Chinese language surveys.
The specific objectives of the survey are to:

I provide accurate information regarding the demographic characteristics
of visitors to Guam;

I provide accurate information regarding the motivations of these visitors,
as well as their specific purpose for their current trip;

I provide accurate information on visitor activities — what they did on

Guam;
I provide accurate information on satiyaction ratings for a wide range of
activities and various components of Guam '3 visitor industry

infrastructure; and

I to collect and analyze expenditure data. (Iverson, 199 7)

The data were collected through exit surveys of international tourists at Guam
International Airport using research methodology developed by the GVB’s Research
Committee and Market Research and Development, Inc., a Guam-based market research
firm. Surveys are collected during a one-week period within a quarter. The sampling

period is moved to different months within a quarter from year-to-year. The specific

37

week within the month in which the survey is conducted has varied due to staff,
schedule- and administration-related issues that occasionally have arisen.

All flights departing for Japan, Taiwan, Hong Kong and Korea during the
sampling period were surveyed. Every nth passenger (with “n ” being a randomly chosen
number) coming through security for a particular flight was asked to participate in the
survey. Based on the airline loads for the week, the “n” was adjusted to allow for a
reasonable number of respondents per day of the week. Each day could thus be treated as
a panel within the one-week survey period. Thus, the sampling procedure is properly
described as a stratified pseudo-random methodology.

Potential respondents were screened to see if they: 1) had stayed on Guam
overnight, 2) were Japanese, Korean, Taiwanese or Hong Kong citizens and 3) were over
the age of eighteen. Those that qualified were then asked to participate in the survey.
Participants filled out the survey at a nearby table. When turning in the completed survey,
participants were asked if they needed further clarification on any item in the survey.

Approximately eighty percent (80%) of the individuals encountered met the
criteria, with relatively few individuals refusing to participate. Screeners were asked to
identify important group patterns among the non-respondents. However, no such patterns

were found. No further documentation regarding the data collection procedures can be

offered in this dissertation given the secondary nature of the data.

38

Sample Characteristics

For this dissertation, a sample of 3,108 cases was randomly-generated from
GVB’s database. This sample represents seventy-five percent (75%) of all interviews
conducted by the GVB during the period from June 1998 to November 1999. The sample
also represents a cross-sectional view of Guam’s most important sources of international
tourist traffic: Japan, South Korea, Taiwan and Hong Kong.

A little more than half of the sample (52.4%) is made up of Japanese respondents.
Koreans make up the second largest national group (21.6 %) in the sample. Respondents
from Taiwan and Hong Kong accounted for approximately fifteen (15.1 %) and seven
percent (7.1 %) respectively. Given the focus on international arrivals, the United States
was not specifically targeted for the exit survey. Nevertheless, a small number of
respondents from the United States were included in the survey for purposes of
comparison.

As shown in Table 1, the various national groups differed in their composition
according to gender. Overall, the sample was almost equally split by gender, with females
forming a slight majority (51.5 %). The Japanese, Korean and Taiwanese groups tended
to follow this general pattern, with the Korean group having an almost equal number of
males (333) to females (338). Females accounted for almost fifty-five percent of both the
Japanese and Taiwanese segments. In contrast, there were more males than females fiom

the Hong Kong and US sub-groups.

39

Table 1. Distribution of Respondents by Nationality and by Gender

 

 

 

 

 

 

 

 

 

 

 

 

No. of Males No. of Females Total
Nationality (% within (% within (% by
National National Nationality)
Group) Group)
Hong Kong 130 (58.8) 91 (41.2) 221 ( 7.1)
Japan 737 (45.3) 891 (54.7) 1,628 ( 52.4
Korea 333 (49.6) 338 (50.4) 671 ( 21.6)
Taiwan 211 (45.0) 258 (55.0) 469( 15.1)
United States 97 (82.2) 21 (17.8) 118 ( 3.8
Total (% by Gender) 1,508 (48.5) 1,599 (51.5) 3,107 (100.0)

 

When examined as a whole, the sample can be described as relatively young, with
half of the respondents below thirty years of age. Including people in the 30 to 39 year
bracket, eighty percent of the sample was less than forty (40) years of age. Only seven
percent (7.4%, to be exact) of the respondents was at least 50 years old.

As with gender, the national groups also differed in terms of their distribution by
age groups (see Table 2). Compared to the other groups, the Japanese segment had a
distinctly younger profile as almost sixty percent (59.7%) fell into the youngest age
bracket. In addition, almost one-fourth (24.0%) of the Japanese respondents were
between 30 to 39 years of age. Thus, more than eighty percent (83.7%) of this segment
was younger than forty years.

As with the Japanese market, close to eighty percent of the Korean and Taiwanese
markets also fell within the 18 to 39 year age range. However, for these two markets,
there was a more equitable split between below-30 and over-30 year old respondents.

In contrast, the Hong Kong and American segments exhibited slightly older age
profiles. Respondents from the ages of 30 to 39 made up the largest grouping from Hong

Kong, accounting for almost forty percent (40.3%) of the segment. Furthermore, twenty-

40

two percent (22.2%) of the Hong Kong group was made up of people from 40 to 49 years
of age. In the American sub-sample, 40 to 49 year olds made up the largest grouping
accounting for some thirty percent (30.2%) of all responses. In fact, more than sixty

percent (61.2%) of the American segment was reported to be forty years or older.

Table 2. Distribution of Respondents by Nationality and by Age Group

 

 

 

 

 

 

 

 

 

 

 

Nationality Age Grogas (% within National Grou s) Total (% by
18-29 30-39 40-49 50-59 60 & up Nationality)
Hong Kong 64 89 49 10 9 221
(29.0) (40.3) (22.2) (4.5) (4.1) (7.2)
Japan 968 389 158 87 30 1,622
(59.7) (24.0) (9.7) (5.4) (1.2) (52.4)
Korea , 312 249 69 26 13 669
(46.6) (37.2) (10.3) (3.9) (1.9) (21.6)
Taiwan 1 86 1 77 73 22 7 465
(40.0) (38.1) (15.7) (4.7) (1.5) (15.0)
United States 17 28 35 23 13 116
(14.7) (24.1) (30.2) (19.8) (11.2) (3.8)
Total (% by 1,547 932 384 168 62 3,093
Age) (50.0) (30.1) (12.4) (5.4) (2.0) (100.0)

 

 

 

 

 

 

Table 3 shows the distribution of respondents, whether they traveled alone or with
companions. Less than one-tenth (7.7%) traveled alone, as almost ninety percent (91.2%)
reported traveling with family members or friends (including office mates) to Guam.

Only a very small percentage of Japanese visitors (1.4%) traveled to Guam
without any travel companions. The majority of Japanese respondents (61.5%) said that
they came with friends, while thirty-five percent said that they were accompanied by their
spouses or other family members.

In contrast, family members were the most common travel companions for
respondents from Hong Kong (53.4%), Korea (59.6%) and Taiwan (57.1%). However,

the proportion of people traveling alone from Hong Kong (14.9%) and Korea (13.0%)

41

was much higher than that for Taiwan (5.1%). In this regard, the American segment
displayed the most atypical profile with a little more than sixty percent (61.5%) traveling

without any travel companions.

Table 3. Distribution of Respondents by Nationality and by
Travel Companion First Mentioned

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Nationality Travel Companion Total (% by
Alone Family Friends Other Nationality)
33 1 18 64 6 221
Hong Kong (14.9) (53.4) (29.0) (2.7) (7.1)
23 580 997 22 1,622
Japan (1.4) (35.8) (61.5) (1.4) (52.3)
87 400 180 4 671
Korea (13.0) (59.6) (26.8L (0.6) Q17)
24 267 177 0 468
Taiwan (5.1) (57.1) (37.8) (0.0) (15.1)
72 16 26 3 1 17
United States (61.5) (13.3 (22.2) (2.6) (3.8)
Total (% by 239 1,381 1,444 35 3,099
ompanion) (7.7) (44.6) (46.6) (1 . 1) (100.0)

 

 

The composition of travel parties from each country of origin may be related to
differences in purpose of travel by nationality. As shown in Table 4, it appears that Guam
is primarily seen as a’pleasure travel destination. Almost sixty-five percent (64.4%) of all
respondents cited "pleasure" as their primary purpose for visiting Guam. The profile of
American respondents is conspicuously different from the other national groups in that
only 3 out of 117 (2.6%) individuals claimed to be traveling for pleasure.

Guam appears to be an important honeymoon destination for the Korean market,
as one-fourth of all Korean respondents (25.3%) gave this response. This statistic is
consistent with the relatively high proportion of family companions reported for the

Korean segment in Table 3. Of the four hundred respondents who had traveled with

42

family members, almost three-fourths (71.4%) said that they were traveling with their
spouses.

The responses fiom Hong Kong, Japan and Taiwanese tourists appear to be
similar in that pleasure travelers accounted for at least seventy percent of each of these
national groups. However, it is also noticeable that a relatively large number of Japanese
(19.6%) chose "company trip" as their main purpose of travel. Again, this is consistent
with findings in Table 3, which showed a higher proportion of friends in the Japanese

travel party. This may be a reflection of the importance of "Office Ladies" and other

work-related segments within the Japanese market.

Table 4. Distribution of Respondents by Nationality and by Purpose of Travel

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Purpose of Travel Total
Honey- Comp. Visit (% by
moon Pleas. Bus. Trip Friends Other Nation)
ong Kong 5 174 25 5 8 4 221
(2.3) (78E (11.3) (2.3) (3.6) (1.8) (7. 1)
Japan 88 1,148 11 315 13 51 1,626
(5.4) (70.6) (0.7) (19.4) (0.8) (3.1) (52.5)
orea 169 326 49 18 74 32 668
(25.3) (48.8) (7.3) 12.7) (11.1) (4.8) (21.5
Taiwan 52 345 25 40 6 0 468
(11.1) (73.7) (5.3) (8.5) (1.3) (0.0) (15.1)
United States 14 3 2 0 0 98 117
(12.0) (2.6) (1.7) (0.0) (0.0) (83.8) (3.8)
Total 328 1,996 112 378 101 185 3,100
(% by Purpose) (10.6) 64.4) (3.6) (12.2) (3.3) (3.4) (100.0)

 

Limitations of the Da_1t_a

 

 

In undertaking the analysis of the Guam Visitor Bureau’s survey data set, the
researcher recognizes that the original purposes for collecting this information were

different fi'om those of this current study. The researcher thus acknowledges the potential

43

for invalidity. According to Kiecolt and Nathan (1985) this is the major drawback to
secondary analysis of existing survey data: “Surveys rarely contain all the variables of
interest to the secondary researcher, and even when they do there may be too few
indicators of a concept for reliable measurement” (p. 14).

In the GVB’s questionnaire, there were more variables associated with
component-level satisfaction than any other construct in the proposed model. Fewer
variables appeared to be associated with constructs such as destination quality, image and
behavioral intentions. This required that assumptions be made regarding which variables
in the data set best corresponded to these constructs. In the absence of explicit measures
for this study’s constructs, variables in the secondary data set were utilized as proxies for
the missing variables.

How well this approach yields good measures of the central concepts of this
dissertation can be evaluated with covariance analytic techniques afforded by some
specific software packages. Kiecolt and Nathan (1985) identified the strengths of
programs like LISREL and EQS. First, these programs allow the researcher to identify
items that are poor measures of a construct, and which ones are central to the construct.
Furthermore, by allowing constructs to be correlated, these techniques point to potential
problems of collinearity if the correlated constructs are used as independent variables.
Finally, covariance analytic techniques correct for the measurement error associated with
many surveys. The procedures for conducting these analyses will be discussed in detail

later in the chapter.

44

Research Models and Hypotheses

Elements of the Model.

The key constructs of the model are satisfaction, destination quality, image and
behavioral intentions. Destination quality is defined as the tourist’s overall evaluation of
the excellence or inferiority of a destination’s attributes. Although it is very likely that
tourists do employ a disconfirmation-like process in order to determine destination
quality, the definition used here is analogous to direct measures of performance or quality
such as Cronin and Taylor’s (1992) SERVPERF. In this study, four variables will be used
as indicators of destination quality: 1) cleanliness, 2) safety, 3) communication, and 4)
ease of travel within the destination.

Satisfaction refers to the tourists’ assessment of their response to the performance
of tourist services (in the case of component satisfaction) or the destination as a whole (in
the case of overall satisfaction). This assessment is transactional in nature in that it refers
only to a specific encounter with the product or service.

Satisfaction is measured at two levels in this study. At the component level,
tourists were asked how satisfied or dissatisfied they were with 1) their accommodations,
2) dining in Guam, 3) the shopping establishments they visited, and 3) the various tourist
attractions or activities that they experienced during their stay. Hotel satisfaction was
measured in terms of satisfaction with eight attributes of hotel service, both tangible (e.g.,
facilities) and intangible (e.g., staff service). Dining satisfaction did not refer to a single
dining establishment but to a general evaluation of Guam restaurants as a whole based on

five attributes of dining service.

45

In contrast to the hotels, satisfaction with shopping establishments and with
tourist attractions were not rated on specific product attributes. Instead 16 shopping
establishments were identified by name and satisfaction for each store was then solicited.
Similarly, the survey did not attempt to break down 17 different attractions into particular
service attributes. Instead each type of attraction was rated on an overall basis.

However, it is highly unlikely that tourists would be able to form satisfaction
ratings for all shops and attractions in the list because very few (if any) would have been
able to experience every single one. Thus, the number of items a tourist would respond to
would vary by the number of shopping establishments and tourist attractions he or she
actually visited. To allow for useful comparisons across the sample, mean satisfaction
ratings for each respondent were computed based on the number of shops and attraction
items that they had valid responses for. The assumption is tourists would not be able to
rate their satisfaction with a shop if they did not have any experience with the shop to
begin with.

Aside from component-level satisfaction, the proposed model also includes an
overall satisfaction construct. The data set does not include any variable that specifically
asks about satisfaction. However, there is one specific item which asks whether “Guam
was better than expected.” This statement appears to be reflective of disconfirmation, a
recognized antecedent of satisfaction. Another item in the survey asks whether or not
"Guam was enjoyable." These two variables echo Hunt’s (1977) conception of
satisfaction as a quasi-cognitive construct in that it is an evaluation of whether the

experience was at least as good as it was supposed to be as well as an evaluation of an

46

emotion. For this reason, the "disconfirmation" and "enjoyable" variables will be used as
indicators of the overall satisfaction construct.

Image refers to the mental picture that tourists hold about a destination or product.
In contrast to satisfaction, image is not limited to a specific past experience but is also
assumed to be an enduring evaluation that can carry over to the next purchase decision
and experience. For this study, there is only one measure for the image construct. The
item asked respondents to compare Guam to other international sites on a seven-point
”least desirable to most desirable" scale. Measured in this manner, image is arguably an
attitude in that it is a disposition to respond favorably or unfavorably toward the object
(Ajzen, 1988) and that it represents an enduring positive or negative feeling about the
destination (Petty & Cacioppo, 1981).

Intentions are indications of the willingness and commitment of tourists to
perform particular behaviors or actions. The behavioral intentions to be considered here
involve 1) returning to the destination and 2) recommending Guam to friends and
relatives back home. Table 5 summarizes the key elements of the model and the specific

variables to be used as their indicators.

47

Table 5. Constructs of Interest and their Associated Measures

 

 

 

 

 

 

 

 

 

 

 

 

 

Construct Variables How the Variables were Measured
(# of Measures)
Hotel Satisfaction Hotel rooms 1 = Very dissatisfied
(8) Hotel view 4 = So-so
Hotel furnishings 7 = Very satisfied
Hotel cleanliness
Hotel food
Hotel staff friendliness
Hotel service
Hotel value for money
Dining Satisfaction Dining cleanliness 1 = Very dissatisfied
(5) Dining food quality 4 = So-so
Dining staff friendliness 7 = Very satisfied
Dining service
Dining value for money
Shopping 16 specific shopping 1 = Very dissatisfied
Satisfaction (2) establishments plus 1 "Other. . . " 4 = 80-50
7 = Very satisfied;
A mean shopping satisfaction score was
computed for each respondent by summing up
all the ratings and then dividing by the number
of items with valid responses.
“Shoppingjrljoyable or not” 1 = not enjgyable; 7 = enjoyable
Satisfaction with Beach resorts; Dinner cruises, 1 = Very dissatisfied
Attractions (1) Dinner shows; Dog races; 4 = 80-50
Fishing, Golf, Gun shooting; 7 = Very satisfied;
Helicopter tour; Jet skiing, Mini-
golf; Museums; Nature tours; A mean score for attractions was computed for
Para-sailing; Scuba diving, each respondent by summing up all the ratings
Sightseeing; Theme restaurants; and then dividing by the number of items with
Water parks. valid responses.
Destination Quality ”Guam clean or not" All rated on a 7-point scale.
(4) "Guam was safe or not”
"No problems with Cleanliness: l = unclean; 7 = clean
communication"
”Going around was easy” Safety: 1 = unsafe; 7 = safe
Communication and Easy:
1 = strongly disagree; 7 = strongly agE
Overall Satisfaction ”Guam was better than expected.” Discontinuation:
(2) I = strongly disagree; 7 = strongly agree
”Guam was enjoyable or not."
(Enjoyable) Enjoyable:
l = not enjoyable; 7 = erjoyable
lmgge (1) "Rate Guam as a destination" 1 = least desirable; 7 = most desirable
Intentions (2) ”Plan to visit Guam again" Both rated on 7 pt scale where:
"Will recommend Guam." 1 = strongly disagree; 7 = strongly am

 

48

Relationships of the Elements

Hoyle (1995) defines a model as "a statistical statement about the relations among
variables." Figure 2 thus depicts the hypothesized relationships between the various
constructs and measures introduced in the preceding section.

The model suggests that tourists evaluate and derive satisfaction from the various
components of a destination. Thus, the model has four exogenous factors corresponding
to the tourists' satisfaction with Guam's hotels; dining establishments; shopping and
attractions. It is hypothesized that each of these component satisfactions will have direct
effects on perceived quality of the destination, and on overall satisfaction with the
destination. Furthermore, destination quality is envisioned to be an antecedent of overall
satisfaction.

After tourists have experienced a destination, the model suggests that tourists'
overall satisfaction and their perception of the destination's overall quality lead to the
formation or modification of their image of Guam. Finally, this post-experience image
influences the tourists‘ intentions toward Guam in terms of recommending Guam or
planning to return to Guam. In other words, the effects of perceived destination quality

and overall satisfaction on tourist intentions are mediated by the new or modified image.

49

  

22885
233m

 
 

 

  

couoémsam
826822

  

cozoémuam
=So>O

  
     

:o roam tam
wcaqonm

own E—

5 team cam

 

  

3:20
555600

  

cozoamflsm

.26:

  

 

re 2. H othesized Relationshi s between Constructs

Fi

50

Hypotheses

Figure 2 is, in effect, the visual representation for this dissertation's first
hypothesis regarding the process involved in post-visit evaluations. Stated as a formal
research statement, it can be written as:

ijothesis 1. The proposed model in Figure 2 will fit the data perfectly.

Thus, Figure 2 illustrates a hypothesis about the entire set of variables, constructs
and relations that make up the model. To better understand how the model works,
hypotheses about specific relationships in the model must also be proposed and tested.

Hypothesis 2 presupposes that, in general, tourist activities or attractions offered
at a destination will more accurately reflect the tourists’ motives for selecting Guam than
do its hotels, restaurants or shopping establishments. In most tourist decisions, attractions
should function as pull factors in the manner described by tourism motivation researchers
(Witt & Wright, 1992; Dann, 1977; Crompton, 1979). Thus, it is feasible that attractions
will have a larger impact on overall satisfaction than will the hotels and shops of Guam.
However, this may not necessarily be true for perceived destination quality because it is a
more objective evaluation which does not give as much import to personal motives and
needs.

Related to this, the study’s definitions of the two constructs also assume that
emotional responses to the destination are incorporated into overall satisfaction rather
than the quality evaluation. For this reason, overall satisfaction is believed to play a
bigger role in attitude formation than destination quality. Thus, the second and third

hypotheses can be stated as follows:

51

ijothesis 2. Satisfaction with activities and attractions will have a stronger
direct eflect on overall satisfaction than satisfaction with any of the other
components.

Hwothesis 3. Overall satisfaction will have a stronger direct effect on image
than will destination quality.

Hypothesis 4 is based on the literature on attitude and behavior which recognizes
that the link between the two is less than perfect. Factors such as the amount of time
between attitude and behavior as well as perceived behavioral control (Ajzen, 1985/1988;
Ajzen & Driver, 1992) can affect the predictive power of attitude on behavior. For many
tourists, the decision to return to Guam is likely to involve a fairly long period of time
between visits as well as a recognition of potential constraints on time, money and the
availability of suitable travel companions. In contrast, word-of-mouth recommendations
may be initiated fairly soon after the trip and — in the case of negative word-of mouth —
limited only by personal notions of propriety or courtesy. Thus:

Hwothesis 4. The model will predict the intention to recommend to friends
better than the intention to return to Guam.

The model presented here, as well as the secondary data with which the model
will be tested, involves post-visit evaluations of Guam. However, the literature suggests
that constructs like satisfaction and perceived quality are also determined in large part by
the visitors’ prior expectations. Thus, the final set of hypotheses attempts to examine,
albeit indirectly, the effect of expectations on the post-visit model.

The effects of expectations must be assessed indirectly because the data set used

for this study did not include any expectations variables. However, there are categorical

52

variables such as purpose of travel, nationality, and previous experience (i.e. repeat
visitors vs. first-time tourists) that may be used to find an indirect answer to the problem.

Purpose of travel is interpreted in this study as suggestive of the tourists’
motivations for visiting Guam. Differences in motivation have been found to
subsequently influence customer or visitor satisfaction (Mayer, Johnson, Hu, &
Chen, 1998; Lounsbury & Hoopes, 1985). Similarly, purpose of travel may factor into
Spreng and Mackoy’s (1996) conception of desires congruency.

Komba (1997) also suggests that nationality may influence satisfaction and
behavioral intentions. He argued that economic, political and cultural similarities of
national groups may determine their prior norms and expectations, thus influencing the
subsequent evaluation of the tourist destination or establishment. Prior experience has
also been found to significantly influence visitor satisfaction and intentions (Mazursky,
1989) in much the same manner — by influencing consumer norms and expectations.

Interestingly, visitor bureaus often use these variables (purpose of travel,
nationality, and prior experience) as bases for segmenting their target markets. The
assumption is that the resulting segments are distinct from each other in terms of product
needs and preferences. Differences in the way that promotional programs are directed at
the segments may also affect the type and amount of information received by each.

The approach taken here will be to hypothesize that these variables exert
moderating effects on the various elements of the general model. In other words, if these
segments do differ in terms of expectation, it may be reasonable to expect the model to

perform differently between groups segmented by nationality, purpose or previous

53

experience. The variables can thus be used as virtual proxies for the missing expectation
variables. The specific hypotheses to be tested are listed below:

Hmothesis 5. The structural model will drfler significantly between
nationalities.

Hwothesis 6. The structural model will dtfler significantly between groups with
varying motivations as implied by stated purpose of travel.

Hypothesis 7. The structural model will differ significantly between groups with
varying levels of knowledge about the destination as implied by prior
experience.

Statistical Procedures

Overview of SEM and the EOS progpam.

The major analysis for this dissertation will involve structural equations modeling
(SEM) using the EQS program. Thus, a short introduction to the technique is appropriate
at this time.

Structural equation modeling (SEM) is a comprehensive approach to testing
hypotheses about relations among observed and latent variables (Hoyle, 1995). Published
research articles may also use other names to refer to the family of models covered by
SEM. Alternative labels for the procedure include covariance structure analysis, latent
variable analysis, confirmatory factor analysis or LISREL analysis (after the name of the
most popular SEM software package in the market). According to Hair, Anderson,
Tatham, and Black (1998), all SEM techniques, regardless of their label, are
characterized by:

1) estimation of multiple and interrelated dependence relationships, and

2) representation of unobserved or latent concepts in these relationships,
accounting for measurement error in the process.

54

Hoyle (1995) writes that SEM has at least four major similarities with standard
statistical approaches such as correlation, multiple regression and ANOVA. First, all
these techniques are based on linear statistical models. Second, the statistical tests
associated with all these techniques are valid only if certain assumptions about the
variables are met. For SEM, these assumptions are independence of observations and
multivariate normality. Third, none of these approaches offer statistical tests of causality.
They merely offer necessary, but not sufficient, evidence of causality. Finally, post-hoe
adjustments to the initial hypothesis dramatically increase the likelihood of sample-
specific results. This raises the necessity of cross-validating the model.

However, Hoyle (1995) also points out that SEM differs from the standard
approaches in three important ways. First, any model to be tested must be formally
specified beforehand. Second, only SEM can estimate and test relations between latent
constructs or factors. Hoyle argues that this is probably the most compelling attribute of
SEM.

However, he also notes that the third distinguishing attribute of SEM can be
considered as its relative weakness. Tests of structural equation models are rather
ambiguous, compared to the relatively straightforward tests of the standard models. He
attributes this to the effects of data and model characteristics on the )(2 variate, on which
most SEM fit indices are based. Given this, the recommended approach has been to rely
on multiple fit indices and evaluate the x2 goodness-of-fit test with reference to the power
of the test, given the characteristics of the model as well as the data.

Models constructed through SEM involve two types of variables. Latent variables

are hypothesized and unobserved constructs that can only be approximated by some

55

measured indicator. Latent variables correspond to factors or principal components in
exploratory factor analysis. Manifest variables are the observed variables collected from
respondents through various methods. In this instance, the manifest variables were
gathered as part of a survey.

For purposes of simplicity, this study will rely on the labeling conventions of the
EQS program to distinguish these two variables from each other. Thus, the term
"variable" in any model to be analyzed refers to a measured or observed variable.

Unobserved variables, i.e. latent constructs, in these models will be called "factors".

About EQS

Researchers engaged in SEM can choose between several computer programs for
testing and evaluating their models. The most widely used SEM program appears to
LISREL (Llnear Structural Rflations). Other popular packages include AMOS, CALIS
and EQS (Bentler, 1995). This dissertation is based on analysis using the EQS program.

In general, the results of SEM conducted using EQS and LISREL should be
consistent across the two programs (Byme, 1995). However, there are some differences
between the two programs that may be useful to note.

In contrast to LISREL, EQS does not use Greek symbols and notations to label
the various SEM elements. Rather, it uses the letters V, F, and E to describe variables,
factors, and error terms, respectively. Relationships between variables are written as
standard regression equations, in contrast to the matrix-based approach of LISREL. For
some researchers, EQS's approach facilitates model specification as well as the reporting

of the analysis.

56

More substantively, the EQS approach in correcting the x2 statistic yields more
reasonable results under conditions of nonnormality (Byrne, 1995). In other words, EQS
appears to be more robust to nonnormality than does LISREL. As noted earlier, SEM
assumes that the variables are multivariate normally distributed. However, this may be an
unreasonable assumption, in practice (Micceri, 1989). The EQS approach loosens this
requirement and allows a wider range of data to be analyzed. This feature is particularly
useful in this dissertation because many satisfaction studies tend to have skewed
distributions (Peterson & Wilson, 1992). A more detailed discussion of EQS's approach
to nonnormality and the underlying elliptical theory can be found in Bentler (1995).

SEM Procedures

The application of SEM to this project will be adapted from the procedures
outlined by Reisinger and Turner (1999) and Hair, Anderson, Tatham, and Black, (1998).
This general approach can be described as a six-stage process of structural equations
modeling. These stages are:

Model conceptualization

Model specification

Model identification and parameter estimation
Assessment of model fit

Model modification
Cross-validation of the model

99:59.“?

57

Analysis of the General Model

The model conceptualization stage describes hypothesized relationships between
the latent constructs (i.e. component satisfaction, overall satisfaction, destination quality,
image, and behavioral intentions) and their measured variables or indicators (see Table
5). A theoretical measurement model reflective of these relationships will first be
developed using exploratory factor analysis (EFA) under SPSS. Latent constructs will be
extracted fi'om the dataset using maximum likelihood analysis with oblique rotation — an
approach which closely approximates that of SEM in that the factors are assumed to be
correlated.

The results of the EPA will provide some information regarding the validities of
the constructs to be included in the SEM. Cronbach’s coefficient alpha (0:) will be used
to assess the reliabilities of the scale constructs.

Model specification involves the construction of a path diagram of causal
relationships between the latent factors and converting this diagram into a set of
structural and measurement equations. This path model reflects the hypotheses described
in detail in the preceding section, and illustrated in Figure 2.

Model identification and parameter estimation will then be done using the EQS
software package. Model identification addresses the extent to which the available data
are sufficient to enable parameter estimation. If there are no identification problems, the
parameters of the model will be estimated using the maximum likelihood (ML) method.

The assessment of model fit (for both the measurement and structural models) can
be done using a variety of methods. Applied to the measurement model, the technique

corresponds to a confirmatory factor analysis (CFA). The evaluation of the structural

58

model, on the other hand, addresses the hypothesized relationships between the latent
constructs and their various indicators as described in Hypotheses 1 through 3. As
suggested by the literature, multiple measures of goodness-of-fit will be evaluated.

A model modification stage may be suggested by the results of the assessment of
model fit. That is, Stage 4 may suggest a need to modify the hypothesized model.
Specifically, the standardized residuals and the modification indices (MI) calculated by
EQS will be examined to determine any possible modifications to the model. Such
modifications may be made if they can be theoretically justified, with the aim of
improving model fit. At this point, the researcher acknowledges that the analysis will
become exploratory in nature.

Hair et al. (1995) suggest that the final stage of SEM involve a cross-validation
of the model. This requires that the modified model (or models) be cross-validated with a
new data set. This was done by randomly dividing the sample into two parts. Assessment
of the model’s fit for both samples could be done simultaneously as EQS has the ability

to analyze multiple samples and test for significant differences between the samples.

Multiple Group Analysis of the Model

The multiple-sample method used for cross-validation can also be used to
compare differences of model fit between specific sub-groups in the sample. However, in
this instance, groups were identified a priori rather than through random selection. The
model was run separately for each of the three segmentation bases, each run
corresponding to a statistical test of Hypotheses 5 to 7. In other words, the model was

simultaneously tested on different national groups. Then, the model was tested on groups

59

according to their stated purpose of travel. Finally, multiple group analysis was done by
dividing the sample into first-time visitors and repeat or previous visitors to Guam.

The model fit indices for each group were then compared. In addition, the
investigator also examined the various parameters of the model to see if there were
significant differences between the groups on these parameters (e.g., path coefficients,

variances, etc.).

60

CHAPTER IV

FINDINGS AND DISCUSSION

Initial Assignment of Variables and Constructs

Th; Working Data Set

An initial examination of the survey instrument revealed fifty-six (56) different
items that appeared to be relevant to the investigation. Of these, forty-six (46) items
referred to visitor satisfaction with specific components of Guam’s tourist product. Eight
items pertained to specific attributes of the respondent’s hotel. Five items asked the
respondent to rate their satisfaction with specific attributes of their dining experience in
Guam, without reference to any one single establishment. The survey also contained
questions about satisfaction with seventeen (17) different tourist activities or attractions
as well as their satisfaction with sixteen (16) individual shopping establishments.
Respondents rated only those activities and shops that they had actually visited or
participated in. In addition, one item asked the respondents to rate how enjoyable or
unenjoyable their overall shopping experience in Guam had been. Finally, the list of
variables included nine items that reflected the tourist’s overall evaluation of Guam —
seven of these appear to be related to issues of overall quality and satisfaction, with two
other items related to future intentions.

Although there were many more items pertaining to component level satisfaction,

this project’s main interests were with the factors associated with the nine overall

61

evaluation items described above. Thus, it was necessary to reduce the list of component
satisfaction items into a more manageable subset through exploratory factor analysis and,

where feasible, by averaging sets of related variables.

MReduction grid Simplification.

Respondents were asked to rate only those shops or activities that they actually
experienced during the trip. Thus, it was unlikely for them to have rated every single
shop or activity on the list. To enable valid comparisons between individuals, composite
scores for satisfaction with shops and satisfaction with tourist activities were developed
using the mean of all shops or attractions rated by each respondent.

Exploratory factor analysis was also employed on the component satisfaction
variables as an additional aid to data reduction. As suggested by Hair et al. (1998), the
factors were extracted through common factor analysis (maximum likelihood) routines
with oblique rotation (oblimin) using the Statistical Package for Social Sciences program.
Common factor analysis is used primarily to identify the underlying dimensions of a list
of variables. In contrast, principal component analysis is used when the objective is to
summarize the original information into a minimum number of factors. Oblique rotation
methods, unlike orthogonal methods, allow the factors to be correlated. This, according to
Hair et al. (1998) is a more realistic approach that leads to theoretically more meaningfirl
factors or constructs (p. 1 10).

The EPA was conducted on two levels. First, the various sets of variables
described in Table 5 were evaluated for unidimensionality — unidimensionality being a

necessary condition for assigning meaning to constructs (Anderson & Gerbing, 1988).

62

Each set was analyzed independently to see if only one factor or dimension was
underlying the variables. Reliabilities for each variable set were evaluated with
Cronbach’s coefiicient alpha. After the separate EF As, all the variables in the working set
were factor analyzed as one merged superset to see if the combined set would yield the
same number of dimensions as did the individual EFAs.

Based on a latent root criterion of eigenvalue > 1.0, only one factor could be
extracted from the five variables for dining satisfaction. Similarly, only one factor could
be extracted from the eight items associated with hotel satisfaction. However, an
examination of the scree plot for the hotel satisfaction variables (see Figure 3) suggested
that a two—factor solution (eigenvalue = .902) might be worth exploring. A second factor

analysis was then undertaken, this time specifying two factors for extraction.

 

Eigenvalues
0)

 

 

 

Number of Factors

F igure 3. Scree Plot of Hotel Satisfaction VMles

 

63

The pattern matrix generated from this second analysis is shown in Table 6. Two
variables — hotel food and hotel value — did not load cleanly. That is, these variables

had statistically significant loadings (greater than .30) on Factors 1 and 2.

Table 6. Pattern Matrix from Factor Analysis of Hotel Satisfaction Variables

 

 

 

 

 

 

 

 

 

 

 

Factor 1 Factor 2
Rate hotel furnishings 0.948 0.021
Rate hotel rooms 0.904 0.028
Rate hotel cleanliness 0.787 -0. 109
Rate hotel view 0.716 0.015
Rate hotel service -0.030 -0.973
Rate hotel staff friendliness -0.053 -0.942
Rate value for money 0.325 -O.460
Rate hotel food 0.319 -0.447

 

 

 

 

A re-run of the factor analysis, this time without the two problematic variables,
mirrored the factor loadings of the initial run. That is, furnishings, rooms, cleanliness and
view again loaded on one factor, while staff and service both loaded on the second factor.
These results appeared, on their face, to be more theoretically consistent with the
literature that describe services in terms of their tangible and intangible elements. Thus, it
was decided to exclude the two suspect items from subsequent analyses. Factor 1 was
labeled hotel tangibles (fumishings, rooms, cleanliness and view) while hotel intangibles
(staff friendliness and service) was the label for Factor 2.

Factor analysis was then conducted on a merged set of variables composed of the
six remaining hotel satisfaction items, the five dining satisfaction variables, and the two
items for shopping satisfaction. In the initial factor analysis, three factors were extracted

— two corresponding to the tangible and intangible hotel factors identified earlier, and the

64

third factor being composed of the five dining satisfaction and the two shopping items.
However, the two shopping items had very weak loadings (less than 0.300) on this third
factor.

A second run, this time specifying a four-factor solution yielded the pattern
matrix shown in Table 7. This result now mirrors the results from the separate EFAs

conducted previously, with two hotel satisfaction factors, one dining satisfaction plus one

factor for satisfaction with shopping.

Table 7. Pattern Matrix from Factor Analysis of Hotel, Dining and Shopping

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Satisfaction
Factor
1 2 3 4
Dining service 0.957 -0.063 -0.107 -0.079
Dining staff satisfaction 0.944 -0.075 -0. 106 -0.085
Restaurant food 0.808 0.044 0.065 0.089
Cleanliness of dining 0.796 0.138 0.058 0.009
Dining value for money 0.740 0.017 0.048 0.140
Hotel furnishings -0.014 0.904 -0.024 0.019
Hotel room 0.002 0.878 0.029 0.011
Hotel cleanliness 0.025 0.774 -0.11 1 -0.021
Hotel view 0.009 0.729 0.005 -0.015
Hotel service 0.002 0.064 -0.874 0.027
Hotel staff friendliness 0.025 0.019 -0.853 0.039
Enjoyable shopping 0.002 -0.026 -0.017 0.718
Mean satisfaction with shops 0.016 0.023 -0.028 0.656

 

 

 

 

 

Table 8 shows the reliabilities (Cronbach’s alpha) for each of the factors. For

factors with only two items, the bivariate correlations are also reported.

65

Table 8. Factor Reliabilities of Hotel, Dining and Shopping Satisfaction

 

 

 

 

 

 

 

 

 

No. of Items Reliability
(Correlation)

Factor 1: Dining 5 .917
Factor 2: Hotel tangibles 4 .902
Factor 3: Hotel intangibles 2 .926

(.862)
Factor 4: Shopping 2 .655

(.502)

 

In evaluating scale reliability, the lower limit for Cronbach’s alpha is generally
accepted to be .70 (Peter, 1979). The reliabilities for the first three factors are well above
this standard. The shopping factor is marginally below the .70 benchmark. However, this
is an acceptable result because Cronbach’s alpha is known to have a positive relationship
to the number of items in a scale. Thus the fact that there are only two items in the
shopping scale could have resulted in a lower alpha. The significant inter-item correlation
between the two shopping variables provides additional support for the scale’s reliability
(Hair et al., 1998).

The final data reduction step was based on Yuan, Bentler and Kano’s (1997)
recommendation to average variables when the number of items becomes large. They
noted that the most frequently used methods for model evaluation — maximum
likelihood (ML) and asymptotically distribution free (ADF) generalized least squares —
may not give reliable inferences under such conditions. They wrote that averaging the
observed variables (which are assumed to depend on only one latent variable) increases
the accuracy of estimators and tests, and improves normality of the averaged error.

Because this project was not as concerned with attribute level satisfaction as with

the overall evaluations, averaged variables were deemed appropriate. The hotel

66

satisfaction and dining satisfaction factors were all reduced to averaged variables. The

shopping factor was retained because one of the items, satisfaction with shopping

establishments, was already a composite item representing the mean of all shops rated by

each respondent. Thus the final set of variables to be included in the structural equations

exercise are as follows“:

H

8.

9.

.Hotel tangibles — the mean of satisfaction ratings for hotel furnishings,

cleanliness, view and room;

. Hotel intangibles — the mean of the respondent’s satisfaction ratings for hotel staff

friendliness and hotel service;

. Dining satisfaction — the mean of five items rating the respondent’s satisfaction

with dining in Guam (food, cleanliness, value for money, staff friendliness, and
service);

. Enjoyable shopping;

. Satisfaction with shops -— a composite score derived by computing the mean of all

shops or stores actually visited by the respondent;

. Satisfaction with tourist attractions — a composite score derived by computing the

mean of all activities or attractions actually experienced by the respondent;

. Destination cleanliness;

Destination safety;

Easy to get around;

10. No problems with communication;

11. Guam was better than expected — used as a proxy for overall satisfaction;

12. Rate Guam - rates Guam as a desirable or undesirable destination;

13. Enjoyable destination;

14. Plan to return to Guam again; and

15. Will recommend Guam to friends and relatives back home.

 

4 Descriptive statistics (means and standard deviations) for these variables are listed in Appendix A.

67

The procedure for conducting the SEM followed the recommended two-step
approach of Anderson and Gerbing (1988). Using this approach, the measurement model
is first estimated. Then, the full model is tested with the measurement and structural
submodels being estimated simultaneously.

They argue that this two-step approach provides a comprehensive assessment of
construct validity. The measurement model provides a test of convergent and
discriminant validity. An assessment of nomological validity is then offered by the test of
the structural model.

In general, this exercise in SEM tests Hypothesis 1 that “the model will fit the
data perfectly.” This general hypothesis incorporates several hypotheses about
relationships between latent variables. These are that:

1. Tourist satisfactions with the various components of the destination flrotels,

dining, shopping and attractions/activities) directly affect tourist conceptions
of both destination quality and overall satisfaction;

2. Destination quality and overall satisfaction are related but distinct construct.
More specifically, destination quality is a predictor of overall satisfaction,
thus partially mediating the effects of the various component satisfactions on
overall satisfaction; and

3. Image fully mediates the effects of overall satisfaction and destination quality
on future intentions.

In addition to the general Hypothesis 1, three other hypotheses were submitted in
Chapter 3. These have to do with the relative impacts of certain latent variables on other
constructs in the model. To review, these are:

I Satisfaction with activities and attractions will have a stronger direct effect
on overall satisfaction them satisfaction with any of the other components.

I Overall satisfaction will have a stronger direct eflect on image than will
destination quality.

68

I The model will better predict the intention to recommend to friends than the
intention to return to Guam.

The Measurement Model

In SEM, the measurement model corresponds to a confirmatory factor analysis
wherein the researcher specifies, a priori, which variables define the factors. In CFA, the
variables are assumed to be indicators of unobserved factors or latent constructs. This
approach differs from exploratory factor analysis where there is no pre-assignment of
variables to factors by the researcher. Rather, the researcher “interprets” the factors based
on the loadings of the variables.

There were a total of 15 variables included in the analysis, based on a sample size
(after listwise deletion of cases) of 1,846. These variables were posited to be indicators of
eight latent constructs: hotel satisfaction, dining satisfaction, satisfaction with shopping,
satisfaction with tourist attractions and activities, destination quality, overall satisfaction,
image, and future intentions. Figure 4 shows the hypothesized relationships between the
observed variables and the latent factors.

For identification of the measurement model, the variances of the factors were
fixed at 1.0. The error terms for factors with only one indicator were fixed at zero

(i.e., single indicator factors were assumed to be without measurement error).

69

 

 

5&2 s 5a-:
”88.5 9 86563-: 638892 £93892 ...aa :55; _ ”35:86.2 Seams as.§§e&8 as, 3636:. 62.”
68% em 2 name as 28658 ....a as? usage 6336.5-.. ”3.8.” ..8émasa some exams: .362; ”$35.23,

286.5 I a £62 . 5 am :55 1 a 5:30 I m ..am 2885. I 438 usasm - m :8 3.5 1 N 288.3% .26: I _ ”805$.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

m m N m N N N N m N N N w m w
FF F DP P a E; F a; a :
2_: 2 S : 2 a_w e e n a m N_
>> > > > >>>> > >> > >>
mm hm mm mm em mm mm Tm

e 4. The Megsurement Model

 

Figpr

70

Assessment of Model Fit.

Using programs like EQS and LISREL, researchers can utilize several different
tests to evaluate model goodness-of-fit. Measures of model fit are generally classified
into three types: (1) absolute fit measures, (2) incremental fit measures, and (3)
parsimonious fit measures.

Absolute fit measures directly assess how well a model reproduces the sample
data. In contrast, incremental fit indices compare a proposed model with a more restricted
baseline or null model. The index then measures the proportionate improvement in fit
between the proposed and reference models (Hu & Bentler, 1995). Finally, parsimonious
fit measures compare models with different numbers of coefficients (therefore, different
degrees of freedom) and determine the fit achieved with each coefficient (Reisinger &
Turner, 1999).

However, there is no single “best” measure of model fit. Because each test or
index has its own inherent limitations, Hair et al. (1998) recommend using one or more
measures from each type. This provides a consensus across different indices as to the
proposed model’s goodness of fit.

Table 9 lists the various goodness-of-fit measures used in this analysis along with
the levels of acceptable fit suggested for each index by the literature. A major criticism of
the x2 test is that as sample size increases, the measure has a greater tendency to indicate
statistical significance. In other words, if sample size is large enough, a significant p-
value will be reported for any specified model (Hair et al., 1998). Given the relatively

large sample size here, the significant x2 test was not altogether unexpected.

71

Nevertheless, the normed chi-square statistic derived by dividing x2 by the model’s

degrees of fi'eedom was still unacceptably high.

Table 9. Goodness-of-Fit Measures for the Proposed Model

 

 

 

 

 

 

Recommended Level of Results
Acceptable Fit. of First CFA
Absolute Fit Measures
Likelihood ratio chi-square Nonsignificant p-value X2 = 825.622
statistic (x2) df = 65
p <.001

Goodness-of-fit index (GFI) > 0.90 .935

(Bagozzi & Yi, 1988)
Root mean square residual < 0.08 .091
(RMR) (Reisinger & Turner, 1999)
Root mean square error of < 0.08 .080
approximation (RMSEA) (Gundersen, Heide, &

Olsson, 1996)

 

Incremental Fit Measures

 

 

 

 

 

 

 

 

 

 

Normed Fit Index (NFI) > 0.90 .929
Non-Normed Fit Index > 0.90 .893
(NNFI) ’

Comparative Fit Index (CFI) > 0.90 .934
Adjusted Goodness-of-Fit > 0.90 .880
Index (AGFI)

Parsimonious Fit Measure
Normed chi-square (x2 + cm 2.0 to 5.0 12.70

 

* - Reference: Hair et al., 1998 unless specified

However, the other fit indices appeared to offer better results. Acceptable levels
of fit were suggested by the GFI, NFI and CFI — all of which were greater than the .9
standard — as well as RMR. The NNFI and AGFI — at .893 and .880, respectively —
were marginally lower than the standard. Marginal fit was also suggested by an

examination of RMSEA.

72

After the assessment of overall model fit, the different measurement equations
were examined. EQS estimates the parameters for the equations that relate the measured
dependent variables to their posited latent constructs. These coefficients can then be
evaluated for significance using a standard z-test. For this initial test, all the estimated
parameters of the measurement model were significant (i.e. z > i- 1.96).

EQS provides two types of tests for identifying means of modifying the proposed
model to improve its overall fit. The Wald or W-test — equivalent to the modification
index in LISREL —— evaluates the effect of adding restrictions (dropping free parameters)
to the model. On the other hand, the Lagrange multiplier (LM) test evaluates the effect of
reducing the restrictions on the model (adding parameters to the model).

The LM test for this initial run suggested large increases in chi-square could result
from allowing the variables “ease of movement” and “communication” to load on several
construct other than that originally specified for them (destination quality). In addition to
the hypothesized loading on destination quality, these two variables appeared to cross-
load on the satisfaction, image and intention factors.

Anderson and Gerbing (1988) identified four basic ways of dealing with
indicators that do not “work out as planned”: relate the indicator to a different factor,
delete the indicator from the model, relate it to multiple factors or use correlated
measurement errors. The first two methods, according to them, are preferred because
these preserve the potential for unidimensional measurement. It was decided to opt for
deletion of the offending variables because the LM tests suggested that they could be
related to any other endogenous factor. Furthermore, it may have also been possible that

the two variables were indicators of some other related but unspecified latent construct.

73

Estimation of this modified measurement model produced fit indices (other than
x2) that generally exceeded all the recommended standards for good fit, e.g.,
12(40, N=1846) = 202.323, p<.001; GFI=.984; AGFI=.963; NFI=.980; NNFI=.968;
RMR=.038; RMSEA=.047. The normed chi-square —— the ratio of chi-square to degrees
of fi'eedom — was much improved. The normed chi-square of 5.058 was just slightly
over the upper limit of the preferred 2 to 5 range.

After evaluating the overall fit of the model, the internal fit of the model was
assessed. First, the convergent and discriminant validity of the model was evaluated.
Then, construct reliabilities and average variances extracted were computed for each
latent factor;

Convergent validity was assessed by examining the regression coefficients of the
observed variables on their respective constructs. Z—tests on these parameters all proved
significant at the .05 level, thus providing support for convergent validity in the model.
Discriminant validity was assessed by constraining two estimated constructs to be equal
(correlation fixed at 1.0) and then conducting a LM test (equivalent to chi-square
difference tests) on the constrained and unconstrained models. This was done separately
for each pair of constructs. Significant improvements in chi-square for the unconstrained
versus the constrained model were observed in all the individual tests, thus suggesting
discriminant validity (Anderson & Gerbing, 1988).

The fit of the model’s internal structure can also be examined through the
computation of construct reliabilities (CR) and average variances extracted (VE). The

recommended standards for structural equation models are to have construct reliabilities

74

greater than or equal to .6 and average variance extracted of at least .5 (Bagozzi and Yi,

1988)

The composite reliability of a construct is calculated as:

. .. Z(standardized loading )2
Construct Relrabrlrty = —

Z (standardized loading )2 + Z a].

 

where a". = measurement error for each indicator.

Average variance extracted is another measure of reliability. It reflects the amount
of variance in the indicators accounted for by the latent construct (Hair et al. 1998) and is

calculated as:

Z (standardized loading 2)
Z (standardized loading 2 )+ z a}.

 

Variance Extracted =

where a". = measurement error for each indicator.

As shown in Table 10, the composite reliabilities and variances extracted for all
the multiple-indicator constructs were better than the minimum standard described
earlier. The shopping satisfaction construct produced the lowest VB and CR (.510 and
.675, respectively) among all the factors, while the best results were obtained from the

quality construct (VE=.688, CR=.815).

75

Table 10. Assessment of Construct Reliabilities and Variances Extracted

 

 

 

 

 

 

 

Factor Indicators Standardized Variance Construct

Loading" Extracted Reliability

Hotel Tangibles .742 0.623 0.766
Intangibles .836

Shopping Enjoyable shopping .766 0.510 0.675
Store satisfaction .661

Quality Safety .804 0.688 0.815
Cleanliness .858

Satisfaction Better than expected .804 0.682 0.811
Enjoyable .824

Intentions Recommend .882 0.613 0.757
Plan to return .698

 

 

 

 

 

 

" Measurement error = (1 - Standardized Loading)

The Structural Model

The model shown in Figure 5 shows the hypothesized relationships between the
constructs defined earlier in the measurement model. To simplify the illustration, only the
latent variables are shown in the figure. The measured variables and error terms
associated with each construct are omitted from the figure. Nevertheless, all of these
parameters were still included in the actual test of the 8-factor model. All the component
satisfaction constructs (hotel, dining, shopping and attractions) were allowed to covary.
As suggested by the literature, an alternate model was also tested. This alternative model
is basically the same as the hypothesized model with one additional path — direct path
from overall satisfaction to intentions (designated with a dashed line in the figure).

Because there were two competing models to be tested, each model was subjected
to cross-validation using a split sample. The test sample of 1,846 cases was split into two
equal groups. One group was designated as the calibration sample and was used to test
the models first. The second, validation sample was then used to cross-validate the

results.

76

 

 

Beat? 283 O

32:: 033583 E 32:“ £3 4 .......

:ocoflmuam

£3 manic—ton»: All 838:?

QZm—Om:

 

 

 

  

cowoflmsum
=Eo>o

   

cozoflmzmm
meanesm

‘ cozoflmzmm

engagemen—

  

25 ESE
233m

 
 

 

  

 

  

cozoémuam

38:

  

 

tru ral Model

r5Th

Fi

77

As shown in Table 11, the original hypothesized model provided less than
acceptable results in both samples. Chi-square was large and significant, with the normed
chi-square well above the preferred ratio. Other than the GFI, all fit indices were
below .9. RMSEA was also unacceptably high.

Table 11. Results of Cross-Validation Procedures; Comparison of Goodness-of-Fit
Measures for Hypothesized and Alternative Models

 

 

 

 

 

 

 

 

 

Chi-Square p Normed RMSEA Ranges of
(df) Chi- Overall Fit
Square Indices
Calibration Sample (n=923)
Hypothesized Model 759.283 (50) <.001 15.19 .124 .791 - .912
Alternative Model 135.509 (49) <.001 2.77 .044 .960 — .984

 

In the alternative model, a path was suggested from overall satisfaction to intentions

 

 

Validation Sample (n=923)

 

Hypothesized Model 666.775 (50) <.001 13.36 .116 .801 - .923

 

 

 

 

 

 

Alternative Model 121.163 (49) <.oor 2.47 .040 964-985

 

In the alternative model, a path was suggested from overall satisfaction to intentions

 

 

 

Note: RMSEA = Root Mean Square Error of Approximation. Overall model fit indices
include Normed Fit Index (NFI), Non-normed Fit Index (NNFI), Comparative Fit
Index (CF I), Goodness-of-fit Index (GFI), and Adjusted Goodness-of-fit Index
(AGFI)

 

In contrast, the indices for the alternate model, from both the calibration and
validation samples, suggested a much better fitting model. Chi-square was still significant
but greatly reduced from that of the original model. More conspicuously, the normed chi-
square was now within the acceptable range of 2 to 5. All overall fit indices were better
than .9, with small RMSEA for both samples.

Despite the good overall fit for Model 2 (the alternate model), the Wald test
indicated several statistically nonsignificant paths. In order of estimated improvement of

Chi-square (based on the calibration sample), these were: quality 9 image; dining

78

 

satisfaction '9 overall satisfaction; image -) intention; and attractions satisfaction -)
quality.

Table 12 presents the solutions for the dependent constructs in equation form (See
Figure 6 for'the graphical representation) based on the estimates of the standardized
coefficients from the calibration sample. The four nonsignificant paths are reflected by
the relatively low coefficients for the respective independent latent variables within the
equations. Thus, in the destination quality equation, F4 or satisfaction with attractions has
a weight of only .055. Other dependent variables in the other solutions with weights less

than .10 were also considered statistically insignificant.

Table 12. Standardized Solutions for Dependent Constructs

 

 

 

 

 

 

Construct Standardized Solution R2
Destination quality (F5) .311(F1) + .222(F2) +.229(F3) + .055(F4) + .754(D5) .432
Overall satisfaction (F6) .118(F1) + .015(F2) + .351(F3) + .173(F4) + .403(F5) + .559(D6) .688
Image (F7) .014(F5) + .449(F6) + .888(D7) .211
Intentions (F8) .879(F6) + .016(F7) + .463(D8) .786

 

 

 

 

Notes: F l=hotel satisfaction; F 2=dining satisfaction; F3=shopping satisfaction; F4=attractions
satisfaction; D = error term for construct

 

That attractions satisfaction did not have a significant direct effect on destination
quality may be an indication that destination quality may reflect tourist evaluations of
facilities per se. Note that the indicators for destination quality in the model are
cleanliness and safety. Since many attractions in Guam are based on the natural rather
than the built environment, these may not be perceived as contributing greatly to the

destination’s overall quality.

79

 

 

 

“cows—em £3 “defines—ES.“ .8 3 03am. com «.53
“585:3?5: 30358 365 cosmon— .mo. 3
335:9.“ >=aoa£§m 2a 328 2.8 =< ”mm—102

 

 

ohm.
ave.

322.85

83:..—

 

:owofimwmm

     

=Eo>O

mew.

 

£an

:0 mad: Smog

cocoflmgm

cocoamflfim

mcozoactas

 

   

meager—m

couoflmzum
mesa

eczema... tam
gem

         
        

J

_

 

6 tructural Model with Factor-to-Factor Path Coefficients

Figure , S

80

In contrast, attractions had a stronger and statistically significant effect on overall
satisfaction. This is compatible with our conception of satisfaction as including personal
characteristics and motivations in the overall evaluation. Nevertheless, satisfaction with
shopping had the greatest effect (both direct and total) on overall satisfaction among all
the components. One interpretation of this could be that shopping should be interpreted as
a primary tourist activity or attraction of Guam on the same level of importance as beach
activities, for example. Dining, it appears, is less important to the attainment of overall
satisfaction — perhaps because it is not considered to be a primary tourist activity.

There may be tendency for planners and managers to assume that sun, sand and
water activities are the main activities in Guam simply because of its island nature.
However, there may be a cultural component (particularly among Japanese) that dictates
the inclusion of shopping as a necessary activity in any trip, particularly to a foreign
destination (Iverson, 1997).

Destination quality did not have a statistically significant effect on the tourist’s
overall image of Guam. This may be interpreted as a reflection of the way in which
image was operationalized. Cleanliness and safety (the indicators of destination quality)
may not be critical ingredients of a “desirable” destination. However, this does not mean
that tourists would not wish for cleanliness and safety in their next vacation destinations.

It can be observed that R2 for image was relatively low. The regression function
explained only some twenty-two percent of the variance in the image construct. This was
in contrast to the R2 for the other endogenous constructs which ranged from .432 to .786.
Again, this suggests shortcomings in the specification of the image construct as defined

by the available set of measured variables. There may be variables, outside the existing

81

data set, that can contribute to a more complete operationalization of the tourist’s image
of Guam as a vacation destination.

These could include measures of the tourist’s prior knowledge, beliefs or attitudes
toward Guam as a vacation destination. These could probably be included in future
models and may be expected to explain a much greater proportion of image variance than
what was achieved here.

Of particular interest in this study were the tourists’ future intentions toward
Guam. Comparing the two behaviors listed in the survey, the model appears to predict
visitor intentions to recommend (R2=.759) better than their intentions to visit again
(R2=.468). One explanation for this would be that it is easier for tourists to commit to the
former behavior than to the latter. Based on the Theory of Planned Behavior (Ajzen,
1985; Ajzen & Driver, 1992), tourists may perceive that they have less control over their
ability to travel to Guam again than over their ability to spread word-of-mouth. Telling
friends about one’s vacation is not as constrained by time, money and the availability of

travel companions as planning another vacation would be.

82

The Multi-Group Structural Equations Model (MSEM)

In cases where data are gathered from individuals belonging to different groups,
Bentler (1995) suggests that it may be appropriate to question whether one is dealing with
a single population or with multiple populations. He writes:

“If a model having identical parameters in all groups can fit acceptably,

the resulting model covariance matrices are identical ...and the samples

can be treated as arising from the same population. If the models of the

various groups have parameters that are different, the resulting model

covariance matrices will be different and the various samples must be

treated as arising from different populations. One can then conclude that

there is an interaction between population membership and structural
model.” (p. 150)

In this particular sample, at least three bases for grouping have been identified: by
nationality, by purpose of travel and by previous experience with Guam. Descriptive
statistics by group — i.e., means and standard deviations by nationality, purpose and
experience — are reported in the appendices.

Three separate analyses based on these grouping variables were conducted in
EQS. Although this type of analysis could have been done using linear regression path
analytic models, MSEM was preferred because of its ability to test the applicability of a
model to different groups simultaneously without having to bother with interaction terms
and nested models (Scott-Lennox & Lennox, 1995).

MSEM allows researchers to test different types of hypotheses regarding group
differences. This study focused on only one type of hypothesis. The general hypothesis
for the three MSEMs was that of equal factor regression coefficients (Factor-to-factor
paths). Thus, the analyses tested whether the path coefficients among the latent factors

was the same across groups. If so, then the latent causal process in the model would be

83

similar across groups. This can occur even if the variances or covariances of the factors

are not equal (Bentler, 1995).

By Country of Origin

For the cross-national analysis, three sub-groups were identified — one group
from Japan (n=979), another made up of visitors from South Korea (n=212) and a merged
grouping of visitors from Hong Kong and Taiwan (302). Although the Japanese sub-
group was distinctly larger than the other two, this was not expected to be a problem
since both the Korean and ethnic Chinese groups each had at least 200 cases. This was
well above the minimum of 100 cases typically recommended in the SEM literature.

The multi-sample model appeared to fit the data very well. Although the chi-
square statistic was significant —— 12(173, N=1493) = 358.470, p<.001 — it was small for
the degrees of freedom involved (Normed x2=2.072). Overall fit indices for the multi-
sample model ranged from .945 to .973 while the error or residual-based indices were
small (RMR=.073; RMSEA=.028).

Three factor-to-factor paths were not statistically significant (or = .05) in all
groups: dining satisfaction 9 overall satisfaction; quality -) image; and image 9
behavioral intentions (refer to Figure 7). However, satisfaction with tourist attractions
had a statistically significant, albeit small (.051, Z = 2.171; p <.05) effect on quality. In

the initial SEM, this path was not significant at the .05 level.

84

 

 

3388 :2“ com ufioficwfi

82 826886 52. x30 5 ease; as 5%. 5258
8 :03 an A3 .v 3 e23— uea :25. 5223 coup—otzo we u 0

SN . v 5 533% can :83. 5953 Bandage me u m

 

A3. v $ se.—0M can :83 503.2. ooceogc .wfi n < co gamma—Wm

 

 

8288.2

 

cosoflmgm
=So>O

 

cowoémzum
wcanosm

      

32285
22:..—

 

owes:

couoémcam
mesa

 

/ Ego

coceczmon

couoummuam
$8:

                

re 7. M 11 le- am 1 Structural E uations Model b Nationalit

Fi

85

The Lagrange multiplier (LM) test revealed four significant factor-to-factor path
differences, all of which involved differences between the Japanese sub-group and one
or both of the other national groups. Specifically, the Japanese differed from the Koreans
with regard to the path from overall satisfaction to image (a = .05) and from the Hong
Kong/'1’ aiwan group with regard to the effect of shopping satisfaction on overall
satisfaction (a = .10). Furthermore, significant differences involving the path from image
to behavioral intentions were found to exist between the Japanese group and the two
other groups. However, as noted earlier, this particular image-) intention path was not
statistically significant in any of the groups.

An analysis of the unconstrained multi-sample model revealed the magnitude of
these differences. The groups produced significantly different weights on two specific
paths. Regarding the effect of shopping satisfaction on overall satisfaction, the estimated
coefficients were slightly lower for Japanese respondents (.323) than for the other groups
(all at .424). However, it was the Korean group (.855) which differed fiom the other

groups (.674) in terms of the effect of overall satisfaction on destination image.

By Purpose of Travel

The multi-sample SEM according to purpose of travel involved three groups. The
largest of these (1,008; 70.05%) was made up of the respondents who signified
“pleasure” as their main purpose for traveling to Guam. Respondents traveling for
business or on a company-sponsored trip were lumped together to form a “corporate”
group (243; 16.89%). The third group included in this analysis was made up of

“honeymooners” (1 88; 1 3.06%).

86

As with the earlier cross-national analysis, this multi-sample model appeared to fit
the data very well. Again, the chi-square statistic was significant:
12(173, N=1493)=452.617, p<.001. However, the normed chi-square (xZ/df = 2.616)
was small and within the recommended range of 2 to 5. Overall fit indices for the multi-
sample model ranged from .931 to .960 while the error or residual-based indices were
small (RMR=.078; RMSEA=.034). All of these suggested excellent fit with the data.

Four factor-to-factor paths were not statistically significant (at = .05) in all groups:
dining satisfaction 9 overall satisfaction; attractions satisfaction 9 quality; quality 9
image; and image 9 behavioral intentions (refer to Figure 8). These were consistent with
the original findings from the single sample SEM involving all respondents.

The Lagrange multiplier (LM) test revealed significant differences between the
honeymooners and one or both of the other groups in terms of factor-to-factor
coefficients. The results of the LM test suggested significant differences (a=.05)
between the honeymoon and pleasure groups with regard to the coefficients for two
paths: hotel satisfaction 9 destination quality and destination quality 9 image.
However, the latter was not statistically significant in either group. Honeymooners also
differed from corporate travelers on two paths: attractions satisfaction 9 destination

quality (a = .05), and dining satisfaction 9 destination quality (a2. I).

87

 

 

A3. v 5 3.2.7.3 28 coo—cacao: 5233 3535—6 .ma u 0
A3. v 5 $053; new £58.35: 5223 oocuootfi .me u m

A3. v 5 053...:— vca 508.35; 5933 oofitobfi .mfi u <

 

 

:ocoflmzem

=Eo>o

 

33:85

OMB—dun ”mm EH

 

 

< 3:30

cow—Eamon

   

58¢QO
22852

cozoflmgm
wean—96

cozoflmzam
mean

so cementum
$8:

             
           

r 8 Multi 1 -Sam le tructuralE uations Modelb Pu ose of Visit

Fi

88

The significant constraints were released and the model was then retested.
Significant differences across groups were found in terms of the effect of hotel
satisfaction on destination quality. Business travelers had the largest coefficient (.589)
followed by honeymooners (.462). Pleasure visitors had the smallest coefficient at .246.
Dining satisfaction also had a significant effect on destination quality for both
honeymooners and pleasure travelers (.176 for both groups). In contrast, dining
satisfaction did not have a significant effect on destination quality among business

travelers with a coefficient near zero (.021).

By Previous Experience

The final MSEM divided the sample according to whether they were first-time
visitors or whether they had previously visited Guam. Almost eighty percent (1,203 of
1,493 cases) of the respondents were visiting Guam for the first time. Only 290
people (19.42%) responded that they had been to Guam previously.

As with the two other multi-sample models, the measures of model fit were all

satisfactory. Although the chi-square test was significant — 12(110, N=1493) = 383.081,

p<.001 — the small Normed x2 (3.483) suggested that this was acceptable for the degrees
of fi'eedom involved. All overall and incremental measures were greater than .9
(NFI=.962; NNFI=.961; GFI=.964; AGFI=.940). In addition, the root mean squared
residual (RMR=.068) and root mean squared error of approximation (RMSEA=.041)

were both small.

89

Five factor-to-factor paths differed significantly between the two groups (see
Figure 9). Statistically significant differences (or = .05) were detected involving the paths
related to the effects of attractions satisfaction, hotel satisfaction and destination quality
on overall satisfaction. Furthermore, the two groups also appeared to differ in terms of
the effects of dining satisfaction and shopping satisfaction on destination quality.
However, for both these paths, statistical significance could only be claimed at or = .10.

For first-timers, neither hotel satisfaction nor dining satisfaction had a statistically
significant effect on overall satisfaction. In the full sample SEM, only the latter (dining
9 overall satisfaction) was not statistically significant. Additionally, the following paths
could not be considered to be statistically significant: attractions 9 destination quality;
destination quality 9 image; and image 9 intentions.

In contrast, the returning visitors model had only four nonsignificant paths. As
with all previous analyses, attraction satisfaction 9 quality; dining satisfaction 9 overall
satisfaction; and image 9 intentions were not statistically significant. However, contrary
to the results of the original model, hotel satisfaction among returning visitors did not
have significant direct effect on overall satisfaction. Conversely, destination quality had a
statistically significant effect on image for the returning group (this path was not

significant in the original model).

90

22285

83:..—

 

 

 

oEImE .8 “cube 80.6 595.5 A. . . .

 

9333 Home. 8.5 Soto 826 cow—85m All

 

comaoflmcam
=So>0

omeE—

 

€20

cote—Ewen

 

coaoémgm

28.88314

 

  

cowoémuam
mafia—25

cowoémzam
33:

First-Time vs.

      

r 9. M lti le-Sam le StructuralE uations Model Com arin

  

 

tVi it r

R

Fi

91

A model wherein the significant paths were released from the equality constraint
was then analyzed to estimate the coefficients for the paths on which there were
significant inter-group differences. Two paths, in particular, exhibited significantly larger
coefficients for the repeat visitors group than for the first-timers. With regard to the effect
of shopping satisfaction on destination quality, the coefficient for the repeaters (.413) was
almost twice that of the first-timers (.225). Similarly, the effect of destination quality on
overall satisfaction was much greater for repeat visitors (.733) than for first—time visitors

(379)

92

CHAPTER V

SUMMARY AND CONCLUSIONS

A major assumption of the marketing literature is that higher levels of satisfaction
with a product or service lead to more fi'equent purchases of that product or service. In
the tourism field, this means that satisfaction or dissatisfaction with a destination
experience can influence future intentions to return to that destination as well as other
behaviors related to that place (e. g., word-of-mouth recommendations).

This dissertation can be said to represent a critical look at that assumption. How
valid is it, considering the unique elements of the tourist product? This study thus
addressed the question of how tourists’ evaluations of a recent visit to a place influence
their future intentions regarding that destination. In this particular study, that question
was posed to international visitors to the island of Guam.

The methods and findings related to this specific research problem are
summarized in this chapter. First, the procedures or research methodologies employed
will be presented. The specific findings of the data analysis, as they relate to the specific
research hypotheses will then be summarized and conclusions presented. Following the
summary of findings will be a summary section on the conclusions reached based on the
findings, as well as a discussion of their implications. Finally, the chapter — and the

dissertation — will conclude with a section outlining recommendations for further study.

93

Summary of Procedures

The dissertation was based on analysis of secondary data collected through exit
surveys of international tourists at Guam International Airport. A sample of 3,108 cases
— representing seventy-five percent (75%) of all interviews conducted by the GVB
during the period from June 1998 to November 1999 ——was randomly-generated from
GVB’s database. The sample also represents a cross-sectional view of Guam’s most
important sources of international tourist traffic: Japan, South Korea, Taiwan and Hong
Kong.

The first objective of the project was to develop a model of tourist behavior that
defines and distinguishes between the key post—experience constructs of satisfaction,
quality and image. An initial examination of the survey instrument revealed fifty-six (56)
different items that appeared to be applicable to this objective.

Forty-six (46) of these items referred to visitor satisfaction with specific
components of Guam’s tourist product. One item asked the respondents to rate how
enjoyable or unenjoyable their overall shopping experience in Guam had been. Finally,
the list of variables included nine items that reflected the tourist’s overall evaluation of
Guam. Seven of these seemed to be related to issues of overall quality and satisfaction,
with two other items related to future intentions.

Given the large number of variables available, it was necessary to reduce the list
to a more manageable subset. Exploratory factor analysis was employed on the variables
related to hotel and dining satisfaction. As a result, these variables were reduced to three
composite variables, two related to hotel satisfaction and one representing satisfaction

with dining in Guam.

94

Respondents were asked to rate only those shops or activities that they actually
experienced during the trip. Thus, it was unlikely for them to have rated every single
shop or activity on the list. To enable valid comparisons between individuals, composite
scores for satisfaction with shops and satisfaction with tourist activities were developed
using the mean of all shops or attractions rated by each respondent.

Thus the final set of variables to be included in the analysis were the following:

hotel tangibles;

hotel intangibles;

dining satisfaction;

enjoyable shopping;

satisfaction with shops;

satisfaction with tourist attractions;
destination cleanliness;

destination safety;

case of getting around;

10. communication;

1 1. overall satisfaction;

12. desirability (of Guam as a vacation destination);
13. how enjoyable Guam was;

14. plan to return to Guam; and

15. to recommend Guam to friends and relatives.

PWSQS‘EP’P?

Structural equations modeling (SEM) using the EQS program was then used to
develop and test a model based on the research problems and hypotheses. The procedure
for conducting the SEM followed the recommended two-step approach of Anderson and
Gerbing (1988). Using this approach, a measurement model was first estimated. Then, the
full model was tested with the measurement and structural submodels being estimated
simultaneously.

According to Anderson and Gerbring, this two-step approach offers a

comprehensive assessment of construct validity. The measurement model provides a test

95

of convergent and discriminant validity. On the other hand, a test of the structural model
presents an assessment of nomological validity.

After the model was tested using the data from the entire sample, multi-group
SEM was employed to test for differences between different groups. Because the survey
did not include measures of the visitors’ prior knowledge of Guam nor of their
motivations for traveling to Guam, group memberships were used as proxies for differing
levels of information, as well as differences in travel motives.

In this particular sample, three bases for grouping were identified: by nationality,
by purpose of travel and by previous experience with Guam. Three separate analyses
were conducted using these bases. Although this type of analysis could have been done
using linear regression path analytic models, MSEM was preferred because of its ability
to test the applicability of a model to different groups simultaneously without having to
bother with interaction terms and nested models (Scott-Lennox & Lennox, 1995).

The MSEM specifically addressed the general hypothesis of equal factor
regression coefficients (factor-to-factor paths) between groups. That is, the analyses

tested whether the path coefficients among the latent factors was the same across groups.

96

Summary of Findings

As recommended by Anderson and Gerbing ( 1988), a two-step approach to SEM
was used. First, a measurement model — corresponding to a confirmatory factor analysis
— was estimated. There were a total of 15 variables included in the analysis, based on a
sample size (after listwise deletion of cases) of 1,846. These variables were posited to be
indicators of eight latent constructs: hotel satisfaction, dining satisfaction, satisfaction
with shopping, satisfaction with tourist attractions and activities, destination quality,
overall satisfaction, image, and future intentions. Figure 10 shows the hypothesized
relationships between the observed variables and the latent factors.

For identification of the measurement model, the variances of the factors were
fixed at 1.0. The error terms for factors with only one indicator were fixed at zero, i.e.

single indicator factors were assumed to be without measurement error.

97

 

 

638 2 55-2
..aeeoE o. assesses £9892 £33892 as :85: xaoéeaaé Sea? ”8:32:58 55 nausea oza
ease a» 2 name. 38 28858 ....a 223 usages seesaw-.. usage .aosaess .233 .8338 some @5335

886.5 i a vase i N. am :55 i o 5:55 i m Jam 86:65.... l 4 ...am asaaoem i m :8 3:5 1 ~ ..eoeosaeom sex i _ 308E

 

 

H

 

 

 

 

 

 

 

m—

v-III
F!

> :2 <—~
4__m
{—93
> Q 4—9
4—m
4—_m
> ox G—N
4——m
> h {—9

>~n

>0 For
{—m
>V {—m
{—w

>
>2

>F‘4——N

 

>2.
>

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

>m
>N

 

>00

 

 

 

 

 

 

 

 

mm mm on mm on mm mm E

re 10. The Meesurement Model

 

Lisa

98

Given that there is no single “best” measure of model fit, several indices were
used to assess the measurement model following Hair et al.’s (1998) suggestion. Table 9
(see page 72) lists the various goodness-of-fit measures used in this analysis along with
the levels of acceptable fit suggested for each index by the literature.

Given the relatively large sample size here, the significant )8 test was not
altogether unexpected. Nevertheless, the normed chi-square was still unacceptably high.
The other indices, particularly the NNFI, AGFI and RMSEA suggested a marginally
acceptable fit. However, a Lagrange multiplier (LM) test suggested significant
improvements in model fit could result from allowing the variables “ease of movement”
and “communication” to load on several constructs other than that originally specified for
them (destination quality).

In other words, they could be associated with any other endogenous factor in the
model. Alternatively, these two variables could have also been indicators of some other
related but unspecified latent construct. Thus, it was decided to modify the model by
deleting these two offending variables. The subsequent estimation of the modified
measurement model produced fit indices (other than x2) that generally exceeded all the
recommended standards for good fit.

Next, the internal fit of the model was assessed. Convergent validity was assessed
by examining the regression coefficients of the observed variables on their respective
constructs. Z-tests on these parameters all proved significant at the .05 level, thus
providing support for convergent validity in the model. Discriminant validity was
assessed by constraining two estimated constructs to be equal (correlation fixed at 1.0)

and then conducting a LM test (equivalent to chi-square difference tests) on the

99

 

constrained and unconstrained models. Significant improvements in chi-square for the
unconstrained versus the constrained model were observed in all the individual tests, thus
suggesting discriminant validity (Anderson & Gerbing, 1988).

In addition, the fit of the model’s internal structure was examined through the
computation of construct reliabilities (CR) and average variances extracted (VE). Based
on minimum recommended levels of .6 and .5 for construct reliabilities and average
variance extracted, respectively, the computations provided additional support for the
model’s internal fit.

One conspicuous statistic generated from the exercise was the high correlation
coefficient between the intention and overall satisfaction constnrcts (.914). Ideally, a
researcher would expect, from the beginning, to detect a strong association between these
two constructs. However, a score relatively close to one may suggest to some relations
that overall satisfaction and firture intentions should belong to one rather than two
independent but related constructs.

Nevertheless, such a conclusion would run contrary to the body of literature that
holds otherwise. It is surmised that this extremely high correlation is simply an artifact of
the data collection process. Since the survey was conducted immediately after the visit,
the responses, particularly those related to future intentions, may have been influenced by
the mood of the tourists at that time. This might heighten the correlation between the

primarily affective construct that is satisfaction and the behavioral intentions construct.

100

Structural Model

The model shown in Figure 11 shows the hypothesized relationships between the
constructs defined in the measurement model. To simplify the illustration, only the latent
variables are shown in the figure. The measured variables and error terms associated with
each construct were omitted from the figure. All the component satisfaction constructs
(hotel, dining, shopping and attractions) were allowed to covary. As suggested by the
literature, an alternate model was also tested. This alternative model is basically the same
as the hypothesized model with one additional path -— a direct path from overall
satisfaction to intentions (designated with a dashed line in the figure).

Because there were two competing models to be tested, each model was cross-
validated using a split sample. The test sample of 1,846 cases was split into two equal
groups. One group was designated as the calibration sample with which to first test the
models. The second half of the sample was then used to cross-validate the results for both
models.

The original hypothesized model provided less than acceptable results in both
samples (See Table 11, page 78). By contrast, the indices for the alternate mode —— for
both the calibration and validation samples — suggested a much better fitting model.
Chi-square was Still significant but greatly reduced from that of the original model.

Despite the good overall fit for Model 2, the Wald test indicated that overall fit
(based on the Chi-square statistic) could probably be improved by releasing the following
paths: quality 9 image; dining satisfaction 9 overall satisfaction; image 9 intention;

and attractions satisfaction 9 quality.

101

 

 

05352, 883 C

«.62: 033883 E 323 fine 4 .......
£8 wofimofiogm Allll
DZmOmi—

 

 

      

2228:—
23.5

 

    
    

cowoémsam
=So>o

 

 

:235309

cowoémzam
22552

:ozoémsnm
wcmaaosm

cowoémawm
mesa

cowoémsam
68$

 

Fi r 11.The St cturalModl

102

 

An assessment of the standardized solutions for each of the dependent latent
constructs revealed that R2 for image was relatively low. The standardized function
explained only some twenty-two percent of the variance in the image construct. This was
in contrast to the R2 for the other endogenous constructs which ranged from .432 to .786.
This suggests limitations in the specification of the image construct as defined by the
available set of measured variables. In particular, the non-inclusion of pre-visit
expectations, motivations or levels of information could explain the relatively poor
specification of the image construct.

The model appeared to predict visitor intentions to recommend Guam better than
it did their intentions to visit Guam again. One explanation for this was that tourists may
feel less constrained to commit to the former behavior than to the latter. At that stage of
the journey, visitors already may have fully-formed plans to recount their visit to fiiends,
relatives and colleagues back home. For some, the opportunity to tell their social circle
about their experiences in a foreign destination may even have been an ego-enhancing
motivation for coming on the trip (Dann, 1977).

In contrast, the notion of a return visit to Guam, let alone another foreign trip to
any destination, may not yet be very concrete to many respondents. Further, for such a
trip to be realized, tourists would probably need to address several constraints (e. g., cost,
availability of travel companions, lack of fi'ee time, etc.) before being able to declare their
true travel intentions.

Another issue that was examined was the relative contribution of the various
component satisfactions to overall satisfaction. Satisfaction with shopping had a greater

effect (both direct and total effects) on overall satisfaction than did their satisfaction with

103

attractions. This may suggest -— for the market segments included in the study — that
shopping should be interpreted as a primary tourist activity or attraction of Guam on the
same level of importance as beach activities, for example. In addition, there may be a
cultural component (particularly among Japanese) that dictates the inclusion of shopping
as a necessary activity in any trip, particularly to a foreign destination (Iverson, 1997).

Multiple sample analysis was also conducted to determine differences in model fit
between groups. This was done by dividing the sample according to three common
market segmentation criteria: by country of origin, by purpose of travel, and by previous
experience with Guam — and then testing for differences.

The practice of market segmentation assumes that segments have different
interests and needs. Furthermore, segmentation by purpose and by experience suggests
differences in motivation and level of knowledge about Guam, respectively. Motivation
and level of information have been said to influence satisfaction and perceptions of
quality through their roles in determining visitor expectations.

Because there were no expectation variables in the database, the effect of
expectations on perceived quality and satisfaction could not be modeled directly.
However, using the three segmentation variables in the MSEM was seen as a way of
examining the moderating effects of differences in expectations by groups. In other
words, the segmentation criteria were used in the MSEM as proxies for the missing
expectations construct.

Multiple-sample structural equations modeling was conducted separately for each

grouping variable. The objective was to test for differences in the factor-to-factor path

104

coefficients between groups. Significant differences were found in each multi-sample
run, using each of the segmentation criteria.

In the MSEM by country of origin, most of the significant differences involved
the Japanese respondents and one or both of the other national groups. No significant
differences were found between the Korean and Hong Kong-Taiwan groups. The exercise
indicated that the Japanese segment differed significantly from the Hong Kong/Taiwan
group with regard to shopping satisfaction 9 overall satisfaction and from the Korean
group along the overall satisfaction 9 image path. Furthermore, significant differences
involving the image 9 intentions were found between the Japanese group and the two
other segments.

Specifically, the MSEM showed that the effect of shopping satisfaction on overall
satisfaction was slightly lower for Japanese visitors than for either of the two other
segments. Could it be that Japanese visitors, as a group, are more experienced
international shoppers than other visitor groups? If so, Japanese visitors may be more
jaded with regard to the shopping activity. Shopping, for the Japanese, may be more of a
cultural obligation than an activity that they look forward to compared to how they are
excited by other types of tourist activities.

Similarly, the MSEM by purpose of travel found significant between-group
differences involving the honeymoon segment and the two other segments. No significant
differences were detected between the business and pleasure segments. The results
indicated that the effect of hotel satisfaction on destination quality was higher for

honeymooners than it was for pleasure visitors to Guam. Assessing the magnitude of this

105

difference, the effect of hotel satisfaction on destination quality was much lower for
pleasure travelers than for business travelers and honeymooners.

Honeymooners may be more conscious about service quality in hotels than
pleasure travelers because the honeymoon is a much more significant life experience than
typical pleasure trips. In addition, the hotel in which they spent their honeymoon might
hold much sentimental value for a visiting couple. Similarly, business travelers may view
hotels as both a base for their business activity, as well as a place where they can rest and
refi'esh themselves before another day of work.

Thus, the hotel is a much more important element of the overall visit to
honeymooners and businesspeople than for pleasure visitors. For this reason,
honeymooners and business travelers may be more sensitive to differences in hotel
service quality than pleasure travelers. For the latter group, the activities and services
inside the hotel might not be as critical to the Guam experience as the tourist services
found outside the hotel.

How does the model differ for honeymooners compared to business travelers?
The MSEM that satisfaction with attractions and dining had stronger effects on the
overall assessment of quality for honeymooners than for business travelers. The result
probably reflects the importance of the work activities to the business traveler, as much
as it reflects the relevance of attractions and dining establishments to a honeymooning
couple. Again, this may be expected since the opportunity to engage in recreational or
tourist activities would not be a primary motivation for most business travelers.

These differences may have important implications for marketing and promotions.

For hotels, the findings may suggest that promotional strategies built around the hotel’s

106

in-house services and physical facilities might be more appropriate for honeymooners
and business travelers. On the other hand, marketing campaigns that sell Guam as tourist
destination might be better for pleasure travelers. Hotel brochures might show the
convenience of their location relative to the key tourist attractions or they could stress
other features or facilities of the hotel and how they facilitate access to the different
recreational opportunities on the island. Business travelers may also pay attention to
communications that show how the hotels can facilitate their business activities while in
Guam. For the GVB, differentiated marketing campaigns for the pleasure, honeymoon
and business segments may also be in order. For pleasure travelers and honeymooners,
the emphasis should be on tourist attractions and activities (including shopping), perhaps
adding a little bit more information on hotel facilities for the honeymooners.

In the comparison of first-time and repeat visitors, two paths exhibited
significantly larger coefficients for the repeat visitors group than for the first-timers. First,
the coefficient for first-timers was almost half that of repeat visitors on the effect of
shopping satisfaction on destination quality. This suggests that the shopping activity may
not be as critical for first time visitors as it is for repeat visitors. First-time visitors to
Guam may be more concerned with discovering and evaluating the other aspects of the
destination such as the hotels and tourists attractions. Given their previous experience
with Guam hotels, restaurants and attractions, repeat visitors might no longer be
particularly surprised by the quality level of these components. However, the shopping
experience could be significantly changed for each new visit when the tourist chooses to

buy new items or to buy for different people each trip.

107

 

By contrast, first-timers would be experiencing — and evaluating — everything
for the first time. Thus, satisfaction with hotels, shopping and attractions all would have a
bearing on their overall assessment of Guam’s quality.

The second key observation fi'om the comparison of first-time and repeat visitors
was that the effect of destination quality on overall satisfaction was much greater for
repeat visitors than for first-time visitors. A possible explanation for this could be that
there is an element of surprise related to the first-timers’ overall satisfaction’ that is
related to factors other than the supplier-based components that factor into destination
quality. For repeat visitors, the novelty of the experience may no longer be as significant.
Therefore, they could rely on a more cognitive, quality-oriented assessment of overall

satisfaction.

Implications

The first objective of this study was to distinguish satisfaction, quality and image
— all post-visit constructs — from each other through a proposed model of tourist
behavior. The exercise of testing the model that described the relationships between these
constructs was thus equivalent to testing Hypothesis 1 that the model would fit the data
perfectly. Overall, the results of the structural equations modeling exercise provided
some support for this hypothesized model. However, some paths in the proposed model

were not validated.

 

5 This corresponds to Spreng and Mackoy’s (1996) conception of expectations disconfirmation, as well as
Crompton and Love’s (1995) assertion that satisfaction is influenced as much by specific supply-based
attributes (e.g. facilities, services and attractions) as by the tourist’s socio-psychological state (mood,
needs, disposition).

108

In particular, it appeared that image did not fully mediate the effects of overall
satisfaction on intentions. To the contrary, the direct effect of satisfaction on intentions
was even stronger than that of image on intentions. The measurement model already
hinted at this result given the strong correlation between satisfaction and intentions. Does
this mean that the hypothesized mediation should be discarded altogether? Or is this
result simply an artifact of the data collection method — an illustration of the halo effect
with the survey coming immediately at the conclusion of a satisfying trip?

The second study objective was to determine the relative contributions of the
various components of tourism (hotels, restaurants, shopping establishments, attractions,
etc.) to overall satisfaction with a destination. Thus, Hypothesis 2 stated that tourist
satisfaction with the different attractions around Guam would have the strongest effect on
perceived quality and overall satisfaction. It was assumed that the attractions best
represented the motivations or factors that "pulled" the tourists to Guam. The other
components — accommodations, dining and shopping — were interpreted as providing a
layer of support services around the core tourist activities.

Hypothesis 2 was supported only in that tourists’ satisfaction with attractions had
a stronger effect on overall satisfaction than did satisfaction with either accommodations
and dining. However, the hypothesis was not supported by the stronger direct effect on
overall satisfaction by shopping satisfaction. The exercise suggests that shopping should
be treated as a tourist activity in its own right. At least for the markets included in this
study, and for the island of Guam, shopping plays a large role in determining both

perceived quality and overall satisfaction.

109

The SEM has also demonstrated that perceived destination quality and overall
satisfaction are distinct constructs. The model has thus provided additional support for
the definition of quality in terms of those attributes that the visitors perceive to be beyond
their control (Crompton & Love, 1995). In comparing the effects of both constructs on
image, strong support for Hypothesis 3 was found. Hypothesis 3 had suggested that
overall satisfaction would have a stronger direct effect on image than perceived
destination quality.

Furthermore, the model has shown that overall satisfaction has a greater impact
on future visitor intentions than does perceived quality. This is consistent with the
marketing concept, which suggests that a marketing entity's profits are gained through
consumer satisfaction. Although it may be convenient and tempting to focus only on
those attributes that are directly controllable by management, destinations should be
forewarned that perceived quality is not the only source of visitor satisfaction. That is,
there does not appear to be a one-to-one correspondence between higher ratings for
quality and ultimate visitor satisfaction. Thus, in seeking to improve quality, managers
should aim to improve the quality-enhancing attributes such that satisfaction is
maximized.

If higher ratings for perceived quality do not automatically translate to
satisfaction, what are destinations to do? The results suggest that improving quality
should not be an end in itself. Other factors must be considered as well. What are the
other antecedents of satisfaction? The literature suggests that visitors themselves
contribute much to their own satisfaction. Destinations should thus be aware of visitors'

travel motivations as well as the benefits they expect to derive from the different

110

attributes of a destination. Some researchers (Spreng & Mackoy, 1996) have suggested
that perceived quality should be management's focus because it involves those attributes
within management's control.

While the motives of visitors and their expectations of personal benefits may
indeed be beyond a destination's control, the satisfaction of these motives and
expectations may still be manageable, albeit indirectly. Research into how visitors derive
benefits from destination attributes (such as those based on a means-end framework) may
better show destination managers how particular elements can contribute to visitor
satisfaction. Such research may not explicitly demonstrate how to manage visitor
satisfaction. Nevertheless, it may enable destinations to identify the means for achieving
satisfaction by pinpointing specific attributes to work on.

There also was an interest in learning about the effects of satisfaction and
perceived quality on the behavioral intentions of visitors to Guam. Recognizing that
tourists may be predisposed differently toward different actions, the study sought to
compare the predictive power of the hypothesized model on behaviors over which
tourists have varying degrees of control. Hypothesis 4 suggested that the model would
predict visitor intentions to recommend Guam better than it would their intentions to
return to Guam.

The findings supported Hypothesis 4, and thus also appeared to uphold the theory
of planned behavior (Ajzen & Driver, 1992). The theory suggests that the strength of
association between attitude and actual behavior is determined by the level of control
people believe they have over the behavior. The more that the tourists recognize

constraints against behavior (visiting or recommending), the less likely it is that the

111

positive attitude will result in action. In this case, the tourists may feel less constrained
about recommending Guam to their friends and relatives. On the other hand, returning to
Guam may be constrained by factors such as time, expense and the availability of travel
companions.

Thus, the GVB, as well as other similar organizations, should consider strategies
for encouraging word-of-mouth (WOM) behavior from their visitors. Encouraging them
to purchase Guam-labeled merchandise, or even to bring home promotional materials
from the visitors bureau could facilitate the story-telling process among fi'iends and
relatives. Governments and other sponsors of travel bureaus should also review their
policies for evaluating the performance of destination promotion boards. The number of
returning visitors (particularly for long-haul destinations) may not be a valid mechanism
for assessing the effectiveness of a tourism program.

These findings do support the principal contention of the marketing concept that a
business (in this case, a destination) should seek to satisfy customer needs at a profit.
However, the implication that this profit will come from a process of repeat visits by
satisfied tourists does not appear to be as valid for long-haul travel destinations like
Guam. Destinations such as Guam should consider a marketing model based more on
generating visitation from referrals than on repeat visitation from satisfied visitors.

The study’s final research objective was to test the effects of motivation and level
of information on the different post-experience constructs in the model. Specifically, the
factor-to-factor relationships described in the model were simultaneously tested for

differences between groups identified by nationality (Hypothesis 5), stated purpose of

112

visit (Hypothesis 6), and previous visitation (Hypothesis 7). These are common criteria
used by destination marketers to segment their target markets.

For this analysis, the tourist’s prior experience with Guam was used as a proxy for
his or her level of information about Guam. The tourist’s stated purpose of visit was then
used as a proxy for motivation. It was assumed that country of origin, as a reflection of
cultural differences in tastes and preferences, would also be possible proxy for differing
motivations. Furthermore, since visitor bureaus run different marketing campaigns for
each country, the levels of information about Guam in each country market could also be
assumed to vary.

Subsequent multiple sample analyses of the hypothesized model supported
hypotheses 5, 6, and 7 by detecting significant differences between groups. Country of
origin, purpose of visit, and previous visitation experience thus appeared to validate their
use as market segmentation criteria. The moderating effects of these variables on the
model suggest that they could be used as proxies for varying levels of information,

motivation and, by extension, expectations.

Implications for Destination Marketers

One key finding from the study is that tourists’ perceptions of a destination’s
quality are not the same as their satisfaction with their visit to the place. Furthermore,
overall satisfaction has a greater impact on future visitor intentions than perceived
quality. Although it is probably desirable for a destination to be perceived as offering
high quality accommodations and tourist services, it may still fall short on satisfaction if

tourists perceive it as sterile, bland or overpriced.

113

Both perceived quality and satisfaction are based on the tourists’ prior
expectations. In cultivating these expectations through marketing, destinations like Guam
may also need to provide more than just quality-related information. For example,
marketing material should not talk only about the whiteness of the sand nor the clarity of
the ocean waters. Rather, the copy or imagery in the material should also convey the
opportunities for enjoyment of these resources. This enjoyment can come directly
through specific beach activities, or indirectly, as in the case when the beach is simply the
setting for a honeymoon or some other special occasion.

This may be even more appropriate when the marketing objective is to encourage
return or repeat visitation. Facilities like hotels and restaurants —— or more accurately, the
quality of these facilities — may represent known quantities for people who have already
been to the destination. Marketing materials may need to point out less familiar elements
of the destination in catering to the tourists’ desire for novelty, even from a familiar
destination. Examples would be out-of-the-way dining establishments, rare shopping
bargains, and other activities off the beaten path.

Popular hotels, restaurants and sights need not be ignored. However, the
marketing organization might want to emphasize the intangible elements of these
components, e.g., sentimental associations, convenience of location, etc., rather than the
standard recitation of features, rates and facilities. This principle may also be applied to
the marketing of specific tourist establishments. A hotel should not just describe the size
or luxuriousness of its rooms to prospective tourists. Rather, it should present the large,
luxurious suites as venues for tourists to indulge or to escape, depending on the

anticipated motivations of target market groups.

114

It has been suggested that marketing organizations — in this case, destinations —-
should try to lower the expectations of tourists. These would then be easier to match or
exceed. This, the argument goes, would improve the destination’s ability to achieve
satisfaction. However, by introducing motive-oriented information to prospective
visitors, is there a risk of raising expectations beyond the destination’s ability to satisfy
them?

Marketers should acknowledge that tourists already have particular travel
motivations and preferences even before they are exposed to promotional material from a
given destination. Thus, in addressing these motivations through marketing is not the
same as creating new expectations. Rather, a destination like Guam will be providing
information that can affirm, validate or build on these latent expectations.

Perhaps the more appropriate admonition with regard to the management of
expectations is that marketers should help prospective visitors develop realistic
expectations. If tourists hold very high -— almost unrealistic — expectations, it may be
difficult for a destination to meet these expectations. However, if they opt to use
marketing as a mechanism for consciously toning down visitor expectations, visitor
bureaus might likewise dampen enthusiasm for the destination even before tourists have
decided on a place to visit.

For destinations like Guam, it may be to their advantage that their target markets
are relatively long-haul travelers. They have potentially greater control over the
information that is delivered to the target markets in that most of the information
processed by tourists will likely come from the visitor bureaus. In contrast, short-haul or

cross-border travel markets (e. g., US to Mexico, Hong Kong to Taiwan) may have much

115

more convenient access to news, and other bits of information about the destination from
sources other than the visitors bureau. However, there will be no assurance that all of this

information will be what the visitors bureaus would consider reliable nor positive.

Recommendations for Further Study

This study has been based on secondary analysis of data gathered by the Guam
Visitors Bureau. As such, the results of the study have been constrained by artifacts from
the original data collection. These limitations should be addressed in future studies.

First, it appears that there are issues arising from the fact that the survey was
administered immediately upon exit frOm Guam. It would be interesting to have studies
with longer periods between the actual experience and the administration of the survey.
Studies that track visitors over time would also be usefirl. Although recall bias will be a
concern, longitudinal studies like these could be used to see if the model decays over
time. For example, will the high correlation between overall satisfaction and future
intentions, detected in this study, decrease with time? Conversely, will the relatively low
effect of image on intention increase when the tourists have had some time to reflect on
and process the information from their vacation?

Regarding this particular issue, two possible approaches come to mind. One
approach would be to apply a constraint-negotiation fi'amework to the problem. Jackson,
Crawford, and Godbey’s (1993)6 fiamework suggests that tourists (assuming that they
want to do so) will return to Guam if they can successfirlly negotiate a hierarchy of

constraints between them and their desired activity. Conversely, failure to mitigate the

 

‘ This builds on their earlier work in Crawford, Jackson and Godbey (1991).

116

effects of these constraints will result in a decision not to travel to Guam. Thus, intentions
to return may be strengthened or formed when constraints can be successfiilly negotiated.

Researchers may hypothesize that immediately after a trip (as was the case in this
study), tourists will still be euphoric over the success of the vacation. In such a state, they
may not be as aware of any possible constraints such as time, money or lack of a suitable
travel companion, as they would probably be after some time has passed. However, when
some time has passed, the tourists may look at a return trip more objectively, recognizing
possible constraints to travel. Thus, even if they were extremely satisfied with their trip,
they might not as easily claim any intention to return if that option appears realistic from
a constraint-negotiation perspective.

Another approach might be to consider that the tourists’ emotional states at
departure will be different from their emotional states after a month. Bagozzi, Gopinath
and Nyer’s (1999) review suggests that this difference in emotion and mood may have
important effects on the tourists’ appraisal of Guam as well as their evaluation of a
possible return to Guam. They note that the influence of mood on evaluation has already
been demonstrated for a variety of stimuli. In general, researchers have found that
individuals in positive-mood states evaluate objects more positively than individuals in
neutral- or negative-mood states.

A month or so after a visit to Guam, it may be expected that tourists will have
moved toward a more neutral-mood state than what they will have been in when
departing from Guam. Based on Bagozzi et al. (1999), tourists in this mood-state might

then lean toward a more neutral evaluation of Guam.

117

Another limitation of the study has to do with the fact that the original survey
apparently focused on tourist evaluations at the business or service level. Thus, were
relatively few variables were included that are associated with quality at the destination
level. This model might be strengthened by adding more variables that could be used as
indicators of perceived quality at the destination level (e. g., “beauty of scenery”, “clean
air”, etc.)

There were two variables, originally assigned to the destination quality construct,
that were subsequently dropped from the model. These were “communication” and “ease
of getting around.” These were dropped because the confirmatory factor analysis
suggested that they might be indicators of an unspecified, unknown construct. Both
variables were originally thought to be indicators of destination quality because the
destination can provide facilities and infrastructure that enhance these attributes. Thus,
the destination exercises some degree of control over the provision of these attributes.

However, upon further reflection, it could also be argued that the tourists also
enjoy some control over these two variables. The degree to which tourists are conversant
in English, or their level of travel experience could influence how they perceive these two
variables. For example, English-speaking visitors or people who have recently been to
Guam might report fewer problems with communication or with getting around. Thus,
the control exhibited by the destination can be described as partial, rather than full
control. This might suggest that another quality construct — one that recognizes tourist
inputs +- should be added to the model.

Researchers may also wish to investigate how the model can be applied or

modified for other types of destinations. As a tourist spot, Guam is a tropical destination

118

heavily dependent on its natural resources. However, much like Hawaii and Saipan, it
caters to a relatively traditional tourist market made up of mass tourists fiom Asia and
the United States. Thus, the built component of its tourist product (hotels, restaurants,
shops, etc.) is also highly developed.

In contrast, a destination like Boracay Island in the Philippines would be much
less developed. Although it has relatively similar tourist assets (beach, warm weather,
marine activities), its tourism infrastructure is definitely not at the same level as Guam’s.
Resorts in Boracay, as well as many other emerging destinations in the Asia-Pacific
region, are generally small, and follow the traditional design of island huts. Dining and
shopping establishments on the island may be described as basic relative to the more
urbanized restaurants, shops and malls on Guam.

In such a destination, the attractions should play an even more important part in
the travel experience. Will Boracay now provide better support for hypothesis 2 — that
satisfaction with attractions and activities have the strongest direct effects on overall
satisfaction? The analysis might be extended further by acknowledging that Guam and
Boracay are at different stages in their life-cycles, Guam being more mature than
Boracay.

Researchers might therefore ask, whether the relative importance of attractions
decreases as a destination matures. As markets — through increased marketing exposure
or through actual visitation - become more familiar with Boracay’s attractions, these
become basic elements to be expected from any trip to the island. On the other hand

unique lodging facilities, and other visitor amenities may be seen as extras that go beyond

119

 

the basic expectation of a Boracay trip. These extras might then provide a greater
contribution to the post-visit evaluation of the destination.

Finally, future investigations following this dissertation’s line of inquiry may
employ qualitative methods to complement the exit survey data. Focused group
interviews or unstructured conversations with tourists may enable researchers to extract
more detailed recollections of visitor experiences. Such in-depth information may
uncover facets of perceived quality and satisfaction that survey questionnaires would not

be able to tap.

Conclusion

Any study based on analysis of secondary data must recognize the limitations of
secondary data. This investigator acknowledges the weaknesses of using data collected
by others. The data was collected for purposes that were different from - though related
to — this dissertation’s research objectives. Inevitably, some factors were not
operationally defined as well as they could have been if measured with primary data
expressly collected for the purpose. Furthermore, the sample used for the survey
represented the target markets of Guam. The visitor markets for other destinations in the
Asia-Pacific region may not be similarly structured. Finally, visitor evaluations of
destinations — Guam, in the case of this particular survey - reflect the unique
characteristics or attributes of the place.

Thus, the results of this dissertation must be treated with caution. But, they can be
related to previous, as well as ongoing, research on quality and satisfaction such that

future directions for extending the model can already be identified. However, more

120

primary research is needed to validate the assumptions that needed to be made for this
study, particularly those involving the indicators of destination quality and image.
Nevertheless, there are a few practical and theoretical conclusions that can be made at
this point.

First, this study has highlighted some of the opportunities for conducting
sophisticated analysis using secondary tourism data. More specifically, it has
demonstrated the benefits from strengthening academe-GVB partnerships in research. A
more open attitude towards sharing data can be a mutually-beneficial arrangement for
both parties. Academics are given access to information and research materials that they
may not have the resources to collect on their own. On the other hand, visitor bureaus can
tap the technological expertise and alternative perspectives that outside researchers have
to offer.

The partnership might be even more fiuitfiil if academic researchers and industry
leaders can collaborate even earlier in the research process. For example, researchers
could work with their counterparts from the CVB in designing or modifying survey
instruments such that the items correspond to the specific research interests of both
academe and industry. This should minimize the need to rely on assumptions about
variable and construct meanings that had to be made for this study.

Second, this dissertation has elaborated on the difference between quality and
satisfaction, as well as the relationship between the two constructs. For both managers
and researchers, these findings indicate a need to view destination products holistically.

This perspective understands that tourist evaluations of a visited destination go beyond

121

the physical attributes of the place. They also assess the feelings that result from
experiencing these place attributes.

Thus, it is important that both managers and researchers recognize the experiential
nature of the destination product. For managers, this implies a return to the basics of the
marketing concept. They should resist a tendency to focus solely on the quality-related,
internal elements of their product such that visitor satisfaction stops being the primary
objective of the organization. Finally, the symmetric notions of “experience as product”
and “product as experience”, hopefully, represent a promising direction in research that
enables students of consumer behavior to elaborate more completely on the concepts of

quality, satisfaction, image, and intentions with regard to visitor destinations.

122

 

APPENDICES

123

APPENDIX A.

Means and Standard Deviations of Variables

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Variable Mean Standard

Deviation
Hotel tangibles 5.25 1.37
Hotel intangibles 5.43 1.32
Dining satisfaction 5.23 1.16
Enjoyable shopping 5.49 1.40
Shopping satisfaction 5.35 1.10
Attractions satisfaction 5.78 1.11
Cleanliness 5.41 1.38
Safety 5.72 1.27
Easy to get around 5.35 1.60
No problems with communication 5.56 1.43
Better than expected 5.50 1.40
Enjoyable destination 5.75 1.25
Rate Guam 5.86 1.53
Plan to return to Guam 5.52 1.78
Recommend to friends 5.47 1.46

 

 

Notes:
1. Total n = 3,108; Valid (listwise) n = 1,846

2. All items rated on 7-point scales where 1 represents the most negative rating (e.g.,
“very dissatisfied”, “strongly disagree”, “least desirable”, “unclean”, etc.) and 7 is

the most positive rating (e. g., “very satisfied”, “strongly agree , most desirable”,
“clean”, etc.)

124

 

APPENDIX B

Means and Standard Deviations of Variables by Nationality

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

I-IK-Taiwan Japan Korea
(n=690) (n=1628) (n=672)
Variables Mean s.d. Mean s.d. Mean s.d.

Hotel tangibles 5.647 1. 172 5.000 1.408 5.295 1.432
Hotel intangibles 5.618 1.246 5.214 1.337 5.725 1.421
Dining satisfaction 5.450 1.086 5.086 1.204 5.316 1.218
Enjoyable shopping 5.184 1.298 5.585 1.360 5.070 1.500
Shopping satisfaction 5.373 1.141 5.335 1.073 5.361 1.260
Attractions satisfaction 5.587 1.050 5.899 1.103 5.619 1.225
Cleanliness 5.569 1.362 5.205 1.487 4.616 1.564
Safety 5.367 1.378 5.572 1.323 5.012 1.495
Easy to get around 5.888 1.131 5.575 1.266 5.434 1.242
No problems with 5.829 1.173 5.090 1.329 5.280 1.308
communicatlon

Better than expected 5.455 1.354 5.789 1.257 5.512 1.331
Enjoyable destination 5.066 1.449 5.613 1.288 4.982 1.410
Rate Guam 5.103 1.452 5.764 1.146 5.175 1.451
Plan to return to Guam 4.969 1.699 5.476 1.489 4.959 1.676
Recommend to friends 5.226 1.503 5.429 1.303 5.013 1.459

 

 

 

 

 

 

 

 

Note: All items rated on 7-point scales where 1 represents the most negative rating (e.g.,
“very dissatisfied”, “strongly disagree”, “least desirable”, “unclean”, etc.) and 7 is the
most positive rating (e.g., “very satisfied”, “strongly agree”, “most desirable”, “clean”,

etc.)

125

Means and Standard Deviations of Variables by Purpose of Travel

APPENDIX C

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Honeymoon Pleasure Corporate
(n=328) (n=1996) (n=490)
Variables Mean s.d. Mean s.d. Mean s.d.
Hotel Tangibles 5.475 1.257 5.173 1.417 5.279 1.315
Hotel intangibles 5.807 1.187 5.389 1.360 5.345 1.319
Dining satisfaction 5.337 1.108 5.211 1.204 5.207 1.173
Enjoyable shopping 5.231 1.439 5.427 1.394 5.347 1.357
Shopping satisfaction 5.256 1.191 5.350 1.110 5.391 1.106
Attractions satisfaction 5.901 0.934 5.775 1.109 5.713 1.180
Cleanliness 4.875 1.513 5.292 1.484 5.022 1.503
Safety 5.114 1.403 5.491 1.371 5.360 1.345
Easy to get around 5.569 1.262 5.700 1.216 5.431 1.284
No problems with communication 5.449 1.272 5.336 1.330 5.177 1.299
Better than expected 5.766 1.111 5.672 1.314 5.584 1.310
Enjoyable destination 5.386 1.225 5.379 1.407 5.278 1.347
Rate Guam 5.620 1.124 5.565 1.306 5.299 1.334
Plan to return to Guam 5.242 1.502 5.287 1.604 5.165 1.584
Recommend to friends 5.362 1.303 5.336 1.382 5.164 1.427

 

 

 

 

 

 

 

 

Note: All items rated on 7-point scales where 1 represents the most negative rating (e. g.,
“very dissatisfied”, “strongly disagree”, “least desirable”, “unclean”, etc.) and 7 is the
most positive rating (e. g., “very satisfied”, “strongly agree”, “most desirable”, “clean”,

etc.)

126

APPENDIX D

Means and Standard Deviations of Variables by Previous Experience with Guam

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Repeat visit First time
(n=620) (n=2308)
Variables Mean s.d. Mean s.d.

Hotel Tangibles 5.137 1.408 5.233 1.382
Hotel Intangibles 5.291 1.381 5.451 1.345
Dining Satisfaction 5.227 1.177 5.217 1.195
Enjoyable shopping 5.436 1.383 5.369 1.399
Shopping satisfaction 5.343 1.204 5.350 1.103
Attractions satisfaction 5.738 1.133 5.768 1.124
Cleanliness 5.317 1.514 5.123 1.506
Safety 5.487 1.369 5.383 1.397
Easy to get around 5.597 1.248 5.626 1.239
No problems with communication 5.203 1.286 5.332 1.333
Better than expected 5.735 1.316 5.628 1.301
Enjoyable destination 5.405 1.345 5.337 1.395
Rate Guam 5.572 1.326 5.482 1.316
Plan to return to Guam 5.648 1.456 5.141 1.619
Recommend to friends 5.413 1.351 5.260 1.406

 

 

 

 

 

 

Note: All items rated on 7-point scales where 1 represents the most negative rating (e. g.,
“very dissatisfied”, “strongly disagree”, “least desirable”, “unclean”, etc.) and 7 is the
most positive rating (e. g., “very satisfied”, “strongly agree”, “most desirable”, “clean”,

etc.)

127

BIBLIOGRAPHY

128

BIBLIOGRAPHY

Ajzen, I. (1985), From intentions to actions: A theory of planned behavior. In Kuhl, J. &
Beckmann, J. (Eds) Action control: From cognition to behavior. (pp. 11-39).
New York: Springer.

Ajzen, I. (1988). Attitudes, personality and behavior. Chicago, IL: Dorsey Press.

Ajzen, I., & Driver, BL. (1992). Application of the theory of planned behavior to leisure
choice. Journal of Leisure Research, 24, 3: 207-214.

Anderson, J ., & Gerbing, D. (1988). Structural equation modeling in practice: A review
and recommended two-step approach. Psychological Bulletin, 103, 3: 411-423.

Baker, D., & Crompton, J. (2000). Quality, satisfaction and behavioral intentions. Annals
of Tourism Research, 27: 785-804.

Bagozzi, R., Gopinath, M., and Nyer, P. (1999). The role of emotions in marketing.
Journal of the Academy of Marketing Science, 2 7, 2: 184-206.

Bagozzi, R., & Yi, Y. (1988). On the evaluation of structural equation models. Journal of
the Academy of Marketing Science, 16, 1: 74-94.

Baloglu, S., & McCleary, K. (1999). US international pleasure travelers' images of four
' Mediterranean destinations: A comparison of visitors and nonvisitors. Journal of
Travel Research, 38, 2: 144-152.

Barsky, J. (1992). A strategy for customer satisfaction. Cornell Hotel and Restaurant
Administration Quarterly, 33, 5: 32-38.

Bentler, P. (1995). EQS structural equations program manual. Encino, CA: Multivariate
Software Inc.

Bitner, M. J ., Booms, B., & Tetreault, M. (1990). The service encounter: Diagnosing
favorable and unfavorable incidents. Journal of Marketing, 54, 1: 71-84.

Bolton, R., & Drew, J. (1991). A multistage model of customers' assessments of service
quality. Journal of Consumer Research, 17, 4: 375-3 84.

Boulding, W., Kalra, A., Staelin, R., & Zeithaml, V. (1993). A dynamic process model of
service quality: From expectations to behavioral intentions. Journal of Marketing
Research, 30: 7-27.

Brown, T., Churchill, G., & Peter, J. (1993). Improving the measurement of service
quality. Journal of Retailing, 69, 1: 127-139.

129

 

 

 

Bultena, G., & Klessig, L. (1969). Satisfaction in camping: A conceptualization and
guide to social research. Journal of Leisure Research, I, 4: 348-3 54.

Byrne, B. (1995). One application of structural equation modeling from two perspectives:
Exploring the EQS and LISREL strategies. In Hoyle, R., ed. Structural equation
modeling. Concepts, issues and applications. (pp. 138-157). Thousand Oaks, CA:
SAGE Publications.

Cadotte, E. R., Woodrufi’, R. B., & Jenkins, R. (1987). Expectations and norms in models
of consumer satisfaction. Journal of Marketing Research, 24: 305-314.

Carman, J. (1990). Consumer perceptions of service quality: An assessment of the
SERVQUAL dimensions. Journal of Retailing, 66, 1: 33-55.

Childress, R., & Crompton, J. (1997). A comparison of alternative direct and discrepancy
approaches to measuring quality of performance at a festival. Journal of Travel
Research, 35, 2: 43-57.

Chon, KS. (1990). The role of destination image in tourism. A review and discussion.
Tourist Review, 45, 2: 2-9.

Crawford, D., Jackson, E., & Godbey, G. (1991). A hierarchical model of leisure
constraints. Leisure Sciences, 13, 4: 309—320.

Crompton, J. (1979). Motivations for pleasure vacations. Annals of Tourism Research, 6:
408-424.

Crompton, J ., & Love, L. (1995). The predictive validity of alternative approaches to
evaluating quality of a festival. Journal of Travel Research, 34, 1: 11-24.

Crompton, J ., MacKay, K., & Fesenmaier, D. (1991). Identifying dimensions of service
quality in public recreation. Journal of Park and Recreation Administration,
9: 15-27.

Cronin, J ., & Taylor, S. (1992). Measuring service quality: A reexamination and
extension. Journal of Marketing, 53, 3: 55-68.

Danaher, P., & Arweiler, N. (1996). Customer satisfaction in the tourist industry: A case
study of visitors to New Zealand. Journal of Travel Research, 35, 1: 89-93.

Dann, G. (1977). Anomie, ego-enhancement and tourism. Annals of Tourism Research,
4: 184-194

Fick, G., & Ritchie, J RB. (1991). Measuring service quality in the travel and tourism
industry. Journal of Travel Research, 30, 2: 2-9.

Floyd, M. (1997). Pleasure, arousal, and dominance: Exploring affective determinants of
recreation satisfaction. Leisure Sciences, 19: 83-96.

130

 

 

Gotlieb, J ., Grewal, D., & Brown, S. (1994). Consumer satisfaction and perceived
quality: Complementary or divergent constructs? Journal of Applied Psychology,
79, 6: 875-885.

Graefe, A., & Fedler, A. (1986). Situational and subjective determinants of satisfaction in
marine recreational fishing. Leisure Sciences, 8, 3: 275-295.

Gronroos, C. (1984). A service quality model and its marketing implications. European
Journal of Marketing, 18, 4: 36-44.

Gundersen, M., Heide, M., & Olsson. O. (1996). Hotel guest satisfaction among business
travelers. What are the important factors? Camel] Hotel and Restaurant
Administration Quarterly, 3 7, 2: 72-81.

Hair, J ., Anderson, R., Tatham, R., and Black, W. (1998). Multivariate data analysis.
Upper Saddle River, NJ: Prentice Hall.

Heung, V., & Cheng, E. (2000). Assessing tourists' satisfaction with shopping in the
Hong Kong Special Administrative Region of China. Journal of Travel Research.
38, 2: 396-404.

Howard, J ., & Sheth, J. (1969). The theory of buyer behavior. New York: John Wiley &
Sons, Inc.

Hoyle, R. (1995). The structural equation modeling approach: Basic concepts and
fundamental issues. In Hoyle, R., ed. Structural equation modeling. Concepts,
issues and applications. (pp. 1-15). Thousand Oaks, CA: SAGE Publications.

Hu, L.T., & Bentler, P. (1995). Evaluating model fit. In Hoyle, R. (ed.) Structural
equation modeling. Concepts, issues, and applications. Thousand Oaks, CA:
SAGE Publications, Inc.

Hudson, S. (1999). Consumer behavior related to tourism. New York: The Haworth
Hospitality Press.

Hunt, K. H. (1977). Consumer satisfaction and dissatisfaction: Perspectives and
overview (Excerpted from proceedings of a conference on "Conceptualization and
measurement of consumer satisfaction/dissatisfaction" conducted by the
Marketing Science Institute with support of National Science Foundation No. 77-
112). Marketing Science Institute.

Iverson, T. (1997). Japan visitors to Guam: Lessons from experience. In Crotts, J. (ed.)
Marketing issues in Pacific area tourism. pp. 41-54. New York: The Haworth
Press, Inc.

Jackson, E., Crawford, D., & Godbey, G. (1993). Negotiation of leisure constraints.
Leisure Science, 15: 1 — 11.

131

 

 

Kiecolt, K]. and Nathan, LE. (1985). Secondary analysis of survey data. Beverly Hills:
SAGE Publications.

Komba, L. (1997). The impacts of national character on critical factors of customer
satisfaction and on customers’ behavioral intentions in service organizations : a
case of tourist hotels in Tanzania. Augsburg : Wissner.

Kotler, P., Bowen, 1., and Maken, J. (1996). Marketing for hospitality and tourism. Upper
Saddle River, NJ: Prentice-Hall.

Locke, E. (1969). What is job satisfaction? Organizational Behavior and Human
Performance, 4: 309-336.

Lounsbury, J ., & Hoopes, L. (1985). An investigation of factors associated with vacation
satisfaction. Journal of Leisure Research, 17, 1: 1-13.

MacKay, K., & Crompton, John (1990). Measuring the quality of recreation services. ,
Journal of Park and Recreation Administration, 8, 2: 47-5 5. ‘

Manell, R., & Kleiber, D. (1997). A social psychology of leisure. State College, PA:
Venture Publishing, Inc.

Martilla, J ., & James, J. C. (1977). Importance-performance analysis. Journal of
Marketing, 41, 1: 77-79.

Mayer, K., Johnson, L., Hu, C., & Chen, S. (1998). Gaming customer satisfaction: An
exploratory study. Journal of Travel Research, 37, 4: 178-183.

 

Mazursky, D. (1989). Past experience and future tourism decisions. Annals of T ourisrn
Research, 16: 333-344.

Micceri, T. (1989). The unicorn, the normal curve, and other improbable creatures.
Psychological Bulletin, 105: 156-166.

Murphy, P., Pritchard, M., & Smith, B. (2000). The destination product and its impact on
traveller perceptions. Tourism Management, 21: 43-52.

Oh, H., & Parks, S. (1997). Customer satisfaction and service quality: A critical review of

the literature and research implications for the hospitality industry. Hospitality
Research Journal, 20, 3: 3 5-64.

Oliver, R. (1980). A cognitive model of the antecedents and consequences of satisfaction
decisions. Journal of Marketing Research, I 7: 460-69.

Oliver, R. (1993a). A conceptual model of service quality and service satisfaction:

Compatible goals, different concepts. Advances in Services Marketing and
Management, 2: 65-85.

132

 

Oliver, R. (1993b). Cognitive, affective, and attribute bases of the satisfaction response.
Journal of Consumer Research, 20: 418-430.

Oppermann, M. (1999). Predicting destination choice. A discussion of destination
loyalty. Journal of Vacation Marketing, 5, 1: 51-65.

Parasuraman, A., Zeithaml, V., & Berry, L. (1985). A conceptual model of service
quality and its implications for fixture research. Journal of Marketing, 49, 4: 41-
50.

Parasuraman, A., Zeithaml, V., & Berry, L. (1988). SERVQUAL: A multiple-scale item
for measuring consumer perceptions of service quality. Journal of Retailing, 64,
1: 12-40.

Peter, J .P. (1979). Reliability: A review of psychometric basics and recent marketing
practices. Journal of Marketing Research, 16: 6-17.

Peter, J. P., Churchill, G., & Brown, T. (1992). Caution in the use of difference scores as
measures of marketing constructs. A. C. Nielsen working paper #91-2, University
of Wisconsin-Madison, Madison, WI.

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

Petty, R., & Cacioppo, J. (1981). Attitudes and persuasion-classic and contemporary
approaches. Dubuque, IA : W.C. Brown Co. Publishers.

Pizam, A., & Milman, A. (1993). Predicting satisfaction among first-time visitors to a
destination by using the expectancy disconfirmation theory. lntemational Journal
of Hospitality Management, 12, 2: 197-209.

Pizam, A., Neuman, Y., & Reichel, A. (1978). Dimensions of tourist satisfaction with a
destination area. Annals of Tourism Research, 5: 314-322.

Reisinger, Y., and Turner, L. (1999). Structural equation modeling with LISREL:
Application in tourism. Tourism Management, 20, 1: 71-88

Ritchie, J. R. B. (1996). Beacons of light in an expanding universe: An assessment of the
state-of-the-art in tourism marketing/marketing research. Journal of Travel and
Tourism Marketing, 5, 4: 49-84.

Scott-Lennox, J., & Lennox, R. (1995). Sex-race differences in social support and
depression in older low-income adults. In Hoyle, R. (ed.) Structural equation
modeling. Concepts, issues, and applications. pp. 199-216. Thousand Oaks, CA:
SAGE Publications, Inc.

Sherif, M., & Hovland, C1. (1961). Social judgment: Assimilation and contrast effects in
communication and attitude change. New Haven, CT: Yale University Press.

133

 

Sirgy, M. J. (1984). A social cognition model of CS/D: An experiment. Psychology and
Marketing, 1, Summer: 27-44.

Spreng, R., & Mackoy, R. (1996). An empirical examination of a model of perceived
service quality and satisfaction. Journal of Retailing, 72, 2: 201-214.

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

Swan, J ., & Combs, L]. (1976). Product performance and consumer satisfaction: A new
concept. Journal of Marketing, 40, 2: 25-33.

Teare, R. (1994). Marketing in hospitality and tourism - A consumer focus. London:
Cassell.

 

Teas, R. K. (1993). Expectations, performance, evaluation and consumers' perceptions of
quality. Journal of Marketing, 57, 4: 18-34.

Teye, V., & LeClerc, D. (1998). Product and service delivery satisfaction among North
American cruise passengers. Tourism Management, 19, 2: 153-160.

Tse, D., & Wilton, P. (1988). Models of consumer satisfaction formation: An extension.
Journal of Marketing Research, 25: 204-212.

Um, S. & Crompton, J. (1990). Attitude determinants in tourism destination choice.
Annals of Tourism Research, 1 7: 43 2—448.

Witt, C., and Wright, P. (1992). Tourism motivation: Life after Maslow. In Johnson, P.,
and Barry, T. (eds) Choice and demand in tourism. (pp. 33-56) London: Mansell
Publishing Limited.

Woodside, A. G. & Lysonski, S. (1989). A general model of traveler destination choice.
Journal of Travel Research, 27, 4: 8—14.

Yi, Y. (1990). A critical review of consumer satisfaction. Review of Marketing, 4: 68-
123.

Yokoyama, F. (1991) Causal modeling of tourist satisfaction: Application to Michigan's
Northwestern coastal tourism region. Ph.D., Michigan State University. Park,
Recreation and Tourism Resources.

Yuan, K.H., Bentler, P., & Kano, Y. (1997). On averaging variables in a confirmatory
factor analysis model. Behaviormetrika, 24, 1: 71-83.

134